Integrating Meta-Analysis Into a Specific Study (InMASS) for Estimating the Target-Population Average Treatment Effect.

Randomized controlled trials remain the benchmark for estimating treatment effects, yet practical constraints often restrict sample sizes in a prespecified target population. In such settings, evidence from completed trials is frequently available only in aggregate form, which precludes direct individual-level integration. We propose InMASS, an inferential framework for estimating the target-population average treatment effect (TATE) for a prespecified target population by integrating aggregate evidence from multiple external trials. InMASS reconstructs pseudo individual-level information from routinely reported summary statistics via a meta-regression model and transports this information to the target population using density-ratio weighting. Under a weak covariate shift assumption, we show that reconstructing second-order moments is sufficient for consistent and asymptotically normal estimation of the target-population TATE, irrespective of the covariate distribution in external trials. Importantly, efficiency gains depend on the total amount of external information rather than on increasing the size of the target trial. Simulation studies and a real-data application demonstrate that InMASS materially improves precision and increases statistical power relative to analyses based solely on the target trial, including settings with unbalanced allocation or single-arm designs. These findings underscore the practical utility of InMASS for target-specific inference when individual-level data are limited.