Inconsistency for Arm-Based Models in Network Meta-Analysis

• Activity Number: 331
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Hong Zhao*+ and James S. Hodges and Haijun Ma and Qi Jiang and Bradley P. Carlin
• Companies: University of Minnesota and University of Minnesota and Amgen and Amgen and University of Minnesota
• Keywords: Multiple treatment comparisons (MTCs) ; Bayesian analysis ; Inconsistency

Abstract

Network meta-analysis (NMA), also known as multiple treatment comparisons (MTCs), is commonly used to incorporate and compare direct and indirect evidence. With recent advances in methods and software, Bayesian approaches to NMA become quite popular and allow models of far greater complexity. When direct and indirect evidence conflict in a NMA, the model is said to suffer from inconsistency. Current loop-based methods for measuring inconsistency using contrast-based (CB) models are complex and require case-by-case investigation of the evidence network, especially when many multi-arm trials exist in the NMA. We propose methods to detect discrepancy of direct and indirect evidence for comparing two treatments using fixed effects, together with a method for exploring for outlying trials using random effects, both in arm-based (AB) model framework. They permit users to address issues addressed using CB models and are much simpler to implement, without sacrificing any information about indirect comparisons. We compare sources of inconsistency identified by our approach and loop-based method for example and simulated datasets, and our method offers more powerful inconsistency detection.