A1046
Title: A comparison of two inconsistency detecting models for network meta-analysis
Authors: Ke Yang - Beijing University of Technology (China) [presenting]
Lu Qin - Jilin University (China)
Tiejun Tong - Hong Kong Baptist University (Hong Kong)
Wenlai Guo - The Second Hospital of Jilin University (China)
Shishun Zhao - Jilin University (China)
Abstract: The application of network meta-analysis is becoming increasingly widespread, and detecting consistency assumptions has always been one of the most concerned issues. The detection results can serve as a criterion for evaluating the effectiveness of network meta-analysis results. Several methods to detect inconsistency have been proposed. Among them, the design-by-treatment interaction model and the side-splitting models are most commonly used. These two types of models are compared within a frequentist framework. By simple examples of networks with three treatments, it is found that the side-splitting models are specific instances of the design-by-treatment interaction model with additional assumptions. The side-splitting models perform better when these assumptions hold. On the other hand, the design-by-treatment interaction model exhibits robust performance across different data structures. Based on the findings, it is suggested to employ the design-by-treatment interaction model in practical use, with the side-splitting models serving as a supplementary method for inconsistency detection in network meta-analysis.