[Poisson regression in epidemiology].

The goals of Poisson regression are to estimate the effects of risk factors on incidence or mortality rates and to evaluate the dose-response relationship for variables representing quantitative levels of exposure. This statistical model presents, over standardization techniques, several advantages which are recalled here. The Poisson assumption, approximating the exact sampling distribution of rates is discussed. The model equation with internal and external standard rates is introduced, as well as summary measures of fit. Two examples are given: one from a cohort study, the other one from a geographical study, to help in the interpretation of Poisson regression results. Lastly, recent and on-going developments of these models are briefly considered.