Test H

Test

The Impact of Dietary Intervention on the Cognitive Development of Kenyan School Children

Shannon E. Whaley, Marian Sigman, Charlotte Neumann, Nimrod Bwibo, Donald Guthrie, Robert E. Weiss, Susan Alber, Suzanne P. Murphy, The Impact of Dietary Intervention on the Cognitive Development of Kenyan School Children, The Journal of Nutrition, Volume 133, Issue 11, November 2003, Pages 3965S–3971S, https://doi.org/10.1093/jn/133.11.3965S 

ABSTRACT


Previous observational studies in developing countries have suggested that diet quality, particularly increased animal source food (ASF) consumption, is positively associated with child cognitive development. This report presents findings from a study in rural Kenya, designed to test the impact of three different diets on the cognitive development of school children. Twelve schools with a total of 555 Standard 1 children (equivalent to U.S. Grade 1) were randomized to one of four feeding interventions: Meat, Milk, Energy or Control (no feeding). Feeding continued for seven school terms (21 mo), and cognitive tests were administered before the commencement of feeding and during every other term of feeding. Hierarchical linear random effects models and associated methods were used to examine the effects of treatment group on changes in cognitive performance over time. Analyses revealed that children receiving supplemental food with meat significantly outperformed all other children on the Raven's Progressive Matrices. Children supplemented with meat, and children supplemented with energy, outperformed children in the Control group on tests of arithmetic ability. There were no group differences on tests of verbal comprehension. Results suggest that supplementation with animal source food has positive effects on Kenyan children's cognitive performance. However, these effects are not equivalent across all domains of cognitive functioning, nor did different forms of animal source foods produce the same beneficial effects. Implications of these findings for supplementation programs in developing countries are discussed.

INTRODUCTION

Research in the past decade has documented significant associations between malnutrition and the cognitive and behavioral development of children (17). School-aged children who suffer from severe malnutrition (those with height or weight-for-height <−2 Z-scores, or with marasmus or kwashiorkor) exhibit significantly compromised reasoning and perceptual-spatial functioning, poorer school grades, reduced attentiveness and unresponsive play behavior, as compared to their adequately nourished peers (8). In addition, children suffering from mild-to-moderate malnutrition, a condition that affects over 30% of the world's children and occurs in both developed and developing countries, show significant deficits in intellectual and behavioral functioning. Deficits include compromised development in multiple domains, including verbal and spatial reasoning (5,911).

Although there is general agreement that mild to severe malnutrition significantly compromises children's developmental outcomes, the scientific community has gradually evolved in its thinking about the nutritional factors most responsible for impacting child cognition. Protein deficiency was stressed in the 1970s, lack of energy or total food intake was emphasized in the 1980s and micronutrient deficiencies and poor dietary quality are now being highlighted (12). The designs of supplementation studies in populations around the world reflect these shifts. The Institute of Nutrition of Central America and Panama (INCAP)4 study in Guatemala in the mid 1970s, for example, focused on the impact of a protein supplement on prenatal, infant and child outcomes. Individuals receiving a high energy protein-rich daily supplement (atole) were compared to those receiving a low energy supplement with no protein (fresco) (1315). Results of this study suggested that there were moderately beneficial impacts of the atole on motor scores in infancy, on perceptual and verbal skills in childhood and on later tests of knowledge, numeracy, reading and vocabulary in adolescence (4).

The INCAP study is the largest feeding intervention study carried out to date and has provided critical data about nutritional interventions in children that has had widespread application. However, whether the impacts of the high energy, high protein supplement diet were due to the protein, the energy or micronutrients remains an open question. Multiple studies have provided nutritional supplements to pregnant women, infants and/or children (1621). Although the majority of these studies share a common dependent variable—child cognitive development—the dietary intervention and additional independent variables (age at intervention, duration of intervention, etc.) have varied widely. Thus, although these studies have provided critical information about the timing of intervention (18,21), the duration of intervention (19) and the multiplicative effects of nutritional supplementation plus environmental stimulation (16,17,20), they do not enable conclusions to be made about the effect of the quality of the diet on child outcomes.

