Contents - Previous - Next

Malnutrition and cognitive development: evidence from experimental/quasi-experimental studies among the mild-to-moderately malnourished¹

Specificity of effects
Critical periods
Differential responsivity
Nutrients
Mechanisms
Avenues for future research
Policy implications
Literature cited

KATHLEEN S. GORMAN²

¹Paper prepared for the International Dietary Energy Consultative Group (IDECG) Task Force workshop on malnutrition and behavior at the University of California, Davis, CA, December 610, 1993. This workshop was supported by IDECG, the Nestle Foundation, Kraft Foods and the International Union for Nutritional Science. Guest editor for this supplement was Ernesto Pollitt, Department of Pediatrics, University of California, Davis, CA 95616.

²To whom correspondence should be addressed: Psychology Department, John Dewey Hall, University of Vermont, Burlington, VT 05405

Psychology Department, University of Vermont, Burlington, VT 05405

The purpose of this paper is to present what is currently state-of-the-art knowledge on the relationship between malnutrition and behavior. The particular focus is on the literature from experimental and quasi-experimental studies, i.e., supplementation studies. The data discussed are primarily from interventions conducted in the late 1960s and early 1970s in Guatemala (Pollitt et al. 1993), Bogota, Colombia (Wader et al. 1981), Cali, Colombia (McKay et al. 1978), Taiwan (Joos et al. 1983), Mexico (Chavez 1975), New York City (Rush et al. 1980) and, more recently, Jamaica (Grantham-McGregor 1991) and Indonesia (Husaini et al. 1991).

Until the late 1980s, research in the malnutrition and behavior literature focused primarily on determining the validity of this hypothesized relationship. Despite flaws in methodology, design and implementation, results from these earlier studies provided enough evidence to conclude that early supplementary feeding of nutritionally at-risk children resulted in small, yet statistically significant, effects on some developmental outcomes (Pollitt 1988). More recent data from studies with stronger designs and from follow-up analyses have contributed to a significant shift in emphasis from determining the existence of such effects to understanding the processes underlying these effects.

This paper, then, reviews the information from these studies in light of this new evidence in an attempt to understand the processes by which malnutrition may affect cognition and behavior. All the studies were selected from endemic populations of mild to moderately, rather than clinically, malnourished children. In most studies, subjects were identified by populationwide indicators of poverty and malnutrition (e.g., anthropometric indicators). With the exception of the Cali study, in which nutritional and educational components were combined, we will address the studies to the extent that data are available on nutritional intervention alone. Before proceeding, we will briefly summarize the studies, which are identified by geographical location.

The Guatemala study (Freeman et al. 1980, Klein et al. l 976, Townsend et al. 1982) consisted of an early supplementary feeding trial in four rural Guatemalan villages between 1969 and 1977. The major hypothesis being tested was whether protein deficiency during the prenatal, infancy and early childhood periods affects cognitive development among moderately malnourished subjects. Randomization occurred at the village level; two villages received a high-energy, high-protein supplement (Atole), whereas two villages received a low-calorie supplement (Fresco). Supplements were administered twice daily ad libitum to all pregnant women, infants and children 7 y of age and under, and data were collected from over 2300 subjects.

The Bogotá study (Waber et al. l 981) consisted of supplementation of nutritionally at-risk pregnant women and their infants up to 3 y of age. Food (e.g., oil, dried milk and bread) was distributed weekly in quantities sufficient for entire families. Subjects (n=433) were randomly assigned to experimental groups that received supplementation during different development periods; two groups received early stimulation as well. The major hypothesis tested was whether the timing of the supplementation (i.e., prenatal vs. postnatal) is associated with cognitive development during the first 3 y of life. A secondary focus of the study was to test the independent and interactive effects of educational stimulation and nutritional supplementation on cognition.

The Cali study (McKay et al. 1978) was a combined nutritional, health and educational program for preschool-age, chronically malnourished children. Preschoolage children (n = 133) were randomly assigned to treatment groups consisting of health care, preschool education and nutritional supplementation (e.g., meals served at the preschool). Treatment groups varied by duration of the intervention with subjects receiving between 9 and 42 mo of intervention. The major focus of this study was to test the effects of a fairly comprehensive intervention on cognitive development and school achievement in children beyond the critical period. The study design also tested the differential effects of the duration of the intervention.

