[The epidemiologist facing assessment of low risk].

Epidemiological studies on low risks involve a number of major methodological difficulties. This review defines the concept of low risk and emphasises the need to control all large, possibly disease-related risk factors, before low risk factors can be evaluated. This procedure is also essential in identifying confounding factors. There is a special need to evaluate statistical power in studies on low risk factors, and the representativeness of both cases and controls should be carefully verified, particularly in case-control studies. The need to control the researcher's wish bias and the main requirements for meta-analysis are also presented.