Inferences about the effect of dietary quality on child outcomes are more often based on observational studies in regions where mild to moderate malnutrition is endemic. For example, the Nutrition Collaborative Research Support Program (NCRSP), a longitudinal observational study in Kenya, Mexico and Egypt in the 1980s, found positive and statistically significant associations between children's usual intake of animal source foods (ASF) and their physical growth, cognitive function and school performance (5,9). These associations remained significant even after controlling for total energy intake, socioeconomic status (SES), parental education and social factors.

These findings showed that dietary quality, as opposed to the quantity of food energy and protein consumed, was a significant predictor of children's cognitive and motor development. The effect was apparent even among children whose energy intakes were adequate, suggesting that the micronutrient composition of animal source foods was of primary importance to optimal development. Animal source foods are nutrient dense and provide protein of high biological value, energy, fat and micronutrients. Meat in particular is rich in heme iron, zinc, riboflavin, vitamin B-12 and other micronutrients essential for normal growth and function, yet is low in vitamin A and folate. Milk is a good source of vitamin A, calcium, vitamin B-12, riboflavin and folate, yet is low in zinc and iron (22). Although adding milk to the diets of children in developing countries has been shown to increase growth (16,17), milk does not support iron and zinc status. There have been no randomized intervention studies in which meat has been fed to children and cognitive function has been measured, nor are there data comparing the effects of meat versus dairy products on cognitive function.

We have had the unique opportunity to study the impact of ASF on child cognitive development within the framework of the Global Livestock Collaborative Research Support Program (GL-CRSP) project titled “Role of Animal Source Foods to Improve Diet Quality and Growth and Cognitive Development in East African Children.” Previous research within the study area in the 1980s showed that children consumed over 75% of their dietary energy from maize and beans, 1% from milk and <1% from meat (23). Children had daily energy intakes that were only 78% of the recommended levels, with only 6% of their energy intake coming from ASF. Iron, zinc and vitamin B-12 intakes were below two-thirds of the RDA in 67, 100 and 96% of children, respectively. Stunting, anemia and low serum ferritin concentrations were prevalent.

It was clear that children who consumed the lowest amount of ASF performed least well on cognitive tests measuring verbal comprehension and perceptual abilities (5,9,24,25). In addition, those children consuming the fewest animal products were the least attentive in the classroom (5) and had the greatest deficits in linear growth (26). Thus, evidence from previous observational studies with school-aged children in Kenya strongly suggests a link between the intake of ASF and cognitive, social and physical development.

Although these results suggest that ASF are beneficial for cognitive development, a randomized intervention was necessary to prove this causal link. Previous results were based on observational studies in which other important factors were confounded with the availability of ASF. Although statistical controls were utilized to separate these factors, one can never be certain with an observational study that all confounds have been measured and controlled. Second, the question as to whether ASF are important because of their energy or micronutrient content can only be determined by comparison with an intervention providing an equivalent amount of energy alone. Third, we do not know whether milk products are sufficient to bring about improved cognitive performance or whether meat products are more beneficial. There are sound biological reasons to believe that milk and meat are not interchangeable, because they not only contribute different micronutrients but differ in their ability to enhance micronutrient bioavailability from plant source foods (22).

This study reports the results of a controlled intervention with school children in Embu, Kenya that was designed to confirm observation that ASF play a key role in the optimal cognitive development of children in this setting. The primary hypothesis addressed in this article is that children receiving supplements of ASF will perform better on cognitive measures than children supplemented with an equivalent amount of energy alone. Whether supplementation with meat or milk leads to equivalent effects is an open question. Additionally, children supplemented with meat, milk or energy are hypothesized to cognitively outperform a fourth group of children receiving no supplementation.

METHODS

Description of the study area.

The area of study includes three sublocations of Embu District in Kyeni South, 120 miles northeast of Nairobi. The total area is ∼60 km2 and is ∼30 km northeast of the town of Embu. The study area is on the southeastern slopes of Mt. Kenya, the elevation ranging from 1200 to 1460 m. Homesteads, called shambas, are scattered and are usually near or adjacent to their fields. The buildings are constructed primarily of mud with wood frames and occasionally of stone or brick. The roofs are of thatch, with the better homes having tin roofs and windows (∼10% of homes). There is no home electrification, and a gravity-fed, piped water system reaches only some of the shambas. Embu district is an area of mixed agriculture, with mostly subsistence agriculture and some cash crops, primarily coffee, cotton and tobacco. The primary dish consumed in the region is githeri, a vegetable stew composed of maize, beans, vegetable oil and some greens. There are ∼2600 households in the area of study, and the average household includes six people. Almost all individuals in the area belong to the Embu tribe and speak Kiembu. Most families are monogamous and nuclear rather than extended.

