Abstract
BACKGROUND: Non-significant randomized controlled trials can hide subgroups of good responders to experimental drugs, thus hindering subsequent development. Identifying such heterogeneous treatment effects is key for precision medicine and many post-hoc analysis methods have been developed for that purpose. While several benchmarks have been carried out to identify the strengths and weaknesses of these methods, notably for binary and continuous endpoints, similar systematic empirical evaluation of subgroup analysis for time-to-event endpoints is lacking. METHODS: This work aims to fill this gap by evaluating several subgroup analysis algorithms in the context of time-to-event outcomes, by means of three different research questions: Is there heterogeneity? What are the biomarkers responsible for such heterogeneity? Who are the good responders to treatment? In this context, we propose a new synthetic and semi-synthetic data generation process that allows one to explore a wide range of heterogeneity scenarios with precise control on the level of heterogeneity. RESULTS: Methods are overall comparable when it comes to detecting heterogeneity, but interaction test-based methods demonstrate better statistical power in harder-to-detect heterogeneity settings. Cox-based multivariate and interaction test-based methods are best at identifying variables predictive of heterogeneity. Methods able to estimate the Conditional Average Treatment Effect (CATE), such as S-learners based on Cox- and tree-based multivariate algorithms, are well suited to identify subgroup of good responders, with Cox-multivariate method being especially effective in low and intermediate heterogeneity settings, but outperformed in higher heterogeneity settings. CONCLUSIONS: Not all methods are suited for all heterogeneity investigations. We recommend employing interaction test-based methods to detect heterogeneity and identify predictive variables, and leveraging multivariate CATE estimation-based approaches for subgroup identification. Moreover, we recommend adopting a two-step strategy consisting of (i) establishing the existence of heterogeneity and finding responsible covariates using interpretable methods and (ii) looking for subgroups using more complex methods.