N.G. NORGAN*
* Department of Human Sciences, University of Technology, Loughborough, Leicestershire, LE11 3TU, U.K.
1. Introduction
2. Definitions and descriptions
3. Approaches to the study of the effects of energy supplementation
4. Supplementation studies
5. Some possible explanations for the small effects
6. Contemporary models
7. Gaps in our knowledge
8. Concluding comments
References
Chronic energy deficiency
is the most widespread nutritional deficiency, it is said to
affect half the world's children. There is much concern about the
situation and considerable efforts are being made to alleviate
this state of affairs. Although nowadays many observers feel that
tackling poverty by a redistribution of income and resources is a
prerequisite to a reduction in the extent of malnutrition, food
distribution programmes may ameliorate the problem. In many
countries the bulk of the budgets for child nutrition goes
towards nutritional supplementation. The proposition that
providing food to the undernourished ipso facto improves
nutritional status and well-being has not been seriously
questioned until recently. There are now numerous observations
and several studies indicating that, for one reason or another,
the effects of nutritional supplementation are small and much
less than expected. This may be a result of poorly designed and
implemented feeding programmes and studies. Alternatively, it may
be that our hypotheses and models of the processes involved have
been unduly simplistic and inappropriate and we may be missing
important effects. We are rarely fortunate to have single causes
and effects, so that a combination of these two explanations is
possible, indeed probable.
The aims of this paper are:
1. to review the effects of energy supplementations in chronic energy deficiency;
2. to consider why supplementation has such small effects, in particular whether better designed and implemented studies will show greater effects, or whether other outcomes should be investigated;
3. to identify gaps in our knowledge.
As a first step, the variables and components of the processes are described.
2.1. Chronic energy deficiency
2.2. Energy supplementation
2.3. Targets of supplementation
2.4. The effects of supplementation
Chronic energy deficiency
(CED) refers to an intake of energy less than the requirement,
for a period of several months or years. It is difficult to
identify directly, as measurements of energy intakes in
free-living individuals and populations are fraught with
problems, a predicament exacerbated in the determination of
requirements. In the latter case, the practical difficulty of
allowing for possible adaptations, behavioural, mechanical,
physiological and biochemical, and varying reproductive status
render this fundamental nutritional characteristic elusive. As a
consequence, CED is usually identified by proxy variables such as
deficits in anthropometry, body composition or growth. The fact
that these vary according to the intensity and duration of CED
would seem an advantage in, for example, identifying those most
in need of treatment, but as the range of normality is wide,
deviations may occur before being identified as such, or normals
may be picked up as deviations. In severe cases of CED, the
picture presented leaves no doubt as to the existence of the
condition. In milder forms, the proxy variables may be less
accurate in identifying the condition, and the sensitivity and
specificity of these criteria may leave much to be desired.
Nutritional supplementation
is an addition to the diet to make up all or part of a
deficiency. It differs from refeeding, where the total provision
of foods is undertaken, and from nutritional intervention, where
non-nutritional components such as socialisation, health care,
etc., are included, intentionally or otherwise.
Supplementation varies according to the type, amount and duration of the supplementation. It is the independent variable, and controlled manipulation gives rise to high internal validity. Supplements may be administered daily at a feeding centre which has to be close to the subjects and requires frequent attendance and high staffing, or distributed weekly and taken home for consumption. This is less efficient in targeting the recipient, as sharing of the supplement may occur. Supplements are invariably based on manufactured, possibly imported, products rather than locally produced or fresh foods, to guarantee homogeneity of composition and adequate supplies.
Some
studies have utilised the process of "natural"
supplementation, that is the naturally occurring covariation in
nutrition and an outcome, as in observations of the previously
malnourished. Such retrospective, non-experimental designs have
limitations in that conclusions of cause and effect cannot be
drawn with confidence.