Families depend mostly on their own food production to feed themselves. Children begin regularly attending school around age 6, at which time they enter Standard 1 (equivalent to U.S. Grade 1). The school year in Kenya begins in January and is comprised of three 3-mo terms, with 1-mo breaks between terms (April, August and December). School fees and uniforms are required for all children and must be paid for by families. There are frequent solicitations for contributions for school improvements. The poorest families are unable to send their children to school on a regular basis. The school day lasts from 08:00 to 13:00 h for children in Standards 1 and 2, with a 30-min playground break around 10:30 h. Children in Standards 3–8 have a 1-h lunch break and attend school until 16:00 h. Before this study, no meals were provided at any school and children infrequently brought a snack or lunch to school.

Study design and supplemental diet.

All protocols developed and used for this study were reviewed and approved by the UCLA Institutional Review Board, as well as the Office of the President and Ministries of Health and Education in Kenya. Twelve schools in the region were selected to participate. Schools were selected based on their size and location: there had to be between 15 and 90 children in Standard 1 classrooms in the first year of the study (1998) and be accessible to a vehicle traveling on the dirt roads in the area. One school had three Standard 1 classrooms, two schools had two Standard 1 classrooms and the remaining nine schools had one Standard 1 classroom. From May to August, 1998, baseline assessments were made of all children at the twelve schools.


In August 1998 schools were randomly assigned to one of four types of feeding interventions: githeri + meat (Meat group), githeri + milk (Milk group), basic githeri with additional oil equivalent to the energy provided in the Meat and Milk groups (Energy group) or no feeding (Control group). All three supplements initially provided 240 kcal. After 1 y, the amount of the supplement was increased to 313 kcal as the children grew in height and weight. Details regarding the supplements and nutrient intakes are provided by Neumann et al., and Murphy and Gewa, elsewhere in this supplement.


Three schools were randomized to each feeding condition, and children from these schools remained in their feeding group for the duration of the study (through December 2000, the end of the Standard 3 school year). The randomization process was slightly restricted such that the three large schools with more than one Standard 1 classroom could not be randomized to the same feeding condition. In the Fall term of 1998, the nine intervention schools began receiving daily mid morning supplemental food. The timing of the intervention was designed so that the snack would not supplant either breakfast or lunch and actually supplement the child's diet.


Food preparation began at midnight from Sunday to Thursday in a centrally located kitchen developed for this study. Ingredients were measured and prepared at the kitchen by a staff of 10 cooks and weighed out into individual sealed bowls for delivery to the children in each of the nine intervention schools. Each child had his or her own numbered bowl. The kitchen staff worked closely with the drivers to ensure that the correct meal was delivered to each school. At each school, a feeding assistant supervised and observed the school feeding, took attendance and weighed and recorded any food not consumed. This procedure continued through the Standard 2 and Standard 3 school years for this cohort of children, for a total of seven school terms of feeding (21 mo).

Participants.

A total of 555 Standard 1 children were enrolled into the study at baseline (May–July 1998). After randomization of schools into treatment conditions, 134 children were assigned to the Meat group, 144 to the Milk group, 148 to the Energy group and 129 to the Control group. Thirteen children had inadequate data collected (e.g., never received cognitive testing due to school absence) and were excluded from analyses. A number of children did not receive the full seven terms of feeding, but their data was kept in the analysis up to the point at which they stopped receiving continuous feeding. Eighteen children changed schools and treatment groups mid study, 14 children refused the assigned diet mid study (eight from the Milk group and six from the Meat group). Two children died, one of severe malaria and the other from cirrhosis of the liver with chronic hepatitis.

Measures and procedures.

Multiple measures were taken from the children and families participating in this study. Although some measures were taken only once, the majority were repeated over time. For each of the following measures, the frequency of measurement is noted.

