Despite the fact that population databases and health insurance databases are increasingly used for designing comparative pharmacoepidemiological studies, SRs remain the main source for pharma-covigilance decisions. In a classical study, e.g. a comparative cohort, the incidence rates a1/N1 and a2/N2 measured in each group (exposed and not exposed to the studied drug, respectively) are compared. In the framework of SR, the validity of such a comparison is jeopardized (i) because of the absence of information on the actual number of cases which have occurred during the considered period of time, and (ii) because of the questionable character of the estimates of the denominators Ni and N2. We will address both issues separately.
Was this article helpful?