BACKGROUND: Clinical trials for novel therapeutics are expensive, time-consuming, and have an end-to-end success rate of under 8%. Several applications of artificial intelligence and machine learning (AI/ML) to trial design and conduct have the potential to increase trial efficiency, and the use of real world data (RWD) is often critical to the development of these novel applications.

OBJECTIVES: The objective of this study was to comprehensively review applications of AI/ML that used RWD to improve clinical trial design and conduct.

METHODS: We conducted a review of published work on AI/ML-based technologies for clinical trials, focusing on AI/ML applications trained using RWD. The review provides a survey of the field, with concrete examples of applications and explanations of the machine learning task. Additionally, we describe open challenges and opportunities for the use of AI/ML based on real world data to improve clinical trials.

RESULTS: RWD is essential to training AI/ML algorithms used for clinical trial optimization; over 50 published examples were found by the review. We identified four overarching categories of applications: (1) improved participant recruitment; (2) enhanced trial management; (3) enabled remote patient monitoring (e.g., signal processing from wearables); and (4) in silico trials. Lack of freely available RWD, limited model generalizability, privacy considerations, potential fairness issues, lack of familiarity with AI/ML methods, and limited prospective evaluations of novel approaches pose challenges for the expanded use of AI/ML in clinical trial optimization. Opportunities for advancing applications include transfer learning to accommodate limited domain-specific data, increased use of multi-modal data to train models, and applications of federated learning to develop AI/ML models when data is distributed.

CONCLUSIONS: AI/ML technologies based on RWE have the potential to reduce costs and optimize clinical trials, and hold promise to continue to improve study efficiency in a way that directly impacts the patient experience.