Considerations on causality in pharmacovigilance

Abstract

Causality has been a topic of debate by philosophers, scientists, lawyers and for centuries. It is essential to define as precisely as possible all steps in the logical chain of events, since each may strengthen or confound an argument. Almost always there are issues of missing and conflicting data that need to be addressed specifically. In pharmacovigilance, as in many other situations, there is not just one possible causation for an effect but several. Each must be evaluated in the given context for probability. There is also likely to be a causal chain of events to the adverse effect under consideration, and each must be considered. In an individual patient diagnosis the components of patient history, clinical findings and various laboratory test findings are combined to point to the probability of the patho-physiological diagnosis, which in turn is related to possible causes with a strength determined by the constellation of findings. The established Bradford-Hill criteria are valuable in considering all the possible causal factors. Pharmacoepidemiology allows for population incidences of causes for particular effects and therefore provides an a priori probability listing for competing possible causes, or at least of one possible cause against the background of all others in a control group. Since adverse effects of medicines are generally rare, it is not possible to exclude drug causation in an individual by reliance on epidemiological evidence alone, only to argue that the incidence is below a level determined by statistical power, of the study or studies combined. Other areas of society are concerned with the process of causal inference, and this is especially true in legal cases in which judgements are made on possible personal injury by drugs.