Previous - Next
E. POLLITT*.
* Department of Applied Behavioral Sciences, University of California, Davis, CA 95616, U.S.A.
1. Background
2. The main-effect model
3. Deficiencies of the main-effect model
4. Suggestions for future research
References
Three decades of research
have passed, involving costly, lengthy and elaborate attempts to
test the hypothesis that early chronic protein energy
malnutrition influences cognitive and social-emotional
development. A cursory review of this research provides a gloomy
view of its accomplishments. Without an in-depth analysis, the
data gathered in the last 30 years are, at best, inconclusive;
admittedly, not a single study can be cited that satisfies all
the requirements of experimentation and at the same time provides
a clear and distinct picture of the effects of early chronic
undernutrition on mental development. Serious attempts to test
the hypothesis in question failed because of conceptual and
methodological problems in the criteria that define the subjects
to be studied, in the sampling procedure, in the delivery of the
experimental treatment, or in other aspects of research design.
For example, intervention studies included subjects not likely to
be nutritionally at risk (RUSH, STEIN and SUSSER, 1980); in
others, the frequency of the treatment was hopelessly confounded
with self-selection in the sample (KLEIN, et al., 1976),
or the treatment represented not only a nutritional intervention,
as intended, but a potent social intervention as well (CHÁVEZ
and MARTÍNEZ, 1982).
This dim view of the past three decades brightens, however, when the studies are pooled and the evidence is examined in the light of new relevant theoretical and empirical work in developmental biology and developmental psychology. This paper examines critically and synthesizes the conceptual approaches used and the data derived from studies on undernutrition and behavioral development- particularly cognitive development1. This critique takes advantage of current thinking on the nature of behavioral development and of the role attributed to biological risk factors in shaping cognitive function. A final objective is to underline what are in my view the most important findings on which we can draw, and point to new approaches that can lead future research in this area.
1 Most research on the developmental consequences of early chronic protein and energy deficiency has focused on aggregate measures of intelligence (e.g., IQ) or specific measures of cognitive development (e.g., attention; POLLITT, 1984). There are some notable exceptions that have focused on social-emotional development, or mother-child interaction, or school behavior, among others (see references). These are, however, exceptions.
Early studies of the
effects of prenatal and early postnatal undernutrition on
behavioral development were guided by a static main-effect model,
which has its roots in the biomedical model of disease causation.
In its simplest form the biomedical model is a reductionistic
approach in which agent and disease are conceptualized as
causally and linearly related (ENGEL, 1979). A simple version of
this model in relation to the hypothesis of an adverse effect of
early protein energy malnutrition on mental development is
represented in the following bivariate equation:
BD = f (NR)
where BD = behavioral deviation and NR = nutritional risk factor (e.g., wasting during the second six months of life). That is, the probabilities of a delay in behavioral development will vary as a function of severity and timing of the nutritional disorder.
This main-effect model can be traced in the developmental literature to the early-trauma, later-deficit hypothesis, which goes back to the mid-19th century (LITTLE, 1982). A basic assumption is that exposure to biological risk factors2 during critical periods of brain growth results in structural lesions in the brain, which in turn leave sequelae such as mental retardation. The emphasis is on the measurement of an input (e.g., chronic fetal hypoxia) and an output (e.g., IQ; see, for example NAEYE, 1987). Besides epidemiological studies using large sample sizes and establishing bivariate correlations, clinical trials and the measurement of a dose-response are sharp expressions of this conceptual and methodological approach.
2 In the present context risk is defined as the probability of an individual's developing a given disease or experiencing a health-status change over a specified period of time (KLEINBAUM, KUPPER and MORGENSTERN, 1982). Biological risk factors are adverse environmental circumstances or events that occur in pregnancy (e.g., intra-uterine growth retardation), at birth (e.g., hypoxia), during lactation, or during the preschool period (e.g., lead intoxication) and increase the probability of diverting a child's growth and developmental trajectory from a course typically followed when physiological and emotional needs are adequately met.
