Breast-feeding

and Diarrheal

Barry M. Popkin,
Robert

Black,

Wilhelm

MS,

PhD*;

MD, MPHt;

Flieger,

ABSTRACT.
than

This study used a unique

longitudinal
3000 mother-infant
pairs
observed

pregnancy

through

infancy.

The

of infants

from the Cebu

region

sequencing

of

breast-feeding

sample

and

diarrheal

of feeding

patterns.

range

of environmental

Because

suits

can

be generalized

the

work

causes
to other

The

morbidity

analysis
effects

controlled

for

of diarrhea,

populations

the rewith

and tDepartment
of International
Bank, Washington,
Cebu, Philippines

some

foods,

confidence.
The addition
to the breast-milk
diet of even
water, teas, and other nonnutnitive
liquids doubled
or
tripled
the likelihood
of diarrhea.
Supplementation
of
breast-feeding

broths).

further

increased

benefits
nutritive

additional

nutritive

significantly

of breast-feeding
foods/liquids

foods

882; breast-feeding,

on liquids

the risk of diarrhea;

most

alone on in combination
became
small during
the

with
second

half of infancy.
Benefits
of breast-feeding
greater in urban environments.
Pediatrics
diarrheal

morbidity,

were

slightly

disease,

a major

cause

for publication
Sep 29, 1989;
requests
to (B.M.P.)
Carolina

versity

Square

Chapel

Hill,

East,
Chapel

PEDIATRICS

CB

8120,

Hill,

NC

(ISSN

University

mor-

Academy

874

PEDIATRICS

at

27516.

0031 4005).
of Pediatrics.

American

Carolina

Copyright

1990

Vol. 86 No. 6 December

by

(eg,

number

milks

evaporated

PhDt;

in

infancy

gruels,

nonnutritive

feeding
make

ages

as

rapid

regimens
simplistic

the

nutritive

well

as

a large

(teas,
and

water,

often

occurring

subtle

throughout

definitions

of bottle

misleading.23

and

Nonetheless,

categorized
the inor bottle-fed
and,
the complexity
of
time and at differ-

infant.4

of a given

certainly

vary

infants

nutritional

with

infant-feeding

the

infants

needs

vary

method

age.

Not

almost

only

at different

do an

ages,

but

so does the infants

vulnerability
to the harm that
can result
from feeding-pattern
changes.
Whereas
a 3-week-old
neonate
may suffer from being weaned
completely
from breast
milk, a 9-month-old
infant
may be minimally
affected.
Thus,
it must
be recognized
that a number
of
possible
mechanisms
of diarrhea
found

exist
for
in breast-feeding

the

lower
frequency
infants
and the

higher
risk observed
when
other
fluids
and foods
are introduced.
First,
breast-fed
infants
have less
exposure
to entenic
pathogens
transmitted
by contaminated
fluids and foods. Second,
breast-feeding
may offer protection
against
diarrhea
because
of
humoral

man

1990

condensed

other

liquids
the

feeding

of the

as breast-milk
and

milk),

and

Furthermore,

shifts

used

formulas

juices),

of

Effects

accepted
Jan 2, 1990.
Population
Center,
Uniof North

S. Akin,

and

previous
research
has generally
fant as breast-fed,
mixed-fed,
therefore,
has failed to address
feeding
patterns
at any specific

bidity and mortality,

is the focus of extensive
public
health
program
efforts
in numerous
low-income
countries.
Most recent
efforts
to alleviate
the problems of diarrhea!
disease
have focused
on promoting
its case management
primarily
through
the use of
oral rehydration
therapy.
The value of preventive
measures,
such as the promotion
of appropriate
infant-feeding
practices,
has been
little
emphasized.1 There
has been extensive
documentation
of
the protective
value of breast-feeding
against
diarrhea in settings
where
poor hygiene
is prevalent.
Received
Reprint

and

supplemental

1990;86:874Philippines.

of infant

various
(commercial

liquids

ent
Diarrhea!

John

of Economics,
University
of North
Health, Johns Hopkins University,
DC and IlOffice of Population

including

substitutes

sweetened

with

PhD*;

MS, PhD;

Nevertheless,
important
issues related
to the effects
of types of food supplements
and the risks associated with their
use in different
types
of environments
remain
largely
unaddressed.
The vast majority
of infants
who are not exclusively breast-fed
consume
a wide variety
of other

from

is representative

events was carefully

examined
in a longitudinal
which allowed for the examination
of age-specific
wide

survey

of the Philippines.

Adair,

Briscoe,

PhD II

MA,

From the *Depment

of Nutrition
Carolina at Chapel Hill; epartment
Baltimore,
Maryland;
The World
Studies, University
of San Carlos,

of more

Linda

John

Morbidity

and

milk.

cellular

Third,

antiinfective

breast

milk

properties

may

promote

in hu-

an

intestinal

environment

and

micnoflora

proliferation
of entenic
pathogens.
feeding
can enhance
the infants
thereby

reducing

that

susceptibility

surveys.
Additional
details
data appear
eisewhere.57

inhibits

Finally,
nutritional

breaststatus,

specific

rheal
wide

effects

examined.
Longitudinal
for the examination

of different

feeding

morbidity.
Because
range of environmental

results
some

MATERIALS
Study

the

can be generalized
confidence.

AND

design

and

to diarrhea.

