Comparisons in pharmacoepidemiology: New challenges and limitations of current approaches

Background: Undertaking robust comparisons in pharmacoepidemiological (PE) and comparative effectiveness (CE) studies are a major challenge. In these designs certain subject characteristics may influence the probability of exposure and the researcher cannot control allocation to treatment or the conditions under which treatment is given. Hence there is the potential for non-random selection into a cohort, and if appropriate, the comparator group(s). Methodological advances in statistical analyses allow exploration of observed (and unobserved) characteristics affecting selection, but have their limitations. Problems with comparisons have become more difficult with the widespread application of clinical guidelines (national/ regional), pharmacoeconomic policies and policies for reimbursement, all of which determine treatment choices thus introducing selection biases which are beyond the capabilities of existing methods to handle. Objectives: To explore current challenges to the design of postmarketing studies of medicines for which there are significant external influences governing use and discuss possible solutions.