Supplementation studies
have focused on pregnant and lactating women, and the infant and
young child. Occasionally older children and workers have been
targeted, and in some cases the whole family may have been
recipients. In supplementation programmes, as opposed to research
studies, the most needy may have to be identified and targeted.
The planning and policy issues of targeting are not considered in
this paper.
The effects of
supplementation, the outcome variables, are sought in raised
birth weights, improved infant and child growth, and decreased
morbidity and mortality. In older children, growth has been the
cardinal feature, but considerable attention has been given to
whether deficits in behaviour and mental performance are
overcome. The importance of activity, play and discretional
activities is being recognised and included in studies. The
effects on physical work capacity and productivity allow an
econometric analysis of supplementation, a point that may impress
funding agencies. Evaluation of the effects should be a
fundamental part of any supplementation activities.
The simplest model of the
effects of energy supplementation is:
Energy supplementation (r) Raised energy intake (r) Raised energy stores
Such a model has been described as a direct intervention model (BEATON, 1982) or a main-effect model (ADAIR and POLLITT, 1985; POLLITT, 1987). Its characteristics are a direct, proportional, linear relationship between the intervention and a specific outcome, with few confounding variables. It portrays supplementation in a dose-response manner with similar responses expected in different populations and is a variant of the clinical trial model.
In CED, the outcome variable would be a reduced incidence in CED or improved nutritional status as shown by the proxy variables of, for example, increase in growth and size in the population. However, already the outcome variables are now less specific and may be influenced by many other environmental factors, such as illness, and poor social and economic circumstances. When the outcome is a functional variable such as work capacity or performance, there are many more stages between supplementation and outcome. With a multistage model, which is also a direct and main-effect model, the effects are expected to be smaller and less specific, and more influenced by intervening and confounding variables. The multistage model has been the common conceptual approach to supplementation effects until recently.
Chronic
energy deficiency is characterised by falls in body weight and
fatness, falls in resting metabolic rate and habitual physical
activity, and in physical working capacity. Such changes were
recorded in the classic Minnesota study (KEYS et al.,
1950) and are well-recognised outside the laboratory. CED has
effects on behaviour and socialisation as well as on energetics,
and a single outcome is obviously not representative of all the
effects. In addition, some of the changes are interactive. Falls
in metabolic rate reduce the extent of depletion of energy
stores; reduced size and reduced activity may both influence work
capacity. If the effects of supplementation in reversing some or
all of the characteristics of CED are to be identified, an
appropriate conceptual and methodological approach is required.
However, it is the heuristics of models that are important rather
than the models themselves. The development and testing of models
should provide useful feedback and confirmation to guide research
and planning efforts.
4.1. The INCAP study
4.2. The Gambian studies
4.3. The Bacon Chow Study, Taiwan
4.4. Conclusions
Three typical
supplementation studies are described. They are then examined
along with other studies such as the Bogota, Colombia study (MORA
et al., 1978; 1979; 1981; WABER et al., 1981), the
New York study (RUSH, STEIN and SUSSER, 1980) and the USDA Women,
Infants and Children (WIC) programme (BERKENFIELD and SCHWARTZ,
1980; HEIMENDINGER et al., 1984; KENNEDY and KOTELCHUCK,
1984; METCOFF et al., 1985) to examine whether the results
were affected by inappropriate models, experimental design and
execution, and which model best describes the processes and
outcomes of supplementation.
A number of
reviews and critical appraisals of the various aspects of these
supplementation studies have been published recently: AEBI and
WHITEHEAD (1980); RUSH et al., (1980); LECHTIG (1981);
SUSSER (1981); BEATON (1982); BEATON and GHASSEMI (1982); BROZEK
and SCHÜRCH (1984); BARRETT (1984); JOOS and POLLITT (1984);
MARTORELL (1984); ADAIR and POLLITT (1985); SCHÜRCH and FAVRE
(1986). The effects of supplementation on cognitive performance
in children and on body composition, work capacity and
productivity are described in the Meeting papers of POLLITT and
SPURR respectively.