The main-effect model does not account for either the social context in which development occurs, for the previous and subsequent health history of an individual, or for the interactions3. Most importantly, the model, within the context of developmental research, has failed to incorporate two critical developmental characteristics: plasticity and canalization (MCCALL, 1981). Plasticity refers to the notion that, within the limits of individual differences and structural functional capacities, the organism has the elasticity or flexibility to be influenced by, and adapt to, new environmental contingencies (KUO, 1967; GOLLIN, 1981). Canalization refers to the organism's characteristic of following a species-specific developmental path that allows it to withstand a great amount of environmental stress before any significant deviation is observed4.
3 There are, among others, two categories of interactions that have been identified in human developmental research (RUTTER, 1983). One refers to the interactions (synergistic or antagonistic) between variables outside the organism that affect developmental outcome. The other (ordinal interactions) refers to interactions between an environmental variable and an organismic variable that determine different effects from similar experiences in different people, or from similar experiences in the same people at different ages.
4 At least three different definitions of canalization have been advanced (see GOTTLIEB, 1983). In the present context reference is made to canalization as defined by the developmental geneticist WADDINGTON (1971).
The main-effect model guided correlational and experimental (or quasi-experimental) research in the field and in the laboratory. Correlational studies measured the associations between retardation in physical growth and delays in mental development (i.e., DQ or IQ; KLEIN et al., 1972) or in tests of specific cognitive function (CRAVIOTO, DELICARDIE and BIRCH, 1966). Height-forage, weight-for-age or weight-for-height were taken as indicators of present nutritional status or of nutritional history. Statistically significant correlations between, for example, height-for-age and low intelligence test performance were interpreted to mean that exposure to undernutrition increased the probabilities of developmental risk. Experimental studies of nutrition supplementation measured the variance in developmental outcome accounted for by the variance in dietary intake (Jogs and POLLITT, 1984).
A striking illustration of the influence of the main-effect model on the conduct and interpretation of research on undernutrition and behavior was the calculation of regression coefficients to measure the changes in protein intake presumably required for a given change in an intelligence quotient (TAYLOR and SELOWSKY, 1973). This econometric analysis might not be representative, but it does convey the flavor of the conceptual approach that dominated the field.
Animal and basic research, and influential theoretical concepts such as the critical period of behavioral development (SCOTT, 1962) 5, strengthened the face validity 6 of the main-effect model. The critical period hypothesis in connection with a nutritional insult predicted that protein and energy malnutrition during a time of accelerated brain growth (i.e., the first 24 months of life) resulted in a deviation from the normal trajectory of central nervous system development (DOBBING and SANDS, 1979). Implicit in this definition is the idea that past a certain "critical point" (duo, 1967), there is little chance for rehabilitation from the insult produced by undernutrition. A corollary is that past such critical point of brain growth the presence of undernutrition would not have severe developmental consequences.
5 A critical period is defined as a limited period during development in which a particular stimulus will have a profound effect upon the organism. The same stimulation before or after this interval will have little or no effect upon the organism.
6 Face validity is a subjective evaluation by judges as to what a measuring device appears to measure.
Experimental
and basic research data in part supported the main-effect model
and fed into the critical period hypothesis.
Experimentally-induced malnutrition in early life on laboratory
rodents (BARNES et al., 1966; FRANKOVA and BARNES, 1968)
and pigs (BARNES, MOORE and POND, 1970) adversely affected
learning in later development. Early undernutrition in rats
retarded the division process of every type of proliferating
brain cell and delayed migration of cells and myelination (WINICK
and NOBLE, 1966). Curtailment of brain cell division was also
documented in infants who died of severe undernutrition (WINICK
and ROSSO, 1969).