We present
a study of the dynamic
relationship
between
infant-feeding
patterns
and diarrhea!
disease.
Data
were collected
through
a longitudinal
survey
in the Philippines,
designed
specifically
to
study
the relationship
of diet and health.
The sequencing
of breast-feeding
and diarrhea!
morbidity
events
is carefully
tical methods
allow

on the survey

statisof the age-

patterns

on diar-

work controls
causes of diarrhea,

to other

for a
the

populations

with

Variables

Diarrhea!

mothers
report
sode of diarrhea

is represented
variables

of whether
her infant
in the 7 days before

used

by a set of
reflect the

had an epieach
longitu-

in the

literature.

The

percentages

of infants

with diarrhea
in the 7 days before each survey
were
7, 12, 20, 24, 24, and 23 for ages 2, 4, 6, 8, 10, and
12 months,
respectively.
The percentage
of study
children
who
experienced
at least
one diarrhea!
at some

point

a number
comparable
ported
for the whole

Design

that

dna!
survey.
This
variable
is responsive
to the
independent
variables
of interest,
was measured
with a high degree
of reliability,
occurs
with sufficient
frequency
in the sample,
and is commonly

episode

METHODS

morbidity
dichotomous

time-varying

during

the

with
nation

the rate of 52.9%

of the Philippines

first

year

was

60.3,

refor

1983.8

The

study

site

was

includes

the

cities

of

Lapu; coastal
Cebu towns;
a number
of villages
in
mountainous
and coastal
areas of Cebu; and inhabited areas
on some
smaller
islands.
We used
a

Infant
feeding
data
were
obtained
by trained
interviewers
using survey
instruments
designed
to
gather
precise
information
on patterns
of feeding
and food processing.
Interobserver
reliability
studies, preceded
by a 2-week
dietary
data collection

stratified

training

two-stage

metropolitan
Cebu,

Cebu,

Mandaue,

sampling

Households
births

were

between

Baseline

pregnancy

pregnant

women

month

period.

surveys

and

were

gave

Ofthese

to select

to collect

1, 1983,

who

Lapu

or barangays

surveyed

May

and

procedure

random
sample
of 33 communities
urban
and 16 rural).

which

on

30,

obtained

birth

women,

data

April

all

1984.

for 3327

during

this

had

stillbirths,

1.1%

(17

12-

0.4% had miscarriages,

0.8% had twin births,
4.6%
either
migrated
before
the infants
birth or refused
birth interviews,
and 0.7% represented
observations
whose
data
files had faulty
information
(such
as
baseline
interview
data obtained
after birth).
For
this analysis
of infant
health,
initially
of the 3080
women
births
and for whom
baseline

were
available.
An
sample
communities
month

period

but

the sample
consisted
who
had
single
live
and birth
information

additional
57 women
gave
birth
during
either

did

not

live

in
the

in the

the
12com-

munities
during pregnancy
on were somehow
missed
during
the screening
for pregnant
women.
Data
on the mothers
and infants
came
from
baseline
seventh

surveys
month

and gestational
and in the birth
gitudinal
surveys
price and health
health
personnel,

of the
of pregnancy,

mothers

in
from

their

sixth

or

anthropometnic

age information
obtained
at birth
information
survey,
and from ionrepeated
bimonthly.
Community
facility
data came from a set of
health
facility,
and food market

course,

were

used

to

ensure

high-quality

data. Additional
training
and reliability
tests were
conducted
several
times
during
the survey
period.
Using
food models
and measuring
aids,
trained
interviewers
recorded
the frequency
of breast-feeding and the types, amounts,
and feeding
frequencies
of all liquids,
semisolids,
and solids fed to the infant
on the day preceding
the survey.
Information
was
also

obtained

on

the

and whether
infants
categories
(eg, water,
solid

For

or solid

this

of mutually
breast-feeding
liquids.

An

frequency

foods,

etc)

analysis,
and

7 days

infants

exclusive
infant

of

were given
breast
milk

consumption
was

before

were

feeding
considered

breast-feeding

foods in specified
substitutes,
semieach

grouped

categories,

of other

survey.

into

a set

reflecting

foods

and

to be exclusively

breast-fed
if its entire
intake
consisted
only of
breast
milk from its own mother.
This definition
excluded
consumption
of any foods or liquids
that
could potentially
introduce
pathogens
to the infant.
A second
category
included
breast-feeding
in cornbination
with nonnutnitive
liquids
such as water
and herbal
teas. More than
95% of the infants
in
this category
were fed water.
A third
category
encompassed
breast-feeding
in combination
with nutritive
liquids
(eg, juices, milks) or foods (sernisolids
or solids).
The
fourth
category
represented
no
breast-feeding

In Table

at all.

1 we present

the

prevalence

ARTICLES

of infant-

875

TABLE
pines,

1. Breast-feeding
1983_1985*

Patterns

During

Infancy,
Bre ast-feeding

Infants
by Age
and Residence

Breast

Milk

Breast

Cebu

Metropolitan

Philip-

Patterns
.

