[Confounding and causality in epidemiology]

Since the beginning of the discipline epidemiologists developed a specific methodology to assess causal relationships, even opposed to statistics. A special effort to assure unbiasedness explains the importance of the concept of confounding in epidemiology. However, there is an intrinsic conflict in the formal definition of confounding based on association criteria. This conflict is solved in practice by gathering information on the subject specific matter. Causal criteria for confounding generalize with regards to identifiability of causal relationships in observational studies.