Presentation (CTI)

OBJECTIVES: To evaluate the impact of sample size and baseline covariate imbalance on MAIC and STC performance in estimating objective response rate (ORR) treatment effects under mild (SMD≈0.15) and moderate (SMD≈0.30) imbalance across IPD sizes of 50, 150, and 250 patients.
METHODS: Trial A (IPD) and Trial B (aggregate data) were simulated with three baseline covariates (age, sex, ECOG) and a binary ORR outcome. Outcomes were generated from a logistic model calibrated to ORRs of ~10% in Trial A and ~15% in Trial B. Imbalance was induced to achieve prespecified SMDs. Each sample size was evaluated over 2,000 replicates. MAIC reweighted Trial A to match comparator covariates, while STC fitted a logistic model and transported predictions. Treatment effects were summarized as risk ratios (RR), with performance assessed by bias, RMSE, and 95% CI coverage.
RESULTS: MAIC and STC produced broadly comparable estimates, with performance influenced by sample size and imbalance. At N=50, weighted MAIC and STC more often detected treatment effects; however, STC showed instability under moderate imbalance, whereas MAIC produced finite but sometimes extreme estimates. At N=150, both methods consistently detected effects, with STC generally estimating smaller effects. At N=250, estimates converged across methods and imbalance levels. Greater imbalance caused larger effective sample size loss under MAIC.
CONCLUSIONS: MAIC and STC yielded similar estimates, with reliability depending on sample size and imbalance. MAIC provided more stable estimates in small samples with moderate imbalance, while differences between methods decreased with larger sample sizes.