Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Selective Information Borrowing for Region-Specific Treatment Effect Inference under Covariate Mismatch in Multi-Regional Clinical Trials

Chenxi Li, Ke Zhu, Shu Yang, Xiaofei Wang

TL;DR

This work tackles RSATE estimation in multi-regional clinical trials when the target region is small and covariate mismatch and potential outcome drift complicate information borrowing. It develops a three-part framework: (i) an inverse-variance weighted (IVW) estimator that combines a target-region, covariate-rich estimator with a full-borrowing doubly robust estimator, (ii) conformal selective borrowing (CSB) that uses individual-level conformal p-values to adaptively select compatible auxiliary data under outcome drift, and (iii) a Fisher randomization test for finite-sample, model-free type I error control that accommodates selection uncertainty. Simulation studies show CSB-IVW yields 10–50% reductions in mean squared error and higher power than no-borrowing or full-borrowing approaches, while maintaining valid inference; an empirical POWER trial application demonstrates improved RSATE precision and narrower confidence intervals. Collectively, the framework provides robust, estimand-based inference for region-specific effects in MRCTs, with practical relevance for local regulatory decision-making and potential extensions to time-to-event outcomes and augmented data settings.

Abstract

Multi-regional clinical trials (MRCTs) are central to global drug development, enabling evaluation of treatment effects across diverse populations. A key challenge is valid and efficient inference for a region-specific estimand when the target region is small and differs from auxiliary regions in baseline covariates or unmeasured factors. We adopt an estimand-based framework and focus on the region-specific average treatment effect (RSATE) in a prespecified target region, which is directly relevant to local regulatory decision-making. Cross-region differences can induce covariate shift, covariate mismatch, and outcome drift, potentially biasing information borrowing and invalidating RSATE inference. To address these issues, we develop a unified causal inference framework with selective information borrowing. First, we introduce an inverse-variance weighting estimator that combines a "small-sample, rich-covariate" target-only estimator with a "large-sample, limited-covariate" full-borrowing doubly robust estimator, maximizing efficiency under no outcome drift. Second, to accommodate outcome drift, we apply conformal prediction to assess patient-level comparability and adaptively select auxiliary-region patients for borrowing. Third, to ensure rigorous finite-sample inference, we employ a conditional randomization test with exact, model-free, selection-aware type I error control. Simulation studies show the proposed estimator improves efficiency, yielding 10-50% reductions in mean squared error and higher power relative to no-borrowing and full-borrowing approaches, while maintaining valid inference across diverse scenarios. An application to the POWER trial further demonstrates improved precision for RSATE estimation.

Selective Information Borrowing for Region-Specific Treatment Effect Inference under Covariate Mismatch in Multi-Regional Clinical Trials

TL;DR

This work tackles RSATE estimation in multi-regional clinical trials when the target region is small and covariate mismatch and potential outcome drift complicate information borrowing. It develops a three-part framework: (i) an inverse-variance weighted (IVW) estimator that combines a target-region, covariate-rich estimator with a full-borrowing doubly robust estimator, (ii) conformal selective borrowing (CSB) that uses individual-level conformal p-values to adaptively select compatible auxiliary data under outcome drift, and (iii) a Fisher randomization test for finite-sample, model-free type I error control that accommodates selection uncertainty. Simulation studies show CSB-IVW yields 10–50% reductions in mean squared error and higher power than no-borrowing or full-borrowing approaches, while maintaining valid inference; an empirical POWER trial application demonstrates improved RSATE precision and narrower confidence intervals. Collectively, the framework provides robust, estimand-based inference for region-specific effects in MRCTs, with practical relevance for local regulatory decision-making and potential extensions to time-to-event outcomes and augmented data settings.

Abstract

Multi-regional clinical trials (MRCTs) are central to global drug development, enabling evaluation of treatment effects across diverse populations. A key challenge is valid and efficient inference for a region-specific estimand when the target region is small and differs from auxiliary regions in baseline covariates or unmeasured factors. We adopt an estimand-based framework and focus on the region-specific average treatment effect (RSATE) in a prespecified target region, which is directly relevant to local regulatory decision-making. Cross-region differences can induce covariate shift, covariate mismatch, and outcome drift, potentially biasing information borrowing and invalidating RSATE inference. To address these issues, we develop a unified causal inference framework with selective information borrowing. First, we introduce an inverse-variance weighting estimator that combines a "small-sample, rich-covariate" target-only estimator with a "large-sample, limited-covariate" full-borrowing doubly robust estimator, maximizing efficiency under no outcome drift. Second, to accommodate outcome drift, we apply conformal prediction to assess patient-level comparability and adaptively select auxiliary-region patients for borrowing. Third, to ensure rigorous finite-sample inference, we employ a conditional randomization test with exact, model-free, selection-aware type I error control. Simulation studies show the proposed estimator improves efficiency, yielding 10-50% reductions in mean squared error and higher power relative to no-borrowing and full-borrowing approaches, while maintaining valid inference across diverse scenarios. An application to the POWER trial further demonstrates improved precision for RSATE estimation.
Paper Structure (17 sections, 1 theorem, 8 equations, 5 figures, 1 table)

This paper contains 17 sections, 1 theorem, 8 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Preliminary
  3. Target region, estimand, and data structure
  4. Target-region-only identification and estimation
  5. Auxiliary-region-integrated identification and estimation
  6. Information borrowing under covariate mismatch
  7. Conformal selective borrowing under outcome drift
  8. Conformal p-value
  9. MSE-guided selection threshold
  10. Fisher randomization test for type I error rate control
  11. Simulation studies
  12. Simulation design
  13. Simulation results
  14. Real data illustration
  15. Data preparation
  16. ...and 2 more sections

Key Result

theorem 1

Under the null hypothesis $H_0$, for any $\alpha\in(0,1)$, $\mathbb{P}_{\boldsymbol{A}}\!\left(p^{\text{FRT}}\le \alpha \mid \boldsymbol{A}_{\mathcal{E}}\right)\le \alpha,$ where $\mathbb{P}_{\boldsymbol{A}}$ denotes probability with respect to the randomization distribution of the assignment vector

Figures (5)

  • Figure 1: Data structure of a multi-regional randomized clinical trial with target and auxiliary regions.
  • Figure 2: MSE% vs log(Precision) across different signal of $X$ with correlated $X$ and $U$.
  • Figure 3: Type I error rates of FRTs under varying magnitudes of hidden bias and statistical power under varying true treatment effects and hidden bias.
  • Figure 4: Outcome (LBM on Day 84) VS. Sampling score distribution
  • Figure 5: Comparison of estimated treatment effects and p-values across borrowing strategies. The upper (lower) numbers report asymptotic (FRT) p-values.

Theorems & Definitions (6)

  • remark 1: Conformal score
  • remark 2: Split conformal p-value
  • remark 3: Test statistic
  • remark 4: Monte Carlo approximation
  • remark 5: One-sided hypothesis and randomization-based interval
  • theorem 1