Confounding is a great challenge in pharmacoepidemiology. Confounding may be present in observational studies of any type if a predictor of the outcome is imbalanced across the exposure groups, a situation often seen when treatment allocation is not random but driven by clinical decisions. There are multiple strategies to deal with confounding. In this session, two teams of two experts each will present their best knowledge on whether confounding should be addressed from a subject matter perspective or from a data-driven one. The teams will present arguments and rebuttals, and the audience will have the opportunity to ask questions and interact.
