Use of Generalized Linear Mixed Models for Network Meta-analysis

In the past decade, a new statistical method—network meta-analysis—has been developed to address limitations in traditional pairwise meta-analysis. Network meta-analysis incorporates all available evidence into a general statistical framework for comparisons of multiple treat-ments. Bayesian network meta-analysis, as proposed by Lu and Ades, also known as ‘‘mixed treatments compari-sons,’ ’ provides a flexible modeling framework to take into account complexity in the data structure. This article shows how to implement the Lu and Ades model in the frequentist generalized linear mixed model. Two examples are provided to demonstrate how centering the covariates for random effects estimation within each trial can yield correct estimation of random effects. Moreover, under the correct specification for random effects estimation