Improved Inference for Heterogeneous Treatment Effects Using Real-World Data Subject to Hidden Confounding
share
The heterogeneity of treatment effect (HTE) lies at the heart of precision medicine. Randomized clinical trials (RCTs) are gold-standard to estimate the HTE but are typically underpowered. While real-world data (RWD) have large predictive power but are often confounded due to lack of randomization of treatment. In this article, we show that the RWD, even subject to hidden confounding, may be used to empower RCTs in estimating the HTE using the confounding function. The confounding function summarizes the impact of unmeasured confounders on the difference in the potential outcome between the treated and untreated groups, accounting for the observed covariates, which is unidentifiable based only on the RWD. Coupling the RCT and RWD, we show that the HTE and confounding function are identifiable. We then derive the semiparametric efficient scores and integrative estimators of the HTE and confounding function. We clarify the conditions under which the integrative estimator of the HTE is strictly more efficient than the RCT estimator. As a by-product, our framework can be used to generalize the average treatment effects from the RCT to a target population without requiring an overlap covariate distribution assumption between the RCT and RWD. We illustrate the integrative estimators with a simulation study and an application.
READ FULL TEXT
