Consistency Assessment of Regional Treatment Effect for Multi-Regional Clinical Trials in the Presence of Covariate Shift

Multi-Regional Clinical Trials (MRCTs) play a central role in the development of new therapies by enabling the simultaneous evaluation of drug efficacy and safety across diverse global populations. Assessing the consistency of treatment effects across regions is a fundamental aspect of MRCTs. Existing methods typically focus on region-specific marginal treatment effects. However, when treatment effect heterogeneity arises due to effect-modifying baseline covariates, distributional differences in these covariates can lead to erroneous conclusions. In this paper, we explicitly account for this phenomenon in the consistency assessment by considering the conditional average treatment effect. We propose a two-step assessment strategy that complements existing methods and mitigates the impact of treatment effect heterogeneity. Results from numerical studies demonstrate the effectiveness of the proposed approach.

💡 Research Summary

Multi‑Regional Clinical Trials (MRCTs) are increasingly central to drug development, allowing simultaneous assessment of efficacy and safety across diverse populations. Regulatory guidance (ICH‑E17, FDA, Japan’s MHLW) stresses not only overall trial success but also the consistency of treatment effects across regions. Existing consistency assessment methods, typified by the MHLW‑type procedures, rely on marginal average treatment effects (ATEs) within each region and compare them to the overall ATE or to the ATE of the complementary regions using a pre‑specified fraction q. While useful for exploratory purposes, these methods ignore the distributional differences of baseline covariates that modify treatment response. When such effect‑modifying covariates exhibit “covariate shift” across regions, marginal ATEs can diverge even if the underlying conditional average treatment effect (CATE) is identical, leading to false conclusions about consistency.

The authors address this gap by proposing a two‑step consistency assessment framework that explicitly incorporates CATEs. First, they retain the conventional one‑step MHLW‑type test as a screening tool. If the region passes this test, consistency is declared and the procedure stops. If the region fails, the second step investigates the source of inconsistency. They compare the region‑specific CATE, Δ_r(X), with the CATE of the complementary regions, Δ_{‑r}(X). To assess similarity, they estimate individual treatment effects (ITEs) using the Leave‑One‑Out Potential outcomes (LOOP) estimator, which leverages information from all other participants to obtain low‑variance, model‑robust ITE estimates. These ITEs are then entered into a regression model that includes main effects of covariates, a region indicator, and region‑by‑covariate interaction terms. A Wald or likelihood‑ratio test on the interaction coefficients (β_{RX}) determines whether Δ_r(X) differs from Δ_{‑r}(X). A significant interaction leads to a final declaration of inconsistency.

If the interaction test is non‑significant, the authors attribute the observed ATE discrepancy primarily to covariate distribution differences. They then construct a covariate‑shift‑adjusted ATE for the region, δ*r = E{F_r}