SDA-RF-CH-3.1 NEW.pdf

Media

extracted text
SDA-RF-CH-3.1

Bulletin ofthe World Health Organization, 61 (6): 967-979(1983)

© World Health Organization 1983

Interventions for the control of diarrhoeal diseases
among young children: supplementary feeding
programmes
R.G. Feachem*1
The effect of supplementary feeding programmes on diarrhoeal disease morbidity and
mortality among preschool children is reviewed using data from field studies in developing
countries. The supplementary feeding programmes considered are those that provide food
to preschool children on a continuing and community-wide basis. Nutritional rehabilitation
of sick children and feeding programmes in disasters and emergencies are not considered.
The evidence that poor nutritional status predisposes to increased diarrhoeal disease
incidence, or that supplementary feeding programmes can reduce diarrhoeal disease
incidence, is not strong. There is evidence that poor nutritional status predisposes to more
severe diarrhoea and to higher case fatality, and that supplementary feeding programmes
can reduce the severity of the diarrhoea and the mortality. However, supplementaryfeeding
programmes entail high costs and considerable logistic and managerial complexity and it is
unlikely that they are a cost-effective intervention for national diarrhoeal diseases control
programmes. Prospective studies into the effect of nutritional status on the severity of
etiology-specific diarrhoeas and the resulting deaths are warranted.
A synergism between diarrhoeal diseases and nu­
tritional status has been accepted for many years (35)
and substantial research efforts have been directed
towards unravelling the nature of this synergism.
Some authorities (41) have suggested that the
enhancement of nutritional status by supplementary
feeding programmes may be an effective intervention
to reduce the rates of diarrhoeal diseases in preschool
children. Supplementary, feeding programmes refer
here to the planned distribution of foodstuffs to
improve the dietary intake of preschool children (6
months to 5-6 years is the usual age range of target
children). The food distribution may be on a ‘takehome’ basis or it may involve supervised feeding at
Teeding centres. Supplementary feeding programmes
may aim to reach the whole of a stated age-sex group,
or they may be targeted at children with particular
levels of nutritional deprivation. Several recent
reviews of supplementary feeding programmes pro­
vide a useful background to this more focused
analysis (/, 8, 20, 23). The therapeutic feeding, over
short periods, of children suffering from severe mal­
nutrition and diarrhoea is undoubtedly an essential

part of case management and can reduce the case
fatality rate among such children. In this review,
however, such nutritional rehabilitation of sick chil­
dren is not considered and only data on the preventive
effect of more broadly targeted supplementary feed­
ing programmes are considered. This review of the
role of supplementary feeding programmes in diar­
rhoeal disease control is the second in a series of
reviews of potential anti-diarrhoea interventions
being published in the Bulletin of the World Health
Organization (15, 16).

EFFECTIVENESS

For supplementary feeding programmes to be an
effective diarrhoea control intervention, it must be
true that:

1 Reader in Tropical Public Health Engineering and Head of
Department of Tropical Hygiene, London School of Hygiene and
Tropical Medicine, Keppel Street, London WCIE 7HT, England.
Requests for reprints should be addressed to the Programme
Manager, Diarrhoeal Diseases Control Programme, World Health
Organization. 1211 Geneva 27, Switzerland.

4362

—967—

either

supplementary feeding
programmes can
improve the
nutritional status
of young children

hypothesis
I

968

R. G. FEACHEM

and

young children with
improved nutritional
status have reduced
diarrhoea morbidity
rales and/or mortality
rates and/or severity

hypothesis
2

supplementary feeding
programmes can reduce
diarrhoea morbidity
rates and/or mortality
rales and/or severity
in young children

hypothesis
3

Most of the literature on this topic has been addressed
to one or more of these specific hypotheses. The
potential effectiveness of supplementary feeding pro­
grammes would be suggested by a demonstration
either of the correctness of hypotheses 1 and 2 or of
the correctness of hypothesis 3. The evidence for and
against the three hypotheses is now examined.

Hypothesis 1. Supplementary feeding programmes
can improve the nutritional status ofyoung children
Beaton & Ghassemi (<8) reviewed the impact on nu­
tritional status of 43 supplementary feeding pro­
grammes. Major differences in the nature and goals
of the feeding programmes, and in the methods for
evaluating their impact, make it difficult to draw
generalized conclusions from these programmes (7).
There is no doubt that supplementary feeding pro­
grammes dan have a major impact on the nutritional
status of individual participating children, and that
the worse the initial nutritional state of the child, the
greater is the improvement in anthropometric indices
that may be attributed to the feeding programme.
Impact on the nutritional status of the target group as
a whole, however, is typically low or non-existent.
This lack of impact may be due to the following
factors:

structure. Coverage in programmes reviewed by
Beaton & Ghassemi was typically less than 10%.
(ii) Low levels ofsupplementation: in programmes
reviewed by Beaton & Ghassemi, the supplementation
was designed to meet 40-70% of the estimated energy
gap, but in practice only 10-25% of the gap was
closed. Energy intakes from supplementary food
ranged from 19 to 431 kcal (79.5 to 1803 kJ) per child
per day.
(iii) Food sharing: in take-home feeding pro­
grammes, only 40-60% of the food distributed
appeared to reach the targeted children, the re­
mainder being consumed by other family members or
being sold (8).
(iv) Food substitution: there is often a compen­
satory decrease in the intake of other foods. In 5 out
of 7 programmes reviewed by Beaton & Ghassemi in
this regard, the net increase in energy intake among
the participating children, as a percentage of the
energy in the supplementary food ingested, was less
than 100% and in one programme it was as low as
16%.

Many of the feeding programmes reviewed by
Beaton & Ghassemi (<S) were research or pilot inter­
ventions rather than routine feeding programmes,
and the former tend to have greater impact than the
latter. Some of the research interventions, for
instance in Colombia, Guatemala and India, pro­
duced mean weight gains in project children that were
0.5-1.0 kg/year greater than in the control children.
These marked impacts in research interventions may
serve to define maximum impacts, but they do not
reflect the generally more modest achievements of
routine supplementary feeding programmes.
The programmes reviewed were, in general, not
effective in reaching children aged 6-23 months. This
is probably due to traditional practices of late wean­
ing and to maternal attitudes towards appropriate
feeding regimens for children of this age. Since diar­
rhoea rates are at their highest in this 6-23-month
period (56), this finding is important in the context of
diarrhoea control and suggests that attention may be
better directed towards improved weaning practices
than supplementary feeding programmes.
Hypothesis 2. Young children with improved nu­
tritional status have reduced diarrhoeal morbidity
rates and/or mortality rates and/or severity.

Associations have been frequently reported be­
tween poor nutritional status and increased diarrhoea
(i) Low coverage: a considerable proportion morbidity, mortality, and severity. This literature is
(20-75%) of children enrolled in supplementary comprehensively reviewed by Leslie." These associ* LESLIE, J. Child malnutrition and diarrhea: a longitudinal
feeding programmes fail to participate, while many
from north-east Brazil. Doctor of Science thesis, School of
more children in the region or country are denied study
Hygiene and Public Health, Johns Hopkins University, Baltimore
access because of limitations in the programme infra­ MD, USA, 1982.

969

SUPPLEMENTARY FEEDING FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

ations could be due to one or more of the
following:

— diarrhoea causes poor nutritional status (propo­
sition 1);
— poor nutritional status predisposes to diarrhoea
(proposition 2);
— both poor nutritional status and diarrhoea are
associated with other factors such as a recent measles
attack or poverty (proposition 3).
The evidence for the first proposition (diarrhoea
causes poor nutritional status) is considerable (14,28,
29, 32).b This implies that diarrhoea control may be
an effective anti-malnutrition intervention. The third
proposition is also true, it being well established that
both diarrhoea and poor nutritional status are found
disproportionately among the deprived and under­
privileged sections of the community (40). The pre­
cise cause-and-effect relationships are not understood
See footnote a, page 968.

Table 1. Association between nutritional status and
diarrhoea incidence in rural Bangladesh"
Type of data collected

Finding

1.

Discharge weight-for-age
of 811 children (0-4 years)
treated for diarrhoea at
rural hospital compared
with weight-for-age of 882
children of the same age
range in the community.

For both males and females
the prevalence of severely
malnourished children
(< 60% weight-for-age)
was significantly higher
among diarrhoea cases
than among village
children.

2.

Diarrhoea treatment rate
over 2 years for 2019
children (12-23 months)
according to nutritional
status (weight-for-age,
weight-for-height, heightfor-age) at the start of
the 2-year period.

No association was found
between nutritional status
and diarrhoea treatment
rate, either for individual
anthropometric measures
or combined measures.

Diarrhoea incidence by
community surveillance
over 1 year among 207
children (0-4 years)
according to nutritional
status (weight-for-age)
at the start of 1 year's
surveillance.

No association was found
between nutritional status
and diarrhoea incidence
rate.

Probability of experiencing
diarrhoea in a given
4-week period among 207
children (0-4 years)
according to weight gain
(weight-for-age % increase,
body weight % increase,
body weight kg increase) in
previous 4-week period.

No association was found
between weight gain and
probability of diarrhoea.

3.

4.

' Summary of Chen et al. (13}.

but the general conclusion is that overall socio­
economic development may gradually reduce both
diarrhoea and malnutrition. For the second propo­
sition (poor nutritional status predisposes to diar­
rhoea) there is uncertainty and controversy. It is only
if this proposition is correct that nutritional supple­
mentation could be an effective anti-diarrhoea inter­
vention.
Most studies into the association between diar­
rhoea and nutrition have failed to separate propo­
sition 1 from proposition 2, and have also failed to
allow for the confounding variables of proposition 3.
Only a few prospective studies have adequately
explored proposition 2. The study by Chen et al. (13)
in rural Bangladesh is summarized in Table 1, in
which the first finding is typical of many previous
studies and fails to disentangle propositions 1 and 2 or
to control for confounding variables; the remaining
findings are of greater interest and fail to show that
poor nutritional status predisposes to increased inci­
dence of diarrhoeal illness.
Another prospective study was that of 343 children
aged 6-32 months in rural northern Nigeria (38). The
heights and weights of each child were recorded in
April and the incidence and duration of diarrhoea
were recorded during May-July by weekly home
visits. The period May-July coincides with the end of
the dry season and the start of the rains and is the
period of peak diarrhoea incidence. The results of this
study are shown in Table 2. Being underweight or

Table 2. Association between nutritional status and
diarrhoea incidence and duration among children aged
6-32 months in rural northern Nigeria"

Nutritional
status

No. of
children

Diarrhoea
attack rate
per child
over 3
months6

Time with
diarrhoea
*
(%)

Weight/age:

> 75%

220

1.25

8.5|

<75%
(underweight)

123

1.52

11.3 J

>90%

245

1.37

< 90% (stunting)

98

1.45

> 80%

302

1.29)

7.el „

< 80% (wasting)

41

1.90)

i3.ej

Height/age:

7,9l ..

10.8/

Weight/height:

Data from Tomkins (38).
Significance values of comparisons within anthropometric
groups: •P<0.02; •"PcO.OI; •••P<0.001.

970

R. G. FEACHEM

stunted did not predispose to increased diarrhoea inci­
dence but being wasted was associated with a signifi­
cantly higher attack rate. Being underweight, or
stunted or wasted were all associated with a signifi­
cantly increased duration of diarrhoea, and this effect
was most marked for children who were wasted. The
major defect of this study is that socioeconomic and
environmental variables were not controlled. It could
be that children who were undernourished in April
tended to come from more crowded, less educated,
poorer and dirtier homes than other children (40).
Such disadvantaged children might experience in­
creased exposure to diarrhoea-causing organisms
(and thus higher attack rates) and receive less
adequate care when sick (and thus have a longer
duration of illness) than other children. The Nigerian
report (38) provides no data on the degree of familial
clustering of the children with poor nutritional status.
Since the confounding variables mentioned above are
all family attributes, future studies of this type should
control for familial variation, perhaps by comparing
poorly nourished children with well nourished chil­
dren in the same families. Trowbridge et al. (39)
found a strong association between poor nutritional
status and diarrhoea incidence in the subsequent 12
months in El Salvador, but recognized that this
association could be confounded by socioeconomic
status.
Other studies have recorded a relationship between
poor nutritional status and increased duration of diar­
rhoea. Palmer et al. (30) found increased cholera
duration among hospitalized male Bangladeshis of
low weight-for-height, although, since anthropo­
metric measures were made at discharge, it was not
clear whether the low weight caused prolonged
cholera or vice versa. In a prospective study in San
Jose, Costa Rica, the average duration of diarrhoea
episodes in children aged 12-59 months was signifi­
cantly longer among those with low weight-for-age
than among others (22).
Two studies of children admitted to hospital with
acute diarrhoea of known etiology in Bangladesh
failed to find an association of poor nutritional status
with diarrhoea duration, but did find an association
with diarrhoea severity as measured by the rate of
stool output and degree of dehydration. In the first
study (70), which was of children with rotavirus
diarrhoea, those with low (<60%) weight-for-age
had a higher frequency of severe dehydration, but a
similar duration of diarrhoea, than others. In the
second study/ no difference was found in the rate of
stool output, total stool volume, or diarrhoea dur­
ation between children of poor (< 80% weight-forheight) and better (> 80% weight-for-height) nuc BLACK, R. E. ET AL. Nutritional status, body size and severity
of diarrhoea associated with rotavirus or enterotoxigenic Escherichia
coli. (Unpublished).

tritional status who had been admitted to hospital
with acute rotavirus or enterotoxigenic Escherichia
coli diarrhoea. However, the rate of stool output (per
kg of body weight) was significantly higher among
children with small body size (low weight- and heightfor-age).
If poor nutritional status predisposes to more
severe diarrhoea, particularly in the form of more
dehydrating diarrhoea or diarrhoea of longer
duration, then it would be expected that poor nu­
tritional status would predispose to diarrhoea mor­
tality. Chen et al. (11) measured the heights and
weights of 2019 children aged 12-23 months in rural
Bangladesh and then recorded mortality among these
children over the following 2 years. A striking associ­
ation between nutritional status (weight-for-age) and
subsequent diarrhoea mortality was recorded (Table
3), with children < 65% weight-for-age having a
diarrhoea mortality rate 3.8 times higher than
children > 65% weight-for-age. In this and another
report (12) a marked association between mortality
rate and socioeconomic status (measured by housing
floor area) in the same children was noted. Unfor­
tunately the data on diarrhoea mortality by nu­
tritional status are not controlled for socioeconomic
status and the possible importance of this confound­
ing variable is not elucidated.
Other prospective studies have also reported an
association between poor nutrition and diarrhoea
mortality. In rural Punjab, India, 71% of under-3year-old children dying from diarrhoea were < 70%
weight-for-age in the two months preceding death
(27). The mean prevalence of < 70% weight-for-age
in children from the same community at the same time
was significantly lower (25%). In rural Bangladesh,
children aged 0-9 years who died of diarrhoea had a
pre-morbid mean weight-for-height of 74% of the
standard, compared with 83-86% of standard for
children who died of other causes and 88% of
standard for living controls (26).
In summary, the evidence that poor nutritional
status predisposes to increased diarrhoea incidence is

Table 3. Association between nutritional status and
diarrhoea mortality among children aged 12-23 months
(at start of study) in rural Bangladesh"

Nutritional
status (weightfor-age)

No. of
children

Diarrhoea
mortality
per 1000
over 2 years

All children

2019

20

Children < 65%

742

38

Children > 65%

1277

10

' Oata from Chen et al. (11).

971

SUPPLEMENTARY FEEDING FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

not strong. There is good evidence, however, that
children with poor nutritional status are more likely
to have dehydrating or prolonged diarrhoea, and are
more likely to die from diarrhoea, than well
nourished children. Data from Bangladesh suggest
that children < 65% weight-for-age have a diarrhoea
mortality rate 3.8 times higher than other children
(//). There is a need for more studies to clarify
further the association between poor nutritonal status
and risk of severe diarrhoea and diarrhoea death.
Such studies should be etiology-specific and pro­
spective, and control for family socioeconomic
status. No studies so far reported meet these three cri­
teria and very little is known of the possible differing
effects of nutritional status on diarrhoeas of differing
etiology.

Hypothesis 3. Supplementary feeding programmes
can reduce diarrhoea morbidity rates and/or mor­
tality rates and/or severity in young children
Only a small minority of supplementary feeding
programmes collect data on their impacts, and only a
minority of these evaluate impact by measuring the
changes in mortality or morbidity. Annex 1 sum­
marizes the results of evaluations of 9 supplementary
feeding programmes, which included measurement of
mortality and/or morbidity changes. Of the 9 pro­
grammes, 6 investigated mortality changes and, of
these, 5 found that mortality was reduced while 1
found that it was not. Of the 5 programmes with a
mortality reduction, 2 included both nutritional
supplementation and curative health care and it is not
possible to separate out the impact of the nutritional
intervention alone. Three studies (in Guatemala in
1959-64; in Punjab, India; and in Peru) documented
a significant reduction in mortality that was
attributed to supplementary feeding alone, but no
data are presented on the impact on diarrhoea mor­
tality specifically. It is reasonable to assume that the
percentage reduction in diarrhoea mortality is at least
as great as the reduction in mortality from all causes.
•With this assumption, the reduction in diarrhoea
mortality due to the supplementary feeding pro­
grammes in Guatemala and the Punjab may have
been 19-31% among children under 1 year of age and
around 50% among children aged 1-4 years. In Peru
there was a striking (> 50%) reduction in diarrhoea
mortality among infants under 1 year of age.
Theoretical reductions in diarrhoea mortality
achieved by a supplementary feeding programme may
be computed (Table 4). Six different initial preva­
lences of severe malnutrition are considered and
supplementary feeding programmes are assumed to
either eliminate severe malnutrition or to cut its preva­
lence by half. Reductions in diarrhoea mortality are
then computed by assuming that severely mal-

Table 4. Estimated reductions in diarrhoea mortality
among children under 5 years of age due to supple­
mentary feeding programmes of differing effectiveness
in communities with differing levels of severe malnu­
trition

% of children <: 65% weight-for-age
before nutritional
intervention

after nutritional
intervention

% reduction
in diarrhoea
mortality rate
caused by
nutritional
intervention0

12

5

0

5

2.5

6

10

0

22

10

5

11

15

0

30

15

7.5

15

20

0

36

20

10

18

25

0

41

25

12.5

21

30

0

46

30

15

23

“ Assuming that children < 65% weight-for-age have a 3.8
times higher diarrhoea mortality rate than other children (see
Table 31.

nourished children have a diarrhoea mortality rate 3.8
times higher than other children and that only
severely malnourished children are predisposed to
diarrhoea mortality (//). The reductions in diarrhoea
mortality thus estimated are in the range 6-46% and
are greater in communities with a higher prevalence of
severe malnutrition. These calculations are greatly
simplified and take no account of age-related effects.
Children aged under 2 years have the highest diar­
rhoea mortality rate (36) but are least affected by
supplementary feeding programmes. The relative risk
of diarrhoea mortality for severely malnourished chil­
dren (3.8) was taken from a single study in Bangla­
desh of children between 12 and 47 months of age.
Table 4 can be recomputed with respect to a particular
community in which the prevalence of severe malnu­
trition and the differential diarrhoea mortality rates
of severely malnourished children are known.
Six of the 9 programmes were evaluated for their
impact on diarrhoea incidence: 1 showed an impact, 3
showed no impact, and 2 could not be interpreted
owing to deficient study methods. The one study that
showed an impact on diarrhoea morbidity (in
Madhya Pradesh, India) evaluated the combined

972

R. G. FEACHEM

impact of supplementary feeding of 6-35-month-old
children plus a comprehensive package of curative
and preventive health care. The effect of the nu­
tritional intervention alone cannot be determined.
One study (in Guatemala, 1959-64) documented a
reduced case-fatality ratio for diarrhoea (1.7% vs
0.8%) and a reduced proportion of severe diarrhoeas
(20% vs 10%) that were attributed to the supple­
mentary feeding intervention.
In summary, there is evidence from three studies
and from theoretical calculations that supplementary
feeding programmes can reduce diarrhoea mortality
in children under 5 years of age. The magnitude of
this reduction may be as high as 50% in some age
groups and in communities where severe malnutrition
is common. There is no evidence that supplementary
feeding can reduce diarrhoea incidence, and some evi­
dence that it can reduce the case-fatality ratio and the
severity of diarrhoea episodes. These findings corre­
spond closely to the conclusions on hypothesis 2.

reported costs of USS 21-147 (1982 dollars) per
participating child. Food accounted for between 54%
and 92% of these costs. When costs are jomputed not
per participating child, but per malnourished child
Gudged nutritionally or anthropometrically), the
figures increase considerably (Table 5).
If the annual cost per participating child is USS 100
(Table 5), and if a 20% reduction in diarrhoea mor­
tality is achieved among participating children (Table
4) having a pre-intervention annual diarrhoea mor­
tality rate of 1.4 per 100 children (36), the cost per
diarrhoea death averted is USS 36 000. Under the
most favourable conditions, with an annual cost per
participating child of USS 20 and a 50% diarrhoea
mortality reduction achieved, the cost per diarrhoea
death averted is USS 2900. These calculations under­
estimate the cost-effectiveness of supplementary feed­
ing programmes by assuming that they achieve no
benefits other than prevention of diarrhoea mor­
tality.

FEASIBILITY

CONCLUSIONS

Supplementary feeding programmes have been
implemented in many developed and developing
countries. There is considerable accumulated oper­
ational experience of such programmes and this has
been reviewed (7, 8). The major operational diffi­
culties are to attain and sustain a high coverage of
targeted children and to ensure that the participating
children regularly ingest a sufficient amount of
supplementary energy. These and other difficulties
are discussed above in the section dealing with
hypothesis 1.

The evidence that poor nutritional status pre­
disposes to increased diarrhoeal disease incidence, or
that supplementary feeding programmes can reduce
diarrhoeal disease incidence, is not strong. There is
evidence that poor nutritional status predisposes to
more severe diarrhoea and to higher case fatality, and

Table 5. Annual feeding programme costs in four
countries*
Annual cost (in 1982 US$)
per
participating
child

per child
with energy
deficit

per child
with anthro­
pometric
deficiency

39

45

406

dry, take-home

21

23

103

wet, take-home

24

26

116

37

45

71

147

175

452

COSTS

Country

The costs of operational supplementary feeding
programmes reviewed by Beaton & Ghassemi (8) were
USS 23-39 (1982 dollars) per child enrolled per year.
These costs were for the provision of 300-400 kcal
(1255-1674 kJ) per child per day and include food
plus delivery and administration. Total costs in super­
vised feeding programmes were somewhat higher
than in take-home programmes. The cost of food
accounted for about 70% of the total cost of takehome programmes. Since non-participation is com­
mon (20-75% of children enrolled fail to participate),
the costs per participating child are increased by
1.25-4 times, thus giving a maximum cost range of
USS 29-156 (1982 dollars) per participating child per
year.
In a review of CARE preschool feeding pro­
grammes in five countries, Anderson et al. (1)

Colombia
take-home
Dominican
Republic

Pakistan
take-home

Costa Rica

on-site

* Data from Anderson et al. If). Costs converted to 1982
USS using GNP deflators computed from data In Internationa!

financial statistics.

SUPPLEMENTARY FEEDING FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

that supplementary feeding programmes can reduce
diarrhoea severity and mortality.
Supplementary feeding programmes, however,
entail high costs and considerable logistic and
managerial complexity and have typically failed to
achieve a significant improvement in the nutritional
status of children aged 6-23 months, who have the
highest rates of sickness and death from diarrhoea. It
is unlikely that supplementary feeding programmes
are a cost-effective intervention for national diar­
rhoeal diseases control programmes. Prospective
studies into the effect of nutritional status on the
severity of etiology-specific diarrhoeas and the result­
ing deaths (using the methods recommended above)

973

are warranted.
Therapeutic feeding has not been included in this
review. Such feeding can be life-javing for the
severely malnourished child. Also not included are
feeding programmes implemented in emergency or
disaster situations where inadequate food availability
may be the cause of malnutrition for a majority of
children. Reviewed here are supplementary feeding
programmes designed to enhance food intake of pre­
school children over several years and on a com­
munity-wide basis. Such programmes are shown to be
costly, to be ineffective in reducing diarrhoea mor­
bidity rates, and to be of uncertain effectiveness in
reducing diarrhoea mortality rates.

ACKNOWLEDGEMENTS
The author is grateful for the constructive criticisms on earlier drafts of this paper provided by L. C. Chen, I. de Zoysa,
R. Hogan, R. Martorell, M. Merson, A. Pradilla, P. M. Shah and A. Tomkins. Secretarial and bibliographic assistance was
most ably provided by Maelorwen Jones and Dianne Fishman.

RESUME
INTERVENTIONS POUR LA LUTTE CONTRE LES MALADIES DIARRH&IQUES
CHEZ LES JEUNES ENFANTS : PROGRAMMES D’ALIMENTATION COMPLfiMENTAIRE

Ce document est le deuxieme d’une sArie d’Atudes sur les
interventions possibles en vue de rAduire la morbidity et la
mortality par maladies diarrhAiques parmi les enfants de
moins de 5 ans dans les pays en dAveloppement. On admet
depuis de nombreuses annAes qu’il existe une synergie entre
les maladies diarrhAiques et 1’Atat nutritionnel, et d’importantes recherches ont AtA effectuAes en vue d’Alucider la
nature de cette synergie. D’aprAs certains experts, 1’amAlioration de I’etat nutritionnel grace A des programmes d’ali­
mentation complementaire pourrait constituer une strategic
efficace pour rAduire les taux de maladies diarrhAiques chez
les enfants d’Age prescolaire. Par programme d’alimenta­
tion complAmentaire, on entend ici la distribution planifiAe
de denrAes destinies A amAliorer I’apport alimentaire chez
les enfants d’Age prAscoIaire (la fourchette d’Age habituelie
des enfants cibles est de 6 mois A 5-6 ans). Les aliments
distribuAs peuvent etre soit «A emporter A domicile#, soit
destines A etre consommes sous surveillance dans des centres
d’alimentation. Ces programmes d’alimentation compIAmentaire peuvent viser la totalitA d’un groupe d’Age et de
sexe dAfinis ou bien des enfants atteints de degrAs particuliers de carence nutritionnelle.
Les preuves qu’un Atat nutritionnel mAdiocre predispose A
une incidence accrue de maladies diarrhAiques ou que des
programmes d’alimentation complAmentaire peuvent rAduire cette incidence sont assez fragiles. Toutefois, des ob­
servations ont montrA qu’un Atat nutritionnel mediocre pre­

dispose A des diarrhees plus graves et A une lAtalitA plus
elevee, et que les programmes d’alimentation complemen­
taire sont capables de reduire la gravite des diarrhees et la
mortalite qu’elles causent. Ces programmes entrainent des
depenses elevees et sont trAs complexes au point de vue de la
logistique et de la gestion; en outre, ils n’ont pas reussi A
ameiiorer de maniAre notable 1’Atat nutritionnel d’enfants
agAs de 6 A 23 mois, c’est A dire de ceux justement parmi lesquels les taux de morbiditA et de mortalite par diarrhees sont
le plus AlevAs. Il est peu probable que les programmes d’ali­
mentation complementaire soient une intervention d’un bon
rapport coQt/rendement pour les programmes nationaux de
lutte contre les maladies diarrhAiques.
L’alimentation thArapeutique n’a pas AtA incluse dans
cette Atude. Une telle alimentation peut sauver la vie d’un
enfant atteint de malnutrition sAvAre. Ne sont pas non plus
considArAs les programmes d’alimentation mis. eh oeuvre
dans les situations d’urgence ou de catastrophe, dans lesquelles la pAnurie d’aliments peut Aire une cause de malnu­
trition pour la majoritA des enfants. Sont AtudiAs ici les pro­
grammes d’alimentation complAmentaire visant A accroitre
I’apport alimentaire des enfants d’Age prAscolaire sur plusieurs annAes et sur la base d’une collectivitA tout entiAre. De
tels programmes sont coflteux, incapables de rAduire les taux
de morbiditA par diarrhAe, et leur efficacitA est douteuse en
ce qui concerne la rAduction des taux de mortalitA par
diarrhAe.

974

R. G. FEACHEM

REFERENCES
Anderson, M. A. et al. Supplementary feeding. In: 17. Gopaldas, T. et al. Project Poshak: an integrated
Austin, J. E. & Zeitlin, M., ed. Nutritional intervention
health-nutrition macro pilot study for preschool chil­
in developing countries. Cambridge, MA, Oelgedren in rural and tribal Madhya Pradesh, Vol. I, New
schlager, Gunn and Hain, 1981, pp. 25-48.
Delhi, CARE India, 1975.
2. Arole, R. S. Comprehensive rural health project, 18. Gordon, J. E.et al. Nutrition and infection field study
Jamkhed, India. Contact, No. 10, August 1972, pp.
in Guatemalan villages, 1959-1964. VI. Acute diarrheal
1-9.
disease and nutritional disorders in general disease
3. Arole, R. S. Comprehensive rural health project,
incidence. Archives of environmental health, 16:
Jamkhed. In: Gopalan, C., ed. Alternative approaches
424-437 (1968).
to health care systems. New Delhi, Indian Council of
19.
Guzman, M. A. et al. Nutrition and infection field
Medical Research, 1978, pp. 95-101.
study in Guatemalan villages, 1959-1964. VII. Physical
4. Arole, M. & Arole, R. A comprehensive rural health
growth and development of preschool children.
project in Jamkhed. In: Newell, K. W., ed. Health by
Archives of environmental health, 17: 107-118 (1968).
the people. Geneva, World Health Organization, 1975,
20. Gwatkin, D. R. et al. Can health and nutrition
pp. 70-90.
interventions make a difference? Monograph 13,
5. Ascoli, W. et al. Nutrition and infection field study in
Washington, DC, Overseas Development Council,
Guatemalan villages, 1959-1964. IV. Deaths of infants
1980.
and preschool children. Archives of environmental 21. Hofvander, Y. & Eksmyr, R. An applied nutrition
health, 15: 439-449 (1967).
program in an Ethiopian rural community. American
6. Baertl, J. M. et al. Diet supplementation for entire
journal of clinical nutrition, 24: 578-591 (1971).
communities: growth and mortality of infants and chil­ 22. James, J. W. Longitudinal study of the morbidity of
dren. American journal of clinical nutrition, 23:
diarrheal and respiratory infections in malnourished
707-715(1970).
children. American journal of clinical nutrition, 25:
7. Beaton, G. H. Evaluation of nutrition interventions:
690-694 (1972).
methodologic considerations. American journal ofclin­ 23. Kielman, A. A. et al. Nutrition intervention: an
ical nutrition, 35: 1280-1289 (1982).
evaluation of six studies. Studies in family planning, 13:
8. Beaton, G. H.&Ghassemi, H. Supplementary feeding
246-257 (1982).
programs for young children in developing countries.
24. Kielman, A. A. et al. The Narangwal experiment on
American journal of clinical nutrition, 35 (4, sup­
interactions of nutrition and infections. II. Morbidity
plement): 863-916 (1982).
and mortality effects. Indian journal of medical
9. Becroft, T. & Bailey, K. V. Supplementary feeding
research, 68 (suppl.): 21-41 (1978).
trial in New Guinea highland infants. Journal of 25. Kielmann, A. A. et al. The Narangwal nutrition
tropical pediatrics, 11: 28-34 (1965).
study: a summary review. American journal of clinical
10. Black, R. E. et al. Glucose vs sucrose in oral rehy­
nutrition, 31: 2040-2052 (1978).
dration solutions for infants and young children with
26. Koster, F. T. et al. Synergistic impact of measles and
diarrhoea on nutrition and mortality in Bangladesh.
rotavirus-associated diarrhea. Pediatrics, 67: 79-83
(1981).
Bulletin of the World Health Organization, 59:901-908
(1981).
11. Chen, L. C. et al. Anthropometric assessment of
27.
McCord, C. & Kielmann, A. A. A successful pro­
energy-protein malnutrition and subsequent risk of
mortality among preschool-aged children. American
gramme for medical auxiliaries treating childhood diar­
rhoea and pneumonia. Tropical doctor, 8: 220-225
journal of clinical nutrition, 33: 1836-1845 (1980).
(1978).
12. Chen, L. C. et al. Epidemiology and causes of death
among children in a rural area of Bangladesh. Inter­ 28. Martorell, R. et al. Acute morbidity and physical
growth in rural Guatemalan children. Americanjournal
national journal of epidemiology, 9: 25-33 (1980).
1^. Chen, L. C. et. al. A prospective study of the risk of
of diseases in children, 129: 1296-1301 (1975).
diarrheal diseases according to the nutritional status of 29. Martorell, R. et al. Diarrheal diseases and growth
children. American journal of epidemiology, 114:
retardation in preschool Guatemalan children. Amer­
284-292(1981).
ican journal of physical anthropology, 43:: 341-346
U975).
14. Cole, T. J. & Parkin, J. M. Infection and its effect on
the growth of young children: a comparison of the
30. Palmer, D. L. et al. Nutritional status: a determinant
Gambia and Uganda. Transactions of the Royal Society
of severity of diarrhea in patients with cholera. Journal
of infectious diseases, 134: 8-14 (1976).
of Tropical Medicine and Hygiene, 71:196-198 (1977).
15. Feachem, R. G. et al. Diarrhoeal disease control: 31. Parker, R. L. et al. The Narangwal experiment on
interactions of nutrition and infections. III. Measure­
reviews of potential interventions. Bulletin of the World
ment of services and costs and their relation to outcome.
Health Organization, 61: 637-640 (1983).
Indian journal of medical research, 68 (suppl.): 42-54
16. Feachem, R. G. & Koblinsky, M. A. Interventions
(1978).
for the control of diarrhoeal diseases among young chil­
dren: measles immunization. Bulletin of the World
Health Organization, 61: 641-652 (1983).

1.

SUPPLEMENTARY FEEDING FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

975

Rowland, M. G. M. et al. A quantitative study into 37. Taylor, C. E. et al. The Narangwal experiment on
interactions of nutrition and infections. I. Project
the role of infection in determining nutritional status in
design and effects upon growth. Indian
*
journal of
Gambian village children. British journal of nutrition,
medical research, 68 (suppl.): 1-20 (1978).
37: 441-450(1977).
33. Scrimshaw, N. S. et al. Nutrition and infection field 38. Tomkins, A. Nutritional status and severity of diar­
rhoea among pre-school children in rural Nigeria.
study in Guatemalan villages, 1959-1964. III. Field
Lancet, 1: 860-862(1981).
procedure, collection of data and methods of measure­
39. Trowbridge, F. L. et al. Nutritional status and
ment. Archives of environmental health, 15: 6-15
severity of diarrhoea. Lancet, 1, 1375 (1981).
(1967).
Wittmann, W. et al. An evaluation of the relationship
34. Scrimshaw, N. S. et al. Nutrition and infection field 40.
between nutritional status and infection by means of a
study in Guatemalan villages, 1959-1964. V. Disease
field study. South African medical journal (sup­
incidence among preschool children under natural
plement— South African journal of nutrition), 41:
village conditions, with improved diet and with medical
664-682
(1967).
and public health services. Archives of environmental
41.
Wray, J. D. Direct nutrition intervention and the
health, 16: 223-234 (1968).
control of diarrhoeal diseases in preschool children.
35. Scrimshaw, N. S. et al. Interactions of nutrition and
American journal of clinical nutrition, 31: 2073-2082
infection. American journal of medical sciences, 237:
(1978).
367-403 (1959).
36. Snyder, J. D. & Merson, M. H. The magnitude of the
global problem of acute diarrhoeal disease: a review of
active surveillance data. Bulletin of the World Health
Organization, 60: 605-613 (1982).

32.

976

R. G. FEACHEM

Annex 1

SUMMARY OF SUPPLEMENTARY FEEDING PROGRAMMES THAT WERE EVALUATED
FOR IMPACT ON MORBIDITY OR MORTALITY

Country

Rural/
urban

Dates of
feeding
programme

Colombia

urban

Ethiopia

rural

Educational
programme
for mothers

Other
intervention

56% of children
enrolled failed
to participate
regularly
throughout 1 year
of programme
and were excluded
from impact
analysis

Yes

No

Weekly collection
of take-home
allocation to
provide 335 kcal
and 14 g of
protein per
child per day

Only children
receiving > 50%
of weekly
distributions
were included
in study

Yes

DPT, TB. and
and smallpox
vaccination;
general curative
care provided

Target
children

Feeding
programme

1964-65

Malnourished
children.
0-71 mo; plus
their siblings
and pregnant or
lactating mothers

Weekly collection
of take-home
allocation of
1 pound of dried
skimmed milk
per child per
week

1965-67

All children
0-11 years

Participation

Guatemala

rural

1959-64

All children
0-59 mo.

Daily supervised
feeding providing
350 kcal and 15 g
of protein per
child

52% of children
of 6-59 mo.
participated
> 50% of time.
Participation of
< 6 mo. children
was not expected.
Participation
declined during
5 yrs of
programme

Guatemala

rural

1969-76

All children
0-83 mo.

Twice daily
supervised
feeding at
feeding centres

No data

i

No

No

Curative and
preventive
health services.
supplementary
feeding for
pregnant or
lactating women

977

SUPPLEMENTARY FEEDING FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Control
community

Changes in
nutritional
status

Changes in
mortality'

Causes
of death

Morbidity
survey.
methods

Changes in diarrhoeal disease rates

*
Cost

Source

No

Malnutrition,
judged by weightfor-age decreased
substantially during
the 1 yr feeding
programme

No data

No data

Yes, 1-week
recall at visits
to nutrition
centre

Data on diarrhoea
incidence cannot be
interpreted owing to
lack of controls

No data

41

No

Clinical signs of
PCM decreased and
upper arm tircumfcrence-for-agc
increased. Weightfor-age and heightfor-age did not
change

No data

No data

Yes, 3-monthIy
or 6-monthly
examinations
and stool
microscopy

Data on diarrhoea.
Entamoeba histo­
lytica, Ascaris,
and hookworms
cannot be
interpreted
owing to lack
of controls

No data

21

Yes

Proportion of
0-59 mo. children
< 90% weightfor-age declined
from 91% in 1959
to 83% in 1964.
A child in the
feeding village
could expect to
be 30 mm taller
and 1 kg heavier
at age 5 yrs
than a control
child

9 yrs prior to
intervention
compared with
5 yrs of inter­
vention. Crude
mortality
declined from
25 to 16 per 1000
(36% reduction)
0-11 mo.
mortality
declined from
182 to 146 per
1000(19%
reduction), 12-59
mo. mortality
declined from
56 to 24 per
1000 (56%
reduction)

During 5 yrs of
programme, the
proportion of
deaths due to
diarrhoea was
around 20%

Yes, 2-week
recall at
home visits

Incidence of diar­
rhoea cases, among
0-59 mo. children,
per year rose from
48 per 1000 in 1959
(control village 123) to 219 per 1000
in 1964 (control
village - 165). This
was attributable to
falling participation
in the feeding
programmes. Diar­
rhoea case-fatality
ratios during 195964 were 0.8% in the
feeding village and
1.7% in the control
village. The pro­
portions of cases
judged to be severe
(blood or mucous or
duration > 3 days)
were 10% in the
feeding village and
20% in the control
village

No data

5,18,19,
33,34

internal
controls

Children of
9-59 mo. who
regularly took
supplement had
weight and
height gains
10-15% higher
than other
children

Using beforeand-after
comparisons,
0-11 mo.
mortality
declined from
150 to 55 per 1000
(63% reduction)
and 12-59 mo.
mortality declined
from 28 to 6
per 1000 (79%
reduction).
70% of these
*
reduction
were
attributed to
health care and
30% to nutrition,
especially
nutrition of
pregnant mot hen

No data

No data

US$ 7.2 per
capita per year
for health
services. Nutrition
component costs
not available

No

20

All mortality rates for 0-11 mo. children are per 1000 live births. Other mortality rate
*
are per 1000 children in the stated age range.
All costs are given in 1982 USS. Costs in other currencies and years have been convened to 1982 USS using exchange rate
*
and ONP deflators computed from
data in International financial statistics.

978

R. G. FEACHEM

Country

Rural/
urban

Dates of
feeding
programme

*
Indi
(Punjab)

rural

1970-73

India (Madhya
Pradesh):
exploratory
phase

rural

intensive
phase

Target
children

Educational
programme
for mothers

Other
intervention

Feeding
programme

Participation

All children
0-35 mo. and
< 70V. weight
for age

Twice daily
supervized
feeding to provide
up to 400 kcal
and 11 g of
protein per child
per day

Attendance by
target children
averaged 22V. for
those aged 0-12
mo. and 41V. for
those aged
13-36 mo.

Yes

Prenatal care,
supplementary
feeding of
underweight
mothers

1971-72

All children
6-35 mo.

Weekly collection
of take-home
allocation to
provide 377 kcal
and 19 g of
protein per child
per day

53V. of
available food
was collected

Yes

Preventive
and curative
health care.
Supplementary
feeding for
pregnant and
lactating mothers

rural

1972-74

as above

as above

< 50V. of
available
food collected

No

No

India
(Maharashtra)

rural

1971-79

‘‘Deserving’’
children
0-59 mo

Daily supervised
feeding

No data

Yes

Curative and
preventive
health services,
supplementary
feeding for
pregnant women

Papua
New Guinea

rural

1961-62

All children
6-11 mo.

Weekly collection
of take-home
allocation
to provide
100-280 kcal and
10 g of protein
per child per day

35V. of children
enrolled missed
> 15 weeks and
were excluded
from impact
analysis

No

No

Peru

rural

1962-67

All children
and all adults

Weekly collection
of take-home
allocation to
provide 250 kcal
and 12.5 g of
protein per day
to all family
members

95 V. of families
collected their
supplement
every week

No

No

SUPPLEMENTARY FEEDING FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Control
community

Changes in
nutritional
status

Changes in
*
mortality

Causes
of death

Morbidity
survey,
methods

Changes in diar­
rhoeal disease rates

*
Cost

979

Source

Yes

Target children
> 17 mo. were
significantly
heavier, and
target children
>21 mo. were
significantly
taller, than
controls

0-11 mo.
mortality was 89
per 1000 com­
pared with 129
per 1000 in
control villages.
12-35 mo.
mortality was
10 per 1000
compared with
19 per 1000 in
control villages

44
*7
.
of deaths
due to diarrhoea
in 1970

Yes, l-week
recall at
home visits

No effect on diar­
rhoea incidence
or duration
demonstrated

USS 47 per
0-35 mo. child
per year for
nutrition
intervention

24, 25, 27,
31,37

Yes

Significant
improvement in
weight-for-age
for regular
(> 50
*/.
of
collections)
collectors of food

No data

No data

Yes, 2-month
recall at start
and end of
intervention

Significant
reduction in diar­
rhoea incidence
among 24-35 mo.
age group, but not
among younger
children

USS 57 per child
per year for the
complete inter­
vention package

17

Yes

Significant
improvement
in weight-for-age
but not as great as
achieved by
complete package
(see above)

No significant
reduction

4.5'5'. of
deaths among
0—47 mo.
children
attributed to
diarrhoea

No

No data

USS 39 per child
per year for the
nutrition inter­
vention alone

Yes

No data

In 1976,0-11 mo.
mortality was 39
per 1000 in
project area and
90 per 1000 in
control area

No data

No

No data

USS 2.5 per
capita per year
for complete
intervention
package

Yes

Supplementary
feeding did not
affect height or
weight gain

No data

No data

Yes, monthly
examination and
stool microscopy

Supplementary
feeding did not
influence the value
of a composite
"illness score" that
included diarrhoea
and dysentery

No data

9

Yes

No clear evidence
that the supple­
mentary feeding
improved the
nutritional status
or growth of any
age group. This
poor response was
attributed to the
supplementary
foods displacing
traditional foods

0-11 mo.
mortality was
significantly lower
in intervention
villages (44-52
per 1000) than
in control
villages (104-165
per 1000). 12-59
mo. mortality was
not significantly
reduced

No data

No

No data

No data

6

2, 3.4

* All mortality rates for 0-11 mo. children are per 1000 live births. Other mortality rates are per 1000 children in the stated age range.
• All costs are given in 1982 USS. Costs in other currencies and years have been convened to 1982 USS using exchange rates and GNP deflators computed from
data in International financial statistics.

CH 3-2.
© World Health Organization 1983

Bulletin of the H arid Health Organization, 61 (4): 641 - 652 (1983)

Interventions for the control of diarrhoeal diseases
among young children: measles immunization
*
R. G. Feachem1 & M. A. Koblinsky2
The effects of measles immunization on diarrhoea morbidity and mortality are
reviewed using data from field studies and theoretical calculations. Two types of measlesassociated diarrhoea are distinguished: with-measles diarrhoea, which starts between 1
week pre-rash-onset and 4 weeks post-rash-onset, and post-measles diarrhoea, which starts
4-26 weeks post-rash-onset. The etiology of these measles-associated diarrhoeas is
unknown, but some evidence points towards a frequently severe and dysenteric form of
disease, with Shigella playing a major role. Theoretical calculations indicate that measles
immunization, at the age of 9-11 months, with coverage of between 45 % and 90% can avert
44-64% of measles cases, 0.6-3.8% of diarrhoea episodes, and 6-26% of diarrhoea deaths
among children under 5 years of age. The cost of measles immunization is in the range of
USS 2-15 (1982 prices) per child vaccinated. The impact of measles immunization on
diarrhoea mortality may be partly additional to the impact of oral rehydration because it
averts deaths that are not prevented by oral rehydration. Community research is urgently
needed to confirm or reject these theoretical suppositions, to clarify the etiology of measlesassociated diarrhoea, and to determine the cost-effectiveness of measles immunization as
an intervention to reduce diarrhoea mortality.
Measles immunization is now an integral part of the
Expanded Programme on Immunization (EPI) which
many developing countries are implementing with the
support of the World Health Organization. In some
developed countries, e.g., Czechoslovakia and the
USA, the elimination of measles may soon be
achieved (12). There is a marked association between
measles and diarrhoea in developing countries today,
as there was in developed countries in earlier times
(20). Since cases of measles complicated by diarrhoea
have a high fatality rate, measles immunization is a
potential intervention for diarrhoea control. This
review of the role of measles immunization in
diarrhoeal disease control is the first in a series of
reviews of potential anti-diarrhoea interventions
which will be published in the Bulletin of the World
Health Organization (9).
EFFECTIVENESS

either

a considerable pro­
portion of diarrhoea
morbidity or mortality
in young children in
developing countries is
associated with measles

hypothesis
1

and

measles immunization
can reduce the
incidence of measles

hypothesis
2

measles immunization
can reduce diarrhoea
morbidity or mortality
rates in young children

hypothesis
3

For measles immunization to be an effective
diarrhoea control intervention, the following hypoth­
eses have to be tested and proved:
• Requests for reprints should be sent to the Programme
Manager, Diarrhoeal Diseases Control Programme, World Health
Organization, 1211 Geneva 27, Switzerland.

1 Reader in Tropical Public Health Engineering, London School
of Hygiene and Tropical Medicine, Keppel Street, London WC1E
7HT, England.
2 WHO Consultant.

4322

—641 —

642

R. G. FEACHEM & M. A. KOBLINSKY

If hypotheses 1 and 2 are true, it could be appropriate
to field test hypothesis 3; whereas if hypothesis 3 is
known to be true, measles immunization could be
recommended as an operational component of
national diarrhoeal diseases control programmes.
The evidence for and against these hypotheses is
examined below.

Hypothesis 1. A considerable proportion of diar­
rhoea morbidity or mortality in young children in
developing countries is associated with measles
Definitions

Two kinds of measles-associated diarrhoea are
distinguished in this review; with-measles diarrhoea
and post-measles diarrhoea. With-measles diarrhoea
is diarrhoea occurring in close association with an
attack of measles and is defined, arbitrarily, by the
period of time separating the appearance of the
measles rash and the onset of diarrhoea. Authors
differ in their definition of this time period. Scrim­
shaw et al. (25) reported diarrhoea episodes in 127
measles-infected Guatemalan children, under 5 years
of age, which occurred from 2 weeks pre-rash-onset
to 2 weeks post-rash-onset and found that the great
majority of diarrhoea episodes started between 8 days
pre-rash-onset and 5 days post-rash-onset. These
authors subsequently defined the time limits of withmeasles diarrhoea as that starting between 7 days prerash-onset and 7 days post-rash-onset. In a study in
Nigeria of 142 children suffering from measles and
diarrhoea (20), it was found that the diarrhoea started
14-7 days pre-rash-onset in 5% of cases, 6-1 days prerash-onset in 19% of cases, 0-6 days post-rash-onset
in 25% of cases, 7-14 days post-rash-onset in 11% of
cases, and more than 14 days post-rash-onset in 40%
of cases. This high (40%) proportion of diarrhoea
occurring more than 2 weeks after the rash-onset may
be related to the phenomenon of post-measles diar­
rhoea discussed below. Koster et al. (14), after study­
ing 119 village children in Bangladesh who had
measles and diarrhoea over a 6-month period, re­
ported a marked increase in diarrhoea starting be­
tween 1 week pre-rash-onset and 4 weeks post-rash­
onset. With-measles diarrhoea is defined, in the
present review, as that starting between 1 week pre­
rash-onset and 4 weeks post-rash-onset, although it
must be cautioned that most studies do not report the
exact time frame but imply that with-measles diar­
rhoea occurs shortly before or during the exanthem.
A second type of measles-associated diarrhoea,
called here post-measles diarrhoea, was hinted at by
some earlier writers and has recently been investigated
in Bangladesh. Morley et al. (20), for example,
reported from Nigeria that children “were susceptible
to diarrhoea for a long period after the measles itself

had subsided” and that “the diarrhoea may continue
or reappear over a period of many weeks”. More
recently, in Bangladesh, a substantial predisposition
to dysentery, though not to watery diarrhoea, was
found in children for a period of five months follow­
ing an attack of measles.0 There is thus some evidence
for a phenomenon of post-measles diarrhoea, which
is defined here, arbitrarily, as diarrhoea occurring in
the period 4-26 weeks post-rash-onset, over and
above that which could occur in that age group at any
time.
In summary, the total of measles-associated diar­
rhoea cases is defined here as the sum of with-measles
diarrhoea cases (starting between 1 week pre-rashonset and 4 weeks post-rash-onset) and post-measles
diarrhoea cases (starting 4-26 weeks post-rash-onset).
There is, as yet, very little information on the magni­
tude or nature of post-measles diarrhoea.

Measles-associated diarrhoea morbidity
The proportion of diarrhoea episodes that are
measles-associated can be derived theoretically and
from the results of field studies.
Among children aged 0-59 months in developing
countries, there are on average 2.2 episodes of diar­
rhoea per year per child (26). If we assume that every
child also has measles before 5 years of age and that
every measles case has one associated diarrhoea
episode, then in the first five years of life an average
child will have 11 episodes of diarrhoea, of which 1 is
measles-associated diarrhoea. Therefore a theoretical
upper proportion of measles-associated diarrhoea to
all diarrhoea in the first five years of life is 9%.
*
A
review of 10 community-based studies in 5 countries0
shows that only 15-63% of measles cases in children
have accompanying diarrhoea. Using these extreme
figures, the proportion of diarrhoea in the first 5 years
of life that is with-measles diarrhoea is 1.0-4.0%,
assuming that 70% of children have measles before
they are 60 months old (70% is the median figure
from 18 studies in 10 countries0). In addition, data on
post-measles diarrhoea from Bangladesh suggest that
18% of children who have measles will experience an
episode of diarrhoea in the following 6 months, which
they would not otherwise have had.rf By the addition
of these post-measles diarrhoea episodes, it is calcu­
lated that 2.1-5.2% of diarrhoeal episodes in the first
5 years of life are measles-associated, assuming that
70% of children contract measles before the age of 60
° Shahid, N. S. ET al. Long-term complication in measles in
rural Bangladesh. International Centre for Diarrhoeal Disease
Research, Bangladesh, 1982 (unpublished report).
b This is not the theoretical maximum proportion since it is
possible that each measles attack is associated with more than 1 diar­
rhoea episode.
& Tabulated data and sources arc available on request from
d See footnote a, above.

644

R G. FEACHEM & M. A. KOBLINSKY

demies occur in the cool, dry season when diarrhoea
incidence is at its lowest.
In summary, theoretical considerations indicate
that 1-7% of diarrhoea episodes in under-5-year-old
children may be measles-associated (Table 1), while
studies in Nigeria and Guatemala both report a figure
of 6%.
Measles-associated diarrhoea mortality

As with measles-associated diarrhoea morbidity,
the proportion of diarrhoea deaths that are measlesassociated may be derived theoretically and from field
pata.
Theoretically computed measles-associated diar­
rhoea incidence rates are presented in Table 1. They
range from 3 to 14 cases per year per 100 children aged
0-59 months. By the application of case-fatality rates
to these incidences, mortality rates for measlesassociated diarrhoea may be derived. The case­
fatality rates for measles-associated diarrhoea are
high. A review of 13 studies from 11 countries'’shows
that in hospitals, which tend to see more severe cases,
5-29% of young children with measles and diarrhoea
die, whereas fatality rates reported from community
studies are between 2% and 9%. In Table 2, three
possible case-fatality rates (3%, 6% and 9%) are
applied to three possible measles-associated diar­
rhoea incidence rates (4, 8 and 12 per 100 children
aged 0-59 months) to obtain a range of measlesassociated diarrhoea mortality rates. These mortality
rates are then compared with an overall diarrhoea
mortality rate of 1.4 per 100 under-5-year-old chil­
dren per year (26) to obtain the proportions of diar­
rhoea deaths that are measles-associated. The range
of proportions is considerable (9-77%), and is subject
*to the important cautions given in the footnote to
Table 2.
Only one field report giving the proportion of diar­
rhoea deaths that are measles-associated has been
located (14). It describes a one-year surveillance of
5775 rural Bangladeshi children under 10 years old, 29
of whom died of diarrhoea alone and 15 of diarrhoea
with measles (death within one month of rash-onset).
Thus, 34% of these diarrhoea deaths were measlesassociated. In this same study, a further 8 children
died with measles complicated by a combination of
diarrhoea and respiratory symptoms or by various
other complications. If 2 of these 8 deaths are
included as deaths from measles-associated diar­
rhoea, the proportion of diarrhoea deaths that were
measles-associated rises from 34% to 37%. These
data relate to 0-9-year-oId children, not 0-4 years as
in Table 2. The effect of considering this older age
group may be to increase the proportion of diarrhoea
deaths that are measles-associated because the more Tabulated data and sources are available on request from
R.G.F.

Table 2. The proportion of diarrhoea deaths in the first
five years of life that is measles-associated, based on
various assumptions
Measlesassociated
diarrhoea
incidence
Case-fatality
per 100 chilrate for
dren aged
measlesassociated
0-59 months
per year
diarrhoea (%)

Measles
associated
diarrhoea
mortality
per 100
children aged
0-59 months
per year

Proportion of
all diarrhoea
deaths among
children aged
0-59 months
that are
measlesassociated (%)°

4

3

0.12

8.6

4

6

0 24

17.1

4

9

0.36

25.7

8

3

0.24

17.1

8

6

0.48

34.3

8

9

0.72

51.4

12

3

0.36

25.7

12

6

0.72

51.4

12

9

1.08

77.1

" These calculations assume that the annual mortality rate
from diarrhoea in the first five years of life is 1.4 per 100 children
(26). In areas where children die more frequently from diarrhoea
than this, the proportion of deaths that are measles-associated
will be lower, while in areas where the diarrhoea death rate is
lower, the proportion will be higher. It is further assumed that
measles-associated diarrhoea deaths are recorded as diarrhoea
deaths. In the case of post-measles diarrhoea deaths this is a
reasonable assumption, but many with-measles diarrhoea
deaths will be recorded as deaths from measles rather than
deaths from diarrhoea. If all with-measles diarrhoea deaths were
recorded as measles deaths, the proportions given in the final
column should be reduced by 50-86% (see Table 1). These
calculations also assume that the case fatality rates for withmeasles diarrhoea and post-measles diarrhoea are the same;
there is no evidence on which to judge the correctness of this
assumption.

tality rate from all diarrhoeas declines more rapidly
after 5 years of age than does the mortality rate for
measles. Expressing the mortality rate among 6-9year olds as a percentage of the rate among 4-5-year
olds, the same researchers (14) found that it was 6%
for diarrhoea and 23% for measles. This effect may
be partly counteracted by the exclusion from these
data of deaths due to post-measles diarrhoea.
In summary, theoretical considerations indicate
that perhaps 9-77% of diarrhoea deaths in the first 5
years of life are measles-associated, while the one
field study providing data on this suggests a figure of
37% (for 0-9-year-old children).

Etiology of measles-associated diarrhoea

The etiology of measles-associated diarrhoea
remains largely unknown. The high case-fatality
rates, however, suggest a severe form of the disease.

MEASLES IMMUNIZATION AND CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

months. This calculation is sensitive to the assump­
tions made about the proportion of measles cases that
have with-measles or post-measles diarrhoea and to
the assumption that 70% of children have measles
before they are 60 months old. Table 1 presents the
proportions of diarrhoeas in the first 5 years of life
that may be measles-associated, based on different
values of these parameters. The range of proportions
is 1.4-6.6%, and is subject to the cautions given in the
footnote to Table 1.
Data from community studies support this theor­
etical range. In Imesi, Nigeria, 259 children under 5
years old averaged 2.3 episodes of diarrhoea per year
per child over a 3-year period, of which 0.14 (or 6%)
were measles-associated (20). In Santa Cruz Balanya,

643

Guatemala, each child under 5 years of age was found
to suffer an average of 1.65 episodes of diarrhoea in
the year commencing April 1963 (JO). During a 4month period of that year, there was a measles epi­
demic in the village and, during that time, a child
suffered on average 0.6 episodes of diarrhoea of
which 0.098 were with-measles diarrhoea. Therefore,
computed over the year and assuming that no measles
occurred outside the period of the epidemic, the pro­
portion of diarrhoea episodes that were measlesassociated was 6%. The proportion over the period of
the epidemic only was 16%, thus illustrating the
manner in which measles-associated diarrhoea be.
comes more prominent during a measles epidemic anc
especially when, as commonly occurs, measles epi­

Table 1. The proportion of diarrhoea episodes in the first five years of life that is measles-associated, based on
various assumptions
Proportion of
measles cases
having post­
measles diarrhoea
(that they
would not other­
wise have had)
(%)

Proportion
of children
contracting
measles before
age 60 months
(%)

20

10

20

10

20

Proportion of
measles cases
having withmeasles diarrhoea
(%)

Measles-associated
diarrhoea incidence
per 100 children
aged 0-59
months per year

Proportion of all
diarrhoea episodes
among children aged
0-59 months that is
measles-associated
(%)“

90

5.4

2.6

70

4.2

1.9

10

50

3.0

1.4

20

20

90

7.2

3.3

20

20

70

5.6

2.6

20

20

50

4.0

1.8

40

10

90

9.0

4.1

40

10

70

7.0

3.2

40

10

50

5.0

2.3

40

20

90

10.8

4.9

40

20

70

8.4

3.8

40

20

50

6.0

2.7

60

10

90

12.6

5.7

60

10

70

9.8

4.5

60

10

50

7.0

3.2

60

20

90

14.4

6.6

60

20

70

11.2

5.1

60

20

50

8.0

3.6

j

' These calculations assume that, on average, a child has 2.2 episodes of diarrhoea per year in the first five years of life 126}. In
areas where children have more frequent diarrhoea, the proportion that is measles-associated will be lower, white where diarrhoea is
less frequent, the proportion that is measles-associated will be higher. It is further assumed that measles-associated diarrhoea is
recorded as diarrhoea in surveillance data and thus forms a part of the 2.2 episodes of diarrhoea per child per year. In the case of post­
measles diarrhoea this is almost certainly a correct assumption, but with-measles diarrhoea may be recorded as a complication of
measles rather than a case of diarrhoea.

646

R. G. FEACHEM & M. A. KOBLINSKY

Table 3. Reported coverage rates for measles immuniz­
ation in some developing countries reporting > 50%
coverage as at 1982
Reported
coverage
(%)“

WHO Region

Country or area

Africa

Botswana

63

Gambia

61

United Republic of Tanzania

82

Argentina

60

Americas

Brazil

58

Chile

88

Costa Rica

68

Ecuador

67

Maldives

98

Mongolia

98

Europe

Turkey

52

Eastern
Mediterranean

Bahrain

63

Egypt

63

Islamic Republic of Iran

81

Israel

69

South-East Asia

Western Pacific

Kuwait

71

Libyan Arab Jamahiriya

65

Tunisia

65

American Samoa

86

Brunei

*
73

Hong Kong

*
74

Niue

*
77

Singapore

86

Trust Territory of the
Pacific Islands

*
53

° Coverage rates refer to the proportion of children vaccinated
by 12 months, except in cases marked * where they describe
immunization coverage by 60 months. For most countries the
figures refer not to the whole country but to selected regions
where measles immunization has been initiated on a pilot basis.

