Contents - Previous - Next

Relation of mild-to-moderate malnutrition to human development: correlational studies¹

Anthropometry studies
Direct assessment of food intake
Future directions
Literature cited

THEODORE D. WACHS²

¹Prepared for the International Dietary Energy Consultative Group (IDECG) Task Force workshop on malnutrition and behavior at the University of California, Davis, CA, December 6-10, 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 corrspondence should be addressd: Department of Psychological Sciences, Purdue University, Psychological Sciences Building, West Lafayette' IN 47907-1364.

Department of Psychological Sciences, Purdue University, West Lafayette, IN 47907-1364

The present paper reviews correlational studies relating mild-to-moderate malnutrition to human behavior and development. Besides traditional correlational studies, the paper also encompasses quasi-correlational studies, i.e., studies that transform continuous variables (e.g., weight for height) into categories and compare the behavior of individuals in different category groupings. Readers should consult other sections of this supplement

American Institute of Nutrition. for state-of-the-art knowledge in areas such as nutrition and central nervous system functioning, impact of severe malnutrition, nutritional supplementation studies and trace mineral deficits. A number of areas potentially relevant to the study of mild malnutrition will not be considered in this review. Specific review articles are noted for readers interested in these topics. These areas include the short-term impact of nutrients on behavior and central nervous system functioning (for reviews see Krassner 1986, Wurtman 1987) eating disorders (Garner and Garner 1992, Price 1989) and the numerous studies on food sensitivity or impact of refined foods grouped under the general heading of orthomolecular approaches (Kaplan 1988, Toller 1992).

For the purposes of the present review, mild malnutrition is defined as a level of intake of energy or specific nutrients that is below the recommended daily allowance, which is associated with less than adequate physical growth and/or changes in metabolism, but not to the degree that would lead to significant wasting, stunting or clinical symptoms. Although chronic mild malnutrition may be related to an overall lack of availability of food or to a lack of availability of specific foods, in many cases this may not be the only cause (Buzina et al. 1989). Chronic mild malnutrition may also be associated with factors such as morbidity (Bentley et al. 1991, Neumann et al. 1992), maternal employment patterns (Leslie 1989), child caregiving arrangements (Engle 1991), education level (Zipp and Holcomb 1992) and, particularly for the elderly, memory functioning (Silver 1992).

Over the past decade a number of reviews have appeared on nutrition and development, many of which include correlational studies on the relation between chronic mild-to-moderate malnutrition and behavioral development (e.g., Allen 1995, Barrett 1986, Buzina et al. 1989, Lozoff 1988, Lozoll 1989, Pollitt 1988a, Read 1982, Simeon and Grantham-McGregor 1990). The present review builds on the conclusions of these earlier reviews, with particular emphasis on studies that were not considered in prior reviews or that have appeared since these reviews were published. Five major conclusions from these reviews concerning the behavioral impact of chronic mild-to-moderate malnutrition can be summarized as follows: 2) Relations between indexes of prenatal malnutrition and subsequent postnatal development are inconsistent. 2) There appear to be low but significant correlations between anthropometric measures thought to be indexes of chronic mild malnutrition and measures of cognitive development in childhood. Modest associations were also demonstrated between measures of anthropometry (particularly weight) and adult activity (work capacity).3) Moderate-to-severe deficits in certain micronutrients such as iodine, iron, zinc or B vitamins may be associated with behavioral deficits. However, the relevance for behavior and development of mild subclinical micronutrient deficits is inconclusive at present. 4) Chronic mild-to-moderate malnutrition typically covaries with a host of other developmental risk factors, including higher risk of morbidity, inadequate psychosocial rearing and increased exposure to environmental stressors. 5) Particularly for development, the majority of reviews stressed indirect influences of chronic mild-to-moderate malnutrition. Specifically, even if there are no structural changes in the central nervous system, chronic mild malnutrition may reduce the child's interactions with the environment (e.g., activity and exploration) as well as the child's ability to elicit appropriate stimulation from the environment.

This review focuses on whether these conclusions need to be modified in light of new evidence or evidence not previously considered. The review concludes with a discussion of future research directions for correlational studies on mild-to-moderate malnutrition.