3.1. Outcomes of primary and secondary malnutrition
3.2. Effects of the environment and experience
3.3. Outcomes of monofocal and multifocal interventions
The results of
correlational and supplementation research studies in humans and
experimentation in animals, however, were insufficient to justify
the main-effect model. Gradually, most investigators recognized
that a bivariate approach of simple linear causality was not
conducive to an understanding of the developmental effects of
undernutrition among poor children (POLLITT and RICCIUTI, 1969;
RICHARDSON, 1974, 1980; RICCIUTI, 1981; RUSH, 1984;
GBANTHAM-MCGREGOR, 1984). It became apparent that undernutrition
was a multifactorially determined human condition, far too
complex to be reduced to the blueprint of the main-effect model.
The model constricted the alternatives for the recognition and
measurement of factors that coexisted and interacted (see
footnote 3) with protein and energy malnutrition and contributed
to the nature of the final developmental outcome. In other words,
the model lacked the sensitivity to account for what was observed
in the field as well as for some of the findings of studies on
human populations. Some of these findings are reviewed below,
underscoring the basic issues that evidenced the need to drop the
main-effect model and pointed in the direction of a new paradigm.
The review is primarily intended to uncover contradictions in the
findings of different types of studies, which reveal the
inadequacies of the bivariate model. Three main issues will be
addressed:
1. discrepancy in the findings between the developmental outcome of primary and secondary malnutrition;
2. effects of favorable and unfavorable social environmental circumstances in the developmental outcome of undernourished children; and
3. the differences in the effects produced by nutritional supplementation with and without health care and education stimulation.
There is a striking
contradiction between the results from behavioral studies of
children with a history of malnutrition living in economically
impoverished environments in developing countries, and those from
studies of children whose malnutrition is secondary to an organic
illness in developed countries. The former consistently showed an
association between undernutrition in early life and later
developmental delay (see reviews in GALLER, 1984; BROZEK and
SCHÜRCH, 1984; POLLITT and THOMSON, 1977). Conversely, studies
on children with organic illness generally showed that these were
not developmentally delayed. In these cases, chronic
undernutrition in early life, independent of socioeconomic
deprivation, did not correlate with poor developmental or
intelligence test quotients or else the correlations found were
very low (BEARDSLEE, et al., 1982; BERGLUND and RABO,
1973; KLEIN, FORBES and NADER, 1975). Similarly, a study in
Holland on the effects of famine during World War II showed no
effects of early undernutrition on performance on a non-verbal IQ
test at age 18 (STEIN, et al., 1975).
One of the
suggestions derived from this contradiction in the nature of the
findings from these two types of studies is that undernutrition
in early life did not leave developmental sequelae. The
developmental delays observed among the undernourished children
in developing countries were due to the contextual factors in
which malnutrition occurs. An alternative explanation is that the
environment of children with secondary malnutrition acted as a
buffer and neutralized the effects of the nutritional deficiency.
Evidence of potent
advantageous and disadvantageous effects of the social
environment and of experience on behavioral development has
confirmed the inadequacy of the main-effect model (MCKINNEY,
1986). This evidence has accumulated gradually from different
sources and in relation to different biological risk factors (see
FARRAN and MCKINNEY, 1986). Children with similar types of early
trauma differed in their developmental outcomes according to the
nature of their social and familial environment (WERNER, 1986).
Early signs of intellectual impairment associated with biological
trauma such as intra-uterine growth retardation disappeared in
some circumstances because of favorable qualities of the
environment and the educational experiences to which children
were exposed (SAMEROFF and CHANDLER, 1975).