(Source: Data reported to WHO)

figure of 75%, bearing in mind that the relevant
coverage figure is not what is now being achieved in
countries with only limited immunization pro­
grammes, but what might be reasonably expected in
countries committed to measles immunization and
with a national programme built up over several
years.
Thus, of the 756 children requiring measles
vaccine, 567 (75%) will be vaccinated. The remaining

189 will not be vaccinated and an estimated 66% / or
125 of these, will go on to contract measles before
month 60.
Of the 567 vaccinated children, a proportion would
have been immunized successfully. The not success­
fully immunized ones will be those who were suscept­
ible but failed to seroconvert, plus those who were not
susceptible because of persisting maternal antibodies.
The seroconversion rate at 9-11 months, based on
reports from developing countries, is generally
> 90% although lower figures have also been
reported (Table 4). High seroconversion rates are
reported from carefully organized research studies in
which the vaccine’s potency is well maintained. Sero­
conversion rates in on-going immunization pro­
grammes are probably lower because of operational
factors. In addition, a small proportion of children
(probably 1-5%) at ages 9-11 months are not suscept­
ible owing to persisting maternal antibodies. Taking
these considerations into account, it may be assumed
for the present example that 90% of children without
a history of measles who are vaccinated at ages 9-11
months will seroconvert and be immune to measles.
In Table 5, two seroconversion rates have been
adopted for comparison (80% and 90%). Thus, of the
567 vaccinated children, 510 (90%) would have been
effectively immunized and 57 would not; 38 (66%) of
these 57 would then go on to contract measles before
month 60?
In summary, out of 1000 children born alive, the
following will get measles between months 0 and 60:
3 will have measles and then die before 9-11 months,
94 will have measles and recover before 9-11 months,
125 will not be vaccinated and will go on to get
measles before month 60, and 38 will be vaccinated
but will nevertheless get measles before month 60.
This makes a total of 260 measles cases.
The number of measles cases in the absence of an
immunization programme may be computed in a
similar manner. Of 1000 children born alive, 150 will
die before 12 months and 3 of these will have measles
before they die. Of the remaining 850, approximately
70% (or 595) will have measles before the age of 60
months. This makes a total of 598 measles cases in the
first five years of life. Therefore, in the first five years
of life, the number of measles cases averted per 1000
live births by a measles immunization programme will
be 338 (598-260), or 57%.
This calculation is presented in Table 5, using four
levels of programme coverage (45%, 60%, 75% and
90%), and two rates of successful seroconversion
after measles immunization at 9-11 months of age
(80% and 90%). A correction using infant mortality
J If 11 °7o have measles before 12 months, and 70% before 60
months, the proportion of those who have not had measles al 12
months but who will have it by 60 months is 66%.

MEASLES IMMUNIZATION AND CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Indications that with-measles diarrhoea may be pri­
marily dysenteric in nature are based on reports from
several studies. Thus, blood was commonly present in
the stools of Gambian village children suffering from
with-measles diarrhoea (16). Hospital data from East
and West African countries found 62% (21) and 65%
(22), respectively, of with-measles diarrhoea to be
mucoid or mucoid and bloody, or to be associated
with anal prolapse. In the above reports, bloody and
mucoid diarrhoeas with measles were more fatal than
those described as mucoid only (11% vs 6% in East
Africa; 18% vs 11% in West Africa), although the
mucoid diarrhoeas were the majority of the dysenteric
diarrhoeas (68% and 72% respectively). Bacterio­
logical examination detected Shigella species in
“approximately 50%” of faecal samples from chil­
dren having with-measles diarrhoea in Bangladesh
(14). Although this observation was uncontrolled, it
may be contrasted with the typical 4-10% isolation
rates of Shigella from paediatric diarrhoea cases in
the same area of Bangladesh (2). In Bangladesh it was
also observed (14) that the with-measles diarrhoea
cases were of significantly longer average duration
(51 % of episodes lasting more than 6 days) than other
diarrhoeas (only 25% of episodes lasting more than 6
days). Immunological factors that may account for
the association between measles and diarrhoea have
been reviewed by Greenwood & Whittle (11).
With regard to post-measles diarrhoea, recent evi­
dence from Bangladesh suggests the predominance of
dysentery/ In a comparison of children in the five
months after measles with matched controls, it was
found that the attack rates for watery diarrhoeas were
47% in the post-measles cases and 56% in the controls
(not significant), whereas the attack rates for mucoid
diarrhoeas were 42% among the cases and 19%
among the controls (P < 0.01) and the attack rates for
bloody diarrhoeas were 21% among cases and 9%
among controls (P < 0.05).
In contrast to these findings, a hospital study in
Kenya (23) reported that, of 200 rectal swabs taken
from children having with-measles diarrhoea, only 3
were positive for Shigella and 1 for Salmonella-, the
stools of these patients were described as bloody and
mucoid. The case-fatality rate in this Kenyan series
was, however, high at 29%. A study of 54 children
hospitalized with measles and diarrhoea in Rwanda
found Shigella in the stool of only 1 child, whereas
Salmonella was found in 16(7). A retrospective study
in Thailand found diarrhoea to be a common compli­
cation of hospitalized measles (26%), but the diar­
rhoea was reported to be rarely severe and almost
always responded to oral rehydration.8 Interestingly,
1 See footnote a, page 642.
r WARD, N. a. Survey to determine mortality and morbidity
patterns in measles, Thailand. Unpublished WHO document SEA/

EPI/5, 1979.

645

measles in general was not considered to be especially
life-threatening in Thailand, and the hospital case­
fatality rate for all measles cases was only 2.3%.
The etiology of measles-associated diarrhoea, and
in particular the role of Shigella and other agents of
dysentery, requires urgent investigation. Such studies
are under way, with WHO support, in Rwanda and
should be conducted in other sociocultural and
environmental settings as well.
Hypothesis 2. Measles immunization can reduce the
incidence of measles

The experience of. some developed countries, such
as Czechoslovakia and the USA which are close t|
measles elimination, clearly shows that measles
immunization can reduce the incidence of measles
very greatly (12). In developing countries, however,
many factors intervene to reduce the overall impact of
a measles immunization programme.
Consider, for example, a group of 1000 children
born alive. By the age of immunization (usually 9-11
months), 150 will have died/ and a proportion of
those surviving will have had measles. Data from 18
studies in 10 developing countries show that the pro­
portion of children contracting measles before the age
of immunization is 4-25% and that the figure is gen­
erally higher in urban than in rural areas;' the figure
of 11 % will be adopted here as it is the median figure
from the data reviewed. Therefore, of the 850 chil­
dren surviving to immunization age, 11 % or 94 would
have had measles and it will be assumed that a further
2%, or 3 of the 150 children who died earlier, had
measles before they died. Therefore, out of the 1000
children born in month 0, there will be 97 measles
cases by month 12.
During months 9-11, the 850 surviving children will
be the subjects of a measles immunization prc«
gramme; 756 (850-94) would not have had measles
and a proportion of these will receive the vaccine.
This proportion depends on the programme cover­
age. Theoretically, this coverage can be 100% but in
practice it is lower. As many countries do not have an
effective measles immunization programme, the
coverage figures for a whole region are generally less
than 10%. The coverage figures for individual
selected countries can be higher (Table 3), although
some of the figures reported are very optimistic esti­
mates. In subsequent tabulations (Tables 5 and 6),
four possible coverage figures have been adopted for
comparative purposes (45%, 60%, 75% and 90%).
For the present example, we shall assume a coverage
h The infant mortality rate varies greatly among countries.
A figure of 150 is adopted for the purposes of this illustrative
example. No allowance is made for mortality after 1 year of age.
' Tabulated data and sources are available on request from
R.G.F.

648

R. G. FEACHEM & M. A. KOBLINSKY

portion of measles cases averted in under 5-year-old
children, as a result of the measles immunization
programme.

Hy pothesis 3. Measles immunization can reduce diar­
rhoea morbidity or mortality rates in young children
Field reports on the effect of measles immunization
on diarrhoeal disease rates have been looked for with­
out success, although studies of this type are under
way in Bangladesh and possibly elsewhere.
The only approach, at present, to assessing hypoth­
esis 3 is, therefore, a theoretical one using infor­
mation computed during the assessment of hypoth­
eses 1 and 2. If, among under-5-year-o!d children, 4%
of diarrhoea morbidity (Table 1) and 30% of diar­
rhoea mortality (Table 2) were measles-associated,
and if 54% of the measles cases could be averted by
measles immunization with a coverage of 75% (Table
5), then measles immunization might reduce diar­
rhoea morbidity by 2.2% and diarrhoea mortality
by 16% in children in this age group. Table 6 presents
the results of these computations for various values of
measles-associated diarrhoea morbidity (2%, 4% and

6%), measles-associated diarrhoea mortality (20%,
30% and 40%), and immunization programme cover­
age (45%, 60%, 75% and 90%). It is concluded that
measles immunization among children aged 0-59
months may reduce diarrhoea morbidity by 0.6-3.8%
and diarrhoea mortality by 6-26%. These calcu­
lations all depend on the assumption that, if a case of
measles is averted by immunization, the diarrhoea
(both with-measles and post-measles) associated with
that case of measles will also be averted. This is only
true if measles-associated diarrhoea is actually
“caused” by measles (not necessarily in the sense that
the measles virus causes the diarrhoea but in the more
general sense that measles infection leads to diarrhoea
by still unknown immunological or pathological
mechanisms) and not because the susceptibilities to
measles and diarrhoea are both dependent on some
other factor.
The proportions of diarrhoea morbidity and mor­
tality averted in children under 5 years of age are
likely to increase as the immunization programme
continues and herd immunity rises, with a consequent
decrease in the proportion of children acquiring
measles before the age of immunization and an in­

fable 6. Reduction of diarrhoea morbidity and mortality in the first five years of life by measles immunization at
various levels of coverage

Coverage
of measles
immunization
1
(%)

Proportion
of measles
cases averted
by measles
immunization
(see Table 5)
(%)

45

32

2

0.6

20

6.4

45

32

4

1.3

30

9.6

45

32

6

1.9

40

12.8

60

44

2

0.9

20

8.8

60

44

4

1.8

30

13.2

60

44

6

2.6

40

17.6

75

54

2

1.1

20

10.8

75

54

4

2.2

30

16.2

75

54

6

3.2

40

21.6

90

64

2

1.3

20

12.8

90

64

4

2.6

30

19.2

90

64

6

3.8

40

25.6

Proportion
of diarrhoea
episodes that
are measlesassociated
(see Table 1)
(%)

Proportion of
of diarrhoea
episodes averted
by measles
immunization0
(%)

Proportion
of diarrhoea
deaths that
are measlesassociated
(see Table 2)
(%)

Proportion
of diarrhoea
deaths averted
by measles
immunization0
(%)

a These calculations assume that measles-associated diarrhoea occurs with equal frequency among measles cases that are and
are not averted by measles immunization. If failure to be immunized and measles-associated diarrhoea are both associated with an
independent factor (say, low socioeconomic status), then this assumption does not hold and the impact on diarrhoea of measles
immunization will be lower than predicted here.

647

MEASLES IMMUNIZATION AND CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 4. Age-specific seroconversion rates following measles immunization
Seroconversion rate (%) at:
Country

Reference


6

7

8

9
months

10

52

72

86

95

98

12

11

Africa
Kenya

29
94°

Nigeria

South Africa

23

45

57

Upper Volta

76

Zimbabwe

59

90

80

86
74

63

44

United Republic of Tanzania

71

86

97

13
8
28
17

97

North America
80-85

USA

15

South America
Chile

57

59

3
5

99

3 countries:b

well nourishedc

59

69

84

87

93

94

97

undernourished4*

81

91

93

94

90

96

86

a Children aged 12-23 months
b Brazil, Chile and Ecuador
c > 85% weight-for-age
d 60-85% weight-for-age

rates is not shown in Table 5 because these rates affect
equally the communities both with and without a
measles immunization programme and they do not
change the computed proportions of measles cases
averted. The proportion of measles cases in children
aged 0-59 months averted by measles immunization
(Table 5) is between 30% and 68%, depending on the
programme coverage and, to a lesser extent, on the
assumed seroconversion rate. The reader who is
interested in a particular locality and knows that the
programme coverage, seroconversion rate and age at
immunization are different from those assumed
above may use his figures and make a calculation to
obtain the percentage of measles cases averted.
These calculations reflect the situation of a
community having no measles immunization and the
same community immediately after the implement­
ation of a measles immunization programme with a
specified level of coverage. As this programme
continues, the age distribution of measles cases will
shift upwards, reducing the proportion of children
who acquire measles before the age for immunization
and increasing the proportion of unimmunized chil­
dren who acquire measles after their fifth birthday.
Both these changes will cause an increase in the pro­

Table 5. Proportion of measles cases averted in the first
five years of life by measles immunization, based on vari­
ous assumptions on coverage and the proportion of chil­
dren successfully immunized’

Measles
immunization
coverage
(%)

Proportion of
children
successfully
immunized ®
(%)

Proportion of
cases averted
among 0-4-yearold children
(%)

45

80

30

45

90

34

60

80

41

60

90

46

75

80

51

75

90

57

90

80

61

90

90

68

' The logic and assumptions underlying these computations
are set out in an illustrative example in the text.
’ The proportion of children not having had measles who,
when vaccinated at 9-11 months of age, are subsequently im­
mune to measles.

650

R. G. FEACHEM & M. A. KOBLINSKY

most reliable. Unfortunately these costs refer to
BCG and DPT (rather than measles) immunization.
Measles immunization may be cheaper than BCG plus
DPT, since it is a single-dose vaccine. As only 10-20%
of the cost of an immunization programme is due to
the vaccines (6), the additional cost of adding measles
immunization to an existing EPl programme may be
low. A reason for the apparently high cost figure from
Ivory Coast (Table 7) is that the reverse approach was
taken, 75% of all shared EPI costs being allocated to
measles immunization. This has the effect of making
measles immunization appear expensive, while other
immunizations appear relatively cheap.
h Future research will generate improved data on the
Tost and impact of measles immunization and will
permit cost-effectiveness comparisons of measles
immunization with other diarrhoea control interven­
tions. Care is required in the allocation of costs and
benefits between diarrhoeal disease control pro­
grammes (CDD) and EPI programmes and it must be
recognized that many with-measles diarrhoea deaths
are recorded as measles deaths (see footnote to Table
2). On the other hand, measles immunization may
avert a considerable number of post-measles diar­
rhoea deaths that are not currently accounted for in
evaluations of the impact of measles immunization
programmes (7).

CONCLUSIONS

A theoretical case has been made out that measles
immunization can substantially reduce diarrhoea
mortality among children under five years of age.
Immunization with a coverage of 60% may reduce
diarrhoea mortality in children aged 0-59 months by
9-18%, coverage of 75% may reduce diarrhoea mor­
tality by 11-22%, and coverage of 90% may reduce
diarrhoea mortality by 13-26% (see Table 6). The
impact of measles immunization on diarrhoea mor­
tality is likely to increase as the measles immunization
programme continues and the age distribution of
measles cases shifts upwards. The impact of measles
immunization on diarrhoea mortality may be partly
additional to the impact of oral rehydration because it
averts deaths that are not prevented by oral re­
hydration. Community research is urgently needed
to confirm or reject these theoretical suppositions and
to clarify the etiology of measles-associated diar­
rhoea.
If community studies confirm that measles immu­
nization can reduce diarrhoea mortality, over and
above the reduction achieved by oral rehydration,
detailed cost-effectiveness analyses are required to
compare measles immunization with other possible
interventions for averting the same diarrhoea deaths.

ACKNOWLEDGEMENTS
The authors are grateful for the constructive criticisms of earlier drafts of this paper provided by 1. de Zoysa, S. Foster,
A. Galazka, R. Henderson, R. Hogan, M. Merson, D. Morley, D. Shepard and A. Tomkins. Secretarial and bibliographic
assistance was most ably provided by Ms Maelorwen Jones.

RfiSUME
STRATEGIES POUR LA LUTTE CONTRE LES MALADIES DIARRHEIQUES CHEZ LES JEUNES ENFANTS:
VACCINATION ANTIROUGEOLEUSE

Cette etude est la premiere d’une serie consacree aux stra­
tegies susceptibles d’abaisser la morbidity et la mortality
dues aux maladies diarrheiques chez les enfants de moins de
cinq ans dans les pays en ddveloppement. On constate
actuellement dans ces pays, comme jadis dans les pays
aujourd’hui d6veloppes, une nette association entre la rougeole et la diarrhee; la vaccination antirougeoleuse represente done une arme possible dans la lutte antidiarrheique.
Sur la base de donndes provenant d’enquetes sur le terrain et
de calculs theoriques, les auteurs de 1’etude passent en revue
les effets de la vaccination antirougeoleuse sur la morbidity
et la mortalite dues aux maladies diarrhlques. De 1 Vo & 7 Vo
des episodes diarrheiques et de 9 Vo a 77Vo des deeds par
maladie diarrheique chez les enfants de moins de cinq ans

pourraient etre lies A la rougeole. Une distinction est faite
entre deux types de maladies diarrheiques assocides & la
rougeole: la diarrhee concomitante, dont le debut se situe
entre la semaine pr6c6dant 1’apparition de Irruption et les
quatre semaines consdcutives A celle-ci, et la diarrhee postrougeoleuse qui commence de 4 A 26 semaines aprds 1’ap­
parition de reruption. L’etiologie de ces diarrhees liees A la
rougeole est inconnue, mais certains indices laissent penser
qu’il pourrait s’agir d’une forme frequemment severe et
dysenterique de la maladie, dans laquelle Shigella jouerait
un role primordial. Selon des calculs theoriques, la vacci­
nation antirougeoleuse pratiquee entre 9 et 11 mois, avec
une couverture vaccinale comprise entre 45Vo et 90Vo, serait
a meme d’empecher de 44Vo i 64Vo des cas de rougeole, de

649

MEASLES IMMUNIZATION AND CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

crease in the proportion of unimmunized children
acquiring measles after their fifth birthday.
As the estimated mortality reduction is consider­
able, there is a need for field studies to assess the role
of measles immunization in reducing diarrhoea mor­
tality. If a considerable proportion of measlesassociated diarrhoea is dysenteric in nature, it may
respond poorly to oral rehydration therapy in the
home or at a peripheral health centre. Thus the mor­
tality averted by measles immunization (6-26%) may
be, in part, additional to that averted by oral rehy­
dration. This would make measles immunization an
especially attractive intervention, since certain other
interventions might only reduce the mortality that is
largely averted by oral rehydration.

Possibility that diarrhoea is caused by measles
immunization. It is necessary to consider whether the
above computations require modification because
diarrhoea in some children may be caused by measles
immunization. Only one controlled study on post­
immunization diarrhoea using the further-attenuated
vaccine strains (Schwarz and Moraten) has been
located. This study, which was carried out in Israel
(27) among children aged 9-47 months, showed that
the period prevalence of diarrhoea during days 6-14
after immunization was 12-14% among children
receiving measles vaccines (Schwartz and Moraten)
and 12% among children receiving a placebo (sterile
saline). A survey of 10 035 children aged between 10
and 18 months, who had been immunized against
measles in England, showed that only one child (with
a congenital oesophageal atresia) was admitted to
hospital with diarrhoea and vomiting one week after
the immunization (IS). A recent review of experience
with 131 million cases of measles immunization in the
USA did not mention diarrhoea as an adverse reac­
tion (19). This limited evidence suggests that diar­
rhoea is not an adverse reaction to the measles
vaccines in current use.
The effect of undernutrition. The role of poor
nutritional status in the synergism between diarrhoea
and measles has not so far been mentioned in this
review. There is evidence that poor nutritional status
predisposes to death from both diarrhoea and measles
and so it is likely that the death rate from measlesassociated diarrhoea is higher among children who
are nutritionally deprived. The calculations in this
paper are based on data from Africa, Asia and Latin
America and reflect the situation in communities with
widely differing levels of undernutrition. The cal­
culations are thus generalized across individual
countries and environments. In some communities
the key variables may approximate the median figures
used in this paper, in which case the impact of measles
immunization on diarrhoea may be as calculated

here. In other communities the key variables may
differ considerably from the values used here, in
which case the impact of measles immunization on
diarrhoea may be considerably greater or less than
our calculations. It is likely that in communities where
the prevalence of undernutrition is exceptionally
high, the impact of measles immunization on diar­
rhoea mortality will be greater than estimated in this
review.

FEASIBILITY

Measles immunization is well established atw
highly successful in some developed countries (72)7
Some developing countries have initiated measles
immunization programmes that are achieving reason­
able coverage in selected regions. The main oper­
ational difficulties are maintaining vaccine potency
and achieving high coverage of children within the
target age range (typically 9-11 months). It is
expected that, as experience in measles immunization
in developing countries grows, and as its benefits to
child health become more widely appreciated, these
operational difficulties will be overcome.

COSTS

Few developing countries as yet have national
measles immunization programmes and there is little
reliable information on the cost of such programmes.
Table 7 provides cost data for five countries, of which
the comparative costings of the EPI programmes in
Indonesia, Philippines and Thailand are probably the

Table 7. Costs of immunization
Cost
per child
vaccinated"
(1982 US$)

Country

Immunizations

Indonesia

DPT, BCG

4

Ivory Coast

measles

15

Philippines

DPT, BCG

6

Thailand

DPT, BCG

14

Zambia:
rural
urban

measles
measles

8-14
2-5

Reference

6
30
6
6
24

“ Costs converted to 1982 USS using exchange rates and
GNP deflators computed from data in International Financial

Statistics.

MEASLES IMMUNIZATION AND CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

0,6% a 3,8% des episodes diarrheiques, et de 6% & 26% des
deces dus a des maladies diarrheiques chez les enfants de
moins de cinq ans. Une couverture vaccinale de 75% permettrait d’eviter 2% des Episodes diarrheiques et 16% des
deces dus & des maladies diarrheiques. Il n’existe pas de
donnees fiables sur le cout de la vaccination antirougeoleuse, mais d’apres des statistiques emanant de cinq pays il
pourrait se situer entre 2 et 15 dollars (prix de 1982) par
enfant vaccine.
L’impact de la vaccination antirougeoleuse sur la morta­
lity d’origine diarrh&que devrait augmenter & mesure que se

651

developpera le programme de vaccination et que diminuera
la proportion des cas de rougeole concernant des enfants.
Cet impact viendrait s’ajouter a celui de la rehydratation par
voie orale puisque la vaccination tend & prevenir des deces
que n’dviterait pas le traitement. Il est urgent de procider A
des enquetes et etudes de population pour confirmer ou
infirmer ces suppositions theoriques, clarifier 1’etiologie de
la diarrhee li6e & la rougeole et determiner !’efficacit£, par
rapport au cout, de la vaccination antirougeoleuse en tant
que strategic visant d reduire la mortality due aux maladies
diarrheiques.

REFERENCES
Aaby, P. et al. Measles vaccination and child 13. Ifekwunigwe, A. E. et al. Immune response to
measles and smallpox vaccinations in malnourished
mortality. Lancet, 2: 93 (1981).
children. American journal of clinical nutrition, 33:
2. Black, R. E.etal. A two-year study of bacterial, viral,
621-624(1980).
and parasitic agents associated with diarrhoea in rural
14. Koster, F. T. et al. Synergistic impact of measles and
Bangladesh. Journal of infectious diseases, 142:
diarrhoea on nutrition and mortality in Bangladesh.
660-664 (1980).
Bulletin of the World Health Organization, 59:901-908
3. Borgono, J. M. & Greiber, R. Measles: a five-year
experience with a vaccination program in Chile. In:
(1981).
15. Krugman, S. Present status of measles and rubella
International Conference on the Application of Vac­
immunization in the United States: a medical progress
cines Against Viral, Rickettsial, and Bacterial Diseases
report. Journal of pediatrics, 90: 1-12 (1977).
of Man, Washington, 14-18 December 1970. Washing­
ton, Pan American Health Organization, 1971, pp. 20316. McGregor, I. A. Measles and child mortality in the
Gambia. West African medical journal, 13: 251-257
206 (Scientific Publication 226).
4. Burrowes, J. & Cruikshank, J. G. At what age should
(1964).
17. Meyer, H. M.et al. Response of Volta children to live
measles vaccine be given? Report of a small trial in
attenuated measles virus vaccine. Bulletin of the World
Bulawayo. Central African journal of medicine, 22:
Health Organization, 30: 769-781 (1964).
45-47 (1976).
5. Collaborative study by the Ministries of Health of 18. Miller, C. Surveillance after measles vaccination in
children. The Practitioner, 226: 535-537 (1982).
Brazil, Chile, Costa Rica, Ecuador and the Pan
19. Measles prevention. Morbidity and mortality weekly
American Health Organization. Seroconversion rates
report, 31 (17): 217-231 (1982).
and measles antibody titers induced by measles vaccine
20. Morley, D. et al. Measles in Nigerian children.
in Latin American children 6-12 months of age. Bulletin
A study of the disease in West Africa, and its manifes­
of the Pan American Health Organization, 16: 272-285
tations in England and other countries during different
(1982).
epochs. Journal of hygiene, Camb., 61: 115-135
6. Creese, A. L. et al. Cost-effectiveness appraisal of
(1963).
immunization programmes. Bulletin of the World
21. Morley, D. C. et al. Measles in East and Central
Health Organization, 60: 621-632 (1982).
Africa. East African medical journal, 44: 497-507
7. De Mol, P. & Bosmans, E. Campylobacter enteritis in
(1967).
Central Africa. Lancet, 1: 604 (1978).
8. Dick, B. et al. A minimum age for measles vaccine 22. Morley, D. C. et al. Measles in West Africa. West
African medical journal, 16: 24-31 (1967).
administration to coloured children. South African
23. O’Donovan, C. Measles in Kenyan children. East
medicaljournal, 49: 1951-1954 (1975).
African medical journal, 48: 526-532 (1971).
9. Feachem, R. G. et al. Diarrhoeal disease control:
24. POnnighaus, J-M. The cost-benefit of measles immu­
reviews of potential interventions. Bulletin of the World
nization: a study from southern Zambia. Journal of
Health Organization, 61: 637-640 (1983).
tropical medicine and hygiene, 83: 141-149 (1980).
10. Gordon,}. E.etal. Nutrition and infection field study
25. Scrimshaw, N. S. etal. Studies of diarrhoeal disease in
in Guatemalan villages, 1959-1964. VI. Acute diar­
Central America. VIII. Measles, diarrhoea and nu­
rhoeal disease and nutritional disorders in general dis­
tritional deficiency in rural Guatemala. American
ease incidence. Archives of environmental health, 16:
journal of tropical medicine and hygiene, 15: 625-631
424-437 (1968).
(1966).
11. Greenwood, B. M. & Whittle, H. C. Immunology of
26. Snyder, J. D. & Merson, M. H. The magnitude of the
medicine in the tropics. London, Edward Arnold,
global problem of acute diarrhoeal disease: a review of
1981.
active surveillance data. Bulletin of the World Health
12. Hinman, A. R. World eradication of measles. Reviews
Organization, 60: 605-613 (1982).
of infectious diseases, 4: 933-939 (1982).

1.

652

R. G. FEACHEM & M. A. KOBLINSKY

Swartz, T. et al. A comparative study of four live 29. Expanded programme on immunization: the optimal
age for measles immunization. Weekly epidemiological
measles vaccines in Israel. Bulletin of the World Health
record, 57 (12): 89-91 (1982).
Organization, 39: 285-298 (1968).
28. Expanded programme on immunization: serocon­ 30. Expanded programme on immunization: cost-effective­
version after measles immunization. Weekly epidemio­
ness. Weekly epidemiological record, 57 (22): 170-173
(1982).
logical record, 56 (30): 234-237 (1981).
27.

CH 3.3

/ Research
/ Recherche
Bulletin of the Wbrld Health Organization, 61(4):637 - 640(1983)

>

/
/
© World Health Organization 1983

Diarrhoeal disease control: reviews of potential
„ interventions
R. G. Feachem,1 R. C. Hogan,12 & M. H. Merson2
Diarrhoeal diseases are a major cause of sickness and death among young children in
most developing countries. Since effective interventions to control these diseases are
available, they are a priority target for the primary health care programmes being planned
or implemented in many countries. Governments and international agencies, including the
World Health Organization, have emphasized oral rehydration as an effective intervention
for reducing diarrhoeal disease mortality. Other interventions are, however, needed to
reduce morbidity, to reduce mortality not averted by oral rehydration, and to develop a
multifaceted approach in which oral rehydration is one of several anti-diarrhoea measures
being implemented simultaneously with mutally reinforcing and complementary impacts.
This paper presents a classification of potential interventions for the control of diarrhoeal
disease morbidity and/or mortality among children under 5 years of age and introduces a
series of reviews of these interventions. The first of these reviews, on measles immuniz­
ation, also appears in this issue of the Bulletin of the World Health Organization.
The concept of primary health care involves the
delivery of a package of curative and preventive
health services at the community level. Various health
service needs have to be satisfied but, owing to man­
power, budgetary and other resource constraints, it is
necessary to select those few that meet the priority
health needs and are, at the same time, affordable. A
number of approaches may be used to design an
appropriate primary care package in a specific
country or region. A rational approach is first to
define the major health problems and then select the
most cost-effective means of ameliorating them. This
approach has been discussed fully elsewhere (/, 2).
Identifying and ranking the major health problems
may be done, in defined age groups, by the use of
objective measures of the burden of death or illness
attributable to specific diseases. If this is done, diar­
1 Reader in Tropical Public Health Engineering, London School
of Hygiene and Tropical Medicine, Keppel Street, London WC1E
7HT, England.
2 Diarrhoeal Diseases Control Programme, World Health Organ­
ization, 1211 Geneva 27, Switzerland. Requests for reprints should be
sent to Dr M. H. Merson.

4321

rhoeal diseases emerge in most developing countries
as a major cause of sickness and death in young chil­
dren (2). Recent estimates show that diarrhoeal dis­
eases cause nearly 5 million deaths per year in children
under 5 years old in developing countries (excluding
China) where in every 100 children in this age group
there are, on average, 220 diarrhoeal episodes and 1.4
deaths from diarrhoea every year (5). For a health
problem to be a target for selective primary health
care, it must not only be a major cause of sickness and
death but it must also be controllable at a reasonable
cost. Diarrhoeal disease mortality can be effectively
reduced at reasonable cost by oral rehydration (4) and
possibly other measures.
INTERVENTIONS FOR DIARRHOEAL DISEASE CONTROL

In circumstances where diarrhoeal diseases have
been identified as a priority health problem, and a
commitment has been made to combat the problem, it
is necessary to decide how to reduce the mortality and

—637—

63S

R. G. FEACHEM ET AL.

morbidity they cause. The funds available in many
countries for the total primary health care package
are in the order of only USS 1-5 per capita per year,
and only a proportion of this may be devoted to the
control of diarrhoeal diseases. How should this
amount be spent?
This is not a new question. The Fifth Caribbean
Health Ministers’ Conference, meeting in Dominica
in 1973, called for a plan of action against diarrhoeal
diseases and malnutrition in children under 2 years
old. A group of experts was convened and recom­
mended a 10-point programme designed to achieve
specified reductions in mortality and morbidity.'1 This
plan of action, although embracing several important
aspects of diarrhoea control, did not provide
guidance to governments on the comparative cost­
effectiveness of the many different interventions
recommended.
More recently, Walsh & Warren (2) identified diar­
rhoeal diseases as a priority target for selective pri­
mary health care and emphasized oral rehydration as
the key intervention for the reduction of diarrhoeal
disease mortality. Chen (5) drew attention to the
importance of diarrhoea morbidity and to the need
for interventions (other than oral rehydration) that
are specifically addressed to this facet of the overall
diarrhoea problem.
It is widely agreed that oral rehydration, delivered
within a primary health care programme, is an effec­
tive and relatively inexpensive intervention for the
reduction of mortality due to dehydrating diarrhoeas.
Other interventions are also required, however, for
three main reasons. First, like all health services
delivered at the community level, oral rehydration
programmes face operational constraints that may
militate against the achievement of their full potential
impact. Second, oral rehydration is of limited use in
the treatment of chronic or dysenteric diarrhoeas and,
in areas of the world where these are responsible for a
considerable proportion of diarrhoeal disease mor­
tality, the effect of oral rehydration programmes on
the overall mortality from diarrhoeal diseases may be
modest. Third, oral rehydration can be expected to
have little or no impact on diarrhoea morbidity rates.
A multifaceted strategy is therefore preferable, in
which oral rehydration is but one of several anti-diarrhoea measures being implemented simultaneously,
with mutually reinforcing and complementary
impacts.
The Diarrhoeal Diseases Control (CDD) Pro­
gramme of the World Health Organization has, since
its inception in 1978, advocated the following fourpart strategy for diarrhoea control:
a Strategy and plan of action to combat gastroenteritis and mal­
nutrition in children under two years of age. Report of a Technical
Group Meeting on Malnutrition and Gastro-cnteritis, St Vincent,
8-11 January 1974 (Unpublished document).

— improved case management, with particular
emphasis on the early use of oral rehydration therapy
in acute diarrhoea and on appropriate feeding during
illness and convalescence;
— improved maternal and child health care, with
particular emphasis on breast-feeding, weaning prac­
tices, personal and domestic hygiene, and maternal
nutrition;
— improved use and maintenance of drinkingwater and sanitation facilities, and improved food
hygiene;
— detection and control of epidemics.

In the first years of the CDD programme, greatest
emphasis was placed upon oral rehydration as the
primary intervention for reducing diarrhoeal disease
mortality among young children (5). The CDD pro­
gramme has developed detailed recommendations for
oral rehydration therapy& and the production of oral
rehydration salts/ and has worked with governments
of Member countries in the planning, implementation
and evaluation of ora! rehydration and other diar­
rhoea control measures.d With the implementation of
CDD programmes in over 35 countries, it is now
appropriate to supplement the emphasis on oral rehy­
dration by developing, in detail, other interventions
for diarrhoea control and undertaking the necessary
field research and evaluation to establish their feasi­
bility and cost-effectiveness.
The CDD programme has therefore undertaken a
systematic and comprehensive review of the effective­
ness, feasibility and cost of the many possible anti­
diarrhoea interventions available for the reduction of
morbidity and/or mortality among children under 5
years of age. A classification of such interventions is
shown in Table 1. Its purpose is to guide and sys­
tematize the process of review and not to provide a
recommendation for diarrhoea! diseases control.
Each intervention listed in Table 1, and possibly
others that may subsequently be proposed, will be
reviewed using a standard format which places
emphasis on information concerning the effectiveness
of the intervention. If the intervention is known or
believed to be effective, available data on its feasi­
bility and cost are also presented.
As a result of these reviews, each intervention listed
in Table 1 will be assigned to one of three categories,
each having different requirements for follow-up
action by the CDD programme. First are inter­
ventions that are clearly shown to be effective, feas­
ible and affordable. For these, the next step will be for
b A manual for the treatment of acute diarrhoea. Unpublished
document WHO/CDD/SER/80.2, 1980.

r Guidelines for the production of ora! rehydration salts.
Unpublished document WHO/CDD/SER/80.3, 1980.

'' Manual for the planning and evaluation of national diarrhoeal
diseases control programmes. Unpublished document WHO/CDD/
SER/81.5, 1981.

INTERVENTIONS FOR DIARRHOEAL DISEASE CONTROL

Table 1. Potential interventions for reducing diarrhoeal
morbidity or mortality among children under five years of
age
I.

III.

A.

By reducing transmission of the pathogenic agents of diar­
rhoeal diseases
A.

By case management

1. Administration of oral rehydration in the home
Administration of oral rehydration at a medical
facility.
B.

B.

Non-oral rehydration therapy

Appropriate feeding

C

1. Promoting the appropriate feeding of children during
diarrhoeal illness and convalescence.
D.

Chemotherapy
D.

2. Administration of therapeutic agents at a medical
facility.

E.

2. Improving prenatal and postnatal
improve the quality of breast milk.

B.

nutrition

to

Child nutrition

Fly control
1. Control of flies, especially flies breeding in associ­
ation with human or animal faeces.

Ma ternal nutrition
1. Improving prenatal nutrition to reduce the incidence
of low birth-weight.

Control of zoonotic reservoirs
1. Control of infection of domestic and farm animals by
pathogens causing diarrhoea in man.

By increasing host resistance to infection and/or Illness and/
or death
A.

Food hygiene
1. Promoting improved practices for the preparation
and storage of foods, both commercially and in the
home, and especially emphasizing the hygienic prep­
aration of weaning foods.

1. Administration of therapeutic agents in the home.

II.

Personal and domestic h ygiene
1. Promoting specific features of personal and dom­
estic hygiene, such as hand-washing, by appropriate
educational campaigns.

1. Administration of rehydration by intravenous or
other routes at a medical facility.
C.

Water supply and excreta disposal
1. Constructing water supplies that improve the quality
and availability of water for domestic purposes, and
improved excreta disposal facilities; and providing
the necessary educational support to ensure use and
maintenance of these new facilities.

Ora! rehydration therapy
2.

639

By controlling and/or preventing diarrhoea epidemics

IV.

A.

Epidemic surveillance, investigation and control
1. Improving the ability to identify and investigate an
epidemic early in its course and the capacity to
implement effective control activities.

1. Promoting exclusive breast-feeding up to age 4-6
months and partial breast-feeding thereafter.

2. Improving weaning practices for children aged 4-18
months (introducing non-milk foods not later than
the sixth month, continuing breast-feeding for as
long as possible, and using nutritious and locally
available weaning foods).

C.

3.

Supplementary feeding to improve the nutritional
status of children aged 6-59 months.

4.

Promoting the use of growth charts by mothers as
an aid to proper child nutrition and child care.

Immunization
1. Rotavirus and/or cholera immunization (when effec­
tive and tested vaccines are available) of the child
and/or mother.
2. Measles immunization to reduce measles-associated
diarrhoea.

D.

Chemoprophylaxis
1. Chemoprophylaxis of children at special risk, such
as contacts of known cases, to reduce the incidence
and/or severity of diarrhoea.

the CDD programme to develop detailed guidelines
for their implementation within national primary
health care programmes and to promote any oper­
ational research needed to improve their delivery or
impact. Second are interventions for which there is
good theoretical evidence of effectiveness but in­
sufficient field experience to predict impacts precisely
or to judge feasibility and cost. For these the next step
will be for the CDD programme to promote field
research designed to fill the gaps in knowledge. The
results of this research will determine whether these
category 2 interventions are moved to category 1 or
category 3. Third are interventions which are shown
to be either ineffective, unfeasible or too costly. These
interventions will not be recommended by the CDD
programme as important elements of diarrhoeal dis­

640

R. G. FEACHEM ET AL.

ease control activities or priorities for diarrhoeal dis­
ease research. It is hoped that this review will help
focus the attention of governments, researchers and
international agencies on a few interventions of
known or suspected effectiveness which, if imple­
mented along with oral rehydration therapy, could
markedly reduce the rates of both morbidity and
mortality due to diarrhoeal diseases among young
children.
Reviews of some of the interventions listed in
Table 1 will be published in the Bulletin of the World
Health Organization. The first of these is published in

this issue (pp. 641-652) and deals with measles immu­
nization as an intervention for diarrhoeal disease con­
trol (II.C.2 in Table 1). It is hoped to publish in
forthcoming issues of the Bulletin reviews on
supplementary feeding (II.B.3), breast-feeding
(II.B.I), and personal and domestic hygiene (III.B. 1).
Readers having suggestions to make on the classifi­
cation of potential interventions in Table 1, or on the
individual reviews as they are published, are invited to
write to the Programme Manager, Diarrhoeal Dis­
eases Control Programme, World Health Organiz­
ation, 1211 Geneva 27, Switzerland.

RESUME
ANALYSE DES MODES D’INTERVENTION POSSIBLES DANS LA LUTTE CONTRE LES MALADIES DIARRHEIQUES

Les maladies diarrheiques constituent dans la plupart des
pays en developpemem 1’une des causes principals de mor­
bidity et de monalite chez les jeunes enfants. Du fait que
1’on dispose pour luiter centre ces maladies de modes
d’intervention efficaces, ils doivent representer un objectif
prioritaire des programmes de soins de same primaires
planifies ou mis en oeuvre dans nombre de pays. Des gouvernements et des organismes internationaux, y compris
1’Organisation mondiale de la Same, ont souligne que la rehydratation par voie orale etait une intervention essentielle
pour reduire la mortalite due aux maladies diarrheiques.
Mais il est necessaire cependam de disposer d’autres fa?ons
d’intervenir encore pour diminuer la morbidity, abaisser le

nombre de deces que ne peut empecher la rehydratation par
voie orale et mettre au point une approche conjuguye, dans
laquelle la ryhydratation orale represente 1’une seulemem
des diverses mesures amidiarrheiques appliquees simultanement el dont les effets se renforcem et se completem
mutuellement. L’article qui suit contient une classification
des interventions possibles pour lutter contre la morbidity
et/ou la mortalite dues aux maladies diarrheiques chez les
enfants de moins de 5 ans, et annonce le d6but d’une sdrie
d’analyses de ces strategies, dont la premiere — sur la vacci­
nation antimorbilleuse — est publiye dans ce meme numyro
du Bulletin de [’Organisation mondiale de la Sant#.

REFERENCES
1.

Evans, J. R.et al. Health care in the developing world:
problems of scarcity and choice. New England journal
of medicine, 305:1117-1127 (1981).
2. Walsh, J. A. & Warren, K. S. Selective primary
health care: an interim strategy for disease control in
developing countries. New England journal of medi­
cine, 301: 967-974 (1979).
3. Snyder, J. D. & Merson, M. H. The magnitude of the
global problem of acute diarrhoeal disease: a review of
active surveillance data. Bulletin of the World Health
Organization, 60: 605-613 (1982).

Mahalanabis, D. et al Oral rehydration therapy:
recent advances. World health forum, 2: 245-249
(1981).
5. Chen, L. C. Control of diarrheal disease morbidity and
mortality: some strategic issues. American journal of
clinical nutrition, 31: 2284-2291 (1978).

4.

CH 3- h
© World Health Organization 1984

Bulletin of the World Health Organization, 62 (2): 271 -291 (1984)

Interventions for the control of diarrhoeal diseases
among young children: promotion of breast-feeding
R. G. Feachem1 & M. A. Koblinsky12
The literature on the relative risks of diarrhoea morbidity to infants on different
feeding modes suffers from several methodological problems. Thirty-five studies from 14
countries were reviewed; 83% ofstudies found that exclusive breast-feeding was protective
compared to partial breast-feeding, 88% that exclusive breast-feeding was protective com­
pared to no breast-feeding, and 76% that partial breast-feeding was protective compared to
no breast-feeding. When infants receiving no breast milk are contrasted with infants on
exclusive or partial breast-feeding, the median relative risks are 3.0 for those aged 0-2
months, 2.4 for those aged 3-5 months, and 1.3-1.5 for those aged 6-11 months. Above 1
year of age no protective effect of breast-feeding on diarrhoea morbidity is evident. When
infants receiving no breast milk are contrasted with those on exclusive breast-feeding,
median relative risks are 3.5-4.9 in thefirst 6 months of life. The literature does not suggest
that the relative risks of diarrhoea morbidity for bottle-fed infants are higher in poor
families than in more wealthy families. The protective effects of breast-feeding do not
appear to continue after the cessation of breast-feeding. There is evidence of considerably
increased diarrhoea severity among bottle-fed infants.
There is a limited, and mostly pre-1950, literature on the relative risks of diarrhoea
mortality to infants on different feeding modes. Nine studies from 5 countries were
reviewed, most of which showed that breast-feeding protects substantially against death
from diarrhoea. When infants receiving no breast milk are contrasted with those on
exclusive breast-feeding, the median relative risk of death from diarrhoea during thefirst 6
months of life is 25. When partially and exclusively breast-fed infants are contrasted, the
median relative risk of death from diarrhoea is 8.6.
Breast-feeding can be promoted by changes in hospital routine and by giving
information and support to mothers. A review of21 studiesfrom 8 countries shows that, by
such promotion, the most likely reductions in the prevalence of non-breast-fed infants are
40% among infants aged 0-2 months, 30% among those aged3-5 months, and 10% among
those between 6 months and 1 year old. Theoretical calculations based on these data show
that such promotion can reduce diarrhoea morbidity rates by 8-20% and diarrhoea
mortality rates by 24-27% in the first 6 months of life. For children aged 0-59 months,
diarrhoea morbidity rates would be reduced by 1-4% and mortality rates by 8-9%. A recent
study in Costa Rica has documented a substantial impact of breast-feeding promotion on
neonatal diarrhoea morbidity and mortality, and on diarrhoea morbidity in infants aged
0-5 months. The Costa Rican data show good agreement with the theoretical computations
presented in this paper.
Several important aspects of breast-feeding and diarrhoea remain to be clarified by
research. However, the need for this research should not delay action to promote breast­
feeding and to monitor its effects upon feeding practice and upon diarrhoea.

The long debate on the merits of breast-feeding
initially focused on the differences in mortality rates
1 Reader in Tropical Public Health Engineering and Head of
Department of Tropical Hygiene, London School of Hygiene and
Tropical Medicine, Keppel Street, London WC1E 7HT, England.
Requests for reprints should be sent to Programme for Control of
Diarrhoeal Diseases, World Health Organization, 1211 Geneva 27,
Switzerland.
2 Scientific Coordinator, International Centre for Diarrhoeal
Disease Research in Bangladesh, Dhaka, Bangladesh.

4401

between breast-fed and bottle-fed infants. Since
about 1930, and especially since 1955, increased
attention has been paid to differential morbidity
rates. Some studies on mortality and morbidity in
relation to feeding mode singled out particular infec­
tious causes of death or illness — most commonly,
diarrhoeal and respiratory diseases. The literature
on the relationships between breast-feeding and

—271—

R. G. FEACHEM & M. A. K0BL1NSKY

diarrhoea is now substantial and permits an assess­
ment of breast-feeding promotion as an intervention
for the reduction of diarrhoea morbidity or mortality
in infants. Several recent studies and reviews of
breast-feeding provide a useful background to the
present, more focused analysis (2, 17, 47, 51, 64, 66,
81, 86)“'c of the role of breast-feeding promotion in
diarrhoeal disease control. This review is the third in a
series of reviews of potential anti-diarrhoea inter­
ventions now being published in the Bulletin of the
World Health Organization (19-21).

EFFECTIVENESS

For breast-feeding promotion to be an effective
diarrhoea control intervention, it must be true that:
either

breast-fed infants have
reduced diarrhoea
morbidity rates, mortality
rates, or severity

hypothesis
1

and
the prevalence of breast­
fed infants can be
increased by appropriate
promotional activities

hypothesis
2

Definitions
th review of breast-feeding and diarrhoeal disease
rates requires clear-cut definitions of the various
feeding modes that are to be contrasted. The number
of feeding modes defined should be small (say, 3 to 5)
in order to increase the sample size of infants on each
feeding mode and in order that the operational signi­
ficance of the comparisons will be apparent. How­
ever, to assign all of the many permutations of infant
feeding practice to only 3-5 categories inevitably
introduces a degree of imprecision to the definitions.
In this review the following three categories have been
adopted:

— exclusive breast-feeding, which applies to
infants receiving only breast milk (and is thus not
usual in infants over 6 months old);
— no breast-feeding, which applies to infants
receiving no breast milk;
— partial breast-feeding, which applies to infants
who receive breast milk plus other milk or foods.
The feeding modes described in the literature are
assigned to one of these three categories. In several
studies the feeding modes have not been clearly
defined or do not correspond to any of the categories
listed above. In these cases, the feeding mode is
designated so as to minimize the computed protective
effect of breast-feeding.

Methodological problems

or

the promotion of breast­
feeding can reduce
diarrhoea morbidity rates,
mortality rates, or severity
in infants

Hypothesis 1. Breast-fed infants have reduced
diarrhoea morbidity rates, mortality rates, or
severity

hypothesis
3

Most of the literature on this topic is addressed to
hypothesis 1 or 2. The potential effectiveness of
breast-feeding promotion would be suggested by a
demonstration either of the correctness of hypotheses
1 and 2 or of the correctness of hypothesis 3. The
evidence for and against these hypotheses is examined
below.
“ Evensen, S. Relationship between infant morbidity and
breast-feeding versus artificial feeding in industrialized countries:
a review of the literature. Copenhagen, WHO Regional Office for
Europe, 1982 (unpublished document 1CP/NUT 010/6).

6 Joint WHO/UNICEF meeting on infant and young child
feeding, Geneva, 1979 (unpublished document).
c WHO/UNICEF. Infant and young child feeding: current
issues, Geneva, 1981 (unpublished document).

The studies on breast-feeding and diarrhoea re­
viewed here made use of different methods for the
collection and analysis of their data. Studies that were
judged to have serious methodological flaws were
rejected but the quality of those that were included
varies widely. The most commonly encountered
problem other than definition of feeding mode (see
above) was failure to control possible confounding
variables. The age of the infant is a confounding
variable encountered in all the studies; older infants
are less likely to be exclusively breast-fed than
younger infants and also have different diarrhoeal
disease rates for reasons other than feeding mode.
Emphasis has therefore been placed here on studies
that analyse breast-feeding and diarrhoea rates by
narrow age ranges (e.g., 0-2 months, 3-5 months,
etc.); studies that examine a wider age range of
infants, e.g., aged 0-11 months, are judged to be of
little value.
A second common set of confounding variables is
socioeconomic status and child care. Mothers who
breast-feed may be more or less educated and

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

wealthy, and may take more or less care of their child
than other mothers. Two distinct patterns emerged
from the studies that analysed the effect of these con­
founding variables. In some poor communities,
especially in developing countries and in the devel­
oped countries before 1930, breast-feeding was more
common among lower socioeconomic strata. In such
communities the confounding socioeconomic vari­
ables tend to increase the diarrhoea rates and there­
fore minimize the apparent protective effect of breast­
feeding. In an uncontrolled study in such communi­
ties the protection afforded by breast-feeding may be
underestimated. In certain more prosperous com­
munities. especially in developed countries, breast­
feeding is at present more common among the middle
classes because it is fashionable. Here the confound­
ing socioeconomic variables tend to decrease the
diarrhoea rates and thus an uncontrolled study may
overestimate the protective effect of breast-feeding.
Several of the more recent studies that were reviewed
controlled for confounding socioeconomic and child
care variables, and such control should be regarded as
essential for any future studies of breast-feeding and
diarrhoea. These and other methodological problems
inherent in studies of breast-feeding and health are
reviewed elsewhere (9, 70, 84).d
Mechanisms of protection
If breast-fed infants experience less diarrhoeal ill­
ness or death than other infants, it may be due to one
or more of the following factors:
— the immunological and antimicrobial properties
of breast milk;
— the “bifidus factor” (exclusively breast-fed
infants have an intestinal flora composed largely of
Gram-positive anaerobic bacteria (Bifidobacterium
species), which may inhibit colonization by Gram­
negative facultative species such as Escherichia coli);
— infants receiving bottle-milk feeds are at risk
from contamination of the milk, the bottle, or the
teat, and infants receiving solid foods are at risk from
contamination of the food (these risks may apply es­
pecially to bacterial pathogens that multiply in milk
and some foods);
— breast-fed infants may have a better nutritional
status than other infants, and thus a lesser risk of
death from diarrhoea.

The protective mechanisms of breast-feeding are a
complex subject, in which there is substantial ongoing
research. This is especially true of the first of the four
items listed above. It is probable that the mechanisms
of protection, and their relative importance, vary by
pathogen and by the age of the infant. It is not the
purpose of this paper to review the literature on prod See footnote a on page 272.

273

tective mechanisms, and the interested reader is refer­
red to other publications (3, 4, 14, 17, 26, 30, 33,
40-42, 55, 58, 60, 65, 83).
Morbidity

Annex 1 lists 35 studies in 14 countries from which
the relative risk of diarrhoeal illness for infants on one
feeding mode, compared to infants on another
feeding mode, can be computed. Both methods and
findings vary greatly among these studies. For partial
breast-feeding compared to exclusive breast-feeding,
30 relative risks are computed of which 25 (83%) are
> 1. For no breast-feeding compared to exclusive
breast-feeding, 25 relative risks are computed of
which 22 (88%) are > 1. For no breast-feeding com­
pared to partial breast-feeding, 45 relative risks are
computed of which 34 (76%) are >1. Thus the
pooled results from 35 studies indicate that in most
circumstances breast-feeding protects against diar­
rhoea morbidity. More analysis on these pooled data
is not useful since the results are strongly influenced
by the age of the infants under study.
The age-specific relative risks derived are sum­
marized in Fig. 1 and 2, which show clearly the effect
of age. When infants receiving no breast milk are
contrasted with infants on exclusive or partial breast­
feeding (Fig. 1), the median relative risks are 3.0 for
ages 0-3 months, 2.4 for ages 3-5 months,e and
1.3-1.5 for ages 6-8 and 9-11 months. Above 1 year
of age, no protective effect of breast-feeding on
f The 0-3-months age group includes studies of both the 0-2months group and the 0-3-months group. This age group therefore
overlaps by one month (month 3) with the 3-5-months age group.
These remarks also apply to Fig. 1 and 2.

Fig. 1. Median relative risks of diarrhoea morbidity for
infants receiving no breast-feeding compared to infants
with partial or exclusive breast-feeding (data from
Annex 1).

274

R. G. FEACHEM & M. A KOBLINSKY

The discussion above deals only with the protective
effects of a particular feeding mode during the period
when this feeding mode is applied. It is conceivable
that breast-feeding confers some protection against
diarrhoea after breast-feeding has been discontinued.
Few studies present data that allow this possibility to
be analysed. The findings of Ferguson et al. (23) in
New Zealand are summarized in Table 1. All infants
described in Table 1 had ceased breast-feeding and
their period prevalences of diarrhoea are compared
according to their past experience of breast-feeding.
The data are weakened by the use of period preva­
lence rather than incidence but, none the less, the
results do not suggest that a longer duration of breast­
feeding was associated with a lower period prevalence
of diarrhoea. This cessation of protection after dis­
continuation of breast-feeding has been reported also
in studies of all significant illness episodes (9) and of
total hospital admissions (15).
Severity

Fig. 2. Median relative risks of diarrhoea morbidity by
feeding mode (data from Annex 1).

diarrhoea morbidity is evident. When infants receiv­
ing no breast milk are contrasted with those on exclu­
sive breast-feeding (Fig. 2), the median relative risks
are 3.5-4.9 in the first 6 months of life. Beyond 6
months of age, exclusive breast-feeding is not a nutri­
tionally recommended feeding mode and may be a
risk factor, rather than a protective factor, for diar­
rhoea. This possibility is illustrated by the data from
Ethiopia and Uganda (Annex 1).
It is possible that the relative risks of diarrhoea for
bottle-fed infants are greatest in families of lowest
socioeconomic status where diarrhoea incidence is
high and nutritional status is low, and where there are
many opportunities for the contamination of bottle
milk. The data in Annex 1 were analysed (not shown)
to test this possibility and gave no grounds for
supposing that the relative risks of diarrhoea for
bottle-fed infants are lower in more wealthy families.
Breast-feeding appears to protect against diarrhoea
irrespective of the levels of hygiene. This suggests
either that the main protective mechanisms are the
immunological and antimicrobial properties of the
breast milk together with the “bifid” flora of the gut
or that contamination of bottle milk is an important
cause of infant diarrhoea even in families of upper
and middle socioeconomic status. The possible lack
of association between the relative protection af­
forded by breast-feeding and socioeconomic status
should be further studied.

Only a few studies provide data from which the
relationship between breast-feeding and diarrhoea
severity may be assessed. Table 2 summarizes 5
studies of case-fatality ratios. Four of these (from
England in the 1930s, USA in the 1920s, and Rwanda
recently) show major differences in case-fatality
ratios by feeding mode. These results could be con­
founded by other clinical problems of the children
who died; those not breast-fed could have been under­
nourished or chronically ill, for instance. The data
from Birmingham, England, in the 1930s (73) show
that the prevalences of undernutrition (<80%
Table 1. Period prevalence of treated diarrhoeas and all
diarrhoeas in specified age groups, by duration of
previous breast-feeding"

Age group
(months,
inclusive)

4-11

12-23

Duration
of previous
exclusive
breast-feeding
(months)

Treated
*diarrhoeas

All
diarrhoeas'

Period prevalence (%)

0

22

44

0-3

25

48

0

29

63

0-3

32

64

4-7

31

64

8-11

29

65

" Data from Ferguson et al. (23).
h Diarrhoeas reported to a medical practitioner or hospital for
treatment.
r All diarrhoeas, including those managed at home.

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

275

Table 2. Case-fatality ratios for diarrhoea cases in four countries, by feeding mode

Country
and place

Date
of study

Socio­
economic
status

Age
group
(inclusive)

Total
number of
diarrhoea
cases
(deaths)

Part BF°

No BF°

Ratio of
case­
fatality
ratios
(No BF/
Part BF)

14

15

1.1

13

Case-fatality ratios (%)
Excl. BF°

Reference

Canada
*
Toronto

1939

?

0-11 mo.

England
Liverpool

1936

Lower/
middle
?

3-26 wk

130 (9)

5

9

1.8

69

?

0-9 mo.

500 (240)

26

77

3.0

73

Rwanda
*
Kigali

1977-78

?

0-23 mo.

849 (95)

7

22

3.1

50

USA
Chicago

1924-29

Lower

0-3 mo.
4-8 mo.

1877 (22)

0.9
0.3

20
2

22
6.7

35
36

*
Birmingham

314 (46)

0.6

“ For definitions of feeding modes, see footnote b to Annex 1
* Study of hospitalized diarrhoea cases only.

weight-for-age) among non-breast-fed infants who
died of diarrhoea, and among all diarrhoea cases,
were 41% and 37% respectively, suggesting that the
apparent effect of feeding mode on case-fatality ratio
(Table 2) was not confounded by malnutrition.
Grulee (35), commenting on the Chicago (1924-29)
study, wrote that “1924 to 1929 were the years of
plenty” and that “undernutrition does not enter into
the picture”. The Chicago study used data from the
surveillance of over 20 000 infants. Such data are less
likely to be confounded by undernutrition, and much
less likely to be confounded by chronic illness, than
data derived from hospitalized cases only.
An investigation of shigellosis in Bangladesh (77)
found that severity was related to feeding mode.
Breast-fed and non-breast-fed children under 2 years
old with shigellosis were compared; among the breast­
fed group, fewer required intravenous therapy (16%
vs 38%) and fewer were admitted as inpatients (5% vs
19%).
Breast-feeding may play a role in reducing the
nutritional consequences of diarrhoea episodes in
young children. A study of children aged 6-35
months, hospitalized with acute watery diarrhoea in
Bangladesh, showed that those being breast-fed had
calorie and protein intakes (per kg of body weight) 1.5
and 2.5 times greater, respectively, than those not
breast-fed (45).
More information is required on the effects of
feeding mode on diarrhoea severity. Such studies
should control for age, nutritional status, and chronic
illness and should investigate the severity of diar­
rhoeas of known etiology.

Mortality
Annex 2 lists 9 studies in 5 countries from which the
relative risks of diarrhoea mortality for infants on one
feeding mode, compared to infants on another
feeding mode, can be computed. The data are less
extensive than those on morbidity (Annex 1) and do
not permit such detailed analysis by age. For partial
breast-feeding compared to exclusive breast-feeding,
7 relative risks are computed, ranging in value from
1 to 10. For no breast-feeding compared to exclu­
sive breast-feeding, 13 relative risks are computed,
ranging in value from 3 to 43. For no breast-feeding
compared to partial breast-feeding, 9 relative risks are
computed, ranging in value from 2 to 19. Thus the
pooled results from 9 studies suggest that breast­
feeding may protect substantially against death from
diarrhoea.
The age-specific relative risks of diarrhoea mor­
tality, from 2 studies in which age-specific analysis
is possible, are presented in Fig. 3. The relative risk
of death from diarrhoea for non-breast-fed infants
is greatly increased in the early months of infancy
compared to the later months. The relative risks of
diarrhoea mortality for infants 0-5 months old are
summarized in Fig. 4 by feeding mode. When infants
receiving no breast milk are contrasted with those on
exclusive breast-feeding the median relative risk of
death from diarrhoea during the first 6 months of life
is 25. When infants on mixed feeding modes (partial
breast-feeding) are contrasted with those on exclusive
breast-feeding the median relative risk of death from
diarrhoea is 8.6. It must be cautioned that these

R. G. FEACHEM & M. A. KOBLINSKY

276
30 -

Fig. 3. Relative risks of diarrhoea mortality for infants
receiving no breast-feeding compared to infants exclu­
sively breast-fed (data from Annex 2).
No breast-feeding
vs
Exclusive breast-feeding
Partial breast-feeding
vs
Exclusive breast-feeding

No breast-feeding
vs
Partial breast-feeding
Number of results

-0-5
Age in months

Fig. 4. Median relative risks of diarrhoea mortality by
feeding mode (data from Annex 2, including studies of
0-2, 3-5 and 0-5-months age groups).

results are derived from only two studies (62, 69) and
mainly from the study of Newman (62). Comparison
of Fig. 2 and 4 shows that the relative risks of diar­
rhoea mortality by feeding mode are 2-6 times greater
than those of diarrhoea illness. This implies a similar
difference in the case-fatality ratios by feeding mode.

