Contents - Previous - Next

Differentiating among the contributions of energy and other nutrients

Some analysts have attempted to separate the specific contributions of energy from those of protein and other nutrients in the supplements.

Pregnant women. This has been done most successfully in pregnant women who ingested similar quantities of energy from the Atole and the Fresco by ingesting much greater quantities of the latter. The dose-response on energy was the same in both the Atole and Fresco groups. This would not have been the case if protein or other nutrients present in Atole but not in Fresco (e.g., calcium, phosphorous) had been limiting in which case the dose response of birth weight to energy would have been higher (steeper slope) in the Atole group. Nor would it have been the case if micronutrients added to the supplements had been limiting. These measured micronutrients (Table 2 in Martorell et al. 1995) were added in equivalent concentrations per volume to both supplements in 1971. Therefore, the dose-response would have been higher in the Fresco group if the measured micronutrients had been limiting because the micronutrient to energy concentration was higher in the Fresco group. Furthermore, there would have been no dose response in the Fresco group previous to 1971. This indicates that neither protein nor the other nutrients were limiting factors for fetal growth in the home diets of these mothers, but total energy was.

As discussed previously, the same conclusion results from step-wise multiple-regression analysis that reveals that energy is still statistically associated with birth weight, even when protein and the other measured nutrients are taken into account, whereas the converse is not true. Protein and the other nutrients are not statistically associated with birth weight when energy is taken into account. The above inference that energy is more deficient than protein in these diets is borne out by direct examination of the home diets. The mean of the home diet intakes for utilizable protein was slightly above the recommended daily allowances (7.4%) in contrast to the mean energy intake which was 39% less than the recommended daily intake (Lechtig et al. 1975b). Although the reliability of the individual measures of home diet is poor, that of the means is good (Habicht and Martorell 1992). Given this fact, plus some assumptions about the distribution of intakes around these means and about the dietary requirements, and above all, given the results about the relative effects of supplemental energy and protein, we conclude that energy was likely to be much more limiting than protein for these women.

Children. In children there was much less overlap between Atole and Fresco groups in energy supplement consumption (Schroeder et al. 1992). Atole children consumed much greater amounts of energy in the first three years of life. Analyses in Fresco children found similar or larger growth responses to energy intake than in Atole children, resulting in the conclusion that energy and not protein was limiting (Yarbrough et al. 1978). However, the results do not exclude the possibility that energy is limiting at lower levels of supplement intake as seen in the Fresco villages, but that protein may be limiting at higher levels of intake as seen in the Atole group.

The difference in slopes could not be because of a protein effect at lower levels of supplementation. They might be due to the flattening of the dose-response curve (Yarbrough et al. 1978) as energy intake approached adequacy. The higher response in Fresco also could be interpreted as evidence that another nutrient was limiting because their concentrations in relation to energy were much greater in the Fresco compared with the Atole.

In conclusion, none of these competing hypotheses have found resolution to date, possibly because resolution may not be possible in this data set. This lack of clarity about the exact nutrients that were responsible for the supplements' impact does not affect in any way the inference about a causal effect of the supplementation program on child growth to 3 y of age.

One claim that the effect of the supplement on growth was solely due to protein in the supplement (Balderston et al. 1981) is based on incorrect interpretations, as discussed below, from analyses using the home diet.

Taking home diet into account in dose-response analyses. The objective of the INCAP longitudinal study was to improve nutrition. It is therefore important to know whether the energy and nutrients from the supplements were supplemental or simply displaced home diet consumption. A direct approach would appear to be the use of total dietary intake (home diet + supplement) in the analyses. Alternatively, home diet and supplementation may be used as separate variables in a multivariate equation explaining the outcome. Both approaches would capture the net improvement in energy or nutrient intakes when comparing children from Atole and Fresco villages. Unfortunately, the low reliability of the home diet data usually precludes finding associations between the home diet and the outcomes. For example, Schroeder et al. (1995) found home diet energy to be much less related to growth increments than supplement energy. Rivera et al. (1995) increased the reliability of home diet energy by combining as many as eight separate surveys per subject and by creating a dummy variable, above or below the median; even though the diet variable was statistically significant and in the expected direction, analyses showed its inclusion in the model did not affect the regression coefficient for supplement. This indicates that the range in energy and protein intake from the home diets is too small to be important in the analyses. For this reason, home diet is also usually a de facto constant when included in analyses involving supplement because of the imprecision of its regression coefficients with the outcomes. Nonetheless, some analyses (Rivera et al. 1995, Schroeder et al. 1995) include home diet to increase persuasiveness as many readers would be troubled by the omission of this variable.

