Artificial intelligence in clinical trials-state of the evidence, gaps, and next steps
Papers on AI in trials, patient-facing generative AI and a geriatrician's commentary converge on evaluation as the binding constraint.
TL;DR
- A PubMed review examines the state of the evidence for AI in clinical trials, its gaps and next steps.
- Related papers evaluate patient-facing generative AI and an AI preoperative planning system for high tibial osteotomy; an ISPOR task force reports good practices for generative AI in systematic reviews.
- A STAT commentary argues AI for older adults deserves careful scrutiny.
A PubMed review titled "Artificial intelligence in clinical trials-state of the evidence, gaps, and next steps" surveys where AI has been used in trials and where the evidence is thin. A second paper studies the interpretive influence and system-level evaluation of patient-facing generative AI, shifting scrutiny from model accuracy to how patients read its output. [1] [2]
Applied evaluations continue in surgical and evidence-synthesis settings. One study reports that an AI-based preoperative planning system for high tibial osteotomy significantly enhances planning speed with accuracy comparable to surgeons, and an ISPOR special task force sets out good practices for using generative AI in systematic literature reviews. [3] [5]
Commentary in the science and health press adds a population-specific caution. A STAT piece argues that AI for older adults deserves careful scrutiny, as told by a geriatrician, focusing on where deployment assumptions may not hold for that group. [4]
Why it matters
AI is entering care through trials, patient-facing tools and evidence synthesis at once. The recurring finding across these items is that evaluation design, not model capability, determines whether the tooling helps.
Editor's note
Abstracts are not stored in the collection, so each item is described from its title and venue; no specific performance figures are asserted beyond the wording of the titles.