Table 2 shows the ratios of no breast-feeding to
partial breast-feeding case-fatality ratios to be 1.1,
1.8, 3.0, 3.1, 6.7 and 22.
As with the case-fatality ratios (Table 2), the
mortality rates by feeding mode (Annex 2) could be
confounded with undernutrition or chronic illness.
A factor that causes the abandonment of breast­
feeding, such as a chronic illness, could also increase
the risk of death from diarrhoea. Four studies listed in
Annex 2 controlled for these, or related factors, or
excluded infants who were disadvantaged from birth
owing to prematurity, low birth-weight, or congenital
defects. In addition, four of the studies in Annex 2
derived their mortality rates from records of over 100
deaths from diarrhoea. With sample sizes of this
magnitude it is unlikely that rare events such as
chronic illness would substantially alter the computed
effect of feeding mode.
Despite these comments, the data from which to
draw conclusions about the relative risks of diarrhoea
mortality by feeding mode are limited in extent and
quality. The studies are old (only one since 1947); all
except one are from what are now wealthy and tem­
perate countries (Canada, England, Sweden, and the
USA); they provide poor breakdown of mortality
rates by age; and they do not adequately control for
potentially confounding variables. The age of the
studies is of concern first because their designs are
inadequate as judged by today’s epidemiological
standards, and secondly because the non-breast-fed
infants were not receiving modern infant milk for­
mulae. The importance of this last point is uncertain.
Of the four potential protective mechanisms listed
above, only nutritional status is likely to be affected
by the type of breast-milk substitute.
Reliable data on the effect of feeding mode on
diarrhoea mortality rates in developing countries
today are not available. The collection of such data
through correctly designed studies is a priority item
for research programmes in both diarrhoeal diseases
and maternal and child health. Evidence that such
studies will find a significant relative risk of diarrhoea
mortality among bottle-fed infants comes from
studies of overall infant mortality by feeding mode.
For instance, a study in north-eastern Brazil found
that, when numerous confounding variables were
controlled, children who were never breast-fed were
1.7 times more likely to die in infancy than other
children (29).
Etiology-specific diarrhoea and breast-feeding
There is a growing literature on substances in
human colostrum and milk that act against specific
agents of diarrhoea. Most attention has been focused
on antibodies against rotavirus (71, 72, 90), anti­
bodies against Escherichia coli and its toxins (1, 28,

277

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

48), and antibodies against Vibrio cholerae and its
toxin (55, 72).
Little is known of the relationships between feeding
mode and the epidemiology of diarrhoeas of known
etiology. High relative risks of cholera in Bahrain,
and of salmonellosis in Arkansas, USA, in infants
receiving no breast milk have been documented
(Annex 1). In a study of 5-day-old babies in London,
England, the prevalences of rotavirus infection
among breast-fed and non-breast-fed infants were
22% and 58% respectively (6). Only 14% of infected
breast-fed infants excreted > 109 virus particles per
gram of faeces, compared to 52% of non-breast-fed
infected infants. In a study of children under 1 year
old, hospitalized with diarrhoea in Mexico City,
Mexico, 5% of those with rotavirus infection were
breast-fed compared to 16% of those with other
diarrhoeas (76). In a study of children under 2 years
old with diarrhoea in Dhaka, Bangladesh, the propor­
tion who were breast-fed was 59% in those with
shigellosis compared to 78% in those with other
diarrhoeas (77). A similar association with breast­
feeding was not found for the other common enteric
pathogens.
This information is incomplete and partially con­
tradictory. It is uncertain whether the protection
against diarrhoea morbidity and mortality associated
with breast-feeding (Fig. 1-4) extends to all diar­

rhoeas or is due to high protection against some
agents and low protection against others. Studies on
feeding mode in relation to the incidence and severity
of specific diarrhoeas are needed to clarify the
situation and to guide intervention policy towards
geographical areas in which breast-feeding may be
especially important.
Hypothesis 2. The prevalence of breast-fed infants
can be increased by appropriate promotional acti­
vities

The most recent and comprehensive account of
patterns of breast-feeding is provided by a WHO col­
laborative study of 22 857 mothers in 9 countries
(86). The reader concerned with the details of breast­
feeding trends and practices should study this report.
Simplified data on three contrasting patterns of
breast-feeding (dubbed here patterns A, B and C)
have been abstracted and are set out in Table 3.
Pattern A may be found in relatively urbanized and
wealthy communities in developing countries and in
some developed countries. It is a pattern where non­
breast-feeding is prominent and it may be an increas­
ingly common pattern in many developing countries,
especially in urban areas. Pattern C is a pattern of pre­
dominant breast-feeding and may be found in many
poor and traditional societies in developing countries.

Table 3. Three contrasting patterns for breast-feeding1'

Pattern
of breast­
feeding b

A

B

C

Countries and the
socioeconomic strata of the population studied

Urban
poor

All
strata

Upper/
middle

Hungary
Sweden

Chile
Ethiopia
Guatemala
Nigeria
Philippines

Chile

India
Nigeria

Guatemala
Philippines

Zaire

Ethiopia
India
Nigeria
Zaire

Rural
poor

Chile
Philippines

Ethiopia
Guatemala
India
Nigeria
Zaire

Type of
feedingf

Prevalence (%) of breast­
feeding by age of infant

0-2 mo.

3-5 mo.

6-11 mo.

Excl. BF

30

10

0

Part BF

45

35

15

No BF

25

55

85

Excl. BF

50

30

5

Part BF

40

45

55

No BF

10

25

40

Excl. BF

75

50

20

Part BF

25

45

70

No BF

0

5

10

" All data drawn from Fig. 3 and Tables 3 and A5 of the WHO breast-feeding study (861. The data have been greatly simplified for
the purposes of this analysis on breast-feeding and diarrhoea; for the details of breast-feeding patterns, please consult the full report
(88).
b Patterns A, B, and C may be expected in 'modern', 'transitional', and 'traditional' communities, respectively. It is assumed that
these communities suffer low, medium, and high incidences of diarrhoea, respectively (see Tables 6 and 8).
c For definitions of feeding modes, see footnote b to Annex 1.

278

R. G. FEACHEM & M. A. KOBLINSKY

Pattern B is intermediate between patterns A and C
and may be found in societies that are transitional, in
terms of both social norms and wealth. This threepart categorization of breast-feeding patterns is a
considerable simplification of the actual global
picture as described by WHO (86). This simplifi­
cation is necessary for the compulations on breast­
feeding promotion and diarrhoea reduction that
follow below. The reader interested in a specific
community, where the breast-feeding pattern is
known, can insert local data in place of those given in
Table 3 and repeal the computations.
In some countries, most notably certain developed
countries (with breast-feeding pattern A, Table 3),
growing public awareness of the advantages of breast­
feeding has caused substantial increases in the preva­
lence of breast-fed infants over a short time period
(64). In Oslo, Norway, the prevalence of breast­
feeding in infants aged 3 months was 14% in 1972-73
and 31 % in 1974-75. In Copenhagen, Denmark, the
median duration of breast-feeding doubled, from 2 to
4 months, between 1975 and 1977. In the USA the
prevalence of infants who were only breast-fed rose
from an all-time low of 26% in 1973 to 54% in
1980.
Promotion of breast-feeding can reduce the
prevalence of non-breast-fcd infants substantially.
Table 4 summarizes the impact on breast-feeding
of local breast-feeding promotions in Brazil, Czech­
oslovakia, England, Guatemala, Scotland, Singa­
pore, Sweden, and the USA. Because of the very
varied pre-promotion levels of breast-feeding, the
impacts are expressed as percentage reductions in the
prevalences of infants receiving no breast milk. In the
computations (see below) the most likely reductions
in the prevalences of non-breast-fed infants of
various ages due to breast-feeding promotion are
taken as 40% for the 0-2-months age group, 30% for
the 3-5-months age group, and 10% for the 6-11months age group (Table 4).
Table 4, The effectiveness of breast-feeding promotion
programmes in reducing the prevalence of infants
receiving no breast milk"
Age of
infant
(months,
inclusive)

No. of
results

% reduction in prevalence
of non-breast-fed infantsh

Range

Median

0-2

23

8-100

42

3-5

8

7-43

28

6-11

6

2-43

11

0 Data derived from Winikoff & Baer (85) (summary of 20
studies in 8 countries) and from Hardy et al. (43).
h If the prevalences of non-breast-fed infants of a given age
before and after the promotion are 70% and 50% respectively,
the % reduction is 28.6%.

Hypothesis 3. The promotion of breast-feeding can
reduce diarrhoea morbidity rates, mortality rates,
or severity in infants
Il is possible to calculate the theoretical impacts of
breast-feeding promotion using the data assembled
above on hypotheses 1 and 2. Median relative risks
of both morbidity and mortality by age have been
derived (Fig. 1-4), three contrasting patterns of
breast-feeding have been described (Table 3), and
the impact of breast-feeding promotion on breast­
feeding has been reviewed (Table 4). The next step is
to characterize the impact on breast-feeding of breast­
feeding promotions having three levels of success:
high impact, medium impact, and low impact. The
consequences to breast-feeding patterns of these three
levels of success are quantified in Table 5. The data on
medium impact are derived from Table 4, while high
and low impacts are assumed to influence twice and
one half, respectively, of the proportion of target
mothers.
From the data assembled in Fig. 1-4 and Tables 3-5
it is possible to compute the impact on age-specific
diarrhoea morbidity and mortality rates of breast­
feeding promotions with a given level of success (say,
medium impact) in a community of known breast­
feeding pattern (say, pattern A). For morbidity rates,
key assumptions are presented in Table 6 and the
reductions in age-specific diarrhoea morbidity rates
are set out in Table 7. Considering breast-feeding
promotions of medium impact, diarrhoea morbidity
rates in the first 3 months of infancy may be reduced
by 8-20%, and in the second 3 months of life by
9-17%, with the highest percentage reductions being
achieved in communities where breast-feeding is
initially least common (pattern A). Negligible reduc­
tions in diarrhoea morbidity rates are achieved among
children over 6 months old. The computed reductions
in total diarrhoea morbidity for the first five years of
life are only 1-4%.
For mortality rates, key assumptions are presented
in Table 8 and the reductions in age-specific diarrhoea
mortality rates are set out in Table 9. Considering
breast-feeding promotions of medium impact, diar­
rhoea mortality rates in the first 6 months of infancy
may be reduced by 24-27%, and diarrhoea mortality
rates for the first five years of life by 8-9%. Zero
reductions in diarrhoea mortality rates are computed
for children over 6 months old (Table 9) because of
the conservative assumption of a relative risk of 1.0
for this age group (see footnote d to Table 8). The
impacts on children aged 0-59 months are similar,
whatever the initial level of breast-feeding (patterns
A, B and C) because of the high diarrhoea mortality
rates assumed for 0-5-month old infants in com­
munities having breast-feeding pattern C (Table 8).
The computed morbidity and mortality reductions
(Tables 7 and 9) for the total 0-59-months age group

279

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

Table 5. Data on the three levels of impact of breast-feeding promotion programmes"
Percentage changes in breast-feeding prevalence by age of infant and type of feeding

6-11 mo.
*

3-5 mo.

0-2 mo.
Level of
impact

From Part BF
to Excl. BF

From No BF
to Part BF

From No BF
to Excl. BF

From Part BF
to Excl. BF

From No BF
to Part BF

40

80

30

30

60

20

20

40

15

15

30

10

10

20

7.5

7.5

15

5

From No BF
to Part BFf

From No BF
to Excl. BF

High

40

Medium

20

Low

10

“ Explanation : Consider a promotion with medium impact on the 0 - 2 mo. age group. It is assumed that the behaviour of 40% of the
target mothers can be changed (Table 4). Thus 40% of Part BF switch to Excl. BF and 40% of No BF switch to Part BF or Excl. BF
(20% to each mode). For low impact it is assumed that only 20% of target mothers can be influenced, while for high impact it is
assumed that 80% of target mothers can be influenced.
b Since most authorities recommend supplementary feeding of infants over 5 months, it is assumed that the breast-feeding
promotion would encourage Part BF and not Excl. BF in this age group.
e For definitions of feeding modes, see footnote b to Annex 1.

Table 6. Assumptions made in calculating the impact of breast-feeding promotions on diarrhoea morbidity

Age group
inclusive)

Episodes of diarrhoea
per child per year, by breastfeeding pattern (Table 3IC

Relative risk of diarrhoea
morbidity by feeding mode0

Proportion (%) of
0-59 month-old
children falling
stated age group

Part BF vs
Excl. BF
*

No BF vs
Excl. BF
*

A

B

C

0-2

2.0

4.0

1.0

2.0

3.0

3-5

2.0

4.0

1.0

2.0

3.0

6

6-11

1.3

1.6

1.5

30

4.0

11

12-59

1.0

1.0

0.5

1.5

2.0

76

7

° See Annex 1 and Fig. 1 and 2.
b For definitions of feeding modes, see footnote b to Annex 1.
c Estimates derived from Snyder & Merson (74).

depend on the assumptions made on age-specific
morbidity and mortality rates in Tables 6 and 8 (see
footnotes b to Tables 7 and 9). These rates are based
on the assumption that breast-feeding pattern A
will most commonly be found in relatively wealthy
communities having relatively low diarrhoea rates,
whereas pattern C will be typical of very poor com­
munities having the highest diarrhoea rates. Com­
munities with breast-feeding pattern B are assumed to
have intermediate diarrhoea rates. This generaliz­
ation does not reflect the situation in all countries.
There are areas having low breast-feeding prevalence
but high diarrhoea rates (for instance, in urban slums
in some Latin American cities), and there are areas
having high breast-feeding prevalences but low diar­
rhoea rates (for instance, in Sweden).
The former case, of low breast-feeding prevalence
and high diarrhoea rates, is of particular interest.

A community was defined having breast-feeding
pattern A (Table 3), but with diarrhoea morbidity and
mortality rates the same as those assumed for com­
munities with breast-feeding pattern C (Tables 6'and
8). Such a community might be found in an urban
slum in Latin America. The age-specific percentage
reductions in diarrhoea morbidity and mortality rates
due to breast-feeding promotion in this community
are exactly as calculated in Tables 7 and 9 for a
community with breast-feeding pattern A. This is
because the age-specific morbidity and mortality rate
reductions are not dependant on the assumed
diarrhoea rates; they depend rather on the relative
risks by feeding mode and it has been assumed
throughout these computations that these relative
risks are the same in communities having breast-feed­
ing patterns A, B and C. The diarrhoea rate
reductions for the total 0-59-months age group are

R. G. FEACHEM & M. A. KOBLINSKY

280

Table 7. Percentage reductions in diarrhoea morbidity
rates, by age of child, due to breast-feeding promotions
of varying effectiveness

Pre-intervention
pattern
of breast­
feeding"

Age of
children
(months,
inclusive)

A

B

C

Reduction (%) in diarrhoea
incidence by effectiveness
of breast-feeding promotion b

High
*"impact

Medium
*"
impact

Low
impact *"

0-2

39

20

10

3-5

34

17

8

6-11

4

2

1

12-59

0

0

0

0-59

8

4

2

0-2

31

15

8

3-5

29

14

8

6-11

2

1

0

12-59

0

0

0

0-59

5

2

1

0-2

16

8

4

3-5

21

9

6

6-11

1

0

0

12-59

0

0

0

0-59

3

1

1

0 See Table 3.
b For reductions in the 0-2, 3-5, 6-11, and 12-59 months
age groups the given percentage reductions are applicable,
whatever the incidence of diarrhoea or the age structure in the
community. The percentage reductions in the 0-59 months age
group are calculated using the assumptions on diarrhoea inci­
dence and age structure given in Table 6.
e See Table 5.

dependent on the age-specific diarrhoea rates
assumed. However, similar percentage reductions are
computed for the urban community in Latin America
as for the more typical pattern A community (Tables 7
and 9), because the higher diarrhoea rates apply not
only to the first 6 months of life when changes in
breast-feeding are effective, but also to the next 4.5
years when they are not.
Clearly the absolute, rather than the proportional,
reduction in diarrhoea morbidity and mortality rates
will be higher in communities having higher initial
diarrhoea rates. For instance, a community having
breast-feeding pattern A and diarrhoea rates for that
pattern as shown in Table 6, will have 67 diarrhoea
episodes per 100 children under 5 years old per year.
Following a breast-feeding promotion of high im­
pact, there will be 61.5 episodes per year, an 8%
reduction (Table 7) with only 5.5 episodes averted
annually per 100 children under 5 years old. The Latin
American slum community characterized above
(breast-feeding pattern A, but diarrhoea rates as
assumed for communities with pattern C), will have
initially 233 episodes of diarrhoea per 100 children
under 5 years old per year. Following a breast-feeding
promotion of high impact, there will be 216 episodes
per year, only a 7% reduction but with 17 episodes per
year averted. Precisely the same arguments hold for
mortality reductions.
A final note of caution is necessary. The impacts
computed in Tables 7 and 9 depend on median relative
risk data drawn from Annexes 1 and 2 and sum­
marized in Fig. 1-4. In the relative risks of morbidity
(Fig. 1 and 2) one may have considerable confidence.
The studies are numerous, some are recent and some
have used sophisticated epidemiological methods. In
the relative risks of mortality one must have much less

Table 8. Assumptions made in calculating the impact of breast-feeding promotions on diarrhoea mortality

Age group
(months,
inclusive)

Diarrhoea deaths per 1000
children per year, by breast­
feeding pattern (Table 3)f

Relative risk of diarrhoea
mortality, by feeding mode0

Proportion (%) of
0-59 month-old
children falling
within the
stated age group

Part BF vs
Excl. BF
*

No BF vs
Excl. BF
*

A

B

C

0-5

8.0

25.0

11

24

40

6-11

1.0''

1.0''

9

16

20

11

12-59

1.0

1.0

3

6

12

76

13

* See Annex 2 and Fig. 3 and 4.
b For definitions of feeding modes, see footnote b to Annex 1.
c Estimates derived from Puffer & Serrano (67) and Snyder & Merson (74).
d It is possible that the relative risk in this age group is > 1, but there is inadequate data on which to estimate a value and so a
conservative value of 1 has been adopted.

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

Table 9. Percentage reductions in diarrhoea mortality
rates, by age of child, due to breast-feeding promotions
of varying effectiveness

Pre-inter­
vention
pattern
of breast­
feeding0

Age of
children
(months,
inclusive)

A

B

C

Reduction (%) in diarrhoea
mortality rate by effectiveness
of breast-feeding promotion b
High
impacte

Medium
impactr

Low
*"impact

0-5

56

26

13

6-11r

0

O

0

12-59

0

0

0

0-59

17

8

4

0-5

54

27

16

6-11'1

0

0

0

12-59

0

0

0

0-59

18

9

5

0-5

44

24

14

6-1 r'

0

0

0

12-59

0

0

0

0-59

14

8

4

“ See Table 3.
b For reductions in the 0-5, 6-11, and 12-59 months age
groups the given percentage reductions are applicable, whatever
the diarrhoea mortality rates or the age structure in the com­
munity. The percentage reductions in the 0-59 months age
group are calculated using the assumptions on diarrhoea mor­
tality rates and age structure given in Table 8.
e See Table 5.
d See comment in footnote d to Table 8.

confidence for the reasons discussed above. Modern
studies are needed on diarrhoea mortality by feeding
mode to confirm or deny the substantial mortality
reductions predicted in Table 9.
A recent study in Costa Rica (56) has documented a
dramatic impact of breast-feeding promotion on
neonatal diarrhoea morbidity and mortality. Between
1976 and 1980 hospital routines were changed so as to
promote early breast-feeding and close mother-child
contact. Over the same period neonatal diarrhoea
morbidity fell by 91 % from 17.7 to 1.6 cases per 1000
live births, and neonatal diarrhoea mortality fell from
3.9 to 0 deaths per 10 000 live births. These changes
were attributed mainly to the emphasis given to the
ingestion of colostrum by neonates. Another report
of the same study (57) showed that the incidence of
diarrhoea among infants aged 0-5 months was 36%
lower in a population receiving intense breast-feeding
promotion (only 15% of infants aged 5 months
receiving no breast milk) than in a population receiv­
ing less intense promotion (41% of infants aged 5
months receiving no breast milk).

281

These morbidity reduction data from Costa Rica
(57) agree closely with the theoretical calculations
described above. The level of breast-feeding pro­
motion in this population (where the prevalences of
non-breast-fed infants were 10-19% at 0-2 months,
27-41% at 3-5 months, and 63% at 9 months) stands
mid-way between breast-feeding patterns A and B
in Table 3. It is predicted in Table 5 that a breast­
feeding promotion of high impact would reduce these
prevalences to 2-4% at 0-2 months, 10-16% at 3-5
months, and 50% at 9 months. The actual prevalences
of non-breast-fed infants in the population where
there was intense promotion in Costa Rica were
5-11% at 0-2 months, 13-16% at 3-5 months, and
31% at 9 months. The reduction in diarrhoea mor­
bidity among infants aged 0-5 months, caused by a
high-impact breast-feeding promotion, is predicted to
be 34-39% in communities having breast-feeding pat­
tern A, and 29-31 % in communities having pattern B
(Table 7). The actual reduction in morbidity among
children aged 0-5 months in Costa Rica was reported
to be 36%. This agreement between theoretical
calculations and the Costa Rican experience suggests
that the morbidity and mortality reductions predicted
in Tables 7 and 9 may be achievable in practice,
especially the more modest reductions predicted for a
breast-feeding promotion with only medium impact.

FEASIBILITY AND COSTS

The promotion of breast-feeding, which has been
reviewed elsewhere (39, 51, 85)/ is generally of two
kinds: information and support programmes and
changes in hospital routine. The most cost-effective
designs for these interventions in various societies are
not known. Changes in hospital routine have to be
made only once and they may not involve any increase
in operating costs. However, such interventions will
be effective only in societies where a substantial pro­
portion of deliveries take place in hospitals where they
are more likely to affect the initiation of breast­
feeding than its duration. It is likely that a combi­
nation of information and support programmes,
together with changes in hospital routine, will prove
to be the most cost-effective intervention in many
societies.
Costs of breast-feeding promotion activities have
not been documented but they are probably low in
comparison with most other anti-diarrhoea inter­
ventions. Studies of both the financial and the
economic costs of breast-feeding compared to bottle­
feeding have shown that breast-feeding is the cheaper
alternative (57, 52).
See footnote c on page 272.

2S2

R, G. FEACHEM & M. A. K0BL1NSKY

CONCLUSIONS

The literature on breast-feeding and diarrhoea is of
varied quality; sometimes the findings are contra­
dictory and substantial areas of ignorance remain.
The purpose of this review is to attempt to draw a
consensus from the literature.
Breast-feeding, whether exclusive or partial, ap­
pears to offer protection to children up to one year of
age, but not beyond (Fig. 1 and 4). Protection is great­
est in the first 3 months of life and falls thereafter
(Fig. I and 3). During the first half year of life, exclu­
sive breast-feeding is more protective than partial
breast-feeding and partial breast-feeding is protective
compared to no breast-feeding (Fig. 2 and 4).
The data summarized in Fig. 2 and 4 are suggestive
of the possible mechanisms of protection of breast­
feeding. If protection were due solely to the immuno­
logical and antimicrobial properties of the breast milk
itself, then the relative risk of no breast -feeding versus
partial breast-feeding might approximate that of no
breast-feeding versus exclusive breast-feeding. In fact
these relative risks are substantially different (Fig. 2
and 4). If protection were due solely to the contamin­
ation of foods other than breast milk, then the relative
risk of no breast-feeding versus exclusive breast­
feeding might approximate that of partial breast­
feeding versus exclusive breast-feeding. This would
also be the case if protection was caused solely by the
dominant colonization of the intestine by Bifido­
bacterium (a feature only of exclusively breast-fed
infants). In fact these relative risks are substantially
different (Fig. 2 and 4). Thus the relative risks com­
puted suggest that the protection is caused neither by
breast-milk properties alone, nor by the “bifid”
factor alone, nor by food contamination alone, nor
by a combination of the last two mechanisms alone.
Some combination of these three mechanisms, to­
gether with the non-specific nutritional benefits of
breast-feeding, may be responsible for the observed
degree of protection. The evidence that protection is
not caused by food contamination alone is further
supported by the protective effect of breast-feeding
for infants in families of high socioeconomic status
in developed countries, such as Canada, England,
Finland, New Zealand and the USA (Annex 1). It is
probable that the mechanisms of protection, and their
relative importance, vary by pathogen and by the age
of the infant.
The literature on breast-feeding promotion shows
that breast-feeding may become substantially more
common following changes in hospital routine com­
bined with information and support programmes for
mothers. Theoretical calculations show that a typical
breast-feeding promotion may reduce diarrhoea mor­
tality by 24-27% among infants aged Q-5 months

and by 8-9% among children under 5 years of age
(Table 9).
Only one study on the actual impact on diarrhoea
of breast-feeding promotion has been located (56,
57). This study, from Costa Rica, found substantial
reductions in neonatal diarrhoea morbidity and mor­
tality and a 36% reduction in diarrhoea morbidity
among infants aged 0-5 months. A detailed com­
parison between the Costa Rican data and the theor­
etical calculations presented in this paper shows good
agreement and gives confidence that the predicted
morbidity and mortality reductions set out in Tables 7
and 9 can be achieved in practice.
This review has highlighted several areas of ignor­
ance that require further research. The highest re­
search priority is to determine the level of protection
against diarrhoea mortality afforded by partial or
exclusive breast-feeding among infants in various
socioeconomic settings in developing countries. Well
designed studies of diarrhoea morbidity by feeding
mode in developing countries are also urgently
needed. Studies are required into the relationships
between diarrhoea severity and feeding mode and into
the possible lack of association between the relative
risk of diarrhoea in non-breast-fed infants and the
socioeconomic status of their families. The relation­
ships between breast-feeding and chronic diarrhoea,
and the protection against nosocomial diarrhoea that
may be afforded by the continued breast-feeding of
hospitalized infants, are worthy of investigation.
All these studies should be etiology-specific in
order tc clarify the undoubted differences in the levels
of protection provided by breast-feeding against
diarrhoeas of different etiology. The design of these
studies requires careful and detailed planning. Fine
age ranges must be used and several important con­
founding variables must be controlled. These require­
ments will tend to lead to study designs having many
cells and large overall sample sizes. A prospective
study may often be unduly expensive and complicated
and a case-control approach will be preferable in
these situations.
As regards operational research, more information
is needed on the design, effectiveness, and cost of
breast-feeding promotion in developing countries.
Where possible, effectiveness should be measured not
only by impact on breast-feeding patterns but also by
impact on diarrhoea rates. This latter kind of impact
measurement will typically require major prospective
studies, like that reported from Costa Rica. For these
studies to be worthwhile they must be very carefully
designed and must incorporate a detailed analysis of
the financial and economic costs of the breast-feeding
promotion.
Despite the limitations in the mortality data, and
the need for continued research as discussed above,

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

the evidence that breast-feeding protects young in­
fants from diarrhoea is strong. Governmental and
other agencies with responsibility for diarrhoea con­
trol should act to promote breast-feeding on the basis

283

of the evidence now available. Research will generate
new understandings, both fundamental and oper­
ational, which will improve the effectiveness of these
breast-feeding promotions.

ACKNOWLEDGEMENTS
The initiation of this review was greatly assisted by the preparatory work of M. Pollack during her time as a WHO
Consultant. The authors are grateful for the constructive criticisms on earlier drafts of this paper provided by A. Ashworth,
M. Belsey, D. Blum, M. Carballo, I. de Zoysa, R. C. Hogan, N. J. Hoover, M. H. Merson, M. Moore, P. I. Nieburg and
P. M. Shah. Secretarial, bibliographic and editorial assistance was most ably provided by Maelorwen Jones, Lynne Davies
and Dianne Fishman.

RESUME
INTERVENTIONS POUR LA LUTTE CONTRE LES MALADIES DIARRHEIQUES CHEZ LE JEUNE ENFANT:
ENCOURAGEMENT DE L’ALLAITEMENT MATERNEL

Le present article est le troisieme d’une serie d’etudes sur
ce qu’on peut faire dans les pays en developpement pour
abaisser la morbidite et la mortalite dues a la diarrhee chez
1’enfant de moins de cinq ans. 11 existe unc documentation
ires fournie sur les risques relatifs de morbidite enfantine
selon les differents modes d’alimentation. Cette documen­
tation souffre de plusieurs problemes mdthodologiques. On
a examine 35 etudes emanant de 14 pays. Quatre-vingt-trois
pour cent de ces etudes constatent que I’allaitement maternel total est plus protecteur que I’allaitement maternel partiel, 88% des eludes constatent que I’allaitement maternel
total est plus protecteur que 1’absence d’allaitement mater­
nel et 76% constatent que I’allaitement maternel partiel est
plus protecteur que 1’absence d’allaitement maternel. Si 1’on
compare les enfants que ne re?oivent pas de lait maternel
avec ceux qui sont nourris au sein totalement ou partiellement, le risque median relatif est de 3 entre 0 et 2 mois, de
2,4 entre 3 et 5 mois et de 1,3 i 1,5 entre 6 et II mois. Audela d’un an d’age, il n’y a pas d’effet protecteur visible de
I’allaitement maternel contre la morbidite diarrheique. Si
1’on compare les enfants qui ne re?oivent pas de lait mater­
nel avec ceux qui sont nourris entierement au sein, le risque
median relatif est de 3,5 A 4,9 dans les six premiers mois de la
vie. D’apres la documentation etudiee, rien n’indique que le
risque relatif de morbidite pour les enfants nourris au
biberon soil plus eleve dans les families pauvres que dans les
families plus riches. Par ailleurs, 1’effet protecteur de
I’allaitement maternel ne semble pas subsister apres la
cessation de cet allaitement. Par contre, il y a des signes
d’augmentation considerable de la gravite de la maladie
chez 1’enfant nourri au biberon.
Il existe peu d’ouvrages, et ils sont pour la plupart ant£rieurs a 1950, sur les risques relatifs de mortalite du nourrisson selon le mode d’alimentation. On a examine neuf

etudes emanant de cinq pays, et la plupart montrent que
I’allaitement maternel est une protection substantielle
contre le risque de mortalite. Si 1’on compare les nourrissons
qui ne re?oivent pas de lait maternel avec ceux qui sont
nourris uniquement au sein, le risque median relatif de deces
est de 25 dans les six premiers mois de la vie. Si 1’on compare
des enfants nourris totalement au sein et des enfants nourris
partiellement au sein, le risque tombe a 8,6.
On peut encourager I’allaitement maternel en changeant
les habitudes hospitalieres, ainsi qu’en eduquant et en
aidant les inches. L’examen de 21 etudes Emanant de 8 pays
montre que selon toute probability on peut ainsi abaisser la
prevalence de la maladie chez les sujets ne recevant pas de
lait maternel de 40% entre 0 et 2 mois, de 30% entre 3 et 5
mois et de 10% entre 6 mois et un an. Des calculs theoriques
fondes sur ces chiffres montrent qu’une action d’envergure
moyenne d’encouragement de I’allaitement maternel peut
faire diminuer la morbidite diarrheique dans une proportion
de 8% a 20% et la mortalite dans une proportion de 24% a
27% au cours des six premiers mois de la vie. Pour les
enfants ages de 0 a 59 mois, la morbidity serait reduite dans
une proportion de 1 % & 4% et la mortalite dans une propor­
tion de 8% & 9%. Une etude rycente faite au Costa Rica a
demontre une incidence substantielle de 1’encouragement de
I’allaitement maternel sur la morbidite et la mortality du
nouveau-ny, et sur la morbidite du nourrisson de 0 b 5 mois.
Les donnyes costa-riciennes concordent avec les calculs
theoriques prysentys ici.
Plusieurs aspects importants de I’allaitement maternel et
des maladies diarrheiques doivent encore etre yclairys par la
recherche. Toutefois, la nycessite de cette recherche ne doit
pas retarder Taction en vue d’encourager I’allaitement
maternel et d’en surveiller les effets sur les pratiques alimentaires et sur la diarrhee.

284

R. G. FEACHEM & M. A. KOBLINSKY

REFERENCES
Aust-Kettis, A. et al. Antibodies to heat-labile Esch­ 18. Fallot, M. E. et al Breast-feeding reduces incidence
of hospital admissions for infection in infants. Pedi­
erichia co/z enterotoxins in human milk and sera: a study
atrics, 65: 1121-1124(1980).
of Ethiopian and Swedish mothers and their children.
19. Feachem, R. G. Interventions for the control of diar­
Tropical and geographical medicine, 33: 225-230
rhoeal diseases among young children: supplementary
(1981).
feeding programmes. Bulletin of the World Health
2. Baer, E. C. & Winikoff, B. Breastfeeding: program,
Organization, 61: 967-979 (1983).
policy, and research issues. Studies in family planning
20.
Feachem, R. G. et al Diarrhoeal disease control:
(special issue), 12(4): 123-206 (1981).
reviews of potential interventions. Bulletin of the World
3. Behar, M. The role of feeding and nutrition in the
Health Organization, 61: 637-640 (1983).
pathogeny and prevention of diarrheic processes.
21. Feachem, R. G. & Koblinsky, M. A. Interventions for
Bulletin of the Pan American Health Organization, 9:
the control of diarrhoeal diseases among young chil­
1-9(1975).
dren: measles immunization. Bulletin of the World
4. Bullen, C. L.& Willis, A. T. Resistance of the breast­
Health Organization, 61: 641-652 (1983).
fed infant to gastroenteritis. British medical journal, 3:
22. Ferguson, D. M. et al. Infant health and breast­
338-343 (1971).
feeding during the first 16 weeks of life. Australian
5. Chandra, R. K. Prospective studies of the effect of
paediatric journal, 14: 254-258 (1978).
breast-feeding on incidence of infection and allergy.
23. Ferguson, D. M. et al. Breast-feeding, gastrointes­
Acta paediatrica Scandinavica, 68: 691-694 (1979).
tinal and lower respiratory illness in the first two years.
6. Chrystie, I. L. et al. Asymptomatic endemic rotavirus
Australian paediatric journal, Yl: 191-195 (1981).
infections in the newborn. Lancet, 1: 1176-1178 (1978).
7. Cunningham, A. S. Morbidity in breast-fed and arti­ 24. France, G. L. et al. Breast-feeding and Salmonella
infection. American journal of diseases of children,
ficially fed infants. Journal of pediatrics, 90: 726-729
(1977).
134: 147-152(1980).
8. Cunningham, A. S. Morbidity in breast-fed and arti­ 25. French, J. G. Relationship of morbidity to the feeding
ficially fed infants. II. Journal of pediatrics, 95:
patterns of Navajo children from birth through twenty685-689(1979).
four months. American journal of clinical nutrition, 20:
9. Cunningham, A. S. Breast-feeding and morbidity in
375-385 (1967).
industrialized countries: an update. In: Jelliffe, D. B.
26. Gerrard, J. W. Breast-feeding: second thoughts.
Pediatrics, 54: 757-764 (1974).
& Jelliffe, E. F. P., ed., Advances tn international ma­
ternal and child health. Vol. 1, Oxford, Oxford Univer­
27. Ghaj, O. P. et al. Infantile diarrhoea in relation to
sity press, 1981, pp. 128-165.
feeding practices. Indian journal of pediatrics, 38:
10. Cushing, A. H.& Anderson, L. Diarrhea in breast-fed
354-357 (1971).
and non-breast-fed infants. Pediatrics, 70: 921-925
28. Gindrat, J.-J. etal. Antibodies in human milk against
(1982).
E. coli of the serogroups most commonly found in neo­
II. Davis, W. H. Prevention of infant mortality by breast­
natal infections. Acta paediatrica Scandinavica, 61:
feeding. American journal of public health, 11: 67-71
587-590 (1972).
(1912).
29. Goldberg, H. I. et al. Infant mortality and breast­
12. Douglas, J. W. B. The extent of breast-feeding in
feeding in northeastern Brazil. In: Proceedings of the
Great Britain in 1946, with special reference to the
Population Association of America Annual Meeting,
health and survival of children. Journal of obstetrics
San Diego, California, 29 April-1 May 1982.
and gynaecology, 57: 335-361 (1950).
30. Goldman, A. S. & Smith, C. W. Host-resistance
13. Ebbs, J. H. & Mulligan, F. The incidence and mor­
factors in human milk. Journal of pediatrics, 82:
tality of breast- and artificially-fed infants admitted to
1082-1090(1973).
hospital with infections. Archives of disease in child­
31. Gordon, J. E. et al. Weanling diarrhea. American
hood, 17: 217-219(1942).
journal of the medical sciences, 245: 345-377 (1963).
14.
Elegbe, I. A. et al. Pathogenic bacteria isolated from 32. Gordon, J. E. et al. Acute diarrhoeal disease in less
infant feeding teats. American journal of diseases of
developed countries. Bulletin of the World Health
children, 136: 672-674 (1982).
Organization, 31: 9-20 (1964).
15.
Ellestad-Sayed, J. et al. Breast-feeding protects 33. Gothefors, L. & Winberg, J. Host-resistance factors.
against infection in Indian infants. Canadian Medical
Journal, of tropical pediatrics and environmental child
Association Journal, 120: 295-298 (1979).
health, 21: 260-263 (1975).
16. Espejo, R. T. et al. Rotavirus gastroenteritis in 34. Grantham-McGregor, S. M. & Back, E. H. Breast­
hospitalized infants and young children in Mexico City.
feeding in Kingston, Jamaica. Archives of disease in
Revista Latinoamericana de microbiologia, 20: 239-246
childhood, 45: 404-409 (1970).
(1978).
35. Grulee, C. G. et al. Breast and artificial feeding:
17. European Society for Paediatric Gastroenter­
influence on morbidity and mortality of twenty thou­
ology and Nutrition, Committee on Nutrition.
sand infants. Journal of the American Medical Associ­
Guidelines on infant nutrition. III. Recommendations
ation, 103: 735-739 (1934).
for infant feeding. Acta paediatrica Scandinavica,
302(Suppl.) 1-27 (1982).

1.

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

285

Grulee, C. G. etal Breast and artificially fed infants: 56. Mata, L. et al. Promotion of breast-feeding, health
and growth among hospital-born neonates, and among
a study of the age incidence in the morbidity and
mortality in twenty thousand cases. Journal of the
infants of a rural area of Costa Rica. In: Chen, L. C. &
American Medical Association, 104: 1986-1988 (1935).
Scrimshaw, N. S., ed., Diarrhoea and malnutrition:
37. Gunn, R. A. et al. Bottle feeding as a risk factor for
interactions, mechanisms and interventions. New York,
cholera in infants. Lancet, 2: 730-732 (1979).
Plenum Press, 1982, pp. 177-202.
38. Gurwith, M. et al. A prospective study of rotavirus 57. Mata, L. et al. Diarrhoea and malnutrition: breast­
infection in infants and young children. Journal of
feeding intervention in a transitional population. In:
infectious diseases, 144: 218-224 (1981).
Holme, T. et a!., ed., Acute enteric infections in
39. Hales, D. J. Promoting breastfeeding: strategies for
children. New prospects for treatment and prevention.
changing hospital policy. Studies in family planning, 12:
Amsterdam, Elsevier North Holland Biomedical Press,
167-172(1981).
1981, pp. 233-251.
40. Hanson, L. A. et al. Antibodies — potential capacity 58. Mata, L. J. & Urrutia, J. J. Intestinal colonization of
in human milk. In: Proceedings of the International
breast-fed children in a rural area of low socioeconomic
Pediatric Association Symposium on Infant and Young
level. Annals of the New York Academy of Sciences,
Child Feeding, Ankara, 27-29 November, 1982.
176: 93-109(1971).
41. Hanson, L. A.etal Immune defense factors in human 59. Mata, L. & Vives, M. The impact of human rotavirus
milk. Modern problems in paediatrics, 15: 63-72
diarrhea in Latin America. In: Proceedings of the First
(1975).
International Conference on the Impact of Viral
42. Hanson, L. A. & Winberg, J. Breast milk and defence
Diseases on the Development of Latin American
against infection in the newborn. Archives of disease in
Countries and the Caribbean Region, Rio de Janeiro,
childhood, 47: 845-848 (1972).
Brazil, 21-26 March, 1982.
43. Hardy, E. E. et al Breastfeeding promotion: effect of 60. Mata, L. J. & Wyatt, R. G. Host resistance to infec­
an educational program in Brazil. Studies in family
tion. American journal ofclinical nutrition, 24:976-986
planning, 13: 79-86 (1982).
(1971).
44. Howarth, W. J. The influence of feeding on the 61. Mittal, S. K. et al. Gut flora in breast- and bottle-fed
mortality of infants. Lancet, 2: 210-213 (1905).
infants with and without diarrhea. Indian pediatrics, 20:
45. Hoyle, B. etal. Breast-feeding and food intake among
21-26 (1983).
children with acute diarrheal disease. American journal
62. Newman, G. Infant mortality: a social problem,
of clinical nutrition, 33: 2365-2371 (1980).
London, Methuen, 1906.
46. Kanaaneh, H. The relationship of bottle-feeding to 63. Paine, R. & Coble, R. J. Breast-feeding and infant
malnutrition and gastroenteritis in a pre-industrial
health in a rural U.S. community. American journal of
setting. Journal of tropical paediatrics and environ­
diseases of children, 136: 36-38 (1982).
mental child health, 18: 302-306 (1972).
64. Petros-Barvazian, A. Prevalence and duration of
47. Knodel, J. Breast-feeding and population growth.
breastfeeding in different parts of the world: an over­
Science, 198: 1111-1115 (1977).
view. In: Proceedings of the International Pediatric
48. Larguia, A. M. etal. Fresh human colostrum for the
Association Symposium on Infant and Young Child
prevention of E.coli diarrhea — a clinical experience.
Feeding, Ankara, 27-29 November, 1982.
Journal of tropical pediatrics and environmental child
65. Pittard, W. B. Breast-milk immunology: a frontier in
health, 23: 289-290 (1977).
infant nutrition. American journal of diseases of
49. Larsen, S. A. & Homer, D. R. Relation of breast
children, 133: 83-87 (1979).
versus bottle feeding to hospitalization for gastro­
66. Popkin, B. M. et al. Breast-feeding patterns in lowenteritis in a middle-class U.S. population. Journal of
income countries. Science, 218: 1088-1093 (1982).
pediatrics, 92: 417-418 (1978).
67. Puffer, R. R. & Serrano, C. V. Patterns of mortality
50. Lepage, P. et al. Breastfeeding and hospital mortality
in childhood: Report of the Inter-American investi­
in children in Rwanda. Lancet, 2: 409-411 (1981).
gation of mortality in childhood. Washington, DC,
Pan American Health Organization, 1973 (Scientific
51. McCann, M. F. et al. Breast-feeding, fertility, and
Publication, No. 262).
family planning. Population reports. Series J(24):
68.
Research Sub-Committee of the South-East
525-575 (1981).
England Faculty of the Royal College of General
52. McKigney, J. Uniqueness of human milk: economic
Practitioners The influence of breast-feeding on the
aspects. American journal of clinical nutrition, 24:
incidence of infectious illness during the first year of
1005-1012(1971).
life. Practitioner, 209: 356-362 (1972).
53. Majumdar, A. S. & Ghose, A. C. Protective prop­
69. Robinson, M. Infant morbidity and mortality: a study
erties of anticholera antibodies in human colostrum.
of 3266 infants. Lancet, 1: 788-794 (1951).
Infection and immunity, 36: 962-965 (1982).
54. Mannheimer, E. Mortality of breast-fed and bottle-fed 70. Sauls, H. S. Potential effect of demographic and other
variables in studies comparing morbidity of breast-fed
infants: a comparative study. Acta genetica et statistica
and bottle-fed infants. Pediatrics, 64: 523-527 (1979).
medica, 5: 134-163 (1954).
71.
Schoub, B. D. et al. The role of breast-feeding in the
55. Mata, L. Breast-feeding: main promoter of infant
prevention of rotavirus infection. Journal of medical
health. American journal of clinical nutrition, 31:
microbiology, 11: 25-31 (1977).
2058-2065 (1978).
36.

Annex 1

Relative risk of diarrhoea morbidity by feeding mode
Relative risk of diarrhoea morbidity0
Socio­
economic
status

Age group
(inclusive)

Bahraine

1978

7

0-11 mo.

7.0

Canada
Toronto

1939

7

0-11 mo.

5.80

1972-75

Lower

0-11 mo.

?

?

0-23 mo.

Manitoba
Indian
reservation

Urban

*
Winnipeg

Pan BF
vs
Excl. BF6

No BF
vs
Excl. BF"

No BF
vs
Part BF"

10 32

3.20

Factors controlled

Case-control study. Pairs
matched for age and place of
residence (rural/urban). Waterrelated habits and child care
practices were not risk factors
for cholera.

Reference

37

13
Marital status of mother,
occupation of father, and
family income were not
associated with breast-feeding.
Women who breast-fed for
> 12 mo. were older, had
more children, less education,
and lived in more crowded
houses than those who breast­
fed for < 12 mo.

15

Feeding modality controlled
for socioeconomic status,
parental education, and family
size.

5

1976-79

All

0-16 mo.

1.41

38

Colombia

1964-65

Lower

0-5 mo.
6-11 mo.

2.37
2.37

2.91
1.64

1.33
0.69

88

Costa Rica

1979-82

Lower

0-2 mo.
3-5 mo.
6-8 mo.
9-11 mo.

1.37
3.31

6.47
6.92

4.73
2.09
1.29
1.00

59

England
Liverpoool

1936

Lower/
middle

3-26 wk

6.36

12.85

2.02

69

Nationwide

1946-48

Lower

0-3 mo.
4-8 mo.
9-23 mo.

2.57

3.20

1.24
0.79
0.91

12

7

7

2-11 mo.

2.98

2.33

0.78

76

Oxford

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

Date of
study

Country/place

South-east

1968-69

Upper/
middle

0-11 mo.

3.89

Sex, date of birth, birth
weight, ages of parents,
number of persons per house,
number of sibs, and number
of sibs at school, were not
associated with breast-feeding.

68

Nationwide

1970-75

All

0-11 mo.

1.55

Feeding modality controlled
for maternal age, child's sex,
birth weight, birth rank,
maternal smoking, and socio­
economic status.

78

Ethiopia
Addis Ababa

1979

All

0-5 mo.
6-11 mo.

Finland
4 towns

1949

7

0-11 mo.

Guatemala

1964

Lower

3-5 mo.

0.85

6-8 mo.

1.14

9-11 mo.

1.32

9.14
0.67

80

1.48
1.99

89

3.04

32

1955-59

Lower

0-2 mo.
3-5 mo.
6-8 mo.
9-11 mo.

1.26
1.39
1.17
1.55

31

New Delhi
rural

?

Lower

3-5 mo.
6-8 mo.
9-11 mo.

1.39
1.02
0.95

27

urban

7

Middle

3-5 mo.
6-8 mo.

1.14
0.97

Rural

7

?

0-11 mo.

New Delhi

7

Lower

0-5 mo.
6-11 mo.

Israel

1971

Lower

0-5 mo.

Jamaica

1967-68

7

0-3 mo.

3.01

5.53
1.20

10.47

*
3.39

Feeding modality controlled
for socioeconomic status,
parental education, occu­
pation, and family size.

5

Cases and controls were socio­
economically matched and
both groups were hospitalized.

61

Twins and infants with low
birth-weight or birth defects
were excluded.

46

Lower socioeconomic status,
non-working mothers at 6 mo.
post-birth, and no cribs,- were
associated with breast-feeding.

34

i. FEACHEM & M . A. KOBLINSKY

India
Punjab

6.19
0.34

New Zealand
Christchurch

Christchurch

1977-79

Upper/
middle

Upper/
middle

(treated
diarrhoea)
0-3 mo.

2.41

1.60

1.49

(all
diarrhoea)
0-3 mo.

3.92

1.99

1.27

(treated
diarrhoea)
0-3 mo.
4-11 mo.
12-23 mo.

2.06
1.23
0.88

(all
diarrhoea)
0-3 mo.
4-11 mo.
12-23 mo.

2.81
1.21
0.97

22

Relative risks for 0-3 mo.
group remained significant
when controlled for child's
sex, maternal age, race,
education, smoking, family
size, number of parents, and
quality of child care.

23

1.71
0.64

82

1955

All

0-5 mo.
6-11 mo.

USA
Chicago

1924-29

Lower

2 wk-8 mo.

Chicago

1924-29

Lower

0 mo.
1 mo.
2 mo.
3 mo.
4 mo.
5 mo.
6 mo.
7 mo.
8 mo.

Boston

1930-40

All

0-11 mo.

1960

Lower

0-2 mo.
3-5 mo.
6-8 mo.
9-11 mo.

California

1973-75

Upper/
middle

0-2 mo.

5.77

Iowa

1973-78

Upper/
middle

0-5 mo.

4.15

Navajo Indian
reservation

Relative risks remained
significant when controlled
for maternal age, parity,
parental education, race,
number of parents, living
standard, gestation period.
and birth weight.

2.24
3.00
3.71
2.55

0.95
0.23
1.90
2.01

1.83

35

0.42
0.08
0.51
0.79
1.67
2.29
2.63
2.72
4.34

36

1.20

75

4.20 '
2.78'
1.38'
1.98'

25

49
Number of parents,
number of sibs, number of
clinic visits of infants with
no sibs and those with one
or more, parental education,
urban vs rural, were not
significantly associated with
breast-feeding.

63

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

Uganda
Villages near
Kampala

1977

Cooperstown,
NY

1974

Upper/
middle

0-11 mo.

1.97

Arkansas7

1977-78

0-11 mo.

96.0

Syracuse

1978

Upper/
middle
?

New York
City

?

Lower

0-5 mo.

2.72

Feeding modality was
controlled for mother’s age,
parity, ethnicity, and
education.

1979

All

0-11 mo.

2.01

Breast-feeding was associated
with maternal age and
education, home-ownership,
and ethnicity, but not with sex.
maternal career, family size,
water source, medical care
source, or day care.

24

18

2.70

0-3 mo.

7, 8, 9

Footnote g

10

e The relative risk of feeding mode X compared to feeding mode Y is computed by dividing the incidence of diarrhoea for children on feeding mode X by that for children on feeding
mode Y. When incidence data were not reported, prevalence was used or the relative risk was computed from data on the distribution of feeding modes among infants with diarrhoea
compared to healthy infants.
b Categories of feeding mode are defined as follows: Excl. BF = children receiving only breast milk throughout the stated age range; Part BF = children receiving breast milk and
other milk and/or food throughout the stated age range and children receiving only breast milk or breast milk plus other milk/food for part of the stated age range; No BF = children
receiving no breast milk throughout the stated age range.
c This study relates to cholera diarrhoea only.
d This study relates to rotavirus diarrhoea only.
' Children described as No BF included some who were Part BF and some who were No BF, but these two groups could not be separated.
7 This study relates to Salmonella diarrhoea only.
g SOLIMANO, G. ET AL. Morbidity patterns in breast-fed and non-breast-fed infants in a low socioeconomic urban U.S. population Poster session. 12th International Congress of

Nutrition, San Diego, CA, 1981.

Annex 2
Relative risk of diarrhoea mortality by feeding mode0
Relative risk of diarrhoea mortality6

Country/place

Date of
study

Socioeconomic
status

Age group
(inclusive)

Part BF
vs
Excl. BFC

No BF
vs
Excl. BFf

No BF
vs
Part BFC

Factors controlled

Reference

.. G. FEACHEM & M . A. KOBL1NSKY

Albuquerque,
New Mexico

Relative risks (of all illnesses)
remained significant when
parental education, maternal
age, family size, birth weight,
sex, Apgar score, and birth
month were controlled. Higher
relative risks (of all illnesses)
were associated with male
infant, lower maternal age,
larger family size, and lower
birth-weight.

Canada
Toronto

1939

?

0-11 mo.

Egypt
Menoufia

1979-80

Lower

0-11 mo.

3.33d

England
Derby'

1900-03

Lower

0-11 mo.

2.53

5.83

Liverpool

1884-86

1901-04

?
7

0-2 mo.

Finsbury'

0-2 mo.
6-8 mo.

9.82

25.24
10.99

2.57

62
62

Brighton

1903-05

7

0-2 mo.
3-5 mo.

7.33

29.15
43.24

3.98

62

13

6.48

79
2.31

Children born prematurely, or
with defects or malformations,
were excluded. Better housing
associated with bottle-feeding.

Croydon

1904

?

0-5 mo.

Liverpool

1936-42

Lower/
middle

3-26 wk

3/0z

6/0z

3.50

Premature and ‘weakly’
children were excluded.

Sweden
Stockholm

1943-47

All

1-11 mo.
2-11 mo.

1.08
3.25

6.79
18.75

6.27
5.77

Breast-feeding associated
with belter child care. Relative
risks (of total mortality) not
influenced by age of mother
or birth-weight.

1911
7

All

2 wk-11 mo.

8 cities'

All

1-8 mo.
9-11 mo.
0-11 mo.

Chicago

1924-29

Lower

2 wk-8 mo.

USA
Boston'

17.22

13.02

Relative risks (of total mor­
tality) remained substantial
when race, nationality,
prematurity, maternal death,
and plural birth were
controlled. Relative risks (of
total mortality) was higher in
lower income groups.

11.3
3.5
7.71

18.82

62
69

54

11
87

35, 36

° All the studies summarized here contain severe methodological flaws and the relative risks derived should be regarded as indicative only.
b The relative risk of feeding mode X compared to feeding mode Y is computed by dividing the diarrhoea mortality rate of children on feeding mode X by that for children on feeding
mode Y. Where mortality rates were not reported, the relative risk was computed from data on the distribution of feeding modes among infants who died from diarrhoea compared to
healthy infants.
e For definitions of feeding modes see footnote b to Annex 1.
d To compute this relative risk it was necessary to assume that the distribution of feeding modes among infants who died from causes other than diarrhoea was
identical to that of infants in the general population. Provided that breast-feeding is not a risk factor for non-diarrhoea deaths, this assumption can only lead to an underestimation
of the relative risk.
' Rates based on over 100 diarrhoea deaths.
z There were 0 deaths in the Excl. BF group.

PROMOTION OF BREAST-FEEDING FOR CONTROL OF DIARRHOEAL DISEASES

15

44

Bulletin of the World Health Organization, 63 (I): 165 - 184 (1985)

© World Health Organization 1985

Interventions for the control of diarrhoeal diseases
among young children: prevention of low birth weight
*
Ann Ashworth1 & R. G. Feachem*
12

The effect of low birth weight (LB IV) on diarrhoea morbidity and mortality is analysed
and interventions to increase birth weights are reviewed. Birth weight is a major
determinant of infant mortality and, in developed countries at least, its effect on neonatal
mortality is independent of socioeconomic status. We have located no satisfactory data on
LBW as a determinant of diarrhoea mortality or morbidity. The strong association between
LBW and mortality, however, makes it likely that there is an association between LB W and
diarrhoea mortality in developing countries where diarrhoea is a major cause of infant
death. Poor maternal nutrition, certain infections, pre-eclampsia, arduous work after mid­
pregnancy, short birth intervals, and teenage pregnancy are likely to be causally associated
with LBW in developing countries. Tobacco and alcohol consumption are additional risk
factors.
Of the interventions examined, maternal food supplementation has been the most
studied. If targeted to mothers at nutritional risk, and if thefood is consumed in addition to
the usual diet, the prevalence of LBW can be expected to be reduced. However, food
supplementation can be expensive and the results from carefully supervised feeding trials
may be better than those that can be achieved tn national programmes. The effect ofsupple­
mentation with iron, zinc or folate requires further study. If it were possible to intervene in
maternal nutrition, health and life-style in a developing country in a way that reduced the
prevalence of LBWfrom around 30% to around 15 %, a fall in the infant mortality rate of
around 26 % would be expected. Thefall in infant diarrhoea mortality rate might be similar.
The scarce data on relative risk of morbidity by birth weight do not allow any comparable
computations for morbidity reductions to be made.
This review confirms that whatever its association with diarrhoea, LBW is an
important determinant of infant mortality. For the more general goal of reducing infant
mortality it is necessary to know more about the nature, etiology, and prevention of LBW
in developing countries.

In this review, the effect of low birth weight on
diarrhoea morbidity and mortality is analysed and
interventions to increase birth weight are reviewed.
While recognizing that the etiology of low birth
weight is multifactorial, emphasis is given to those
maternal factors that are believed to be of greatest
importance in developing countries and that may be
amenable to change in the short term. These include
inadequate dietary intake, infections, arduous

workloads, teenage pregnancy, short birth intervals,
and excessive tobacco or alcohol consumption. This
paper examines whether interventions that reduce the
prevalence of low birth weight might be effective in
reducing morbidity or mortality from diarrhoeal
diseases among young children. This review is the
fifth in a series of reviews of potential anti-diarrhoea
interventions being published in the Bulletin of the
World Health Organization (33-37).

• Requests for reprints should be sent to the Director, Diarrhoeal
Diseases Control Programme, World Health Organization, 1211
Geneva 27, Switzerland.

1 Lecturer, Department of Human Nutrition, London School of
Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT,
England.
2 Reader in Tropical Public Health Engineering and Head of
Department of Tropical Hygiene, London School of Hygiene and
Tropical Medicine, London, England.

4513

EFFECTIVENESS

If the prevention of low birth weight, by improving
the nutritional status, health or life-style of pregnant
women, is to be an effective diarrhoea control

-165—

Table 1. Percentage distribution of live births by birth weight in four countries
Percentage of live births

USA

< 1501

0.8

0.8

1.0

1.0

1.2

0.9

1.6

2.2

1.7

3.9

3.0 )

1501-2000

0.9

1.1

1.3

1.5

1.4

1.3

1.8

2.5

2.8

3.4

3.0 J•

Oregon USA (5
USA
(Portland) regions) (white)

New
York

USA
(non­
white)

Brazil
(Recife)

Colorado
(Denver)

New
York
(black)

India

New
Delhi

2.8

North
Arcot

Palghar

Guatemala
(Santa
Maria
Cauqud)

1.1

5.2

1.2

5.5

9.4

6.5

2001-2500

3.5

4.2

4.5

5.1

5.1

5.7

6.6

8.3

10.0

8.0

9.8

20.2

25.3

23.2

34.0

2501-3000

15.6

17.8

17.2

19.1

18.5

22.7

23.9

25.3

33.0

24.9

28.4

45.8

41.2

37.4

48.1

3001-3500

38.1

39.2

38.1

39.5

38.0

40.6

39.6

37.1

35.0

36.9

36.6

25.8

21.5

22.4

10.0

3501-4000

30.4

27.7

28.2

26.0

26.8

22.6

21.0

18.9

14.8

18.4

15.8

4.8

4.8

>4000

10.6

9.1

9.6

7.8

9.0

6.2

5.5

5.8

2.6

4.5

3.5

0.6

0.7

2.4 |
>
0.0 )

0.2

% LBW

5.2

6.1

6.8

7.6

7.8

8.0

10.0

12.9

14.5

15.2

15.7

23.0

31.9

37.8

41.7

Reference

10

110

20

91

20

21

21

20

82

55

21

91

93

103

70

.. ASHWORTH & R.G. FEACHEM

California

New
York
(white)

Birth weight
(0)

167

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

intervention, it must be true that:

Hypothesis 1. A considerable proportion of
diarrhoea morbidity or mortality in young children
in developing countries is due to low birth weight.

either
a considerable proportion of
diarrhoea morbidity or mortality in
young children in developing
countries is due to low birth weight

hypothesis
1

and

improving the nutritional status,
health or life-style of pregnant
women can reduce the prevalence
of low birth weight

hypothesis
2

or

improving the nutritional status,
health or life-style of pregnant
women can reduce diarrhoea
morbidity or mortality rates in
young children

hypothesis
3

The literature on hypothesis 2 is extensive, whereas
hypotheses 1 and 3 have been little studied. The
effectiveness of low-birth-weight prevention as an
intervention to reduce diarrhoea morbidity or
mortality would be suggested by a demonstration
either of the correctness of hypotheses 1 and 2 or of
the correctness of hypothesis 3. The evidence for and
against these hypotheses is examined below.

Prevalence and distribution of low birth weight.
The term low birth weight (LBW) is used to describe
infants who weigh less than 2500 g at birth. From
sample surveys and country reports, WHO has esti­
mated that over 20 million LBW infants are born each
year. At the global level this represents 16% of all
births, but the proportion is not uniform and national
rates range from 4% in Scandinavia to around 50% in
parts of India and Bangladesh (124, 125). Table 1
compares birth weight distributions in 15 populations
with different LBW prevalence rates, ranging from
5% to 42%.
Infants with a low birth weight may be divided into
two broad subgroups: (a) those who are born
preterm, that is of less than 37 weeks’ gestation; (b)
those who are growth-retarded in ulero and are born
small for gestational age (SGA). Investigators are not
consistent in their definition of SGA, but it may be
defined as a birth weight of 2SD or more below the
mean birth weight for gestational age. Although most
preterm infants are appropriate for gestational age
(AGA), some LBW infants are both preterm and
SGA. In the developed countries the majority of
LBW infants are the result of a preterm delivery. In
contrast, in developing countries it would seem from
the limited information available that the majority of
LBW infants are small for gestational age. For
example, in 18 reports from the Indian subcontinent,
south-east Asia and Latin America, between 65% and
96% of LBW infants were small for gestational age
(124). In Africa, there are data for only 5 cities and
these show a more diverse pattern with between 34%

Table 2. Association of low birth weight and infant mortality in developing countries

Region or
country

Place

Prevalence
of LBW (%)

Percentage of deaths weighing
< 2500 g at birth

Reference

Neonatal

Latin America and
Caribbean

a

Postneonatal

Infant



47



42

11

-

Hyderabad

21

84





North Arcot

32

56

42

48

Santa Maria Cauqu6

42

87

58

70

Nigeria

Igbo-Ora

India
India

Guatemala

— denotes data not available.