Child measures.

Cognitive assessments.

Cognitive assessments were carried out on all children at baseline and in terms 1, 2, 4 and 6 of feeding. The cognitive battery included three tests used extensively in our previous work in Embu. The Raven's Colored Progressive Matrices (27) were used to assess performance-type abilities. The child was presented with a matrix-like arrangement of symbols and completed the matrix by selecting the appropriate missing symbol from a group of symbols. The test measures the child's ability to organize perceptual detail and to reason by analogy and form comparisons. The Verbal Meaning test was similar to the Peabody Picture Vocabulary test (28), but with pictures used previously in Eastern Africa. The child was presented with four pictures and asked to point to the one named by the tester. A total of 36 items was administered to children; simpler items involved discrimination between common nouns, whereas more advanced items required knowledge of abstract concepts. An Arithmetic test adapted from the Wechsler Intelligence Scales for Children-Revised was used to assess basic knowledge of arithmetic. The 19-item test began by asking the child to add and subtract simple numbers and then proceeded to more difficult items involving division, multiplication and decimals. The total testing time for the three tests averaged ∼30 min.


Testing was carried out in the Kiembu language by six local testers. Three of the testers carried out the cognitive testing during our previous work in Embu, and all six received extensive training for this study. Testing was carried out in an empty classroom at each school in the afternoons. Daily, 12–15 children were asked to return to school after their lunch break for testing. When the children reached Standard 3 and were in class in the afternoon, they were removed from the classroom for the 30 min testing session. Occasionally, testing was carried out on Saturdays so that all study children could be tested within the 3-mo school term. Across terms, children were tested in approximately the same order so that time elapsing between testing sessions was equivalent for all children.
For all three cognitive tests, all items were administered to each child and the total number correct on each was calculated. Sixteen percent of the testing was observed simultaneously by two testers for reliability purposes. All six testers rotated partners on a weekly basis to insure reliability across all testers, and reliability testing was carried out during each round of data collection. Correlations between testers ranged from 0.99 to 1.00 for the Raven's matrices and from 0.98 to 1.00 for the Verbal Meaning and Arithmetic tests.

Anthropometry.

Children's height was measured every 4 mo and weight was measured monthly. Twelve enumerators trained in anthropometric measures carried out these measurements at school during school hours or at home in the case of absenteeism. The enumerators worked in teams of two. Intrateam and interteam measurement error was monitored by independently repeating all anthropometric measurements during the same session in a random sample (29). Further details of the collection and analysis of anthropometric data are provided by Grillenberger et al. in this volume.

Consumption of the supplemental food.

Classroom aides were hired for the feeding classrooms to assist in watching the children eat the supplement and ensure there was no food sharing or spillage. When children were absent from school, their uneaten portion was returned to the kitchen. Any uneaten food was sealed in the child's bowl and weighed in the kitchen. Aside from the 14 children who refused the diet mid study, children generally ate the meals provided.

Family measures.

Maternal literacy.

Maternal reading and writing skills were assessed during the first year of the study on 490 (88.3%) of the mothers of study children. Passages for both the reading and writing tests were selected from Kiembu textbooks used locally and reflected grade levels of difficulty. Tests were administered in either Kiembu (92.9%) or English (7.1%), based on the mother's language preference, by one of the six local testers who were trained in cognitive assessment. In the reading test, the adult read passages out loud and responded orally to comprehension questions. Writing passages were dictated twice. For both the reading and writing tests, a grade level score was calculated based on the highest level passed by the mother. Responses of 83 mothers (15.1%) were scored simultaneously by two testers, and identical scores were obtained in all instances.

Socioeconomic status.

Socioeconomic status of all families was assessed at baseline. Enumerators visited households of all study participants to complete a survey with the mother or father regarding the number and types of possessions owned by the family, years of education completed by both parents, family income, the extent of involvement in community organizations, leadership positions and the use of banks, the post office and telephones. A summary SES score was then calculated for each family based on all variables.

Data analysis.