Milk

Area,

Breast
Milk + Nutritive
foods,t
Maybe Nonnutnitive Liquids

No

Breast

Milk

Only

Nonnutritive
Liquids
Only

Rural
Urban

53.3
24.0

17.3
19.4

23.1
38.0

18.7

2186

Both

31.1

18.9

34.4

15.7

2886

4 mo
Rural

31.7

19.3

40.4

8.6

16.3
20.1

18.9
19.0

40.6
40.4

24.2
20.3

Rural

4.6

6.3

79.2

Urban

2.8

Both

3.2

4.7
5.1

63.9
65.6

Rural

0.2

0.8

86.9

12.2

662

Urban

0.2

0.9

66.4

32.5

2007

Both

0.2

0.9

71.4

27.5

2669

0.0
0.2

0.6
0.5

84.1
62.2

15.3

654

37.1

1976

0.1

0.5

87.7

31.7

2630

Rural

0.2

0.2

78.0

21.7

650

Urban
Both

0.1

0.1

56.4

43.5

1950

38.0

2600

2 mo

Urban
Both

698

6.3

688
2118

2806

6 mo

668

9.9

2052
2720

28.6
24.0

8 mo

10 mo

Rural
Urban
Both
12 mo

Based

on

0.1

0.1

feeding

behavior

on

liquids

including

the

61.8

day

preceding

the

survey.

Horizontal

columns

total

100%.

t Nutritive

other

milks

or semisolid

feeding
patterns
for urban
and rural infants
from
birth to 12 months.
Urban
and rural patterns
were
examined
separately
because
previous
analyses
of
the Cebu data have found important
differences
in
health-related
behaviors
and their determinants
between these samples,
justifying
their separate
consideration.67

Even

though

many

continued

to

breast-feed,
infants
were introduced
to other foods
and liquids,
particularly
nonnutritive
ones,
very
early. At 2 months,
approximately
19% of infants
were fed breast
milk in combination
with nonnutnitive
liquids
and 34% were fed breast
milk and
nutritive
foods or liquids
(possibly
in combination
with nonnutnitive
liquids).
Expressed
in another
way, approximately
63% of breast-fed
infants
had
their diets supplemented
with some other liquid or
food by the age of 2 months.
By 8 months
essentia!ly
liquids

all breast-fed
and foods.

infants

were

receiving

We note several
rnent of breast-feeding

other

points
related
to the measurein this study. First, mothers
may
move
in and
out of exclusive
breast-feeding.
They
may exclusively
breast-feed
their
infants
at
one age, add foods at a later age, and then,
even
later,
return
to exclusive
breast-feeding.
For example,
116 urban (5.3%) and 25 rural infants
(3.6%)

876

BREAST-FEEDING

AND

DIARRHEAL

and solid foods.

made

transitions

terns

at

2 months

at 4 months
1988.

EZ,

teas,

Popkin

manuscript).

water,

or some

gruels,

nutritional
infants

value
received

and

juices,

a sporadic

sumption

more complex
feeding
patto exclusive
breast-feeding

(Bisgrove

Unpublished

ceived

on

from
back

BM,
Most

other

nonnutnitive

or other

foods

at an early
age.
these
non-breast-milk
basis,

is not

and

likely

in

Barba

re-

broths,

or liquids

with

However,
items

many
only

those

to have

C.,

infants

cases

the

a long-term

con-

effect

on

infant
morbidity.
Second,
some infants
were fed by a wet nurse.
At
each 2-month
interval,
2% to 4% of the infants
were nursed
at least once in the day preceding
the
survey
by a woman
other than their mother.
In no
case
breast

in

breast

the data,
milk from

however,
a wet nurse

milk

four

or

own

mother.

The

possible

not

considered

feeding
birth.

boiled,

MORBIDITY

this

is initiated
Before

and

result
from
ing is not

patterns.9

in

more

this,

did an
without
times

the

no effect

that

effects
study.

in the

infant
also

Third,

their
consumption;
considered
in the

from

most

to fifth
fed

on neonatal

day

its

of wet nursing

third

liquids

receive
receiving

to

are

breast-

day

newborns

health

thus prelacteal
definition
of

after
are

seems

to

feedfeeding

Statistical

Controls

To understand
the precise
relationship
between
feeding
and infant
morbidity,
it is necessary
to
control
for various
biologic
and behavioral
variables
that affect susceptibility
to illness
and exposure
to
pathogens.

We

methods

used

designed

for analysis

man
populations.
in cases
where
and

or the

ural

settings

reasons)

to

research

must

to

observe

during

the

Susceptibility
the
status

first

biologic

at

birth

in hu-

appropriate
(for ethical

conduct

experimental

be conducted
in

in nata

realistic

allow for the evaluation

of
susceptibility
and exposure
be significantly
related
to

variables

infants

of behaviors

behaviors

context.
These
methods
a number
of important
factors
considered
to
diarrhea

multivariate

These
methods
are
either
it is not possible

financial

studies,

longitudinal

year

of the

infants

capture

information

endowments,

including

(measured

by

life.

about
health

gestational

age

and

birth

weight)
and subsequent
healthiness
(measured
by weight
velocity
and weight
velocity
interacted with weight).
Sex of the infant
is an exogenous variable
that may represent
susceptibility
differences
because
of genetic
endowments
of the
infant

or the

effects

of the

haviors

of different

mother

when

health-related

dealing

be-

with

males

vs

females.

