Interpretation of associations in pharmacoepidemiology

Pharmacoepidemiology is an important tool for understanding and documenting the relationships between the use of drugs and adverse events. Side effects are an important consideration for clinicians making treatment decisions, and they also figure prominently in action by regulatory agencies. Unlike the results of randomized trials, epidemiologic findings are based on observational data subject to various types of bias; on the other hand, epidemiology has the advantage of monitoring "real world" situations. This chapter is intended to provide readers with a practical treatment of the issues that are encountered in the pharmacoepidemiology literature so that they can arrive at their own interpretation of the findings. The discussion is focused on issues that are of particular relevance to the epidemiologic study of drug associations with hematologic outcomes, including validity of exposure information, appropriate definition of exposure, confounding by indication and from concomitant use of multiple drugs, sample size considerations, and the relationship between size of an association and rarity of the outcome.