What can artificial intelligence do for randomized controlled trials?: A scoping review
Ling Wang, Xufei Luo, Jie Zhang, Meihua Wu, Xu Wang, Junxian Zhao, Janne Estill, Ruitai ShaoBackground:
Randomized controlled trials (RCTs) generate reliable evidence for evaluating healthcare interventions. However, their design, conduct, and evidence synthesis remain resource-intensive and methodologically challenging. Recent advances in artificial intelligence (AI) have created new opportunities to enhance the efficiency, scalability, and quality of RCT-related processes.
Objective:
To map and summarize the applications of AI across different stages of RCTs.
Methods:
A scoping review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews. MEDLINE, Embase, Web of Science, PsycINFO, trial registries, preprint servers, Google Scholar, and reference lists were searched through January 2, 2025. Studies using AI to support, optimize, or accelerate RCT-related processes were included.
Results:
A total of 59 studies were included, of which 26 studies used AI to support the conduct of new RCTs, and 33 studies to support the review of existing RCTs. In studies related to the conduct of new RCTs, AI was primarily applied to participant recruitment and allocation (n = 12) and data analysis (n = 6). For studies related to the review of existing RCTs, information extraction was the most common application (n = 18), followed by literature search and screening (n = 6) and risk-of-bias (RoB) assessment (n = 4). Applications to support standardized reporting were limited (n = 3), and trial design, sample size estimation, quality control, and semantic analysis were each addressed in only 2 studies.
Conclusions:
AI applications in RCTs are concentrated in data-intensive and relatively standardized tasks, particularly participant recruitment, data analysis, and evidence extraction. The use of AI in core methodological decision-making remains limited. Future research should prioritize rigorous validation, transparency, interpretability, and responsible integration of AI into RCT workflows.