A scoping review of artificial intelligence applications in clinical trial risk assessment
Problem
Assessing safety, efficacy, and operational risks in clinical trials is difficult; AI is increasingly applied to improve it.
Why it matters
Trial failures and safety risks impose major costs and patient risks in drug development.
What the AI did
A scoping review of 142 studies (2013–2024) analyzing AI techniques for adverse drug event prediction, treatment effect estimation, and phase transition prediction.
Results
Some models achieve AUROC up to 96%; 142 studies reviewed across safety (55), efficacy (46), operational (45) risks.
Impact
Documents high-performing AI models but highlights selection bias, limited prospective studies, and data quality issues.
Record
| Venue | npj Digital Medicine |
|---|---|
| Published | 2025 |
| Work type | article |
| Authors | Douglas Teodoro, Nona Naderi, Anthony Yazdani, Boya Zhang, Alban Bornet |
| Citations | 46 (OpenAlex, at import time) |
| References | 191 |
| Model or method | Multiple reviewed: machine learning, graph neural networks, transformers, causal ML, LLMs |
| Organization | not stated in source |
| Primary source | https://doi.org/10.1038/s41746-025-01886-7 |
| Open access | gold |
| Keywords | clinical trials, risk assessment, machine learning, scoping review, AUROC |
Importance rating
significant — thousands to millions affected
Clinical trial risk assessment affects drug development safety, efficiency, and patient risk; 142 studies show a substantial active field, but the abstract does not quantify the problem's human/economic scale.