|
Abstract:
|
In this talk, I will discuss a Bayesian approach to meta-analysis of safety data from multiple clinical trials. Meta-analysis of safety data is a critical tool for combining evidence, but poses several challenges. In particular, many adverse events are rare, and adverse events which occur infrequently may be omitted from official trial reports. The observed data therefore correspond to low counts which may be subject to left censoring. As an additional challenge, there are many risk factors of interest, including subtypes of disease, classes of drugs, and their interactions. Consideration of all possible combinations makes accurate effect estimation difficult and raises issues with multiplicity. We address these challenges in a Bayesian hierarchical model for rare and censored events. To identify high-risk subgroups, we impose sparsity on the interaction terms through the use of horseshoe priors. We demonstrate through simulation studies that our method enables the identification of relevant interactions and improves the accuracy of risk prediction over existing methods. Finally, we illustrate our method by applying it to a meta-analysis of adverse events in cancer immunotherapy.
|