This longitudinal study was
designed to observe the effects of chronic malnutrition on
physical growth and mental development (LECHTIG et al.,
1979; ENGLE et al., 1979). It was conducted in four
Spanish-speaking villages in eastern Guatemala. At the time the
study began, 15% of infants died before the age of one year.
During pregnancy, energy and protein intakes were 6 MJ and 40 g.
Mothers were small and light (1.49 m, 49 kg), and weight gain
during pregnancy was half that of well-nourished women.
Two types of supplement were provided, atole, a protein-energy supplement, and fresco, which had no protein and one third the energy of atole. Both preparations contained minerals and vitamins in similar amounts. Two villages were randomly assigned to receive atole and two fresco. Supplements were distributed twice daily at a supplementation centre at which attendance was voluntary, so that there was considerable variation in levels of supplementation. Preventive and curative medical care were provided. The target groups were all the mothers and all the children less than seven years. New cohorts of children were included from the start of the study in 1969 until 1973. The study was terminated in 1977.
Birth weights were significantly lower (2997 ± 471 g versus 3114 ± 426 g) in 192 women who had on average a low level of supplementation (180 kJ/d) compared with 165 women with higher levels of supplementation (975 kJ/d). Energy appeared more important than protein and carbohydrate. Although birth weight gains were only 29 g per 42 MJ supplement, the incidence of low birth weight (<2500 g) was halved in the higher level group. Infant mortality fell from 135 to 51/1000.
In 453 0-3-year-old children (MARTORELL, KLEIN and DELGADO, 1980) greater supplement intake was associated with better growth in supine length and weight but not limb circumferences. Skinfold thicknesses decreased with supplementation. At three years, children in the highest supplementation group of more than 800 kJ/d were 1 kg heavier and 4 cm longer than children in a baseline group. The differences equal the pooled SD but were statistically significant nonetheless. Similar differences were observed between atole and fresco villages, but that of height was smaller although still significant.
Behavioural
development was tested in over 450 children at 6, 15 and 24
months of age. A Composite Infant Scale, with subscales for
mental and motor development, was used. At 6 months of age,
children classified as of low (<21 MJ per 3 months), medium
(21-42 MJ) and high (>42 MJ) supplement intake showed
statistically significant differences only for the mental
subscale. At 15 and 24 months, differences were greater and
significant for both subscales. Testing continued until 5 years
with the INCAP Preschool Battery (ENGLE et al., 1979). A
composite score and separate psychological test scores of verbal
reasoning, learning and memory, visual analysis and motor skill
showed significant difference in the three supplementation groups
at 3, 4 and 5 years. More recently, when measured at 6 to 8
years, high-supplementation children showed more interest in and
exploration of a novel environment, more involvement in a
competitive game, greater persistence on a frustrating task,
better motor impulse control and greater initiative across
several group tasks than low-supplementation children (BARRETT
and RADKE-YARROW, 1985).
In Keneba, in The Gambia,
there is marked seasonality in nutritional status of pregnant and
lactating women, as indicated by changes in energy intake and
body weights. In the wet season, energy intakes are lowest, down
to 5.2 MJ/d, and energy expenditure is highest due to the demands
of agricultural activities. The amplitude of the weight change is
some 4 kg. In the dry season, intakes are higher, 6.2 MJ/d in
pregnant women and 7.0 MJ/d in lactating women, but are less than
65% of suggested standards (PRENTICE et al., 1981). During
1980-82, all pregnant women were offered supplements of
groundnut-based biscuits and a vitamin-fortified tea drink
(PRENTICE et al., 1983). Supplements were available six
days a week, consumption was carefully supervised and measured,
and home food intakes were assessed. Overall the daily
consumption of supplement was 2.8 MJ/d of which 1.0 MJ/d
substituted some of the normal diet. Net increases in intake were
higher than in most supplementation studies. Compared to baseline
measurements of 1978-80, supplementation increased mean birth
weight by 100 g, after adjusting for sex, birth order and month
of birth, and decreased the incidence of low birth weight from 20
to 6%. However, the effects were mainly restricted to births in
the wet season when energy intakes are less adequate (PRENTICE et
al., 1981). Birth weights in the months July to January were
224 g higher.
