BACKGROUND: The most consistent result of epidemiological studies on diet and cancer is that a diet rich in vegetables, fruit and, more generally, in plant foods is associated with a reduced risk of cancer at several anatomical sites. Epidemiological studies have been less consistent regarding the putative increase in risk related to consumption of fat or meat. In addition it has not been possible to identify clearly the biological role of specific nutrients or non-nutrient food components in the prevention or causation of cancer. Limitations in the precision and validity of traditional dietary intake measurements and limited use of biomarkers combined with narrow ranges of variations in dietary habits within single populations, have been the main reasons for the limited success in identifying more specific diet and cancer links. METHODS: EPIC is a multi-centre prospective cohort study designed to investigate the relation between diet, nutritional and metabolic characteristics, various lifestyle factors and the risk of cancer. The study is based in 22 collaborating centres in nine European countries and includes populations characterized by large variations in dietary habits and cancer risk. Data are collected on diet, physical activity, sexual maturation and reproductive history, lifetime consumption of alcohol and tobacco, previous and current illnesses and current medication. Following a common protocol and using identical equipment, blood samples are collected, aliquoted into plasma, serum, white blood cells and erythrocytes, and stored in liquid nitrogen at -196 degrees C for future laboratory analyses on cancer cases and matched healthy controls. Anthropometric measurements are taken according to a standard protocol. It is planned to include around 400,000 middle-aged men and women. RESULTS AND CONCLUSIONS: The collection of questionnaire data, anthropometric measurements and blood samples is under way. Almost 340,000 subjects had been included in the study by mid-1996, and recruitment is expected to be almost complete by 1997. Follow-up for cancer incidence and total mortality has started and it is expected that about 23000 cancer cases will be identified during the first 10 years of follow-up.