Some difficulties inherent in the interpretation of dietary data from free-living populations

Abstract

Isocaloric adjustments of diet in metabolic experiments tend to induce negative correlations. In free-living populations, on the other hand, the absolute amounts of major nutrients consumed are positively correlated with total calorie intake and with each other. In general, these positive correlations are fairly strong. Relative amounts of nutrients consumed tend to be weakly and negatively correlated with total calories and each other. Because of these intercorrelations conclusions drawn from cross-sectional population data are ambiguous, at best, and misleading, at worst. Beyond that, the correlation between the intakes of specific nutrients with total calorie intake introduces the possibility that dietary variables may to some degree be surrogates of some nondietary variables; in particular, of physical activity. Unless the questions being asked in population studies are precisely formulated, with a careful consideration of the biological issues and the statistical limitations, biased conclusions are likely to result.