The European Prospective Investigation into Cancer and Nutrition (EPIC): plans and progress.

The hypothesis that diet and related metabolic, anthropometric and hormonal factors could play a role in cancer etiology was originally supported by a series of early case-control studies, geographical correlation studies (also called ecological studies) (NRC 1982) as well as pioneering work on rodents in experimental laboratory studies carried out in the 1940s (Tannenbaum 1940a, 1940b, 1942a and 1942b). During the first half of the century, several researchers noticed that overfed rodents were less resistant to the growth of transplanted tumors (Rous 1914) and more sensitive to the effect of chemical carcinogens (Tannenbaum 1942a). Several decades later, correlation studies showed that incidence of (and mortality from) cancer of the breast, colorectum and prostate were positively correlated with the foods most typically consumed in Western societies (i.e., meat, total and animal fat, simple sugars) and negatively correlated with the consumption of various vegetable foods (grains, cereals and vegetable fiber) (Armstrong and Doll 1975). Geographical correlation studies can indicate only that disease risk and the prevalence of a given factor are correlated across different populations; they are limited by the methodological and practical possibility of taking into account confounding factors that may create spurious correlations at the population level. During the past 20 years, a considerable number of retrospective case-control and, more recently, prospective cohort studies have been conducted to investigate whether, in each given population, these dietary factors were effectively related to cancer risk at the individual level. Results of epidemiologic and experimental studies on nutrition and cancer have been reviewed in depth in recent years by three independent expert committees as follows: