The article presents a generalized linear modeling (GLM) framework for synthesizing data from randomized controlled trials, particularly for use in probabilistic decision-making. The framework is designed to handle various outcome types and data transformations, including normal, binomial, Poisson, and multinomial likelihoods, with appropriate link functions such as identity, logit, log, complementary log-log, and probit. The core models can be applied to pairwise meta-analysis, indirect comparisons, multiarm trials, and network meta-analysis (NMA) without distinction. The authors use a Bayesian Markov Chain Monte Carlo (MCMC) approach with WinBUGS software to estimate parameters and assess model fit. Key aspects of the framework include the use of exchangeability assumptions, the development of models for different data types, and the extension to NMA. The article provides detailed examples and illustrations to demonstrate the application of the framework in various scenarios, such as binomial data with logit link, rate data with Poisson and binomial likelihoods, competing risks with multinomial likelihood, continuous data with normal likelihood, and ordered categorical data with probit link. The framework is modular and flexible, allowing for the synthesis of different types of evidence in a unified manner.The article presents a generalized linear modeling (GLM) framework for synthesizing data from randomized controlled trials, particularly for use in probabilistic decision-making. The framework is designed to handle various outcome types and data transformations, including normal, binomial, Poisson, and multinomial likelihoods, with appropriate link functions such as identity, logit, log, complementary log-log, and probit. The core models can be applied to pairwise meta-analysis, indirect comparisons, multiarm trials, and network meta-analysis (NMA) without distinction. The authors use a Bayesian Markov Chain Monte Carlo (MCMC) approach with WinBUGS software to estimate parameters and assess model fit. Key aspects of the framework include the use of exchangeability assumptions, the development of models for different data types, and the extension to NMA. The article provides detailed examples and illustrations to demonstrate the application of the framework in various scenarios, such as binomial data with logit link, rate data with Poisson and binomial likelihoods, competing risks with multinomial likelihood, continuous data with normal likelihood, and ordered categorical data with probit link. The framework is modular and flexible, allowing for the synthesis of different types of evidence in a unified manner.