This is the old United Nations University website. Visit the new site at http://unu.edu
Early supplementary feeding, child development, and health policy
Ernesto Pollitt and Se-Young Oh
Selection of studies
Four criteria were used to select the studies for inclusion in the meta-analysis: subjects 24 months old or younger, random assignment to a treatment (i.e., supplementary feeding) or a control group of subjects or populations, experimental treatment restricted to the administration of a nutritional supplement, and published data allowing between-group statistical comparisons. Six studies met these criteria. They were carried out in Bogota, Colombia [13]; eastern Guatemala [10]; West Java, Indonesia [11]; Kingston, Jamaica [12]; Sui Lin, Taiwan [14]; and the Harlem district of New York City, USA [15].
Research designs
Randomization
Five of the studies were double-blind, randomized clinical trials (table 1). The sixth, in Guatemala, had random assignment of villages, rather than individuals, to one of two nutritional treatments.
TABLE 1. Research designs and statistical methods used in data analysis in the six studies analysed
Bogota, Colombia
[13]
6 treatment groups
Randomly assigned individuals
N= 141a
Griffith and Einstein tests at 4, 6,12, and 18 months
t test
El Progreso, Guatemala [10]
2 treatments
Randomly assigned villages
N=270-292 at 15 months, 219-238 at 24 monthsa
Composite infant scales
Multiple regression (SES, mother's education participation, age)
West Java, Indonesia [11]
2 treatment groups
Randomly assigned individuals
N=113
Bayley scales
Multiple regression (age, pre-intervention developmental measures
and height)
Kingston, Jamaica [12]
4 treatment groups
Randomly assigned individuals
N = 65a
Griffith tests
Multiple regression (age, sex, pre-intervention developmental and
anthropometric measures, SES, mother's age, birth weight)
Sui Lin, Taiwan [14]
2 treatment groups
Randomly assigned individuals
N=198
Bayley scales at 8 months
t test
New York City, USA
[15]
3 treatment groups
Randomly assigned individuals
N= 600
Bayley scales at 12 months
Multiple regression (age, sex)
a These numbers do not represent the total number of subjects in the original study but are restricted to those included in the analyses done here. See reference for complete documentation.
The studies in Colombia, Jamaica, and New York also included groups that combined educational and nutritional interventions (i.e., educational stimulation). Those groups were excluded from the present analysis.
Sample selection
The children in Colombia, Jamaica, Taiwan, and New York were selected by age and dietary or anthropometric criteria (table 2). The only criterion used in Guatemala and Indonesia was age; however, pregnant women also participated in the feeding programme.
TABLE 2. Characteristics of populations and subjects in the six studies
Bogotá, Colombia
Poor families
< 6 months pregnant
50% of children under 5 years old malnourished
Mother's daily intake: 1,600 kcal, 36 g protein
El Progreso, Guatemala
4 rural Spanish-speaking villages
Mother's daily intake: 1,400 kcal, 45 g protein
Children < 7 years old, pregnant and lactating women
West Java, Indonesia
Day-care centres on 6 tea
plantations
Children 6-20 months old
Kingston, Jamaica
Poor families
Children 9-24 months old with lengths below - 2 SD of NCHS
reference
Sui Lin, Taiwan
Poor women 19-30 years old
Third trimester pregnant
1 normal male child
Good maternal health
Mother's daily intake: 1,200 kcal, <40 g protein
New York City, USA
Poor black women
< 30 weeks pregnant
At risk for low-birth-weight infant
Mother's daily intake: 2,065 kcal, 79 g protein
In Colombia, Guatemala, Indonesia, Taiwan, and New York, nutritional risk was defined by a low protein and energy intake, whereas in Jamaica the criterion for inclusion was anthropometry (< 2 SD below the reference norm from the US National Center for Health Statistics).
For the purposes of the present analyses, the subjects in the six studies were classified into two groups according to age: infants and children 8-15 months old, and children 18-24 months old. The age of the subjects in Indonesia ranged from 6 to 20 months, with the mean ages of the experimental and control subjects being 12 and 13 months respectively [11]. As the mean age for all subjects (12 months) fell within the range of 8-15 months, the subjects in the Indonesia study were included in the younger group.
The meta-analysis for the younger group covered all six studies, while that for the older group was restricted to the studies in Colombia, Guatemala, and Jamaica.
Experimental intervention
The goal of the supplementary feeding in all the studies except that in Indonesia was to fill the subjects' energy and protein gap; however, the nutrient composition and volume of the supplements varied among the studies (table 3). The supplement in Indonesia was primarily characterized by its high energy (approximately 400 kcal per day) and low protein content. The role of micro-nutrients was not assessed in any of the studies; however, the experimental and control groups in Guatemala, Taiwan, and New York received equivalent amounts of vitamins and minerals per unit of volume.
TABLE 3. Recipients and composition of the nutritional supplements
Bogotá, Colombia |
|
El Progreso, Guatemala |
|
West Java, Indonesia |
|
Kingston, Jamaica |
|
Sui Lin, Taiwan |
|
New York City, USA |
|
The supplement was restricted to mothers in two of the studies: in New York it was administered only during pregnancy, and in Taiwan it was administered during pregnancy and lactation. Supplements were also provided to mothers during pregnancy and lactation in Colombia and Guatemala, but infants and young children also received them. In Indonesia and Jamaica the treatment was restricted to infants.
