|
Abstract:
|
Meta-analysis of interventions usually relies on randomized controlled trials (RCTs). However, when the dominant source of information comes from the pool of single-arm studies, or when the results from RCTs lack generalization, methods synthesizing both evidence are reasoned and important. Undeniable, single-arm studies may be less reliable compared with RCTs due to selection biases and so forth, which brings new challenges to the synthesis process. In this paper, several Bayesian hierarchical methods are proposed to synthesize RCTs and single-arm studies directly, incorporating heterogeneity across trials, differences between designs and potential biases from single-arm studies. The proposed methods are applied to two motivating data sets and their performance is evaluated through extensive simulations.
|