This study is a four-condition design with three schools randomized to each condition, multiple children in each school and five longitudinal measurements per child. Hierarchical linear random effects models and associated methods (30,31) were used to examine the effects of treatment group on changes in cognitive performance over time. Random effects models of the baseline cognitive measures revealed no differences among treatment groups, nor was there a significant random school effect. The five longitudinal measurements were analyzed using a random intercept and slope (RIAS) model, which naturally accommodates child-to-child variation in initial status and in slope. Maximum likelihood methods as implemented in SAS PROC MIXED (32) were used to obtain estimates, standard errors and contrasts and to test their statistical significance.


The intervention was initiated at the same time in all three intervention groups between the first and second longitudinal measurements. Before treatment initiation, defined as time zero, all subjects were effectively in the Control group. Therefore there can be no treatment difference at time zero and any treatment effect will be manifested as a difference in the rate of increase in performance after time zero. This is accommodated for in the model by including a treatment by time interaction and not including a main effect of treatment, which would inappropriately allow for treatment differences at time zero. All subjects were also coded as controls before time zero. In addition to time and treatment by time, the models also included sex and baseline age as predictors. As a secondary set of analyses, child height, maternal literacy and family SES were also examined as covariates. These variables were added separately to the main models to retain the largest possible sample sizes for each analysis.

RESULTS

Table 1 presents means and standard deviations of baseline cognitive measures for all children and by treatment group. Analyses revealed no significant differences between treatment groups on baseline measurements, indicating that the randomization of schools to treatment conditions was successful in obtaining comparability across groups. Children's heights and weights reflected significant stunting and underweight in the population across all groups (33). Twenty-three percent of the boys and 15.5% of the girls were stunted (height-for-age Z-scores at or below −2). Approximately 30% of the children (30.2% of boys and 30.6% of girls) were moderately to severely underweight (weight-for-age Z at or below −1).

TABLE 1Baseline characteristics of the sample: mean (sd)

Entire sample (n1 = 555)Meat group (n = 134)Milk group (n = 144)Energy group (n = 148)Control group (n = 129)F2
Age, y 7.63 (1.43) 8.07 (1.51) 7.51 (1.35) 7.54 (1.53) 7.41 (1.22) 1.62 
Sex, % male 52.1 50.0 54.2 52.0 51.9 0.48 (χ2
SES 82.15 (24.98) 84.33 (22.96) 77.74 (23.73) 81.59 (22.91) 85.46 (29.67) 1.02 
Weight, kg 20.14 (3.00) 20.70 (4.02) 19.91 (2.54) 19.97 (2.86) 20.02 (2.30) 0.42 
Height, cm 115.77 (6.26) 116.91 (6.28) 115.57 (5.96) 115.12 (6.99) 115.60 (5.56) 0.10 
Mother's reading, grade 6.81 (5.25) 6.47 (5.02) 7.44 (5.18) 6.74 (5.33) 6.59 (5.47) 0.43 
Mother's writing, grade 5.13 (4.96) 4.47 (4.82) 5.80 (4.99) 5.10 (4.94) 5.14 (5.07) 1.14 
Raven's (0–30) 17.31 (2.56) 17.28 (2.35) 17.12 (2.61) 17.26 (2.91) 17.61 (2.28) 0.29 
Verbal Meaning (0–40) 27.04 (4.85) 27.57 (4.71) 26.29 (4.98) 27.09 (5.22) 27.27 (4.36) 0.73 
Arithmetic (0–19) 7.03 (1.68) 7.14 (1.75) 6.80 (1.50) 6.78 (1.79) 7.44 (1.60) 1.67 
SES, socioeconomic status.
1 Number of subjects assigned to each condition. Actual n for each analysis varied depending on missing data.
2 All P > 0.10, df = 3.
Using the analyses described previously, longitudinal growth curves were calculated for all children on each of the three cognitive tests across the five time points when cognitive tests were administered. Given the a priori hypotheses that intake of ASF would lead to better cognitive performance, children from the Meat and Milk groups were expected to outperform children in the Energy group, who were expected to outperform the children in the Control group. With these unidirectional hypotheses, one-tailed tests were used for analyses. Two-tailed tests were used to compare the Meat and Milk groups, as no prior data suggested one form of ASF would lead to better performance than the other.