Anthropometry studies

Prenatal undernutrition. A number of recent studies seem to suggest a stronger role for mild prenatal malnutrition than did prior reviews. Specifically, less than adequate postnatal cognitive performance was found to be associated with a variety of anthropometric indexes of prenatal malnutrition, including low birth weight (Kirksey et al. 1991, Lynn and Hattori 1990), small size for gestational age (Simopoulos 1986, Walther 1988, Xu 1993), slower head growth in utero as measured by ultrasound (Harvey et al. 1982) and body symmetry in a low-birth weight population (Gorman and Pollitt 1992). In addition, small size for gestational age was associated with suboptimal (Watt 1987) and less stable (Riese 1992) patterns of infant temperament. However, a closer look at the available data reveals that caution is warranted in deducing a direct relation between postnatal development and prenatal malnutrition as indexed by anthropometric measurements. A number of studies reported that there are no unique predictive relations between size for gestational age and subsequent cognitive performance from infancy through the grade school years (Dowling and Bendell-Estroff 1991, Hack et al. 1989, Haddres-Algra and Touwen 1990, Iyer et al. 1989, Vohr et al. 1989). Other studies reported that when predictive relations are found, they are restricted to a specific subsample of smallfor-gestational-age infants as characterized by very early preterm birth (Smedler et al. 1992) or associated developmental disabilities (Robertson et al. 19901. The possibility of a sensitive period is suggested by evidence indicating that growth stunting occurring only before 26 weeks' gestational age is likely to have an influence on later cognitive performance (Harvey et al. 1982, Lynn and Hattori 1990). Further, relations between prenatal anthropometry and postnatal development do not appear at all postnatal ages when subjects are followed longitudinally (Gorman and Pollitt 1992).

A number of studies indicate that relations between anthropometic proxies of prenatal undernutrition and subsequent development can be strongly reduced or eliminated in the case of infants from more advantaged backgrounds (Vohr et al. 1989, Wilson 1985), for infants with fewer postnatal biomedical complications (Eckerman et al. 1985) or for infants receiving postnatal cognitive stimulation (Padin-Rojas et al. 1991). Other evidence suggests the possibility that relations between anthropometric indexes of prenatal malnutrition and subsequent performance may be moderated by racialcontextual factors. Specifically, Niestroj (1991) noted that relations between anthropometric indexes of mild prenatal malnutrition and subsequent performance are positive rather than negative in a sample of Zulu neonates. The overall pattern of results suggests continued caution in drawing conclusions about relations between indexes of prenatal malnutrition and subsequent cognitive and noncognitive development.

Postnatal undernutrition. Recent research indicates that there are significant relations between postnatal anthropometry and concurrent cognitive performance. Strongest relations appear for height for age, which has been associated with cognitive performance in Brazilian school-age children (Paine et al. 1992), visual motor integration in disadvantaged U.S. preschool children (Karp et al. 1992) and measures of academic achievement in both U.S. preschool children (Karp et al. 1992) and mainland Chinese school-age children (Jamison 1986). Weight for height has been associated with older children's cognitive performance in Brazil (Paine et al. 1992) and grade placement in China (Jamison, 1986), as well as with teacher ratings of Phillipine adolescents' ability to concentrate in school (Popkin and Lim-Ybanez 1982). Weight for age has been associated with both visual motor integration and preschool achievement (Karp et al. 1992). Small but consistent correlations with concurrent cognitive performance and adaptive behavior were reported for both height (Humphreys et al. 1985, Nagoshi and Johnson 1987, Sigman et al. 1989a) and weight (McCullough et al. l 990, Sigman et al. l 989a), and predictive correlations have been reported for height (Martorell et al. l 992). Postnatal head size was found to be moderately correlated with childhood IQ (Lynn 1989), whereas indexes of fat fold and arm circumference were found to be associated with preschool achievement (Karp et al. 1992). A similar pattern of findings was repotted when anthropometric indexes were used to group children into adequately and poorly nourished groups (e.g., Singh et al. 1976). Recent evidence also indicates that measures of height and weight taken at infancy predict cognitive performance 5 y later (Sigman et al. 1991).

