Statistical methods to assess and minimize the role of intra-individual variability in obscuring the relationship between dietary lipids and serum cholesterol

Investigations involving comparisons of populations, as well as intervention studies on groups, both under free-living and metabolic ward conditions, unequivocally demonstrate significant relationships between lipid components of the diet and serum cholesterol levels. However, low order or no correlations have been reported when these dietary variables and serum cholesterol are related for individuals. Several factors underlie this apparent paradox. This paper discusses at length the role of intra-individual variation in this apparent paradox, with particular emphasis on the errors caused by intra-individual variation in classification and estimation of correlation coefficients. Methods are presented for estimating the number of measurements required to achieve a suitable degree of accuracy for both classification and correlation analysis. Based on these theoretical considerations and practical examples illustrating their application, an assessment is made of commonly used dietary survey methods in terms of their appropriateness for measuring individual dietary intake. Suggestions are made for improving the design and methods of nutritional surveys, especially with regard to reducing errors introduced by intra-individual variation.