An earlier study of similar design on lactating women (PRENTICE et al., 1980) provided a mean net increase of 3 MJ/d over 12 months and resulted in supplemented women being 1-2 kg heavier. However, the pattern of seasonal weight change was maintained, albeit at a lower amplitude. No consistent improvement in breast-milk intake, as evidenced by test-weighings, was found nor were differences in milk energy content apparent.
In a recent
smaller study, supplemented pregnant women gained 2 kg more fat
than controls from another village (LAWRENCE et al.,
1987).
This was a longitudinal
double-blind nutrition intervention of rural women at risk
nutritionally in an economically depressed area of west-central
Taiwan (ADAIR, 1984; ADAIR and POLLITT, 1985). After the birth of
a first child, women were randomly assigned to a nutrient-rich
dietary supplement group (A) or a placebo group (B) for the
lactation period, the gestation period of the subsequent
pregnancy and its lactation period. The supplements were handed
out twice daily by nurses who monitored intakes. Both groups
received a multivitamin and mineral tablet daily and were
provided with medical care. The outcome variables assessed
included infant birth measurements, postnatal physical growth,
motor and mental development, morbidity, and maternal weight and
skinfold changes during pregnancy and lactation.
Energy
intakes during lactation were significantly different in the two
groups (5 and 7.5 MJ/d) but no significant effects were found in
maternal weight, weight gain and skinfold thicknesses (ADAIR,
1984). Indeed, few A-B differences in mean values of the outcome
variables were found except in some subgroups. Male infants born
after a supplemented pregnancy weighed more at birth than
brothers born earlier. Sibling correlations were lower in the
supplemented group, suggesting that maternal supplementation
affected the components of variation in infant anthropometry more
than the means (MUELLER and POLLITT, 1982), and that there are
differential risk and probability of response in populations
(ADAIR and POLLITT, 1985). Important mediators of supplement
effects included maternal body size (greater weight gain in
mothers with lower weight-for-height), sex of offspring (males
being more susceptible to supplementation than females) and
season of the year. There were no treatment group differences in
mental scores of 8-month-old infants but motor development scores
were marginally significantly different. Supplementation had no
effect on the Stanford Binet IQ evaluation of 5-year-old
children. Thus, supplementation had limited effects on the
population as a whole despite the apparent marginal nutritional
status.
These studies and those
referred to earlier show only small (100 g) increases in birth
weight with supplementation. A more important effect is the
reduction in the proportion of low-birth-weight infants.
Supplementation appears to have little effect on lactation, and
supplements may be better directed to infants than lactating
women. In children, mean differences in weight gain are some
0.5-1.0 kg/y (BEATON and GHASSEMI 1982). Significant changes in
growth rate measured as height have been demonstrated. These data
come from research studies which are likely to be more successful
than feeding programmes because of higher levels of commitment.
The effects on behavioural development are similarly restricted
and, during the first two years, occur primarily in motor items
and subscales (Jogs and POLLITT, 1984). In general, although the
level of response is conditioned by the relative severity of the
CED, deprived children are never brought up to the level of
economically and socially advantaged children. The effects of
supplementation are disappointing and the fate of most of the
supplement is unexplained.
The
observed effects do not fit what is expected in the direct
intervention main-effect model. The hypothesis that
supplementation will increase the intake and improve nutritional
status resulting in improved function and performance in growth,
behaviour, pregnancy, lactation and work is not borne out.
Whether this is a failure of the model or poor experimental
design and implementation is considered next.