89
4
80
93
69

Table 3. Relative risks of neonatal mortality by birth weight compared to birth weight of 2501-3000 g
Relative risk of neonatal mortality

USA
(1960)

47.5

55.0

32.8

52 6

94.2

46.2

41 4

16 T

20.6

19.6

9.4

18.2

30.8

13.9

23.3

4.4

27.3

5.2

4.4

4.4

2.1

4.2

4.5

3.3

4.0

1.4

3.4

1.0

1.0

1.0

1.0

1.0

1.0

1.0

1.0

1.0

1.0

1.0

0.5

0.4

0.5

0.4

04

0.3

0.7

0.5

0.8

0.3

0.2
0.2 )

0.4

0.3

0.3

0.3

0.5 >
0.4 I

0.6

0.4

0.4 1

0.7
1.1 )

0.4

0.2
0.3

0.8

0.9

1 7 /

0 5

1001-1500

52.7

99.9°

41.1

80.5

1501-2000

13.5

17.6

16.6

19.8

2001-2500

3.0

3.8

4.5

2501-3000

1.0

1.0

3001-3500

0.3

3501-4000
4001-4500

0.3
0.3 )

>4500

0.4 j

0.4

0.3 J

0.6

1

14

5.6

8.2

8.5

12.7

13.8

16.9

17.5

18.4

21.2

26.7

28.2

34.8

39.0

% LBW

unstated

6.1

5.2

unstated

5.2

6.8

15.2

7.8

23.2

12.9

8.7

31.9

41.7

No. of
births

290 000

234 000

700 000

44 700

40 000

3.6
million

14 400

4.26
million

4 590

657 100

18 200

4 220

416

Reference

123

110

30

90

10

20

55

20

41

20

90

93

69

Includes all births < 1500 g.
Deaths per 1000 live births, except for ref. 123 (per 1000 single vaginal births), ref. 30 (per 1000 total births), and ref. 10 (per 1000 single, white, live births).

.. ASHWORTH & R.G. FEACHEM

Colorado
(Denver)
(1974-80)

California
(1977)

Neonatal
mortality
rate6

Brazil
Guatemala
USA (non­ (Ribeirao
India
(Santa Maria
India
(N. Arcot)
Cauqu6)
(Delhi)
white)
Preto)
(1969-72)
(1960)
(1968-70) (1969-75) (1964-72)

USA
(white)
(1960)

USA (5
Oregon
regions
Norway
California (Portland)
(1974-75) (1967-78) (1969-70) (1959-66)

Birth weight
(g)

169

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 4. Relative risks of post-neonatal mortality by birth weight compared to birth weight of 2501-3000 g
Relative risk of post-neonatal mortality

Guatemala
(Santa Maria
Cauqu6)
(1964-72)

Norway
(1967-68)

USA
(5 regions)
(1974-75)

USA
(white)
(1960)

USA
(1960)

USA
(non-white)
(1960)

India
(Delhi)
(1969-72)

India
(N. Arcot)
(1969-75)

1001-1500

2.6

12.2'

6.3

6.1

4.8

9.3

4.5'

1501-2000

3.0

3.7

3.5

3.4

3.0

5.3

2.3

7.0

2001-2500

1.8

1.9

1.9

1.9

1.8

2.2

1.1

0.8

Birth weight
(g)

2501-3000

1.0

1.0

1.0

1.0

1.0

3001-3500

0.6

0.6

0.6

0.6

3501-4000

>
(

4001-4500

0.5
0.4

0.7
0.7

)

(

>4500

0.5

J

Post-neonatal
mortality
rate6

3.2

4.3

% LBW

5.2

6.1

No. of births

700 000

234 000

Reference

30

110

0.5
0.5

1.0

1.0

1.0

0.6

0.5

0.4

0.5

0.5

0.5

0.4

0.7

0.5

0.5

0.7

5.4

6.9

15.1

25.3

51.6

60.0

6.8

7.8

12.9

23.2

31.9

41.7

657 100

4 590

4 220

416

20

41

93

69

*
3.6
million 4.26million

20

20

0.7

a Includes all births < 1500 g
6 Deaths per 1000 survivors, except for ref. 30 (per '1000 total births) and ref. 110 (per 1000 live births).

and 73% of LBW infants being small for gestational
age. More research is required to determine the rela­
tive proponion of these two subgroups of LBW
infants in developing countries.
Association of low birth weight with mortality. In
developed countries, LBW infants comprise the
majority of infant deaths. For example, in a recent
report from the USA, although LBW infants repre­
sented only 6% of live births they comprised 55% of
infant deaths (110)- Data from developing countries
are limited since death registration is frequently
incomplete and birth weights are rarely recorded.
Indications are, however, that LBW infants similarly
comprise a large proportion of infant deaths as shown
in Table 2.
Birth weight has a marked association with neo­
natal mortality rates in both developed and
developing countries. The relative risks of neonatal
death among infants of different birth weights,
compared with infants weighing 2501-3000 g, are
shown in Table 3. The most favourable birth weights
were in the range of 3500-4500 g. Although there
were 7-fold differences in overall neonatal mortality
rates among the 13 studies, the relative risks in each
500-g birth-weight interval were similar, except for
Denver and North Arcot, where infants with birth
weights of < 2500 g had lower relative risks than in
the other places studied.

The association between birth weight and mortality
extends into the post-neonatal period and birth
weights of 3500-4500 g were similarly associated with
the lowest risk. Table 4 shows that among LBW
infants the relative risk of post-neonatal death in
North Arcot was somewhat lower than that observed
in developed countries, whereas in Delhi the relative
risk among LBW infants was higher than in
developed countries. In Santa Maria Cauque, the
relative risks were somewhat anomalous, perhaps
because of the small number of subjects in this
study.
It is possible that there is also an association
between birth weight and mortality in the 1-4-year
age group. In the Gambia, child mortality is signifi­
cantly higher in children born in the wet season (when
the incidence of LBW is high) than in children born in
the dry season (55), and there is some indication from
Guatemala that SGA infants may have an increased
I-4-year mortality risk (6S, 71). In view of the limited
data available, however, it has been conservatively
assumed in the calculations which follow that low
birth weight confers no increased risk of death
beyond the first year of life.
In populations where the whole Gaussian distri­
bution of birth weights is shifted to the left, leading to
a lower mean birth weight, it is possible that the defi­
nition of LBW based on a cut-off point of 2500 g is

170

A. ASHWORTH & R.G. FEACHEM

Table 5. Neonatal and post-neonatal mortality rates by birth weight and gestational age
Mortality rate
Birth weight
(g)

Gestational
age

Santa Maria
CauquS

Delhi

North
Arcot

New York
City

Neonatal mortality rate
(deaths per 1000 live births)

< 2500

Preterm

323

87

73

93

Term {SGA)

28

27

41

44

Preterm

10

30

Term

4

21

> 2500

8

__b
_b

Post-neonatal mortality rate
(deaths per 1000 survivors)

Preterm

286

46

66

16

Term (SGA)

58

35

66

13

< 2500

> 2500

Term

Reference

_b

18

50

42

13

42

_b

69

41

93

117

Preterm

“ No infants in this category.
6 No data available.

too rigid (95). In such populations the relative risk of
neonatal or post-neonatal death among births of
< 2500 g may be lower than in populations with
higher mean birth weights. There is evidence for this
from the USA in populations living at high altitude
(e.g., Colorado, Tables 1 and 3) and from the United
Kingdom in populations differing in racial origin
(27). At present, however, there are insufficient birth­
weight-specific mortality data to determine whether,
or to what extent, the relative risk of neonatal and
post-neonatal death among LBW infants in
developing countries differs from that in developed
countries.
Many of the factors affecting mortality are inter­
related. Few studies have attempted to determine the
effect of one factor while controlling for the inter­
related factors. In Baltimore, during 1960-64, a
cohort of 108 852 single live births of known birth
weight were followed for one year after birth. When
the effects of birth weight, race, socioeconomic
status, maternal age, birth order, and prenatal care
were separated, birth weight was the most important
factor in neonatal mortality, and factors such as race,
socioeconomic status and maternal age were
important only because they were related to birth
weight {102). Likewise birth weight was the most
important factor in post-neonatal mortality. How­
ever, in this period the other factors (except for race)
were also important. In New York City during

1976-78, socioeconomic circumstances were also
found to have little effect on the probability of neo­
natal death once birth weight was controlled (83). No
comparable analyses from developing countries have
been located.
Mortality among preterm versus SGA infants.
Although the data are limited, it would appear that in
both developed and developing countries preterm
infants have a considerably higher risk of neonatal
death than term SGA infants of similar birth weight
(9, 110 and Table 5). The reason for this is the greater
difficulty experienced by preterm infants in adapting
to the extrauterine environment owing to immaturity
of many of the body’s systems (117).
In the post-neonatal period, preterm infants may
also experience higher mortality rates than term SGA
infants (Table 5), but the difference is less marked and
less consistent than in the neonatal period.
The data on childhood mortality are very limited.
However, in New York, term SGA infants appeared
to be at slightly greater risk, their 12-23-month
mortality rate being 2.7 per 1000 compared with 2.3
for preterm infants (117). In Santa Maria Cauque,
term SGA infants had a higher mortality rate during
the second, third, and fourth years of life than
preterm infants (68).
Studies in the USA and United Kingdom have
shown that term SGA infants grow more slowly than

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

preterm infants (25, 32, 81, 117), the difference in
their mean weight being 0.7 kg at 12 months (25,117)
and 5 kg at 10 years of age (32). Slower growth in
SGA infants compared with preterm infants has also
been reported from New Delhi (100). In conclusion, it
would seem that preterm infants are at greater risk of
death, especially neonatal death, in the first year of
life than term SGA infants, but that preterm infants
who survive may grow better than SGA infants and
have lower mortality rates after 1 year of age.

Association of low birth weight with infant
morbidity. At present only limited data are available,
although several long-term prospective studies of
preterm infants are in progress in developed
countries. In Oakland, California, LBW infants did
not experience significantly more episodes of acute
diseases in the first two years of life than infants of
higher birth weight (116).
Association of low birth weight with diarrhoea
morbidity or mortality. The association, if any,
between LBW and diarrhoea morbidity or mortality
is poorly documented. In the Inter-American Investi­
gation of Mortality in Childhood (89), LBW was
examined as an associated cause of diarrhoea
mortality, but only in relation to the neonatal period.
Of 1269 neonatal diarrhoea deaths recorded in 8
countries, LBW was a contributory cause in 49%,
with a range from 28% in Bolivia to 80% in Jamaica.
In a prospective study from 1964 to 1972 in the Guate­
malan village of Santa Maria Cauque, LBW infants
show’ed greater rates of diarrhoeal disease and greater
prevalence rates of infection with Shigella,
Entamoeba histolytica and Giardia in the first six
months of life than infants of higher birth weight (68,
69). These findings are not reported in a way that
allows diarrhoea incidence by birth weight to be cal­
culated.
Since epidemiological data are limited, an alter­
native approach is to examine physiological function.
It is known that the immune response of LBW infants
is severely compromised (19, 38) and is more
adversely impaired than that of postnatally
malnourished infants (38). In addition to defects in
cellular immunity, LBW infants have been found to
have a significant reduction in maternal-fetal transfer
of IgG (19, 100) and impaired synthesis of IgA, IgM,
and the Ca component of complement (100). In a
follow-up study of 30 LBW infants in New Delhi,
impaired immunity was associated with recurrent
attacks of diarrhoea in the first 6 months of life. The
impairment was most marked in SGA infants (700). It
may be tentatively concluded from these limited data
that LBW infants, especially if they are small for
gestational age, may be predisposed to increased
diarrhoea morbidity rates. Furthermore, since mal­

171

nourished infants are more likely than well nourished
infants to die from diarrhoea (33), it is reasonable to
expect that LBW infants, especially if they are SGA,
will experience increased diarrhoea mortality rates.
Further research is warranted to confirm these
tentative conclusions.
Evidence suggests that breast-feeding protects
young infants from diarrhoea (36). In some coun­
tries, especially where babies are born in hospital, a
greater proportion of LBW infants may be fed artifi­
cially compared with infants of heavier birth weight.
Feeding mode may therefore also contribute to an
increased risk of diarrhoea among LBW infants.

Conclusions on hypothesis 1. Birth weight is a
major determinant of infant mortality and, in
developed countries at least, the effect of birth weight
on neonatal mortality is independent of socio­
economic status. Where information is available, and
this shows considerable local and regional variations,
the majority of LBW infants in developed countries
are preterm, whereas in developing countries the
majority are small for gestational age. Preterm AGA
infants suffer higher infant mortality rates than SGA
infants, and this difference is especially marked in the
neonatal period. On the other hand, preterm AGA
infants may grow better post-neonatally than SGA
infants.
We have located no satisfactory data on LBW as a
determinant of diarrhoea mortality or morbidity. The
strong association between LBW and mortality
(Tables 3 and 4) makes it likely that there is an
association between LBW and diarrhoea mortality in
developing countries where diarrhoea is a major cause
of infant death. Immunity is severely impaired in
LBW infants, especially SGA infants. Data from
India and Guatemala indicate that SGA infants may
be predisposed to increased diarrhoea morbidity.

Hypothesis 2. Improving the nutritional status,
health or life-style of pregnant women can reduce
the prevalence of low birth weight.

Etiology of low birth weight. Low birth weight is
caused by factors that shorten the length of gestation
and/or impair fetal growth. Factors affecting the
latter are more pertinent to this review since SGA
infants probably comprise the majority of low-weight
births in developing countries. Some factors, for
example altitude, are not amenable to change. Other
factors including socioeconomic status, multiple
pregnancy, maternal height, and pregravid weight
can only be influenced by long-term interventions. In
this review, only factors that are more readily
amenable to change and might respond to short-term

172

A. ASHWORTH & R.G. FEACHEM

interventions in developing countries are examined.
The factors selected arc poor nutritional status,
physically demanding work, maternal health,
maternal age, birth interval, and tobacco and alcohol
consumption. Only the first of these is discussed in
any detail.

Nutritional status. There is ample evidence that an
inadequate dietary intake during pregnancy adversely
affects birth weight. In the conditions of severe food
shortage that occurred in parts of Europe during the
Second World War, average birth weights fell by
185 g to 600 g (3, 28, 105).
In developing countries with seasonal food short­
ages, considerable fluctuations in birth weight occur.
In the village of Keneba in the Gambia, for example,
the mean monthly weight gains of pregnant women
were 0.4 kg in the wet season (July-October) and
1.4 kg in the dry season, and the corresponding
prevalences of LBW were 35% and 13% (86). In July
the average birth weight was 480 g less than in May
(94). However, the rainy season is not simply a time
of food shortage, it is also a period of increased
prevalence of maternal infection, including malaria,
and a period of increased agricultural activity. The
seasonal decrease in birth weight is thus likely to be
the result of several adverse factors which together
affect both maternal nutritional status and health.
A major focus of maternal and child health pro­
grammes has been the provision of additional food to
pregnant women. Unfortunately, few such interven­
tions have been evaluated. One exception is the
nationwide special supplemental food programme for
low-income American women, infants and children
(the WIC programme), which encourages clinic atten­
dance and provides individual nutrition counselling in
addition to the WIC foods (milk or cheese, eggs, ironfortified cereal, and fruit juices) through the Food
and Nutrition Service of the US Department of Agri­
culture. This programme has been associated with an
increased weight gain in pregnancy, an increase in
birth weight (+ 136 g, if supplemented for more than
6 months), and a decrease in the prevalence of LBW
from 10% to 6% (11, 29, 54). Of the services pro­
vided, only food supplementation had a significant
effect on birth weight (54).
Several small-scale feeding trials have been
evaluated and the results of nine such studies are
summarized in Table 6. In four of the studies the
observed improvement in birth weight was statisti­
cally significant although not always large (Mexico,
+ 213 g; Keneba, + 120 g; Guatemala, + 117 g; and
Montreal, +40 g). In the three developing countries
where the prevalence of LBW was > 15% (the
Gambia, Guatemala, and Mexico), supplementation
led to a statistically significant reduction in the pro­

portion of low-weight births." In China (Province of
Taiwan), the existing prevalence of LBW was low and
prc-supplementation birth weights were relatively
good, especially if one takes into account maternal
short stature (65). The participants in this study may
therefore not have been at much nutritional risk. It
may be concluded that, where maternal nutritional
status is poor, food supplementation that effectively
increases net intakes can improve birth weights and
reduce the prevalence of LBW.
In Harlem (New York City) a decrease in mean
birth weight was observed with a high-protein, highmineral supplement (99). Since similar findings have
been reported from San Francisco with the same sup­
plement (2) and from Motherwell (Scotland) when
pregnant women were advised to cat a predominantly
protein diet (44), caution has been expressed about
possible deleterious effects of excessively high protein
intakes in pregnancy. The apparent decrease in mean
birth weight with a protein-free supplement in
Birmingham (England) was probably due to imper­
fect matching of the groups (119).
The provision of a food supplement, however, does
not necessarily lead to its consumption. In some
societies pregnant women may purposely restrict their
intake in anticipation of an easier delivery, and thus
traditional dietary customs may limit any effort to
improve maternal nutrition, whether by dietary
advice or by direct food supplementation. For
example, in Project Poshak in India, food collection
rates were poor in pregnancy and 90% of the ration
which was collected was subsequently dispersed
among other family members (43). In Colombia
where food supplements were provided for each
family member, pregnant women consumed only
57% of their allocation. Moreover, the supplement
replaced some of their regular diet so that despite very
generous rations there was a relatively small incre­
ment in net intake (50). In contrast, in the Gambia a
high uptake of the supplement was achieved which
was attributed to its palatability and to the fact that it
was offered in the early morning when the women
would not normally have eaten at home (88). Its highenergy density assisted in achieving a substantial net
increase in energy intake.
The magnitude of the response in birth weight to
supplementation may appear disappointingly low.
However, where the data have been disaggregated,
the response among inadequately nourished mothers
is substantial. For example, supplementation
increased the mean birth weight during the wet season
in the Gambia by 225 g and decreased the prevalence
of LBW from 28% to 5% (88). In Colombia,
0 Similar results have also been reported from India, where
supplementation in the third trimester was associated with a
significant increase in birth weight (+170 g) and a 48°7o reduction in
the prevalence of LBW (129).

Habitual
daily intake

Country

Daily allocation
of supplement

Place
kcal

protein
(g>

Canada

Montreal

2250 (9.41)°

68

Colombia

Bogota

1610 (6.74)

Gambia

Keneba

1470 (6.15)

35
_/

Guatemala

4 villages

1500 (6.28)

40

Mexico

Rural areas

1950 (8.16)

50

China
(Province
of Taiwan) Sui-Lin

United
Kingdom

USA

kcal

protein
(g)

Net increase
in energy
intake
(kcal/day)

Duration of
supple­
mentation
(weeks)

Increase in
mean birth
weight (g)

Dose response
g/10 000 kcal
or 41.84 MJ'

Reference

_/

self-selected

40“

16

_/

97

860 (3.60)a

38

155 (0.65)“

13

51

21

41

1000 (4.18)

37

430 (1.80)

24

120“

68c

17

149 (0.62)

self-selected

117“

41d

29

275 (1.15)

34

213“

80c

28

50, 76
86-88
46, 58, 60,
61
22, 23

> 40

16

48

individualized

self-selected

300 (1.26)

20

800 (3.35)

40

Aberdeen

2060 (8.62)

70

290 (1.21)

15

189 (0.79)

12

37

Birmingham





425 (1.78)

0



10

-120

425 (1.78)

11



10

2060 (8.62)

80

320 (1.34)

6

207 (0.87)

> 10

470 (1.97)

40

261 (1.09)

> 10

Harlem

Reduction in
LBW births
(%)

12, 65
17
]

-



330





(

41

28



1

-32

-

-

1

° Figures in parentheses are equivalent units in megajoules (MJ), which is the approved SI unit to replace the thermochemical kilocalorie (kcal).
b Statistically significant increase in mean birth weight (P< 0.05).
e Statistically significant decrease in % of LBW births (P< 0.01).
d Statistically significant decrease in % of LBW births (P< 0.05).
f Grams of birth weight per 10 000 additional kcal (or 41.84 MJ) consumed.
1 — denotes data not available.

119

MENTION OF LOW BIRTH W EIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 6. Effect of food supplementation on mean birth weight and prevalence of LBW

174

A. ASHWORTH & R.G. FEACHEM

supplementation increased the mean birth weight by
181 g among mothers of low weight-for-height (50).
When the Guatemalan data were disaggregated by
socioeconomic score, supplementation reduced the
prevalence of LBW only among the more dis­
advantaged mothers. The reduction was from 29% to
13% (59). These examples emphasize the importance
of targeting food supplements to those most at risk.
Unfortunately, relatively little attention has been
directed towards identifying meaningful and practical
criteria for selecting pregnant women who would
benefit from dietary supplementation. Anthro­
pometric indices are likely to be the most appropriate
criteria (46, 121). Dynamic criteria such as weight
gain are preferable, but the necessity for at least two
measurements limits their feasibility.
Energy, rather than protein, appears to be the main
factor limiting fetal growth, and the total additional
calories consumed during pregnancy seem more
important than the trimester in which supplement­
ation is initiated (62). Where data permit, dose­
response values have been calculated in terms of the
mean increase in birth weight for every additional
10 000 kilocalories (41.84 MJ) consumed during
pregnancy (Table 6). The average values vary between
17 gand44 gper41.84 MJ (10 000 kcal). Where data
are disaggregated, the response increases with
decreasing nutritional status of the mother.
Space does not permit a full discussion of the
possible relationships between other specific nutri­
tional deficiencies and birth weight here. Anaemia in
pregnancy is common in developing countries (5) and
severe anaemia (haemoglobin of <3.7 mmol/1 or
60 g/1) is associated with LBW (104). Although there
may be a relationship between haematological status
and birth weight (40, 104, 122, 128), studies in dev­
eloped countries (49) and in India (109) suggest that
the administration of iron during pregnancy has no
detectable effect on birth weight or length of
gestation. These studies, however, have not examined
the effect of iron supplementation on birth weight in
severe anaemia, as opposed to mild anaemia. The
relationship between maternal folate status and birth
weight is not clear (96, 122, 128), but folate supple­
mentation may enhance birth weight (8, 52) and
further research is warranted. Poor maternal zinc
status is associated with SGA births (72,84). Prenatal
supplemental iron or folate may adversely affect
maternal zinc status (47, 74, 108).
Physically demanding work. A seasonal increase in
the energy expenditure of pregnant women may have
a greater impact on birth weight than a seasonal
decrease in energy intake. For example, in Keneba in
the Gambia, the decrease in birth weight preceded the
decrease in energy intake and mirrored the increase in
physical work (94), and in the village of Ikwiriri in the

United Republic of Tanzania no seasonal decrease in
birth weight was observed in 1979 when flood rains'
delayed field work (7). In a review of the pattern of
work during pregnancy in 112 traditional societies, in
45% of them the women continued full duties until
the onset of labour (53).
In the US Collaborative Perinatal Project, continu­
ation of employment during the third trimester was
associated with a reduction in mean birth weight of
between 150 g and 400 g. The effect was greatest for
women whose work involved standing, and those with
a low pregravid weight, hypertension, or a low weight
gain (78). In a Bombay cotton mill during 1925-28,
the mean birth weight was 139 g lower among women
mill-workers than among the non-working wives of
mill-workers, all of whom lived in similar grossly
overcrowded conditions. During 1928-29, when there
was a general strike for 6 months and the women
workers had an enforced rest, the mean birth weight
increased by 151 g and the prevalence of LBW
(< 5 pounds or 2268 g) decreased from 23% to 13%
(6). In Dakar, birth weights were found to be
relatively low in women involved in strenuous
physical work but whose nutritional status was
considered adequate (13). It is known that exercise
and an upright posture adversely affect placental
blood flow and it is for this reason that statutory
maternity leave was first introduced in the United
Kingdom. If exercise is undertaken in the heat, blood
has to be diverted to the skin as well as to the muscles,
thus reducing placental perfusion still further. Thus
certain types of work after mid-pregnancy may
impair fetal growth by reducing placental blood flow
and/or by affecting energy balance.
In many developing countries it is likely that a
reduction in the work load of pregnant women would
reduce the prevalence of LBW. Clearly this would
involve technological innovations and a radical
change in social attitudes and in the work patterns of
both men and women (31). Changes in social
attitudes and work patterns are unlikely to be
achieved in the short term.
Maternal health. Untreated maternal infections,
hypotension, and hypertension may affect fetal
growth. In endemic regions malaria is associated with
LBW. In a study in southern Nigeria, 24% of par­
turient women had malarial infection of the placenta
and the mean birth weight of their infants was 145 g
lighter than the infants of non-infected women (14).
Parasitaemia rates are highest among primigravidae
(120). In the Gambia, the difference in mean birth
weight between the infants of infected and non­
infected women was approximately 150 g for
firstborn infants and 60 g for all other births (67). In
the Solomon Islands, the mean birth weight rose sub­
stantially within months of starting anti-A nopheles

175

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

spraying (66). Between 1969 and 1971 the mean birth
weight increased by 252 g for babies of primigravidae
and by 153 g for all other babies. The overall
prevalence of LBW fell from 21% to 12%. Infections
of the urinary tract are associated with preterm
delivery (77) and with SGA births (48). In rural
Guatemala, the incidence of urinary tract infection
was 27 per 100 pregnancies (115). A number of
studies have reported a reduction in the prevalence of
LBW by antibiotic treatment of pregnant bacteriuric
women (7). Amniotic fluid infections are common
especially in undernourished gravidae, and infection
within the amniotic cavity may initiate preterm labour
(75). Poor maternal nutrition appears to interfere
with the normal antibiotic activity of the amniotic
fluid. In Ethiopia, antimicrobial activity was absent
in 75% of randomly selected urban women at term,
possibly as a result of zinc deficiency (113), and 31%
of women had localized chorioamnionitis at delivery
(79).
Maternal hypotension and hypertension are associ­
ated with reduced uteroplacental perfusion. In the
British Perinatal Mortality Survey, severe pre­
eclampsia
(diastolic
pressure
of
14.7 kPa
(110 mmHg) or more, or of 12 kPa (90 mmHg) or
more with proteinuria) was associated with a 3.3
relative risk of a LBW delivery and with a mean birth
weight reduction of 225 g (75). In Ethiopia, severe
growth retardation was found in fetuses whose deaths
were ascribed to pre-eclampsia (112).
Antenatal care of good quality is likely to improve
maternal health and thereby reduce the prevalence of
LBW. Such an effect has been documented in India
(80, 103), the United Republic of Tanzania (7), and
the USA (45). Unfortunately the existing coverage of
antenatal services is very low in most developing
countries and of the few women who use the services,
most do so only in the last trimester.
Maternal age. Teenage mothers have a higher
frequency of low-weight births for any given parity
(90). In 1976 in the USA, teenage mothers had a
relative risk of LBW of approximately 1.5 compared
with mothers aged 20-24 years (114). In New Delhi
the relative risk was 1.4 (97). A reduction in the pro­
portion of teenage pregnancies should reduce the
prevalence of LBW, but will be difficult to achieve in
the short term in some societies.
Birth interval. Studies in the United Kingdom (18),
the USA (Table 7), and Guatemala (69) have shown
that a short birth interval, especially of < 12 months,
is associated with an increased risk of LBW. In
developing countries, short birth intervals are becom­
ing more common especially in urbanized areas where
reduced breast-feeding is associated with a shortened
duration of postpartum amenorrhoea (118).

Table 7. Percentage of low-weight births among Black
and White populations, by interval since last birth, in 43
States of the USA, 1976°
Birth interval
(months)

Prevalence of LBW (%)
Black

White

<12

26.9

14.8

12-23

12.1

5.0

24-35

9.9

3.9

36-47

9.7

3.8

>48

9.8

4.7

° All data from ref. 114.

Tobacco smoking and chewing. It is well estab­
lished that maternal smoking reduces birth weight
(42) and that there is a linear dose-effect relationship
in which the more cigarettes are smoked, the greater is
the reduction (16, 39). The putative mechanisms
responsible are discussed elsewhere (26,51,57,63,64,
73, 127). Smoking more than 15 cigarettes/day
doubles the incidence of LBW. It has been shown that
the adverse effect of smoking is independent of social
class, maternal age, and parity (16,24,42). In a recent
review, average reductions in birth weight ranging
from 120 g to 430 g are reported (57). Tobacco
chewing also reduces birth weight (56), although this
has been little studied. In a prospective, randomized
experiment (101), a specific antismoking campaign
directed at pregnant smokers was associated with a
significant increase in mean birth weight ( + 92 g).
There was also a 24% reduction in the prevalence of
LBW, from 8.9% to 6.8%, although this was not
statistically significant.
Alcohol consumption. Animal studies suggest that
alcohol is embryotoxic and teratogenic. The ‘fetal
alcohol syndrome’ is found only among infants of
mothers who regularly consume over 80 g alcohol/
day. Characteristically these infants are growth
retarded at birth and show a consistent pattern of
congenital anomalies (107). Whether more moderate
alcohol consumption, before or during pregnancy,
adversely affects birth weight is less clear. Failure to
allow for confounding variables and difficulties in
obtaining accurate drinking histories may have con­
tributed to the apparently conflicting results. A recent
prospective investigation of 900 pregnancies in
London, which was well controlled for confounding
variables, has found that women consuming more
than 100 g alcohol/week (1-2 drinks/day), around
the time of conception, were more than twice as likely
as light drinkers to have a LBW infant (126). There
was no apparent benefit, in terms of birth weight,

176

A. ASHWORTH & R.G. FEACHEM

from reducing such drinking once pregnancy was con­
firmed. The effects of heavy binge drinking are cur­
rently being analysed. Moderate drinking (50-100 g
alcohol/week) was not associated with a significantly
increased risk of LBW.
Conclusions on hypothesis 2. Poor maternal nutri­
tion, certain infections, pre-eclampsia, arduous work
after mid-pregnancy, short birth intervals, and teen­
age pregnancy are likely to be causally associated with
LBW in developing countries. Tobacco and alcohol
consumption are additional risk factors. It follows,
therefore, that interventions that reduce the preval­
ence of these ‘causes' or their relative risks will reduce
the prevalence of LBW. Of the interventions
examined, maternal food supplementation has been
the most studied. If targeted to mothers at nutritional
risk, and if the food is consumed in addition to the
usual diet, the prevalence of LBW can be expected to
be reduced. However, food supplementation can be
expensive and the results from carefully supervised
feeding trials may be better than those that can be
achieved in national programmes. The effect of sup­
plementation with iron, zinc, or folate requires
further study.

Hypothesis 3. Improving the nutritional status,
health or life-style of pregnant women can reduce
diarrhoea morbidity or mortality rates in young
children.

The only true test of this hypothesis would come
from a study in which pregnant women received food
supplementation, improved health care, or some
other relevant intervention, and where the impact of

this on LBW and on the diarrhoea rates in their
children was monitored. A randomized controlled
trial would be ideal, but might encounter ethical and
logistical difficulties. No study of this type has been
located, although in the Gambian prenatal supple­
mentation study the preliminary findings
*
show
that there has been a sustained improvement in the
nutritional status of the wet season cohort. The most
likely explanation is that their increased birth weight
resulted in fewer, or shorter, episodes of diarrhoea. A
few studies have been reported in which food supple­
mentation for pregnant women, combined with other
interventions, was introduced and the impact of the
combined intervention on diarrhoea rates was
recorded. These results cannot be used to test
hypothesis 3 because the other interventions, such as
food supplementation for infants, are ones that may
well have independent effects upon diarrhoea rates.
Hypothesis 3 must be examined, therefore, by theor­
etical calculations of the reductions in diarrhoea
morbidity and mortality rates that might be achieved
by levels of birth-weight enhancement that may result
from improved maternal nutritional status, health or
life-style.
Three hypothetical populations are defined. First,
a relatively wealthy population having a prevalence
rate of LBW of < 10%; we call this population 1.
Second, an intermediate population, such as a poor
community in a developed country or a more wealthy
urban community in a developing country, having a
prevalence rate of LBW of 10-20%; we call this popu­
lation 2. Third, a relatively poor community having a
- PRENTICE, A. M. ET AL. Effect of prenatal supplementation
on birth weight and subsequent growth of infants. Paper presented at
the Fourth Asian Congress of Nutrition, Bangkok, 1-4 November
1983.

Table 8. Standardized birth-weight distributions for populations at three different socioeconomic levels
Percentage of live births
Birth weight
(g)

< 1501

1.0 X

1501-2000

1.5 >
5.0 '

2001-2500

Population 2
(intermediate)
% LBW= 10-20

Population 1
(relatively wealthy)
% LBW =<10

Population 3
(relatively poor)
% LBW= > 20

1.5 x

7.5

3.0 >

10.0

1.5 x

14.5

5.0 >
25.0 ’

2501-3000

20.0

28.0

45.0

3001-3500

40.0

36.0

20.0

3501-4000

25.0

17.0

3.0

>4000

7.5

4.5

0.5

3200

3000

2700

Estimated mean
birth weight (g)

31.5

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 9. Standardized relative risks of neonatal mortality
and post-neonatal mortality by birth weight compared to
birth weight of 2501-3000 g
Relative risk

Birth weight (g)

Neonatal
mortality"3

Post-neonatal
*
mortality

50.0

6.1

177

term this will clearly not be true and as LBW preval­
ence rates decline, so also will birth-weight-specific
mortality rates. It is the correct assumption to make
here, however, in order to separate out the impact of
interventions to increase birth weight from the impact
of other curative and preventive interventions.

These calculations do not take full account of the
fact that the amenability of LBW to change, and the
choice of interventions that will cause that change,
34
18.0
1501-2000
depend on three characteristics of the low-weight
1.8
4.0
2001-2500
births under consideration. First, amenability to
1.0
1.0
2501-3000
change and choice of intervention will depend on the
0.6
0.5
3001-3500
degree of LBW. Births of < 1500 g are unlikely to be
reduced by the improvements in maternal nutritional
0.4
0.5
3501-4000
status, health and life-style discussed here. To
0.5
0.6
> 4000
account for this, the prevalence rates of < 1500 g
° Median values from Table 3.
births have been set at 1.0% in population 1, and
b Median values from Table 4.
1.5% in populations 2 and 3 (Table 8), despite the fact
that the rates are actually > 2% among non-whites in
the USA (Table 1). Second, amenability to change
and choice of intervention will depend on the ratio of
prevalence rate of LBW of > 20%; we call this popu­
preterm to SGA births. SGA births are likely to be
lation 3. Using the data on birth-weight distributions
more responsive to the interventions discussed here
in Table 1, each of these three populations is assumed
than are preterm births. Third, and closely related to
to have a birth-weight distribution as set out in Table
the last point, as the overall LBW prevalence falls, the
8.
ratio of preterm to SGA births will rise and so the
The reductions in infant mortality rates in popu­
LBW pattern will become less amenable to change
lations 2 and 3 that may result from interventions that
increase birth weight can now be calculated on the through improvements in maternal nutritional status,
health and life-style. Thus, it is likely to be much
basis of the following assumptions.
easier to shift the birth-weight distribution of
(1) The effect of birth weight on risk of death does population 3 to population 2, than to shift that of
not extend beyond the first year of life. This may not population 2 to population 1 (Table 8).
be true but the assumption is necessary owing to lack
No data on the risk of diarrhoeal death by birth
of data. It is a conservative assumption.
weight have been located. It is therefore assumed that
(2) A LBW prevention programme in population 3 the relative risks of diarrhoeal death by birth weight
will shift the birth-weight distribution to that of popu­ are the same as those for all death (Tables 3, 4 and 9).
lation 2 (Table 8).
This assumption is unlikely to be correct for infants of
(3) A LBW prevention programme in population 2 very low birth weight (say, < 1500 g). However, such
will shift the birth-weight distribution to that of babies are assumed to make up a small and relatively
population 1 (Table 8). This is much less likely than unchanging proportion of all births in the 3 popula­
assumption 2, as discussed below.
tions being considered (Table 8). For higher birth
(4) The relative risks of neonatal and post-neonatal weights, and especially for post-neonatal mortality,
mortality by birth weight are as set out in Table 9, the assumption that all deaths and diarrhoeal deaths
derived from the median values in Tables 3 and 4. have the same risk pattern by birth weight is not
This may bias the relative risk towards that prevailing unreasonable and is supported by data from the
in developed countries where LBW infants are pre­ Dutch famine study where infant diarrhoeal deaths
dominantly preterm. Using the median value may and overall infant mortality increased in a similar
overestimate the impact of a LBW prevention pro­ manner in the SGA cohort born during the famine
gramme in North Arcot for example, where the UH).
relative risks appear to be less than the median. Until
The effects of shifting the birth-weight distribution
more data are available, this assumption is unavoid­ of population 3 to population 2, and that of popula­
able.
tion 2 to population 1, on the neonatal, post-neo­
(5) A LBW prevention programme changes the dis­ natal, infant, and 0-4-year age group diarrhoea death
tribution of birth weight (see assumptions 2 and 3) but rates are shown in Table 10. A population 3 to popu­
the neonatal and post-neonatal death rates for a speci­ lation 2 shift is what might be anticipated by a LBW
fied birth weight remain constant. In the medium prevention programme in a poor community in a
< 1501

178

A. ASHWORTH & R.G. FEACHEM

Table 10. The impact on diarrhoea mortality rates of shifting the distribution of birth weights11
% reduction in diarrhoea

birth-weight
distribution
shiftb

Population 3 to
population 2

Population 2 to
population 1

Neonatal
mortality

Post-neonatal
mortality

Infant
mortality^

% of 0-4 years
diarrhoea deaths
that occur in
infants

% reduction in
diarrhoea mortality
among 0-59-monthold children^

30

25

26

40

10

32

17

19

60

16

80

21

40

8

60

11

80

15

0 All calculations assume that the relative risks of diarrhoea death by birth weight are the same as the relative risks of all death by
birth weight (see text) and are as set out in Table 9.
b See Table 8 for the birth-weight distributions of populations 1, 2 and 3.
c Calculated by assuming that 14% of Infant diarrhoea deaths occur in the neonatal period (median figure obtained in 8 studies).
d Calculated by assuming that either 40%, 60% or 80% of diarrhoea deaths in children under 5 years old occur in children under
1 year, and that birth weight has no effect on the risk of diarrhoea death beyond 1 year of age (see text).

developing country. It presumes a reduction in LBW
from 31.5% to 14.5% (a 54% reduction, see Table 8),
which has been achieved in the Gambia and Mexico
(Table 6). It further presumes an increase in mean
birth weight of 300 g, something that appears far
harder to achieve on the evidence presented in Table
6. A population 2 to population 1 shift is what might
be anticipated by a LBW prevention programme in a
Third World city or in a relatively wealthy, develop­
ing country. It presumes a reduction in LBW from
14.5% to 7.5% (a 48% reduction, see Table 8), which
has been achieved in China (Province of Taiwan)
(Table 6). It further presumes an increase in mean
birth weight of 200 g, something that appears harder
to achieve on the evidence of Table 6.
Expected reductions in diarrhoea mortality rates
are 30-32% in the neonatal period, 17-25% in the
post-neonatal period, and 19-26% for the entire first
year of life (Table 10). Expected reductions in
diarrhoea mortality rate in the 0-4-year age group are
8-21%, assuming that LBW confers no excess risk of
diarrhoea death after 12 months of age (Table 10).
This last expected reduction is sensitive to the
proportion of 0-4-year diarrhoea deaths that occur in
the first year of life. Widely differing proportions are
reported; for instance, 90% in Recife, Brazil (91),
81% in Latin America (89), 68% in Ludhiana, India
(91), 55% in North Arcot, India (92), and around
40% in a number of studies in Asia and Latin America
where active surveillance was employed (106). Three
proportions (40%, 60% and 80%) are adopted for
comparison in Table 10.
There is no good evidence to suppose that either
a population 3 to population 2, or a population 2

to population 1, birth-weight-distribution shift is
achievable on a national scale by interventions in
maternal nutrition, health and life-style. As noted
above, the second of these shifts is especially unlikely
in the short term. The impacts computed in Table 10
are those expected if the birth-weight changes under
discussion were achieved. Lesser birth-weight
increases would result in lesser reductions in
mortality.
Conclusion on hypothesis 3. If it were possible to
intervene in maternal nutrition, health and life-style
in a developing country in a way that reduced the
prevalence of LBW from around 30% to around 15%
(Table 8), a fall in infant mortality rate of around
26% would be expected (Table 10). The fall in infant
diarrhoea mortality rate might be similar. The scarce
data on the relative risk of morbidity by birth weight
do not allow any comparable computations for
morbidity reductions to be made.

FEASIBILITY AND COST

In the consideration of hypothesis 2 above, a range
of interventions were shown to be likely to decrease
the prevalence of LBW. These interventions would
seek to increase energy intake, decrease arduous
work, improve antenatal health care, and reduce
tobacco and alcohol consumption in pregnant
women. Interventions would also seek to discourage
teenage pregnancies and increase the birth intervals.
We have reviewed five small-scale feeding trials in
developing countries (Table 6). We have no know­

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

ledge of any regional or nationwide programmes for
LBW prevention in developing countries using any of
the interventions mentioned above that have been
evaluated. We can therefore say little about the feasi­
bility and cost of such programmes.
Energy supplementation for pregnant mothers at
nutritional risk is perhaps the most likely of the inter­
ventions listed above to achieve reductions in LBW
prevalence in the short term. However, such an inter­
vention suffers from the very considerable cost,
logistical and other disadvantages of all supple­
mentary feeding programmes (33). Changes in work
patterns and attitudes towards work during
pregnancy may be extremely resistant to change in the
short term. Antenatal care is gradually improving in
most developing countries, and it is unlikely that a
LBW prevention programme could significantly
accelerate this process. Vigorous educational pro­
grammes may be able to reduce tobacco and alcohol
intake during pregnancy, but we know of no docu­
mented examples of such campaigns in developing
countries. Discouraging teenage pregnancy, and pro­
moting longer birth intervals, are part of existing
birth control programmes in several developing coun­
tries. They should have a beneficial effect upon birth
weights but this is secondary to the primary goal of
reducing the birth rate.
CONCLUSIONS

A substantial reduction in the prevalence of low
birth weight is theoretically possible and would lead

179

to a substantial reduction in infant mortality rate
(Table 10). The effect on diarrhoeal mortality or
morbidity rates is unknown, however. In this paper it
has been assumed that the relative risk of diarrhoeal
death by birth weight is the same as that for all death
(Table 9). Data are so scarce on the risk of morbidity,
diarrhoeal or other, by birth weight that no
computations have been possible.
Prospective studies are required that record from
birth the diarrhoea morbidity rates, and if circum­
stances permit, the diarrhoea mortality rates, of
groups of infants having known birth weights. These
infants should be followed for at least 12 months, and
preferably 24 months to determine if LBW confers an
excess risk of diarrhoea in the second year of life. The
relative risk values obtained may be used to compute
the reductions in diarrhoea morbidity and mortality
that would follow from a given improvement in birth­
weight distribution.
At the same time, greater attention should be paid
to the problem of LBW in developing countries. This
review has confirmed that, whatever its association
with diarrhoea, LBW is an important determinant of
infant mortality. For the more general goal of
reducing infant mortality it is necessary to know more
about the nature (preterm vs small for gestational
age), etiology, and prevention of LBW in developing
countries. In particular, it is necessary to know
whether interventions to reduce LBW are feasible at a
national or subnational scale in developing countries
and whether such interventions are cost-effective in
comparison with other strategies for reducing infant
mortality rates.

ACKNOWLEDGEMENTS
The authors are grateful for the constructive criticisms of earlier drafts of this paper provided by D. Blum, M. Campbell
Brown, I. de Zoysa, R. Hogan, B. McCarthy, L. Mata, M. Merson, A. Pradilla, A. Prentice, P. Shah, and D. Silimperi.
Editorial, bibliographical and secretarial assistance was most ably provided by Lynne Davies, Dianne Fishman, Maelorwen
Jones, Caprice Mahalla and Suzanne O’Driscoll.

RESUME
INTERVENTIONS CONTRE LES MALADIES DIARRHPlQUES CHEZ LE JEUNE ENFANT: PREVENTION D’UN FAIBLE

POIDS DE NAISSAANCE

Le present article constitue la cinquidme mise au point
d’une s£rie sur les interventions possibles en vue de reduire la
morbidity et la mortality associies aux maladies
diarrheiques chez les enfants de moins de cinq ans dans les
pays en developpement. On itudie ici 1’influence d’un faible

poids de naissance (FPN) sur la morbidity et la mortality par
diarrhies ainsi que les interventions visant A augmenter ce
poids. Tout en reconnaissant que 1’itiologie du FPN est
multifactorielle, 1’accent est mis sur les facteurs matemels
dont on pense qu’ils ont lq plus d’importance dans les pays

180

A. ASHWORTH & R.G. FEACHEM

en developpement et qu’ils sont susceptibles d’Svoluer dans
un proche avenir, moyennant des interventions convenables.
Sur 1’ensemble des naissances, on compte 16% de FPN
(< 2500 g). D’apres les observations, la majorite des cas de
FPN seraient des prematures dans les pays d6velopp6s, et
des enfants petits par rapport A leur age gestationnel (PAG)
dans les pays en developpement. Le poids de naissance
conditionne largement la mortalite infantile et, au moins
dans les pays developpes, son influence sur la mortality
neonatale ne depend pas de la situation socio-dconomique.
Chez les prematures de poids normal pour leur age
gestationnel (NAG), on observe un taux de mortalite
infantile plus elev£ que chez les nourrissons PAG mis &
terme, la difference 6tant particulierement marquee & la
periode neonatale. Aucune donn^e convaincante n’a pu etre
trouvee quant au role etiologique du FPN dans la morbidity
ou la mortalite d’origine diarrheique. L’association 6troite
entre FPN et mortalite fait qu’il existe probablement un lien
entre FPN et mortalite d’origine diarrheique dans les pays
en developpement ou les diarrhdes sont I’une des grandes
causes de mortalite infantile. L’immunite est gravement
amoindrie chez les nourrissons de faible poids de naissance,
specialement les nourrissons PAG.
Une nutrition mediocre chez la mere, certaines infections,
la nephropathie gravidique, un travail penible au-dete du
milieu de la grossesse, des naissances rapprochees et le jeune
age de la future mere, encore adolescente — tous ces facteurs
ont certainement un lien de cause & effet avec le FPN dans
les pays en developpement. L’usage du tabac et la consommation d’alcool sont des facteurs de risque supplementaires.
Il s’ensuit que les interventions qui rendent ces “causes”
moins frequentes ou qui abaissent le risque relatif
correspondant, sont de nature A r£duire la prevalence du
FPN. Parmi les interventions examinees, radministration ci
la mere de supplements nutritionnels a et6 la plus 6tudiee.
Cette intervention est en principe efficace si elle vise les

femmes en cause du fait de leur nutrition et si les aliments
distribu^s viennent completer leur ration habituelle.
Cependant, la distribution de supplements nutritionnels
peut etre couteuse, et des essais soigneusement supervises
pourraient donner de meilleures resultats que ceux qu’on
obtient dans les programmes nationaux. L’effet de
supplements de fer, de zinc ou de folate necessite des etudes
compiementaires.
Si 1’on pouvait agir sur la nutrition, 1’etat de sante et le
mode de vie des meres dans un pays en developpement de
fa?on & ramener la prevalence du FPN d’environ 30% &
15%, on devrait obtenir une baisse du taux de mortalite
infantile de 1’ordre de 26%. Le taux de mortalite infantile
par diarrhee devrait enregistrer une chute similaire. Les
rates donnees dont on dispose sur le risque relatif de
morbidite associe & un faible poids de naissance ne
permettent pas de faire des calculs analogues quant i la
reduction A attendre pour le taux de morbidite.
Il est indispensable de realiser des etudes prospeclives ou
seront enregistres des la naissance le taux de morbidite par
diarrhee (et, si les circonstances le permettent, le taux de
mortalite correspondant) dans des groupes de nourrissons
ayant un faible poids de naissance, de valeur connue. En
meme temps, il faudra se preoccuper davantage du
probleme du FPN dans les pays en developpement. La
presente etude a confirme que le FPN constitue, quel que
soit son lien avec la diarrhee, un determinant important de la
mortalite infantile. S’agissant de i’objectif, plus general,
d’un recul de la mortalite infantile, il faudrait en savoir plus
sur la nature (prematures ou enfants petits pour leur age
gestationnel), 1’etiologie et la prevention du FPN dans les
pays en developpement. En particulier, il convient de savoir
si les interventions visant ci limiter les cas de FPN sont
praticables dans ces pays, A l’6chelle nationale ou infra­
nationale, et si elles ont un rapport coOt/efficacite favorable
par comparaison & d’autres strategies envisageab’.es pour
faire reculer la mortalite infantile.

REFERENCES
1.

Abramowicz, M. & Kass, E. H. Pathogenesis and
prognosis of prematurity. New England journal of
medicine, 275: 1001-1007 (1966).
2. Adams, S. O. et al. Effect of nutritional supplemen­
tation in pregnancy. Journal of the American Dietetic
Association, 72: 144-147 (1978).
3. Antonov, A. N. Children born during the siege of
Leningrad in 1942. Journal of pediatrics, 30: 250-259
(1947).
4. Ayeni, O. & Oduntan, S. O. The effects of sex,
birthweight, birth order and maternal age on infant
mortality in a Nigerian community. Annals of human
biology, 5: 353-358 (1978).
5. Baker, S. J. & DeMaeyer, E. M. Nutritional
anaemia: its understanding and control with special
reference to the work of the World Health
Organization. American journal of clinical nutrition,
32: 368-417 (1979).

Balfour, M. I. & Talpade, S. K. The maternity
conditions of women mill-workers in India. Indian
medical gazette, 65: 241-249 (1930).
7. Bantje, H. Seasonal variations in birthweight distri­
bution in Ikwiriri village, Tanzania. Journal of tropical
pediatrics, 29: 50-54 (1983).
8. Baumslag, N. et al. Reduction of incidence of pre­
maturity by folic acid supplementation in pregnancy.
British medical journal, 1: 16-17 (1970).
9. Bhargava, S. K. et al. Outcome of babies with severe
intra-uterine growth retardation. I. Maternal factors,
congenital malformations, mortality and survival
pattern. Indian journal of medical research, 62:
367-374(1964).
10. Behrman, R. E. et al. Fetal and neonatal mortality in
white middle class infants. American journal of
diseases of children, 121: 486-489 (1971).
6.

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Berkenfield, J. & Schwartz, J. B. Nutrition inter­
vention in the community —the “WIC” program.
New England journal of medicine, 302: 579-581
(19S0).
12. Blackwell, R. Q. et al. Prospective maternal nutri­
tion study in Taiwan: rationale, study design, feasi­
bility, and preliminary findings. Nutrition reports
international, 7: 517-532 (1973).
13. Briend, A. Fetal stunting, fetal wasting and maternal
nutritional status. In: Aebi, H. & Whitehead, R.G.,
ed., Maternal nutrition in pregnancy and lactation.
Bern, Hans Huber, 1980, pp. 150-159.
14.
Bruce-Chwatt, L. J. Malaria in African infants and
children in southern Nigeria. Annals of tropical
medicine and parasitology, 46: 173-200 (1952).
15. Butler, N. R. & Alberman, E. D. Perinatal
problems: the second report of the 1958 British
perinatal mortality survey. Edinburgh, Livingstone,
1969.
16. Butler, N. R. et al. Cigarette smoking in pregnancy:
its influence on birth weight and perinatal mortality.
British medical journal, 2: 127-130 (1972).
17. Campbell Brown, M. Protein-energy supplements in
primigravid women at risk of low birthweight. In:
Campbell, D. M. & Gillmer, M. D. G., ed., Nutrition
in pregnancy. London, Royal College of Obstetricians
and Gynaecologists, 1983, pp. 85-98.
18. Chamberlain, G. et al. British births 1970. Vol 2,
Obstetric care. London, Heinemann, 1978.
19. Chandra, R. K. Fetal malnutrition and postnatal
immunocompetence. American journal of diseases of
children, 129: 450-454 (1975).
20. Chase, H. C. Infant mortality and weight at birth:
1960 United States birth cohort. American journal of
public health, 59: 1618-1628 (1969).
21. Chase, H. C. A study of risks, medical care and infant
mortality. American journal of public health, 63
(suppl.): 1-16(1973).
22. Chavez, A. Effects of malnutrition on infant body
morphology. In: Sterky, G. & Mellander, L., ed.,
Birth-weight distribution —an indicator of social
development. Uppsala, SAREC Report No. R:2, 1978,
pp. 18-20.
23. Chavez, A. & Martinez, C. Effects of maternal
undernutrition and dietary supplementation on milk
production. In: Aebi, H. & Whitehead, R. G., ed.,
Maternal nutrition during pregnancy and lactation.
Bern, Hans Huber, 1980, pp. 274-284.
24. Comstock, G. W. et al. Low birthweight and
neonatal mortality rate related to maternal smoking
and socioeconomic status. American journal of
obstetrics and gynecology. 111: 53-59(1971).
25. Cruise, M. O. A longitudinal study of the growth of
low birthweight infants. I. Velocity and distance
growth, birth to 3 years. Pediatrics, 51: 620-628
(1973).
26. Davies, D. P. et al. Cigarette smoking in pregnancy:
associations with maternal weight gain and fetal
growth. Lancet, 1: 385-387 (1976).
27. Dawson, I. et al. Birthweight by gestational age and
its effect on perinatal mortality in white and in Punjabi
births: experience at a district general hospital in West
11.

181

London 1967-1975. British journal of obstetrics and
gynaecology, 89: 896-899 (1982).
Dean, R. F. A. The size of the baby at birth and the
yield of breast milk. In: Studies of undernutrition,
Wuppertal 1946-9 (MRC Special Report Series, No.
275). London, HMSO, 1951, pp. 346-378.
29. Edozien, J. C.et al. Medical evaluation of the special
supplemental food program for women, infants, and
children. American journal of clinical nutrition, 32:
677-692(1979).
30. Erickson, J. D. & Bjerkedal, T. Fetal and infant
mortality in Norway and the United States. Journal of
the American Medical Association, 247: 987-991
(1982).
31. Fagley, R. M. Easing the burden of rural women: a
16-hour workday. Assignment children, 36: 9-28
.
(1976)
32. Falkner, F. Maternal nutrition and fetal growth.
American journal of clinical nutrition, 34: 769-774
.
(1981)
33. Feachem, R. G. Interventions for the control of
diarrhoeal diseases among young children: supple­
mentary feeding programmes. Bulletin of the World
Health Organization, 61: 967-979 (1983).
34. Feachem, R. G. Interventions for the control of
diarrhoeal diseases among young children: promotion
of personal and domestic hygiene. Bulletin of the
World Health Organization, 62: 467-476 (1984).
35. Feachem, R. G. & Koblinsky, M. A. Interventions
for the control of diarrhoeal diseases among young
children: measles immunization. Bulletin of the World
Health Organization, 61: 641-652 (1983).
36. Feachem, R. G. & Koblinsky, M. A. Interventions
for the control of diarrhoeal diseases among young
children: promotion of breast-feeding. Bulletin of the
World Health Organization, 62: 271-291 (1984).
37. Feachem, R. G. et al. Diarrhoeal disease control:
reviews of potential interventions. Bulletin of the
World Health Organization, 61: 637-640 (1983).
38. Ferguson, A. C. Prolonged impairment of cellular
immunity in children with intrauterine growth retard­
ation. Journal of pediatrics, 93: 52-56 (1978).
39. Fielding,}. E. Smoking and pregnancy. New England
journal of medicine, 298: 337-339 (1978).
40. Garn, S. M. et al. Hematological status and
pregnancy outcomes. American journal of clinical
nutrition, 34: 115-117 (1981).
41 Ghosh, S. et al. Mortality pattern in an urban birth
cohort. Indian journal of medical research, 69:
616-623 (1979).
42. Goldstein, H. Smoking in pregnancy: some notes on
the statistical controversy. British journal ofpreventive
and social medicine, 31: 13-17 (1977).
43. Gopaldas.T. etal. Project Poshak. In: CARE India.
New Delhi, 1975, Vol. 1, pp. 77-81.
44. Grieve, J. F. K. et al. Dieting in pregnancy: a study
of the effect of a high protein, low carbohydrate diet on
birthweight in an obstetric population. In: Sutherland,
M. W. & Stowers, J. M., ed.. Carbohydrate metab­
olism in pregnancy and the newborn. Berlin, Springer
Verlag, 1979, pp. 518-533.
28.

182

A. ASHWORTH & R.G. FEACHEM

Guyer, B. et al. Birth-weight-standardized neonatal 63. Lehtovirta, P. & Forss, M. The acute effect of
smoking on intervillous blood flow of the placenta.
mortality rates and the prevention of low birthweight:
British journal of obstetrics and gynaecology, 85:
how does Massachusetts compare with Sweden? New
729-731 (1978).
England journal of medicine, 306: 1230-1233 (1982).
64.
Longo, L. D. Carbon monoxide: effects on
46. Habicht, J. P. & Yarbrough, C. Efficiency in select­
oxygenation of the fetus in utero. Science, 194:
ing pregnant women for food supplementation during
523-525 (1976).
pregnancy. In: Aebi, H. & Whitehead, R. G., ed.,
65. McDonald, E. C. et al. The Bacon Chow study:
Maternal nutrition during pregnancy and lactation.
maternal nutritional supplementation and birth weight
Bern, Hans Huber, 1980, pp. 314-336.
of offspring. American journal of clinical nutrition,
47. Hambidge, K. M. et al. Zinc nutritional status during
34: 2133-2144(1981).
pregnancy: a longitudinal study. American journal of
66. MacGregor, J. D. & Avery, J. G. Malaria transmis­
clinical nutrition, 37: 429-442 (1983).
sion and fetal growth. British medical journal, 3:
48. Harris, R. E. et al. Asymptomatic bacteriuria in
433-436(1974).
pregnancy: antibody-coated bacteria, renal function,
67. McGregor, 1. A. et al. Malaria infection of the
and intrauterine growth retardation. American journal
placenta in the Gambia, West Africa; its incidence and
of obstetrics and gynecology, 126: 20-25 (1976).
relationship to stillbirth, birthweight and placental
49. Hemminki, E. & Starfield, B. Routine administration
weight. Transactions of the Royal Society of Tropical
of iron and vitamins during pregnancy: review of con­
Medicine and Hygiene, 77: 232-244 (1983).
trolled clinical trials. British journal of obstetrics and
68. Mata, L. J. Malnutrition-infection interactions in the
gynaecology, 85: 404-410 (1978).
tropics. American journal of tropical medicine and
50. Herrera, M. G. et al. Maternal weight/height and
hygiene, 24: 564-574 (1975).
the effect of food supplementation during pregnancy
69. Mata, L. J. The children of Santa Maria Cauquti a
and lactation. In: Aebi, H. & Whitehead, R. G. ed.,
prospective field study of health and growth.
Maternal nutrition during pregnancy and lactation.
Cambridge, MIT Press, 1978.
Bern, Hans Huber, 1980, pp. 252-263.
51. Herriot, A. et al. Cigarette smoking in pregnancy. 70. Mata, L. J. et al. Survival and physical growth in
infancy and early childhood. American journal of
Lancet, 1: 771-773 (1962).
diseases of children, 129: 561-566 (1975).
52. Iyengar, L. Folic acid requirements of Indian
71. Mata, L. J. et al. Antenatal events and postnatal
pregnant women. American journal of obstetrics and
growth and survival of children in a rural Guatemalan
gynecology, 111: 13-16(1971).
53. Jimenez, M. H. & Newton, N. Activity and work
village. Annals of human biology, 3: 303-315 (1976).
during pregnancy and the postpartum period: a cross72. Meadows, N. J. etal. Zinc and small babies. Lancet,
cultural study of 202 societies. American journal of
2: 1135-1137 (1981).
obstetrics and gynecology, 135: 171-176(1979).
73. Meyer, M. B. How does maternal smoking affect
54. Kennedy, E. T. etal. Evaluation of the effect of WIC
birth weight and maternal weight gain? American
supplemental feeding on birth weight. Journal of the
journal of obstetrics and gynecology, 131: 888-893
American Dietetic Association, 80: 220-227 (1982).
(1978).
55. Koops, B. L. etal. Neonatal mortality risk in relation 74. Milne, D. B. et al. Effect of oral folic acid supple­
to birth weight and gestational age: update. Journal of
ments on zinc, copper, and iron absorption and excre­
pediatrics, 101: 969-977 (1982).
tion. American journal of clinical nutrition, 39:
56. Krishna, K. Tobacco chewing in pregnancy. British
535-539 (1984).
journal of obstetrics and gynaecology, 85: 726-728
75. Minkoff, H. Prematurity: infection as an etiologic
(1978).
factor. Obstetrics and gynecology, 62: 137-144
57. Landesman-Dwyer, S. & Emanuel, I. Smoking
(1983).
during pregnancy. Teratology, 19: 119-126 (1979).
76. Mora, J. O. etal. Nutritional supplementation and
58. Lechtig, A. & Klein, R. E. Prenatal nutrition and
the outcome of pregnancy. I. Birth weight. American
birth weight: is there a causal association? In:
journal of clinical nutrition, 32: 455-462 (1979).
Dobbing,
J.,
ed.,
Maternal
nutrition
in
77. Naeye, R. L. Causes of the excessive rates of perinatal
pregnancy—eating for two? London, Academic
mortality and prematurity in pregnancies complicated
Press, 1981, pp. 131-156.
by maternal urinary-tract infections. New England
59. Lechtig, A. etal. Causes of low birthweight in Latin
journal of medicine, 300: 819-823 (1979).
America. Archives latinoamericanos de nutrition, 27
78. Naeye, R. L. & Peters, E. C. Working during
(Suppl. I, Part 2): 28-77 (1977).
pregnancy: effects on the fetus. Pediatrics, 69:724-727
60. Lechtig, A. et al. Effect of food supplementation
.
(1982)
during pregnancy on birthweight. Pediatrics, 56:
79. Naeye, R. L. et al. Amniotic fluid infections in an
508-520(1975).
African city. Journal of pediatrics, 90: 965-970
61. Lechtig, A. et al. Food supplementation during
.
(1977)
pregnancy, maternal anthropometry and birth weight
80. National Institute of Nutrition. Annual report
in a Guatemalan rural population. Journal of tropical
1979. Hyderabad, Indian Council of Medical
pediatrics, 24: 217-222 (1978).
Research, 1979, pp. 59-61.
62. Lechtig, A. et al. Effects of maternal nutrition on 81. Neligan, G. A. etal. Born too soon or born too small
infant health. Implications for action. Journal of
(Clinics in Developmental Medicine, No. 61), London,
tropical pediatrics, 28: 273-286 (1982).
Heinemann, 1976, pp. 54-63.