Again, however, closer inspection of the available data suggests that the results are not as consistent as they would appear at first glance. First, not all studies show significant effects. A number of studies reported essentially insignificant relations between standard anthropometric measures and cognition (Church and Kagitbak 1991, Wachs et al. 19951 or achievement (Dugdale and Siew 1977, Popkin and Lim-Ybanez 1982, Richards et al. 1985, Wagstaff et al. 1987). Although in some cases the lack of relations may be attributed to insufficient statistical power (e.g., Popkin and Lim-Ybanez 1982), this does not appear to be true for the remaining studies. In addition, even when significant relations are found between anthropometry and cognitive performance, there is always some degree of ambiguity because anthropometry scores may be associated with a variety of factors other than inadequate nutrition. For example, a relation between height and intelligence may be a function of pleiotropy (i.e., the same genes influence both height and IQ), genetic linkage (i.e., the genes controlling two traits may be in close proximity) or parental assortive mating (Nagoshi and Johnson 1987).

Finally, a number of studies suggest that relations between anthropometry and cognitive performance may be moderated by other parameters, making it difficult to ascribe such relations purely to inadequate nutrition. Moderators that were identified include birth order (Belmont et al. 1975), family size (Belmont et al. 1975, Dugdale and Siew 1977), parental occupation (Paine et al. 1992), sex of child (Humphreys et al.1985, Lynn 1989, Nagoshi and Johnson 1987), race (Nagoshi and Johnson 1987), generation (Nagoshi and Johnson 1987) as well as the particular component of anthropometry utilized (e.g., leg length vs. trunk length when assessing height (Humphreys et al. 1985). Perhaps the most consistent moderator appears to be the quality of the child's psychosocial rearing environment, whether indirectly assessed via demographic measures such as social class (Singh et al. l 976, Vasir et al. 1988), maternal education (Popkin and LimYbanez 1982) or family income (Dugdale and Siew 1977), or more directly via the indexes of school quality (Popkin and Lim-Ybanez 1982) and the quality of the child's physical and psychosocial rearing environment (Church and Kagitbak 1991, Jachack et al. 1983).

These complex patterns of findings do not necessarily negate the hypothesis that there are relations between cognitive development and chronic mild postnatal malnutrition as indexed by anthropometry. Perhaps the most appropriate conclusion to be drawn from the anthropometry data is that chronic mild-to-moderate postnatal malnutrition may be a necessary but insufficient influence for explaining variability in cognitive performance. Certainly, the results do not support some of the extravagant claims that were made regarding relations between anthropometric indexes of nutritional status and cognitive development. This is particularly true for claims that inadequate nutrition can explain secular changes in intelligence over the short term (Schoenthaler et al. 1986) or across generations (Lynn 1989).

The evidence is much more scarce for noncognitive outcomes. Previous reviews (e.g., Lester 1979) hypothesized that inadequate nutrition may cause disturbances in parent-child interactions. Arya (1989) reported that undernourished Indian infants were more passive and their mothers were less interactive. Valenzuela (1990) reported that, in Chile, low-weight-forage children were substantially less securely attached than were adequate-weight-for-age children. However, because neither study provided data on the nutritional status of the mothers, it is difficult to know whether observed relations were a function of inadequate nutrition of the mother, the child, or both. An answer to this question will be suggested in the following section. At the other end of the age spectrum, Peterson (1989) noted that elderly Americans reporting nutritional problems also tended to report being less involved and less interested in political processes, suggesting a potential relation between inadequate nutrition and apathy in older populations. These results held true even when educational level was taken into account.

Direct assessment of food intake

Recent studies in Egypt and Kenya related measures of food intake, as assessed by dietary recall or food weighing, to cognitive and noncognitive performance of infants and children. The evidence from these correlational studies, indicating the salience of animal source foods as well as specific micronutrients, suggests the need for an expanded definition of what is meant by chronic mild-to-moderate malnutrition.