The study conducted in Taiwan by Bacon Chow (Jogs et al. 1983) consisted of nutritional supplementation of pregnant and lactating women (n = 294). Women were randomly assigned to two treatment groups soon after the birth of one infant and then followed through lactation, the interpregnancy period, and the pregnancy and lactation of their next child. The supplementation consisted of a high-calorie, high-protein drink, with a low calorie beverage as placebo. All women were provided with a vitamin and mineral supplement. The hypothesis tested was whether improved maternal nutrition was associated with improved growth and development in infants.

The study in Mexico (Chavez et al. 1975) was a supplementation study of nutritionally at-risk infants during the first year of life in a small rural community. Pregnant women were provided with milk, vitamin and mineral supplementation starting at 45 d gestation. Infants (n = 20) began milk and food supplementation between the 12th and 16th wk of life, in quantities sufficient to maintain growth. Twenty additional dyads of mothers and infants born during the previous year were included as controls. This study focused on the effects of improved nutrition on mother-child interaction.

The New York City trial (Rush et al. 1980) identified nutritionally at-risk pregnant women and randomly assigned them to one of three treatment groups: high protein and calorie supplementation, moderate protein and calorie supplementation and a control. All women (n = 770) received standard multivitamin and mineral tablets. The major hypothesis tested was whether improved maternal nutrition is associated with improved infant outcomes, including mortality rate, growth and cognitive development.

The Indonesia study (Husaini et al. 1991) consisted of a 90-d supplementary feeding trial of nutritionally atrisk infants (n = 113) in day care centers located on tea plantations throughout West Java, Indonesia. Day care centers were randomly assigned to either continuing their normal dietary practices or to allowing intervention with dietary supplements in the form of nutritionally prepared snacks. Infants between 6 and 20 mo of age were included. The main hypothesis was that improved nutrition was associated with improved growth and mental and motor development.

The Jamaica study (Grantham-McGregor et al. 1991) was a 2-y nutritional supplementation of growth-stunted children between 9 and 24 mo of age. Subjects (n = 129) were randomly assigned to either a control group, a nutritional supplementation group, a psychosocial stimulation group or a supplementation plus stimulation group. Supplementation consisted of a milk-based formula provided weekly. At issue were the effects of nutritional supplementation and psychosocial stimulation on children's mental development.

As is clearly illustrated in these summaries, none of the studies is directly comparable with any of the others in its critical elements: type of supplement, duration of supplementation, recipient of supplement or outcomes assessed. (See Table 1 for descriptive data on each of the studies reviewed.) Therefore, a basic limitation of any inferences drawn is the design and methodological characteristics of the studies. For example, any attempt to make inferences about the specificity of outcomes (e.g., motor development) is limited by differences in the recipient of the supplement (e.g., mother vs. infant). This limitation in turn contributes to the difficulty of making inferences about the mechanisms involved. Similarly, our understanding of sensitive periods is limited by the specificity inherent in each study design, e.g., all but one study included supplementation during at least a portion of the sensitive period. Finally, inferences about the effects of specific nutrients will be limited by the precision with which actual subject intakes were measured.

In subsequent sections, we will address what is known about the specific outcomes affected by malnutrition, the significance of the timing of such insults, the issue of differential responsivity (i.e., individual differences), and the role of specific nutrients. We will then discuss potential mechanisms that may account for these data and conclude with a discussion of their implications for policy and future directions for research.

Specificity of effects

The first question refers to what is known about the specific outcomes or processes affected by malnutrition during each particular stage of development.

Infancy. One of the strongest conclusions we can draw regarding the specific effects of supplementary feeding is on motor development during infancy. Despite variation in the measures of motor development across studies (e.g., Bayley Scales of Psychomotor performance, Griffiths Subscales), a strong and consistent association has been reported (Grantham-McGregor et al. 1991, Husainietal. 1991, Joosetal. 1983, Pollitt et al. 1993, Waber et al. 1981). This finding is reinforced by the results of a meta-analysis that concluded beneficial effects on motor development among younger (8-15 mo) as well as older infants (18-24 mo) (Pollitt and. Oh 1994).

