OBJECTIVES: As suggested by the National Institute for Health and Care Excellence Decision Support Unit (DSU) 18, two methods for population-adjusted indirect treatment comparisons (PAIC) are considered standard when individual patient data are available from at least one trial; matching adjusted indirect treatment comparison (MAIC) and simulated treatment comparison (STC). Recent studies evaluated the performances of these methods or proposed alternative novel PAIC methods. Our study aims to provide a full picture of published PAIC methods and summarize their advantages and disadvantages.

METHODS: A targeted literature review following PICOS eligibility criteria was conducted. Peer-reviewed publications in English that were published after January 2010 (i.e., the year of the first key MAIC publication) which compared or introduced novel PAIC methods were included. Three different databases (i.e., Medline, Cochrane database of systematic reviews, and Cochrane methodology register via Ovid) were searched.

RESULTS: A total of twenty-four studies met the eligibility criteria (i.e., fifteen were identified through the database search and nine through cross-referencing). Fourteen studies assessed the performance of standard PAIC methodologies, four studies explored the potential extensions of the standard MAIC approach, and six studies included novel population adjustment methodologies. The advantages of MAIC/STC over traditional unadjusted indirect treatment comparison (ITC) methods (i.e., Bucher) in the presence of effect modifiers are well confirmed in the literature; however, these approaches also have limitations. Attempts to extend the PAIC approach are mainly related to weight schemes, e.g., weights based on entropy balancing or via a polynomial-based non-linear optimization. Recently proposed novel PAIC methods include multilevel network meta-regression and parametric G-computation.

CONCLUSIONS: The performance of different PAIC methods is still controversial and not clear. Future research should focus on guidelines on PAIC methods selection depending on the target estimand, presence of effect modifiers, covariate overlap, and the size of the treatment network.