OBJECTIVES: This work aims to provide a presentation of the recently suggested Multi-level Network Meta-Regression (ML-NMR) method in layman’s terms and to illustrate its benefits based on a simulated example in presence of heterogeneity.

METHODS: The ML-NMR methodology, introduced by Phillippo in 2019, is presented as part of the HTA Coordination Group Guidelines for Quantitative Evidence Synthesis. This method, which is an extension of regression-based approaches leveraging individual patient data, aims at providing a more methodologically robust population-adjusted indirect comparison alternative to estimate relative treatment effects. A simulated dataset was generated to illustrate the concepts and benefits of the ML-NMR, considering three randomized clinical trials, comparing respectively the drugs B, C and D to A. The endpoint of interest was a time-to-event outcome, and three covariates were considered for each trial. Two scenarios were analyzed: one with and one without heterogeneity in these covariates.

RESULTS: The ML-NMR allows for controlling for observed differences in treatment effect modifiers across trials through a regression approach, while maintaining the comparison at a network level. The adjustment for heterogeneity through the ML-NMR is illustrated through the HR [95% CrI] from the simulated dataset when compared to a standard NMA. HRs were closer to the true values with the ML-NMR.

CONCLUSIONS: This study offers a simple overview of the ML-NMR, advocating its integration in future Health Technology Assessment submissions and enhancing understandability for a non-statistical audience. ML-NMR enables population-adjusted indirect treatment comparisons according to multiple factors at the network level, a significant advancement over previous methods that permitted population adjustments solely in a pairwise context (Matched Adjusted Indirect Comparison, Simulated Treatment Comparison) or according to a limited set of factors (network meta-regression).