As shown on Figure 1 and Table 2, there were significant group differences on the Raven's Progressive Matrices [F(3,1509) = 3.98, P = 0.01]. Post hoc analyses revealed that children who received the supplement with meat showed significantly greater gains on the Raven's Progressive Matrices than all other groups. Contrary to expectation, children in the Energy and Milk groups did not outperform children in the Control group. As shown on Figure 2 and Table 2, the four groups of children did not differ in their performances on the Verbal Meaning test, [F(3,1509) = 0.69, P > 0.10]. Group differences were evident on the Arithmetic test [F(3,1509) = 3.34, P = 0.019]. Children in the Energy and the Meat groups outperformed children in the Control group. Unexpectedly, children in the Energy group outperformed those in the Milk group (see Table 2 and Figure 3).


FIGURE 1Raven's scores by group over time.Raven's scores by group over time.
Raven's scores by group over time.


FIGURE 2Verbal Meaning scores by group over time.Verbal Meaning scores by group over time.
Verbal Meaning scores by group over time.


FIGURE 3Arithmetic scores by group overtime.Arithmetic scores by group overtime.
Arithmetic scores by group overtime.



TABLE 2Base models for the Raven's, Verbal Meaning and Arithmetic tests

Growth rate/slope (Points per year)Standard errort-valueP-value
Raven's:     
    All groups 1.05 0.08 13.92 <0.001 
    Meat 1.41 0.15 9.59 <0.001 
    Milk 0.73 0.14 5.19 <0.001 
    Energy 1.00 0.14 7.04 <0.001 
    Control 1.07 0.14 7.47 <0.001 
Treatment slope comparisons     
    Meat vs. Control 0.34 0.20 1.69 0.045 
    Meat vs. Energy 0.41 0.20 2.06 0.020 
    Meat vs. Milk1 0.68 0.20 3.44 <0.001 
    Milk vs. Control2 −0.34 0.20 −1.69 ns 
    Milk vs. Energy −0.27 0.20 −1.40 ns 
    Energy vs. Control −0.07 0.20 −0.33 ns 
Verbal Meaning:     
    All groups 3.97 0.09 42.33 <0.001 
    Meat 4.03 0.17 23.72 <0.001 
    Milk 3.89 0.16 23.94 <0.001 
    Energy 4.12 0.17 24.83 <0.001 
    Control 3.83 0.17 23.06 <0.001 
Treatment slope comparisons     
    Meat vs. Control 0.20 0.23 0.88 ns 
    Meat vs. Energy −0.09 0.22 −0.39 ns 
    Meat vs. Milk1 0.14 0.22 0.65 ns 
    Milk vs. Control 0.06 0.22 0.25 ns 
    Milk vs. Energy −0.23 0.22 −1.07 ns 
    Energy vs. Control 0.29 0.23 1.28 ns 
Arithmetic:     
    All groups 0.96 0.03 27.64 <0.001 
    Meat 1.02 0.07 14.75 <0.001 
    Milk 0.87 0.07 13.17 <0.001 
    Energy 1.10 0.07 16.37 <0.001 
    Control 0.85 0.07 12.54 <0.001 
Treatment slope comparisons     
    Meat vs. Control 0.18 0.10 1.84 0.033 
    Meat vs. Energy −0.08 0.09 −0.83 ns 
    Meat vs. Milk1 0.15 0.09 1.59 ns 
    Milk vs. Control 0.03 0.09 0.30 ns 
    Milk vs. Energy2 −0.23 0.09 −2.47 0.014 
 Energy vs. Control 0.26 0.10 2.69 0.004 
ns, not significant.
1 Two-tailed test.
2 Hypothesized as one-tailed test—results were counter to expectation, thus two-tailed tests were used.

The impact of additional factors on cognitive outcomes.

As expected, children improved on all three tests over time, as evidenced by highly significant positive slopes (see Table 2). In addition to the main analyses, multiple covariates were entered in the models for the three cognitive outcomes and slopes and significance tests are shown on Table 3. Child's age at baseline showed a trend toward significance on all three outcomes, with older children exhibiting slightly higher scores than younger children. Older children also showed greater increases over time than younger children on the Verbal Meaning test. Boys outperformed girls on the Raven's and Verbal Meaning tests, and boys showed a higher increase in Raven's scores over time than girls. Child height was a significant predictor of Verbal Meaning scores, with taller children exhibiting higher scores, but shorter children showed greater gains over time on this test than taller children. Child height had no relation with performance on the Raven's or Arithmetic tests.





