Strategies for analyzing nutritional data for epidemiological purposes--conceptual framework.

The relation between nutritional factors and health investigated in epidemiological studies are often inconsistent. One of the reasons for such findings can be the improper addressing of the multitude of nutritional dimensions in the specific study situation such as physiological individuality of human beings, different living conditions, or numerous interdependencies between nutritional variables. Epidemiological research in nutrition and health should recognize such facts and work with appropriate study models and adequate data analyses. Instead of investigating heterogeneous populations it is advisable to concentrate on specific "types" of people. Under consideration of the study goals such "types" can be compiled according to physiological properties, e.g., cholesterol sensitivity, or biological-constitutional factors such as body build, life-style entities, or other factors. The variety of nutrition factors far beyond the commonly applied nutrient values can be expressed in integrated indices of "food patterns". Such "food patterns" can be derived in many ways. They can be deduced from theories by using specific criteria, but also explored by modern multivariate statistical analyses. The ways leading to "food patterns" are discussed. The ideas presented and discussed in this paper lead to an improved model for research in the field of nutrition and health with integrated indices of "food patterns" as the critical point. It is assumed that using this approach will generate new insight in the relation of nutrition and health, a currently still diffuse research area.