Exposure
coming
into
measures

variables
contact
of personal

quality,
excreta
and the mothers
(evidenced
rhea tends

each

variables
or within
affect

the

and

the

The

presence

in

or under

The
will

care).
patterns

varies

other

contaminated

more

probability

season-

transmission

of animals

(chickens,

house

also
that

represents
can

con-

of pathogens.
livestock,

be transmitted

to
from

presents

diarrhea

the probability
as the

result

that

of feeding

an infant

is not

combined

with

various

and

biomedical

model,

of logical

relationships,

the model
processes

beginning

with

Inasmuch
reduce
the

the

as breast-feeding
may not only directly
incidence
of diarrhea,
but also increase

infants

severity

nutritional

status,

incidence

and

in turn

reducing

of diarrhea,3

the

these

mdi-

rect
effects
must also be considered.
Prior
growth
velocity
(grams
per day during
each measurement
period)
and growth
velocity
interacted
with prior
nutritional
status
(the interaction
term is included

growth

velocity

because

a given

effects
status

on health
when
the

diarrhea
effect

analysis
to capture
of breast-feeding
on

according

may

have

different

to the childs

growth

began)

are

nutritional

included

in the

the indirect
nutritional
diarrhea!
morbidity.

It

is also possible
that exclusive
breast-feeding
could
have
direct
and indirect
effects
in opposite
directions,
particularly
at older
ages
when
exclusive

breast-feeding

may

be associated

with

growth

fal-

tering.

The

procedure

of each

allowed

of
of the

affecting

each

example,

infant

behavioral

for
ions

can

the

also

probit

normally

for.
and

outcomes
infants
and

the

factors

over

time,

modeled

procedure

(mothers
health

effects
(Cebu

may

and

also

recognize

take

measures

of correlation
Study

of the
biologic
with

statistical
of certain

Team.

For

modeled

of their effects
manuscript).10

encountered
the endogenity

and

processes

accounted

are

on
for.

timing

behavioral
be

net

model

accounted

is recorded

controls

them),

lished

model,
and

feeding

in particular

and

be

the

mu!the

in the

can

diarrhea

longitudinal

several

lems,

diarrhea

of the likely timing

Team.
1989. Unpublished

Study

to calculate

us

variable

biologic

of

This

period just before

the observation
or absence
of diarrhea.
The other

a time

presence
sideration

estimation

probit.

it is a longitudinal

sequencing

and

for

independent

probability

Because

used

is a longitudinal

procedure

effect
the

we

model

con-

(Cebu

accounts

probbehavthreats
to reduce

of unobserved
1989.

Unpub-

manuscript).15

practices

and of the full set of susceptibility

and control
variables
described
above.
To understand
the precise total effect of breast-feeding
on diarrhea,
when
it is and

a series

to their

Methods

behavioral

a set of community
and household
socioeconomic
conditions
and progressing
through
their effects
on
a set of intermediate
behavioral
and biologic
vanables
(eg, food processing
and birth
weight)
to
health
outcomes.

The

pets)

exposure

into

must be considered.
The
in this study was based

in detail elsewhere.#{176} In general

a set of health-affecting

disaggregates

for

Other

of person-to-person

resultant

a coherent

discussed

tivariate

Diaras a

variables.

on

of relationships
of the relationships

diarrhea

humans.

model

have

the

this
or

pathogens
to

Modeling

water

represent
crowding
in the cornthe household,
either
of which

the

additional
animals

health
seasonal
to

because

temperature

tacts,

may

of soap),

practices,
food hygiene,
with preventive
measures

exposures

survey

than

munity

(use

In this case study,

seasonality
was
by a measure
of rainfall
in the 14 days

represented

control

disposal
concern

varying

environment.
before

hygiene

by use of preventive
to follow
distinct

of

result

ally

reflect
an infants
chances
of
with pathogens.
These
include

a number
analysis

other

foods,

RESULTS

in

Results
detail

are derived
from
in the Appendix.

the model
represented
The
coefficients
are

ARTICLES

877

maximum

likelihood

estimates

variable
negative

on the probability
coefficient
means

reduced

by that

cating

the

ing.

was
both

For

variable,

statistical

Diarrhea

days

urban
(no

that

ing the
Similarly,
with

A
is

effect.

of feed-

samples,

liquids

or

exclusive

foods

given)

decreased

would

have

combination

nificant!y

less

not breast-fed.
breast-feeding
for the urban

likely

1 and

Breast

milk

+ nonnutritive

Breast

milk

+ nutritive

breast

milk

Adjusted
probability
of
having
diarrhea

20

.10

diarrhea

on

on diarrhea.

foods

than

ity

infants

information

of

on

the

in breastThe estieffect
of

creased

even

more

foods

tive

a!! other

character-

tended

diarrhea;

breast

milk

Diarrhea!
for

to

fed

and

was

breast

the

nonnutritive

morbidity

infants

have

the likelihood
and

milk

highest

inplus

By

probabil-

D
LI

Breast

who

were

not

exclusively

breast-fed,
who

infants

in Table

that

the

pre-

were

exclu-

2. From

feeding

those

of nonnutri-

of

age,

when

sample
were
other
nutritive

and

benefits

all

breast-feeding

fed a mixed diet of breast

foods
and liquids,
the

of breast-feeding

fell

considerably.

