Juan
A Rivera,
Jean-Pierre
ABSTRACT wasting (< 90% old children assigned
Habichi,
Three-month weight-for-length)
in four
Guatemalan
villages
that
a moderate
(Atole)
or low
10% ofthe
(RDI)
from intake
than
daily
the supplement (including
a comparable
was
portion for.
to the
KEY WORDS
group
increased
energy
and whose
10.5%
ofthe
with
low
group ofthis
confounding 199 1;54:62-8.
recovery trial
from
Subjects
were
wasting (range
large mo
(1). For and
example,
in 1980
5 y in a national
as moderately of the National
controlled
Therefore, mentary
quate analysis effect
that
of Guatemala
in children
cannot
feeding
to
ages
6
classified
weighed < 90% (NCHS)-World
to malnourished
of supplementary
attending
Beaton
trials,
children, and
feeding
3). However, control groups.
used
has
programs
on “anthropometric
in a supplementation
ofwasting
and
recovery
that
and
public
health
pro-
in nutrition
research
design
for this analysis
tween
1969
trial
and
area
1977.
were
randomly
were
collected
during
in rural
Guatemala
conducted
Detailed
selected
allocated (63
g/L)
villages
were
assigned
protein amounts
supplement of vitamins
descriptions
for
centers
and
drink
of
each
session
available community aged 0-7
to the
nearest
times
study. Atole.
The
a low-energy
Fresco. minerals
a conbe-
the
sample,
of the
villages
(3.77
MJ/L),
remaining
two
(1 .38 MJ/L)
non-
Both supplements had similar (Table 1). A preventive and
offered centrally daily,
in all four villages in supplementary on
at two y were 10 mL.
the energy
Two
a high-energy
called
to receive called and
were
members of the takes of children
the
to receive
curative health program was supplements were distributed
a voluntary
(7). The feeding
basis,
to all
3-h sessions. Individual measured and recorded The
is almost
three
children plement
in the Atole villages consumed than did children in the Fresco
I
and
given
Butz
in ade-
a positive
effect
(5) in their
meta-
concluded
that
improvement
metaanalysis
Am
From
the Instituto
CAP), Guatemala,
to supple-
Habicht when
Ghassemi
as
of this on the
content
energy
content
ofFresco. larger villages.
inat
of Atole
Furthermore, volumes of supAs a result, the
nutrition-re(2,
solely
feeding,
surprisingly small.” Much of their to seek reasons for this small effect.
62
were
be attributed
supplementary
In contrast,
ofsupplementation
referred
between
and
supplementation
graphical
sup-
of children their height is
of children
from 40% to 80% centers had adequate
ofsupplementary
amounts
on growth.
range in these
Rrs reported feeding.
In a review
often
wasted because they for Health Statistics
(PEM)
centers studies
(4) concluded
are
programs
objective feeding
methods, and quality control were published elsewhere (6). A brief description follows: Four rural Guatemalan villages that were similar in population, ethnicity, development, and geo-
(WHO) references for weight-for-height observations, 1980). rates (Rrs) from mild to moderate protein-
malnutrition
habilitation none ofthe
25%
sample
to severely Center
Health Organization (INCAP, unpublished Reported recovery energy
children
the
methods
trolled
were
wasted
wasting
indicators
employed
and
The data
pro-
recovery,
feeding programs. The proportion as evidenced by low weight given
to moderate using
Study population
supplemental
trial
to severely
feeding
grams.
Introduction Moderately
mild
commonly
high-protein
supplementary who are wasted
of supplementary
higher
This
malnutrition,
importance
in Guatemala,
are
total
RDI
whose recovery
variables
the
of energy
supplementation.
Protein-calorie field
randomly
intake
group)
was
in the high-Atole recovered. Much
rose after potential Am J Clin Nutr
plementation
diet)
high-Fresco
due
been (Fresco)
dietary
(high-Atole home
had
Given
strategies for the control of undernutrition, study was to quantify the effect ofsupplementary
(Rr) in the Atole villages was villages (P < 0.05). This effect in the Atole villages who con-
recommended
energy intake. All those was due to malnutrition 29-52%)
S Robson
rates from moderate compared in 6-24-mo-
supplement. The recovery rate 12% higher than in the Fresco was above all due to the children
energy
Douglas
recovery were
to receive
sumed
and
from mild
the was
de Nutnci#{243}nde Centro
and the Division
America
of Nutritional
y Panama
Sciences
(IN-
and Bio-
metrics Unit, Cornell University, Ithaca, NY. 2 Supported in part by grant RD1-HD-20029 from the National Institute of Child Health and Human Development to the Rand Corporation and by the Instituto de Nutrici#{243}nde Centro America y Panama (INCAP), Guatemala. Financial support for J R’s doctoral studies was received from the Consejo Nacional de Ciencia y TecnologIa, Mexico, and from the Kellogg Company. 3 Address reprint requests to J Rivera, INCAP, P0 Box 1 188, Guatemala, Guatemala City, CA ReceivedJune 1, 1990. Accepted for publication November 14, 1990.
then
went
on
J C/in
Nutr
l99l;54:62-8.