Dietary quality and quantity. Traditionally, malnutrition has been defined in terms of protein-calorie deficits (dietary quantity). Recent data on Kenyan children related higher energy intake to more positive social and emotional behaviors and higher activity level (Espinosa et al. 1992). Data from both Egypt (Wachs et al. 1993) and Kenya (Sigman et al. 1989a) are consistent in showing that higher levels of toddler mental performance and symbolic play are primarily associated with higher intake of energy and total protein. However, data from Egypt also indicate that neonatal habituation and orientation performance are uniquely related to maternal intake of energy, protein and fat from animal sources rather than total energy intake (Kirksey et al. 1991). Although infant alertness in the first 6 mo is positively related to both animal and total energy intake, the strongest predictions occur for animal source energy intake (Rahmanifar et al. 1993). In Kenya, animal protein intake measured at 1830 months was found to be a stronger predictor of cognitive performance than total protein intake when assessed 5 y later (Sigman et al. 1991). Cognitive performance of Kenyan school-age children relates positively to both dietary quantity (energy, protein) and dietary quality (animal protein intake (Sigman et al. 1989b), although intake of energy, proteins or carbohydrates was not found to predict school achievement in the United States (Thatcher et al. l 984). Animal source protein intake also appears to be a stronger predictor of Egyptian children's cognitive performance and activity level than total protein intake (Wachs et al. 1995). A similar pattern was found in a recent study relating nutrient intake to adult cognitive performance in Egypt (Wachs, T. D., McCabe, G., Moussa,W.,Yunis, F., Kirksey, A. Harrison, G., Jerome, N. and Galal, O., unpublished results).

Noncognitive outcomes. As previously noted, evidence from the anthropometry literature suggests that inadequate nutritional status may have an adverse impact on the quality of caregiver-child relations. However, it is unclear from this literature whether the results are a function of the adequacy of infant nutrition, caregiver nutrition, or a combination of the two. In both the Kenya and Egypt projects, nutrition of both mother and toddler was measured using multiple regression. The results indicated that variability in caregiver behavior is primarily a function of toddler rather than maternal nutritional intake, although some interactions between the two were noted (Wachs et al. 1992). However, as will be seen below, these results may be moderated by background, nutritional and cultural factors.

Micronutrients. Iron. Experimental evidence on the detrimental impact of iron deficiency anemia on growth and development was reviewed in the paper by Pollitt in this supplement. Evidence on the developmental impact of iron deficiency without anemia suggests that the data simply are not sufficient for concluding that iron-deficient children are at elevated developmental risk (Lozoff 1990). Highly complex results are typically found in correlational studies that focused on populations less severely affected by iron status. These studies suggest that the relation between suboptimal (but not anemic iron status and cognition may be moderated by age of measurement or by the outcome measure under study. Wassermann et al. (1992) report that declines in hemoglobin from birth to 2 y of age covary with declines in general mental development over this age period, but significant relations occur only at 18 mo of age. For older children, Pollitt et al. (1983) report that suboptimal iron status is related to poorer original learning but not to variability in memory, auditory learning or IQ. Popkin and Lym-Ybanez (1982) report that suboptimal iron status is related to impaired learning achievement and increased absence from school, but not specifically to either science or math achievement or to teacher ratings of the child's ability to concentrate m class.

At least for nonmanipulative field studies, one explanation for these inconsistent findings may have to do with the nature of the iron source. Specifically, studies from Egypt show positive relations between iron level intake and neonatal habituation (Kirksey et al.), infant alertness (Rahmanifar et al. 1993) and female school-age children's verbal IQ (Wachs et al., in press]. However, what is predictive in these studies is not iron intake per se but, rather, the bioavailability of the iron or the fact that the iron came from animal rather than plant sources. This suggests the need to consider iron source or bioavailability of iron in correlational studies of suboptimal, but not clinical, iron deficiency (see below for an alternative hypothesis).