Only in urban
areas was the probability
of diarrhea
for breast-fed
infants
at this age lower than it was
for infants
who were not breast-fed
and, even there,

liquids
I liquids

foods

infants

with

appear
be seen

months

relative

milk + nutritive

No breast

in

in the

milk
milk
+ nonnutritive

Cebu,

liquids
had important
effects
on diarrhea!
morDepending
on age, the infant
was 2.0 to 3.2
times
as likely
to have
diarrhea
if nonnutritive
liquids
were fed in addition
to breast
milk than
if
he or she was only fed breast
milk. When
nutritive
foods
and/or
liquids
were
added to the breast
milk
diet, the infant
became
4.7 to 13.1 times
as likely
to have diarrhea!
morbidity.
At given times during
the first 6 months,
the infant
who was not breastfed was from 4.7 to 16.8 times
as likely to have
diarrhea
as was the exclusively
breast-fed
infant.
infants

Only breast
Breast milk

morbidity,

sample.

bidity.

infant-feeding

probabilities.
infants

found

not

has

or liquids,

were

only

both

higher.

who

that

morbidity

was

fants

infant

of developing
fed

infants

was

it can

breast-fed

liquids

diarrhea

breast-fed,

varies,
the prevalence
of diarrheal
morbidassociated
with
each
feeding
pattern
is presented.
In other
words,
these
figures
represent
the

low likelihood

of

results,

ity

diarrhea!

rural

breast-fed
at all. Between
one fifth and one third
of all infants
who were not breast-fed
7 days before
a survey
can be expected
to have diarrhea
in the 7
days between
that date and the survey
day.
The relative
risks of diarrhea!
morbidity
for in-

sively

the

diarrheal

to 1985:

and the indirect

effects
and, in turn,
of growth

so

Exclusively

and

1983

in comparisons

pattern

adjusted

Breast-feeding

Philippines,

sented

diarrhea
on growth

If each

controlled

Fig 2.

to be sig-

responsiveness
of diarrhea
to changes
feeding
for the urban
and rural samples.
mated
relationships
reflect
the direct
breast-feeding
of breast-feeding

10

to
fed

The combined
impact
variables
is statistically
and rural samples.
present

6
8
Age (in months)

relative

found

I liquids

milk

No breast

survey

nutritive

not

liquids
foods

dun-

Infants

with

were

to have

2 we

before

infants.

however,

who were
the three
significant

nutritive

Only

D
U

ri

the

diarrhea

of diarrhea,

breast-fed

liquids,

Figs

7 days

probability
in

and/or

for

mdi-

by mode

rural

liquids

fully

breast-milk

istics

of each

week from that

date to the survey
date.
infants
fed breast
milk in combination

except

In

of the

significantly

an infant

decreased

all

and

other

nonnutnitive

had

effect

T statistic

affected

a survey

probability

the

significance

the

before

with

strongly

breast-feeding

of the

of having
diarrhea.
that the probability

milk

the

Adjusted

difference

clear

probability

of
having
diarrhea

inverse

tective

was

small.

The

relationship

results

between

of breast-feeding

effects

on

indicate

age and

the pro-

diarrheal

mor-

bidity.

DISCUSSION
The
2

10

12

Age (in months)

Fig 1.
Philippines,

878

Breast-feeding
and
diarrhea!
1983 to 1985: urban
sample.

BREAST-FEEDING

AND

morbidity,

Philippines,

finding

breast-feeding
Cebu,

ity.4
either

DIARRHEAL

Cebu,

general

MORBIDITY

In

case

of a large
protects

particular,
nonnutritive

against

the

reaffirms

indicate
or

the

of research
diarrhea!

results

liquids

study

body

nutritive

that
morbid-

that

adding
foods

or

TABLE

2.

Feeding

Patterns,

Infants

Adjusted

Relative

Risk

Metropolitan

by Age
.
Residencet

and

and

Cebu,

95%

Confidence

Philippines,

Breast
Milk
Only

Breast

Urban

by

Childs

Age

and

Infant-

Lower

95%

CI
Upper

Limit

Limit

Breast

Milk
+ Nutritive
.
Foods/Solids

Relative
Risk

95%

No

Lower

CI
Upper

Limit

Limit

Breast

Milk

Relative

95%

CI

Risk

Lower
Limit

Upper
Limit

2-6 mo

2 mo

1.00

3.17

2.50

3.83

13.30

6.15

20.44

17.32

6.74

27.89

4 mo

1.00

2.12

1.77

2.46

12.93

6.04

19.81

13.72

6.14

21.30

6 mo

1.00

3.18

2.53

3.84

10.61

4.63

16.60

13.45

5.24

21.66

2 mo

1.00

2.24

2.01

2.47

5.67

4.31

7.03

5.68

4.31

7.05

4 mo

1.00

1.97

1.80

2.13

6.30

4.93

7.66

5.64

4.52

6.76

1.00

2.21

2.07

2.35

4.73

3.95

5.52

4.73

3.97

5.49

sample

6 mo

Age group:
8-12
Urban
sample

mo

8 mo

lOmo
12 mo
Rural
sample
8 mo

10 mo
12 mo

The

Morbidity

sample

Rural

a! of Diarrhea!

Milk
+ Nonnutni.
tive Liquids

Relative
Risk

Age group:

Interv

1983_1985*

relative

risk

reported

here

is the

mean

1.00

1.32

1.22

1.42

1.00
1.00

1.32
1.32

1.23
1.23

1.42
1.42

1.00

1.02

1.02

1.03

1.00
1.00

1.03
1.04

1.02
1.03

1.03
1.04

each

of which

of individual-level

risks,

is the

ratio

of individual

probabilities.
The
individual-level
probabilities
are estimated
as in Figs.
1 and
2. The
standard
deviation
of the
individual
re!ative
risk is multiplied
by 1.96 and the product
is added
and subtracted
to the mean
relative
risk to obtain
the 95% confidence
interval
(CI).
t The relative
risk is the ratio
of the risk of diarrheal
morbidity
or respiratory
infection
for infants
in each age-feeding
category
compared
with
exclusively
breast-fed
infants
for 2 to 6 months
and for those
breast-fed
plus
fed nutritive
foods and liquids
for ages 8 to 12 months.
Risks
are adjusted
for a set of exposure
and susceptibility
factors
mentioned

in the text.
liquids
to the breast-milk
diet is associated
with a
large increase
in the occurrence
of diarrhea
and
that
the protective
effect
of breast-feeding
decreases
with
age. It is important
to note that if the
classification
procedure
used in this study
causes
infants
to be placed
into a breast-feeding
category
to which
they
do not continuously
belong,
that
category
is likely to be either
the only-breast-milk
or breast-milk-plus-nonnutnitive-liquids
category,
because

both

short

represent

lengths

because

of

foods.