TABLE 1
Description of studies included in this review

¹Supplement administered to control group: N. none; MV, multivitamin and mineral supplement; C, calorie.
²Consumption quantifiable.

Education not separate from nutritional intervention.

Less consistent effects have been reported on mental development during this same developmental period with the meta-analysis indicating effects of supplementary feeding on mental development only among the older (18 mo and older) infants (Pollitt and Oh 1994). In this case, mental development is nonprocess specific.

Preschool The available evidence (i.e., Guatemala, Jamaica, Cali and Bogota) indicates that there are significant effects of supplementation on broad measures of cognitive development during the preschool period, despite differences across studies. It is important to note that the Bogota and Guatemala studies included both preand postnatal supplementation, whereas the Jamaica study was solely postnatal. The Cali study, as noted, refers to a combined nutritional and educational intervention. Furthermore, the reported effects are limited by the nature of the outcome measures used, i.e., broad measures of cognitive abilities (e.g., developmental quotient (DQ], general cognitive ability, preschool battery factor scores, vocabulary and various subscales of the Griffiths and Preschool Battery) rather than more specific process variables.

Postintervention follow-up. For three studies, follow-up data are available to test whether there are enduring effects once the intervention had been discontinued. In the first, Taiwan, children were tested on an IQ measure at 5 y of age, and no effects were reported. In Bogota, subjects were followed up at about 6 y of age, and differences were reported on achievement-related abilities (Griffiths) but not on more basic cognitive competencies (e.g., Einstein). In the longest follow-up (e.g., Guatemala) through adolescence, supplementation effects were evident across a wide variety of tests termed psychoeducational reflecting skills in such domains as reading, numeracy and other achievement-related areas. Fewer effects were evident on tests of basic processes: memory, simple and choice reaction time.

It is important to caution that the apparent strength and consistency of effects may be related to types of outcomes most frequently used and the strength of their psychometric properties. Lack of evidence for effects on information-processing measures may reflect weaker statistical properties of the tests used to measure these constructs (e.g., lower reliability and internal consistency) rather than the absence of an effect. Furthermore, it could be expected that information-processing measures, such as attention, would show an immediate sensitivity to supplementation and that these effects could potentially mediate long-term effects such as the achievement-related behaviors observed in the Guatemalan study. Furthermore, it should be pointed out that by limiting outcomes to typical developmental scales and IQ tests, those measures most sensitive to changes associated with malnutrition may have been overlooked (e.g., social and affective behaviors). This is particularly important in light of the research on early interventions that has shown declines in IQ gains over time but long-term effects on more socially mediated behaviors such as teen pregnancy, high school completion, and delinquency (Haskins 1989).

Critical periods

The underlying assumption of a critical (or sensitive) period is that there is a specific time during which development is particularly sensitive or vulnerable to insult, and that period is one characterized by rapid brain growth and development. The actual definition of the boundaries of this period (e.g., gestation, first 2 y of life) and the degree to which it is either critical or sensitive vary from study to study.

Not coincidentally, with the exception of the Cali study, all of the studies addressing malnutrition and behavior focused the intervention on at least some period considered part of the sensitive period, and only the Bo gotá study design allowed for testing contrasting periods of sensitivity. As a consequence, although most of the reported effects occur during periods considered sensitive, the majority of the designs do not allow for discounting other periods or actually determining the boundaries of these effects.

Prenatal Those studies that actually built in a specific test of the critical period hypothesis did not find evidence to support it. For example, data from Bogota (Waber et al. 1981) found that supplementation during the first 6 mo was not critical (e.g., subjects who began supplementation after 6 mo of age did comparatively better) and more importantly, was not sufficient (e.g., if treatment was not continued beyond 6 months, the benefits were lost).

Postnatal The significant effects reported in the Bogota study of supplementation beginning after a critical period (defined as 6 mo) were also supported by data from Jamaica (intervention after 9 mo) and Cali (after 3 y), although, again, the Cali intervention included an educational intervention as well.