5.1. Are the recipients really malnourished?
5.2. Are the target groups being energy-supplemented?
5.3. Are the target groups appropriate?
5.4. Are the outcome variables appropriate?
There are two issues here.
The first concerns the validity of energy intake measurements and
the second the interpretation of low intakes, low body weight and
energy stores.
Considering the expense and effort of supplementation studies, it must be said that a disproportionately small amount of effort seems to have been made to obtain reliable intake data before or during supplementation. Given the intraindividual variation in energy intakes, recall data and weighed intakes over short periods are inadequate. Monitoring of supplement consumption has been satisfactory but the remainder of the intake has not been assessed accurately.
The question of the significance of low intakes and adaptations to these have been considered in detail by James. Low intakes have, over the last 15 years, become acceptable without any real justification for this except that investigators have reported being as careful and as accurate as possible under the field conditions. This has led to a number of hypotheses about adaptations, a term used to mean several different things. Where energy intakes and expenditures have been measured and a discrepancy found, intakes have been regarded as more accurate given the techniques involved (DURNIN, 1984). The new technique of doubly-labelled water is suggesting that, in The Gambia at least, the energy expenditure measurements may be correct, in line as they are with those from indirect calorimetry, and that the error lies in food-intake measurements SINGH et al., in press). This supports the evaluation of our Papua New Guinea data (NORGAN, FERRO-LUZZI and DURNIN, 1974) made earlier by JAMES and SHETTY (1982) and could mean that the need for hypotheses of adaptive processes should be seen in a different perspective. BEATON'S (1984) comment "Either our concepts and understanding are wrong or the data are wrong" remains unanswered.
In the
Taiwan and Gambia studies, intake data (5-7 MJ/d) and
anthropometric data suggest the subjects were at risk
nutritionally. However, reproductive performance was good. In
Taiwan, mean birth weights of 3000 g were high for a developing
country and only 7% of births weights were below 2500 g. In The
Gambia, lactational performance was satisfactory. When
performance and functional capacity data conflict with the proxy
variables of anthropometry or energy intake (a proxy for energy
balance), we ought to accept the former. There must be some doubt
as to whether the subjects of many supplementation studies and
programmes are as chronically energy deficient as is thought.
Supplements are consumed at
distribution centres as in the INCAP, Gambia and Taiwan studies
or taken home for subsequent consumption as in the Bogota, New
York and WIC programme in the USA. Centre-based consumption has
obvious advantages in ensuring and measuring supplementation but
it does have the disadvantage of being expensive, costly in time
with possible low or variable participation. Take-home
supplementation is now more common, but direct measurement of the
intake of supplement is less frequent, and "leakage" of
the supplement away from the intended recipient has been
identified in many studies. Supplements may be given to other
family members or even sold. BEATON and GHASSEMI (1982) estimate
that 30-80% of take-home supplements do not reach the intended
child.
A problem with both types of distribution is that substitution of some part of the habitual intake is common and that the net increase in intake may be only 45-70% of the supplement, although values as low as 10% have been reported. While most studies monitor supplement consumption closely, measurements of home intakes have not yet been frequent or accurate enough.
In The Gambia, at least, total intakes come within 10% of those of pregnant women in Cambridge, U.K., but in most studies only a small part of the apparent energy deficit is being met, which in itself may explain some of the poor results obtained.
Characteristically, energy supplementation is not the only intervention taking place. Protein intakes increase by similar or greater degrees than energy, for, when many of the studies were planned and begun, protein supplementation was the goal. Vitamin and mineral supplementation is the rule, and health care, preventive and therapeutic, is made available. Separation of the effects of each of these is rarely possible. Other confounding variables, associated in the same direction as the independent variable with the outcome, may be subtler and unrecognised. In examining the effects of a supplement or a placebo on the behavioural development of children, attendance at a distribution centre itself has a stimulatory effect, and both groups will differ from the normal uninvolved child. Thus, our conceptualisation of the treatment is approximate as supplementation has non-nutritional components.
