Invited commentary: the challenge of multi-center cohort studies in the search for diet and cancer links.

The investigation of the role of nutrition in cancer etiology and the perception of its importance among concerned scientists has gone through various phases over the past 20 years, ranging from the enthusiasm of Candide, Voltaire's hero (we are in the best possible world and we know everything) to Hume's kind of skepticism (the subject is appealing but causality can never be shown). These ups and downs have been marked by the excitement caused by reports which have emphasized the predominant role of diet in cancer causation (1-3) or the disappointment of large cohort studies that did not find what they were supposed to find, e.g., that fat causes breast cancer (4) or that (vegetable) fiber protects against colorectal cancer (5). In the 1970s, descriptive epidemiology studies clearly showed that the incidence of cancers of the breast, colorectum, and prostate is substantially higher in populations that share economic development and a Western life-style, and that the incidence is positively correlated with the typical Western diet characterized by high consumption of meat, total or saturated fat, and highly refined carbohydrates, and negatively correlated with consumption of starchy foods and plant foods in general. This so-called "ecologic evidence" has been corroborated by the results of the vast majority of case-control studies conducted in different populations around the world. In a recent and comprehensive report by the World Cancer Research Fund (WCRF) (6), almost 200 casecontrol studies on food and nutrient intake were cited (not counting those on alcohol, body mass index, and physical activity) and, from a rough calculation, about two-thirds of them show significant results in the anticipated direction (e.g., that fruit and/or vegetables protect against a variety of cancers, that red meat