Prenatal and postnatal In Guatemala, those with both preand postnatal supplementation over extended periods of time (e.g., minimum 2 y) evidenced the greatest effects, corroborating the data from other studies (e.g., Bogota). Because of design limitations, small sample sizes precluded the ability to test hypotheses with regard to specific groups (prenatal only and postnatal only).

We can conclude, then, that the periods of gestation and the first 2 y of life continue to be important periods in the development of young children and have important implications for subsequent development. The evidence indicates, however, that nutritional interventions can be effective even when initiated outside of this sensitive period and that our consideration of nutritionally at-risk children should be expanded to include children of all ages. Furthermore, given the mixed findings regarding prenatal supplementation alone, efforts at improving infant nutrition should focus on duration as well as early targeting.

Differential responsivity

The most common analytic strategy employed in most of the studies reviewed has been to assume a linear relation between nutrition and outcome, with the focus being on testing between-group differences - very few attempts have been made to examine the role of individual differences eg., potential interactions). However, in keeping with current developmental theory as well as the available evidence, those studies that tested for interactions provide some interesting information regarding important individual differences that may affect responsivity to nutritional interventions.

For example, data from Guatemala show that children from lower socioeconomic status (SES) families were more likely to respond to treatment than those from upper SES families. The question that these analyses raise is whether the effects observed were a function of greater nutritional need or of greater consumption. In other words, did poorer children consume more and hence show greater improvement (i.e., dose response), or did children consume equivalent amounts but improve more because of greater initial deprivation, hence, a catch-up type mechanism (cursory attempts to answer this question suggest support for this latter explanation). The evidence from Bogota - that families with greater resources had children who responded most significantly to the supplementation, although in the opposite direction - raises similar questions. What were the characteristics of the children in the Bogota study (those from families with greater resources) that allowed them to benefit from the supplement? Although it is less likely that the effects were purely nutritional (dietary) - families with greater resources could be expected to have had better diets - it may be that supplementation allowed for greater utilization of either nutritional resources or other behavioral resources. Methodologically, the data do not provide definitive answers to these questions.

In the only other study to explore the issue of interactions (Jamaica), the authors focused their discussion of interactions on differences between treatment types (nutritional vs. psychosocial) without addressing specifics of individual differences. Further analyses are warranted to address these issues.

Beyond issues of differential:responsivity based on SES and family background, individual characteristics reflecting the health and nutritional status of subjects may be important factors to consider. For example, among populations where the incidence of parasitic infection is high, malnutrition rates also tend to be high, and malnourished children are less able to ward off parasitic infections. Furthermore, treatment of parasitic infection has been associated with increased appetite, food intake and growth (Stephenson 1989, Stephenson et al. 1989). Similarly, because the role of micronutrients is particularly relevant to infection (e.g., iron and blood loss), the confounding role of infection and the potential role of interactions between nutritional status and infection must be considered.

Nutrients

One of the most important advances in our understanding of the relations between malnutrition and behavior has been in the shift in recent years from a focus on either calories or protein to a more complex understanding of nutritional needs and diets. Despite these advances, all of the studies reviewed here were designed with the notion of calories (and in some cases protein) as the limiting factor. Consequently, they are seriously limited in their ability to address concerns relating the specific roles of micro-vs. macronutrients or dietary quality. There are three major concerns in terms of our understanding of nutrients that need to be addressed.

The increased recognition of the importance of both quality and quantity of intake has not been paralleled with adequate measurement of intakes. To our knowledge, only two studies coded the actual consumption of the supplement: Guatemala and Taiwan. In Guatemala, analyses led to the conclusion that no one type of nutrient (e.g., protein, calorie or micronutrient) was solely responsible for the observed effects (Oh, S.-Y. and Pollitt, E., unpublished results). In New York City, Bogota and Cali, the consumption data were estimated from dietary recall. In general, these measures have low reliability.

In addition to the issue of the accuracy of measuring dietary and supplementation intakes, the studies reviewed here are limited by the fact that no consideration was given to the potential effects of micronutrient (in most cases, a multivitamin) supplements provided to the control groups. Although at least five of the studies (i.e., Guatemala, Bogota, Cali, New York City and Taiwan) gave supplements that included micronutrients (usually a multivitamin) in addition to protein and calories, most of the comparisons between treatment and control groups did not account for the potentially confounding role of micronutrient supplementation to the control group.

