A Bayesian model for meta analyses of safety studies where the outcome is interval censored in some studies

Meta analyses are routinely used for synthesising evidence from safety studies in the context of health technology assessment (HTA). Thereby, data from systematic literature reviews have to be used and analyses often have to be based on single arm studies. This leads to data bases where frequently not all studies report the outcome of interest but only a subcategory of the event of interest. Specifically, while the number of grade 3/4 adverse events (AE) may be of key interest, some studies may only report results on specific or most frequent AE by type and severity. Instead of ignoring the latter studies, one can use the fact that they provide a lower bound for the outcome grade 3/4 AE. In order to use this information, we implement a Bayesian hierarchical model for exact and interval-censored binomial outcome and discuss potential pitfalls of such model. The proposed model can be used for any situation where only interval censored binomial outcomes are observed.

About the speaker
Manuel Wiesenfarth

Manuel Wiesenfarth is a biostatistician at the German Cancer Research Center Heidelberg (DKFZ) and at Cogitars GmbH, a consulting company for innovative clinical trials. At Cogitars he provides statistical support for adaptive Bayesian phase I/II trials including phase I adaptive dose-escalation for single agent or combination treatments incorporating pre-clinical or existing historical information (evidence synthesis) and phase II basket trials adaptively combining evidence from multiple arms, as well as for Health Technology Assessment.
He received his diploma degree in statistics from the university of Munich (LMU) and his PhD from the Georg August University in Göttingen.

Cookies help us deliver our services. By using our services, you agree to our use of cookies. Learn more