Exact Network Meta-Analysis for Rare Events (306491)Chris Corcoran, Utah State University
Pralay Senchaudhuri, Cytel Software Corporation
*Brinley Zabriskie, Brigham Young University
Keywords: indirect treatment comparisons, mixed treatment comparison, multiple-treatment meta-analysis, heterogeneity, network algorithm
Network meta-analysis is an active area of research in public health and medicine. Network meta-analysis is often part of comparative effectiveness research, which aims to identify the most effective clinical intervention of all the efficacious interventions for a given medical condition. Many of these conditions are considered a rare or adverse event, and very little research has been done to determine to what extent rare events, small sample sizes, and heterogeneity impact the validity of existing network meta-analysis methods. We analyze how existing frequentist methods fare under these circumstances. Additionally, we propose a permutation-based approach to network meta-analysis that provides more reliable results for rare event and heterogeneous data.