Methodological Considerations on Intra- versus Interpopulation Correlation Studies

Abstract

Epidemiological studies on the diet-health relationship have frequently resulted in confusing controversies rather than in clear and consistent findings. As an illustration of this phenomenon the relationship between dietary cholesterol and coronary heart disease (CHD) is used in this presentation. The epidemiological evidence favoring a strong and causal relationship between dietary cholesterol, serum lipids and CHD relied until some years ago largely on studies between populations. However these studies are more susceptible to confounding error. Therefore epidemiological studies have been conducted within populations looking into the relationship between dietary cholesterol, CHD risk and CHD frequency on the individual level. Until recently no consistent association had been observed between dietary cholesterol and serum total cholesterol levels or between dietary cholesterol and CHD mortality in homogeneous groups of individuals; this relates probably to methodological problems in measuring the dietary variable with sufficient precision. More recent studies using a more precise dietary methodology and a prospective study design over longer time periods result in observations that are more in line with the interpopulation study observations. Besides epidemiological evidence it is necessary to compare these findings with those from experimental work and from clinical studies. This approach has the best chance to separate ecological fallacy from real existing associations between diet and health.