such

If the
health

some

milk

periodically

and

breast-feeding
and

herbal

illness

nonnutritive

broths.

that

or

shortage

hypotheses

about

of breast-feeding.
be

are

receive

the

nutritive

exclusively

liquids

relative
liquids

Consumpion

to being
such

as

of these

reported

the

sively

could

infants

most

adequate
breast-feeding
be expected

fed

non-

breast-fed

protective

should

pose

no

conditions

because

hydration

status

even
in warm
to substantially

effect

of

health

risks

infants

while

can

exclu-

climates6
and
reduce
diarrhea

in this age group.

This study
also indicates
that supplementation
of breast-feeding
with
nutritive
foods
or liquids
further
increases
the risk of diarrhea
above
that
results

when

added

to the

breast-milk

nonnutni-

infants

exclusively

smaller

diet

climatic

which

teas,

being

breast-feeding
than actually
existed
in the
samples
analysed.
The relationship
found
in the
Cebu study is statistically
strong
and large in magnitude,
and it holds
true when
a wide range
of
exposure
and susceptibility
factors
are controlled
in the analysis.
Eliminating
nonnutnitive
liquids

or foods.

fed breast

adverse
effect on
also suggest
that

misclassified

as

study is of
of life, of
water,

a large
results

exclusive

maintain

other liquids
as breast-fed
liquids
ac-

based on the case

the first 6 months

probably

from

of

to have
The

that

liquids

under

because

as being

research

nutritive

a
posi-

Benefits

earlier

of

creating

underestimated

classified

for

probably

we are

actually
periodically
receive
and some infants
classified
addition
of only nonnutnitive

A major conclusion
the benefit,
during
exclusive

as

would

who

patterns

periodically,

is correct,

of the

effects

infants

tually

factors

test

breast-feeding

breast-fed
or foods,
with the

occur

assertion

conservative

tive

feeding

of time

tive liquids
seems
diarrhea!
morbidity.

can be
exclusive

only

inferred
that
breast-feeding

breast-fed

infant

promote

adequate

nonnutnitive
diet.

As

liquids

are

a consequence,

it

it is important
to promote
up to the age at which the

requires

growth

complementary

and

foods

to

development.

The

ARTICLES

879

dilemma
created
because
the weanling
is protected
from
diarrhea
by delaying
supplementation,
but
needs relatively
early supplementation
for adequate
growth,
leads to a certain
ambiguity
in this age-ofsupplementation
recommendation.
The
need
for
supplemental
foods almost
certainly
occurs
before
5 to 6 months
of age but could come as early as 3
to 4 months.
The benefits
of receiving
breast
milk alone or in
combination
with other
foods or nutritive
liquids
become
significantly
smaller
during
the second
half
of infancy.
At these
later ages when
the infants
immunologic
system
is more mature,
the quantitative effect of passive
immune
factors
is diminished,
and essentially
all infants
are exposed
to pathogens
in the diet, so the benefits
of breast-feeding
as a
protection
against
diarrhea
become
relatively
small.
Because
of the extensive
controls
for other environmental
and seasonal
exposure
and susceptibility
variables
used in this empirical
study,
the results
can be taken
to indicate
that
exclusive
breastfeeding
will be associated
with reduced
diarrhea!
morbidity
study
has

in a wide
used
more

range
precise

of environments.
categorizations

The
of in-

fant feeding
than past work related
to diarrhea,
has
considered
carefully
the sequencing
of feeding
and
diarrhea!
morbidity,
and has controlled
for a cornplete set of confounding
variables.
That
the estimated
impacts
of breast-feeding
in diarrhea
reduction are so robust,
given the statistical
care taken,
is strong

evidence

of the

positive

value

of breast-

feeding
on infant
health.
The case study has found differences
in the protective
effects
of breast-feeding
in urban
and rural
environments.
Infants
in urban
areas
appear
to
receive
a slightly
greater
benefit
from
exclusive
breast-feeding.
ity of exposure

This
may
to diarrhea!

reflect
a greater
probabilpathogens
in the urban

environment,
perhaps
through
some pathway
left
unmeasured
in this study.
Urban
infants
in the
Philippine
sample
are exposed
to a more crowded
environment
with poorer
excreta
disposal
practices.
Water
quality,
which
generally
is better
in rural
areas,
seems
to have a smaller
effect
on diarrhea
(Cebu
Study
Team.
1989.
Unpublished
manuscript)0
than excreta
disposal
practices.
In general,
studies

infant

880

based

is exposed

in

low-income

to high

BREAST-FEEDING

countries

levels

AND

where

of pathogens

DIARRHEAL

the

have

estimated
than
have

a greater
benefit
studies
from
less

from
breast-feeding
contaminated
environ-

ments.

