|
Many clinical trials fail because diseases are heterogeneous and sample sizes are too small. A promising strategy to increase success rates employs machine learning to identify study participants more likely to benefit from the study intervention.
In this episode of JAMA+ AI Conversations, Joseph Geraci, PhD, of Queen’s University speaks with JAMA+ AI Editor in Chief Roy Perlis, MD, MSc, about efforts to apply AI to support more efficient clinical trials focusing on highly-responsive subtypes.
He describes using AI to re-interpret completed studies and plan future ones, and discusses how regulators may evaluate these trials.
Listen now on Spotify | Apple Podcasts | YouTube | JAMA.com.
Editor’s Picks in this week’s JAMA+ AI:
- Wearable-derived health data are increasingly available in clinical settings, but the value of such data is not always clear: they appear clinically meaningful but lack clear evidence, standardized reference ranges, or established roles in diagnosis and treatment. Health systems need standards for which data belong in electronic health record workflows and how clinicians should respond, write the authors of this JAMA Perspective article. (JAMA)
- Recent FDA guidance on general wellness technologies allows many low-risk digital tools to remain outside traditional medical device oversight, but some products may still influence health-related decisions in ways consumers could interpret as medically validated. The author of this JAMA Viewpoint advocates for clearer evidence standards and labeling for digital health tools that generate individualized health insights. (JAMA)
- Along the same lines, as AI expands the capabilities of devices aimed at enhancing health, the boundary may blur between wellness support and medical device. In a JAMA Editor’s Note, JAMA Executive Editor Gregory Curfman, MD, observes that future regulation may need to focus more closely on the influence these products have on disease diagnosis or treatment, while continuing to avoid unnecessary regulation for low-risk wellness tools. (JAMA)
|