Finally, a third difficulty is the actual role of specific nutrients. There is now evidence that clearly shows that micronutrients may have independent as well as cumulative or interactive effects on behavior (see other articles, this volume for discussions of specific nutrient effects). None of the study designs or analyses considered potential interactions of specific nutrient deficiencies with the nutritional intervention.

Mechanisms

Before discussing the data in terms of mechanisms, it may be helpful to summarize the conclusions from the previous discussion:

1. There is evidence to support nutritional effects on behavior independent of social and environmental factors.

2. Differential responsivity, i.e., that the effects (or response) of nutritional supplementation are not equally distributed across all individuals.

3. Nutrient quality and quantity are not independent contributors but rather they may be additive or interactive components. Malnutrition is the result of the inadequate intake of foods within a diet rather than simply as specific nutrient deficiency.

Despite advances in many areas, the accumulated knowledge was not particularly useful in specifying actual mechanisms (via CNS function, linking nutrition to behavior. The most common explanatory model, nutrition (r) brain development (r) behavior, suggests that malnutrition affects brain growth and development and hence future behavioral outcomes. The literature reviewed here provides no insight as to what brain structures and/or brain functions were affected. A frequently proposed alternative explanation argues that the effects of malnutrition are the result of a complex process of behavioral and socioenvironmental deprivation. Again, none of the studies reviewed here actually collected data to support this hypothesis.

In contrast, more recent studies have taken into account these weaknesses and postulated actual processes by which malnutrition may affect behavioral development. Illustrative of this approach is an ongoing study in Indonesia (Pollitt, E., personal communications]. In part on the basis of the accumulated evidence for consistent effects of supplementation on early measures of motor development, Pollitt and his colleagues recently hypothesized that the association between nutritional status and cognitive development is mediated by motor maturation, activity level and exploratory behavior. The data being collected include both process variables as well as standard outcome measures.

With regard to other types of models that can be hypothesized and tested, it follows that the significance of motor development is at least due in part to the match between timing of the intervention and the sensitivity of development of the particular behavior being measured. For example, motor development may be particularly sensitive to intervention during the period of acquisition of motor milestones and after which no effects would be observed. At later (or earlier) periods, the effects of nutritional deprivation will be manifested in other behaviors.

The data reviewed here raise some important questions regarding the pathway(s) by which nutritional effects occur. For example, how do we reconcile the fact that supplementation to mothers and supplementation to infants result in similar behavioral effects in infants (e.g., improved motor development) or that different nutritional deficiencies (e.g., iron and protein energy) appear to have similar effects? Although one might argue that as the recipient of the intervention changes, so must our interpretation of the mechanisms involved, it is also possible to think about the possibility that there may be common pathways that lead to similar outcomes. The data reviewed here do not provide answers to these questions but rather illustrate the need to explore more closely the implications of such explanations.

One of the most promising avenues for understanding mechanisms comes from the data on environment by organism interactions. The evidence, however, is not easily interpreted: whereas data from Guatemala suggest that children from poorer families benefited more from supplementation, children from Bogota seemed to benefit more when they came from environments that had more resources. It may well be that both need (e.g., poor SES) and resources (e.g., maternal education) interact to influence outcome. In this case then, individual differences based on these characteristics (socioeconomic status, wealth and education) will determine the degree of effects of a totally nutritional nature. Both greater specificity of design and more precise clarification of hypothesized interactions will improve our ability to understand the pathways by which behavioral, environmental and CNS functions interact with nutritional status. Finally, the degree of sophistication in specifying the role of particular nutrients (see Chapters on Iron and Zinc in this volume) is probably one the greatest contributions that this field has made in the past 20 years to our understanding of the nutrition-behavior literature. However, this notable scientific advancement has not been paralleled in terms of solving the problems of the endemically malnourished poor. Although meritorious in its own right, the study of the relation of specific nutrients to specific developmental processes becomes significantly more meaningful when it guides our actions to provide better diets to undernourished children.

Contents - Previous - Next