Toward Valid Generative Clinical Trial Data with Survival Endpoints

TL;DR

The paper tackles the challenge of generating synthetic clinical trial data with time-to-event endpoints under censoring. It introduces HI-VAE, a variational autoencoder that jointly models mixed-type covariates and survival times within a unified latent space, without assuming independent censoring, and optimizes an ELBO-based objective to learn the joint distribution. A calibration-focused evaluation framework assesses fidelity, utility, privacy, and downstream type I error and power, revealing miscalibration in naive generations and demonstrating that a post-generation selection procedure can partially restore statistical validity. Empirical results across simulated and four real phase III datasets show HI-VAE outperforms survival GAN/VAEs on classical metrics yet highlights remaining challenges for safe, regulatory-grade use, particularly in achieving robust calibration and stronger privacy guarantees for public data sharing.

Abstract

Clinical trials face mounting challenges: fragmented patient populations, slow enrollment, and unsustainable costs, particularly for late phase trials in oncology and rare diseases. While external control arms built from real-world data have been explored, a promising alternative is the generation of synthetic control arms using generative AI. A central challenge is the generation of time-to-event outcomes, which constitute primary endpoints in oncology and rare disease trials, but are difficult to model under censoring and small sample sizes. Existing generative approaches, largely GAN-based, are data-hungry, unstable, and rely on strong assumptions such as independent censoring. We introduce a variational autoencoder (VAE) that jointly generates mixed-type covariates and survival outcomes within a unified latent variable framework, without assuming independent censoring. Across synthetic and real trial datasets, we evaluate our model in two realistic scenarios: (i) data sharing under privacy constraints, where synthetic controls substitute for original data, and (ii) control-arm augmentation, where synthetic patients mitigate imbalances between treated and control groups. Our method outperforms GAN baselines on fidelity, utility, and privacy metrics, while revealing systematic miscalibration of type I error and power. We propose a post-generation selection procedure that improves calibration, highlighting both progress and open challenges for generative survival modeling.

Figures (23)

• Figure 1: Performance comparison on simulated and real datasets, using J-S distance, survival curve distance, and $K$-map score. Arrows indicate directions of better performance.
• Figure 2: Type I error and power estimation for independent case (top) and dependent case (bottom). Dashed lines: empirical power. Green: theoretical power with reduced control size. Blue: theoretical power with generated control size.
• Figure 3: Proportion of Monte Carlo replications with at least one generated dataset not rejected by the adjusted log-rank test (at the 5% level) against the original controls.
• Figure 4: Type I error and power estimation after post-generation selection for independent case (top) and dependent case (bottom). Dashed lines: empirical power. Green: theoretical power with reduced control size. Blue: theoretical power with generated control size.
• Figure 5: Type I error and power estimation after post-generation selection for independent case under two training strategies: using only available control samples (top) and using both control and treated arms (bottom). Dashed lines: empirical power. Green: theoretical power with reduced control size. Blue: theoretical power with generated control size.