It has been estimated

elsewhere
that the effective
global
promotion
of breast-feeding
could
reduce
diarrhea!
morbidity
rates 8% to 20% for infants
up
to

the

age

global
year

of 6 months.

diarrhea!
for

Using

morbidity

this

age

1982

rates

group,

and

estimates

of

of 2.3 episodes

pen

an

estimated

1988

global population
size of 70.7 million
infants,
these
authors
estimated
that 6.5 to 16.3 million
episodes
of diarrhea!
illness per year could be averted
worldwide. Although
these
calculations
are only crude
estimates,

they

show

the

magnitude

of what

be achieved
through
from the promotion

even modestly
ofbreast-feeding.

feeding

particularly

promotion,

of exclusive

breast-feeding,

tion

worlds

of

the

almost
certainly
the age at which
which
weanling
greatest

positive
While

the

will

episodes

encouragement

reduce
of

might

results
breast-

only

diarrhea,

a porit

will

reduce
the risk of diarrhea
during
infant
growth
is most
rapid
and in
diarrhea
is believed
to pose
the

threat

to survival.

ACKNOWLEDGMENTS
This
article
is part
of a collaborative
research
project
involving
the Office
of Population
Studies,
University
of
San Carlos,
Cebu,
Philippines,
directed
by Dr Wilhelm
Flieger;
the Nutrition
Center
of the Philippines,
directed
by Dr Florentino
S. Solon;
and a group
from the Carolina

Population

Center,

University

of

North

Carolina

at

Chapel Hill (UNC-CH).

Barry
CH is the project coordinator.

M. Popkin of the UNCFunding

for parts of the

project

and

provided

design,

data

by the

R01-HD19983A,
Nestles

Wyeth
tional

collection,

National

Institutes

R01-HD23137,

Coordinating
International,
Academy
of

computerization
was
of Health
(Contracts

and

Center
the Ford
Sciences,

R01-HD18880),

for Nutrition
Foundation,
the
Carolina

the

Research,
the US NaPopulation

Center,
UNICEF,
the US Agency
for International
Deve!opment,
and The World Bank. Funds for data analyses
for this study were provided
by the National
Institutes
of Health
(R01-HD19983A)
and the Office
of Evaluations,

UNICEF.

We acknowledge
ment
Peter
ticularly,

the support,

of the UNICEF
Greaves,
Julian
Samir

Basta

advice,

staff, including
Lambert,
Roger
and

Margaret

and encourage-

Kathleen
Pearson,
Kyenkya.

Cravero,
and parDavid

Fugate,
Debora
Barnes,
Frances
Dancy,
and Emelita
Wong
provided
invaluable
assistance.
We thank
David
Guilkey
for his central
role in developing
the model and
statistical

MORBIDITY

estimation

techniques.

APPENDIX
Longitudinal

Analysis:

Residence

Structural

in Metropolitan
Explanatory

Equation

for Diarrhea

Incidence

in Week

Preceding

Survey

by Urban-Rural

Cebu, 19831985*

Variables

Urban
Coefficient

Rural
T Statistics

Coefficient

T Statistics

Endogenous
Susceptibility

Lagged

weight

velocity

(g)

Lagged

weight

velocity

inter-

acted
Gestational

with weight
age (wk)

Gestational

age

-7.57

x i0

-1.19

2.65

x 106

2.18

0.12
9.32

x 106

2.98
1.62

-7.33
2.08

i0

106

-0.64
0.91

(g x g/d)

interacted

with

-0.03

-0.64

x iO

-6.30

-0.06

age (wk x d)
Susceptibility/exposure

Feeding

practices

Exclusive

breast-feeding

1.39

-4.87

-0.87

-2.49

-3.20

-0.50

-1.32

(prob)
Breast-feeding
tive liquids

with
(prob)

nonnutni-

-0.91

Breast-feeding
with
foods and liquids
Total calories
(cal)

nutritive
(prob)

-0.17

Health
service
health care

-3.35

use:

preventive

-1.13

x iO

-8.88

-1.40

-0.23

-1.25

-0.30

-3.20

3.42

iO

-0.01

x i0

0.79

0.44

1.65t

Exposure
Personal

and

health

environmental

practice

Good quality

water

source

(prob)
Soap purchases/capita/week

-4.43

x iO

-0.07

-0.23
-1.27

-1.31

X iO

-0.95

(g)

Pathogenic

food processing

0.82

1.69t

0.20

0.43

(prob)
Exposure
Personal
and environmental
health
practice

Poor excreta
Poor excreta

disposal

(pnob)

disposal

inter-

acted
with age
Exogenous
Susceptibility
Childs
age (d)
Childs
sex (0-1)
Childs
sex interacted
(0-1 x d)

0.80
-1.24

4.85

iO

-1.94t

0.45
-1.58

2.08
iO

-1.93t

x d)

(prob

1.28

x i0
-0.04

with

age

6.26

x iO

2.25

2.05 x iO

-0.49
2.31

-0.11
5.23 x

1.96
-0.78

iO

1.01

Exposure
Animals
Animals

Baby

in house
(0-1)
under
house
(0-1)

crawling
animals

interacted
in house

-3.95

x i0

-0.09

-0.01

with

0.06

-0.44

0.04

0.72

-0.02

0.77

-0.22

0.02

0.22

0.02

0.79
0.08

(0-1)

Crowding
No.

of preschoolers

(0-6)

No.

of people/room

(0-9.5)

Community
km2)
Cumulative
before

Cumulative
with

density
rainfall
survey

-0.04

(people/
in last

2 wk

1.10

1.42

6.20

106

6.80

4.62

x iO
x 10_6

7.13

x iO

0.16

1.83

x i0

1.45

0.81

rainfall
interacted
age (in x d)

variable.