Findings from studies on undernutrition and cognitive function concurred with the behavioral observations on other biological risk factors. Social-environmental conditions determined large variability in the developmental outcome of children with similar histories of early nutritional deficiencies (RICHARDSON, 1980). Moreover, the magnitude of the differences between the intelligence test scores of children with and without a history of malnutrition varied as a function of social-environmental factors. Under comparatively favorable environmental conditions the differences were small (e.g., 2 IQ points); on the other hand, under unfavorable environmental conditions, these differences were relatively large (e.g., 10 IQ points; RICHARDSON, 1980). Similarly, environmental conditions that met the developmental needs of children protected against, and in some instances remedied, the cognitive delays associated with the nutritional deficiency (WINNICK, MEYER and HARRIS, 1979; GBANTHAM-MCGREGOR, 1984, 1984a). Thus, early malnutrition, even during so-called critical periods of brain growth, was not a sufficient condition to fix a developmental trajectory 7.
7 Based on data from a long-term follow-up of severely malnourished (marasmic) children on the island of Barbados, GALLER (1984a) has argued strongly that, after controlling for social-environmental variables, malnutrition accounts for a large and significant portion of variance in intelligence test measures, school behavior and school attainment. Her measures of graduated parameters of social structure are illustrated by assessments of household items, quality of housing or father's work. Because of the retrospective nature of the study process variables relevant to the time of the nutritional deficiency, could not be assessed.
Galler's work is indeed commendable because of the amount of follow-up information she has collected and the variety of behavioral measures obtained. However, given the nature of her study design, I fully agree with RICCIUTI'S (1981) comments related to the issue of looking for the effects of malnutrition, controlling for socioeconomic factors: "From this reviewer's perspective, these kinds of analyses are heuristically useful up to a point, but they are of rather limited value in advancing our understanding of the interactive influence of nutritional and socioenvironmental variations on intellectual development. First the indices of nutritional status and of the family and home environment are typically quite simplistic and hence may be capturing only a small portion of the developmentally relevant variations in each domain. Moreover, obtained estimates of the independent contribution of nutritional versus socioenvironmental factors will vary greatly depending on various characteristics of the samples employed such as age, homogeneity or heterogeneity with respect to environmental and nutrition variation and also depending on the particular outcome measures utilized." (p. 402).
Moreover, there was strong evidence of a synergism between nutritional history and social-environmental conditions in connection with developmental outcomes.
The corrective process, determined in part by social-environmental circumstances and related experiences, illustrated the plasticity of the organism in behavioral adaptation. Likewise, the shift toward a normal developmental trajectory following early developmental deviations could be considered akin to canalization, namely, a propensity to move in the direction of a species-typical path toward which its members tend to develop (see WADDINGTON, 1971).
Conversely, cognitive developmental deficits increased as children grew older and were continuously exposed to environments that did not meet their physiological, emotional and educational needs. Evidence from different sources confirmed the validity of the so-called "cumulative deficit hypothesis" (SACO-POLLITT, POLLITT and GREENFIELD, 1985). This pervasive effect of the environment was explained by the fact that under ordinary circumstances key components of such environment remain invariant across time. People tend to remain poor under conditions where the opportunities for change rarely exist, and the support systems required for such a change are scarce.
The high frequency of illness among children who were undernourished and living in economically impoverished environments (MATA, 1978) was also noted by investigators concerned with the effects of undernutrition on cognition. Behavioral observations were made that infection and diarrhea result in anorexia, affect the biological homeostasis of children and limit motor activity (BEATON, 1983). In one study (POLLITT, 1983), a significant relationship was observed between frequency of respiratory and gastrointestinal illness during the first six months of life and mental and motor development scores at eight months of life. The explanation given was that a demand for biological homeostasis leads an organism that is frequently ill to reduce activity and decrease energy expenditure.
Environmental
conditions will determine to a large extent the direction of the
developmental trajectory of children with a history of
malnutrition. They can potentiate the effects of the nutritional
deficit or, conversely, remedy or prevent adverse effects.
Whenever there is a relationship between undernutrition and
developmental delay, it is not likely to have been a simple,
linear causal relationship. Unfortunately, however, with some
notable exceptions (CRAVIOTO and DELICARDIE, 1979) little
research has been done to define the nature of the environmental
contingencies that will determine final outcomes for malnourished
children.