45.

PREVENTION OF LOW BIRTH WEIGHT FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

183

Nunes, R. M. [Study of courses of action relating to 99. Rush, D. et al. A randomized controlled trial of
prenatal nutritional supplementation in New York
human reproduction and nutrition in Recife (Brazil)].
Boletin de la Oficina Sanitaria Panamericana, 81:
City. Pediatrics, 65: 683-697 (1980).
304-312 (1976) (in Spanish with summary in English).
100. Saha, K. et al. A six-months’ follow-up study of
growth, morbidity and functional immunity in low
83. Paneth, N. etal. Social class indicators and mortality
birthweight neonates with special reference to
in low birthweight infants. American journal of
intrauterine growth retardation in small-for-gestaepidemiology, 116: 364-375 (1982).
tional-age infants. Journal of tropical pediatrics, 29:
84. Patrick, J. et al. Zinc and small babies. Lancet, 1:
278-282 (1983).
169-170(1982).
85. Prentice, A. M. Variations in maternal dietary 101. Sexton, M. &Hebel, J. R. A clinical trial of change in
maternal smoking and its effect on birth weight.
intake, birthweight and breast-milk output in the
Journal of the American Medical Association, 251:
Gambia. In: Aebi, H. & Whitehead, R. G., ed.,
911-915
(1984).
Maternal nutrition during pregnancy and lactation.
102. Shah, F. K. & Abbey, H. Effects of some factors on
Bern, Hans Huber, 1980, pp. 167-183.
neonatal and postneonatal mortality. Milbank
86. Prentice, A. M. et al. Long-term energy balance in
Memorial Fund quarterly, 49: 33-57 (1971).
child-bearing Gambian women. American journal of
103. Shah, P. M. & Udani, P. M. Analysis of the vital
clinical nutrition, 34: 2790-2799 (1981).
statistics from the rural community, Palghar —II.
87. Prentice, A. M. et al. Dietary supplementation of
Perinatal, neonatal and infant mortalities. Indian
lactating Gambian women. I. Effect on breast-milk
pediatrics,
6: 651-668 (1969).
volume and quality. Human nutrition: clinical
104.
Singla, P. N. et al. Effect of maternal anaemia on
nutrition, 37C: 53-64 (1983).
the placenta and the newborn infant. Actapaediatrica
88. Prentice, A. M. et al. Prenatal dietary supplemen­
Scandinavica, 67: 645-648 (1978).
tation of African women and birth-weight. Lancet, 1:
105. Smith, C. A. Effects of maternal undernutrition upon
489-492(1983).
the newborn infant in Holland (1944-1945). Journal of
89. Puffer, R. R. & Serrano, C. V. Patterns of mortality
pediatrics, 30: 229-243 (1947).
in childhood. Washington, Pan American Health
106. Snyder, J. D.&Merson, M. H. The magnitude of the
Organization, 1973 (Scientific Publication No. 262).
global problem of acute diarrhoeal disease: a review of
90. Puffer, R. R. & Serrano, C. V. Birthweight,
active surveillance data. Bulletin of the World Health
maternal age, and birth order: three important
Organization, 60: 605-613 (1982).
determinants in infant mortality. Washington, Pan
107. Sokol, R. J. et al. Alcohol abuse during pregnancy:
American Health Organization, 1975 (Scientific
an epidemiologic study. Alcoholism, 4: 135-145
Publication No. 294).
(1980).
91. Puffer, R. R. Mortality of infants and children under
108. Solomons, N. W. & Jacob, R. A. Studies on the bio­
5 years of age in India. New Delhi, USAID India,
availability of zinc in humans: effects of heme and
1982.
nonheme iron on the absorption of zinc. American
92. Rao, P. S. S. & Inbaraj, S. G. Perinatal, infant and
journal
of clinical nutrition, 34: 475-482 (1981).
early childhood mortality in North Arcot district of
109.
Sood, S. K. et al. WHO sponsored collaborative
Tamil Nadu, India. In: Longitudinal studies in human
studies on nutritional anaemia in India. Quarterly
reproduction (Monograph No. 6). Vellore, Christian
journal of medicine, 44: 241-258 (1975).
Medical College, 1973.
Starfield, B. et al. Mortality and morbidity in
93. Rao, P. S. S. & Inbaraj, S. G. A prospective study of 110.
infants with intrauterine growth retardation. Journal
infant mortality and congenital malformations in
of pediatrics, 101: 978-983 (1982).
relation to intra-uterine growth rates in south India.
111. Stein, Z. et al. Famine and human development: the
Indian journal of medical research, 67: 245-254
Dutch
hunger winter of 1944-1945. London, Oxford
.
(1978)
University Press, 1975.
94. Roberts, S. B. et al. Seasonal changes in activity,
112.
Tafari, N. & Naeye, R. L. Perinatal death due to pre­
birth weight and lactational performance in rural
eclampsia in an African city. East African medical
Gambian women. Transactions of the Royal Society of
journal, 55: 462-466 (1978).
Tropical Medicine and Hygiene, 76: 668-678 (1982).
95. Rooth, G. Low birthweight revised. Lancet, 1: 113. Tafari, N. et al. Failure of bacterial growth inhi­
bition by amniotic fluid. American journal of
639-641 (1980).
obstetrics and gynecology, 128: 187-189(1977).
96. Rothman, D. Folic acid in pregnancy. American
114.
Taffel, S. Factors associated with low birthweight:
journal of obstetrics and gynecology, 108: 149-175
United States, 1976. In: Vital and health statistics—
(1970).
Series 21, No. 37. Washington, DC, Department of
97. Rush, D. Nutritional services during pregnancy and
Health, Education and Welfare, 1980 (Publication No.
birthweight: a retrospective matched pair analysis.
(PHS) 80-1915).
Canadian Medical Association journal, 125: 567-576
115. Urrutia, J. J. etal. Infection and low birthweight in
(1981).
a developing country. American journal of diseases of
98. Rush, D. et al. The rationale for, and design of, a
children, 129: 558-561 (1975).
randomized controlled trial of nutritional supplemen­
116.
Van den Berg, B. J. Morbidity of low birthweight
tation in pregnancy. Nutrition reports international, 7:
and/or preterm children compared to that of the
547-553 (1973).
“mature”. Pediatrics, 42: 590-597 (1968).

82.

184

A. ASHWORTH & R.G. FEACHEM

Van den Berg, B. J. & Yerushalmy, J. The 123. Williams, R. L. & Chen, P. M. Identifying the
relationship of the rate of intrauterine growth of
sources of the recent decline in perinatal mortality rates
infants of low birthweight to mortality, morbidity, and
in California. New England journal of medicine, 306:
congenital anomalies. Journal of pediatrics, 69:
207-214(1982).
531-545 (1966).
124. World Health Organization. The incidence of low
1 IS. Van Ginneken, J. K. The impact of prolonged breast­
birthweight. A critical review of available information.
feeding on birth intervals and postpartum
World health statistics quarterly, 33: 197-224 (1980).
amenorrhoea. In: Mosley, W. H., ed., Nutrition and
125. World Health Organization. The incidence of low
human reproduction. New York, Plenum, 1978, pp.
birthweight: an update. Weekly epidemiological
179-195.
record, 59: 205-212(1984).
119. Viegas, O. A. C. et al. Dietary protein-energy 126.
Wright, J. T. et al. Alcohol consumption,
supplementation of pregnant Asian mothers at
pregnancy, and low birthweight. Lancet, 1: 663-665
Sorrento, Birmingham. II. Selective during third
.
(1983)
trimester only. British medical journal, 285: 592-595
127. Wynn, M. & Wynn, A. The prevention of handicap of
(1982).
early pregnancy origin. London, Foundation for
Education and Research in Child Bearing, 1982.
120. Watkinson, M.& Rushton, D. 1. Plasmodial pigmen­
128. Yusufji, D. et al. Iron, folate and vitamin Bu
tation of placenta and outcome of pregnancy in West
African mothers. British medicaljournal, 287:251-254
nutrition in pregnancy: a study of 1000 women from
(19S3).
southern India. Bulletin of the World Health Organiz­
121. Wharton, B. A. et al. Selection, compliance and
ation, 48: 15-22(1973).
logistics —lessons from the Sorrento supplementation
129. Bhatnagar, S. etal. Effect of food supplementation
study. In: Campbell, D. M. & Gillmer, M. D. G., ed.,
in the last trimester of pregnancy and early post-natal
Nutrition in pregnancy. London, Royal College of
period on maternal weight and infant growth. Indian
Obstetricians and Gynaecologists, 1983, pp. 101—111.
journal of medical research, 77: 366-372 (1983).
122.
Whiteside, M. G. et al. Iron, folic acid and vitamin
Bn levels in normal pregnancy, and their influence on
birthweight and the duration of pregnancy. Medical
journal of Australia, 1: 338-342 (1968).
117.

CH 3.5~
Bullerinofthe World Health Organization,62(3):467-476(1984)

© World Health Organization 1984

Interventions for the control of diarrhoeal diseases
among young children: promotion of personal and
domestic hygiene
*
R. G. Feachem*1
The effects of improving personal and domestic hygiene on diarrhoea morbidity are
reviewed using data from studies in hospitals, day-care centres, and communities. There is
evidence that low educational attainment and certain religious customs predispose to diar­
rhoea, presumably because of behavioural factors. The specific hygiene-related behaviour
that has been most studied is hand- washing. Hospital studies suggest that enteric infections
can spread via contaminated hands and that hands can be decontaminated by washing with
soap and water. Three studiesfrom Bangladesh, the USA, and Guatemala on the impact of
hygiene education programmes on diarrhoea are reviewed in detail. Reductions in
diarrhoea incidence rates of between 14% and 48% were documented in these studies. Little
is known on the impact of hygiene education programmes on diarrhoeas ofspecific etiology
or of their impact on diarrhoea mortality. Information is lacking on the optimal design of
such programmes, on their costs, and on their dependence on pre-existing levels ofsanitary
facilities. The available evidence suggests that hygiene education programmes may be a
cost-effective intervention for diarrhoea morbidity reduction. Research is necessary to
fill the current gaps in understanding and to clarify the operational aspects of these pro­
grammes.

Most of the pathogenic organisms that cause diar­
rhoea, and all the pathogens that are known to be
major causes of diarrhoea in many countries, are
transmitted primarily or exclusively by the faecaloral route. For some enteric pathogens, man is the
principal reservoir and thus most transmission orig­
inates from human faeces; examples are enterotoxi­
genic Escherichia coli, Shigella spp., Vibrio cholerae,
Giardia lamblia and Entamoeba histolytica. For
other enteric pathogens, animals are important reser­
voirs and transmission originates from both human
and animal faeces; examples are Campylobacter
Jejuni, Salmonella spp. and Yersinia enterocolitica.
For viral agents of diarrhoea the role of animal reser­
voirs in human disease remains uncertain.
Faecal-oral transmission may be water-borne,
food-borne, or direct. Water-borne transmission may
occur when water contaminated by faeces is drunk.
Food-borne transmission may occur when food con­
taminated by faeces is eaten. Direct transmission is
used here to describe an array of other faecal-oral
• Requests for reprints should be sent to the Director, Diarrhoeal
Diseases Control Programme, World Health Organization, 1211
Geneva 27, Switzerland.
1 Reader in Tropica! Public Health Engineering and Head of
Department of Tropica! Hygiene. London School of Hygiene and
Tropical Medicine, Keppel Street, London WC1E 7HT, England.
4420

routes such as via fingers, or objects such as eating
utensils, or bed linen, or dirt which may be ingested by
young children.
The interruption of water-borne and food-borne
transmission requires specific measures which will be
reviewed separately. Interrupting direct transmission
depends primarily on improved hygiene and on
improved facilities, such as better water supplies and
latrines that facilitate improved hygiene. Such
improved hygienic behaviour may also lessen the
chances of contamination of food by food handlers,
and may therefore reduce food-borne transmission.
Ideally governments will promote educational
measures to improve hygiene, as well as water and
sanitation projects to improve the physical facilities,
in integrated programmes. Such integrated pro­
grammes are advocated as part of the International
Drinking Water Supply and Sanitation Decade. In
practice, however, there are substantial operational
differences between hygiene education programmes
and water supply and sanitation projects: the two are
usually implemented by different ministries and
agencies, they require different types of personnel'
and, in particular, they have very different costs. It is
useful therefore to review the effectiveness of hygiene
education alone as an intervention for the reduction

—467 —

46S

R. G. FEACHEM

of diarrhoea morbidity or mortality. This review of
the role of hygiene education in diarrhoeal disease
control is the fourth in a series of reviews of potential
anti-diarrhoea interventions being published in the
Bulletin of the World Health Organization (6-9).

— studies that show an association between diar­
rhoea rates and levels of education;
— studies of diarrhoea epidemiology that inci­
dentally comment upon behavioural factors in trans­
mission;
— studies of behaviour and the transmission of en­
teric pathogens.

EFFECTIVENESS

These three types of study are considered in turn.

Hypothesis i. The transmission of enteric pathogens,
and thus the incidence of diarrhoea, are increased by
specific behaviours.

Diarrhoea rates and educational levels. The liter­
ature contains many observations that diarrhoea rates
are highest in families with the lowest levels of edu­
cational attainment. Hygiene and literacy may be
closely related (22). Such observations in themselves
are not useful because families with the lowest edu­
cational attainment will tend to be those with the low­
est income, poorest housing, most crowding, and
worst sanitary facilities. These confounding variables
will also promote the transmission of enteric path­
ogens.
In Bangladesh, Levine et al. (15) showed that fam­
ilies with no formal education had a 1.7 times higher
incidence of non-cholera diarrhoea, and a 1.8-3.4
times higher incidence of cholera, than families with
at least one high-school graduate. Although these
comparisons were controlled for tubewell usage, they
were not controlled for socioeconomic factors.
Levine et al. stated that families with high-school
graduates were relatively wealthy, judged by living
space, type of house construction, and possession of a
radio or watch. An interesting finding from earlier
studies on cholera in rural Bangladesh (17) was that
cholera incidence was higher among Hindus than
among Muslims. The incidence among Hindus was
3.0 times higher in 1963-64, 1.1 times higher in
1964-65, and 5.3 times higher in 1965-66 than among
Muslims.
This evidence is fragmentary and inconclusive.
Studies are required that compare diarrhoea rates by
literacy, educational attainment, or religious cus­
toms, with the environmental and wealth variables
controlled. If significant differences are found, the
most likely reason is that the educational or religious
differences cause behavioural differences that affect
the transmission of enteric pathogens. Detailed
anthropological studies will be needed to describe
'these behavioural differences. If such differences are
found, it remains to be shown whether the specific
behaviours can be changed by hygiene education in
the short-term (rather than by general education in
the longer-term or by changing certain religious
customs).

The anecdotal and descriptive evidence on this
topic is extensive; the rigorous and quantified evi­
dence is very limited. The evidence is mainly from
three sources:

Behavioural factors in diarrhoea epidemiology.
Numerous studies of diarrhoeal disease epidemi­
ology, and investigations of diarrhoea outbreaks,
comment on behavioural factors that may have in-

For hygiene promotion to be an effective diarrhoea
control intervention it must be true that:
either

the transmission of
enteric pathogens, and
thus the incidence of
diarrhoea, are increased
by specific behaviours

hypothesis
1

and
these specific behaviours
can be altered by appro­
priate hygiene education
programmes

appropriate hygiene
education programmes
can cause behavioural
changes which can reduce
the transmission of
enteric pathogens and
thereby reduce diarrhoea
morbidity or mortality
rates

hypothesis
2

hypothesis
3

Most of the literature on this topic deals with one or
more of these specific hypotheses. The potential
effectiveness of hygiene education would be sug­
gested by a demonstration either of the correctness of
hypotheses 1 and 2 or of the correctness of hypothesis
3. The evidence for and against these hypotheses is
examined below.

469

PROMOTION OF HYGIENE FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

fluenced the pattern of spread. An exhaustive review
of these comments would be unproductive since in
general they are speculative and do not firmly as­
sociate specific behaviours with specific levels of risk.
Most comments in studies from developing countries
draw attention to a complex of poverty, ignorance,
illiteracy, and crowding and suggest that associated
with these circumstances are behaviours that promote
the transmission of enteric pathogens. In developed
countries the two most commonly mentioned factors
are crowding, as may occur among lower socioecon­
omic groups or during exceptionally cold weather,
and poor hygienic practices of young children and
those who care for them. The risk of pathogen spread
from young children to other members of the family,
either directly or via the hands of the parent who
cleans them, is stressed repeatedly. The importance of
|hand-washing by staff in controlling the spread of
enteric infections in hospitals, day-care centres, and
other institutions is also emphasized.
Behaviour and the transmission of enteric path­
ogens. The specific behaviours that have received
most attention with regard to their role in promoting
the transmission of enteric pathogens are water­
handling behaviour, food-handling behaviour, and
hand-washing. Water-handling and food-handling
behaviours will be treated separately in the context of
interrupting water-borne and food-borne transmis­
sion and will be discussed in later reviews in this series.
Here, information on hand-washing is reviewed.
Concern about the possible role of hospital staff in
the spread of nosocomial infections has led to several
studies of hand-washing behaviour in medical insti­
tutions in developed countries. Three such studies
are summarized in Table 1. In some settings hand­
washing was found to be inadequate, both in fre­
quency and thoroughness. The data from Seattle are
particularly striking and suggest a lax attitude to per­
sonal hygiene even among those, such as physicians.

with a high level of theoretical understanding of the
need to maintain scrupulous hygiene in intensive care
units. This illustrates that knowledge is not neces­
sarily translated into practice. No studies on hand­
washing behaviour in the home, either in developed or
developing countries, have been located.
Studies on the occurrence and survival of enteric
pathogens on the hands are summarized in Table 2.
The hands of hospital staff in developed countries are
commonly contaminated and contamination takes
place easily during a variety of nursing procedures.
The hands of children were readily contaminated
by Shigella sonnei during shigellosis outbreaks in
England. Enteric bacteria on hands survive for at
least 3 hours in detectable numbers and can be trans­
ferred to food and to other hands. A study in Dhaka
(26) found that the hands of attendants of hospital­
ized children with rotavirus diarrhoea were com­
monly contaminated with rotavirus, and that this
contamination was more likely among those who
attended younger children.
Studies on the cleansing of hands by washing with
water and soap are summarized in Table 3. Hand­
washing with water and unmedicated soap removes
90-100% (below the limits of detection) of inoculated
bacteria. Washing with water alone removes a con­
siderable but lesser proportion. Some washing pro­
cedures with disinfectants do not achieve greater bac­
terial removals than washing with water and soap.
The opinion is often expressed in the literature that
the effectiveness of hand-washing is determined more
by its thoroughness (time taken and attention to all
parts of the hands) than by the types of soap or water
used. No data have been located on the effectiveness
of hand-washing in the home, or in developing coun­
tries, or using other procedures such as rubbing the
hands with sand or soil.
The studies summarized in Tables 1-3 are mainly
conducted in hospital settings. They indicate that

Table 1. Hand-washing and enteric pathogen transmission: hand-washing behaviour
Study

Findings

Reference

1

1. Hand-washing frequency by medical staff at
an intensive care unit in Seattle, USA, was
covertly observed.

Hand-washing occurred after only 41 % of
patient contacts. Physicians washed after
significantly fewer contacts (28%) than
nurses (43%).

2. Hand-washing behaviour among staff at a
radiotherapy clinic and a neonatal unit in
Helsinki, Finland, was observed over several
weeks.

Average hand-washing frequencies per person
per 8-hour shift were 10-20 at the radio­
therapy clinic and 27-42 at the neonatal unit.

21

3. Thoroughness of hand-washing by nurses in
England was studied by inviting them to wash
with a dye.

89% of nurses missed some part of the hand
surface; the most neglected areas were the
thumbs, backs of the fingers, and backs of
the hands. Right-handed nurses washed the
left hand better than the right hand, and vice
versa.

28

470

R. G. FEACHEM

Table 2. Hand-washing and enteric pathogen transmission: occurrence and survival of enteric pathogens on the
hands
Findings

Study

Reference

4

1. The transmission of Klebsiella spp. in an
intensive care unit in London, England, was
studied.

Klebsiella were commonly passed from

2. Occurrence and survival of Shigella sonnei on
the hands of children in Southampton,
England, during shigellosis outbreaks were
investigated.

In 4 studies, 0-49% children had S. sonnei on
their hands following a visit to the toilet for
urination. S. sonnet on the hands survived for
at least 3 hours. S. sonnei in stools passed
through double thickness of several brands of
toilet paper onto the hands.

10

3. Contamination by Gram-negative bacteria of
the hands of nurses at an intensive care
nursery in Florida, USA, was studied.

151 hand cultures were made from 13
nurses. 86% of cultures and 100% of nurses
were positive for Gram-negative bacteria.
Klebsiella pneumoniae and Escherichia colt
accounted for 55% of isolates. Evidence was
obtained that some Gram-negative bacteria,
including E. coli, could multiply and persist on
the hands of some nurses.

13

4. Occurrence and survival of enteric bacteria on
finger-tips were studied at the Central Public
Health Laboratory in London, England.

Escherichia coli were not isolated from the

23

A known number of coliforms were placed on
the hands of the author.

After three hours the number of coliforms
was "virtually unchanged".

24

6. Bacterial contamination of the hands of staff
on the general surgical and medical wards at a
hospital in New York, USA, was investigated.

Coliforms were found in 23% of hand rinses
from physicians, 55% from nurses, 67% from
nurse aides, and 67% from other staff. 18%
of cultures revealed > 103 coliforms per 2
hands. 88% of coliforms isolated were in the
Klebsiella-Aerobacter group and the
remainder were Escherichia coli. 92% of
coliforms isolated were resistant to one or
more antibiotics.

25

7. Contamination of the hands of the attendants
of 147 children under 5 years hospitalized
with acute diarrhoea in Dhaka, Bangladesh,
was studied. 70 of the children had rotavirus
diarrhoea.

Rotavirus antigen was detected in the hand
rinses from 79% of the attendants of children
with rotavirus diarrhoea, and from 20% of the
attendants of children with non-rotavirus
diarrhoea.

26

8. Effect of baby handling on hand bacteria of
nurses was studied in a hospital nursery in
New York, USA.

Changing soiled diapers increased the
coliforms on the hands by 10’— 105-fold.

27

5

patients (especially patients' hands) to
nurses' hands during simple and 'clean' nursing
procedures. Over 90% of Klebsiella on dry
hands could survive for at least 2.5 hours.

finger-tips of 100 laboratory staff but were
isolated from the finger-tips of 12% of
butchers in a meat factory. E. coli inoculated
onto finger-tips decreased by 99% or more
a.fter 1 hour. Salmonella inoculated onto
finger tips decreased by 96-99.8% after 1
hour. With an initial inoculum of 530 per
finger-tip, S. anatum were still detectable
after 3 hours. S. anatum were frequently
isolated from corned beef and cooked ham
that had been touched for 5 seconds by
contaminated finger-tips.

knowledge of the importance of hand-washing does
not necessarily lead to adequate hand-washing; that
hands become easily contaminated by faecal bacteria
and viruses even under conditions of good hygiene
and high awareness; that enteric bacteria on hands
can survive for at least 3 hours and can be transferred
to food and other hands; and that washing with soap
and water is an effective method of cleansing the
hands.

Conclusions on hypothesis 1. Much of the evidence
presented bears only indirectly on hypothesis 1. Low
educational attainment and certain religious customs
predispose to diarrhoea, presumably because of be­
havioural factors. The specific behaviour that has
been most studied is hand-washing. Hospital studies
suggest that enteric infections can spread via contami­
nated hands and that hands can be decontaminated by
washing with soap and water. Thus it is probable that

471

PROMOTION OF HYGIENE FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 3. Hand-washing and enteric pathogen transmission: cleansing the hands by washing with water and soap
Study

Findings

Reference

1. Experiments were conducted to determine
whether ordinary toilet soap, without anti­
bacterial additives, could act as a vehicle for
the dissemination of bacteria. Bacteria used
were Escherichia coli. Staphylococcus aureus.
two Gram-positive micrococci and Serratia

Bacteria inoculated onto the surface of soap
bars declined in number by at least 5 logio
units in 15 minutes. Washing massively
contaminated hands transferred bacteria to
the soap bar, but these bacteria did not
subsequently transfer to the hands of the
next user. Bars of soap under ordinary heavy
use did not accumulate appreciable bacterial
populations, although bars kept in non­
draining trays became somewhat more
contaminated than those which were allowed
to drain.

2

2. Effect of hand-washing with water (rubbing
hands together under 45 °C running tap
water for 20 seconds) and soap and water
(rinsing in warm water for 5 seconds,
washing with soap for 15 seconds, rinsing for
5 seconds) on removal of inoculated Klebsiella
from hands of staff in an intensive care unit
was studied.

Washing with water alone removed < 98% of
Klebsiella. Washing with plain soap removed
> 98% of Klebsiella in 50% of experiments
Washing with medicated soap removed
> 98% of Klebsiella in 77% of experiments.

4

3. Hands were contaminated with Staphylo­
coccus aureus or Pseudomonas aeruginosa
and then washed for 30 seconds with soap
and running water. Bacterial counts were
compared with those on unwashed (control)
hands that had been similarly contaminated.

Washing with soap and water reduced the
geometric mean counts of S a. by 99 7%,
and of P a. by 99.8%. Some washing
procedures using disinfectants did not remove
more bacteria than did washing with soap.

16

4. Hands were contaminated with Staphylo­
coccus aureus and then rinsed for 30 seconds
in distilled water. Bacterial counts were
compared with those on unrinsed (control)
hands that had been similarly contaminated.

Rinsing with distilled water reduced the
geometric mean counts of S. a. by 89.8%.
Rinsing with hypochlorite solution did not
remove significantly more S. a. than did
distilled water.

16

5. Effect of hand-washing (with soap and
running water for 15 seconds and then drying
on a paper towel) on removing inoculated
Salmonella anatum from finger-tips was
studied.

Proportion of experiments in which S. anatum
could be isolated from the finger-tips after
hand-washing depended on the initial
inoculum and was 100% for 106 S. anatuml
finger-tip, 30% for 10J-104 S. anatum/Hngertip, and 0% for < 10J S. anafi/m/finger-tip.

23

6. A variety of experiments on the removal of
resident and transient skin flora from hands
by scrubbing with soap and water were
conducted.

Bacterial removal was not affected by water
temperature (24-56 °C), type of soap, drying
on a sterile towel or bacteriological water
quality. Inoculated bacteria were reduced by
50% after washing with soap and warm
water for 30 seconds.

24

7. Effect of rapid hand-washing with soap and
water, or water alone, on removing naturally
acquired coliforms from the hands of nurses
in a hospital nursery was studied.

Hand-washing with soap and water removed
67-100% of coliforms (median, 96%). Hand­
washing with water alone removed 93-100%
of coliforms (median, 98%). The use of
disinfectants in rapid hand-washing did not
improve coliform removal.

27

marcescens.

certain specific behaviours do promote the trans­
mission of enteric pathogens and that failure to wash
the hands is one such behaviour.
Hypothesis 2. Specific behaviours (that promote the
transmission of enteric pathogens) can be altered by
appropriate hygiene education programmes.

The literature on the methods and efficacy of hy­
giene education is mainly comprised of theoretical
discussions of approaches likely to succeed or of

qualitative descriptions of field experience (77, 20,
29). Few reports that quantify the impact of a given
hygiene education programme on a specific set of
personal or domestic hygiene behaviours have been
located.
Torun (30) reports an evaluation of a hygiene edu­
cation programme in a village in the Pacific lowlands
of Guatemala during 1979-80. The programme was
directed at 106 mothers, all of whom had a child
under 6 years old; 32 similar mothers acted as con­
trols. The programme consisted of nine 1-hour
sessions between educators and groups of mothers

472

R. G. FEACHEM

(9-27 per group), using stories and discussions
assisted by radio plays and evocative pictures. The
mothers were encouraged to reflect upon their hy­
giene problems and to commit themselves to specific
actions. The content of the educational programme
covered the recognition and treatment of diarrhoea,
excreta disposal, hand-washing, breast-feeding, food
hygiene, care of drinking water, and diet. The propor­
tions of mothers giving correct answers to questions
on prevention were 56% before the programme, 90%
immediately after the programme, and 88% 6 weeks
later. A significant increase was observed in the pro­
portion of target families that were judged to have
correct hygiene behaviour with respect to diaper dis­
posal, kitchen hygiene, water storage, latrine hygiene,
garbage disposal, and child cleanliness. Hygiene prac­
tices that would have required expenditure (e.g.,
improvements to wells and animal enclosures) were
not significantly changed. Diarrhoea incidence was
reduced in the children of target mothers compared to
control mothers, and these data are reviewed below
under hypothesis 3.
The body of both theoretical and qualitative evi­
dence, taken together with evidence from other
spheres of health education (for instance, smoking,
obesity, breast-feeding), strongly suggests that the
adoption of hygienic behaviour can be achieved by
sustained and culturally appropriate educational pro­
grammes. Research is urgently needed to measure the
behavioural impact of various types of hygiene edu­
cation in various cultural and socioeconomic settings.
Such research should not prove unduly difficult or
expensive.

Hypothesis 3. Appropriate hygiene education pro­
grammes can cause behavioural changes which can
reduce the transmission of enteric pathogens and
thereby reduce diarrhoea morbidity or mortality
rates.
Three studies (from Bangladesh, the USA, and
Guatemala), documenting the impact on diarrhoea
rates of hygiene education programmes, have been
located. In two cases, Bangladesh and USA, the edu­
cation focused exclusively on hand-washing, while in
the third, in Guatemala, the programme sought to
improve several aspects of personal and domestic
hygiene.
The Bangladesh study. In Dhaka, Khan (/2)
selected patients with culture-confirmed shigellosis
attending a clinic and allocated their families to four
groups: a soap and water group that were provided
with 2-4 pieces of soap and 1-3 water pitchers and
were urged to wash their hands after defecation and
before eating; a soap group that was provided with
soap only; a water group that was provided with

pitchers only; and a control group that was provided
with nothing. Rectal swabs of family contacts of the
index shigellosis cases in the four groups were ob­
tained daily for 10 days. Contacts infected by the
same type of Shigella as the index case were termed
secondary infections, and those who were also sick
(3 or more episodes of diarrhoea or dysentery in 24
hours) were termed secondary cases. The secondary
case rate was 2.2% for the soap and water group and
14.2% for the control group, and Khan concluded
that the intervention had lowered the secondary case
rate by 84%. In the soap and water group, secondary
infection rates were significantly higher among those
who used less water for washing and bathing. This dif­
ference was less apparent among the control group.
The reduction in secondary infection rates was less for
Sh. dysenteriae type 1 than for other Shigella species,
possibly because of the lower infectious dose of that
organism. Attack rates of non-Shigella diarrhoea
were 37% lower in the soap and water group than in
the control group over the 10-day period of surveil­
lance.
To make the reduction in the shigellosis secondary
case rate more nearly comparable with the reduction
in the non-Shigella diarrhoea attack rate, the second­
ary case rate reduction was converted to an attack rate
reduction." On the assumption that the hand-washing
promotion had no influence on the incidence rate of
index cases, an 84% secondary case rate reduction is
equivalent to a 35% reduction in attack rate in the
families under study.
The USA study. In Atlanta, GA, Black et al. (3)
investigated the impact of hand-washing on diarrhoea
incidence in four day-care centres. Two groups of
children, one aged 6-17 months and the other aged
18-29 months, were studied at each day-care centre.
Two of the centres were randomly selected to receive a
hand-washing promotion campaign which encour­
aged staff to wash their hands after arriving at the
centre, before handling food, and after helping a
child to use the toilet or using the toilet themselves.
When children entered the centre, used a toilet, had
their diapers changed, or were prepared to eat, staff
washed the children’s hands using bar soap and paper
towels. Children using the toilet were supervised by
staff to ensure that they did not place their hands in
their mouths. These practices were rigorously moni­
tored. The other two day-care centres received no
hygiene promotion and served as controls. It had been
observed, prior to this investigation, that the practice
of hand-washing and of toilet supervision of young
children in these centres was generally lax. The inci­
dence of diarrhoea among the selected children in the
four centres was monitored for 10 months; this incidence among children aged 6-29 months was reduced
" Assumptions and computations used to make this conversion
are available on request from the author.

473

PROMOTION OF HYGIENE FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 4. The effect of hygiene education on diarrhoea incidence and the percentage of days ill with diarrhoea in a
village in Guatemala*1
Proportion of days ill

Diarrhoea incidence
Age
group
(months)

Full year'’
or peak
diarrhoea
seasonr

0-23

Full year

Peak season

0-71

Full year

Peak season

Target or
control
group

Mean
number of
children
studied
per month

Mean
monthly
incidence
(episodes/
100
children/
month)

Percentage
reduction
in incidence

Mean
number of
child-days
studied
per month

Mean
monthly %
of days
with
diarrhoea''

Percentage
reduction
in % of
days with
diarrhoea

14

1433

4.5

24

906

5.9

1752

4.1

872

9.2

4457

3.0

2577

3.4

5378

2.7

2253

5.2

Target

49

36

Control

32

42

Target

60

38

Control

32

59

Target

152

25

Control

92

29

Target

185

25

Control

82

37

36

14

32

55

12

48

4 Reanalysis of data from Torun (30).
6 September 1979 to August 1980.
c March-June 1980.
d (Number of child-days with diarrhoea x 100) - Total number of child-days observed

by 48% in the hand-washing day-care centres com­
pared with the control centres.

The Guatemalan study. In Florida Aceituno, a
village in the Pacific lowlands of Guatemala, Torun
(30) promoted health awareness and hygienic
behaviour among mothers and studied the impact on
their knowledge and behaviour and the effect on diar­
rhoea rates in their children under 6 years old. The
promotion, and its impact on knowledge and hygienic
practice, are reviewed above under hypothesis 2. The
impact on diarrhoea rates in children is summarized
in Table 4. The target group comprised the children of
106 mothers who participated in the educational pro­
gramme, and the control group comprised the chil­
dren of 32 mothers who did not participate. Twelve
(38%) of the 32 control mothers were of above
average socioeconomic status, being the wives of
store-owners, preachers or community leaders. The
impact on the proportion of days with diarrhoea (a
measure combining possible impacts on incidence and
duration of episodes) was higher than the impact on
incidence. For both impact measures, impact was 2-4
times greater in the peak diarrhoea season (March to
June) than throughout the whole year.

Conclusions on hypothesis 3. These 3 studies from
Bangladesh, the USA, and Guatemala provide 5
measures of the impact of hygiene education on diar­
rhoea rates:

— a 35% reduction in the incidence rate of shig­
ellosis among all ages in urban families in Bangla­
desh;
— a 37% reduction in the incidence rate of nonShigella diarrhoea among all ages in urban families in
Bangladesh;
— a 48% reduction in the incidence rate of all
diarrhoea among children aged 6-29 months in day­
care centres in the USA;
— a 14% reduction in the incidence rate of all diar­
rhoea among children aged 0-71 months throughout
the year in a Guatemalan village;
— a 32-36% reduction in the incidence rate of al!
diarrhoea among children aged 0-71 months during
the peak diarrhoea season in a Guatemalan village.
Thus the reduction of diarrhoea incidence rate to be
anticipated from hygiene education lies in the range
14-48%. Other studies support the general conten­
tion that hygiene education can reduce diarrhoea

474

R, G. FEACHEM

rates, but do not allow a calculation of the reduction
in incidence achieved by a clearly defined educational
intervention (14, IS, 19).
It might be supposed that the commonness of direct
person-to-person transmission of Shigella would
make shigellosis especially sensitive to reduction by
hand-washing. The evidence presented here does not
support this. First, the computed reduction in the
shigellosis incidence rate in Dhaka was 35%, whereas
in the same families the reduction in the incidence rate
of non-Shigella diarrhoea was 37%. Second, in the
Atlanta study a 48% reduction was recorded and, out
of 85 diarrhoea stools cultured, none contained
Shigella.
The three studies summarized above suggest that
hygiene education, especially hand-washing pro­
motion, has a marked impact on diarrhoea morbidity
rates. These studies should be repeated in different
socioeconomic and environmental settings and
should also quantify the impact on diarrhoea due to
rotavirus, enterotoxigenic Escherichia coli, Campylo­
bacter jejuni. Shigella. Giardia lamblia and other
agents that are of known local importance.
FEASIBILITY AND COST

There is such little documented experience of hy­
giene education programmes that their feasibility is
difficult to judge and their costs are unknown.
Experience with other types of health education sug­
gests that such programmes are feasible on either a
national or local level and that they can employ a com­
bination of mass media techniques and direct inter­
action between target families and hygiene promoters
(5). Costs of hygiene education are probably low
compared to some other interventions for diarrhoea
morbidity reduction — such as the provision of im­
proved water supplies and sanitation facilities. The
effectiveness of hygiene education may depend, how­
ever, upon the presence of such facilities. In Dhaka
the provision of soap, which would be costly on a con­
tinuing basis, may have been an essential part of the
intervention, and in Atlanta modern facilities for
washing and defecation were already available. Oper­
ational research is needed to clarify the most effective
and feasible types of hygiene education programme,
to detail their costs, and to assess their dependence on
pre-existing levels of sanitary facilities.

CONCLUSIONS

Interest in the role of education in disease control
has increased considerably in recent years. It is prob­
able that better educated communities enjoy relative
protection against several diseases compared to less
educated, but otherwise similar communities. This
protection may be conferred both by general edu­
cation (as measured, for instance, by school attend­
ance, adult literacy or education of heads of house­
holds) and by disease-specific education. Disease­
specific education can be preventive or therapeutic in
content.
The evidence marshalled in this paper suggests that
hygiene education can improve hygiene and can re­
duce diarrhoea morbidity rates by 14-48%. These are
hopeful findings. Many countries, especially in subSaharan Africa and Asia, are having extreme diffi­
culty in sustaining the development of their health
infrastructure. In the water and sanitation sector, for
instance, the rate of construction of new projects
barely keeps pace with population growth and the rate
of breakdown is alarming. In these circumstances,
educational interventions appear especially attrac­
tive. They may be cheap compared to infrastructure
projects and they may achieve lasting changes in
health-related behaviour. Most importantly, as indi­
cated by this review, they may achieve substantial
impacts.
Hygiene education programmes are being con­
ducted in many countries and should continue. Coun­
tries not having such programmes should seriously
consider launching them. Research is necessary, how­
ever, to improve the cost-effectiveness of hygiene edu­
cation. This research is of three main types. First,
more information is needed on the associations be­
tween specific behaviours and risks of diarrhoea mor­
bidity and mortality of known etiology. Second, oper­
ational research is needed to clarify the most effective
and feasible types of hygiene education programme,
to detail their costs, and to assess their dependence on
pre-existing levels of sanitary facilities. Third, impact
studies should be conducted to clarify the impact on
diarrhoea of carefully designed hygiene education
programmes. These impact studies should be eti­
ology-specific and, where possible, should document
impacts on diarrhoea mortality rates as well as
morbidity rates.

ACKNOWLEDGEMENTS
The author is grateful for the constructive criticisms of earlier drafts of this paper provided by R. C. Ballance, R. Black,
D. Blum, I. de Zoysa, R. Hogan, M. U. Khan, M. Merson, A. Moarefi, S. C. Pal, G. Schultzberg, D. Silimperi and
1. Tabidzadeh. Secretarial and bibliographical assistance was most ably provided by Maelorwcn Jones.

PROMOTION OF HYGIENE FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

475

RESUME
LUTTE CONTRE LES MALADIES DIARRHEIQUES CHEZ LES JEUNES ENFANTS:
PROMOTION DE L’HYGIENE PERSONNELLE ET DOMESTIQUE

Cet article est le quatrieme d’une serie portant sur les
mesures qui permettraient de red u ire la mortalite ct la mor­
bidite par diarrhee chez les enfants de moins de 5 ans dans
les pays en developpement. On a etudie les effets d’une meilleure hygiene personnels et domestique sur la morbidite par
diarrhee en se basant sur des donnees recueillies dans des
hopitaux. des garderies d'enfants et des collectivities. Ces
donnees montrent qu’un faible niveau d’instruction et
certaines coutumes religieuses predisposent aux maladies
diarrheiques, probablcment en raison de certains factcurs
comportementaux. La pratique d’hygiene qui a cte le plus
etudiee a cet egard est celle qui consiste a se laver les mains.
Les etudes faites en milieu hospitalier demontrent que les
infections intestinaies peuvent etre propagees par des mains
contaminecs et que Ton peut eliminer cc risque en se lavant
les mains a 1’eau et au savon. Trois etudes faites au Bangla­
desh, aux Etats-Unis d’Ameriquc et au Guatemala sur

1’impact que les programmes de promotion de 1’hygiene ont
sur la diarrhee sont passees en revue dans Particle. Ces
etudes ont mis en Evidence des reductions des taux de mor­
bidite par diarrhee allant de 14% a 48%. On sait peu de
chose des effets des programmes d’hygiene sur les diarrhees
d’etiologic determinee ou sur la mortalite par diarrhee. On
ne dispose pas non plus de renseignements sur la conception
optimale de tels programmes, sur leurs couts, non plus que
sur la mesure dans laquelle leur succes depend de 1’existence
prcalablc d’installations sanitaires. D’apres les donnees
disponiblcs cos programmes constituent probablcment une
mesure d’un bon rapport cout-efficacite pour la reduction
de la morbidite par diarrhee. Des rccherches doivent etre
faites pour combler les lacunes existantes dans les connaissances et pour eclairer les aspects operationncls des pro­
grammes d’cducation en maticre d’hygiene.

REFERENCES
1. Albert, R K. & Condie, F. Hand-washing patterns in
medical intensive-care units. New England journal of
medicine, 304: 1465-1466 (1981).
2. Bannan, E. A. & Judge, L. F. Bacteriological studies
relating to handwashing. 1. The inability of soap bars to
transmit bacteria. American journal of public health,
55: 915-922 (1965).
3. Black, R. E. et al. Handwashing to prevent diarrhea
in day-care centers. American journal of epidemiology,
113:445-451 (1981).
4. Casewell, M. & Phillips, I. Hands as route of trans­
mission for Klebsiella species. British medical journal,
2:1315-1317 (1977).
5. Cerqueira, M. T. etal. A comparison of mass media
techniques and a direct method for nutrition education
in rural Mexico. Journal of nutrition education, 11:
133-137 (1979).
6.
Feachem, R. G. Interventions for the control of diar­
rhoeal diseases among young children: supplementary
feeding programmes. Bulletin of the World Health
Organization, 61: 967-979 (1983).
7. Feachem, R. G. et al. Diarrhoeal disease control: re­
views of potential interventions. Bulletin of the World
Health Organization, 61: 637-640 (1983).
8.
Feachem, R. G. & Koblinsky, M. A. Interventions
for the control of diarrhoeal diseases among young chil­
dren: measles immunization. Bulletin of the World
Health Organization, 61: 641-652 (1983).
9. Feachem, R. G. & Koblinsky, M. A. Interventions for
the control of diarrhoeal diseases among young chil­
dren: promotion of breast-feeding. Bulletin of the
World Health Organization, 62: 271-291 (1984).

10.

Hutchinson, R. 1. Some observations on the method
of spread of Sonne dysentery. Monthly Bulletin of the
Ministry of Health and the Public Health Laboratory
Service, 15: 110-118 (1956).
11. Isely, R. B. Evaluating the role of health education
strategies in the prevention of diarrhea and dehy­
dration. Journal of tropical pediatrics, 28: 253-261
(1982).
12. Khan, M. U. Interruption of shigellosis by hand wash­
ing. Transactions of the Royal Society of Tropical Med­
icine and Hygiene, 76: 164-168 (1982).
13. Knittle, M. A. et al. Role of hand contamination of
personnel in the epidemiology of Gram-negative noso­
comial infections. Journal of pediatrics, 86: 433-437
(1975).
14. Koopman, J. S. Diarrhea and school toilet hygiene in
Cali, Colombia. American journal of epidemiology,
107: 412-420(1978).
15. Levine, R. J. et al Failureof sanitary wells to protect
against cholera and other diarrhoeas in Bangladesh.
Lancet, 2: 86-89 (1976).
16. Lowbury, E. J. L. et al. Disinfection of hands:
removal of transient organisms. British medical
journal, 2: 230-233 (1964).
17. McCormack, W. M. et al. Endemic cholera in rural
East Pakistan. American journal of epidemiology, 89:
393-404 (1969).
18. Nutting, P. A. et al Reduction of gastroenteritis
morbidity in high-risk infants. Pediatrics, 55: 354-358
(1975).

476

R. G. FEACHEM

Odumosi . M. O. The response of mothers to health 25. Salzman, T. C. et al. Hand contamination of per­
sonnel as a mechanism of cross-infection in nosocomial
education and the incidence of gastro-enteritis among
their babies at lle-lfe, Nigeria. Social science and
infections with antibiotic-resistant Escherichia coli and
medicine, 16: 1353-1360(1982).
Klebsiella-Aerobacter. Antimicrobial agents and
20. Ogionwo, \V. Socio-psychological factors in health
chemotherapy, 7: 97-100 (1967).
behaviour: an experimental study on methods and
26. Samadi, A. R. et al. Detection of rotavirus in hand­
attitude change. International journal of health
washings of attendants of children with diarrhoea.
education, 16 (Suppl.): 1-16(1973).
British medical journal, 286: 188 (1983).
21. Ojajarvi, J. et al. Failure of hand disinfection with 27. Sprunt, K. et al. Antibacterial effectiveness of routine
frequent hand washing: a need for prolonged field
hand washing. Pediatrics, 52: 264-271 (1973).
studies. Journal of hygiene, 79: 107-119 (1977).
28. Taylor, L. J. An evaluation of handwashing tech­
22. Pacey, A. Hygiene and literacy, Waterlines, 1: 26-29
niques — 1. Nursing Times, 74: 54-55 (1978).
(1982).
29. Tonon, M. Models for educational interventions in
23. Pether, J. V. S. & Gilbert, R. J. The survival of
malnourished populations. A merican journal of clinical
salmonellas on finger-tips and transfer of the organisms
nutrition, 31: 2279-2283 (1978).
to foods. Journal of hygiene, 69: 673-681 (1971).
30. Torun, B. Environmental and educational interven­
24. Price. P. B. The bacteriology of normal skin: a new
tions against diarrhea in Guatemala. In: Chen, L. C.
quantitative test applied to a study of the bacterial flora
& Scrimshaw, N. S., ed.. Diarrhea and malnutrition:
and the disinfectant action of mechanical cleansing.
interactions, mechanisms and interventions. New York,
Journal of infectious diseases, 63: 301-318 (1938).
Plenum Press, 1982, pp. 235-266.
19.

Bulletin of the H orld Health Organization, 63 (2): 295-315 (1985)

World Health Organization 1985

Interventions for the control of diarrhoeal diseases
among young children: chemoprophylaxis
*
1. DE ZOYSA*1 & R. G. FEACHEM2

A number of situations place young children at increased risk of diarrhoea. Among
these, the best documented in developing countries is contact with a diarrhoea case in a
family or household. The most common application of chemoprophylaxis in developing
countries is to prevent cholera or shigellosis among household contacts of known cases.
There is little evidence that chemoprophylaxis is effective in reducing diarrhoea morbidity
and mortality, except perhaps in travellers. Theoretical calculations in this paper (based on
optimistic assumptions) suggest that chemoprophylaxis of household contacts of known
cholera cases in Bangladesh might reduce overall diarrhoea incidence rates in children
under 5 years of age by 0.02-0.06% and diarrhoea mortality rales by 0.4-1.2%.
Chemoprophylaxis of household contacts of known shigellosis cases might reduce overall
diarrhoea incidence rates by 0.15-0.35 % and diarrhoea mortality rates by 0.3-0.7% in the
same age group. The correct identification of index cases of cholera and shigellosis,
followed by the rapid distribution of drugs to their household contacts, requires skills and
resources that are scarce in the developing countries. Chemoprophylaxis can contribute to
the widespread emergence and dissemination of antimicrobial resistance. The available
evidence suggests that chemoprophylaxis is not feasible in many settings and that, even if
successfully implemented, it is not a cost-effective intervention for national diarrhoeal
diseases control programmes.

The main application of drugs in the control of
diarrhoeal diseases is in the treatment of selected cases
in order to reduce the duration and severity of illness
and prevent death. Additionally, because the
duration of excretion of the infectious agent may
sometimes be reduced, mass chemotherapy, or the
widespread administration of drugs to cases and to
infected asymptomatic persons, has been recom­
mended (90) for the purpose of reducing the pool of
excreters and thereby the potential for transmission.
Drugs may also be used to protect uninfected indivi­
duals from infection or illness. In this way, antimicro­
bials have been given prophylactically to individuals
at high risk, such as close contacts of known cases or
travellers to endemic areas. In practice, mass
chemotherapy and individual chemoprophylaxis
repeated on a large scale merge into each other
because the presence of infection is not always
ascertained or recognized.

In this review, we define chemoprophylaxis of
diarrhoea as the administration of drugs to persons
exposed to a recognized risk, whether infected or not,
to prevent diarrhoea in these persons and to reduce
the sources of infection. We consider here the role of
chemoprophylaxis in national programmes to reduce
diarrhoea morbidity and mortality among children
under 5 years of age. This paper is the sixth in a series
of reviews of potential anti-diarrhoea interventions
published in the Bulletin of the World Health
Organization (3, 38-42).

■ Requests for reprints should be sent to the Director, Diarrhoeal
Diseases Control Programme, World Health Organization, 1211
Geneva 27, Switzerland.

a considerable proportion of
diarrhoea morbidity or mortality
in young children occurs in
hypothesis
children who are exposed to a
1
recognized risk, such as contact
with a known case

EFFECTIVENESS

For chemoprophylaxis to be an effective inter­
vention to control diarrhoeal diseases it must be true
that:
either

1 Save the Children Fund Research Fellow, Department of
Tropical Hygiene, London School of Hygiene and Tropical
Medicine, Keppel Street, London WC1E 7HT, England.
- Reader in Tropical Public Health Engineering and Head of
Department of Tropical Hygiene, London School of Hygiene and
Tropical Medicine, London, England.
4526

— 295 —

296

I. DE ZOYSA & R. G. FEACHEM

Table 1. Additional cases and infections among household contacts of index cases having diarrhoea of known
etiology

Etiology of index case

Country

Infection
rate
(%)

Reference

21

29

10

10

23

Age group
of contacts
(years)

Case
rate
(%)

0-1
2-4

Enterotoxigenic Escherichia coli

ST/LT strains

Bangladesh

ST strains

All strains

Giardia lamblia

Norwalk agent

Rotavirus

15

4

10

0-4

8

42

All ages

1

9

Canada

All ages

2-5

7-13

70

USA

All ages

17

ND"

119

USA

All ages

30-32

ND"

USA

0-4

40

ND"

1
5

Thailand

All ages

19

ND"

Canada

All ages

21

46

New Zealand

0-12

66

75

All ages

38

46

0-14

62

62

All ages

36

44

0-12

15

22

All ages

13

24

34

147
55

59

151

USA

Adults

8

35

66

USA

Adults

8

55

USA

Adults

71

ND"

78
121

USA

All ages

15

ND"

119

United Kingdom

0-14

24

ND"

149

All ages

18

ND"

Bangladesh

type 1
Bangladesh

Bangladesh

Bangladesh

Marshall Islands

Shigella flexneri & sonnei

22

10

USA

Shigella flexneri

11

22

2-4

Sweden

Shigella dysentenae

4

0-1

All ages

Norway

Salmonella spp.

All ages

USA

USA

0-4

30

48

All ages

19

35

0-4

31

31

All ages

13

20

0-4

11

11

All ages

4

7

124
76

77

0-4

0

11

All ages

4

21

0-4

33

50

All ages

13

32

All ages

6

ND"

144

0-4

68

ND"

103

All ages

56

ND"

0-14

50

40

All ages

36

31

75
77

111

297

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 1 {continued}

Etiology of index case

Country

Shigella sonnei

United Kingdom

USA

USA

USA

Shigella spp.

Bangladesh

Age group
of contracts
(years)

Case
rate
(%)

Infection
rate
(%)

0-7

34

45

All ages

16

33

0-4

81

ND"

All ages

51

ND"

Reference

23

86
157

0-5

45-54

ND"

All ages

27-36

ND"

All ages

46

ND"

146

0-4

24

ND"

73

All ages

14

ND"

Bangladesh

All ages

14

32

71

USA

All ages

26

ND"

119

Classical

Bangladesh'■

All ages

4-16

11-24

37

El Tor

Six countries7'

All ages

2-25

4-32

Vibrio cholerae

“ No data.
h Feachem (37) reviewed 20 studies of cholera attack rates in households of index cases in Bangladesh. India. Hong Kong, Israel.
Philippines and China (Province of Taiwan).

Hypothesis 1. A considerable proportion of
diarrhoea morbidity or mortality in young children
occurs in children who are exposed to a recognized
risk, such as contact with a known case.

and
individuals receiving chemo­
prophylaxis have lower
diarrhoea morbidity rates
or mortality rates or severity
than otherwise similar
individuals

hypothesis
2

or
chemoprophylaxis in young
children exposed to a
recognized risk, such as contact
with a known case, can reduce
overall diarrhoea morbidity
rates or mortality rates or
severity in young children

hypothesis
3

The potential effectiveness of chemoprophylaxis
would be suggested by a demonstration either of the
correctness of hypotheses 1 and2orof the correctness
of hypothesis 3. We examine below the evidence for
and against the three hypotheses and we consider the
magnitude of reductions in diarrhoea morbidity and
mortality in young children that may be achieved by
chemoprophylactic measures.

Several recognized situations place young children
at increased risk of diarrhoea. We describe here these
high-risk situations and consider their significance
with regard to overall diarrhoea morbidity and
mortality in young children.

Contact with a known case in the family or
household. Numerous studies have been conducted
on the transmission of diarrhoeal diseases within
households. Data on additional cases and infections
among household contacts of a known index case are
set out in Table 1 (in this context, the index case is the
first recognized case in the household). These studies
used different methods. Most were prospective, with
observation periods ranging from 1 to 65 days. AU
index cases had diarrhoea of known etiology and, in
some studies, only cases and infections caused by the
same organism as the index case were included in the
analysis. In other studies microbiological surveillance
of contacts was not carried out and all additional
cases of diarrhoea were included.
AU additional cases and infections among house­
hold contacts following the identification of an index
case were recorded and therefore both co-primaries
and secondaries are included. We are concerned here
with the risk of disease among contacts after
identification of an index case and it is of no

298

I. DE ZOYSA & R. G. FEACHEM

importance whether the additional cases and
infections resulted from a common-source exposure
or secondary spread. The terms additional case and
additional infection, rather than secondary case and
secondary infection, are used throughout this
paper.
Additional case rates among household contacts
ranged from 1 c7o to 71% and additional infection
rates ranged from 4% to 55%. Both rates tended to be
higher in the younger age groups. When infected,
young children were also more likely to have
symptoms than older children and adults. Not only
were young children at greater risk of contracting
diarrhoea after introduction of the disease into the
household, but they also played a major role in the
introduction and spread of the disease in the house­
hold (70, 52, 74, 77, 95, 149, 157).
Household contacts, therefore, are at considerable
risk of contracting diarrhoea after identification of an
index case. This risk is highest immediately after onset
of diarrhoea in the index case, and wanes rapidly
thereafter. Among household contacts of known
cholera cases, 12-43% of additional cases were
identified on the first of 10 days of observation and
85-88% by the sixth day (8, 95, 116, 139).
In the above studies, the household was the unit of
investigation and it was usually assumed that all
household contacts were members of a single family.
Increased risk of cholera, enterotoxigenic Escherichia
coli (ETEC) diarrhoea, and shigellosis has also been
reported among persons in the same cluster of houses
or neighbourhood as an index case (18, 34, 61).

Attendance at day-care centres and schools. A
number of studies from industrial countries have
examined the role of day-care centres and schools in
the spread of diarrhoeal diseases. It has been shown
that children attending these centres have a higher risk
of contracting diarrhoea than children who remain at
home (9, 25, 28). Black et al. (11) found that children
were at increased risk of diarrhoea two to four weeks
after enrolment in a day-care centre, which suggests
that illness follows contact with other infected or ill
children. Investigations within day-care centres have
shown a pattern of repeated outbreaks caused by a
variety of pathogens and sometimes having high
attack rates (Table 2). Studies from the United
Kingdom and North America provide evidence that
young children in close contact with other children in
day-care centres or schools play a major role in
introducing diarrhoea into their households and
spreading it to other young children (9, 23, 25, 70,
119, 146, 157, 158).
No prospective studies of this kind from developing
countries have been located. Outbreaks of diarrhoeal
disease in day-care centres and schools have been
described in countries such as Brazil (145) and China

Table 2. Attack rates during
recorded in day-care centres

diarrhoea

outbreaks

Etiology

Country

Attack rate
(%)

Reference

Astrovirus

Japan

52

80

Campylobacter
jejuni

Belgium

20-50

83

Clostridium
difficile
Giardia lamblia

USA

20-58

Canada

20-39

79
70

USA

27-35

USA

17

Rotavirus

Shigella spp.

Multiple
etiologies

Belgium

58-78

USA

100

USA

71

United Kingdom

51

USA

73

USA

38-51

USA

33

USA

36-50

USA

51-57

USA

14-92

9
119
43
121
119
150
48
158
119
146
119
35

(136), and it is probable that day-care centres, schools
and informal neighbourhood play groups where
many children congregate constitute high-risk
situations for children everywhere.

Contact with a known case in institutions and
hospitals. Institutions such as residential homes for
children and institutions for the mentally retarded are
associated with a high risk of diarrhoeal disease in
industrial countries, where diarrhoea incidence rates
are low in the community. Despite the prophylactic
and therapeutic use of drugs and the application of
isolation techniques, diarrhoeal diseases (particularly
shigellosis, giardiasis and amoebiasis) continue to be
a significant health problem among institutionalized
children in these countries. Children are at highest
risk of infection and illness shortly after admission
(19, 29).
Hospital-acquired diarrhoeal diseases are well
documented in industrial countries and may well
represent a significant problem in developing
countries. In a recent review of publications on
nosocomial infections originating from developing
countries (159), most reported diarrhoea outbreaks in
nurseries for premature babies and paediatric wards.
Hospital outbreaks may be serious (115) and may
lead to considerable disruption of patient services
(81). In industrial countries at least, a considerable

299

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

urban (95) and rural areas (106, 138). fs recent report
from Tanzania (99) has documented the role of a
hospital in the spread of cholera in an urban centre.

proportion of the diarrhoeal diseases treated in
hospital may be nosocomially acquired (100, 113,
127, 140). The etiological agent implicated is often
rotavirus (127), reflecting its importance as a cause
for hospitalization among young children.