TABLE 3Slope estimates of relations between covariates and cognitive outcomes (estimate, standard error)
CovariateRaven'sVerbal MeaningArithmetic
Child age 0.104 (0.059)3 0.175 (0.105)3 0.069 (0.041)3 
Child age × time 0.051 (0.057) −0.276 (0.070)1 −0.038 (0.027) 
Child sex 0.499 (0.165)1 0.940 (0.290)1 0.090 (0.110) 
Child sex × time 0.309 (0.148)2 −0.081 (0.187) −0.049 (0.072) 
Child height 0.012 (0.017) 0.127 (0.030)1 0.012 (0.011) 
Child height × time 0.023 (0.012) −0.045 (0.016)1 −0.008 (0.006) 
SES 0.008 (0.003)2 0.025 (0.006)1 0.005 (0.002)2 
SES × time 0.001 (0.003) 0.004 (0.004) 0.001 (0.002) 
Maternal reading 0.031 (0.017)3 0.099 (0.029)1 0.019 (0.012)3 
Maternal reading × time 0.002 (0.015) 0.003 (0.019) 0.019 (0.012) 
Maternal writing 0.037 (0.018)2 0.091 (0.031)1 0.010 (0.012) 
Maternal writing × time 0.010 (0.016) 0.020 (0.020) 0.012 (0.007) 
SES, socioeconomic status.
1 P < 0.01.
2 P < 0.05.
3 P < 0.10.
In addition to child level variables, family variables were related to child cognitive performance. Family SES was a significant predictor of cognitive scores, with children from higher SES families exhibiting higher scores on all three tests. Maternal reading and writing scores significantly predicted child Raven's and Verbal Meaning scores. Children of mothers with better reading and writing ability performed better on these two tests, and maternal reading ability showed a trend toward predicting child arithmetic scores.

DISCUSSION

Supplementation with ASF had positive effects on Kenyan children's cognitive performance. However, these effects were not equivalent across all domains of cognitive functioning, nor did all forms of ASF show the same beneficial effects. The most striking finding is the significant impact of supplementation with meat on performance on the Raven's Progressive Matrices. The Raven's is widely used as a culturally reduced test of fluid intelligence, thus tapping into on-the-spot reasoning and problem-solving ability as opposed to accumulated factual knowledge. As would be expected, all children in the study showed improvement in Raven's performance over time, but the children supplemented with meat showed significantly greater increases in general problem-solving ability than any other group. This suggests that increasing energy intake alone is not sufficient for improving cognitive performance, and that the quality of the diet is important. In addition, meat and milk supplementation do not appear to be interchangeable—children supplemented with meat significantly outperformed children supplemented with milk in problem-solving ability.


In contrast, supplementation of any kind had no impact on verbal performance, a skill that is generally believed to illustrate the accumulation of factual knowledge, and in the category of crystallized intelligence. Children showed comparable gains in verbal ability over time, regardless of the type or presence of supplementation. However, the effects of supplementation on arithmetic skills, also a form of crystallized intelligence, showed a different pattern from those for either the Raven's or Verbal Meaning tests. Over time, children supplemented with meat or energy outperformed the children who did not receive supplementation. In addition, children supplemented with energy outperformed the children supplemented with milk. Thus, it appears that both diet quality and diet quantity are important predictors of arithmetic performance.


These data support previous studies in Kenya, Guatemala and other countries showing the importance of diet quality on child developmental outcomes (4,5,7). In addition, these data suggest that meat, milk and energy are not equivalent dietary supplements for children facing mild to moderate undernutrition and multiple micronutrient deficiencies. Although the results do not allow clear conclusions to be made about the mechanisms underlying the effects of supplementation on cognitive development, they suggest that meat provides a more bioavailable source of micronutrients than the other supplements. Alternatively, or in addition, meat may provide micronutrients that are particularly deficient in this population of children. Meat provides a combination of iron, zinc and heme iron that was not available in the other diets. Heme iron improves the bioavailability of iron and zinc from cereals (34). Thus, it is possible that supplementation with meat contributed to improvements in micronutrient status, facilitating gains in specific domains of cognitive functioning.