In the
studies considered here, duration of supplementation does not
seem to have been a direct factor limiting the effects and
magnitude of the outcome. The type of supplement, whether based
on local foods or manufactured products, does not seem to affect
uptake of the supplement although it may influence its fate.
BEATON and GHASSEMI (1982) report that premixed infant formulae
are less shared in the family than ingredients of a mix of local
foods. The success in achieving a large net increase in intake in
The Gambia has been attributed to a palatable supplement made
available at a convenient time, early in the morning.
This question refers to two
issues, the representativeness of the target group and their
suitability in terms of their capacity to change nutritional
status.
Supplementation studies require adequate control groups and random assignment of subjects to target and control groups. Under these circumstances, the individual differences in outcome can be ascribed more confidently to the manipulated independent variable, than to confounding factors. Random assignment has been a feature of most research studies (Bogota, New York, Taiwan). In the INCAP study, villages were randomly assigned to treatments. The retrospective comparison of different groups with and without supplementation, as in The Gambia, is less satisfactory because of possible secular trend effects such as those seen in the Taiwan study (ADAIR and POLLITT, 1985). Concurrent measurement of control and target group is recommended.
Varying participation may qualify the effects of randomisation. Weight gain is often proportional to degree of participation and non-attenders differ from regular attenders. In the early years of the INCAP study, less than two-thirds of the births were weighed although in this case no bias was introduced, as in 1970-73, when coverage was 92%, no differences were found compared to earlier figures. In the Taiwan study, non-attenders were followed up in their homes to avoid this effect and to raise levels of supplementation.
Are the
target groups able to respond to supplementation? Given that
malnutrition exists in the community the choice of pregnant and
lactating women, infants and children and workers seems
appropriate, but children under two years are underrepresented in
supplementation studies (BEATON and GHASSEMI, 1982). Infants will
and should be encouraged to breast-feed but the late introduction
of solid foods may be responsible for much of the increase in
malnutrition that begins at this age, and more work on this group
is required. Results from the Bogota study have been interpreted
to suggest that prenatal supplementation may be less useful than
later supplementation (JOOS and POLLITT, 1984). Similarly, direct
supplementation of children may be more cost-effective than
supplementing lactating women.
Anthropometry of mothers,
infants and children has occupied a central place in assessing
the outcome of supplementation. Weight and fat gain or loss have
been taken to indicate nutritional adequacy in growth, pregnancy
and lactation and in the performance of work. Anthropometric
variables can be measured objectively with sufficiently high
precision and measures of central tendency and dispersion
calculated. However, anthropometric variables are proxy variables
for CED and nutritional status. An abnormal or impaired state is
best represented by a change in function or capacity. As yet, few
of these are sufficiently specific to energy nutrition, or not
enough is known about the intervening and confounding factors to
warrant introduction at this stage.
The ability of Gambian women with low body weight and apparent low energy intake to lactate as efficiently as women of higher weights, energy intakes and energy stores is an example of a function being a more appropriate outcome variable than anthropometry. Changes in anthropometric measures are slow to develop, and other outcome variables may change first. Improvements in social responsiveness may precede increases in weight in supplemented children (BARRETT, 1984).
An overemphasis on anthropometric outcomes has contributed to the longevity of the direct main-effect model and led to a misinterpretation of some other effects of supplementation. For instance, the sharing of supplements is a feature of most studies, particularly take-home distribution studies. With outcome variables such as anthropometry or a functional capacity, sharing represents an inefficiency of supplementation.
Beaton has argued convincingly that leakage, that is, sharing plus substitution of the normal intake, can still confer positive benefits on the intended recipient. If the supplement is consumed by others, more home food is available for the recipient and others in the family have benefited (BEATON, 1982; BEATON and GHASSEMI, 1982). Even if the supplement is sold, the income benefits the family, including the intended recipient. Supplementation can thus have multiple effects, many of which are indirect and have not been considered previously and which may be more important than the small changes in anthropometric variables.
