Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
All of these data and their respective documentation are available at cost on condition that the researchers keep INCAP (Juan Rivera) and Cornell University's Division of Nutritional Sciences, Office of Computing and Statistical Consulting (Edward A. Frongillo) informed of proposed analyses and deliver to each of these institutions a copy of any resulting reports or publications. The two institutions maintain and update an identical master data tape and keep each other informed about the data set.
Releasing data in this fashion has the disadvantage that wrong inferences may be made by users not familiar with the study. Thus, analysts are encouraged to make arrangements to involve somebody who understands the data—either an original team member or an analyst who has analysed the pertinent variables with an original team member. A formal agreement is recommended to prevent conflicts of interest where both parties are analysing the data for the same or too similar an objective. The larger benefits to society of having these data in the public domain outweigh the disadvantages. For example, important inferences made in previous publications can be reanalysed and either corrected or validated. Equally important, many more insights will be gained from the data set if it is widely available.
Prior to making the data set available to the public, considerable work was done in 1984 at Cornell and INCAP to organize, complete, and document its various components. By this time, numerous versions of the tapes were in a variety of institutions, and their documentation was incomplete. The formidable task of creating a unified data set was made possible by a contract from the Office of Nutrition of USAID (AID-TA-C/1224) to J-P. Habicht of Cornell University and a grant from the Rockefeller Foundation (FR-73-40-E7352) to INCAP. The effort at Cornell was managed by Kathleen M. Rasmussen assisted by Nancy Mock, and that at INCAP by Hernán Delgado assisted by Bruce Newman and Peter Russell.
The experimental design had many strengths. First, data from the study showed clearly that protein-energy malnutrition was prevalent among mothers [37] and children [13]. The design itself randomized the treatment to each village to deal with intrinsic differences between villages, excluding as well confounding due to characteristics associated with attendance at the feeding centre and intake of supplement. The above applies only if appropriate analyses are carried out; that is, if the original village-level design is taken into account in specifying the analytic models [38]. When appropriate analyses are applied to the child growth data, a statistically important physiological improvement in the atole villages is shown, as suggested previously [10].
Supplement ingestion data on individuals made it possible to assess whether village-level effects were compatible with dose-response relationships observed across children, such as the recovery of malnourished children taken as a function of supplement intake [39]. These demonstrations increased the persuasiveness of the findings because they excluded the possibility that the measurers were influenced by knowledge of the village treatments. For example, the anthropometrists were unaware of the amount of supplement consumed by individuals. Thus, better growth in the children in the atole villages than in the fresco villages can be ascribed to the ingestion of atole, although it is not possible to be sure to what degree this is due to energy or to protein [10, 38].
The one weakness of method, rather than of design, was the impossibility of ascertaining total dietary intake reliably enough to permit multivariate analyses of the impact of home diet or total diet on most of the outcomes of interest. This weakness is universal but has been generally recognized only recently. Previously, understanding the impact of the supplementation was difficult for nutritionists who could not see why one did not use total dietary intake measures in the analyses.
A key weakness of the design was that it had low statistical power because there were too few units of randomization. Another weakness was that the study was based on the assumption that protein was the cause of malnutrition in both mothers and children. Fortunately, the different contents of energy in the supplements together with very large intakes of fresco permitted the identification of energy as being more limiting than protein in the diets of pregnant women [37]. Such was not possible in the case of children, because the ingestion of energy from fresco was very low at the ages when atole had marked effects on growth [10, 38]. Thus for children the strong evidence for atole's beneficial effect inferred from the randomized design cannot differentiate between an energy and a protein effect.
Another weakness of the design was the inability to separate completely the components of the INCAP presence and its medical and nutritional interventions. Thus, different participation rates across treatments resulted in village-level differences in exposure to some aspects (e.g., contact with supplementation personnel). Fortunately, we can partition out participation rates from the ingestion of the supplement. Thus we can show that the effect on anthropometry in newborns and children is due to the ingestion of supplement and not to factors associated with participation.
It is necessary to demonstrate that the supplement improved the nutrition status of children in order to infer that associations between supplementation and behavioural outcomes are nutritional. The strength of this inference is greatest when the relationship is with energy in maternal supplementation, or with energy or protein in child supplementation. Plausibility is poorer if the association is with the micronutrients in the supplement, because these are confounded with the amount of supplement ingested, a proxy for confounding associated with self-selection. Furthermore, no physiological response has been noted in women that can be associated with the amount of supplement or amount of micronutrients ingested when supplemental energy is taken into account.
In conclusion, the INCAP longitudinal study produced the strongest evidence then available that supplementary feeding in a malnourished could effectively improve the nutrition status of children during life in utero and in the first years thereafter. Some of these findings have since been corroborated by even stronger designs, whereas for others the results of this study remain the most convincing.
Data collection for the INCAP longitudinal study was supported for the most part by contract HD-5-0640 from the US National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Md. Two lesser but nonetheless important sources of support were contract AID-TA-C/1224 from the Agency for International Development, Washington, DC, and grant 73030-E7352 from the Rockefeller Foundation.