Printed
in USA.
© 1991 American
Society
for Clinical
Nutrition
Downloaded from https://academic.oup.com/ajcn/article-abstract/54/1/62/4691088 by East Carolina University user on 13 January 2019
Effect of supplementary feeding on recovery to moderate wasting in preschool children13
SUPPLEMENTARY TABLE
1
AND score,
Ingredients
and nutrient
content
of Atole
and Fresco1 Frescot
Ingredients (g/0. 18 L) Incaparina Dry skim milk Sugar
Contents
per cup; cup volume
benefit.
-
were
90%
-
>
-
was
15.3 4.0
wasted
-
the
1.1
of the
as the
and
The
whose
-
than
in the Fresco
plement daily
intake group
aged
6-24 group.
The
sample
based
18-21,
risk
was
Measures ofsiblings
were
over,
some
thus
contributing
Therefore, holds
Rrs were Rrs)
used
as the
obtained
Fresco
of energy the
Fresco
can
(RDI)
considered
aged
6-24
show
1988).
a nonsupple-
of children supplementary
Fresco
from
the
village-level two
household
children
21-24
mo.
intervals:
Because
participated
was conducted
tional
Institutes
6-9,
the study
in more
study
than
according
mo.
one
The
analyses
tailed
9- 12, 12- 15, 15-18, was longitudinal, child-age
to ethical
interval.
guidelines
The
of the Na-
of Health.
formation vantage were
24-h
take.
method
was collected
every
3 mo for children
as the unit
of the example
tion,
during
signs
of other
The groups used were those children
economic
last
and
of which
height ofboth in the house,
parents, number
breast-feeding.
Three
about
the
the mothers:
house
of signs including
demographic
well as characteristics collected,
2 wk,
the
about
house
the
and ofthe
following
were
scores
were
vocabulary
generated and
infection and
used:
mother’s parity, number of persons per room,
characteristics
a maternal
diarrhea
data
ofthe
of respiratory
fever.
child’s
parents, years
and Socio-
family,
as
were also of school,
The mean
ofpersons living and duration of
from
the information
from
interviews
and
score,
a maternal
tests
modernity
group
were
children ance
of
intake
to as high-Fresco
household subtracted
is the sum high-Fresco
was
mean villages
tested
for one-
of variance
by using
an arc sine
treatment,
about
in the
the
Atole
trans-
of the
will
effects
individual
by taking
children
iningested
in the Atole in the high-Atole
with the percentile
a
benefit
supplement
6-24-mo-old
Fresco
and Atole
villages. groups.
high-Atole groups ofthe distribution
villages.
These
are
referred
groups.
Rrs ofwasted from
in the high-Atole
In addition, the
two
receiving
RDI from the supplement are referred to as being
of supplement
children
from
the
Atole
between
by using
Atole
for comparison above the 55th
to
(1 1). This analysis takes adstudy in which two villages
children
ofthe
villages
Rrs ofwasted
receiving Fresco. was to study household
45%
10% of the These children
infections,
the
information
aged 18-24 mo. Morbidity data were collected every 2 wk during home visits. Mothers were asked about the presence and durathe
to the
that
(mean compared
Fresco
compared
difference and
children
were
subtracted
the variances design ofthe
more than the children The second approach For
This
follow.
of RDI
occurred.
by an analysis
assigned
household that
households
and
from
ofanalysis
to stabilize ofthe original expectation
advantage
AB.
significance
household
of wasted
had
were
children
overall
randomly
priori
The data relevant to this paper were collected as follows: weight and length were obtained every 3 mo for the first 2 y of age. Home dietary intake (excluding breast-feeding) obtained by a dietary-recall
the
(mean
diet
household
villages
house-
rates
Atole
Rrs were
mean
sample.
within
averages
feeding
of wasted
statistical
the village
some
Fresco
Rrs
to compute included
The
the
effects,
The
within
ofthese
effects.
villages.
as the
in the
More-
the percentages
averaged
means
true
Overall and
The
in
for the calculations
as well were
Rrs
a number of wasting,
cluster
ofthese
used
ofthe
children.
