Toward Valid Generative Clinical Trial Data with Survival Endpoints

2025-11-20

AI Summary

Clinical trials—especially in oncology and rare late-stage diseases—face challenges including patient dispersion, slow enrollment, and high costs. A key enabler is generating high-fidelity synthetic control arms incorporating censored survival outcomes; however, existing GAN-based approaches rely on strong independent censoring assumptions, require large datasets, and suffer from training instability. This paper proposes a unified latent-variable framework based on variational autoencoders (VAEs) that jointly models mixed covariates and time-to-event outcomes without assuming independent censoring. It further introduces a post-generation selection strategy to enhance statistical calibration. Evaluated on both real and synthetic data under privacy-preserving constraints, the method significantly outperforms GAN baselines, achieving superior balance among fidelity, practical utility, and statistical validity. It effectively supports control-arm augmentation and cross-institutional data sharing, while systematically correcting type-I error inflation and power distortion.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Game Theory and Economic Paradigms: Adversarial Learning

Application Category

• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• User Modeling, Personalization and Recommendation: Accountability, Transparency, and Ethics for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

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.

Problem

• Generating synthetic clinical trial data with realistic survival endpoints
• Overcoming limitations of GAN-based methods for small censored datasets
• Addressing data sharing privacy and treatment group imbalance issues

Innovation

• VAE generates covariates and survival outcomes jointly
• Unified latent framework without independent censoring assumption
• Outperforms GANs on fidelity, utility, and privacy metrics