106

-6.01
0.11

Log likelihood
The units
for
The

ratio
the explanatory

significance

2.86

2.73

(in)

Others
Constant

-2.31

0.01

levels

variables
for

testing

are

-3.67
6.86

-5294.7
in parentheses.

whether

the

-3.25

coefficient

-1609.9
Prob is the
is zero

iO

-1.26

-0.55
0.15

predicted
is indicated

-0.30
4.42

probability
by

of the

t for 10%,

1 for

explanatory
5%,

and

for 1%.
ARTICLES

881

REFERENCES
9.
1. Feachem

RG,

of diarrhoeal
breast-feeding.
2.

Dimond

Koblinsky

diseases

A. Infant

Mexico
and Malaysia:
fed infant.
Hum Nutr
3.

Gussler

JD,

feeding
4.

Mock

Nevis.
Popkin
BM,
Lasky
The Infant
Feeding

New

York,

5.

Adair

LA,

6.

Guilkey

tionality

for

feeding

the rarity
AppI Nutr

T,

Litvin

J,

D,

Triad: Infant,
Mother,
NY: Gordon
and Breach;
1986
BM.

in Filipino
DK,

Popkin

Birth

10.

terminants
In press

breastof

and

Spicer

of

in Kenya,

exclusively

in Java, the Philippines,

Nutr.
1983;13:75-85

Popkin

control

promotion

practices

of the

the

1987;41A:51-64
A comparative
description

N.

practices
Ecol Food

Interventions

among young children:

WHO. 1984;62:271-291

Bull
Ashworth

HJ,

MA.

11.

infant

St Kitts-

Yamamoto
and Household.

12.

Fagela-Domingo

morbidity

mortality

13.

weight,

FAMILY

F, Gonzales

patterns:

GG.

a report

(PRIMARY

Infant

from

and

the

CARE?)

ofchildren.
Palmer
DL,

preschool

15.

Brown

Black

child

LOSE

and

in
Data.

consumption

sively
breast-fed
1986;108:677-680

DOCTORS

Islam

FAVOR

MR.

severity

rural
New

in

WHEN

York,

warm

882

BREAST-FEEDING

doctors

AND

lose favor

DIARRHEAL

when

it pays.

MORBIDITY

The

Wall

Street

Journal.

of

NY:

models

climate.

IT PAYS

January

2, 1990.

status

with

diarrhoea
Larwet.
Cambridge
using

R, deRoma#{241}a

The popular
labor-relations
image at a time of exploding
medical
expenses
is
that of employers
trying
to shift costs to workers,
who in turn fight to keep as
many
benefits
as possible.
But a new study
by National
Research
Corp.,
a
health-care
research
firm in Lincoln,
Neb., suggests
that many employees
may
be more flexible
than previously
thought.
In a nationwide
survey
of 1,000 adults,
National
Research
found 41% would
be willing to choose
a doctor
from a list approved
by their employer
in exchange
for a lower insurance
premium.
Thats
up from 28% just two years
ago. In
response
to a similar
question,
32% said they would
be willing
to select
a
physician
from a list supplied
by a hospital
if they could save 10% on the cost
of an office visit.
Health
costs are starting
to hit peoples
pocketbooks,
says. . .a senior
vice
president
at National
Research.
Their
premiums
and deductibles
are going
up. While
many
patients
still take
a money-is-no-object
attitude
toward
specialists
or surgeons,
she says, they are growing
more
receptive
to saving
money
at the general-practitioner
level.
In effect,
. . .theyre
telling
their family doctors,
Youre good, but youre not
that good.

Ruffenach
G. Family
Health
Costs.

Nutritional

Nigeria.

and hydration

infants

of the
status

in patients

Press;
1986
GS. Limited
dependent
variable
J Hum Resources.
1987;22:307-338
KH, de Kanashiro
HC, del Aguila

RE. Milk

de-

J Epidemiol

A prospective
study
to the nutritional

of diarrhea

children
ofPanel

Am

in

1981;114:284-292

among

University
Maddala

Study.

AKMJ,

Tomkins

data.

Cebu

E, Huffman
SL.
diseases
according

of severity
J Infect Dis 1976;134:8-14
A. Nutritional
status

1981;1:860-862
Hsiao
C. Analysis

Philip-

the

Am J Epidemiol.
Koster
FT, Alam

14.

16.

health:

a determinant

cholera.

maturity,
and proporinfants.
Hum Biol. 1988;60:319-339
BM,
Akin
JS, Wong
E. Prenatal
care

C, Aguila

and

of child

Chen
LC, Huq
risk of cliarrheal

status:

M.

and pregnancy
outcome
in the Philippines.
J Dev Econ.
1989;30:241-272
7. Wong E, Popkin
BM, Akin JS, Guilkey
DK. Accessibility,
quality of care, and prenatal
care use in the Philippines.
Soc
Sci Med. 1987;24:927-944
8.

pines.
J Singapore
Paediatr
Soc. 1987;29:32-40
Fernandez
ME,
Popkin
BM.
Prelacteal
feeding
patterns
the Philippines.
Ecol Food
Nutr.
1988;21:303-314
Cebu
Study
Team.
The socio-economic
and biomedical

panel
GL,

of excluJ Pediatr.