Inappropriate inclusion of home diet in the analyses and poor interpretation of the results can lead to false inferences as exemplified by the analyses that led to the claim that it was protein and not energy supplementation that produced better growth in children consuming Atole (Balderston et al. 1981). This claim was made on the basis of two findings: The first was: "The large effect [on growth] of increments of Atole supplementation for children eating the same home diet - contrasted with the small effect of increment in home diet for children consuming the same amount of supplement - is not consistent with the hypothesis that the total energy value of the supplement is what accounts for the gains of children in Atole villages" (1981:59). In fact, this contrast was even stronger for protein (not reported) and was solely because of the poorer reliability of home diet compared with that of supplement intake. When the independent variable is very poorly measured, the estimates of magnitude (regression coefficients) are biased so they approximate 0 (Habicht et al. 1979). The correlation coefficients are also small, but this does not depend upon whether or not home diet is an independent or dependent variable.

The second finding in Balderston et al. (1981) was that in multiple regression analysis of the total diet (i.e., sum of home diet and supplement) there was a strong association between total protein ingestion and growth and none between total calorie ingestion and growth. This was entirely because of the fact that the Atole had a greater impact on total protein intake (on the average 35%) than on total energy intake (on the average 17%; estimates from Martorell et al. l982 and from WHO 1985). As noted above, the variability in supplement ingestion is much better correlated with growth than is variability in home diet because of home diet's poor reliability. Therefore, for equal variability of supplement intake, supplemental nutrients that contribute a greater proportion of total dietary intakes will be more highly correlated with growth. This is particularly the case in children under 3 y of age because energy and protein intakes from the supplements are highly correlated with each other; energy ingestion from the Fresco was very low in relationship to Atole.

Home diet data are important, however, in investigating the degree to which the supplement replaced rather than supplemented the home diet. One cannot estimate the amount of supplement used to replace the diet at different levels of diet because this requires using the total diet as the independent variable. On the other hand, the estimates of amount of home diet replaced by the supplement at different levels of supplementation can be estimated without bias because in this analysis supplementation is the independent variable and it is measured almost perfectly. The estimated level of replacement for supplemental energy was 22% for pregnant mothers (calculation from Table V in Lechtig et al. 1975a), but negligible for children (Martorell e t al. 1982).

These figures give the apparent proportion of supplement that substitutes for the diet. It might be that high ingestors of supplement would not have had enough food at home to bring their home diets to the same levels of intake as low ingestors, even if they had not consumed the supplement. This is indeed likely because high-energy ingestion from the supplement is related to lower socioeconomic status (Johnson 1988, Schroeder et al. 1992).

These magnitudes still will be somewhat underestimated because the outcomes are affected by variations in nutrition, not because of the supplement. The most important variation is home diet. This variation reduces the power to find associations between measures of supplementation and outcomes so that both the statistical significance and the regression coefficients relating a nutritional component of the supplement to the outcome understate this nutritional relationship (Habicht et al. 1979). In principle, this effect of home diet is no different from the influence of other factors that affect an outcome and that are randomly distributed across different levels of supplementation. The omission of home diet from the analysis of the impact of supplement on outcome measures is no different from the omission of other variables that affect growth but that are not related to the supplement, if one has taken the confounding because of home diet into account. Such omissions are inevitable. The confounding because of home diet is dealt with by correcting for the apparent substitution of supplement for home diet. For instance, the magnitude of the effect of supplement on birth weight is ~20% higher than the figures published when this correction is made. This means that the response of birth weight to actual energy supplementation is ~35 g per 10,000 kcal (41,840 kJ).


This paper presents the statistical significance of causality for an effect of the supplementation of the growth of 3-y-old children. It presents credible evidence for an effect of the supplement during pregnancy on the birth weight of infants. Finally, it presents evidence and the significance of causality for the effect of the supplement on the recuperation of malnourished children. The combination of the probability tests for causality with tests of association for credibility is necessary to make the most persuasive argument that the supplement had a nutritional effect on the outcome of concern. Credibility analyses are always possible and should always be done. Where the probability analysis for causality cannot be done, more analyses for credibility are necessary.


We thank Catherine Geissler for comments on an early draft of this paper.