Zinc. Experimental studies on the relation to growth and development of moderate-to-severe zinc deficiency were summarized by Golub in a separate paper in this supplement. Again, the few correlational studies available focused on less severe intake deficits. Using hair sample analyses, Thatcher et al. (1984) reported that zinc status is related positively to school achievement Data from Egypt related higher maternal intake of zinc to more optimal neonatal habituation (Kirksey et al. 1995) and infant alertness (Rahmanifar et al. 1993), whereas activity level of male school-age children was shown to be positively related to zinc intake (Wachs et al. 1995). Again, the importance of nutrient source may be critical. For example, animal source rather than total zinc intake is uniquely related to infant alertness (Rahmanifar et al. 1993), and intake of plant source zinc is negatively related to mental development at 6 mo (Kirksey et al. 1995).

Other trace minerals. Data from Egypt also suggest a possible role for other nutrients in behavioral development. Calcium intake positively relates to toddler symbolic play (Wachs 1993), whereas copper intake is positively associated with female school-age children's attention seeking (Wachs et al. 1995). Both calcium and copper are found to be related positively to the activity level of male school-age children Wachs et al. 1995).

Several of the studies cited above are relevant to a prior suggestion by Halas and Eberhardt (1987) that combinations of, rather than single, trace mineral deficits may be critical for behavior and development. Although little research on this question existed at the time of the Halas and Eberhardt paper, a number of the studies cited above lend support to this hypothesis, particularly with regard to micronutrient combinations related to bioavailability of critical nutrients such as iron or zinc. Results from these studies (Kirksey et al. 1995, Wachs et al. 1995) indicate that calcium, copper and/or vitamin C intake appeared as predictors in combination with iron intake, and calcium and copper intake appeared in combination with zinc intake as predictors. Calcium, copper and vitamin C intake were also shown to play a role in iron metabolism (ranks 1988, Hallberg 1988), whereas calcium and copper were implicated as influencing zinc bioavailability (Reinstein et al. 1984). The above pattern of results leads to the tentative suggestion that specific micronutrient deficits that do not reach clinical proportions may still be related to behavioral deficits, when embedded in a set of other low-level micronutrient deficits that influence bioavailability. Although provocative, this hypothesis clearly needs to be tested experimentally, given the tight covariance that exists in nature for many of the nutrient combinations described above.

Vitamins. Data from Egypt also provide a consistent body of evidence on the relation of B vitamins to a variety of cognitive and behavioral outcomes. Neonates who are more responsive to aversive stimuli, more easily consoled and less likely to show extreme distress have mothers with higher levels of vitamin B-6 in their breast milk, and B-6 level is associated with 6-mo infant vocalization as well (McCullough et al. 1990). Maternal intake of vitamins B-6 and niacin, particularly in the second trimester of pregnancy, are positively associated with neonatal habituation (Kirksey et al. 1995). Maternal intake of riboflavin, niacin and vitamin B-6 are also associated with infant alertness (Rahmanifar et al. 1993). Vitamin B-12 was found to be positively associated with degree of toddler symbolic play (Wachs 1993), whereas vitamins B-6 and B- l2 were both found to be positively associated with increased activity of male school-age children (Wachs et al. 1995) In addition, mothers with suboptimal vitamin B-6 intake are more likely to be unresponsive to their infant's vocalization, to be less effective when responding to their infant's distress and to use sibling caregivers (McCullough et al. 1990). Relations between B vitamin intake (particularly vitamin B-6) and development may be a function of the role played by these vitamins in neurotranmission processes, as suggested in a recent paper by Guilarte (1993).