observation
supplementation,
These
attributable
episodes
of analysis.
were
groups that
ofwasted
these
Rr
(10).
because
repeated
the means
supplement
% RDI).
between
(8) for the in children
observations,
be
sup-
1 1% of the
group
unpublished
group
daily
was
the
obtained.
the
oth-
in the estimation
one
unit
actual
initially
from
to the
not independent
than
and
Rr
were
exposure
consideration
variances
who
is analogous
in the group
more
experienced interval was
by subtracting
experienced into
and their
household
were
to take
were
This
included
children
to have of the
the Rr of a nonsupplemented,
Observations
intervals
of
was 90% at the end of 3 or the effect attributable to
(AB),
by a pernicious
ofeffects.
beginning
expected weight, reference stan-
of weight-for-length.
obtained
group. caused
child-age
at the
of children
group
comparable to a disease
from
in the Atole
the mean
children
intake 1% for
on six child-age
and
children
only
is higher
example,
wasted
(J Rivera,
the
mented
For
dietary
and
mo
Therefore,
were
villages.
in moderately
recommended
Atole
the supplement
obser-
rates, and attributable
end
references
benefit
supplementation,
erwise
1.5 18.5 1.2 5.0 0.2
considered at the
weight-for-length
of the supplemented
were from
as wasted
proportion
attributable
(8) from intake
unpublished
ifthey weighed < 90% ofthe according to the NCHS-WHO
To determine
energy
recovery
NCHS-WHO
defined
mo.
t Both preparations were distributed daily from January 1, 1969, to February 28, 1977. Vitamins and minerals were added to the Fresco on October 1, 1971. Institute of Nutrition ofCentral America and Panama, Guatemala City.
average
recovery, classified
dard (9). A wasted child was recovery if its weight-for-length
247
0.18 L.
=
ofwasting,
Children
the 3-mo intervals given their length,
13.3 2.1
682 11.5 0.7 27.8 4.0 0.4 1. 1 0.3 1.5 18.5 1.2 5.4 0.2
(J Rivera,
Analytic approach
-
-
score
the
ofthe mean groups
the group
two
variances compared.
in the high-Fresco
household
to obtain
household was
children
mean
ofthe Rr
Rrs
the AB. Rrs being
of the The
of wasted
The
high-Atole difference
AB vari-
subtracted. and between
of
Downloaded from https://academic.oup.com/ajcn/article-abstract/54/1/62/4691088 by East Carolina University user on 13 January 2019
Flavoring agent Nutrients (per 0.18 L) Energy (U) Protein(g) Fats(g) Carbohydrates (g) Ascorbic acid (mg) Calcium (g) Phosphorus (g) Thiamine (mg) Riboflavin (mg) Niacin(mg) Vitamin A (mg) Iron(mg) fluoride (mg)
a house-quality 1988).
Definition
13.5 21.6 9.0
63
RECOVERY
and
vations,
Atole
S
FEEDING
64
RIVERA
the
means
was
was tested To show
parison old
converted
how
were
the 55th
the
ABs
used,
in the
high-Fresco
nificances
were
Controlfor high-Fresco
computed
the
that
antee
equal
among
ofall
and
Logistic-regression
and
and
for the
which
statistical
groups,
sig-
in terms
groups that
of
were
on recovery
from
confounding
and
other
variables
that
identified
after
Rrs for the well-supplemented (high-Fresco)
(12).
independent
gistic-regression
(high-Atole)
groups
coefficients.
were
The
de-
causes
For this
and
estimated
poorly
by using
estimation
of
supthe lo-
logistic-regression the
clustering
the
regression
coefficient. effect
the
model
used
coefficient.
on the difference and
was
household’s
mean
described standard
above, errors
Therefore,
a correction
to multiply
the
The
between
effects of the
correction
predicted
(CF)
for
errors
of
is based
Rr for the household
Rr values
from
in time
I
(n(MRr x*(1x) H-K
-
J
will
high-Atole
group
ofregression
rates
of the
period,
as a result
from
by
100.
normal
ofthe
weight-for-length
not due This
frequency
areas
the curve
were
values
was estimated
in this
the
available set but
the
requisites
weight-for-length
(B) is the binomial expected variance
data
distributed
in this
ofthe
measured
variance
(16).
