OBJECTIVES: Matching adjusted indirect comparison (MAIC) is a recommended method for population-adjusted indirect comparisons, e.g. indirectly comparing two active treatments (B vs C) via A. However, MAIC estimate (B vs C) depends on the chosen target population (i.e. the aggregated population from AC trial). Potentially, the aggregated population does not reflect the real-world population (RWP) relevant to HTA decision making. The NICE guidance shows MAIC result can be transported into RWP if shared effect modifier (SEM) is assumed. This assumption means both the effect modifiers of the treatments in comparison and the change in treatment effect caused by each effect modifier are the same. As this assumption is possibly unrealistic, the objective of our study is to explore how to calibrate MAIC estimate to RWP without making this assumption.

METHODS: Fictitious datasets were simulated. Subgroup analyses from aggregated trial were used to test the SEM assumption. If the assumption is violated, an additional calibration factor was added to the MAIC estimate calibrating it to RWP. Treatment effects were measured as odds ratios.

RESULTS: When the SEM assumption hold, the MAIC estimate (0.486, [95%CI: 0.404, 0.585]) was close to the estimated result in the RWP (0.485, [95%CI: 0.392, 0.600]). When the assumption was violated, the MAIC result (0.927, [95%CI: 0.782, 1.099]) was not comparable to the RWP estimate (0.690, [95%CI: 0.568, 0.839]), while the calibrated result (0.685, [95%CI: 0.567, 0.833]) was. Additional scenario analyses lead to similar findings.

CONCLUSIONS: Before transporting MAIC result to RWP, testing the SEM assumption based on the available subgroup information is needed. When the assumption is violated, the proposed method helps to calibrate MAIC result to RWP. However, this method is limited by the availability of appropriate subgroup data. Calibrating for multiple effect modifiers or potential effect modifiers with no subgroup data reported are problematic.