Literature cited

Balderston, J. B., Wilson, A B., Freire, M. E. & Simonen, M. S. (1981) Malnourished Children of the Rural Poor. Auburn House, Boston, MA.

Cohen, J. (1988) Statistical Power Analysis for the Behavioral Sciences, 2nd ed.Lawrence Lawerence Associates, NJ.

Habicht, J.-P. & Martorell, R. (1992) Objectives, design and implementation of the INCAP longitudinal study. Food Nutr. Bull. 14: 176-190.

Habicht, J.-P., Yarbrough, C. & Martorell, R. (1979) Anthropometric field methods: criteria for selection. In: Human Nutrition, Nutrition and Growth (Jelliffe, D.G. & Jelliffe, E. F. P., eds.), vol. 2, pp. 365-387, Plenum Publ. Corp., New York.

Johnson, C. S. (1988) The role of participation with nutritional supplementation during pregnancy: a comparison of data from Indonesia and Guatemala. Masters thesis, Cornell University, Ithaca, NY.

Judge, G. G., Griffiths, W. E., Hill, R. C. & Lee, T.-C. (1980) The Theory and Practice of Econometrics. John Wiley and Sons, New York.

Kupper, L. L. (1984) Effects of the use of unreliable surrogate variables on the validity of epidemiologic research studies. Am. J. Epidemiol. 120: 643-648.

Lechtig, A., Habicht J.-P., Delgado, H., Klein, R. E., Yarbrough, C. & Martorell, R. (1975a) Effect of food supplementation during pregnancy on birth weight. Pediatrics 56: 508-520.

Lechtig, A., Yarbrough, C., Delgado, H., Habicht J.-P., Martorell, R. & Klein, R. E. (1975b) Influence of maternal nutrition on birthweight. Am. J. Clin. Nutr. 28: 1223-1233.

Lutter, C. K., Habicht, J.-P., Rivera, J. A. & Martorell, R. (1992) The relationship between energy intake and diarrhea! disease in their effects on child growth: biological model, evidence and implications for public health policy. Food Nutr. Bull. 14: 36-42.

Martorell, R., Habicht, J.-P. & Klein, R. E. (1982) Anthropometric indicators of changes in nutritional status in malnourished populations. In: Proceedings Methodologies for Human Population Studies in Nutrition Related to Health (Underwood, B., ed.), NIH Publication #82-2462, pp. 96-110, U.S. Government Printing Office, Washington, DC.

Martorell, R., Habicht, J.-P, & Rivera J. A. (1995) History and design of the INCAP longitudinal study (1969-77) and its follow-up (1988-89). J. Nutr. 125: 1027S-1041S.

Rivera, J. A. (1988) Effect of supplementary feeding upon the recovery from mild-to-moderate: wasting in children. Doctoral thesis, Cornell University, Ithaca, NY.

Rivera, J. A., Habicht, J.-P. & Robson, D. S. (1991) Effect of supplementary feeding on recovery from mild-to-moderate wasting in preschool children. Am. J. Clin. Nutr. 54: 62-68.

Rivera, J. A., Martorell, R., Ruel, M. T., Habicht, J.-P, & Haas, J. (1995) Nutritional supplementation during preschool years influences body size and composition of Guatemalan adolescents. J. Nutr. 125: 1078S-1089S.

Schroeder, D. G., Kaplowitz, H. & Martorell, R. (1992) Patterns and predictors of participation and consumption of supplement in an intervention study. Food Nutr. Bull. 14: 191-200.

Schroeder, D., Martorell, R., Rivera, J. A., Ruel, M. T. & Habicht, J.-P. (1995) Age differences in the impact of nutritional supplementation on growth. J. Nutr. 125: 1060S-1067S.

Snedecor, G. W. & Cochran, W. G. (1980) Statistical Methods, 7th ed. Iowa State University Press, Ames, Iowa.

World Health Organization (1983) Measuring change in nutritional status. WHO, Geneva, Switzerland.

World Health Organization (1985) Energy and protein requirements. WHO Tech. Rep. Ser. 724.

Yarbrough, C., Habicht, J.-P., Klein, R. E., Martorell, R., Lechtig, A. & Guzman, C. (1978) Response of indicators of nutritional status to nutritional interventions in populations and individuals. In: Evaluation of Child Health Services: The Interface Between Research and Medical Practice (Bosch, S. & Arias, J., eds.). DHEW Publ. No. (NIH) 78-1966, pp. 195-207, U.S. Government Printing Office, Washington, DC.

Contents - Previous - Next