Did My Drug Cause This? a Biostatistical Perspective

Abstract

The role of conditional probability in evaluating causes of adverse reactions is studied. Conditional probability is used in both randomized clinical trials and in postmarketing surveillance for identification of causality. The identification of cause is greatly simplified in randomized controlled trials because direct comparisons may be made on patients taking the drug against those who did not. Determination of cause in postmarketing surveillance is more difficult because there is no comparison group. Conditional probability, as expressed by Bayes rule, may be used in this setting. A method for evaluating the sensitivity of Bayesian methods to the assumptions about prior probabilities will be presented.