Digital Twins as Synthetic Controls in Single-Arm Trials

2026-05-14 · Source: stat.ML updates on arXiv.org · Field: Science & Research — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences, Health & Medical Research · Depth: Expert, extended

Summary

This work proposes digital twins as synthetic controls in single-arm clinical trials to evaluate drug efficacy and safety, offering an efficient, ethical, and practical alternative to randomized controlled trials (RCTs). Digital twins are personalized predictions of disease progression generated from machine learning models trained on historical datasets. The paper reviews doubly robust estimators, specifically Augmented Inverse Probability Weighting (AIPW), which combine outcome models with propensity-based adjustments to debias predictions and improve robustness. It presents power and sample size formulas for these estimators and discusses trade-offs in selecting historical data for training and analysis. The authors also outline practical considerations for deploying digital twins within the framework of recent FDA draft guidance on artificial intelligence in drug development. Case studies reanalyzing data from trials in amyotrophic lateral sclerosis (ALS) and Huntington's disease (HD) demonstrate the proposed methods, showing that model-based approaches, particularly AIPW, offer comparable or smaller bias and substantial variance reduction (40-60%) relative to one-to-one propensity score matching (PSM).

Key takeaway

For AI Scientists and Research Scientists designing single-arm clinical trials, integrating digital twin models with Augmented Inverse Probability Weighting (AIPW) offers a statistically robust and efficient method for estimating treatment effects. You should prioritize rich, diverse training data for digital twins and pre-specify all model and analysis components to align with FDA guidance, ensuring both scientific rigor and regulatory credibility. This approach can significantly reduce required sample sizes and improve the reliability of efficacy conclusions.

Key insights

Digital twins enhance single-arm trials by providing robust, efficient synthetic controls via machine learning and doubly robust estimators.

Principles

• Doubly robust estimators improve reliability by requiring only one of two models to be correctly specified.
• Outcome models leverage heterogeneous data more effectively than propensity-based methods.
• Adding any external control data, even marginally relevant, increases effective sample size.

Method

Construct synthetic control arms using machine learning-based digital twins to predict counterfactual outcomes, then apply doubly robust estimators like AIPW for treatment effect estimation, incorporating power and sample size calculations.

In practice

• Use AIPW for increased efficiency and robustness in single-arm trial analysis.
• Evaluate digital twin models against observed outcomes using standard predictive metrics.
• Pre-specify models, data handling, and analysis plans to align with FDA guidance.

Topics

• Digital Twins
• Single-Arm Trials
• Synthetic Control Arms
• AIPW Estimator
• Machine Learning in Drug Development

Best for: AI Scientist, Research Scientist, Data Scientist

Related on AIssential

• Treatment Response Optimized Clinical Decision Support AI System via Digital Twin Simulation — An online adaptive AI framework integrates TE, DT, and RL for safe, continuously improving clinical decision support.
• From virtual experiments to biomedical insight with synthetic data — Bridging the "simulation to reality" gap is crucial for synthetic data to realize its full potential in biomedical AI.
• From Baseline to Follow-Up: Counterfactual Spine DXA Image Synthesis in UK Biobank Using a Causal Hierarchical Variational Autoencoder — The CHVAE model synthesizes anatomically plausible, causally consistent spine DXA images by intervening on age, addressing anatomical variation challenges.
• Learning Cardiac Electrophysiology Digital Twins Through Agentic Discovery of Hybrid Structure — An LLM agent discovers stable, personalized cardiac electrophysiology digital twin models using structured domain knowledge.
• Transition-Based Digital Twin Modelling for Alzheimer's Disease under Sparse Longitudinal Data — Local transition modeling offers a data-efficient and robust strategy for personalized disease forecasting in neurodegenerative disorders.
• GeoVision-Enabled Digital Twin for Hybrid Autonomous-Teleoperated Medical Responses — A GeoVision-enabled Digital Twin improves remote medical response by integrating real-time geospatial, clinical, and operational data.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by stat.ML updates on arXiv.org.