Something Out of Nothing? The Influence of 0-0 Studies in Bayesian Meta-Analysis of Drug Safety Studies

• Activity Number: 29 - Statistical Issues Specific to Therapeutic Areas, Power and Sample Size Calculations, and Trial Monitoring
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Zhaohu(Jonathan) Fan* and Dungang Liu and Yuejie Chen and Nanhua Zhang
• Companies: University of Cincinnati and University of Cincinnati and North Carolina State University and Cincinnati Children's Hospital Medical Center
• Keywords: meta-analysis; drug safety

Abstract

Examining the safety of new drugs is important in clinical trials. In clinical studies, the occurrence of adverse events (e.g., deaths) is often rare. However, in the case of rare events, a single study may not produce sufficient power to detect meaningful signals. The meta-analysis, which synthesizes multiple studies, is widely used to analyze rare events data. When events are rare, some of the studies may observe zero events in both treatment and control groups. These studies are referred to as double-zero studies. The influence of double-zero studies has been researched in the literature, but it still remains unsettled. Some argued that in theory, these studies contain information for inference, whereas others do not observe their influence in numerical studies. This paper examines when and how double-zero studies contribute to inference in Bayesian analysis. Through extensive numerical studies, we demonstrate that 1) type I error can be significantly inflated, 2) the testing power can be significantly decreased, and 3) bias can be increased if double zeros event studies are excluded from the analysis.