Abstract
In pharmacoepidemiology studies, the nature of the research question will dictate the choice of methodological approach and the conditions for optimizing the level of evidence. Thus, to document the treated population and the modes of use of a new drug in real-life prescribing conditions, a descriptive approach through cross-sectional or longitudinal studies conducted on databases, or else ad-hoc studies, will be preferred. On the other hand, evaluation of the real-life "effectiveness" of a new drug will be based on cohort, case-control or scientific modeling, depending on the drug and the disease of interest. For questions involving drug risks and safety, it is the adverse effects profile that will guide the choice of study design, both for identification of the effect (signal) and assessment of causation. In all cases, in the post-marketing authorization (MA) setting, the evidence acquired in pre-MA studies serves as the basis for generating hypotheses. Whatever the research question and the method chosen to address it, the potential biases and their impact on the results need to be identified. In certain cases, a combination of several complementary approaches may prove preferable to a single study.