Nutrition and context. Recent correlational research highlighted the potential role of animal source nutrients, illustrated the importance of low-level deficits in iron, zinc and B vitamins and supported the concept of nutrient-by-nutrient interactions as a parameter influencing developmental outcomes. In interpreting this research, it is important to remember that chronic mild nutritional deficits rarely occur in isolation from covarying psychosocial, biological and contextual stressors (Beaton 1984, Golden 1991, Pollitt 1988b, Ricciuti 1993). Current correlational research also serves to illustrate the process whereby variability in nutrition and its covariates relates to variability in developmental outcomes. Two points are worth emphasizing. First, many studies show that even when predictors such as socioeconomic status, family education level, literacy or duration of schooling are covaried out, indexes of suboptimal nutrition still contribute unique predictive variance (Espinosa et al. 1993, Kirksey et al. 1991, Kirksey et al. 1995, McCullough et al. 1990, Rahmanifar et al. 1993, Sigman et al. 1989a, Sigman et al. 1991, Wachs et al. 1992, Wachs et al. 1993, Wachs et al. 1995~. However, many of these studies also indicate that the strongest predictions of developmental outcomes can be made when based on the interaction of nutrition with such psychosocial-contextual factors as low parental educational level, low socioeconomic status, illiteracy and specific patterns of caregiver-child transactions (Kirksey et al. 1995, Popkin and Lim-Ybanez 1982, Rahmanifar et al. 1993, Sigman et al. 1989b, Sigman 1991, Wachs et al. 1993, Wachs et al. 1995~. This pattern is congruent with both past and current theoretical models that emphasize the joint role of nutritional, contextual and individual difference parameters on development, rather than viewing these as isolated influences (Beaton 1984, Horowitz 1989, Lester 1979, Pollitt 1988b). A simplified process model illustrating this pattern is shown in Figure 1.

It could be argued that evidence supporting the role of contextual influences is artifactual, perhaps because of the strong covariance between nutrition and context. Three lines of evidence do not support the validity of this alternative interpretation. First, in a number of studies, sociodemographic risk factors continue to add unique predictive variance, even after variability associated with anthropometry or food intake is parceled out (Espinosa et al.1992, Galler and Ramsey 1985, Sigman et al. 1989b, Sigman et al. 1991) (Wachs, T. D., McCabe, G., Moussa, W., Yunis, F., Kirksey, A., Harrison, G., Jerome, N.

Galal, O. unpublished results). Second, infrahuman studies that independently manipulate nutritional intake and rearing conditions consistently show unique predictive variance associated with rearing conditions, over and above the impact of nutritional intake (Chaloupka et al. 1975, Crnic 1983, Frankova 1985, Gamallo et al. 1988, Tonkiss et al. 1987). Finally, human intervention studies involving both nutritional and environmental enrichment show unique effects of environmental enrichment over and above the effects associated with nutritional supplementation (Colombo et al. 1992, Grantham-McGregor 1993, Super et al. 1990).

It is not yet clear whether the observed increases in predictive power are a function of additive coaction among individual nutritional and psychosocial-contextual influences or of nonlinear synergistic interactions between these classes of predictors. For the most part, the little available evidence on this question suggests that additive coaction may be more critical (Galler and Ramsey 1985, Sigman et al. 1991, Wachs et al. 1993). However, statistical and methodological problems (e.g., low power associated with testing for interactions) make further research on this issue critical. It is also warranted because of a small set of studies supporting the possibility of nutritional buffering as a special form of interaction.

Buffering interactions. Buffering refers to a protective factor in the environment attenuating the impact of a known risk factor on a child's development. Buffering is considered a form of interaction in that the impact of a risk factor cannot be estimated without reference to the operation of absence of a bufferer (Rutter 1983). Correlational studies illustrate how adequate nutrition or intake of specific nutrients may act as a buffer against biological or environmental risks. On the basis of analysis of hair samples, data from several studies suggest that children who have higher calcium levels may be less susceptible than are children with lower calcium levels to the cognitive impairments associated with high body-lead levels (Blouin et al. 1983, Fishbein and Meduski 1987, Lester et al. 1986). Similarly, children with higher zinc levels appear to be less susceptible to cognitive impairments associated with high body-cadmium levels than are children with lower zinc levels (Lester et al. 1986, Thatcher Lester 1985). These results may be due to the fact that certain nutrients; such as calcium or zinc, may serve as biological antagonists to the absorption of heavy metal pollutants such as lead or cadmium.

In terms of environmental risk, adequate nutrition was shown to be one of several buffers that can minimize cognitive decrements associated with being reared under low socioeconomic status conditions (Haeussler et al. 1982). Low caregiver vocalization was identified as another risk factor for a variety of early cognitive functions in young children. However, a sample of toddlers receiving only limited caregiver vocalization but with a higher intake of carbohydrates and fats showed more adequate levels of involvement with people and objects in their environment than did toddlers with a lower intake of these specific nutrients (Wachs et al. 1993).

Contents - Previous - Next