Multilevel network meta-regression: methods and implementation: in workshop Time to implement multilevel network meta-regression rather than matching adjusted indirect comparisons

PURPOSE: Multi-level network meta-regression (ML-NMR) extends the standard network meta-analysis framework to leverage individual patient data and aggregate data when comparing multiple treatments while adjusting for differences in populations between trials. Unlike previous population adjustment approaches, ML-NMR is applicable in networks of any size, avoids aggregation bias and issues with non-collapsible effect measures, and crucially for decision-making produces estimates in any target population.

DESCRIPTION: Workshop attendees will obtain a working knowledge of the ML-NMR method, its advantages, and considerations for implementation. Dr. Jansen will chair the session and introduce ML-NMR in the context of the challenges with existing methods (10 min.). Dr. Phillippo will explain the statistical methods for ML-NMR, highlight advantages relative to existing methods, and provide an overview of how to implement the method using the multinma R package in terms of the syntax and features (15 min.). Ms. Cope will illustrate how these methods can be applied in a case study regarding the comparative efficacy of alternative interventions for triple-class exposed relapsed refractory multiple myeloma. This will include audience participation regarding selection of covariates, alternative time-to-event models, conditional vs. marginal estimates, and target populations for prediction (15 min). Mr. Klijn will describe lessons learned and recommendations for implementation of ML-NMR (15 min). Questions from the audience will be addressed (5 min) and this interactive workshop will be valuable to researchers and industry analysts interested in comparative efficacy research for health technology assessments.