To conclude
this section, there have been many problems in the identification
of the undernourished, the degree of supplementation, the
randomness of treatment groups and the choice of outcome
variables such that these have made some contribution to the
observed small effects of supplementation. The possibility of
multiple effects and other evidence considered in the next
section suggest that our conceptualisation of the processes and
effects of supplementation should be modified.
The beneficial effects of
leakage of supplement have been proposed as evidence for a
multiple effect model. In both take-home and centre-based
supplementation there is often a mismatch between the energy
ingested and an anthropometric outcome. Birth weight gains in the
INCAP study were 7 g per 10 MJ, and in the Gambian study weight
change accounted for only 7% of the supplement. Considerable
amounts of energy are missing. BEATON and GHASSEMI (1982) state
that only 15% of the energy distributed to children is utilised
in growth processes. They have attempted to account for the
missing energy of supplementation in children. The best results
are a weight gain of about 1 kg/y requiring about 600 kJ/d. The
extra intakes are some 120800 kJ/d. Increased body size of 1 kg
in a 12-kg child would raise maintenance energy requirements by
about 8% or an average of 170 kJ/d at the midpoint of the year.
This leaves a large apparent disparity between net intake and the
improved growth.
An explanation for missing energy may be that supplementation leads to "de-adaptation", e.g., that basal metabolic rates are increased over and above the size effect, and that physical activity increases. In pregnant Gambian women, LAWRENCE et al., (1984) found the net extra energy cost of tissue maintenance to be 4 MJ. However in women supplemented with amounts sufficient to raise birth weight 190 g but not body weights this increased to 55 MJ. These workers report a 5% higher energy cost of activity in supplemented women, an amount comparable with the extra thermic effect of the ingested supplement (LAWRENCE et al., 1985). It is widely believed by the Gambian villagers that supplemented lactating women have more energy for farming activities (PRENTICE et al., 1980). These observations await verification as REINA, SPURR and BARAC-NIETO (1987) report in their abstract to the Meeting no effect of supplementation on activity energy expenditure in Colombian boys and girls. CHÁVEZ and MARTÍNEZ (1982) have reported differences in behavioural measurements of activity in infants and young children.
Individual differences in responsiveness to supplementation are another argument for refinement of models and hypotheses. The sex of the offspring determines the extent of the response to supplementation. In the Bogota and Taiwan studies, males showed a greater response, probably as a result of their greater vulnerability to environmental conditions. The effects of malnutrition on subsequent mental development vary according to whether the child is brought up in a developed, western country or in environmentally impoverished conditions. The results from the INCAP, Bogota, Taiwan and New York studies, which might be ranked in that order in terms of degree of deprivation, suggest the magnitude of the effects of supplementation (but not the level of significance) depends on the degree of total environmental deprivation. Further factors affecting responsiveness, such as maternal attitude, attendance, distance from distribution centre, etc., illustrate the nature and multiplicity of confounding factors and the need for more detailed models and considerations.
The
importance of the environmental context cannot be overemphasised.
It will influence the nature and severity of CED, the types of
local supplements available, who is affected and what the effects
are. The environment determines the disease pattern, the
responses to infection and the level of health. The proxy
variables of CED and the outcome variables are equally affected
by physical, biological, social and cultural environmental
challenges. The signs of CED and changes in outcome may be
non-nutritional in origin. This mix makes for very complex
situations and it is unlikely that there is one direct linear
pathway linking supplementation to all the outcomes. Our models
should allow for multiple outcomes and non-linear indirect
pathways. These are featured in some contemporary descriptions of
the relationship of nutrition to growth and performance (BEATON
and GHASSEMI 1982; ADAIR and POLLITT, 1985). Our studies should
change from bivariate to multivariate designs, from studies of
outcome variables to studies of the processes occurring, if we
are to better understand the complex processes of human
reactions.