There is a growing research literature suggesting that cognitive functioning does not reflect a unitary neurological variable but is a set of some limited number of different information processes that are brought to bear on a cognitive task (35). Some studies have shown that performance on cognitive tasks such as the Raven's, which tap into fluid intelligence, is mediated by processing speed and working memory (36,37). Individual differences in processing speed are posited to have a direct effect on working memory capacity, which, in turn, is a direct determinant of individual differences in fluid ability (35). It is possible that the micronutrients provided in the meat supplement contributed to increased processing speed, leading to the significant increase in Raven's performance.


Although food intake is a crucial factor contributing to the cognitive development of children in developing countries, one of the primary deficits in many previous studies examining the associations between nutrition and cognitive development is the limited attention paid to contextual factors. In addition to nutritional intake, data from this study also highlight the important effects of SES and maternal literacy on child cognitive outcomes. Regardless of the supplemental diet, children from higher SES families and children with more literate mothers showed superior performance on all the cognitive measures. Correlations between SES and nutrient intake are generally strong, with children from higher SES families likely to have more and better quality food available in the household (7,38). In addition, recent studies suggest that maternal education is an additional predictor of child food intake quality, with more educated mothers providing more ASF to their children (39). Thus, regardless of the intervention provided in this study, the adequacy of children's diets is complex and multidetermined, involving both resource availability and caregiver characteristics.


Another contextual factor to be considered is the potential role of school differences. Data analyses revealed no significant random school effect, suggesting that school-to-school variations did not account for the significant intervention effects. However, a more detailed investigation of school quality would be a significant contribution to the investigation of the impact of nutritional intervention on cognitive performance. Other studies in developing countries (40) have shown school quality indicators such as class size and public versus private education to predict scores on cognitive tests. Although the data analyses for this study do not lead us to conclude that school factors played a central role in the relationship between food intake and cognitive performance, further examination of school quality may shed light on its potential role as a mediator of nutritional intervention effects.


A limitation of the current study was that once the supplemental feeding began, it was not possible to maintain complete blindness of all the enumerators in the field as to children's feeding condition. Although blindness is clearly the most desirable for any study involving the comparison of groups, we are aware of no community-based studies of this magnitude and duration that have successfully maintained blindness of virtually all project staff. Multiple protocols were established to ensure that testers did not alter their procedures for different groups of children or change their methods over time. These ranged from weekly reliability checks between observers designed to ensure that tests were conducted the same for every child in every group, to rotation of observers to all schools and monthly reliability checks with a supervisor to minimize observer bias. Perhaps more importantly, although enumerators were not blind to feeding condition, they were blind to the study hypotheses. The study was known as a feeding intervention, with no discussion in the field of the relative benefits of one type of feeding over another.


To our knowledge, this is among the first interventions to have supplemented children with three different diets across seven school terms. Past studies suggested that ASF were an important correlate of cognitive performance, but this is the first experimental demonstration of the efficacy of meat supplementation for child cognitive performance. Results of this study have significant practical implications for families and policy surrounding the feeding of undernourished children. Poor dietary quantity and quality, and micronutrient deficiencies, are global problems affecting both rural and urban poor in developing countries and the impoverished in industrialized nations (41). In developing countries where food security is low and children are mild to moderately undernourished with multiple micronutrient deficiencies, agricultural interventions to assist families in raising small animals for family consumption are likely to have significant positive impacts on children. Of course, cultural and health considerations must always be considered, but in this region of East Africa, rabbits and chickens are acceptable additions to the diet and can be raised in relatively small areas with minimal risk of disease to the animals or children. Intensive nutrition education would also be required to improve diet quality.


We thank all the Kenyan staff who assisted with this research and the families and children who participated in the project.







Keywords:Test   Doc ID:98697
Owner:MaryGrace E.Group:DS 472 Agriculture Sustainable Development
Created:2020-03-11 15:38 CDTUpdated:2020-03-12 10:04 CDT
Sites:DS 472 Agriculture Sustainable Development
Feedback:  0   0