Developmental variables
The Bayley scales of mental and motor development were used in Indonesia, Taiwan, and New York. Two studies (Colombia and Jamaica) used the Griffiths scale. A new scale was constructed for the specific purposes of the Guatemala study, based on the Bayley and Gesell scales; psychometric data on this custom-tailored scale are published elsewhere [16].
Data analysis
The Rosenthal approach [17] was used to determine whether the studies included in the two age groups analysed tested a similar hypothesis. A test for homogeneity of effect sizes followed the same approach.
Data were analysed using weighted methods because of the wide range in the size of the samples. Results on the developmental outcomes from the different studies were combined according to the method proposed by Mosteller and Bush [18]. Effect sizes were calculated according to the procedures of Hedges [19], which weigh each study by the number of subjects.
The partial correlation coefficient between the intervention and each test was derived from the corresponding effect-size value and was used as a measure of the success rate as proposed by Rosenthal [17].
The hypothesis of homogeneity cannot be rejected for either age group regarding the mental and motor tests (table 4). Accordingly, the studies were pooled to test the hypothesis that the early supplementary feeding of energy and protein had beneficial effects on performance on the developmental scales.
TABLE 4. Results of synthesis of studies
Chi squarea |
Combined Zb |
Combined effect sizec |
|
8-15 monthsd | |||
mental | 3.489 (>.10) |
0579 (>.10) |
0.05 (.02) |
motor | 6.319 (>.10) |
2522 (<.01) |
0.181 (.09) |
18-24 monthse | |||
mental | 1.884 (>.10) |
1.610 (<=.05) |
0.182 (.09) |
motor | 0.541 ( > .10) |
3.790 ( < .0001) |
0.390 (.19) |
a. Tests of homogeneity. Figures in parentheses are p values.
b. Z = standardized normal deviate. Figures in parentheses are p values.
c. Figures in parentheses are partial r (correlation coefficient).
d. Six studies combined.
e. Three studies combined.
For the children 8-15 months old, four of the six studies showed beneficial effects of the supplementary feeding on motor development; none showed effects on mental development (table 5). For the children 18-24 months old, all three studies found significant effects of the supplement on motor development; one showed effects on mental development (table 6).
TABLE 5. Results of six studies used to calculate combined motor score and combined mental score at 8-15 months
Age (mos) |
Control group |
Experimental group |
Za |
p value (one tailed) |
|||
N |
Mean score |
N |
Mean score |
||||
Motor | |||||||
Colombia | 12 |
67 |
94.2 |
67 |
94.6 |
0.155 |
.4364 |
Guatemalab | 15 |
|
9.5 |
|
10.2 |
2.376 |
.0087 |
Indonesia | 6-20 |
38 |
103.8 |
75 |
112.4 |
2.338 |
.0096 |
Jamaica | 12 |
33 |
93.0 |
32 |
101.0 |
2.083 |
.0188 |
Taiwan | 8 |
99 |
3.3 |
99 |
3.8 |
1.934 |
.0286 |
New York | 12 |
216 |
45.81 |
201 |
45.78 |
0.087 |
.4641 |
Mental | |||||||
Colombia | 12 |
67 |
100.7 |
67 |
102.2 |
0.668 |
.2514 |
Guatemalab | 15 |
|
28.2 |
|
29.3 |
1.573 |
.0582 |
Indonesia | 6-20 |
38 |
97.9 |
75 |
96.3 |
0.373 |
.3557 |
Jamaica | 12 |
33 |
92.0 |
32 |
97.0 |
0.976 |
.1635 |
Taiwan | 8 |
99 |
4.4 |
99 |
4.5 |
0.317 |
.3745 |
New York | 12 |
214 |
99.4 |
199 |
99.0 |
-0.652 |
.2578 |
a. Standard normal deviate.
b. Only total numbers of subjects were reported: N = 292 for motor score, 270 for mental score.
TABLE 6. Results of three studies used to calculate combined motor score and combined mental score at 18-24 months
Age (mos) |
Control group |
Experimental group |
Za |
p value (one tailed) |
|||
N |
Mean score |
N |
Mean score |
||||
Motor | |||||||
Colombia | 18 |
67 |
96.6 |
67 |
101.1 |
1.792 |
.0409 |
Guatemalab | 24 |
|
9.8 |
|
10.7 |
2.765 |
.0029 |
Jamaica | 24 |
33 |
108.0 |
32 |
124.0 |
2.603 |
.0047 |
Men | |||||||
Colombia | 18 |
67 |
90 0 |
67 |
93.9 |
2.223 |
.0132 |
Guatemalab | 24 |
|
24.9 |
|
25.1 |
0.301 |
.3821 |
Jamaica | 24 |
33 |
97.0 |
32 |
102.0 |
1.028 |
.1515 |
a. Standard normal deviate.
b. Only total numbers of subjects were reported: IV = 240 for motor scores, 221 for mental scores.
With account taken of the different numbers of subjects in each study, the supplementary feeding had a significant effect on motor development scale in both age groups. Similar statistics on mental development data showed that the effect was restricted to the children 18-24 months old.
The partial correlation coefficients derived from the combined effect-size values indicate that the supplementary feeding improved motor and mental tests by 9% and 2% respectively at 8-15 months. At 1824 months the increases were 19% and 9% respectively.