As artificial intelligence becomes embedded in clinical workflows, trials must accommodate ongoing monitoring and updates. Access to this article via Institution of Civil Engineers Library is not available.
Ellenberg, S. S., Fleming, T.
R. & DeMets, D.
L. Data Monitoring Committees in Clinical Trials: A Practical Perspective (John Wiley & Sons, 2019).
Kilbourne, A., Chinman, M., Rogal, S. & Almirall, D.
Annu. Rev.
Public Health 45 , 69–88 (2024). Support for this work was provided by the Varma Family Chair in Biomedical Informatics and Artificial Intelligence and CIFAR AI Chair to A.G.; the US National Institutes of Health, R37-AI29168 to T.R.F.; the PatientCentered Outcomes Research Institute (PCORI) Award (ME-2022C125619) to J.F.; and a grant from the Agency for Healthcare Research and Quality (R01HS027431) to J.W.
The views presented in this work are solely the responsibility of the author(s) and do not necessarily represent the views of the PCORI or its board of governors or methodology committee. These authors contributed equally: Wouter A.
C. van Amsterdam, Michael Oberst, Jean Feng.
This Research Highlight describes the imperative to adapt clinical trial methods as artificial intelligence (AI) tools become integrated into healthcare operations and are subject to ongoing monitoring and iterative updating.
It highlights that traditional trial designs may not be adequate for AI systems that evolve post-deployment and that trial frameworks should accommodate continuous evaluation and modification.
The piece reports on work published in Nature Medicine (van Amsterdam WA C et al., 2026).
Multiple authors contributed equally.
Funding sources cited include institutional chairs in biomedical informatics and AI, several US federal grants, and PCORI; an explicit disclaimer notes the authors’ views do not necessarily reflect PCORI.
A number of authors disclosed industry affiliations and roles.
The article emphasizes methodological considerations rather than presenting primary clinical data.
It links the need for evolving trial designs to the ongoing monitoring and updating characteristic of many AI systems in clinical workflows.
The content references established resources on trial data monitoring and implementation science, indicating a synthesis grounded in existing methodological literature.
No specific patient population, clinical condition, or single AI product is described in the summary.
The focus is on the class of continuously monitored and updated AI systems used within clinical workflows rather than a tested intervention in a defined cohort.
Explicit outcomes from empirical testing are not reported in the Highlight.
The key messages are conceptual: trials must be structured to permit continuous surveillance of performance and to incorporate updates while retaining rigorous evaluation; monitoring bodies and trial infrastructure may require adaptation.
The highlight does not provide details of empirical results, specific trial protocols, or validated operational frameworks.
It does not report comparative outcomes, safety data, or implementation metrics.