values
for the weight-for-height
the
2-wk
reliabilities
Fieller’s
formula
from
the
for
the
for this study in et al (17). These (18)
weight
to calculate
and
height
2brSS(RR)
-
re-
observed
in reference
+ b2SRJ/i2S
weight, (17),
R is the
S2 is the
reliability
variance,
calculated b is the
weight-for-length, r is the correlation coefficient observed weights and weights from standard given lengths, and i is the mean weight from standard given length (9).
mean
between observed observed
Results Household the
Atole
Fresco if
cumu-
as follows:
y is the
from
for reliability
the
the distribution
proportion ofthe
the curve 0 to Z, both
from
were normally
from
data
obtained
weight and length measurements are available a publication (15) that used data from Martorell meet
mea-
(ZsqR), towards
from
(1 1) because
variability
is not directly
measurement
other
in the
the area under
under
distribution
to week-to-week
reliability
on
and
was also obtained due to regression
by subtracting the area
These
lative
respectively,
ofimprecision
of unreliability
was estimated
0 to ZsqR
where
where ;r indicates summation over households, n is the number of observations per household, H is the total number of households, K is the number of parameters in the logistic-regression model, MRr is the mean Rr for each household, and is the mean ofthe predicted probability ofrecovery for all observations within a household from the logistic-regression model. The numerator (A) is the squared difference between the observed and the predicted (from the logistic model) rate of recovery of the (the cluster). The denominator for the household, which is the
(14),
at one
overestimation of this fraction. This peras follows: 1) the standard normal deviate 90% in the distribution ofpercent weightobtained for each age interval in the highgroups, 2) the product of the standard and the square root of the measurement
the mean from
or lower,
is because
of weight-for-length (sqR) the proportion of recovery
[SR
H-K
to be higher This
reliability 3) finally,
the
I
tend
the mean
or overestimated
percent
household variance
in
Rr
towards
that is under-
surement, is a slight centage was obtained (Z) corresponding to for-length values was Atole and high-Fresco normal deviate (Z)
(
i)2\
the effect
measurement.
liabilities H
In the
weight-for-length
next
data
the logistic
as follows:
CF=
Rr includes
because
population. Reliability
which of the
factor
standard
factor
the observed
ofthe
group.
multiplied
the independent
variables were held constant at their mean values. The estimated Rrs from the model were employed to obtain adjusted AB values after the potential confounding variables were controlled for. The child-intervals used as units of analysis for the logisticregression models have the cluster may result in an underestimate
form.
on the recovery
includes the AB. The part ofthe Rr that is not attributable to supplementation corresponds to the Ri ofthe high-Fresco group. This component the high-atole
ginning
the effect controlling
recovery.
plemented
antilog
lo-
components of unreliability reported for these children elsewhere (15). The fraction of Rr that results from regression towards the mean can be estimated. The percentage ofchildren in the highAtole and the high-Fresco groups who were misclassified as falling below the 90% weight-for-length of the standards (9), at the be-
variables
were
to the the mean
The
corrected for the odds
variables
were
pendent variable was a dichotomous variable indicating a child’s recovery within an interval from wasting as defined above. The independent variables were an indicator variable for treatment groups (high Atole and high Fresco), the potential confounding variables,
transformed toward
(13).
the corresponding confidence limits
point
to assess
wasting
then
by
the
wasting.
employed
were
ofregression
divided
of a cluster effect, both will be 1. The 95% confi-
by an approximation
coefficients and were used to obtain
is therefore
effect,
not guar-
to be or are suspected
from
a cluster
In the absence and the ratio
obtained
factor
correction
the procedure
were those variables with valbetween the high-Atole and
are known
was
The includes
does
confounding
variables
interval
that
which
variance. identical,
gistic-regression standard error ratios
clustering.
variance,
Effect
ABs.
However,
of recovery
potential
of
dence
low-Fresco,
their
similar
The
models
measured
at or
variables. The use of the with the high-Atole group
variables different
causes
of supplementation
were Atole)
the potential
groups.
groups
to be independent
who from
ofcomparison
as potential confounding ues that were significantly
for the
obtained
is no household actual
the binomial variances are
intake
low-Atole,
led to self-selection.
distribution
com-
the Rrs of 6-24-mo-
ofthe
of having
the comparison
high-Fresco
Rrs
as described
for the selection
if other
55th percentile of supplement differences between the Rrs
confounding comparison
likelihood
characteristics employed
the
were
potential group for
maximizes
the The
with
groups
been
of RDI
ofsupplement
there the
have
10%
AL
level
Rrs villages
villages effect
at the were
end 0.49
of 3-mo and
were 0.42 and 0.35 (
ofsupplementary
feeding
supplement
0.52
(
=
intervals
0.50)
0.38). The was 0. 12 (P =