Importance of these high-risk situations with
regard to overall diarrhoea morbidity and mortality.
We have described various situations associated with
a high risk for diarrhoea, but it is not clear, from the
evidence presented so far, what proportion of all
diarrhoea morbidity and mortality occurs in these
situations. There are limited data on the importance
of these high-risk situations in the epidemiology of
cholera and shigellosis. We here summarize some of
these reports:
— Cholera. Detailed studies from Hong Kong,
China (Province of Taiwan), and the Philippines
suggest that cholera did not spread easily, if at all,
within households or among households in a com­
munity (44, 90, 104, 118, 148, 154, 164). Reported
cases were in general sporadic, dispersed throughout
the community, and contact-tracing identified few
additional cases or infections. In Bangladesh, on the
other hand, active transmission within households
and from household to household within neighbour­
hoods has been reported and most cases tended to
appear in short family or community outbreaks in

— Shigellosis. Data from industrial countries
suggest that young children attending day-care
centres and schools may play a major role in the
spread of shigellosis to the community at large by
spreading infection in their own households and from
household to household (103, 126, 146, 150, 157,
158). Institutions, on the other hand, have not been
demonstrated to be a source of infection for the
community at large. Shigellosis in institutions may,
nonetheless, represent a substantial proportion of all
cases (20, 123). Data from developing countries on
the mode of spread of shigellosis through com­
munities are limited to epidemic situations in which
whole communities are suddenly at high risk of
diarrhoeal disease. High attack rates and death rates
have been recorded (94, 120) but clustering by
household or neighbourhood does not appear to be
prominent (47, 144).
Conclusions on hypothesis 1. We have described
several situations that place young children at
increased risk of diarrhoea. Among these, the best
documented in developing countries is contact with a
known case in a family or household. The most
common application for chemoprophylaxis in

Table 3. Detectable index and additional cholera cases for various values of hospitalization rate, household size and
additional case rate
No of detectable
additional cases
per 100 cases
*'

No. of detectable
index plus additional
cases per 100 cases

24

6

30

17

13

30

25

13

17

30

10

5

21

9

30

30

10

15

13

17

30

30

10

25

9

21

30

Additional
case rate"
(%)

No. of detectable
index cases
per 100 cases'’

6

5

6

15

30

6

30

Hospitalization
rate (%)

Household
size
(persons)

30
30

50

6

5

40

10

50

50

6

15

29

21

50

50

6

25

22

28

50

50

10

5

34

16

50

50

10

15

21

29

50

50

10

25

15

35

50

" See Table 1.
h A detectable index case is an index case who is hospitalized.
r A detectable additional case is one occurring in the household of a detectable index case in the 10 days following the reporting of
the index case.

I. DE ZOYSA & R. G. FEACHEM

300

Table 4. Detectable index and additional shigellosis cases for various values of hospitalization rate, household size
and additional case rate
No of detectable
additional cases
per 100 cases1’

No. of detectable
index plus additional
cases per 100 cases

2.5

2.5

5.0

2.0

3.0

5.0

40

1.7

3.3

5.0

10

20

1 8

3.2

5.0

5

10

30

1.4

3.6

5.0

5

10

40

1 1

3.9

5.0

Hospitalization
rate (%)

Household
size
(persons)

Additional
case rate1'
(%)

No. of detectable
index cases
per 100 cases"

5

6

20

5

6

30

5

6

5

10

6

20

5.0

5.0

10.0

10

6

30

4.0

6.0

10.0

10

6

40

3.3

6.7

10.0

10

10

20

3.6

6.4

10.0

10

10

30

2.7

7.3

10.0

10

10

40

2.2

7.8

10 0

” See Table 1.
b A detectable index case is an index case who is hospitalized.
r A detectable additional case is one occurring in the household of a detectable index case in the 10 days following the reporting of
the index case.

developing countries is to prevent cholera or
shigellosis among household contacts of known
cases. We here estimate on a theoretical basis the
proportion of all cholera and shigellosis cases that
occurs among household contacts of known cases,
and is thus potentially preventable by chemo­
prophylaxis.
In most circumstances, a case of cholera or
shigellosis will only be identified if he or she is
hospitalized. The calculations for cholera are set out
in Table 3." Data from Bangladesh and the
Philippines indicate that hospitalization rates for
cholera vary between 23% and 74% (4, 6, 72, 74, 92,
105, 116, 160, 161). Hospitalization rates of 30% and
50% are adopted in Table 3. Two household sizes,
comprising 6 and 10 persons, are used. Six is
approximately the mean family size in many
developing countries while ten represents the larger
extended families, or households comprising, on
average, just under two families. Three additional
cholera case rates (5 %, 15 % and 25 %) are adopted on
the basis of data summarized in Table 1. The pro­
portions of all cholera cases that are detectable
additional cholera cases are 6-21% for a hospitali­
zation rate of 30% and 10-35% for a hospitalization
rate of 50%.
• The formula used to make these calculations is available on
request from R.G.F.

Similar calculations for shigellosis are presented in
Table 4.“ A hospitalization rate for shigellosis of 8%
has been reported from rural Bangladesh (13, 14) and
rates of 5% and 10% are used in Table 4. Three
additional shigellosis case rates (20%, 30% and 40%)
are adopted on the basis of data summarized in Table
1. The proportions of all shigellosis cases that are
detectable additional shigellosis cases are 2.5-3.9%
for a hospitalization rate of 5%, and 5.0-7.8% for a
hospitalization rate of 10%.
The proportions of all cases that are detectable
additional cases, and thus potentially preventable by
chemoprophylaxis, are 6-35% for cholera and
2.5-7.8% for shigellosis. These proportions are
correlated both with the additional case rate and with
the hospitalization rate. Thus for shigellosis, despite
the fact that additional case rates are high (20-40%),
the proportion of detectable additional cases is low
because the hospitalization rates are low (5-10%).
Hospitalization rates depend on the severity of the
symptoms and on the hospital facilities available in
the area. The hospitalization rates adopted in Tables 3
and 4 are mainly derived from studies in the Matlab
area of Bangladesh. The longstanding presence of a
hospital specializing in acute diarrhoeal diseases, and
the availability of ambulance services in this area,
may lead to higher hospitalization rates than in most
other rural areas of developing countries.

Results

Country
(place)

Bangladesh (Dhaka)

Period of
follow-up

10 days

Drug

Tetracycline

V. cholerae (%)

Proportion of samples
from household contacts
which were positive
for V cholerae (%)

4 doses daily x 5 days

0

-

Single dose daily x 5 days

1



Dosage

Single dose
Placebo
India0 (Calcutta)

10 days

Tetracycline

2 doses daily x 3 days

Placebo
India (Calcutta)

15 days

10 days

8

-

13

-

-

1.5

-

3

Sulfadoxine

Single dose

19"

3

Tetracycline

2 doses daily x 3 days

21"

3

42"

7

Placebo
India (Calcutta)

Proportion of household
contacts who excreted

Doxycycline

Placebo

Single dose

15'

2

*
23

4

° V. cholerae was isolated from only 60% of index cases.
b Excluding contacts infected only on day 1 but including contacts infected on day 1 and on subsequent days.
c Excluding contacts infected on day 1.

Reference

91

62

24

134

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 5. Summary of trials of chemoprophylaxis among household contacts of cholera cases

302

I. DE ZOYSA & R. G. FEACHEM

Hypothesis 2. Individuals receiving chemo­
prophylaxis have lower diarrhoea morbidity rates
or mortality rates or severity than otherwise similar
individuals.

Chemoprophylaxis of diarrhoea has been
recommended and used in numerous situations: to
limit the spread of cholera epidemics (22, 46, 90, 98),
to control shigellosis in institutions, such as
institutions for the mentally retarded in the USA (7,
49, 50, 84, 163), and to prevent travellers’ diarrhoea
(141). Yet, despite such widespread use, few
controlled trials have been conducted to assess the
effectiveness of chemoprophylaxis in limiting
transmission and reducing diarrhoea morbidity and
mortality. The evidence is mainly from three
sources:

— studies of chemoprophylaxis among household
contacts of cholera cases;
—studies of the prolonged use of antimicrobials to
prevent infections in young children;
—studies of chemoprophylaxis of travellers’
diarrhoea.
These three types of studies are considered in turn.
Chemoprophylaxis among household contacts of
cholera cases. Four studies, one from Bangladesh and
three from India, have assessed the impact of
chemoprophylaxis on infection among household
contacts of a cholera index case (24, 62, 91, 134).
Different methods were used to analyse the data but
the study design was similar in all cases. Families of
hospitalized cholera cases were assigned to one of
various drug groups or to a control group. Household
contacts were followed bacteriologically over a period
of 10-15 days. Clinical information was reported in
only one study (62) where it is stated that all Vibrio
cholerae excreters were healthy. These studies
therefore assessed the effect of chemoprophylaxis on
the duration of excretion of V. cholerae among
infected contacts and on transmission of infection
within the household. They did not assess the
protective effect of chemoprophylaxis on diarrhoeal
illness. The drugs tested were effective in reducing the
prevalence of infection on successive days and the
proportion of infected household contacts (Table 5).
This effect, however, was of short duration and after
a period of 5-6 days the infection rates were similar in
the treatment and control groups. The only study (24)
that analysed the data by age groups reported that the
maximum impact of chemoprophylaxis was observed
in children under 5 years of age and in adults.
Only one study (72) has been located that tested the
value of chemoprophylaxis in the prevention of
diarrhoea among close contacts of cholera cases. In
this trial from Bangladesh, two doses of tetracycline

were administered to household contacts of hospi­
talized cholera cases. A similar control group was
visited but did not receive any medication. All
families were revisited after 10-12 days to check on
any new cases of diarrhoea and hospitalization. The
attack rate of diarrhoea among household contacts
was similar in the treatment (13%) and control groups
(14%). The occurrence of cases requiring hospitaliz­
ation was, however, significantly lower in the treat­
ment group (4%) than in the control group (8%). The
presence of V. cholerae in the stools of contacts was
not assessed.
Results from these studies suggest that chemo­
prophylaxis is effective in reducing the prevalence of
infection among household contacts of cholera cases.
This effect appears to be greatest shortly after
initiation of the course and wanes rapidly so that after
5 to 6 days the infection rates are similar in the
treatment and control groups. Maximum effective­
ness is observed, nonetheless, during the period of
greatest risk of infection. Presumably the observed
effect is due to a shortening of the duration of
excretion among contacts already infected and to a
reduction in transmission of infection to other
contacts. Khan (72), in contrast, found no impact of
chemoprophylaxis on the attack rates of diarrhoeal
illness among household contacts of cholera cases.
The observed reduction in hospitalization rates for
diarrhoea in the treatment group is ascribed to a
possible effect on the severity of the illness. No
placebo was given in Khan’s study and the increased
tendency to report to the hospital with diarrhoea may
have been associated with the absence of medication
in the control group. If the difference is real, a con­
comitant decrease in the attack rates of milder
diarrhoea would be expected. Only a study that
combines bacteriological surveillance with regular
clinical assessment can clarify this issue.

Prolonged use of antimicrobials in young children.
Low-dose antibiotic feeding is common practice in
livestock and poultry husbandry. The addition to feed
of small daily doses of broad-spectrum antibiotics has
been shown to stimulate growth and to prevent
infections, especially in weak animals reared in
insanitary conditions or fed deficient diets (88, 110).
The prolonged use of antimicrobials in children has
been studied in a number of circumstances, such as
during surveillance of recurrent attacks of rheumatic
fever (93) or in the management of dietary
deficiencies (65). These studies yield conflicting
results, but suggest that the prolonged use of anti­
microbials may occasionally be associated with
improved growth and decreased morbidity and
mortality, especially in malnourished children or in
children suffering from a chronic disease (122).

No. of children
receiving the drug

Duration of
study

Country

Study population

Age group

India

Hospitalized
malnourished
children

6 months
to 7 years

10

Chlortetracycline
or oxytetra­
cycline

2 months

Faster recovery, reduced
incidence of diarrhoea,
reduced mortality.

87

Kenya

Hospitalized
malnourished
children

2 years
(average)

38

Chlortetracycline

2-7 weeks

Faster weight gain, reduced
incidence of intercurrent
infection, reduced frequency
of weight faltering or
weight loss during infection,
reduced duration of diarrhoea

89

Honduras

Village children

6 months
to 6 years

54

lodoxychloroquinoline or
metronidazole

Four 16-week
periods

Reduced incidence of diarrhoea
in children
2 years old.

155

USA

Children in an
Apache community

1-42
months

81

Colistin sulfate

13 weeks

Children < 7 months old,
increased prevalence of
diarrhoea; 7-30 months old,
reduced prevalence;
31-42 months old, no effect.

60

Drug

Results

Reference

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 6. Effect of long-term chemoprophylaxis on diarrhoea morbidity in young children

304

I. DEZOYSA& R. G. FEACHEM

Table 7. Controlled trials of the chemoprophylaxis of travellers' diarrhoea
Diarrhoea
attack rates
(%)

Percentage
reduction

Study
population

Duration
of trial

Drug
groups

Egypt and
Far East

Danish tourists

25 days

Mecillinam

13

75

Placebo

53

-

Honduras

Peace Corps
volunteers

3 weeks

Doxycycline

33

NS"

Placebo

45

-

Peace Corps
volunteers

3 weeks

Doxycycline

32

68

Placebo

100

-

Peace Corps
volunteers

3 weeks

US students

2 weeks

Country

Honduras

Kenya

Mexico

Mexico

Mexico

Mexico

Mexico

Mexico

US students

US students

US travellers

US students

US students

2 weeks

3 weeks

Reference

Doxycycline

6

86

Placebo

43

-

Clioquinol

39

NS"

Neomycin (with kaolin
and pectin)

20

39

Placebo

34

-

Neomycin

16

NS"

Phthalylsulfathiazole

12

50

Placebo

24

-

Bismuth subsalicylate

23

62

Placebo

61

-

For duration of
voyage
(4-13 days)

Erythromycin

0

100

Placebo

29

-

3 weeks

Trimethoprimsulfamethoxazole

16

71

Placebo

55

-

Trimethoprimsulfamethoxazole

2

94

Trimethoprim

14

58
-

2 weeks

Placebo

33

US Navy
personnel

For duration of
exposure
(0.5-2.5 days)

Doxycycline

4

81

Placebo

21

-

Peace Corps
volunteers

3 weeks

Doxycycline

Sri Lanka6
and Kenya

Swiss tourists

2 weeks (duration
of voyage:
2-4 weeks)

Streptotriadc

16

58

Placebo

38

-

Various
destinations6

British airline
personnel and
their families

3 weeks (duration
of voyage:2 days
to over 6 weeks)

Streptotriadc

12

25

Neomycin-sulfonamidesd

19

NS"

Placebo

16

-

Mexico

Morocco

Placebo

" No significant difference in diarrhoea attack rates compared with the control group.
b Poorly controlled and supervised trial.

f Streptomycin + sulfadimidine + sulfadiazine + sulfathiazole.
d Neomycin + sulfadimidine + sulfadiazine + sulfathiazole.

8

83

46

-

15

132
131

129

67

68

30

2

31

32

45

130

141

152, 153

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Four controlled trials have been located that
document the effect of the prolonged administration
of antimicrobials on diarrhoea morbidity in young
children. They are summarized in Table 6. The first
two studies, conducted in India and Kenya, report the
effect of chemoprophylaxis on the recovery of
children hospitalized for malnutrition (87, 89). Both
document faster recovery and reduced incidence of
diarrhoea in children receiving antibiotics. The study
from Kenya also noted a decreased duration of
diarrhoea. Two further studies, from Honduras (155)
and the USA (60), investigated the protective effect of
prolonged administration of antimicrobials on
diarrhoea morbidity among children in the com­
munity. The results suggest that the prolonged
administration of antimicrobials to young children in
endemic areas may have an impact on diarrhoea
incidence or duration in certain age groups. Detailed
etiological investigations were not conducted. Also,
the children were randomized within age groups to the
different treatment cells and we cannot therefore
assess the effect of chemoprophylaxis on the
transmission of diarrhoea within households.
Chemoprophylaxis of travellers’ diarrhoea.
Travellers’ diarrhoea may affect from 10% to 60%
(30, 125) of travellers from low-risk to high-risk areas
during their first few weeks of travel. The condition is
primarily infectious, with ETEC as the most common
pathogen, isolated from 30% to 70% of cases (30, 51,
56, 97, 125, 128, 131, 135). Fourteen controlled trials
that investigated the effectiveness of chemopro­
phylaxis of travellers’ diarrhoea are summarized in
Table 7. Early studies with poorly absorbed anti­
microbials, such as neomycin and phthalylsulfathiazole (67, 68, 152, 153), showed a modest effect.
With improved understanding of the etiology of
travellers’ diarrhoea, doxycycline was tested in a
number of trials. Sack et al. (129, 130) showed that a
single daily dose of doxycycline was highly effective in
preventing diarrhoea in Peace Corps volunteers
travelling to Kenya and Morocco. Protection from
ETEC and other diarrhoeas was observed during the
treatment period and for the first week after cessation
of doxycycline. In the Moroccan study, however, the
volunteers were followed over a longer period of time
and a significant increase in the attack rate was later
observed in the treatment group compared with the
control group. In a study in Honduras (132),
biweekly prophylaxis with doxycycline was only
marginally effective, an outcome that may be due to
the decreased dosage or to the increased prevalence of
resistant Escherichia coli found in Honduras. In a
subsequent study in Honduras (131), a daily dose of
doxycycline was found to be effective in reducing
diarrhoea attack rates, although it did not prevent
diarrhoea caused by doxycycline-resistant ETEC.

305

Doxycycline also significantly reduced the severity of
the illness in those who had diarrhoea. Freeman et al.
(45) demonstrated the effectiveness of daily
doxycycline in US Navy personnel making a short
port call to Mexico. No rebound disease in acute
diarrhoea was noted in the treatment group after
departure from the high-risk area. Other antibiotics
found to be effective in controlled trials are erythro­
mycin (2), mecillinam (15), trimethoprim­
sulfamethoxazole, and trimethoprim alone (31, 32).
In the latter two studies among US students who
remained in Mexico on completion of the trial,
rebound diarrhoea was noted in the first week follow­
ing cessation of the treatment. Finally, a non-specific
drug, bismuth subsalicylate, has been found to have
some protective effect when ingested in doses too
large to be taken routinely (30).
In all of these studies, the drug was administered
only for the first few weeks of residence in the highrisk area, and both effectiveness and the development
of resistance during longer periods of chemoprophy­
laxis were not evaluated. Most studies were conducted
in small groups of similar individuals, such as US
students and Peace Corps volunteers, visiting a single
high-risk area. These individuals were often in close
contact with each other, sometimes eating together or
sharing the same accommodation. Transmission of
diarrhoea within the study groups may have been
facilitated under such conditions. There is some
evidence of this in one study (68). On the other hand,
it is possible that the potential for transmission was
reduced within the study group owing to the adminis­
tration of drugs to some members of the group. No
well controlled trials of chemoprophylaxis in large
groups of travellers to multiple destinations have yet
been carried out.
Conclusions on hypothesis 2. Numerous studies
have been conducted but there is little evidence that
chemoprophylaxis can reduce diarrhoea morbidity,
except perhaps in travellers. No studies have been
located that consider the effect on diarrhoea
mortality. Chemoprophylaxis among household con­
tacts of cholera cases has been shown to reduce the
prevalence of infection during the period of most
active transmission, but no impact on diarrhoea
attack rates has been documented. There is some
indication, however, that the severity of diarrhoea
may be reduced in the treated household contacts.
Broad-spectrum
antibiotics
administered
to
malnourished children have been shown to prevent
nosocomial diarrhoea, but the impact of prolonged
chemoprophylaxis among young children in the
community remains unclear. Studies on travellers’
diarrhoea indicate that chemoprophylaxis may be
highly effective in reducing the incidence of diarrhoea
among adult travellers making short visits to high-risk

306

I. DE ZOYSA & R. G. FEACHEM

Table 8. Reduction in cholera incidence following chemoprophylaxis of household contacts of known cholera
cases

Additional
case rate"
(%)

Proportion of all cholera
cases that occur as
detectable index and
additional cases6
(%)

6

5

30

20

1.2

6

5

30

40

2.4

6

5

30

60

3.6

6

5

50

20

2.0

6

5

50

40

4.0

6

5

50

60

6.0
2.6

Household
size
(persons)

Reduction in additional
case rate due to
chemoprophylaxis
(%)

Reduction in cholera
incidence rate due to
chemoprophylaxis
(%)

6

15

30

20

6

15

30

40

5.1

6

15

30

60

7.7

4.3

6

15

50

20

6

15

50

40

8.6

6

15

50

60

12.8

6

25

30

20

3.3

6

25

30

40

6.7

6

25

30

60

10.0

6

25

50

20

5.6

6

25

50

40

11.1
16.7

6

25

50

60

10

5

30

20

1.9

10

5

30

40

3.7

10

5

30

60

5.6

10

5

50

20

3.1

10

5

50

40

6.2

10

5

50

60

9.3

10

15

30

20

3.4

10

15

30

40

6.9

10

15

30

60

10.3

10

15

50

20

5.7

10

15

50

40

11.4

10

15

50

60

17.2

10

25

30

20

4.2

10

25

30

40

8.3

10

25

30

60

12.5

10

25

50

20

6.9

10

25

50

40

13.8

10

25

50

60

20.8

a See Table 1.
b See Table 3.

307

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Table 9. Reduction in shigellosis incidence following chemoprophylaxis of household contacts of known shigellosis
cases

Additional
case rateJ
(%)

Proportion of all shigellosis
cases that occur as
detectable index and
additional cases''
(%»

6

20

6

20

6

Reduction in additional
case rate due to
chemoprophylaxis
(%)

Reduction in shigellosis
incidence rate due to
chemoprophylaxis
(%)

5

20

0.5

5

40

1.0

20

5

60

1.5

6

20

10

20

1.0

6

20

10

40

2.0

6

20

10

60

3.0

6

30

5

20

0.6

6

30

5

40

1 2

6

30

5

60

1.8

6

30

10

20

1.2

6

30

10

40

2.4

Household
size
(persons)

6

30

10

60

3.6

6

40

5

20

0.7

6

40

5

40

1.3

6

40

5

60

2.0

6

40

10

20

1.3

6

40

10

40

2.7

6

40

10

60

40

10

20

5

20

0.6

10

20

5

40

1.3

10

20

5

60

1.9

10

20

10

20

1.3

10

20

10

40

2.6

10

20

10

60

3.9

10

30

5

20

0.7

10

30

5

40

1.5

10

30

5

60

2.2

10

30

10

20

1.5

10

30

10

40

2.9

10

30

10

60

4.4

10

40

5

20

0.8

10

40

5

40

1.6

10

40

5

60

2.3

10

40

10

20

1.6

10

40

10

40

3.1

10

40

10

60

4.7

" See Table 1.
h See Table 4.

30S

I. DE ZOYSA & R. G. FEACHEM

areas if the appropriate drug is used in the correct
dosage, and if the causative enteric pathogens
(particularly ETEC) are susceptible. The severity of
illness may also be reduced. Of concern, however, is
the rebound increase in diarrhoea incidence that has
been documented when travellers remain in the highrisk area after discontinuation of the drug.
Hypothesis 3. Chemoprophylaxis in young children
exposed to a recognized risk, such as contact with a
known case, can reduce overall diarrhoea
morbidity rates or mortality rates or severity in
young children.

The test of this hypothesis would come from a
study in which chemoprophylaxis was given to young
children exposed to one of the high-risk situations
described in hypothesis 1, and where the impact on
overall diarrhoea rates among young children was
monitored. No study of this kind has been located.
The two studies described in hypothesis 2, which
considered the effect on diarrhoea morbidity of the
prolonged administration of drugs to children in the
community, cannot be used to test hypothesis 3
because the children in the treatment groups were not
selected on the basis of their exposure to high-risk
situations.
Hypothesis 3 must be examined, therefore, by
theoretical calculations of the reductions in diarrhoea
rates that may result from chemoprophylaxis of
young children exposed to a recognized risk. Let us
consider the most common of the identified high-risk
situations: contact in the household with a known
case of cholera or shigellosis, and calculate the
potential impact of chemoprophylaxis on the
incidence of specific diarrhoeas.6
The calculations for cholera are set out in Table 8.
Two household sizes (6 and 10) and three additional
case rates (5%, 15% and 25%) are used as before. For
the purposes of this discussion, optimistic assump­
tions are made on the reduction in additional case
rates due to chemoprophylaxis and three values
(20%, 40% and 60%) are adopted on the basis of the
data summarized in Table 5. The expected reduction
in cholera incidence rates due to chemoprophylaxis,
computed on the basis of these assumptions, ranges
from 1.2% to 16.7% for a household size of 6 and
from 1.9% to 20.8% for a household size of 10.
Similar calculations for shigellosis are presented in
Table 9. The expected reduction in shigellosis
incidence rates due to chemoprophylaxis ranges from
0.5% to 4.0% for a household size of 6 and 0.6% to
4.7% for a household size of 10.
The impact of chemoprophylaxis on overall
b The formula used to make these calculations is available on
request from R.G.F.

diarrhoea incidence rates depends on the prominence
of cholera and shigellosis as a cause of diarrhoea and
this varies greatly from country to country. For
cholera, we may take the extreme example of
Bangladesh, where cholera is endemic and accounts
for approximately 0.4% of all diarrhoea cases in
children under 5 years (12, 14, 92, 102, 106). If we
take values of 5-15% from Table 8 as estimates of the
expected reduction in cholera incidence rates due to
chemoprophylaxis of household contacts of known
cholera cases, the intervention might reduce overall
diarrhoea incidence rates in children under 5 years by
0.02-0.06%. In developing countries shigellosis
accounts for approximately 10% of all diarrhoea
cases in children under 5 years (12, 14, 57, 64, 102).
Chemoprophylaxis of household contacts of known
shigellosis cases might reduce shigellosis incidence
rates by an estimated 1.5-3.5% (Table 9), and there­
fore might reduce overall diarrhoea incidence rates in
children under 5 years by 0.15-0.35%.
In the absence of other information, it may be
assumed that the reductions in cholera and shigellosis
mortality rates caused by chemoprophylaxis of
household contacts of known cases are the same as the
reductions in incidence rates shown in Tables 8 and 9.
We have calculated elsewhere (27) that, in
Bangladesh, cholera may account for 8% of
diarrhoea deaths in children under 5 years of age. If
chemoprophylaxis of household contacts of known
cholera cases reduces the cholera mortality rates in
children under 5 years by 5-15%, then this inter­
vention might, in Bangladesh, reduce the overall
diarrhoea mortality rate in the same age group by
0.4-1.2%. Where shigellosis is responsible for 20% of
diarrhoea deaths in children under 5 years (a
speculative but reasonable assumption), if chemo­
prophylaxis of household contacts of known
shigellosis cases reduces the shigellosis mortality rate
in children under 5 years by 1.5-3.5%, then this
intervention might reduce the overall diarrhoea
mortality rate in the same age group by 0.3-0.7%.

FEASIBILITY

The likely effectiveness of chemoprophylaxis in the
control of diarrhoeal diseases should not be
considered in isolation from the unwanted effects of
the drugs used and obstacles to the widespread
implementation of the intervention.
The use of prophylactic drugs. No prophylactic
drug has yet been identified which is universally safe
and effective. A number of drugs, such as neomycin
(63, 69) and the halogenated hydroxyquinolines (109,
114, 133), previously in common use, are now out of

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

favour because of their potential adverse reactions.
Antimicrobials are popular, but they have selective
action against certain bacteria and protozoa only. A
number of other agents, such as phage preparations
(82, 107, 112), lactobacilli preparations (21, 26), and
bismuth subsalicylate (30, 53) have been considered
for the chemoprophylaxis of diarrhoea but are
probably of limited value.
Antimicrobials are all associated with adverse
reactions, some of them severe, and their use may be
contraindicated in certain persons; for example, the
administration of tetracyclines to children and to
pregnant or lactating women is discouraged because
of their dental staining effect (156). Antimicrobials
may also increase host susceptibility to some enteric
pathogens (16, 96), and alter the activity of pancreatic
enzymes (17) and the metabolism of bile acids (58).
Of greatest concern, however, is that the use of anti­
microbials provides an advantage to resistant strains
and imposes selective pressure to assist in their spread
(36, 85, 101, IOS, 142, 162). There is also the sinister
possibility that the emergence of resistance is
associated with increased virulence or communi­
cability of certain strains (33, 117, 137, 143, 165).
Finally, antimicrobials may mask other bacterial
infections and complicate their therapy. Their
extensive use cannot be recommended in the absence
of monitoring for adverse reactions.

Implementation. The application of chemoprophy­
laxis among family contacts of known cholera and
shigellosis cases requires the correct identification of
the index cases followed by the administration of
drugs to their family contacts. Under normal
circumstances, only hospitalized diarrhoea cases are
likely to be investigated and correctly diagnosed. As
discussed earlier, hospitalization rates vary widely
from area to area and will often be lower than those
adopted in our calculations. Outbreaks of diarrhoeal
disease that need intensified control measures should
be detected as soon as possible, but reporting systems
based on routine data collected from health facilities
are not very sensitive (54) and the laboratory services
needed to make an accurate diagnosis of etiology and
to test for antimicrobial sensitivity may not always be
available (47, 94). Once the index case of cholera or
shigellosis is recognized, the at-risk family members
must be located for the distribution of drugs. Time is
at a premium here, as the additional case rate among
household contacts is highest on the first day of
observation in the studies reported in Table 1 and falls
rapidly thereafter. The distribution of drugs may be a
complex task, especially in rural areas and when a
multiple-dose drug regimen is used.
In short, the detection of cases and the rapid
follow-up of their contacts and distribution of drugs
require skills and resources that are scarce in

309

developing countries. The workload may be
especially heavy in epidemic situations when the
number of contacts to be reached will be very high.
Chemoprophylaxis might then direct attention from
other, more effective, control measures.

COST

The costs of chemoprophylaxis have not been
documented. The costs involved fall into four main
categories:
— The cost of the drugs. Drug bills place a con­
siderable burden on the health budget of most govern­
ments. There are large cost variations between drugs
and even for the same drug, depending on the price
per unit for that particular drug, the quantity
required, the preparation chosen, and the purchasing
and distribution systems used.
— The cost of surveillance (including laboratory
surveillance). This cost is difficult to determine as few
reporting systems for health expenditure list these
items separately. This cost would be shared with other
interventions.
— The cost of the distribution of the drugs,
including manpower and logistics.
—The cost of unwanted effects. These include the
cost of treatment of adverse reactions due to the drugs
used in chemoprophylaxis and the increased cost of
new drugs used in therapy to replace old ones that
become inadequate because of the emergence of anti­
microbial resistance.
CONCLUSIONS

There is little evidence that chemoprophylaxis is
effective in reducing diarrhoea morbidity and
mortality. The most common application of chemo­
prophylaxis in developing countries is to prevent
cholera or shigellosis among selected high-risk
groups, such as household contacts of known cases. A
theoretical case has been made out, based on
optimistic assumptions, that chemoprophylaxis of
household contacts of known cholera cases in
Bangladesh might reduce overall diarrhoea incidence
rates in children under 5 years of age by 0.02-0.06%
and diarrhoea mortality rates by 0.4-1.2%. Chemo­
prophylaxis of household contacts of known
shigellosis cases might reduce diarrhoea incidence
rates by 0.15-0.35% and diarrhoea mortality rates by
0.3-0.7% in the same age group.
The correct identification of index cases of cholera
and shigellosis, followed by the rapid distribution of
drugs to their household contacts, requires skills and
resources that are scarce in the developing countries.

310

1. DE ZOYSA & R. G. FEACHEM

All the drugs currently used have side-effects that
should be carefully monitored. Chemoprophylaxis
can contribute to the widespread emergence and
dissemination of antimicrobial resistance. The costs
of chemoprophylaxis have not been documented but
are likely to be high and no long-term benefits may be

derived. The available evidence suggests, therefore,
that chemoprophylaxis is not feasible in many settings
and that, even if successfully implemented, it is not a
cost-effective intervention for national diarrhoeal
disease control programmes.

ACKNOWLEDGEMENTS
We are grateful for the constructive criticisms on earlier drafts of this paper provided by D. Barua, P.F. Beales,
D. Blum, B. Cvjetanovic, H. G. Dam, R. C. Hogan, M. H. Mcrson, H. Mosley, S. C. Pal and Y. Watanabe. We thank J.
Seaman of the Save the Children Fund, United Kingdom, for his support and encouragement. Secretarial, bibliographical
and editorial assistance was most ably provided by Alison Hinchley and Susanne O’Driscoll.

RESUME
INTERVENTIONS DESTINEES A LA LUTTE CONTRE LES MALADIES DIARRHEIQUES CHEZ LES JEUNES ENFANTS:

CHIMIOPROPHYLAXIE

Cet article est le sixieme d’une serie passant en revue les
interventions possibles en vue de reduire la morbidite et la
mortalite par diarrhee parmi les enfants de moins de 5 ans
dans les pays en developpement. Un certain nombre de
situations font courir aux jeunes enfants un risque accru de
diarrhee. Parmi celles-ci la mieux etudiee dans les pays en
developpement est le contact avec un cas connu dans la
famille ou le menage. L’application la plus frequente de la
chimioprophylaxie dans ces pays vise a pr6venir le cholera
ou les shigelloses parmi les contacts de cas connus dans les
menages. Il ne semble pas que la chimioprophylaxie puisse
reduire efficacement la morbidite et la mortalite
*
par
diarrhee, sauf peut-etre chez les voyageurs. D’apres les
calculs theoriques du present article (fondes sur des
hypotheses optimistes), la chimioprophylaxie appliquee &
des contacts dans les menages comptant des cas connus, au
Bangladesh, pourrait reduire de 0,02-0,06% les taux
d’incidence globaux de la diarrhee et de 0,4-1,2% les taux de
mortalite par diarrhee chez les enfants de moins de 5 ans. En
ce qui concerne les shigelloses, la chimioprophylaxie
administree a des contacts de cas connus dans des menages

pourrait reduire de 0,15-0,35% les taux d’incidence globaux
de la diarrhee et de 0,3-0,7% les taux de mortalite par
diarrhee dans le meme groupe d’age.
Le depistage correct des cas initiaux de cholera et de
shigellose suivi d’une distribution rapide de medicaments A
leurs contacts dans les menages demande des competences et
des ressources qui sont rares dans les pays en developp­
ement. Tous les medicaments actuellement en usage
presentent des effets secondaires qui doivent etre minutieusement surveilles. En outre, point capital, la chimio­
prophylaxie peut contribuer a 1’apparition et d la dis­
semination de la resistance aux antimicrobiens sur une
grande echelle. On ne connait pas les couts de la
chimioprophylaxie, mais ils sont probablement elev6s et 1’on
ne peut en attendre d’avantages A long terme. D’apres les
donnees existantes, la chimioprophylaxie est done irrealisable dans de nombreuses circonstances et, meme si elle est
appliquee avec succes, ce n’est pas une intervention rentable
pour les programmes nationaux de lutte contre les maladies
diarrheiques.

REFERENCES
1.

Adler, J. L. & Zickl, R. Winter vomiting disease.
Journal of infectious diseases, 119: 668-673 (1969).
2. Andremont, A. & Tancrede, C. Reduction of the
aerobic Gram-negative bacterial flora of the gastro­
intestinal tract and prevention of traveller’s diarrhea
using oral erythromycin. Annales de microbiologie,
132 B: 419-427 (1981).
3. Ashworth, A. & Feachem, R. G. Interventions for
the control of diarrhoeal diseases among young
children: prevention of low birthweight. Bulletin of the
World Health Organization, 63 (1): 165-186 (1985).

Azurin, J. C. et al. A controlled field trial of the
effectiveness of cholera and cholera El Tor vaccines in
the Philippines. Bulletin of the World Health
Organization, 37: 703-727 (1967).
5. Baron, R. C. et al. Norwalk gastrointestinal illness.
An outbreak associated with swimming in a
recreational lake and secondary person-to-person
transmission. American journal of epidemiology, 115:
163-172 (1982).

4.

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Bart, K. J. et al. Seroepidemiologic studies during a
simultaneous epidemic of infection with El Tor Ogawa
and classical Inaba Vibrio cholerae. Journal of
infectious diseases, 121: S17-S24 (1970).
7. Belinson, L. & Bellack, S. Use of combined
furazolidone and
tetracycline in controlling
institutional shigellosis. Illinois medical journal, 135:
701-704 (1969).
8. Benenson,
A. S.
et
al.
Person-to-person
transmission of cholera. In: Proceedings of the
Cholera Research Symposium, Honolulu, US
Department of Health, Education and Welfare, 1965,
pp. 332-336.
9. Black, R. E. et al. Giardiasis in day-care centers:
evidence of person-to-person transmission. Pediatrics,
60: 486-491 (1977).
10. Black, R. E. et al. Enterotoxigenic Escherichia coli
diarrhoea: acquired immunity and transmission in an
endemic area. Bulletin of the World Health
Organization, 59: 263-268 (1981).
11. Black. R. E. et al. Handwashing to prevent diarrhea
in day-care centers. American journal of
epidemiology, 113: 445-451 (1981).
12. Black, R E. et al. Incidence and severity of
rotaviruses and Escherichia coli diarrhoea in rural
Bangladesh: implications for vaccine development.
Lancet, 1: 141-143 (1981).
13. Black, R. E. etal. Longitudinal studies of infectious
diseases and physical growth of children in rural
Bangladesh. 1. Patterns of morbidity. American
journal of epidemiology, 115: 305-314 (1982).
14. Black, R. E. etal. Longitudinal studies of infectious
diseases and physical growth of children in rural
Bangladesh. 2. Incidence of diarrhea and association
with known pathogens. American journal of
epidemiology, 115: 315-324 (1982).
15. Black, F. T. et al. Mecillinam, a new prophylactic
for travellers’ diarrhoea. Scandinavian journal of
infectious diseases, 15: 189-193 (1983).
16. Bohnhoff, M. & Miller, C. P. Enhanced
susceptibility to Salmonella infection in streptomycintreated mice. Journal of infectious diseases, 111:
117-127 (1962).
17. BorgstrOm, A. et al. Elevated fecal levels of
endogenous pancreatic endopeptidases after antibiotic
treatment. Scandinavian journal of gastroenterology,
12: 525-529 (1977).
18. Boyce, J. M. et al. Patterns of Shigella infection in
families in rural Bangladesh. American journal of
tropical medicine and hygiene, 31: 1015-1020 (1982).
19. Brown, E. H. Giardia lamblia: the incidence and
results of infestation of children in residential
nurseries. Archives of disease in childhood, 23:
119-128 (1948).
20. Center For Disease Control. Shigella surveillance.
Report no. 39: 4(1977).
21. Clements, M. L. etal. Lactobacillus prophylaxis for
diarrhea due to enterotoxigenic Escherichia coli.
Antimicrobial agents and chemotherapy, 20: 104-108
.
(1981)
22. Cohen, J. et al. Epidemiological aspects of cholera El
Tor outbreak in a non-endemic area. Lancet, 2: 86-89
.
(1971)
6.

311

Davies, J. B. M. Symptomless carriers in home
contacts in Sonne dysentery. British medical journal,
2: 191-192 (1952).
24. Deb, B. C. et al. Effect of sulfadoxine on
transmission of Vibrio cholerae infection among
family contacts of cholera patients in Calcutta.
Bulletin of the World Health Organization, 54:
171-175 (1976).
25. Dingle, J. H. et al. Illness in the home. A study of
25,000 illnesses in a group of Cleveland families.
Cleveland, Press of Western Reserve University, 1964,
pp. 1-32, 188-218.
26. De Dios Pozo-Olano, J. et al. Effect of lactobacilli
preparation on traveler’s diarrhea. A randomized,
double blind clinical trial. Gastroenterology, 74:
829-830(1978).
27. De Zoysa, I. & Feachem, R. G. Interventions for the
control of diarrhoeal diseases among young children:
rotavirus and cholera immunization. Bulletin of the
World Health Organization, (in press).
28. Doyle, A. Incidence of illness in early group and
family day-care. Pediatrics, 58: 607-613 (1976).
29. DuPont, H. L. et al. Shigellosis in custodial
institutions. American journal of epidemiology, 92:
172-179 (1970).
30. DuPont, H. L. et al. Prevention of travelers’
diarrhea
(emporiatric
enteritis).
Prophylactic
administration of subsalicylate bismuth. Journal of
the American Medical Association, 243: 237-241
(1980).
31. DuPont, H. L. et al. Prevention of travelers’
diarrhea
with
trimethoprim-sulfamethoxazole.
Reviews of infectious diseases, 4: 533-539 (1982).
32. DuPont, H. L. et al. Prevention of travelers’
diarrhea with trimethoprim-sulfamethoxazole and
trimethoprim alone. Gastroenterology, 84: 75-80
(1983).
33. Echeverria, P. et al. Antimicrobial resistance and
enterotoxin production among isolates of Escherichia
coli in the Far East. Lancet, 2: 589-592 (1978).
34. Echeverria, P. et al. Identification by DNA
hybridisation of enterotoxigenic Escherichia coli in
homes of children with diarrhoea. Lancet, 1: 63-65
(1984).
35. Ekanem, E. E. et al. Transmission dynamics of
enteric bacteria in day-care centers. American journal
of epidemiology, 118: 562-572 (1983).
36. Farrar, W. E., Jr. Antibiotic resistance in intestinal
bacteria. Clinics in gastroenterology, 8: 803-826
(1979).
37. Feachem, R. G. Environmental aspects of cholera
epidemiology. III. Transmission and control. Tropical
diseases bulletin, 79: 1-47 (1982).
38. Feachem, R. G. Interventions for the control of
diarrhoeal
diseases
among
young
children:
supplementary feeding programmes. Bulletin of the
World Health Organization, 61: 967-979 (1983).
39. Feachem, R. G. Interventions for the control of
diarrhoeal diseases among young children: promotion
of personal and domestic hygiene. Bulletin of the
World Health Organization, 62: 467-476 (1984).

23.

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

313

Khan, M. U. & Mosley, W. H. The significance of 91. McCormack, W. M. et al. Tetracycline prophylaxis
in families of cholera patients. Bulletin of the World
Shigella as a cause of diarrhea in a low economic urban
Health Organization, 38: 787-792 (1968).
community in Dacca. East Pakistan medical journal,
92. McCormack, W. M. etal. Endemic cholera in rural
12: 45-51 (1968).
East Pakistan. American journal of epidemiology, 89:
74. Khan, M. & Shahidullah, M. Cholera due to the El
393-404 (1969).
Tor biotype equals the classical biotype in severity and
attack rates. Journal of tropical medicine and hygiene,
93. McVay, L. V., Jr. & Sprunt, D. H. Aureomycin in
83: 35-39(1980).
the prophylaxis of rheumatic fever. New England
75. Khan, M. & Shahidullah, M. Contrasting
journal of medicine, 249: 387-393 (1953).
epidemiology of Shigella dysenteriae and Shigella
94. Malengreau, M. et al. Outbreak of Shigella
dysentery in Eastern Zaire, 1980-1982. Annales de la
flexneri, Dacca. Transactions of the Royal Society of
Tropical Medicine and Hygiene, 74: 528-533 (1980).
SociNe Beige de Mddecine Tropicale, 63: 59-67
76. Khan, M. et al. Epidemiology of Shigella
.
(1983)
dysenteriae, type 1 infections, in Dacca urban area.
95. Martin, A. R. et al. Epidemiologic analysis of
Tropical and geographical medicine, 31: 213-223
endemic cholera in urban East Pakistan, 1964-1966.
(1979).
American journal of epidemiology, 89: 572-582
77. Khan, M. U. et al. Changes in the trend of shigellosis
(1969).
in Dhaka: family study on secondary infection, clinical
96. Mentzing, L. O. & Ringertz, O. Salmonella
manifestation
and sensitivity pattern:
1980.
infection in tourists. 2. Prophylaxis against
Transactions of the Royal Society of Tropical
salmonellosis. Acta pathologica et microbiologica
Medicine and Hygiene, 78: 151-156 (1984).
Scandinavica, 74: 405-413 (1968).
78. Kim, H. W. et al. Human reovirus-like agent 97. Merson, M. H. et al. Travelers’ diarrhea in Mexico.
infection. Occurrence in adult contacts of pediatric
A prospective study of physicians and family members
patients with gastroenteritis. Journal of the American
attending a congress. New England journal of
Medical Association, 238: 404-407 (1977).
medicine, 294: 1299-1305 (1976).
79. Kim, K. et al. Outbreaks of diarrhea associated with 98. Mhalu, F. S. The problem of resistance to anti­
Clostridium difficile and its toxin in day-care centers:
microbial agents in the treatment and prevention of
evidence of person-to-person spread. Journal of
cholera. In: T. Holme, et al. ed., Acute enteric
pediatrics, 102: 376-382 (1983).
infections in children. Amsterdam, Elsevier/North80. Konno, T. et al. Astrovirus-associated epidemic
Holland Biomedical Press, 1981, pp. 123-126.
gastroenteritis in Japan. Journal of medical virology,
99. Mhalu, F. S. et al. Hospital outbreaks of cholera
9: 11-17 (1982).
transmitted through close person-to-person contact.
81. Kumarasinghe, G. etal. An outbreak of Salmonella
Lancet, 2: 82-84 (1984).
muenchen infection in a specialist paediatric hospital.
100. Middleton, P. J. et al. Viruses associated with acute
Journal of hospital infection, 3: 341-344 (1982).
gastroenteritis in young children. American journal of
82. Lancet Editorial. Phage therapy. Lancet, 2:
diseases of children, 131: 733-737 (1977).
1287-1288 (1983).
101. Moller, J. K. et al. Changing patterns of plasmid83. Lauwers, S. et al. Campylobacter enteritis in
mediated drug resistance during tetracycline therapy.
Brussels. Lancet, 1: 604-605 (1978).
Antimicrobial agents and chemotherapy, 11: 388-391
fc4. LaVeck, G. D. et al. Sulfadimethoxine in the
(1977).
F treatment and prevention of epidemic shigellosis.
102. Mosley, W. H. & Khan, M. Cholera epidemiology —
Antibiotic medicine and clinical therapy, 7: 119-124
some environmental aspects. Progress in water
(1960).
technology, 11: 309-316 (1979).
85. Levy, S. B. Microbial resistance to antibiotics. An 103. Mosley, W. H. et al. Epidemiologic and sociologic
evolving and persistent problem. Lancet, 2: 83-88
features of a large urban outbreak of shigellosis.
.
(1982)
Journal of the American Medical Association, 182:
86. Levy, S. B. et al. Epidemic shigellosis in Minnesota
1307-1311 (1962).
1973. Minnesota medicine, 58: 405-412 (1973).
104. Mosley, W. H. et al. Studies of cholera El Tor in the
87. Lewis, R. A. etal. Antibiotic dietary supplements in
Philippines. 4. Transmission of infection among
the therapy of childhood protein malnutrition.
neighbourhood and community contacts of cholera
American journal of tropical medicine and hygiene, 5:
patients. Bulletin of the World Health Organization,
483-496 (1956).
33: 651-660(1965).
88. Luckey, T. D. Antibiotics in nutrition. In: Goldberg, 105. Mosley, W. H. etal. The relationship of vibriocidal
H. S., ed. Antibiotics: their chemistry and non­
antibody titre to susceptibility to cholera in family
medical uses. Princeton, D. van Nostrand Company,
contacts of cholera patients. Bulletin of the World
1959, chapter 3, pp. 174-321.
Health Organization, 38: 777-785 (1968).
89. MacDougall, L. G. The effect of aureomycin on 106. Mosley, W. H. etal. An epidemiological assessment
undernourished African children. Journal of tropical
of cholera control programs in rural East Pakistan.
pediatrics, 3: 74-81 (1957).
International journal of epidemiology, 1: 5-11
90. Mackenzie, D. J. M. Cholera and its control.
.
(1972)
Proceedings of the cholera research symposium,
Washington, DC, 1965, pp. 341-346.

73.

314

I. DE ZOYSA & R. G. FEACHEM

Mulczyk. M. & Slopek, S. Use of a new phage 124. Rosenstein, B. J. Salmonellosis in infants and
preparation in prophylaxis and treatment of
children:
epidemiologic
and
therapeutic
shigellosis.
Acta
microbiologica
Academiae
considerations. Journal of pediatrics, 70: 1-7 (1967).
125. Rowe, B. et al. An investigation of travellers’
Scientiarum Hungaricae, 21: 115-119 (1974).
diarrhoea. Lancet, 1: 1-5 (1970).
10S. Murray, B. E. et al. Emergence of high-level
126. Ruckarby, G. A. et al. An outbreak of bacillary
trimethoprim resistance in fecal Escherichia coli
dysentery. Medical journal of Australia. 47: 81-85
during oral administration of trimethoprim or
(1960).
trimethoprim-sulfamethoxazole.
New
England
127.
Ryder, R. W. et al. Reovirus-like agent as a cause of
journal of medicine, 306: 130-134 (1982).
nosocomial diarrhea in infants. Journal of pediatrics,
109. Nakae, K. et al. Relation between subacute myelo90: 698-702 (1977).
optic neuropathy (S.M.O.N.) and clioquinol:
128. Sack, D. A. et al. Enterotoxigenic Escherichia coli
nationwide survey. Lancet, 1: 171-173 (1973).
diarrhea of travelers: a prospective study of American
110. National Academy of Sciences. The use of drugs
Peace Corps volunteers. Johns Hopkins medical
in animal feeds. Proceedings of a symposium.
journal,
141: 63-70 (1977).
Washington, DC, 1969 (National Academy of Sciences
129. Sack, D. A. et al. Prophylactic doxycycline for
Publ. No. 1679).
travelers’ diarrhea. Results of a prospective double­
111. Nelson, J. D. et al. Endemic shigellosis: a study of
blind study of Peace Corps volunteers in Kenya. New
fifty households. American journal of epidemiology,
86: 683-689 (1967).
England journal of medicine, 298: 758-763 (1978).
112. New Scientist Editorial. Viruses that can fight 130. Sack, R. B. et al. Prophylactic doxycycline for
travelers’ diarrhea. Results of a prospective double­
diarrhoea. New scientist, 100: 414-415 (1983).
blind study of Peace Corps volunteers in Morocco.
113. Noone, C. & Banatvala, J. E. Hospital acquired
Gastroenterology, 76: 1368-1373 (1979).
rotaviral gastroenteritis in a general paediatric unit.
131.
Sack, R. B. et al. Doxycycline prophylaxis of
Journal of hospital infection, 4: 297-299 (1983).
travelers’ diarrhea in Honduras, an area where
114. Oakley, G. P., Jr. The neurotoxicity of the
resistance to doxycycline is common among
halogenated hydroxyquinolines: a commentary.
enterotoxigenic Escherichia coli. American journal of
Journal of the American Medical Association, 225:
tropical medicine and hygiene, 33: 460-466 (1984).
395-397 (1973).
115. Olarte, J. et al. Resistance of Shigella dysenteriae 132. Santosham, M. et al. Biweekly prophylactic
doxycycline
for travelers’ diarrhea. Journal of
type 1 to ampicillin and other antimicrobial agents:
infectious diseases, 143: 598-602 (1981).
strains isolated during a dysentery outbreak in a
133. Schultz, M. G. Entero-vioform for preventing
hospital in Mexico city. Journal of infectious diseases,
travelers’ diarrhea. Journal of the American Medical
133: 572-575 (1976).
Association, 220: 273-274 (1972).
116. Oseasohn, R. et al. Clinical and bacteriological
134. Sen Gupta, P. G. et al. Effect of doxycycline on
findings among families of cholera patients. Lancet, 1:
transmission of Vibrio cholerae infection among
340-342 (1966).
family contacts of cholera patients in Calcutta.
117. Panhotra, B. R. & Agarwal, K. C. Plasmids
Bulletin
of the World Health Organization, 56:
carrying genes for enterotoxin production and drug
323-326 (1978).
resistance in Escherichia coli of human origin. Indian
135. Shore, E. G. et al. Enterotoxin-producing
journal of medical research, 74: 652-655 (1981).
Escherichia coli and diarrheal disease in adufl
118. Philippines Cholera Committee. Study on the
travelers: a prospective study. Journal of infectious
transmission of El Tor cholera during an outbreak in
diseases, 129: 577-582 (1974).
Can-Itom community in the Philippines. Bulletin of
136. Shu-Cheng, D. Shigellosis in children in China. In:
the World Health Organization, 43: 413-419 (1970).
Shigellosis: a continuing global problem. (Special
119. Pickering, L. K. et al. Diarrhea caused by Shigella,
publication no. 20), Dhaka, International Centre for
rotavirus, and Giardia in day-care centers: prospective
Diarrhoeal Disease Research, Bangladesh, 1983, pp.
study. Journal of pediatrics, 99: 51-56 (1981).
14-25.
120. Rahaman, M. M. et al. An outbreak of dysentery
137. Silva, M. L. M. et al. Plasmid coding for drug
caused by Shigella dysenteriae type 1 on a coral island
resistance and production of heat-labile and heat­
in the Bay of Bengal. Journal of infectious diseases,
stable toxins harbored by an Escherichia coli strain of
132: 15-19(1975).
human origin. Infection and immunity, 39: 970-973
121. Rodriguez, W. J. et al. Common exposure outbreak
(1983).
of gastroenteritis due to type 2 rotavirus with high
138. Sommer, A. & Mosley, W. H. Ineffectiveness of
secondary attack rate within families. Journal of
infectious diseases, 140: 353-357 (1979).
cholera vaccination as an epidemic control measure.
122. Rosenberg, I. H. et al. Infant and child entcritisLancet, 1: 1232-1235 (1973).
malabsorption-malnutrition: the potential of limited
139. Sommer, A. et al. Efficacy of vaccination of family
studies with low-dose antibiotic feeding. American
contacts of cholera cases. Lancet, 1: 1230-1232
.
(1973)
journal of clinical nutrition, 27: 304-309 (1974).
123. Rosenberg, M. L. et al. Shigellosis in the United 140. Spratt, H. C. et al. Nosocomial infantile
States: ten-year review of nationwide surveillance,
gastroenteritis associated with minirotavirus and
1964-1973. American journal of epidemiology, 104:
calicivirus. Journal of pediatrics, 93: 922-926 (1978).
543-551 (1976).
107.

CHEMOPROPHYLAXIS FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Steffen, R. & Gsell, O. Prophylaxis of traveller’s
diarrhoea. Journal of tropical medicine and hygiene,
84: 239-242 (1981).
142. Stenderup, J. et al. Changes in serotype and
resistance pattern of the intestinal Escherichia coli
flora during travel. Results from a trial of mecillinam
as a prophylactic against travellers’ diarrhoea.
Scandinavian journal of infectious diseases, 15:
367-373 (1983).
143. Stieglitz, H. et al. Linkage of heat-stable
enterotoxin activity and ampicillin resistance in a
plasmid isolated from an Escherichia coli strain of
human origin. Infection and immunity, 30: 617-620
(1980).
144. Storch, G. A. et al. Shigellosis in the Marshall
Islands: epidemiologic aspects of an outbreak.
American journal of tropical medicine and hygiene,
29: 456-463 (1980).
145. Sutmoller, F. et al. An oubreak of gastroenteritis
caused by both rotavirus and Shigella sonnei in a
private school in Rio de Janeiro. Journal of hygiene,
88: 285-293 (1982).
146. Tacket, C. O. & Cohen, M. L. Shigellosis in daycare
centers: use of plasmid analysis to assess control
measures. Pediatric infectious disease, 2: 127-130
(1983).
147. Tallet, S. et al. Clinical, laboratory, and
epidemiologic features of a viral gastroenteritis in
infants and children. Pediatrics, 60: 217-222 (1977).
148. Tamayo, J. F. et al. Studies of cholera El Tor in the
Philippines. 3. Transmission of infection among
household contacts of cholera patients. Bulletin of the
World Health Organization, 33: 645-649 (1965).
149. Thomas, M. E. M. & Mogford, H. E. Salmonellosis
in general practice. Observations of cases and their
households in Enfield. Journal of hygiene, 68: 663-671
(1970).
150. Thomas, M. E. M.& Tillett, H. E. Sonne dysentery
in day schools and nurseries: an eighteen-year study in
Edmonton. Journal of hygiene, 71: 593-602 (1973).
151.
Tufvesson, B. et al. Family infections by reo-like
virus. Comparison of antibody titres by complement
fixation and immunoelectroosmophoresis. Scandi­
navian journal of infectious diseases, 9: 257-261
(1977).
141.

315

Turner, A. C. Traveller’s diarrhoea: a survey of
symptoms, occurrence, and possible prophylaxis.
British medical journal, 4: 653-654 (1967).
153. Turner, A. C. Traveller’s diarrhoea. British medical
journal, 5: 118 (1968).
154. Van de Linde, P. A. M. & Forbes, G. I. Observations
on the spread of cholera in Hong Kong, 1961-63.
Bulletin of the World Health Organization, 32:
515-530(1965).
155. Villarejos, V. M. et al. Chemoprophylaxis of
diarrhea. American journal of tropical medicine and
hygiene, 20: 602-607 (1971).
156. Wallman, I. S. & Hilton, H. B. Teeth pigmented by
tetracycline. Lancet, 1: 827-829 (1962).
157. Weissman, J. B. et al. The role of preschool children
and day-care centers in the spread of shigellosis in
urban communities. Journal ofpediatrics, 84: 797-802
.
(1974)
158. Weissman, J. B. et al. Shigellosis in day-care centres.
Lancet, 1: 88-90 (1975).
159. Western, K. A. et al. Hospital infection
control — an international perspective. Infection
control, 3: 453-455 (1982).
160. Woodward, W. E. & Mosley, W. H. The spectrum
of cholera in rural Bangladesh. II. Comparison of El
Tor Ogawa and classical Inaba infection. American
journal of epidemiology, 96: 342-351 (1972).
161. Woodward, W. E. et al. The spectrum of cholera in
rural East Pakistan. I. Correlation of bacteriologic
and serologic results. Journal of infectious diseases,
121: S10-S16 (1970).
162. WHO Scientific Working Group. Antimicrobial
resistance. Bulletin of the World Health Organization,
61: 383-394 (1983).
163. Yannet, H. et al. The use of sulfaguanidine for
prophylaxis in Sonne bacillary dysentery, and in the
control of the carrier state. Yale journal of biology and
medicine, 16: 443-450 (1944).
164. Yen, C. H. A recent study of cholera with reference to
an outbreak in Taiwan in 1962. Bulletin of the World
Health Organization, 30: 811-825 (1964).
165. Yoh, M. et al. Effects of lincomycin and tetracycline
on production and properties of enterotoxins of
enterotoxigenic Escherichia coli. Infection and
immunity, 42: 778-782 (1983).
152.

CH 3.8
Bulletin ofthe H orla Health Organization, 63 (4): 757-772 (1985)

© World Health Organization 1985

Interventions for the control of diarrhoeal diseases
among young children: improving water supplies and
excreta disposal facilities
*
S. A. Esrey,1 R. G. Feachem,*12 & J. M. Hughes3
A theoretical model is proposed that relates the level of ingestion of diarrhoea-causing
pathogens to the frequency of diarrhoea in the community. The implications of this model
are that, in poor communities with inadequate water supply and excreta disposal, reducing
the level of enteric pathogen ingestion by a given amount will have a greater impact on
diarrhoea mortality rates than on morbidity rates, a greater impact on the incidence rale of
severe diarrhoea than on that of mild diarrhoea, and a greater impact on diarrhoea caused
by pathogens having high infectious doses than on diarrhoea caused by pathogens of a low
infectious dose. The impact of water supply and sanitation on diarrhoea, related infections,
nutritional status, and mortality is analysed by reviewing 67 studies from 28 countries. The
median reductions in diarrhoea morbidity rates are 22% from all studies and 27% from a
few better-designed studies. All studies of the impact on total mortality rates show a median
reduction of 21%, while the few better-designed studies give a median reduction of 30%.
Improvements in water quality have less of an impact than improvements in water
availability or excreta disposal.

Of the several interventions that may reduce
diarrhoea morbidity and mortality rates (38), the
improvement of water supply and excreta disposal
facilities has attracted particular interest. These
environmental
improvements,
together
with
improvements in living standards, played a major
role in reducing diarrhoea rates and controlling
epidemic typhoid and cholera in Europe and North
America between 1860 and 1920. It is anticipated that
the improvement of water supply and excreta
disposal in poor communities in developing countries
today will have a substantial impact on diarrhoea
morbidity and mortality rates in those communities.
This expectation provides part of the motivation for
the International Drinking Water Supply and Sani­
tation Decade (1981-1990), the aims of which are to
increase the rate at which new water supply and
excreta disposal facilities are constructed and to maxi­
mize the probability that they will be correctly opera­
• Requests for reprints should be sent to the Director, Diarrhoeal
Diseases Control Programme, World Health Organization, 1211
Geneva 27, Switzerland.
1 Division of Nutritional Sciences, Cornell University, Ithaca,
New York 14853, USA.
2 Head, Department of Tropical Hygiene, London School of
Hygiene and Tropical Medicine, Keppel Street, London WCIE7HT,
England.
3 Director, Hospital Infections Program, Center for Infectious
Diseases, Centers for Disease Control, Public Health Service, US
Department of Health and Human Services, Atlanta, GA 30333,
USA.
4581

ted, maintained and used.
The potential impacts of improved water supply
and excreta disposal on diarrhoea and other waterrelated diseases in developing countries have been
discussed and debated at length over the past decade.
White et al. (86) provided a conceptual framework
for the debate in the context of studies in East Africa;
McJunkin (56) reviewed the topic extensively;
Saunders & Warford (70) summarized the water
supply impact studies; Feachem et al. (39)
summarized the excreta disposal impact studies; and
Blum & Feachem (13) and Esrey & Habicht (28)
considered the methodological difficulties inherent in
attempts to measure the impact of water supply and
excreta disposal projects on diarrhoea.
In this review we analyse the effectiveness of water
supply and excreta disposal improvements for
reducing diarrhoea rates in young children in
developing countries. We also examine their impact
on diarrhoea-related infections, nutritional status,
and mortality. We make no attempt to analyse other
impacts of water supply and excreta disposal
improvements or to compute an overall cost-benefit
ratio for these investments. This paper is the ninth in a
series of reviews of potential anti-diarrhoea inter­
ventions (2, 25, 26, 32, 34-36, 38).“
• ASHWORTH, A. & FEACHEM. R. G. Interventions for the
control of diarrhoeal diseases among young children: improving
lactation. Unpublished WHO document CDD/85.2, 1985.

