All Times EDT
Keywords: adverse reactions, family-wise error rate, influence function, multiplier bootstrap
Identifying and assessing the safety of a medical product, which is evaluated through its adverse event profile, is an essential step in the clinical development program. One particular objective that needs to be answered in the overall clinical development package is to identify and estimate the frequency of the common (usually non-serious) adverse reactions (ARs) that are, or may be, causally related to the use of the drug. The goal of ascertaining common ARs is to compare the incidence and provide appropriate information about absolute risks in perspective in product labeling. Many ARs are correlated since they are directly related to the mechanism of the drug and the disease, leading to challenges in estimating the incidences of ARs with precision. To improve the determination of these ARs, we develop a general framework for constructing the simultaneous confidence intervals (CIs) of the incidence rates or the differences between two drugs. The proposed methodology accounts for covariate effects and differential exposure times. Furthermore, correlations among common ARs are completely unspecified and estimated from the data. The proposed simultaneous CIs have correct coverage probabilities and shorter average lengths than the existing methods that fail to account for the correlations. Extensive simulation studies and real applications are provided.