Population-adjusted treatment comparisons: estimates based on MAIC (Matching-Adjusted Indirect Comparisons) and STC (Simulated Treatment Comparisons)

OBJECTIVES: To review the properties and assumptions of methods for population-adjusted treatment comparison, including Matching-Adjusted Indirect Comparison (MAIC) and Simulated Treatment Comparison (STC), and to provide guidance on their use in health technology appraisal.

METHODS: Standard methods for indirect comparisons and network meta-analysis are based on aggregate data, with the key assumption that there is no difference between trials in the distribution of effect-modifying variables. Two methods which relax this assumption, MAIC and STC, are becoming increasingly common in industry-sponsored treatment comparisons, where a company has access to individual patient data (IPD) from its own trials but only aggregate information from competitor trials. Both methods use IPD to adjust for between-trial differences in covariate distributions. We review the properties of these methods in light of the wider literature on standardisation and calibration based on propensity score reweighting and covariate adjustment, which are the foundation for MAIC and STC respectively, and identify the key assumptions in the context of indirect comparisons.

RESULTS: There is a lack of clarity about how and when the methods should be applied in practice, and both MAIC and STC as currently applied can only produce population-adjusted estimates that are valid for the populations in the competitor trials, rather than the target population for the decision. In addition, the fundamental distinction between “anchored” and “unanchored” forms of indirect comparison – where a common comparator arm is or is not utilised to control for between-trial differences in prognostic variables – is under-emphasised, with the unanchored comparison making assumptions that are infeasibly strong.

CONCLUSIONS: We provide recommendations on how and when population adjustment methods of this type should be used in order to provide statistically valid, clinically meaningful, transparent and consistent results for any given target population, and set out the additional analyses that should be presented to support their use.