— 151 —

75S

S. A. ESREY ET AL.

EFFECTIVENESS

For improved water supply or excreta disposal
facilities to be an effective diarrhoea control
intervention, it must be true that:
either

water supply or excreta disposal
improvements can reduce the inges­
tion by young children of pathogens
causing diarrhoea

hypothesis
1

and

a reduction in the ingestion of these
pathogens by young children can
reduce diarrhoea morbidity or
mortality rates

hypothesis
2

or

water supply or excreta disposal
improvements can reduce diarrhoea
morbidity or mortality rates among
young children

hypothesis
3

The effectiveness of improved water supply and
excreta disposal would be suggested by a demonstra­
tion either of the correctness of hypotheses 1 and 2 or
the correctness of hypothesis 3. In some other reviews
in this series (for instance, 35, 36), most of the
literature bears on hypotheses 1 and 2, and hypothesis
3 must be handled by theoretical calculations. Here
the reverse is the case: there are few data on
hypotheses 1 and 2 and an extensive literature on
hypothesis 3. The evidence for and against the three
hypotheses is examined below.

Hypothesis 1. Water supply or excreta disposal
improvements can reduce the ingestion by young
children of pathogens causing diarrhoea.
There is some evidence to suggest that three types of
water and excreta disposal improvements (improved
water quality, increased water availability and
quantity associated with better hygiene practices, and
improved excreta disposal facilities) may reduce the
ingestion of pathogens causing diarrhoea (33).
All the major infectious agents of diarrhoea are
transmitted by the faecal-oral route, and all can be
transmitted via contaminated water. For most agents

water-borne transmission has been documented. For
some agents there is good evidence that, at least in
some places at some times, water is a major vehicle of
transmission. Notable examples are Salmonella
typhi, Vibrio cholerae, and Giardia lamblia (39).
Traditional water sources are often highly con­
taminated with faecal matter. Improved water
sources may be free of contamination or considerably
less contaminated than unimproved water sources
(30, 77). Uncontaminated source water may become
polluted by the time it is ingested, and water storage
in home containers may result in increased con­
tamination depending on storage conditions (37, 66,
72).
Increased water availability and quantity,
associated with improved hygiene, may reduce faecal
contamination of the hands. Proper cleaning of
utensils, food, and home environments is also likely
to reduce transmission of faecal matter. The
transmission of all the main diarrhoea-causing agents
is probably influenced to some degree by increased
water availability and quantity, but it is Shigella
transmission that has been particularly associated
with poor personal and domestic hygiene (45, 78).
This may be because of the low infectious dose of
Shigella relative to other bacterial enteric pathogens,
or it may be only because Shigella has been most
studied. The relationship between personal hygiene
and the newly-recognized diarrhoea agents
(especially Campylobacter jejuni, enterotoxigenic
Escherichia coli, and rotavirus) should be studied.
All the major infectious agents of diarrhoea are
shed by infected persons via the faeces, and therefore
hygienic disposal of human excreta plays a role in
controlling them. Use of toilets by all members of the
community should reduce faecal contamination of
houses, yards and gardens, and the neighbourhood.
In addition, proper treatment and disposal of human
excreta would prevent faecal contamination of fields,
crops, and receiving water-bodies, which would in
turn further reduce the transmission of faecal
pathogens. The hygienic disposal of the faeces of
children too young to use the toilet is of the utmost
importance, because such children constitute an
important reservoir of several agents of diarrhoea
(for instance, rotavirus and enterotoxigenic E. coli).

Hypothesis 2. A reduction in the ingestion of these
pathogens by young children can reduce diarrhoea
morbidity or mortality rates.

The ingested dose of a pathogen required to cause
diarrhoea depends upon the particular properties of
the pathogen and upon a number of host factors. The
general relationship between ingested dose and
proportion of exposed persons contracting diarrhoea

IMPROVING WATER SUPPLIES AND EXCRETA DISPOSAL FOR CONTROL OF DIARRHOEAL DISEASES

759

Fig. 2. Dose-response relationship for young children
under various levels of exposure to an array of enteric
pathogens.

Fig. 1. Dose-response relationship for a group of
susceptible persons, all exposed to an equal dose of
pathogen X.

is shown in Fig. I. In the dose range below A, no one
becomes ill.
*
In the dose range above B, all
susceptible persons develop diarrhoea. Between A
and B lies an intermediate range in which some
persons become ill and others do not. The dose at
which 50% of challenged persons become ill is known
as the median infectious dose (IDso), which is the
figure generally reported from volunteer studies.
Little is known about the shape of the curve in the
intermediate range and, for this reason, a broken line
is shown in Fig. 1. Both the shape of the curve and the
values of A, B, and the ID50 depend upon the
particular pathogen, its method of ingestion (in water
or in food), and a variety of features of the exposed
group of people, such as age and immunity.
Available ID50 and other infectious dose data have
been recently reviewed for all the major diarrhoea­
causing pathogens (39). For bacterial agents there is a
wide range of lD<o values—from around 103 for
Shigella to 108-10" for Vibrio cholerae. Less is
known about the viral and protozoal agents of
diarrhoea, although there are grounds for assuming
that the IDso values are relatively low (< 102). Nearly
all infectious dose data are derived from studies on
volunteers in which the subjects were healthy adults
from developed countries. The doses necessary to
infect children, particularly malnourished children,
may be very different. ID50 values also depend on the
food or drink with which the pathogens are ingested;
therefore, they may differ among countries with
differing dietary and child-feeding practices. In
* Despite the representation in Fig. 1, it may be that for some
pathogens A = 0.

addition, in situations where large numbers of
persons are exposed to diarrhoea-causing pathogens,
the 1D| or IDo.t values may be of greater
epidemiological relevance than the IDso values. More
reliance can be placed on the relative ranking of
pathogens by IDso than on the absolute dose values
obtained from studies on volunteers.
The relationship depicted in Fig. 1 may be
generalized to a group of young children having
different levels of water supply and excreta disposal
services and, consequently, different levels of
ingestion of enteric pathogens (Fig. 2). Consider first
the incidence rate of mild diarrhoea. At low levels of
ingestion (A-B), there remains an appreciable
incidence of mild diarrhoea, made up of an
irreducible minimum of infectious diarrhoea, plus
diarrhoea not due to enteric pathogens. This situation
is exemplified by children in wealthy communities in
developed countries. When ingestion rises above
point B, the incidence rate of mild diarrhoea also
rises. At point D saturation is reached, and further
increases in the ingestion of pathogens do not result in
an increased incidence rate of mild diarrhoea. As in
Fig. 1, the shape of the curve between points B and D
on the “mild diarrhoea incidence” line in Fig. 2 is
unknown, so a broken line is shown. Poor com­
munities in developed and developing countries, with
their elevated diarrhoea incidence rates, clearly lie to
the right of point B.
The incidence rate of severe diarrhoea, which may
be defined by stooling rate, stool volume, duration,
degree of dehydration or other measures, is indicated
on Fig. 2 by the distance between the two lines (the
shaded band). The incidence rate of severe diarrhoea
is less than that of mild diarrhoea, but it represents an
increasing proportion of the total diarrhoea incidence
rate as the ingestion of enteric pathogens rises above

760

S. A. ESREY ET AL.

the level represented by point C.c The incidence rate
of severe diarrhoea is shown in Fig. 2 to be constant in
the range A-C, to rise in the range C-E, and to be
constant in the range E-F. The breakpoints for severe
diarrhoea incidence rate (C and E) are offset to the
right of the equivalent points for mild diarrhoea inci­
dence rate (B and D), on the assumption that for a
single pathogen a higher ingested dose is necessary to
produce severe diarrhoea than mild diarrhoea. There
is direct experimental confirmation of this for entero­
toxigenic E. coli (27) and for Vibrio cholerae (19).
Indirect evidence is also available for Salmonella, for
which there is an inverse relationship between dose
and incubation period (12) and an inverse rela­
tionship between incubation period and severity (10).
Fig. 2 assumes that the IDso for severe symptoms is
higher than the IDso for mild symptoms for most
enteric pathogens.
The model put forward in Fig. 2 is tentative and
grossly simplified. It may be more applicable to
young children than to a whole community. The
complex role of immunity is not specifically
addressed in the model and, for certain pathogens, the
improvement of hygienic conditions may lead to an
increase in diarrhoea incidence rates in older age
groups. A more complete modelling of the
interrelationships between hygiene levels and
diarrhoea incidence is difficult because of the wide
differences in epidemiology and immunology among
the major diarrhoea-causing agents. This simplified
model provides some theoretical basis for the
explanation of a number of observed features of
childhood diarrhoea. We hope that it may stimulate
others to conduct studies to define more precisely the
complex reality.
The implications of Fig. 2 for diarrhoea control by
reducing pathogen ingestion are as follows. If
pathogen ingestion is reduced within the range F-E,
the incidence rates of mild or severe diarrhoea may
not change. In the range E-D, severe diarrhoea rates
may fall, but mild diarrhoea rates may not. Because
severe cases usually represent a small proportion of all
cases, surveillance of all cases may fail to detect a fall
in incidence rate in this range. For instance, if a water
supply and excreta disposal project reduced the dose
of ingested pathogens from E to D, Fig. 2 suggests
that the incidence rate of severe diarrhoea might fall
by about 44% but the incidence rate of all diarrhoea
by only about 12%. Many diarrhoeal disease studies
are unable to detect a 12% fall in total incidence rate
or to show it to be statistically significant. In the range
D-E an impact is more likely to be documented if data
on the incidence rate of severe diarrhoea (however
f The highest ratios of severe diarrhoea to all diarrhoea have been
recorded in some epidemics, which may be associated with excep­
tionally high average doses of pathogens ingested by an exposed
population.

defined) are collected. Measures of diarrhoea or total
mortality are also more likely to detect an impact in
this range (see below).
In the range D-C in Fig. 2, both mild and severe
diarrhoea incidence rates are falling, and a change
may be detected by surveillance of all cases. Severity,
child growth, or mortality parameters are also likely
to change in this range. In the range C-B, only the
incidence rate of mild cases is declining but, since
most cases are mild, surveillance of all cases may
detect an impact. In the range B-A, reductions in
pathogen ingestion have no effect on diarrhoea of any
type.
This discussion of the implications of the model
presented in Fig. 2 may be restated in two other ways.
First, since it is severe episodes that lead in some
instances to death, the mild diarrhoea incidence rate
in Fig. 2 may be replaced by the total diarrhoea
incidence rate and the severe diarrhoea incidence rate
by the diarrhoea mortality rate. Diarrhoea mortality
rates may therefore be a more responsive indicator in
the range D-E or ranges overlapping with the range
D-E than diarrhoea morbidity rates. Diarrhoea
mortality rates will not, however, be a good measure
of impact in areas where oral rehydration therapy is
widely available and averts most deaths from
dehydration. Second, since there are putative
differences in IDso values, not only between degrees
of severity of diarrhoea caused by a single pathogen,
but also among different pathogens, the mild
diarrhoea incidence rate may be replaced by the
incidence rate of etiologies having low ID$o values,
and the severe diarrhoea incidence rate by the
incidence rate of etiologies having high IDJ0 values/
This model is consistent with known facts, in that
developed countries, which may lie in the range A-C,
have a small proportion of cholera and enterotoxi­
genic E. coli diarrhoea (high IDso), a high proportion
of rotavirus diarrhoea (low IDso), and an
intermediate proportion of shigellosis (intermediate
ID50).f
It must be emphasized that the model presented in
Fig. 2 is hypothetical and grossly simplified. It is,
however, consistent with several established facts
and, as discussed below, it is helpful in explaining
some of the variation in the recorded impacts of water
supply and excreta disposal projects on diarrhoeal
diseases. A somewhat similar model has been
previously published (73)._____________________
d It is not implied here that pathogens having relatively low IDso
values cause relatively mild diarrhoea and vice versa. This is clearly
untrue in the cases of rotavirus and Shigella, both of which have
relatively low IDso values but cause relatively severe diarrhoea.
' This analysis probably only applies to the anthroponotic agents
mentioned here. For the zoonotic diarrhoea agents, such as
Salmonella and Campylobacter, most transmission in developed
countries is from infected animals to man via contaminated food
products, and so levels of pathogen ingestion depend more on
farming methods, food handling practices, and diet than on domestic
water supply and human excreta disposal facilities.

IMPROVING WATER SUPPLIES AND EXCRETA DISPOSAL FOR CONTROL OF DIARRHOEAL DISEASES

Hypothesis 3. Water supply or excreta disposal
improvements can reduce diarrhoea morbidity
or mortality rates among young children.
Numerous attempts have been made to measure the
impact on health of improved water supply or
sanitation. We selected for review those 67 studies
from 28 countries-^ that measured health impact in
J Studies were identified from previous reviews, supplemental
enquiries to workers in the field, computer searches in five
languages, and from unpublished papers presented at the Inter­
national Seminar on Measuring the Health Impact of Water and
Sanitation Programmes, Cox’s Bazaar, Bangladesh, 21-25
November 1983 (sponsored by the International Centre for
Diarrhoeal Diseases Research, Bangladesh, and the Ross Institute,
London).

761

terms of diarrhoea morbidity, Shigella infection or
disease, cholera, Entamoeba histolytica infection,
Giardia lambda infection, nutritional anthro­
pometry, diarrhoea mortality, or total mortality.
These studies are grouped in Table 1 according to the
health impact indicator measured and ordered alpha­
betically by country within each group.
The data abstracted from the studies listed in Table
1 are summarized in Tables 2-5. All the studies in
Table 1 display methodological deficiencies (73, 28)
although some studies were better than others. The
total number of studies is large (67), but only a few are
reported in sufficient detail to allow an objective
assessment of their methodological and analytical

Table 1. Studies on the impact of water supply or excreta disposal on diarrhoea morbidity and mortality, enteric
infections, total mortality, and nutritional status reviewed in this paper
Indicator
Diarrhoea morbidity

Country

Reference

Bangladesh

22, 54, 74,°
24

Chile

Ethiopia

51, 88
62
82
40

Gambia

See below6

Guatemala

16, 72
79
53. 66, 67, 81
82

Colombia

Costa Rica
Egypt

Haiti
India

Iran (Islamic Rep. of)

Lesotho

86,c
37

Mozambique

See belowd

Sri Lanka

82

Saint Lucia

44,g

Sudan

6, 82
17
8, 15, 55, 64, 69, 71, 84

Kenya

United Kingdom

USA
Venezuela
Zambia

Cholera

Entamoeba histolytica
infection

82, 87
4

Philippines

22, 46, 49, 50, 54, 76, 77
3

Costa Rica

62

Egypt
India

20
59, 66

Kenya

See belowr

Libyan Arab Jamahiriya

41
15, 29, 57

Bangladesh

USA

Table 1: continued on next page

762

S. A. ESREY ET AL.

Table 1: continued
Indicator

Country

Reference

Giardia lamblia

Costa Rica

62

Egypt
India

20
66

Kenya

See belowr

Libyan Arab Jamahiriya

41

USA

29

Bangladesh

22, 48, 65“
62

infection

Shigella
infection or disease

Costa Rica
Egypt

Guatemala
India

Iran (Islamic Rep of)

Libyan Arab Jamahiriya

Venezuela

52
82
82
45. 55, 71, 78, 84
82

Panama
Sri Lanka

Sudan
USA

Nutritional status

Diarrhoea mortality

Bangladesh

See below"

Colombia

21

Fiji

See below
*
-

Nigeria

80

Philippines

See below/

Saint Lucia

See below"

Brazil

83
89

India

Total mortality

82
7
66
82
41

Brazil

Costa Rica

61
43

Sri Lanka

85
1
18
60, 63

Sudan

6

Egypt

Guatemala

Malaysia

|

Rahaman, M. M. The Teknaf Health Impact Study: methods and results. Paper presented at the International Workshop on
Measuring the Health Impacts of Water Supply and Sanitation Programmes, Cox's Bazaar, Bangladesh, 21-25 November 1983.
h PICKERING, H. The role of anthropologists in studying diarrhoea epidemiology: a case study from The Gambia. Paper presented at
the International Workshop on Measuring the Health Impacts of Water and Sanitation Programmes, Cox's Bazaar, Bangladesh, 21-25
November 1983.
FENWICK, K. W. H. The short-term effects of a pilot environmental health project in rural Africa: the Zaina scheme re-assessed
after four years, (undated manuscript).
d CAIRNCROSS, S. & Cliff, J. Water and health in Mueda, Mozambique. Paper presented at the International Workshop on
Measuring the Health Impacts of Water and Sanitation Programmes, Cox’s Bazaar, Bangladesh, 21 -25 November 1983.
YEE, V. S. Household level correlates of child nutritional status in Fiji. MPS thesis. Division of Nutritional Sciences, Cornell
University, Ithaca, New York, 1984.
1 MAGNANl, R. J. & TOURKIN, S. C. Impact of improved urban water supplies in the Philippines: methods and results. Paper
presented at the International Workshop on Measuring the Health Impacts of Water and Sanitation Programmes, Cox's Bazaar,
Bangladesh, 21-25 November 1983.
K Henry, F. J. Health impact of water and sanitation interventions in St. Lucia. Paper presented at the International Workshop on
Measuring the Health Impacts of Water Supply and Sanitation Programmes, Cox's Bazaar, Bangladesh, 21 -25 November 1983.

764

S. A. ESREY ET AL.

Table 3. Percentage reductions in diarrhoea morbidity
rates attributed to water supply or excreta disposal
improvements by adult literacy rate of the country and
magnitude of service improvement
Small service
*improvements'

Adult
literacy
rate"
(%)

No. of
results

Median
(%)

Large service
*improvements'

No. of
results

Median
(%)

<40

11

18

7

46

40-75

4

20

8

39

> 75

10

16

13

32

‘ Data on adult literacy, by country, from World development

report, 1983 (Washington. World Bank)
,p See footnote g. page 762.

anticipated that the impact on enterotoxigenic E. coli
will be considerable and on rotavirus negligible. The
latter prediction is indirectly supported by data show­
ing that the incidence rate of rotavirus diarrhoea
among children under 2 years of age is 0.3-0.4
episodes per child per year in both Bangladesh (11)
and Winnipeg, Canada (42).

Impact on nutritional anthropometry. If water
supply and excreta disposal improvements reduce
diarrhoea incidence rates or duration among young
children, then nutritional anthropometric indicators
should also improve because of the inverse relation­
ship between time spent with diarrhoea and child
growth (58, 68). Six studies that investigated the
relationship between water supply or excreta disposal
improvements and nutritional status are summarized
in Table 5. All six studies reported an association
between improved water supply or excreta disposal
and improved nutritional status. In two studies, in
Fiji and the Philippines, attempts to control for
extraneous risk factors reduced the differences
between the control and intervention groups, but
some of these differences were nonetheless found to
be statistically significant.

to small service-level improvements (Table 3). If a
large service-level improvement is made, the
percentage reduction in diarrhoea morbidity rates
achieved is inversely related to the pre-intervention
level of hygiene (Table 3). The greatest impact is
achieved when the pre-intervention hygiene level is
Impact on mortality. Only two studies were located
worst. For small improvements in service level,
impacts appear less dependent upon the pre­ that reported the impact of water supply or excreta
intervention hygiene level. These conclusions are disposal improvements on diarrhoea mortality rates
consistent with the hypothetical model depicted in (Table 1); a 41% median reduction in diarrhoea
Fig. 2 and suggest that the impact of water and mortality rate was calculated from them. Both studies
sanitation improvements depends in part on the were concerned with the impact of improved water
supplies in urban areas, and in neither was the study
presence and interaction of other risk factors.
method well described. A further eight studies
reported impacts of water supply or excreta disposal
Impact on specific infections. When impact on
improvements on mortality from all causes (Table 1),
total diarrhoea morbidity is broken down by etiology,
it is likely, as discussed above, that different specific and they indicated a 21% (range, 0-81%) median
.diarrhoeas will be reduced by different amounts.
reduction in mortality rate.
Table 4 presents data on impacts on cholera, Shigella
infection or disease, and infection by Ent. histolytica
and G. lamblia. Since the distinction between the
severe and mild diarrhoea bands in Fig. 2 is merely
one of infectious dose (ID), in ranges including D-E, Table 4. Percentage reductions in morbidity or infection
any reduction in pathogen ingestion will produce a rates of cholera. Shigella, Entamoeba histolytica, and
greater percentage reduction in the incidence rate of Giardia lamblia attributed to water supply or excreta
the high ID etiologies than the low ID etiologies. Thus disposal improvements
the impact of a water supply and excreta disposal
Percentage
improvement (in ranges including D-E) on specific
reduction
etiologies may be in the following descending order of
magnitude: cholera, enterotoxigenic E. coli, Shigella,
Median
Range
No.
of
Disease or
the protozoa, rotavirus. The relative impacts on infection
results
(%)
(%)
Shigella and the protozoa are supported by the data in
41
0-91
11
Table 4. The anomalous cholera data are discussed Cholera
48
0-81
elsewhere (31).
27
Shigella
Data on the impact of water supply and excreta Entamoeba histolytica
2
0-80
17
disposal projects on enterotoxigenic E. coli and Giardia lamblia
10
0
0-20
rotavirus incidences are not yet available, but it is

IMPROVING WATER SUPPLIES AND EXCRETA DISPOSAL FOR CONTROL OF DIARRHOEAL DISEASES

765

Table 5. Improvements in nutritional status attributed to various types of water supply or excreta disposal improve­
ment
Value of indicator

Country

Nutritional
indicator"

Bangladesh

Percentage with H/A > 90% of standard

Percentage with W/A > 75% of standard

Colombia

Percentage with W/A > 90% of standard

Percentage with H/A > 95% of standard

Fiji

Age
(months)

Type of
intervention''

Control
group

Target
group

Reference

See below d

0-11

Q+A+E

75

76

12-23

Q+A+ E

50

51

24-35

Q+A+E

45

48

0-11

Q+A+E

59

63

12-23

Q+A+E

44

43

24-35

Q+A+E

47

50

6-30

E

26

47

6-30

A

22

51

6-30

E

26

48

6-30

A

20

53

21

See below
*'

Mean percentage of:
standard W/A

urban

0-59

E

102

100

rural

0-59

E

95

102

urban

0-59

E

100

100

rural

0-59

E

99

103

W/A > 75% of standard

6-59

A

50

69

H/A > 90% of standard

6-59

A

80

69

W/H > 80% of standard

6-59

A

63

90

74

75

standard H/A

Nigeria

Philippines

Saint Lucia

80

Percentage with:

See below7

Percentage with W/A > 75% of standard
pre-intervention

6-54

post-intervention

6-54

Q+A

71

80

1-3

A (A + E)

93

92 (86)f

72

90 (88)

Percentage with W/A > 90% of standard

4-6

7-9

51

78 (76)

10-12

44

76 (73)

13-15

50

76 (58)

16-18

51

74 (57)

19-21

53

75 (66)

22-24

54

79 (71)

See below
*

a W/A = weight for age; W/H = weight for height; H/A = height for age.
b Q = water quality improvement, A = water availability improvement, E = excreta disposal improvement.
c Figures not in parentheses refer to the communities receiving water availability improvements (A). Figures in parentheses refer to
communities receiving water availability improvements plus excreta disposal facilities (A + E).
d See footnote a to Table 1.
* See footnote e to Table 1.
7 See footnote / to Table 1.
* See footnote g to Table 1.

766

S. A. ESREY ET AL.

Fig. 2 predicts a larger impact on diarrhoea
mortality than morbidity over a wide range of
conditions. The impact on total mortality will depend
on the proportion of all mortality that is due to
diarrhoea and the degree to which water supply and
excreta disposal improvements affect the causes of
death other than diarrhoea. Since water supply and
excreta disposal improvements will have little impact
on some major causes of death (for instance,
respiratory infection, measles, malaria, and neonatal
tetanus), it is to be expected that the impact on
diarrhoea mortality is considerably greater than the
impact on all mortality.
Results from selected studies. For the analyses so
far, we have used pooled data from all studies listed in
Table 1 to give an overview of all documented
experiences on the impacts of water supply and
excreta disposal improvements on diarrhoea. It is
instructive to compare these findings with the results
of a few of the better studies. Criteria for judging the
quality of each study have been developed and are
reported elsewhere (28).
The first finding from this selective analysis was
that all studies that reported a negative impact were
flawed in one or more major respects. In other words,
the better studies consistently reported positive
impacts.
The best studies were on total mortality (18, 43,
60), and the median impact on total mortality rates
reported from these studies was a reduction of 30%.
Analysis for statistical interactions revealed a range
of reductions in mortality rates of 8-64%, depending
on the type of intervention and on the presence of
other risk factors, such as poor feeding practices and
low literacy rates. For example, in one study (18)
excreta disposal improvements were reported to have
a larger impact on infant mortality rates than water
supply, but the magnitudes of these impacts were
greatly affected by whether the infants were breast­
fed. The impact of the environmental interventions
was greater for non-breast-fed infants than for breast­
fed infants. Thus, it is likely that non-breast-fed
infants were further to the right along the horizontal
axis in Fig. 2 than were breast-fed infants. Another
study reported that excreta disposal improvements
had a greater impact on mortality in families with
literate mothers than in families with illiterate
mothers (60). This may reflect an increased ability of
literate mothers to make correct use of the new
excreta disposal facilities.
Two studies on nutritional status also examined
statistical interactions. In one/ excreta disposal was
reported to have more impact in rural than in urban
* Y EE, V. S. Household level correlates of child nutritional status
in Fiji. MPS thesis. Division of Nutritional Sciences, Cornell

University, Ithaca, New York, 1984.

areas. In the other,' water quality improvements had
an impact only among higher income households.
Studies on diarrhoea morbidity were not as well
controlled as the mortality studies referred to above.
A median reduction in diarrhoea morbidity rates of
27% (range, 0-68%) was found in the studies judged
to be most satisfactory (40, 51, 55, 71). **
''
This brief review of selected studies of superior
design leads to two conclusions. First, the median
reductions in diarrhoea morbidity rates (27%) and
total mortality rates (30%) are a little higher than the
values found by analysing all studies. Second, the
magnitude of the impact depends greatly on the
presence of other risk factors. More knowledge of
these interactions would enable the appropriate type
of intervention to be targeted to families that are
likely to benefit the most.
FEASIBILITY

Nearly all developing countries are currently
engaged in substantial programmes to improve water
supplies in both rural and urban areas. The urban
programmes date back, in many cases, to the 1920s or
earlier, while many of the rural programmes were
initiated in the 1960s. This considerable experience in
water supply programmes throughout the world has
been copiously documented (e.g., 23) and is con­
tinuously monitored by WHO."1 Improved water
supplies can be provided to almost all people in al!
developing countries, and the technologies for
achieving this are, for the most part, well established.
Several problems remain, however, such as poor
operation and maintenance, inappropriate choice of
technology, inadequate revenue collection, failure to
sustain community participation, and high rates of
water leakage and wastage.
All countries have experience in excreta disposal!
MaGNANI, R. J.&TOURKIN, S. C. Impact of improved urban
water supplies in the Philippines, methods and results. Paper
presented al the International Workshop on Measuring the Health
Impacts of Waler and Sanitation Programmes, Cox’s Bazaar,
Bangladesh, 21—25th November 1983.
‘ CAIRNCROSS, S. & CLIFF, J. Waler and health in Mueda,
Mozambique. Paper presented al the International Workshop on

Measuring the Health Impacts of Water and Sanitation Programmes,
Cox’s Bazaar, Bangladesh, 21-25 November 1983.
* PICKERING, H. The role of anthropologists in studying
diarrhoea epidemiology: a case study from the Gambia. Paper

presented al the International Workshop on Measuring the Health
Impacts of Water and Sanitation Programmes, Cox’s Bazaar,
Bangladesh, 21-25 November 1983.
1 RAHAMAN, M. M. The Teknaf Health Impact Study: methods
and results. Paper presented al the International Workshop on
Measuring the Health Impacts of Waler Supply and Sanitation
Programmes, Cox’s Bazaar, Bangladesh, 21-25 November 1983.
See, for example, World Health Statistics Report, Vol. 26, No.
11, 1973; World Health Statistics Report, Vol. 29, No. 10, 1976; and

The International Drinking Water Supply and Sanitation Decade.
Review of national baseline data (as al 31 December 1980), Geneva,

World Health Organisation, 1984 (WHO Offset Publication No.85).

IMPROVING WATER SUPPLIES AND EXCRETA DISPOSAL FOR CONTROL OF DIARRHOEAL DISEASES

programmes in urban areas, in some cases dating
back thousands of years. Rural excreta disposal
programmes are typically a new phenomenon, and
some countries still lack concerted efforts in this
sector. Problems commonly encountered in excreta
disposal programmes include the inappropriate
choice of technology, poor operation and main­
tenance, inadequate revenue collection, and the lack
of perception in many rural communities of the
importance of improved excreta disposal practices.
Research into the technical, economic, and social
aspects of excreta disposal in developing countries
(e.g., see 47) over the past decade has led to some
promising new approaches.
COSTS

Despite the extensive experience with water supply
and excreta disposal projects throughout the world,
the cost data on these projects are often of poor
quality and not strictly comparable. Data on
operation and maintenance costs, on institutional
overhead costs, and on the costs of community
mobilization and education, are especially deficient.
Table 6 presents cost data from 87 developing
countries. Costs vary widely depending upon the
design criteria adopted in different countries and

767

upon the costs of labour and materials. The costs of
operation and maintenance (recurrent costs) must be
added to the construction costs, and data on these
costs are not readily available. One study of excreta
disposal in 12 countries estimated that operation and
maintenance accounted for between 4% and 52% of
the total project costs per year, depending on the
technology under consideration (47). Several widely
used technologies (for instance, sewerage and pour­
flush latrines) had operation and maintenance costs
that comprised approximately 30% of the total
annual project costs. For rural water supplies, data
assembled by a UNDP/World Bank project'1 suggest
that the same figure, 30%, is a reasonable estimate of
the proportion of the total annual costs taken up by
operation and maintenance. The 30% figure is
adopted in Table 6 for both water supplies and excreta
disposal." Reliable data on the costs of software
support (such as the promotion of community par" UNDP project 1NT/81/026.
” The computation of the percentage of total annual costs due to
operation and maintenance requires the adoption of a discount rate
(or opportunity cost of capital). Choosing the discount rate is partly a
matter of judgement, and different economists may advocate
different rates for the analysis of the same project. The choice of a
high discount rate (say, 20%) will reduce the apparent importance of
recurrent costs within total project costs. If a low discount rate (say,
5%) is chosen, perhaps to reflect the scarcity of recurrent funds (5),
the proportion of total costs that is attributed to operation and
maintenance will be greatly increased.

Table 6. Costs for urban and rural water supply and excreta disposal projects

Lifetime
*
(years)

Annual
construction
cost per
capita
(1982, US$)f

Annual
total
cost per
capita
(1982, USSI"

116

20

14

20

66

20

8

11

60

20

7

10

Sewerage

174

50

18

26

Other

66

20

8

11

19

10

3

4

Construction cost
per capita
(1982, US$)a

House connection
Public tap

Type of service

^Vater supply
Urban

Rural'
Sanitation

Urban

Rural

a Median values of the costs from 87 developing countries reported to WHO. For further information, see The International Drinking
Water Supply and Sanitation Decade. Review of national baseline data fas at 31 December 1980), Geneva, World Health Organization, 1984
(WHO Offset Publication No. 85).
b Commonly assumed values.

f Assuming an opportunity cost of capital (discount rate! of 10%.
d Assuming that construction costs are 70% of the total annual costs (see text).

' A variety of technologies are included here, but predominantly public taps and handpumps.

768

S. A, ESREY ET AL.

ticipation or hygiene education) and of apportioned
institutional overheads have not been located, and so
these costs, which are not trivial, have been omitted
from the calculations.
Total annual costs per capita are derived in Table 6,
assuming that recurrent costs comprise 30% of the
total annual costs, that project lifetimes are between
10 and 50 years (depending upon the technology), and
that the appropriate discount rate is 10%. The total
annual costs per capita in Table 6 may be aggregated
to derive the costs of complete water supply and
excreta disposal interventions. For instance, a rural
water supply and excreta disposal project might cost
USS 14 per capita annually (S10 + S4), whereas a
combination of in-house water and sewerage in an
urban area might have an annual cost of USS 46 per
capita (S20 + S26).
A goal of this series of reviews (38) is to derive cost­
effectiveness estimates for each intervention. Special
difficulties are inherent in applying the cost­
effectiveness analysis to interventions having multiple
benefits, and water supply and excreta disposal
interventions present these difficulties in an extreme
form (9, 14). In addition to their impact on diarrhoea
rates among young children, these interventions may
avert diarrhoea in other age groups, reduce the
incidence of other infectious diseases, and have a
variety of benefits unrelated to health. In view of this,
a treatment of the cost-effectiveness of water supply
and excreta disposal in relation to other diarrhoea
control measures will be left to a later publication.

CONCLUSIONS

The results in Tables 2-4 show that substantial
reductions in diarrhoea morbidity and mortality rates
can be expected from investments in water supply and
excreta disposal. Table 2 suggests that investments
that improve both water quality and availability are
especially effective. There are no adequate data on the

impact of improvements in water quality plus
availability together with excreta disposal. Likewise,
the available data do not permit an assessment of the
advantages of adding a hygiene education component
to a project, but analysis of hygiene education alone
suggests that it may further enhance the impact (34).
Taking all this evidence together, and in view of the
impact of large service-level improvements shown in
Table 3, it is possible that well-designed projects
combining water supply, excreta disposal and hygiene
education may achieve diarrhoea morbidity rate
reductions of 35-50%. It is to be expected that, in any
given project, the impact on diarrhoea mortality rates
will be larger than that on diarrhoea morbidity rates,
except in areas where other interventions, such as oral
rehydration programmes, have substantially reduced
the risk of death from diarrhoea.
This review highlights some of the deficiencies in
our knowledge of the impacts of water supply and
excreta disposal on diarrhoea. More studies of these
impacts are required, and it is expected that current
advances in methodology'’ will enable such studies to
be undertaken retrospectively and at reasonable cost.
The model discussed under hypothesis 2 suggests that
there may be advantages in measuring the impact on
severe diarrhoea rather than all diarrhoea. Where
etiology-specific studies are being conducted, they are
more likely to record impacts on diarrhoea due to
agents having high infectious doses than on diarrhoea
due to agents having low infectious doses. The most
pressing research need is to document the impact on
diarrhoea of projects that combine improvements in
water quality, water availability and excreta disposal
with hygiene education that are functioning satis­
factorily and are being utilized by the intended bene­
ficiaries.''
'' BRISCOE, J. ET AL. Measuring (he impacts of water supply and
sanitation projects on diarrhoea: prospects for case-control method^
(in preparation).
fl

'' Minimum evaluation procedure (MEP) for water supply ana
sanitation projects. Unpublished document WHO/ETS/83.1,
1983.

ACKNOWLEDGEMENTS
We are grateful for the constructive criticisms of earlier drafts of this paper provided by R. C. Ballance, D. Blum,
J. Briscoe, A. M. Cairncross, I. de Zoysa, B. H. Dieterich, F. L. Golladay, M. B. Grieveson, R. C. Hogan, J. S.
Koopman, L. Laugeri, D. D. Mara, M. H. Merson, R. N. Middleton, M. Phillips, M. M. Rahaman, G. Schultzberg,
S. Unakul, and A. M. Wright. Secretarial, bibliographic and editorial assistance were most ably provided by Caprice
Mahalla, Susanne O’Driscoll, and Dianne Fishman.

IMPROVING WATER SUPPLIES AND EXCRETA DISPOSAL FOR CONTROL OF DIARRHOEAL DISEASES

769

RESUME
LA LUTTE CONTRE LES MALADIES DIARRHEIQUES DU JEUNE ENFANT: INTERVENTIONS VISANT A AMELIORER

L’APPROVISIONNEMENT EN EAU ET L'ELIMINATION DES EXCRETA

Cei article esi le neuvieme d’une serie d’analyses
consacrees aux mesures possibles pour lutter contre les
diarrhees. Dans les pays en developpemenl, 1’impact des
projeis d’approvisionnemem en eau et d’evacuation des
excreta sur les maladies diarrheiques a 6te au centre de tres
nombreuses discussions et recherches. Quelques observa­
tions donnent a penser que des ameliorations portant sur la
qualite de 1’eau, 1’approvisionnement en eau, 1’hygiene
individuelle et 1’evacuation des excreta, reduiraient
1’ingestion des agents pathogenes a 1’origine de diarrhees.
On expose ici un modele theorique qui etablit une relation
entre la quantite ingeree d’agents pathogenes responsables
de diarrhees et la frequence de ces affections dans la
communaute.
Ce modele montre que, dans des communaul6s pauvres,
mal approvisionnees en eau et ou 1’evacuation des excreta se
fail de fa?on insalubre, la diminution de la quantite d’agents
enieropaihogenes ingeree exerce, a valeur 6gale, un impact
plus grand sur les taux de mortalite que sur ceux de
morbite. Elie ameloire davantage les diarrhees graves que
les diarrhees benignes et plutot cellos dont les agents etiologiques agissent a dose elevee.
Le recensement de 67 etudes effectuees dans 28 pays
permet d’analyser 1’impact de 1’approvisionnement en eau et
de 1’evacuation des excreta sur les maladies diarrheiques, les
infections apparentees, 1’etat nutritionnel et la mortalite.
Toutes ces eludes montrent une diminution mediane de 22%
des taux de morbidite associes aux maladies diarrheiques, la
diminution atteignani 27% dans quelques etudes mieux
con?ues: pour le taux de mortalite global, la diminution
mediane a ete de 21%, atteignani 30% dans les quelques

eludes mieux contjues. L’ameiioration de la qualite de 1’eau
importante moins que celle de 1’approvisionnement ou de
1’evacuation des excreta. Mieux vaut ameliorcr la qualite de
1’eau et 1’approvisionnement qu’un seul de ces elements ou
que I’eiimination des excreta. Les projets d’approvisionnement en eau et d’evacuation des excreta ont un impact plus
important sur la shigellose que sur les infestations a
Entamoeba histolytica ou a Giardia lambda, mais on ignore
quelles peuvent en elre les repercussions sur les diarrhees
causees par Escherichia coli enterotoxigene ou les rotavirus.
Les six etudes portant sur les repercussions de 1’approvision­
nement en eau ou de 1’evacuation des excreta sur 1’etat
nutritionnel ont fait etat d’une amelioration.
L’article presente ensuite une analyse des donnees sur les
couts de construction pour les projets d’approvisionnemem
en eau et d’assainissement dans 87 pays en developpemenl.
Si 1’on ajoute les couts d’exploitation et d’entretien, sans
tenir compte des couts lies a la mobilisation et a reducation
de la communaute ni des frais generaux institutionnels, on
arrive a un total median de 14-46 dollars E.-U. (prix de
1982) par tete, selon 1’importance du service.
Il faudra effectuer davantage d’etudes concernant
1’impact des projects d’approvisionnemem en eau et
d’evacuation des excreta sur les maladies diarrheques. En
lout premier lieu, on etudiera 1’impact de projeis combi­
nant une amelioration de la qualite de 1’eau, de 1’appro­
visionnement en eau et de 1’evacuation des excreta A une
education en matiere d'hygiene, en s’imeressant & des
installations qui donnent satisfaction et sont effectivemenl
utilisees par les personnes visees.

REFERENCES
Ascoli, W. etal. Nutrition and infection field study in
Guatemalan villages, 1959-1964. IV. Deaths of infants
and preschool children. Archives of environmental
health, 15: 439-449 (1967).
2. Ashworth, A. & Feachem, R. G. Interventions for the
control of diarrhoeal diseases among young children:
prevention of low birthweight. Bulletin of the World
Health Organization, 63: 165-186 (1985).
3.
Azurin, J. C. & Alverq, M. Field evaluation of
environmental sanitation measures against cholera.
Bulletin of the World Health Organization, 51: 19-26
(1974).
4.
Bahl, M. R. Impact of piped water supply on the
incidence of typhoid fever and diarrhoeal diseases in
Lusaka. Medical journal of Zambia, 10: 98-99 (1976).
5. Baldwin, G. B. Why present value calculations should
not be used in choosing rural water supply technology.
World development, 11: 1075-1081 (1983).
6. Bannaga, S. E. I. & Pickford, J. Water-health
relationships in Sudan. Effluent and water treatment
journal, 18: 560-569 (1978).

1.

Beck, M. D. et al. Studies on diarrhoeal diseases in
Central America. 1. Preliminary findings on cultural
surveys of normal population groups in Guatemala.
American journal of tropical medicine and hygiene, 6:
62-71 (1957).
8. Berg, L. E. & Mowery, T. M. Health program
evaluation: impact study of the Indian Sanitation
Facilities Construction Act. Tucson, Health Program
Systems Centre, 1968.
9.
Berman, P. A. Selective primary health care: is
efficient sufficient? Social science and medicine, 16:
1054-1059 (1982).
10.
Bille, B. et al. An extensive outbreak of gastro­
enteritis caused by Salmonella newport. I. Some
observations on 795 known cases. Acta medico
Scandinavica, 175: 557-567 (1964).
11. Black, R. E. et al. Incidence and severity of rotavirus
and Escherichia coli diarrhoea in rural Bangladesh.
Implications for vaccine development. Lancet, 1:
141-143 (1981).
7.

770

S. A. ESREY ET AL.

Blaser, M. J. & Newman, L. S. A review of human
salmonellosis: I. Infective dose. Reviews of infectious
diseases, 4: 1096-1106(1982).
13. Blum, D. & Feachem, R. G. Measuring the impact of
water supply and sanitation investments on diarrhoeal
diseases: problems of methodology. International
journal of epidemiology, 12: 357-365 (1983).
14. Briscoe, J. Water and health: selective primary health
care revisited. American journal of public health, 74:
1009-1013 (1984).
15. Brooke, M. M. et al. A sample survey of selected
areas in and near Little Rock, Arkansas, to assess the
prevalence of Entamoeba histolytica. Bulletin of the
World Health Organization, 29: 813-822 (1963).
16. Bruch, H. A. et al. Studies of diarrheal disease in
Central America. V. Environmental factors in the
origin and transmission of acute diarrheal disease in
four Guatemalan villages. A merican journal of tropical
medicine and hygiene, 12: 567-579 (1963).
17. Burr, M. L. et al. Diarrhoea and the drought. Public
health (London), 92: 86-87 (1978).
18. Butz, W. P. et al. Environmental factors in the
relationship between breastfeeding and infant
mortality: the role of sanitation and water in Malaysia.
American journal of epidemiology, 119: 516-525
(1984).
19. Cash, R. A. et al. Response of man to infection with
Vibrio cholerae. I. Clinical, serologic and bacteriologic
responses to known inoculum. Journal of infectious
diseases, 129: 45-52 (1974).
20. Chandler, A. C. A comparison of helminthic and
protozoan infections in two Egyptian villages two years
after the installation of sanitary improvements in one of
them. American journal of tropical medicine and
hygiene, 3: 59-73 (1954).
21. Christiansen, N. et al. Family social characteristics
related to physical growth of young children. British
journal of preventive and social medicine, 29: 121-130
(1975).
22. Curlin, G. T. et al. The influence of drinking
tubewell water on diarrhea rates in Matlab Thana,
Bangladesh. International Centre for Diarrhoeal
Disease Research, Bangladesh, Working paper No. I,
1977.
23. Dangerfield, B. J., ed., Water supply and sanitation
in developing countries. London, Institution of Water
Engineers and Scientists, 1983.
24. Department of Rural Water Supply, National
Sanitary Works, Chile. Drinking water: a determi­
nant of health. World health forum, 4: 172-173
(1983).
25. de Zoysa, I. & Feachem, R. G. Interventions for the
control of diarrhoeal diseases in young children:
chemoprophylaxis. Bulletin of the World Health
Organization, 63: 295-315 (1985).
26. de Zoysa, I. & Feachem, R. G. Interventions for the
control of diarrhoeal diseases among young children:
rotavirus and cholera immunization. Bulletin of the
World Health Organization, 63: 569-583 (1985).
27. Dupont, H. L. et al. Pathogenesis of Escherichia coli
diarrhea. New England journal of medicine, 285: 1-9
(1971).
12.

Esrey, S. A. & Habicht, J. P. The impact of improved
water supplies and excreta disposal facilities on
diarrheal morbidity, growth, and mortality among
children. (Cornell International Nutrition Monograph
Series). Ithaca, Division of Nutrionel Sciences, Cornell
University (in press, 1985).
29. Eyles, D. E. et al. A study of Entamoeba histolytica
and other intestinal parasites in a rural west Tennessee
community. American journal of tropical medicine and
hygiene, 2: 173-190 (1953).
30. Feachem, R. G. Bacterial standards for drinking water
quality in developing countries. Lancet, 2: 255-256
(1980).
31. Feachem, R. G. Environmental aspects of cholera
epidemiology. III. Transmission and control. Tropical
diseases bulletin, 79: 1-47 (1982).
32. Feachem, R. G. Interventions for the control of
diarrhoeal
diseases
among
young
children:
supplementary feeding programmes. Bulletin of the
World Health Organization, 61: 967-979 (1983).
33. Feachem, R. G. Infections related to water and
excreta: the health dimension of the decade. In: Water
supply and sanitation in developing countries, London,
Institution of Water Engineers and Scientists, 1983, pp.
25-46.
34. Feachem, R. G. Interventions for the control of
diarrhoeal diseases among young children: promotion
of personal and domestic hygiene. Bulletin of the World
Health Organization, 62: 467-476 (1984).
35. Feachem, R. G. & Koblinsky, M. A. Interventions for
the control of diarrhoeal diseases among young
children: measles immunization. Bulletin of the World
Health Organization, 61: 641-652 (1983).
36. Feachem, R. G. & Koblinsky, M. A. Interventions for
the control of diarrhoeal diseases among young
children: promotion of breast-feeding. Bulletin of the
World Health Organization, 62: 271-291 (1984).
37. Feachem, R. G. et al. Water, health and develop­
ment, London, Tri-Med Books, 1978.
38. Feachem, R. G. et al. Diarrhoeal disease control:
reviews of potential interventions. Bulletin of the
World Health Organization, 61: 637-640 (1983).
39. Feachem, R. G. et al. Sanitation and disease. Health
aspects of excreta and wastewater management.
Chichester, John Wiley, 1983.
40.
Freij, L. et al. Child health and diarrhoeal disease in
relation to supply and use of water in African
communities. Progress in water technology, 11: 49-55
(1978).
41. Ghannoum, M. A. et al. The incidence of waterrelated diseases in the Brak area, Libya, from 1977 to
1979, before and after the installation of water
treatment plants. Zentralblatt fiir Bakteriologie,
Mikrobiologie und Hygiene, I. Abteilung Orig. B, 173:
501-508 (1981).
42. Gurwith, M. et al. A prospective study of rotavirus
infection in infants and young children. Journal of
infectious diseases, 144: 218-224 (1981).
43. Haines, M. R. & Avery, R. C. Differential infant and
child mortality in Costa Rica: 1968-1973. Population
studies, 36: 31-43 (1982).

28.

IMPROVING WATER SUPPLIES AND EXCRETA DISPOSAL FOR CONTROL OF DIARRHOEAL DISEASES

771

Henry, F. J. Environmental sanitation, infection and 62. Moore, H. A. et al. Diarrheal disease studies in Costa
Rica. IV. The influence of sanitation upon the
nutritional status of infants in rural St. Lucia, West
prevalence of intestinal infection and diarrheal disease.
Indies. Transactions of the Royal Society of Tropical
Medicine and Hygiene, 75: 507-513 (1981).
American journal of epidemiology, 82: 162-184
(1965).
45. Hollister, A. G. etal. Influence of water availability
63. Patel, M. Effects of the health services and
on Shigella prevalence in children of farm labor
environmental factors on infant mortality: the case of
families. American journal of public health, 45:
354-362 (1955).
Sri Lanka. Journal of epidemiology and community
health, 34: 76-82(1980).
46. Hughes, J. M. et al. Epidemiology of eltor cholera in
64. Petersen, N. J. & Hines, V. D. The relation of
rural Bangladesh: importance of surface water in trans­
summertime gastrointestinal illness to the sanitary
mission. Bulletin of the World Health Organization,
quality of the water supplies in six Rocky Mountain
60: 395-404 (1982).
communities. American journal of hygiene, 71: 314-320
47. Kalbermatten, J. M. et al. Appropriate sanitation
(1960).
alternatives: a technical and economic appraisal.
65. Rahaman, M. M. A strategy for control of shigellosis
Washington, DC, World Bank, 1982.
(dysentery) in Teknaf—a rural Bangladesh village.
48. Khan, M. U. Interruption of shigellosis by hand
Progress in water technology, 11: 303-308 (1978).
washing. Transactions of the Royal Society of Tropical
66.
Rajasekaran, P. etal. Impact of water supply on the
Medicine and Hygiene, 76: 164-168 (1982).
incidence of diarrhoea and shigellosis among children in
49. Khan, M. & Curlin, G. T. Urban cholera study, 1974
rural communities in Madurai. Indian journal of
and 1975, Dacca. International Centre for Diarrhoeal
medical research, 66: 189-199 (1977).
Disease Research, Bangladesh, Scientific report No. 7,
67. Raman, V. et al. Assessment and control of water
1977.
quality in a town distribution system with reference to
50. Khan, M. U. etal. The relationship of cholera to water
the incidence of gastrointestinal diseases. Progress in
source and use in rural Bangladesh. International
water technology, 11: 65-71 (1978).
journal of epidemiology, 10: 23-25 (1981).
68.
Rowland, M. G. M. et al. A quantitative study into
51. Koopman, J. S. Diarrhea and school toilet hygiene in
the
role of infection in determining nutritional status in
Cali, Colombia. American journal of epidemiology,
Gambian village children. British journal of nutrition,
107: 412-420(1978).
37: 441-450 (1977).
52. Kourany, M. et al. Prevalence of pathogenic enteric
69. Rubenstein, A. et al. Effect of improved sanitary
bacteria in children of 31 Panamanian communities.
facilities on infant diarrhea in a Hopi village. Public
American journal of tropical medicine and hygiene, 20:
health reports, 84: 1093-1097 (1969).
608-615 (1971) .
53. Kumar, P. et al. Bore-hole disposal of excreta of 70. Saunders, R. J. & Warford, J. J. Village water
supply: economics and policy in the developing world.
children and diarrhoeal morbidity in a rural
Baltimore, Johns Hopkins University Press, 1976.
community. Environmental health, 12: 155-159(1970).
71.
Schliessmann, D. J. Diarrhoeal disease and the
54. Levine, R. J. et al. Failure of sanitary wells to protect
environment.
Bulletin of the World Health Organiz­
against cholera and other diarrhoea in Bangladesh.
ation, 21: 381-386 (1959).
Lancet, 2: 86-89 (1976).
55. McCabe, L. J. & Haines, T. W. Diarrheal disease 72. Shiffman, M. A. et al. Field studies on water, sani­
control by improved human excreta disposal. Public
tation and health education in relation to health status in
health reports, 72: 921-928 (1957).
Central America. Progress in water, technology, 11:
56. McJunkin, F. E.
Water and human health.
143-150 (1978).
Washington, DC National Demonstration Water
73. Shuval, H. 1. et al. Effect of investments in water
Project, United States Agency for International
supply and sanitation on health status: a threshold­
Development, 1982.
saturation theory. Bulletin of the World Health
Organization, 59: 243-248 (1981).
57. Mackie, T. T. et al. Intestinal parasitic infections in
Forsyth county. North Carolina. IV. Domestic 74. Skoda, J. D. et al. A survey in rural Bangladesh on
diarrhoeal morbidity, water usage and related factors.
environmental sanitation and the prevalence of
Dacca, UNICEF, 1st report, 1977.
Entamoeba histolytica. American journal of tropical
75. Snyder, J. D.&Merson, M. H. The magnitude of the
medicine and hygiene, 5: 29-38 (1956).
global problem of acute diarrhoeal disease: a review of
58. Martorell, R. Acute morbidity and physical growth in
active surveillance data. Bulletin of the World Health
rural Guatemalan children. American journal of
Organization, 60: 605-613 (1982).
diseases of children, 129: 1296-1301 (1975).
59. Mathur, T. N. & Kaur, J. The epidemiology of 76. Sommer, A. & Woodward, W. E. The influence of
protected water supplies on the spread of classical Inaba
amoebiasis in an urban area. Indian journal of medical
and El Tor Ogawa cholera in rural East Bengal. Lancet,
research, 60: 1134-1137 (1972).
2: 985-987 (1972).
60. Meegama, S. A. Socio-economic determinants of
infant and child mortality in Sri Lanka: an analysis of 77. Spira, W. M. et al. Microbiological surveillance of
intra-neighborhood El Tor cholera transmission in rural
post-war experience. World fertility, survey report No.
Bangladesh.
Bulletin of the
World Health
8, 1980.
Organization, 58: 731-740 (1980).
61. Merrick, T. The effect of piped water on early
childhood mortality in urban Brazil, 1970-1976. World
Bank working papers, No. 594, 1983.
44.

772

S. A. ESREY ET AL.

Stewart, W. H. et al. The relationship of certain 84. Watt, J. et al. Diarrheal diseases in Fresno County,
environmental factors to prevalence of Shigella
California. American journal of public health, 43:
infection. IV. Diarrheal disease control studies.
728-741 (1953).
American journal of tropical medicine and hygiene, 4:
85. Weir, J. M. An evaluation of health and sanitation in
718-724(1955).
Egyptian villages. Journal of the Egyptian Public
79. Thacker, S. B. et al. Acute water shortage and health
Health Association, 27: 55-114 (1952).
problems in Haiti. Lancet, 1: 471-473 (1980).
86. White, G. F. et al. Drawers of water. Chicago,
80. Tomkins, A. M. et al. Water supply and nutritional
University of Chicago Press, 1972.
status in rural Northern Nigeria. Transactions of the
87. Wolff, H. L. & Van Zijl, W. J. Houseflies, the
Royal Society of Tropical Medicine and Hygiene, 72:
availability of water, and diarrheal disease. Bulletin of
239-243 (1978).
the World Health Organization, 41: 952-959 (1969).
81. Trivedi, B. K. etal. Bacteriological water quality and 88. Wray, J. D. Direct nutrition intervention and the
incidence of water-borne diseases in a rural population.
control of diarrheal diseases in preschool children.
Indian journal of medical science, 25: 795-801 (1971).
American journal of clinical nutrition, 31: 2073-2082
82. Van Zijl, W. J. Studies on diarrhoeal diseases in seven
(1978).
countries by the WHO diarrhoeal disease advisory
89. Zaheer, M. et al. A note on urban water supply in
team. Bulletin of the World Health Organization, 35:
Uttar Pradesh. Journal of the Indian Medical
249-261 (1966).
Association, 38: 177-182 (1962).
83. Wagner, E. G.&Lanoix, J. N. Watersupply for rural
areas and small communities. Geneva, World Health
Organization, 1959 (Monograph Series. No. 42).

78.

CH 3H
Bulletin oj the B arid Health Organization. 63 (3): 569-583 (1985)

fc> World Health Organization 1985

Interventions for the control of diarrhoeal diseases
among young children: rotavirus and cholera
*
immunization
I. deZoysa1 & R. G. Feachem*12
The potential effects of rotavirus and cholera immunization (with an improved
vaccine) on diarrhoea morbidity and mortality among young children are reviewed using
data from field studies and theoretical calculations. In developing countries rotavirus may
be responsible for about 6% of all diarrhoea episodes and 20% of all diarrhoea deaths in
children under 5 years of age. In industrial countries these proportions may be higher.
Rotavirus immunization may reduce overall diarrhoea morbidity rates by 2-3 % and
diarrhoea mortality rates by 6-10% among children under 5 years of age in developing
countries, depending on vaccine efficacy and programme coverage. The impact of
improved cholera vaccines depends on the prominence of cholera as a cause of diarrhoea,
and this varies greatlyfrom country to country. Taking the extreme example of Bangladesh,
where cholera is endemic and may account for about 0.4% of all diarrhoea episodes and 8 %
of all diarrhoea deaths in children under 5 years ofage, cholera immunization might reduce
overall diarrhoea morbidity rates by 0.06-0.13 % and diarrhoea mortality rates by 1-2%
among these children. The similar incidence rates in industrial and developing countries
suggest that rotavirus diarrhoea may not be controlled by improvements in water supply,
sanitation, or hygiene. Control may depend upon the widespread use of an effective
vaccine.
Over recent years substantial resources have been
invested into research to develop a vaccine against
rotavirus diarrhoea and an improved vaccine against
cholera. Rapid progress has been made and field trials
of candidate vaccines against these diseases are under
way. It is timely, therefore, to examine the potential
role of rotavirus and cholera immunization in
national programmes to reduce diarrhoea morbidity
and mortality among children under 5 years of age. In
this review we do not examine the potential benefits of
rotavirus and cholera immunization that might
extend to older children, adults, or especially
susceptible or at-risk groups, nor do we consider the
potential role of cholera immunization in epidemic
control. Several recent reviews on the epidemiology
and control of rotavirus diarrhoea (41, 83) and
cholera (13, 16, 17, 33, 46) provide a useful back­
ground to this more focused analysis. This paper is
the seventh in a series of reviews of potential anti­
• Requests for reprints should be addressed to the Director,
Diarrhoeal Diseases Control Programme, World Health
Organization, 1211 Geneva 27, Switzerland.

1 Save the Children Fund Research Fellow, Department of
Tropical Hygiene, London School of Hygiene and Tropical
Medicine, Keppel Street, London WC1E 7HT, England.
2 Head, Department of Tropical Hygiene, London School of
Hygiene and Tropical Medicine, London, England.
4555

diarrhoea interventions published in the Bulletin of
the World Health Organization (1, 15, 18-22).

EFFECTIVENESS

For rotavirus or cholera immunization to be an
effective intervention for the control of diarrhoeal
diseases it must be true that:
either

a considerable proportion of
diarrhoea morbidity or mortality
in young children is caused by
rotavirus or Vibrio cholerae 01

hypothesis
1

and
vaccines against rotavirus,
V. cholerae 01, or their products
have the potential to reduce
morbidity rates or mortality rates
or the severity of diarrhoea caused
by these organisms

-569-

hypothesis
2

570

I. DE ZOYSA & R G. FEACHEM

rotavirus or cholera immunization
(when effective vaccines are avail­
able) has the potential to reduce
overall diarrhoea morbidity rates
or mortality rates or the severity
of diarrhoea in young children

diarrhoea. The data in Table 1, as far as possible,
refer only to episodes where rotavirus was the sole
recognized enteric pathogen. In some studies, stools
were not examined for some common pathogens such
as enterotoxigenic Escherichia coli (ETEC) (49) and
Campylobacter jejuni (6, 7, 31, 49, 68).
Recorded incidence rates of rotavirus-associated
diarrhoea ranged from 0.2 to 0.8 episodes per child
per year. Incidence rates were low in children aged 0-5
months, reached a peak in children aged 6-11
months, remained high in children aged 12-23
months, and dropped to low levels thereafter. It is
notable that the incidence rate of rotavirus-associated
diarrhoea found in the study from Winnipeg, Canada
(0.3 episodes per child per year among children aged
0-23 months) (31) was similar to the rate reported
from Bangladesh in the same age group (0.4 episodes
per child per year) (6). The incidence of all
diarrhoeas, however, was lower in Canada and the
proportion of diarrhoea episodes associated with
rotavirus was accordingly higher.
We estimate that, in developing countries,
rotavirus accounts for about 8% of all diarrhoea
episodes in children aged 0-5 months, 10°7o in
children aged 6-23 months, 1% in children aged
24-59 months, and 6% in children under 5 years.
These figures are used in the computations below. In a
particular country, where more reliable age-specific
proportions are available, other figures may be
substituted. It will be noted that the incidence rate of
rotavirus-associated diarrhoea found in different
studies (Table 1) did not vary as much as the incidence

hypothesis
3

The potential effectiveness of rotavirus or cholera
immunization in the control of diarrhoeal diseases
would be suggested by a demonstration either of the
correctness of hypotheses 1 and 2, or of the correct­
ness of hypothesis 3. The evidence for and against the
three hypotheses is examined below.
Hypothesis 1. A considerable proportion of
diarrhoea morbidity or mortality in young children is
caused by rotavirus or V. cholerae Ol.

