When conducting a Network Meta-Analysis (NMA) for a Health Technology Assessment (HTA), the submitting company typically will have access to Individual Patient Data (IPD) from their own trials, but only aggregate data (AgD) for the comparator. In this case, they can re-weight the IPD so that the covariate characteristics in the IPD trials match that of the AgD trials, using the increasingly popular method of Matching-Adjusted Indirect Comparison (MAIC).

We carried out a simulation study to investigate this method in a Bayesian setting. We simulated three IPD trials comparing treatments A and B (AB-IPD trials), and one aggregate data trial comparing treatments B and C (BC-AgD trial). We investigated two options of weighting covariates: 1. all three studies are weighted separately to match the BC-AgD trial (MAIC Separate Trials). 2. patients are weighted across all three IPD studies to match the BC-AgD trial, but the NMA still considers each trial separately (MAIC Pooled Trials). We compared the results of the MAIC to a standard NMA and a mixed IPD/AgD NMA. We applied these methods to a network of treatments for multiple myeloma.

MAIC can provide more accurate estimates of the relative treatment effects than a standard NMA in the BC-AgD trial population. However, MAIC may decrease the accuracy of the relative treatment effects in the overall population. Treatment rankings were unchanged when applying MAIC to the multiple myeloma network.

MAIC is beneficial as a sensitivity analysis to demonstrate that results hold across patient populations. If there is a difference in relative treatment effects attributable to population imbalances, then it is useful to be able to quantify this difference. However, we recommend using either a standard NMA or a mixed IPD/AgD NMA for the base case analysis, given the potential bias that can arise in an MAIC.