Rotavirus
Rotavirus-associated morbidity. We have located
in the literature only 7 prospective, community-based
studies that have assessed the importance of rotavirus
diarrhoea among children in the community (Table
1). Four studies (6, 7, 31, 49) report the incidence
rates of rotavirus-associated diarrhoea and rotavirus
isolation rates, and three others (30, 53, 68) give only
the isolation rates. Multiple infections were common
in all the studies: other enteric pathogens were detec­
ted in up to half of all episodes of rotavirus-associated

Table 1. Community-based studies of rotavirus-associated diarrhoea
Country

Bangladesh

Bangladesh

Age
(months)

Number of
person-years

Number of episodes of:

Diarrhoea
from all
causes

Diarrhoea
associated with
rotavirus alone

Proportion of
diarrhoea episodes
associated only
with rotavirus
(%)

Reference

6

0-23

112

377 (3.4)"

43 (0.4)

11

24-59

112

243 (2.2)

0

0

7

2-23

77

497 (6.4)

32 (0.4)

6

24-59

92

444 (4.8)

3 (0.03)

0.7

Brazil

All ages

765

1097 (1.4)

- <0.151“

11

30

Canada

0-23

139

165 (1.2)

40 (0.3)

24

31

El Salvador

0-35

-

-

-

7

68

Guatemala

0-35

132

1050 (7.9)

109 (0.8)

10

49

USA

0-23

-

-

-

10

53

" Figures in parentheses are the number of episodes per person-year.
“ Estimated incidence rate.

rotavirus and cholera immunization for control of diarrhoeal diseases in children

rate of diarrhoea from all causes. The proportion of
all diarrhoea episodes attributable to rotavirus is
likely to be greater than our estimate in settings where
the incidence rate of diarrhoea from all causes is low,
and smaller where the incidence rate of diarrhoea
from all causes is high.
To assess the impact of rotavirus diarrhoea on the
health services, rotavirus isolation rates reported in 77
hospital-based studies from 37 countries have been
reviewed." Caution is necessary in the interpretation
and comparison of these rates. Different diagnostic
methods were used for the detection of rotavirus and
the findings were reported in different age groups.
Many studies did not span an entire year, and large
seasonal fluctuations in rotavirus isolation rates are
well documented. Bearing these factors in mind, we
see that in nearly all the studies rotavirus was the
single most common enteric pathogen identified in
children attending hospital for the treatment of
diarrhoea.
In studies of hospitalized children that lasted at
least one year, the median isolation rate was 34%
(range, 12-71%). The median rate was similar in
studies from developing and industrial countries
(respectively, 35% and 34%; with ranges, 16-71%
and 12-65%). In general, the age-specific isolation
rates were low in infants aged 0-5 months, reached a
peak in infants aged 6-11 months, remained high in
children aged 12-23 months, and dropped rapidly
thereafter. Because a considerable proportion of all
diarrhoea cases that are admitted to hospital occur in
infants, most rotavirus-positive cases were found in
this age group. Indeed, in a number of studies, large
numbers of rotavirus-positive cases were found
among hospitalized children aged 0-5 months,
despite low isolation rates in this age group. Among
the 19 studies that reported these data the proportion
of all rotavirus-positive cases among hospitalized
children aged 0-23 months that occurred in the first 6
months of life ranged from 4% (62) to 73% (57), with
a median of 35%. We have located only 5 studies in
which the data are examined in narrower age-bands.
In these studies, the proportion of all rotavirus­
positive cases among infants (0-11 months) admitted
to hospital for diarrhoea that occurred in the first 3
months of life ranged from 11% to 37%, with a
median of 26% (2, 36, 57, 58, 74).
Studies conducted among children treated for
diarrhoea on an outpatient basis or in short-stay
rehydration units reported lower rotavirus isolation
rates. In studies that lasted at least one year the
median rate found among these patients was 28%
(range, 10-49%). Where outpatients and inpatients
were examined concurrently the rates were found to
" Tabulated data and sources arc available on request from
R.G.F.

571

be lower among outpatients. This finding is difficult
to interpret because the two groups of patients were
not strictly comparable in all studies (S).
Two studies have assessed population-based
hospital case rates. In Washington, DC, USA (66), a
population of about 29 000 children aged 0-14 years
was defined whose primary health care was provided
by a health maintenance organization. Over a period
of 27 months, 38 children from the health mainten­
ance organization were admitted to hospital for the
treatment of diarrhoea. Rotavirus was recovered
from 60% of cases. Yearly population-based
hospitalization
rates
for rotavirus-associated
diarrhoea were 3.7 per thousand in infants and 2.2 per
thousand in children aged 12-23 months, and
dropped to 0.2 per thousand in children aged 24-59
months. Hospitalization for rotavirus-associated
diarrhoea was not observed after 5 years of age. In a
district of Copenhagen County, Denmark (32),
rotavirus was detected among 37% of children aged
0-14 years admitted to the district hospital because
of diarrhoea over a 12-month period. Yearly
population-based hospitalization rates for rotavirusassociated diarrhoea were 5.4 per thousand in infants,
4.1 per thousand in children aged 12-23 months, and
1.4 per thousand in children aged 24-47 months. No
admissions for rotavirus-associated diarrhoea were
observed in children over 4 years of age.
Thus, despite low isolation rates in the community,
rotavirus is responsible for about one-third of
diarrhoea episodes that require hospital admission
among young children, suggesting that diarrhoea
caused by rotavirus is of above-average severity.
Prospective studies of diarrhoea episodes acquired in
the community provide evidence that rotavirusassociated diarrhoea leads more frequently to
dehydration and attendance at a health facility than
all other diarrhoeas in the same age group (6, 7, 49,
81). In these studies, although the incidence of
rotavirus-associated diarrhoea was low, rotavirus was
responsible for about half of all dehydrating
episodes.
Rotavirus-associated
mortality.
Rotavirusassociated diarrhoea may therefore take on severe
forms and is clearly responsible for a proportion of all
deaths due to diarrhoea. Diarrhoea mortality rates
are low in industrial countries, yet fatal episodes of
diarrhoea associated with rotavirus have been
described. Over a 5-year period, 21 deaths associated
with rotavirus were recorded among young children
in Toronto, Canada (10). The severe course of the
diarrhoea was highlighted by the fact that all deaths
occurred within 3 days of onset of symptoms and that
the parents of 16 of the children had had some contact
with a physician during the course of their child’s
illness. Similarly, in the course of an explosive

I. DE ZOYSA & R. G. FEACHEM

572

Table 2. The proportion of diarrhoea deaths in the first five years of life associated with rotavirus in developing
countries, based on various assumptions

Age
(months)

Distribution of
100 diarrhoea
deaths"

Proportion that
are acute watery
diarrhoea deaths'1
(%)

Number of
acute watery
diarrhoea deaths

Proportion of acute
watery diarrhoea deaths
that are associated with
rotavirusc
(%)

Number of
rotavirus-associated
diarrhoea deaths

0-5

29

60

17

20

3.4

6-11

16

60

10

50

5.0

12-23

35

60

21

50

10.5

24-59

20

50

10

10

0-59

100

1.0
19.9

58

° These calculations assume that diarrhoea mortality rates are 20 per 1000 per year in the 0-11 -month age group, 18 per 1000 per
year in the 12-23-month age group, and 4 per 1000 per year in the 24-59-month age group (derived from ref. 67}. Overall mortality
rates vary greatly from country to country. Values adopted here are: infant mortality rate 120 per 1000 live births, mortality rate 40
per 1000 per year in the 12-23-month age group, and 10 per 1000 per year in the 24-59-month age group. It is further assumed that
66% of infant diarrhoea deaths occur in the first 6 months of life (the median figure from 8 studies).
* Estimates based on data from Oberle et al (60) and Chen et al. (72).
c Estimates based on data in the text on the proportion of hospitalized diarrhoea cases and the proportion of dehydrated diarrhoea
cases seen in the community, which were associated with rotavirus.

outbreak of 3439 reported cases of rotavirusassociated diarrhoea in an isolated Pacific island
group (24), 7 children died, all from dehydration due
to diarrhoea and vomiting and all within the first 4
days of illness.
For developing countries, where mortality due to
diarrhoea is considerable (67), we have theoretically
derived the proportion of all diarrhoeal deaths that
are associated with rotavirus. On the basis of the
calculations set out in Table 2, rotavirus may, in dev­
eloping countries, account for about 12% of all
diarrhoea mortality in infants aged 0-5 months, 30%

in children aged 6-23 months, 5% in children aged
24-59 months, and 20% in children under 5 years.
Cholera

The prominence of cholera varies greatly from
country to country. In this review we consider the
extreme case of Bangladesh, where cholera is endemic
and has been studied intensively over a number of
years.
Cholera morbidity in Bangladesh. Table 3 sum­
marizes data from six studies from Dhaka and Matlab

Table 3. Community-based studies of cholera in Bangladesh

Place

Dates of
study

Age
groups

Cholera
incidence rate
(episodes per
1000 per
year)

Proportion of
diarrhoea episodes
associated with
V. cholerae 01
(%)

Dhaka

Sep. 75
to Dec. 76

All ages

3

0.7

Matlab

Nov. 63
to June 66

0-59 months
All ages

10
3

0.7

Matlab

July 69
to June 70

0-11 months
1 2-59 months
All ages

0
9
3





56

Matlab

Dec. 77
to Nov. 78

0-23 months
2-9 years




< 2"
< 3"

6

Matlab

March 78
to March 79

2-59 months

-

0.3

7

Reference

54
50

This proportion refers to alt infections with enteric pathogens other than ETEC. Shigetla and rotavirus, and to mixed infections.

ROTAVIRUS AND CHOLERA IMMUNIZATION FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

573

Table 4. Vibrio cholerae isolation rates among diarrhoea cases in hospitals in Bangladesh
Dates of
study

Type of
patient

Age
groups

V. cholerae 01

Dhaka

July 64
to June 66

Inpatients

All ages

40

48

Dhaka

Dec. 79
to Nov. 80

Outpatients
and inpatients

0-11 months
12-59 months
All ages

1
6
6

69

Matlab

Nov. 63
to June 66

Inpatients

All ages

25

50

Matlab

July 69
to June 70

Outpatients
and inpatients

All ages

27

56

Matlab

Jan. 75
to Dec 75

Inpatients

0-11 months
12-59 months
All ages

1
27
28

60

Matlab

Feb. 77
to Jan. 79

Outpatients
and inpatients

0-23 months
2-9 years
All ages

2
31
13

5

Place

in which surveillance of cholera was conducted in the
community. Three studies recorded cholera incidence
rates, which were very low at 3 episodes per thousand
per year. Rates were highest in children aged 12-59
months and lowest in infants. Overall, the proportion
of all diarrhoea episodes associated with V. cholerae
O1 was less than 1%. For the computations that
follow it will be assumed that, in Bangladesh, cholera
accounts for O.S^o of diarrhoea episodes in children
under 2 years of age, 0.6% in children aged 24-59
months, and 0.4% in children under 5 years of age.
Table 4 lists the V, cholerae O1 isolation rates
reported in hospital-based studies conducted in
Dhaka and Matlab. The proportion of diarrhoea
episodes associated with V. cholerae ranged from 6%
to 40% according to the hospital and the year of
study. We cannot tell from these data whether the
V. cholerae isolation rates were lower in outpatients
than inpatients, but there is evidence to suggest that
patients presenting to hospital with cholera have a
greater risk of moderate or severe dehydration requir­
ing inpatient therapy than other patients (5, 69).
V. cholerae isolation rates were low in the first 2 years
of life and reached a peak in children over 5 years of
age.
Three studies have assessed population-based
hospital case rates. Martin et al. (48) analysed the
admissions to the Dhaka diarrhoeal diseases hospital
between July 1964 and June 1966 and found a
hospitalization rate for classical cholera of 0.4
admissions per thousand population per year with a
peak rate of 0.7 per thousand per year in children
below 5 years of age. The hospitalization rate for

Reference

isolation rates
(%)

cholera (mainly eltor) in the same hospital was 1.7
admissions per thousand population in 1974 and 1.4
in 1975 (44), with the highest rates in children aged
2-9 years. Glass et al. (25) examined the attendance at
Matlab hospital for the 15-year period between
January 1966 and December 1980. In that period,
V. cholerae O1 was isolated from 7141 of the more
than 50 000 patients who presented to the hospital
with diarrhoea. From 1966 to 1973, 97% of the
V. cholerae isolates were of the classical biotype. The
eltor biotype was first identified in the Matlab area in
1969 and was the only biotype present from 1973 to
the end of the study. There was great year-to-year
variation in the number of cases, particularly during
the eltor period. The overall attendance rate for
cholera in the classical period was 1.3 patients per
thousand per year and in the eltor period 2.9 patients
per thousand per year, with the highest rates in
children aged 2-9 years.
Cholera, then, presents an extreme case of a diar­
rhoeal disease that is rare in the community but may
nevertheless place a considerable burden on the health
services. In a study conducted in Matlab (50), cholera
accounted for only 0.7% of all diarrhoea episodes
acquired in the community, yet 25% of all hospital
admissions for treatment of diarrhoea during the
same period were associated with V. cholerae. Other
data from Bangladesh indicate that the proportion of
cholera cases that require hospitalization varies
between 23% and 74% and is higher for classical than
for eltor cholera (3, 42-44, 61, 78, 79). These
findings support the view that cholera is an unusually
severe disease.

574

l.DEZOYSA&R G. FEACHEM

Cholera mortality in Bangladesh. There are no data
on cholera mortality rates in Bangladesh. In the study
areas, deaths due to acute diarrhoea are very rare in
hospital and those that occur in the community have
no etiological diagnosis. For the computations that
follow it will be assumed that, in Bangladesh, cholera
accounts for 5% of diarrhoea deaths in children
under 2 years of age, 20% in children aged 24-59
months, and 8% in children under 5 years of age.
These estimates are based on data summarized in
Table 4 on the proportion of hospitalized diarrhoea
cases that is associated with V.cholerae in different
age groups in Bangladesh.
Hypothesis 2. Vaccines against rotavirus, V.cholerae
Ol, or their products have the potential to reduce
morbidity rales or mortality rates or the severity of
diarrhoea caused by these organisms.

Rotavirus vaccines. Mechanisms of immunity to
rotavirus and advances in rotavirus vaccine develop­
ment have recently been reviewed (41). To date,
research has focused on the development of live
attenuated human rotavirus vaccines, live rotavirus
vaccines from animal hosts, and live attenuated
reassortant vaccines, for delivery by the oral route.
Each of the four recognized human rotavirus sero­
types has now been successfully cultivated (85), and
live attenuated human rotavirus vaccines can be
prepared by conventional tissue-culture methods. A
tissue-culture-adapted mutant of the Wa strain of
human rotavirus (82) is under evaluation for
immunogenicity and safety in susceptible volunteers
(39,40). Cold-adapted strains of human rotavirus are
also under investigation as attenuated human
rotavirus vaccines (Kono, personal communication,
1984). In addition, strains obtained from
asymptomatically-infected neonates are under study
as naturally attenuated strains (Bishop, personal
communication, 1984). These strains are promising
candidate vaccines in view of the evidence that
immunity induced by neonatal infection prevents the
development of clinically severe illness for at least the
first 3 years of life (although it does not protect
against infection) (4).
Another approach is to prepare vaccine material
from animal rotavirus strains that are antigenically
related to human rotavirus (35, 38, 84). Bovine rota­
viruses grow well in tissue culture (51) and have been
used to prepare candidate vaccines. The feasibility of
this approach was demonstrated in experimental
studies in animals (80, 86, 88). A rotavirus vaccine of
bovine origin (strain RIT 4237) was shown to be
immunogenic and safe in young children (75) and a
field trial in a group of Finnish children aged 8-11
months using one dose of vaccine of high titre
(1O8 iTCIDso) showed a protective efficacy of 50%

for all rotavirus-associated diarrhoea and of 88% for
rotavirus-associated diarrhoea lasting more than 24
hours (76). In a subsequent trial in which 2 doses of
vaccine were given one month apart to Finnish
children aged 6-12 months, vaccine efficacy was
found to be 58% for all rotavirus-associated diar­
rhoea and 82% for rotavirus-associated diarrhoea
that was clinically significant (Vesikari, persona!
communication, 1984). Infants who seroconverted
appeared to have the highest level of protection, but
protection was also observed in infants who did not
seroconvert. Seroconversion rates may be improved
by giving a milk feed (infant formula or diluted cow’s
milk) immediately before and after vaccination (77).
Further field trials with this vaccine are under way in
Peru and the Gambia. Another promising candidate
vaccine is a rhesus rotavirus strain (70, 84) which has
been tested for immunogenicity and safety in human
volunteers and is now under field trial in the USA.
Finally, reassortant viruses have been recovered from
mixed tissue-culture infection (27-29), opening the
way to the development of a live attenuated
reassortant rotavirus vaccine.
In conclusion, good progress has been made in the
development of candidate rotavirus vaccines using
different approaches. Further candidate vaccines
may be developed using recombinant DNA tech­
nology. Field trials of rotavirus vaccines of bovine
and simian origin are under way. These vaccines
appear to prevent or modify rotavirus-associated
illness. Studies in progress will assess the duration of
protection induced by these vaccines and their strain
specificity. The optimal age of vaccination must be
determined, bearing in mind the need to vaccinate
infants in the first 6 months of life, and the possible
interference of breast-feeding or of residual maternal
antibodies on seroconversion rates. Information is
also needed on the number of doses required in order
to determine the most appropriate vaccination
schedule, with regard to production and delivery costs
and eventual combination with oral polio vaccine.
Other issues to be addressed concern the stability of
the rotavirus vaccine, its cold-chain requirements,
and its possible interaction with oral polio vaccine.
Cholera vaccines. Advances in cholera vaccine
development have recently been reviewed (46). The
ideal antigenic composition of an effective oral
cholera vaccine is not at present known (13) and work
is under way to identify the major protective somatic
and toxin-derived antigens. The only vaccines
currently available for general use are killed whole­
cell vacci'nes for parenteral administration. Field
trials with these vaccines have established that
induced protection is moderate (50-70%) and of
short duration (3-6 months), depending on the age
group, quality of the vaccines, and dosage schedule

ROTAVIRUS AND CHOLERA IMMUNIZATION TOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

(23). Current research efforts are centred on the
development of a cholera vaccine for oral use, in
order to stimulate local intestinal immunity and to
avoid a possible suppressive effect of parenteral
immunization on local antigenic stimulation (64, 87).
Whole-cell vaccines taken orally afford some (56%)
degree of protection against experimental cholera and
reduce the severity of diarrhoea in ill volunteers (//;
Levine, personal communication, 1984). A whole-cell
vaccine for oral use is currently under field trial in
Bangladesh. Another killed oral vaccine has been pre­
pared by ultrafiltration of the culture supernatant of
two eltor strains of y. cholerae OI. Results from a
field trial in Zaire suggest that this vaccine may have
protective properties against natural disease.
*
The immunogenicity and efficacy of toxin
derivatives have also been evaluated. Field trials of
parenterally administered toxoids, consisting of
purified cholera toxin inactivated with either
formaldehyde or glutaraldehyde, have demonstrated
slight or no protection against cholera (14, 59). In
volunteers, large, multiple oral doses of purified
toxoid are safe and immunogenic but have failed to
provide protection against experimental challenge
(45). Procholeragenoid, a heat-induced aggregate of
cholera toxin (46) and the purified B subunit of
cholera toxin (26, 72) are safe and immunogenic in
volunteers but their protective effects have not yet
been reported.
Animal studies have shown that antibacterial and
antitoxic intestinal antibodies are synergistically
protective (63, 65 , 71), suggesting that a vaccine
prepared from a combination of somatic and toxinderived antigens holds promise. One approach is to
develop attenuated K. cholerae strains for use as oral
vaccines. Studies have been carried out on Texas StarSR, an attenuated strain derived by chemical
mutagenesis from an eltor Ogawa organism (34).
Texas Star-SR produces the B subunit of cholera
toxin but no detectable A subunit. In US volunteers it
was shown to provide a moderate (61%) degree of
protection against experimental challenge with
V'.cholerae and to reduce the stool volume in ill
vaccinees (47). Unfortunately, 24% of vaccinees had
mild or moderate diarrhoea. Of further concern is the
fact that the precise genetic lesion responsible for the
inability of the strain to elaborate cholera toxin is
unknown, and reversion is theoretically possible.
Recombinant DNA techniques have been applied
to develop attenuated K cholerae oral vaccine strains
incapable of genetic reversion (37, 52). Precise
deletions of the genes encoding both the A and B
subunits of cholera toxin have produced a strain
b Bwanga, M. [First controlled trials of oral anticholera
vaccine during a cholera epidemic in the zone of Matemba — Nkulu
(Shaba-Zaire).] Bulletin de la Social# de Pathologie dxotique, 77:
13-16 (1984) (in French).

575

(JBK-70) that protects volunteers from subsequent
challenge to a degree (89%) similar to recovery from
the disease (46). This strain, however, was also
associated with diarrhoea in an unacceptable propor­
tion of recipients. Research is under way to identify
the factors responsible for the observed side-effects.
Another approach is to prepare an oral killed
vaccine containing a combination of somatic and
toxin-derived antigens. Studies in volunteers of
vaccines prepared from killed whole-cell vibrio com­
bined with procholeragenoid or glutaraldehydetreated toxoid have demonstrated complete safety
and protective efficacy ratesof27% (with procholera­
genoid) and 67% (with toxoid) (46). A combined B
subunit and killed whole-cell oral vaccine has been
shown to be safe in volunteers in Bangladesh and
Sweden, and capable of inducing in Bangladeshi
volunteers a local immunological response com­
parable to that evoked by the disease (73).
Subsequent challenge studies among US volunteers
demonstrated a vaccine efficacy of 64% and complete
protection against severe disease (46). Field trials of
the combined B subunit/whole-cell vaccine and
whole-cell vaccine alone are in progress in Bangladesh
to determine the protective efficacy of the vaccines
and the duration of protection in an endemic area. A
number of issues remain unresolved, as with rotavirus
vaccine, concerning especially the influence of age,
breast-feeding practices, and residual maternal anti­
bodies on protection.

Hypothesis 3. Rotavirus or cholera immunization
(when effective vaccines are available) has the
potential to reduce overall diarrhoea morbidity rates
or mortality rates or the severity of diarrhoea in
young children.
At present, the only approach to assessing hypothe­
sis 3 is a theoretical one using information computed
during the assessment of hypotheses 1 and 2.

Rotavirus immunization. The potential impact of
rotavirus immunization on overall diarrhoea rates in
children under 5 years will depend upon the age of
immunization, vaccine efficacy, and programme
coverage. The recommended vaccination schedule
will be determined when data are obtained on dose
requirements, the effect of breast-feeding and
residual maternal antibodies, and the effect of inter­
action with antigens delivered in the context of the
Expanded Programme on Immunization (EPI). The
goal of rotavirus immunization programmes will be
to vaccinate children as early as possible, but
operational and immunological factors may prevent
the achievement of full immunization before 6
months of age. Field trials of the RIT 4237 rotavirus
vaccine in Finland have shown efficacies of 50-58%

576

I. DU ZOYSA & R. G. FEACHEM

Table 5. Maximum impact of rotavirus immunization on diarrhoea morbidity and mortality rates among children
under 5 years of age in developing countries, assuming 100% vaccine efficacy, 100% programme coverage, and an
average age of full immunization of 6 months'7
Age
(months)

Proportion of diarrhoea episodes
Caused by
rotavirus"
(%)

Proportion of diarrhoea deaths

Averted by
rotavirus
immunization
(%)

Caused by
*
rotavirus
(%)

Averted by
rotavirus
immunization
(%)

0 5

8

0'1

12

0"

6-23

10

10

30

30

24 59

1

1

5

5

0-59

6

5'-

20

16'

" The computed proportions of episodes and deaths averted are directly proportional to the vaccine efficacy and the programme
coverage, and thus the effects of different values for these parameters may be readily computed.
h See text.
' See Table 2.
'' Average age of full immunization assumed to be 6 months.
*‘ These calculations assume that diarrhoea morbidity rates are 3 per child per year in the 0-5-month age group, 4 per child per
year in the 6-11-month age group. 4 per child per year in the 12-17-month age group, 3 per child per year in the 18- 23-month age
group, and 2 per child per year in the 24-59-month age group (derived from ref. 67).
' Based on age-specific diarrhoea mortality rates given in footnote a to Table 2.

against all rotavirus-associated diarrhoea and
82-88% against rotavirus-associated diarrhoea of
clinical significance in children aged 6-12 months (76;
Vesikari, personal communication, 1984). In an
ongoing immunization programme aimed at younger
children the efficacy rates may be lower. An efficacy
of rotavirus vaccine of 80% is assumed here. The
proportion of children who would receive the vaccine
depends on programme coverage. Three coverage
figures are adopted here (45%, 60% and 75%),
bearing in mind that these figures are not now being
achieved in countries with limited immunization
programmes, but may be expected in countries
committed to the EPI and with a national programme
built up over several years.
The proportions of all diarrhoea cases and deaths
averted by rotavirus immunization are directly pro­
portional to both vaccine efficacy and programme
coverage, but not to age of immunization. Table 5
presents computations of the maximum impact of
rotavirus immunization at 6 months of age on overall
diarrhoea morbidity and mortality rates in developing
countries, assuming 100% efficacy and 100% pro­
gramme coverage. Under these ideal conditions, rota­
virus immunization might reduce diarrhoea
morbidity rates by 5% and diarrhoea mortality rates
by 16% among children under 5 years of age. With a
vaccine having 80% efficacy and levels of coverage of
45%, 60% and 75%, rotavirus immunization might
reduce diarrhoea morbidity rates, respectively, by

1.8%, 2.4% and 3.0% and diarrhoea mortality rates
by 6%, 8% and 10%. If immunization against
rotavirus is completed at a younger age, say 3 months,
the potential reductions in diarrhoea morbidity and
mortality rates may only be marginally greater:
1.9-3.3% for morbidity and 6-11% for mortality
(assuming that half of the diarrhoea episodes and
deaths that occur in the first 6 months of life occur in
the first 3 months of life; calculations not shown).
The proportional impact of rotavirus immunization
on diarrhoea morbidity rates among young children
in industrial countries is likely to be greater because
the proportion of diarrhoea episodes attributable to
rotavirus in those countries is higher. The difference
may not be marked for reductions in mortality rates,
however, judging from the similar isolation rates for
rotavirus found among young children admitted to
hospital for severe diarrhoea in developing and
industrial countries.
Two major hospital-based studies have been
examined to assess the potential reduction in hospital
reporting rates for diarrhoea that might be achieved
by a rotavirus immunization programme. Data from
a one-year study of patients reporting to the
diarrhoeal diseases hospital in Dhaka, Bangladesh
(69) suggest that rotavirus immunization might have
averted 7-12% of attendances for diarrhoea to the
hospital among children under 5 years of age. Data
from an 8-year study of inpatients in a children’s
hospital in Washington, DC, USA (9) suggest that

ROTAVIRUS AND CHOLERA IMMUNIZATION FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

577

Table 6. Maximum impact of cholera immunization on diarrhoea morbidity and mortality rates among children under
5 years of age in Bangladesh, assuming 100% vaccine efficacy, 100% programme coverage, and an average age of
full immunization of 2 years3
Age
(months)

Proportion of diarrhoea episodes

Proportion of diarrhoea deaths

Caused by
V. cholerae 01 h
(%)

Averted by
cholera
immunization
(%)

Caused by

V. cholerae 01f
(%)

Averted by
cholera
immunization
(%)

0-23

0.2

0"

5

0d

24-59

0.6

0.6

20

20

0-59

0.4

0.3'

8

4'

“ The computed proportions of episodes and deaths averted are directly proportional to the vaccine efficacy and the programme
coverage and thus the effects of different values for these parameters may be readily computed.
b See text.
1 See text.
3 Average age of full immunization assumed to be 2 years.
' Based on age-specific diarrhoea morbidity rates given in footnote e to Table 5.
' Based on age-specific diarrhoea mortality rates given in footnote a to Table 2.

rotavirus immunization might have averted 9-14% of
admissions for diarrhoea over that period. These
calculations assume a vaccine efficacy of 80%,
coverage of 45-75%, and an average age of full
immunization of 6 months.
Cholera immunization. In Bangladesh, cholera
may account for about 0.4% of all diarrhoea
morbidity and 8% of all diarrhoea mortality in
children under 5 years of age. Again, values for the
efficacy of a new cholera vaccine have to be assumed.
A protective efficacy of 64% has been found in
challenge studies among adult US volunteers
receiving 3 oral doses of combined B subunit/wholecell vaccine (46). A value for the efficacy of a new
cholera vaccine of 70% is assumed here. The vac­
cination schedule for this vaccine is unknown at
present. If a new vaccine does not give long-lasting
protection, the optimal age of administration may be
around 24 months, before the peak in age-specific
cholera incidence rate. In this case it will not be
delivered within the existing EPI and so coverage may
be low. Three coverage figures are adopted here:
30%, 45% and 60%.
Table 6 presents computations of the maximum
impact of cholera immunization at 2 years of age on
overall diarrhoea morbidity and mortality rates in
Bangladesh, assuming 100% vaccine efficacy and
100% programme coverage. Under these ideal condi­
tions, cholera immunization might reduce overall
diarrhoea morbidity rates by 0.3% and diarrhoea
mortality rates by 4% among children under 5 years
of age in Bangladesh. With a vaccine having 70%
efficacy and levels of coverage of 30%, 45% and

60%, cholera immunization at 2 years of age might
reduce diarrhoea morbidity rates by 0.06%, 0.09%
and 0.13% and diarrhoea mortality rates by 0.8%,
1.3% and 1.7%. If immunization against cholera is
achieved at a younger age, say 6 months, and
accordingly better coverage figures are adopted
(45%, 60% and 75%), the potential reductions would
be 0.11-0.19% for diarrhoea morbidity rates and
2.1-3.4% for diarrhoea mortality rates (calculations
not shown).
In this review, we do not examine the benefits of
cholera immunization that might extend to older
children and adults in endemic areas if immunity
induced by an improved vaccine is long-lasting, nor
do we consider the potential role of cholera immuniz­
ation in epidemic control' and in the protection of
especially susceptible or at-risk groups. Finally, there
are insufficient data to assess the exciting possibility
that a vaccine prepared from antigens derived from
cholera toxin may induce cross-protection against
ETEC (LT-only or LT-ST) diarrhoea.
Cholera immunization may not have a measurable
impact on hospital attendance rates for diarrhoea
among young children. Data from the hospital-based
study in Dhaka cited above (69) suggest that cholera
immunization might have averted only 0.6-1.2% of
attendances for diarrhoea among children under 5
years of age. Among older patients these potential
reductions in attendances are 2-4%, assuming
repeated vaccinations or long-lasting immunity.
c The role of immunization in the prevention and control of
cholera epidemics will be considered in another review (Blake and
Feachem, under preparation).

I. DE ZOYSA & R G. FEACHEM

578

These calculations assume a vaccine efficacy of 70%,
coverage of 30-60%, and an average age of full
immunization of 2 years.
FEASIBILITY

The delivery requirements of the candidate rota­
virus and cholera vaccines are at present unknown.
The simultaneous administration of rotavirus vaccine
with another vaccine currently included in the EPI
would facilitate its delivery. Three doses of oral polio
vaccine are usually recommended in the first 6 months
of life and it is hoped that rotavirus vaccine can be
combined with one or all 3 doses of oral polio vaccine
without interference with seroconversion to either
virus. The cold-chain and handling requirements of
rotavirus and oral polio vaccines are likely to be
similar. In this optimal case, rotavirus immunization
would require few additional inputs for its delivery.
The delivery requirements of cholera vaccine are
more speculative. A killed cholera vaccine may prove
to be relatively temperature-stable, whereas a live
cholera vaccine may have stringent cold-chain
requirements. Cholera immunization with an
improved vaccine may be indicated in the second year
of life and may involve multiple doses. Cholera
vaccine would not, in this case, be delivered within the
EPl, but would require additional immunization
services. The main operational difficulty would be
achieving high coverage of children in this older age
group. If a new cholera vaccine can induce longlasting protection, its delivery will be greatly simpli­
fied.
Finally, a major constraint on the successful
delivery of oral vaccines against both rotavirus
diarrhoea and cholera would be the need to protect
the vaccine against gastric acidity, in order to assure
its safe passage into the small intestine.

COSTS

Rotavirus and improved cholera vaccines are still
under development, so the cost estimates must be
derived from data on current immunization pro­
grammes. In a recent review, Creese (under prep­
aration) has estimated the likely costs of delivering
these new vaccines on the basis of data from 9 cost­
appraisal studies of immunization programmes in
developing countries. Costs depend heavily on the
extent to which the new vaccines can be incorporated
into an existing immunization programme. The
actual vaccine costs for current EPI vaccines account
for only a small proportion (around 10%) of total
programme costs.

Assuming an optimal delivery strategy for rota­
virus immunization, in which a single dose of
rotavirus vaccine is administered within an existing
EPI programme without increased frequency of
contacts, the likely marginal cost is USS 2 (1982
prices) per fully immunized child. If cholera
immunization requires additional contacts because it
is targetted at a different age group than the EPI,
estimated costs are from USS 5 (if a single dose is
required) to USS 15 (if three doses are required) per
child fully immunized against cholera (1982 prices). It
is here assumed that the cost per vaccinated child is
independent of the number of children vaccinated,
within likely limits.
These cost estimates can be brought together with
the effectiveness data computed above in order to
calculate the likely cost-effectiveness range of the
proposed interventions in developing countries. Two
indicators of cost-effectiveness are considered here:
cost per case averted and cost per death averted/
For rotavirus immunization at age 6 months with a
vaccine having 80% efficacy and a marginal cost of
USS 2 per fully immunized child, the cost-effective­
ness figures are:

— USS 4 (1982 prices) per diarrhoea case averted in a
child aged 0-59 months;
— USS 312 (1982 prices) per diarrhoea death averted
in a child aged 0-59 months.
For cholera immunization at age 2 years in
Bangladesh with a vaccine having 70% efficacy and a
cost of USS 5-15 per fully immunized child, the cost­
effectiveness figures are:
— USS 183-549 (1982 prices) per diarrhoea case
averted in a child aged 0-59 months;
— USS3571-10 714(1982 prices) per diarrhoea death
averted in a child aged 0-59 months.
If cholera immunization is achieved at 6 months of
age at the marginal cost of USS 2-6 per fully
immunized child, the likely cost-effectiveness of the
intervention is greatly improved (cost per case averted
USS 61-183 and cost per death averted USS
872-2617).

CONCLUSIONS

In developing countries, rotavirus may be respon­
sible for about 6% of all diarrhoea episodes and 20%
of all diarrhoea deaths in children under 5 years of
age. In industrial countries these proportions may be
higher. A theoretical case has been made, based on
various assumptions about vaccine efficacy and
d The formulae used to calculate these cost-effectiveness figures
are available on request from R.G.F.

ROTAVIRUS AND CHOLERA IMMUNIZATION FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

programme coverage, thai rotavirus immunization
might reduce overall diarrhoea morbidity by 2-3%
and diarrhoea mortality by 6-10% among children
under 5 years of age in developing countries.
The impact of improved cholera vaccines depends
on the prominence of cholera as a cause of diarrhoea
and this varies greatly from country to country.
Taking the extreme example of Bangladesh, where
cholera may account for about 0.4% of all diarrhoea
episodes and 8% of all diarrhoea deaths in children
under 5 years of age, cholera immunization might
reduce overall diarrhoea morbidity by 0.06-0.13%
and diarrhoea mortality by 1-2% among these
children, based on various assumptions.
These vaccines are under development and vac­
cination schedules and delivery requirements are at
present unknown. If an immunization programme
can incorporate rotavirus vaccine at a marginal cost
of USS 2 (1982 prices) per fully immunized child, the
likely cost-effectiveness values for the intervention
are USS 4 per diarrhoea case averted and USS 312 per
diarrhoea death averted in children under 5 years. For
cholera immunization which may require additional
immunization services at a cost of USS 15 (1982
prices) per fully immunized child, the likely cost­
effectiveness values are USS 549 per diarrhoea case
averted and USS 10 714 per diarrhoea death averted

579

in children under 5 years. These estimates are
extremely tentative and are subject to revision as more
is learnt about the epidemiology of rotavirus
diarrhoea and cholera, and the features of
immunization programmes against these diseases. In
areas where more accurate epidemiological and
economic data are available other estimates of cost­
effectiveness can be computed.
The prominence of rotavirus diarrhoea in
industrial countries, and the similar incidence rate in
industrial and developing countries, suggest that rota­
virus diarrhoea may not be controlled by improve­
ments in water supply, sanitation, or hygiene.
Control may depend upon the development, trial, and
widespread use of an effective vaccine. It is hoped
that rotavirus vaccine can be delivered within existing
EPI programmes, and rotavirus immunization may
prove to be a cost-effective intervention for national
diarrhoeal diseases control programmes.
The cost-effectiveness of cholera immunization, as
an intervention to reduce overall diarrhoea morbidity
and mortality rates among young children, is more
doubtful, even in Bangladesh, although it may have
other applications. More data are required on vaccine
efficacy and the duration of protection before firmer
conclusions can be reached.

ACKNOWLEDGEMENTS
We are grateful for the constructive criticisms on earlier drafts of this paper provided by R. F. Bishop, R. E. Black,
D. Blum, R. I. Glass, R. H. Henderson, R. C. Hogan, J. Holmgren, A. Z. Kapikian, M. H. Merson, N. F. Pierce, and
D. A. Sack. We thank J. Seaman of the Save the Children Fund, United Kingdom, for his support and encouragement.
Secretarial, bibliographical and editorial assistance was most ably provided by Alison Hinchley, Susanne O’Driscoll and
Dianne Fishman.

RESUME
INTERVENTIONS DANS LE CADRE DE LA LUTTE CONTRE LES MALADIES DIARRHEIQUES DU JEUNE ENFANT:
VACCINATION ANTICHOL^RIQUE ET VACCINATION CONTRE LES ROTAVIRUS

Cet article est le septieme d’une serie d’etudes concernant
les interventions susceptibles de reduire la morbiditc et la
mortalite imputables aux maladies diarrheiques chez les
enfants de moins de cinq ans dans les pays en d^veloppcment. Nous avons etudie les effets potentiels de la vacci­
nation contre les rotavirus et de la vaccination anticholerique (a I’aide d’un vaccin ameliore) sur la morbiditc et
la mortalite diarrheiques, en utilisant des donnees provenant
d’enquetes de terrain et de calculs theoriques. Dans les pays
en developpement, les rotavirus sont probablement responsables d’environ 6% de tous les episodes morbidcs et de 20%
de Pensemble des deces d’origine diarrheique chez les

enfants de moins de cinq ans. Ces proportions sont peut etre
plus elevees dans les pays industriels. Les effets potentiels
d’une immunisation a 1’egard des rotavirus sur le taux
gdndral d’atteinte diarrheique chez les enfants de moins de
cinq ans dependra de Page auquel se pratique la vaccination,
de Pefficacite du vaccin utilise et de la couverture assuree
par le programme. Avec un vaccin efficacc i 80% et un taux
decouverture de Pordre de45 & 75%, une vaccination contre
les rotavirus pratiquee a Page de six mois peut diminuer les
taux de mortalite diarrheique de 1,8 a 3,0% et les taux de
mortality de 6 & 10%. Si cette vaccination est pratiqude i un
age moins avance, par exemple trois mois, le gain sur cette

5S0

1. DE ZOYSA & R. G. EEACHEM

baisse des taux peut n’avoir qu’une importance marginale:
1.9-3,3% pour la morbidite et 6-11% pour la monalite
Les effets d’une amelioration des vaccins anticholeriques
dependent du role preponderant du cholera cn tant que
cause des maladies diarrheiques— role qui varie considerablement d’un pays a (’autre. Au Bangladesh, le cholera est
probablement responsable d’environ 0,4% de (’ensemble de
la morbidite diarrheique et de 8% de la totalite des deces
d’enfants de moins de cinq ans imputables a ces maladies.
Dans le cas particulier, on a suppose que le nouveau vaccin
anticholerique etait efficace a 70%. Le schema d’adminis­
tration utilise pour ce nouveau vaccin est actuellement
inconnu. Lorsqu'un nouveau vaccin ne confere pas de
protection durable, 1’age optimal pour son administration sc
situe aux alentours de 24 mois, e’est-a-dire avant
(’apparition du pic d’incidence cholerique specifiquc pour
cette classe d’age. En pareil cas, le vaccin ne sera pas administre dans le cadre du PEV existant et, de la sorte, la
couverture sera peut-etre faible. Avec un vaccin efficace a
70% et un taux de couverture de I’ordre de 30 a 60%, la
vaccination anticholerique pratiquee a Lage de deux ans au
Bangladesh pourrail diminucr les taux de morbidite diar­
rheique de 0,06 a 0,13% et les taux de monalite de 0,8 a
1.7%. Si la vaccination anticholerique est pratiquee a un age
moins avance, par exemple six mois, et si par consequent
une meilleure couverture pent etre assuree (45-75%), la
reduction potentielle serait de I’ordre de 0,11 a 0,19% pour
la morbidite diarrheique et de 2,1 a 3,4% pour la mortalite.
S’il est possible d’englober dans un programme de vacci­
nation une immunisation a 1’egard des rotavirus moyennant
un cout marginal de USS2 (prix de 1982) pour la vaccination

complete de chaque enfant, le rapport cout-efficacite de
cette intervention sera de I’ordre de USS4 pour chaquc cas
de maladie diarrheique qu’il a etc ainsi possible de prevenir,
et de USS312 par deces evite chez les enfants de moins de
cinq ans. Pour une vaccination anticholerique qui peut
exiger un renforcemcnt des services de vaccination d’un cout
s’elevant a USS 15 (prix de 1982) pour la vaccination
complete de chaque enfant, le rapport cout-efficacite au
Bangladesh sera vraiscmblablemcnt de USS549 pour chaque
cas de maladies diarrheiques qu’il sera possible de prevenir
et de USS10 714 pour chaquc deces evite chez les enfants de
moins de cinq ans.
L’importance des maladies diarrheiques a rotavirus dans
les pays industriels, et (’analogic de leur taux d’incidence
ent re pays indust riels et pays en developpement, suggerent
qu’une amelioration des approvisionnements en cau, de
I’assainisscment ou de (’hygiene ne suffit peut-etre pas pour
juguler les affections de cc type. Leur maitrise depend sans
doute de (’elaboration, de 1’essai et de (’utilisation generalisee d’un vaccin efficace. II est permis de penser qu’un
vaccin contrc les rotavirus pourrait etre administre dans le
cadre du Programme elargi de Vaccination et que cette
intervention se revelera interessante, sur le plan du rapport
cout-efficacite, pour les programmes nationaux de luttc
centre les maladies diarrheiques. La vaccination anti­
cholerique, en revanche, peut exiger un renforcement des
services de vaccination entrainant des couts eleves pour un
faible taux de couverture. Meme au Bangladesh, le rapport
cout-efficacite des interventions en faveur d’une diminution
de la morbidite ci de la monalite diarrheiques chez les jeunes
enfants ne sera peut-etre pas positif, bien que d’autres
applications puissent etre envisagees dans ce cadre.

REFERENCES
Ashworth, A. & Feachem, R. G. Interventions for the 7. Black, R. E. et al. Longitudinal studies of infectious
diseases and physical growth of children in rural
control of diarrhoeal diseases among young children:
Bangladesh. 2. Incidence of diarrhea and association
prevention of low birthweight. Bulletin of the World
with known pathogens. American journal of
Health Organization, 63: 165-184 (1985).
epidemiology,
115: 315-324 (1982).
2. Avenda-no, L. F. et al. [Rotavirus infection in breast­
8. Brandt, C. D. et al. Comparative epidemiology of
fed infants with acute diarrhoea.] Revista medica de
two rotavirus serotypes and other viral agents associated
Chile, 111: 240-246 (1983) (in Spanish).
with pediatric gastroenteritis. American journal of
3. Bart, K. J. et al. Seroepidemiologic studies during a
epidemiology, 110: 243-254 (1979).
simultaneous epidemic of infection with El Tor Ogawa
9. Brandt, C. D. et al. Pediatric viral gastroenteritis
and classical Inaba Vibrio cholerae. Journal of
during eight years of study. Journal of clinical micro­
infectious diseases, 121: S17-S24 (1970).
biology, 18: 71-78 (1983).
4. Bishop, R. F. et al. Clinical immunity after neonatal
rotavirus infection. A prospective longitudinal study in 10. Carlson, J. A. K. et al. Fatal rotavirus gastro­
enteritis: an analysis of 21 cases. American journal of
young children. New Englandjournal of medicine, 309:
diseases of children, 132: 477-479 (1978).
72-76 (1983).
5. Black, R. E.etal. A two-year study of bacterial, viral, 11. Cash, R. A. et al. Response of man to infection with
Vibrio cholerae. 2. Protection from illness afforded by
and parasitic agents associated with diarrhea in rural
previous disease and vaccine. Journal of infectious
Bangladesh. Journal of infectious diseases, 142:.
diseases, 130: 325-333 (1974).
660-664 (1980).
6. Black, R. E. et al. Incidence and severity of rotavirus 12. Chen, L. C. et al. Epidemiology and causes of death
among children in a rural area of Bangladesh.
and Escherichia coli diarrhoea in rural Bangladesh:
International journal of epidemiology, 9: 25-33 (1980).
implications for vaccine development. Lancet, 1:
141-143 (1981).
1.

ROTAVIRUS AND CHOLERA IMMUNIZATION FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

13.

581

Craig, J. P. Prospects for the development of 29. Greenberg, H. B. et al. Use of reassortant rotaviruses
improved cholera vaccines. In: Holme, T. et al., ed.,
and monoclonal antibodies to make gene-coding assign­
Acute enteric infections in children. New prospects for
ments and construct rotavirus vaccine candidates. In:
treatment and prevention. Amsterdam, Elsevier/NorthChanock, R. M. & Lerner, R. A., ed., Modern
Holland Biomedical Press, 1981, pp. 461-475.
approaches to vaccines: molecular and chemical basis of
14. Curlin, G. et al. Field trial of cholera toxoid. In:
virus virulence and immunity. New York, Cold Spring
Proceedings of the 11 th Joint Conference on Cholera,
Harbor Laboratory, 1984, pp. 319-327.
New Orleans, November 1975. Bethesda, MD, National
30. Guerrant, R. L. etal. Prospective study of diarrheal
Institutes of Health, 1976. pp. 314-329.
illnesses in northeastern Brazil: patterns of disease,
15. de Zoysa, 1. & Feachem, R. G. Interventions for the
nutritional impact, etiologies, and risk factors. Journal
control of diarrhoeal diseases among young children:
of infectious diseases, 148: 986-997 (1983).
chemoprophylaxis. Bulletin of the World Health
31. Gurwith, M. et al. A prospective study of rotavirus
Organization, 63: 295-315 (1985).
infection in infants and young children. Journal of
16. Feachem, R. G. Environmental aspects of cholera
infectious diseases, 144: 218-224 (1981).
epidemiology. I. A review of selected reports of endemic
32. Hjelt, K. et al. Incidence of hospitalisation and out­
and epidemic situations during 1961-1980. Tropical
patient clinical visits caused by rotavirus and non­
diseases bulletin, 78: 675-698 (1981).
rotavirus acute gastroenteritis. A study of children
17. Feachem. R. G. Environmental aspects of cholera
living in the southern district of Copenhagen county.
epidemiology. III. Transmission and control. Tropical
Danish medical bulletin, 31: 249-251 (1984).
diseases bulletin, 79: 1-47 (1982).
33. Holmgren, J.&Svennerholm, A. M. Cholera and the
18. Feachem, R. G. Interventions for the control of
immune response. Progress in allergy, 33: 106-119
diarrhoeal diseases among young children: supple­
.
(1983)
mentary feeding programmes. Bulletin of the World
34. Honda, T. & Finkelstein, R. A. Selection and charac­
Health Organization, 61: 967-979 (1983).
teristics of a Vibrio cholerae mutant lacking the A
19. Feachem, R. G. Interventions for the control of diar­
(ADP-ribosylating) portion of the cholera enterotoxin.
rhoeal diseases among young children: promotion of
Proceedings of the National Academy of Sciences of the
personal and domestic hygiene. Bulletin of the World
USA, 76: 2052-2056 (1979).
Health Organization, 62: 467-476 (1984).
35. Hoshino, Y. etal. Serotypic similarity and diversity of
20. Feachem, R. G. & Koblinsky, M. A. Interventions for
rotaviruses of mammalian and avian origin as studied
the control of diarrhoeal diseases among young
by plaque-reduction neutralization. Journal of
children: measles immunization. Bulletin of the World
infectious diseases, 149: 694-702 (1984).
Health Organization, 61: 641-652 (1983).
36. Jayavasu, C. et al. A long-term study of rotavirus
21. Feachem, R. G. & Koblinsky, M. A. Interventions for
infection in Thai infants and children with diarrhoea.
the control of diarrhoeal diseases among young
Southeast Asian journal of tropical medicine and public
children: promotion of breast-feeding. Bulletin of the
health, 13: 373-376 (1982).
World Health Organization, 62: 271-291 (1984).
37. Kaper, J. B. et al. Recombinant nontoxinogenic
22. Feachem, R. G. et al. Diarrhoeal disease control:
Vibrio cholerae strains as attenuated cholera vaccine
reviews of potential interventions. Bulletin of the World
candidates. Nature, 308: 655-658 (1984).
Health Organization, 61: 637-640 (1983).
38. Kapikian, A. Z. et al. Antigenic characterization of
23. Feeley, J. C. & Gangarosa, E. J. Field trials of
human and animal rotaviruses by immune adherence
cholera vaccine. In: Ouchterlony, O. & Holmgren, J.,
hemagglutination assay (1AHA): evidence for distinct­
ed., Cholera and related diarrheas, 43rd Nobel
ness of IAHA and neutralization antigens. Infection
Symposium, Stockholm, 6-11 August 1978. Basel,
and immunity, 33: 415-425 (1981).
S. Karger, 1980, pp. 204-210.
39. Kapikian, A. Z. et al. Oral administration of human
24. Foster, S. O. et al. Gastroenteritis due to rotavirus in
rotavirus to volunteers: induction of illness and corre­
an isolated Pacific Island group: an epidemic of 3439
lates of resistance. Journal of infectious diseases, 147:
cases. Journal of infectious diseases, 141: 32-39 (1980).
95-106(1983).
25. Glass, R. I. et al. Endemic cholera in rural Bangla­ 40. Kapikian, A. Z. et al. Studies in volunteers with
desh, 1966-1980. American journal of epidemiology,
human rotaviruses. In: International Symposium on
116: 959-970(1982).
Enteric
Infections
in
Man
and
Animals:
26. Glass, R. I. et al. A randomized, controlled trial of
Standardization of Immunological Procedures, Dublin,
the toxin-blocking effects of B subunit in family
1982. Basel, S. Karger, 1983. Development in biological
members of patients with cholera. Journal of infectious
standards, Vol. 53, pp. 209-218.
diseases, 149: 495-500 (1984).
41. Kapikian, A. Z. & Chanock, R. M. Rotaviruses. In:
27. Greenberg, H. B. et al. Rescue of noncukivatable
Fields, B. N. et al., ed., Virology. New York, Raven
human rotavirus by gene reassonment during mixed
Press, 1985, Ch. 37.
infection with ts mutants of a cultivatable bovine rota­ 42. Khan, M. U. Efficacy of short course antibiotic
virus. Proceedings of the National Academy of Sciences
prophylaxis in controlling cholera in contacts during
of the USA, 78: 420-424 (1981).
epidemics. Journal of tropical medicine and hygiene,
28. Greenberg, H. B. et al. Rescue and serotypic charac­
85: 27-29(1982).
terization of noncultivable human rotavirus by gene
reassortment. Infection and immunity, 37: 104-109
.
(1982)

5S2

I. DE ZOYSA & R. G. FEACHEM

Khan, M. & Shahidullah, M. Cholera due io the El 61. Oseasohn, R. et al. Clinical and bacteriological find­
ings among families of cholera patients. Lancet, 1:
Tor biotype equals the classical biotype in severity and
340-342(1966).
attack rates. Journal of tropical medicine and hygiene,
62. Paniker, C. K. J. et al. Rotavirus and acute
83: 35-39(1980).
diarrhoeal disease in children in a southern Indian
44. Khan, M. U. et al. The eltor cholera epidemic in
coastal town. Bulletin of the World Health Organiz­
Dhaka in 1974 and 1975. Bulletin of the World Health
ation, 60: 123-127 (1982).
Organization, 61: 653-659 (1983).
45. Levine, M. M. et al. Immunity of cholera in man: 63. Peterson, J. W. Synergistic protection against experi­
mental cholera by immunization with cholera toxoid
relative role of antibacterial versus antitoxic immunity.
and vaccine. Infection and immunity, 26: 528-533
Transactions of the Royal Society of Tropical Medicine
(1979).
and Hygiene, 12>: 3-9 (1979).
46. Levine, M. M. et al. Present status of cholera 64. Pierce, N. F.&Koster, F. T. Stimulation of intestinal
immunity. In: Holme, T. et al., ed., Acute enteric infec­
vaccines. Biochemical Society transactions, 12: 200-202
tions in children. New prospects of treatment and
.
(1984)
prevention.
Amsterdam, Elsevier/North-Holland Bio­
47. Levine, M. M. et al. Evaluation in humans of
medical Press, 1981, pp. 413-423.
attenuated Vibrio cholerae El Tor Ogawa strain Texas
65. Pierce, N. F. et al. Oral immunization of dogs with
Star-SR as a live oral vaccine. Infection and immunity,
purified cholera toxin, crude cholera toxin, or B
43: 515-522 (1984).
subunit: evidence for synergistic protection by antitoxic
48. Martin, A. R. et al. Epidemiologic analysis of
and antibacterial mechanisms. Infection and immunity,
endemic cholera in urban East Pakistan, 1964-1966.
37:687-694 (1982).
American journal of epidemiology, 89: 572-582 (1969).
49. Mata, L. et al. Epidemiology of rotaviruses in a 66. Rodriguez, W. J. et al. Rotavirus gastroenteritis in
the Washington, DC, area. Incidence of cases resulting
cohort of 45 Guatemalan Mayan Indian children
in admission to the hospital. American journal of
observed from birth to the age of three years. Journal of
diseases of children, 134: 777-779 (1980).
infectious diseases, 148: 452-461 (1983).
50. McCormack, W. M. et al. Endemic cholera in rural 67. Snyder, J. D.&Merson, M. H. The magnitude of the
global problem of acute diarrhoeal disease: a review of
East Pakistan. American journal of epidemiology, 89:
active surveillance data. Bulletin of the World Health
393-404(1969).
Organization, 60: 605-613 (1982).
51. Mebus, C. A. et al. Cell culture propagation of neo­
68.
Spencer, H. C. et al. Diarrhea in non-hospitalized
natal calf diarrhea (scours) virus. Canadian veterinary
rural Salvadoran population: the role of enterotoxigenic
journal, 12: 69-72 (1971).
Escherichia coli and rotavirus. American journal of
52. Mekalanos, J. J. et al. Cholera toxin genes:
tropical medicine and hygiene, 29: 246-253 (1980).
nucleotide sequence, deletion analysis and vaccine
69. Stoll, B. J. et al. Surveillance of patients attending a
development. Nature, 306: 551-557 (1983).
diarrhoeal disease hospital in Bangladesh. British
53. Monto. A. S. et al. The Tecumseh study. XII. Enteric
medical journal, 285: 1185-1188 (1982).
agents in the community, 1976-1981. Journal of
70. Stucker, G. et al. Antigenic comparisons of two new
infectious diseases, 148: 284-291 (1983).
rotaviruses from rhesus monkeys. Journal of clinical
54. Mosley, W. H. & Khan, M. Cholera epidemiology —
microbiology, 11: 202-203 (1980).
some environmental aspects. Progress in water
71.
Svennerholm, A-M. & Holmgren, J. Synergistic
technology, 11: 309-316 (1979).
protective effect in rabbits of immunization with Vibrio
55. Mosley, W. H. et al. The relationship of vibriocidal
cholerae lipopolysaccharide and toxin/toxoid. Infec­
antibody titre to susceptibility to cholera in family
tion and immunity, 13: 735-740 (1976).
contacts of cholera patients. Bulletin of the World
72. Svennerholm, A-M. et al. Intestinal antibody
Health Organization, 38: 777-785 (1968).
responses after immunisation with cholera B subunit.
56. Mosley, W. H. et al. An epidemiological assessment
Lancet, 1; 305-308 (1982).
of cholera control programs in rural East Pakistan.
73. Svennerholm, A-M. et al. Current status of an oral B
International journal of epidemiology, 1: 5-11 (1972).
subunit whole cell cholera vaccine. In: International
57. Muchinik, G. R. & Grinstein, S. Rotavirus infection
Symposium on Enteric Infections in Man and Animals:
in children hospitalized for diarrhoea in Argentina.
Standardization of Immunological Procedures, Dublin,
A nnals of tropical paediatrics, 1: 167-173(1981).
1982. Basel, S. Karger, 1983. Development in biological
58. Mutanda, L. N. Epidemiology of acute gastroenteritis
standards. Vol. 53, pp. 73-79.
in early childhood in Kenya. HL Distribution of the
74. Truant, A. L. & Chonmaitree, T. Incidence of rota­
aetiological agents. East African medical journal, 57:
virus infection in different age groups of pediatric
317-326 (1980).
patients with gastroenteritis. Journal of clinical micro­
59. Noriki, H. Evaluation of toxoid field trial in the
biology, 16: 568-569 (1982).
Philippines. In: Proceedings of the 12th Joint
75. Vesikari, T. et al. Immunogenicity and safety of live
Conference on Cholera, Sapporo, November 1976.
oral attenuated bovine rotavirus vaccine strain RIT 4237
Tokyo, National Institute of Health, 1977, pp. 302-310.
in adults and young children. Lancet, 2: 807-811
60. Oberle, M. W. et al. Diarrhoeal disease" in
(1983).
Bangladesh. Epidemiology, mortality averted and costs
76. Vesikari, T. et al. Protection of infants against rota­
at a rural treatment centre. International journal of
virus diarrhoea by RIT 4237 attenuated bovine rotavirus
epidemiology, 9: 341-348 (1980).
strain vaccine. Lancet, 1: 977-981 (1984).

43.

ROTAVIRUS AND CHOLERA IMMUNIZATION FOR CONTROL OF DIARRHOEAL DISEASES IN CHILDREN

Vesikari, T. et al. Increased “take” rate of oral rota­
virus vaccine in infants after milk feeding. Lancet, 2:
700(1984).
78. Woodward, W. E. & Mosley, W. H. The spectrum of
cholera in rural Bangladesh. II. Comparison of El Tor
Ogawa and classical Inaba infection. American journal
of epidemiology, 96: 342-351 (1972).
79. Woodward, W. E. et al. The spectrum of cholera in
rural East Pakistan. I. Correlation of bacteriologic and
serologic results. Journal of infectious diseases, 121:
S10-S16 (1970).
80. Wyatt, R. G. et al. Rotaviral immunity in gnotobiotic
calves: heterologous resistance to human virus induced
by bovine virus. Science, 203: 548-550 (1979).
81. Wyatt, R. G.etal. Diarrhea associated with rotavirus
in rural Guatemala: a longitudinal study of 24 infants
and young children. American journal of tropical
medicine and hygiene, 28: 325-328 (1979).
82. Wyatt, R. G. et al. Human rotavirus type 2: culti­
vation in vitro. Science, 207: 189-191 (1980).
S3. Wyatt, R. G. et al. Prospects for development of a
vaccine against rotavirus diarrhea. In: Holme, T. et al.,
ed., Acute enteric infections in children. New prospects

77.

583

for treatment and prevention. Amsterdam, Elsevier/
North-Holland Biomedical Press, 1981, pp. 505-522.
Wyatt, R. G. et al. Definition of human rotavirus
serotypes by plaque reduction assay. Infection and
immunity, 37: 110-115 (1982).
85. Wyatt, R. G. et al. Direct isolation in cell culture of
human rotaviruses and their characterization into four
serotypes. Journal of clinical microbiology, 18:310-317
(1983).
86. Wyatt , R. G. et al. Induction of cross-reactive serum
neutralizing antibody to human rotavirus in calves after
in utero administration of bovine rotavirus. Journal of
clinical microbiology, 18: 505-508 (1983).
87. Yardley, J. H. et al. Local (immunoglobulin A)
immune response by the intestine to cholera toxin and its
partial suppression with combined systemic and intraintestinal immunization. Infection and immunity, 19:
589-597 (1978).
88. Zissis, G. et al. Protection studies in colostrumdeprived piglets of a bovine rotavirus vaccine candidate
using human rotavirus strains for challenge. Journal of
infectious diseases, 148: 1061-1068 (1983).
84.

Position: 1306 (4 views)