CERSI Collaborators: Molly Jeffery, PhD (Mayo Clinic); Joseph Ross, MD, MHS (Yale); Jun (Vivien) Yin, PhD (Mayo Clinic); Xiaoxi Yao (Mayo Clinic); Che Ngufor (Mayo Clinic); Peter Noseworthy (Mayo Clinic)
FDA Collaborators: Ruth Barratt, PhD, DVM; Boris Brodsky, MA; Tala Fakhouri, PhD, MPH; Gordon Gideon, PhD; Qi Liu, PhD, MS; Scott Steele, PhD
Project Start Date: October 18, 2021
Regulatory Science Challenge
Artificial Intelligence (AI) is increasingly used in health care to improve prevention, diagnosis, treatment, and maintenance of health conditions. Trustworthy and transparent approaches to informing patients and providers about AI are needed to address the unique challenges of communicating how AI and Machine Learning (ML) enabled devices and software will be used in patient care and in clinical trials. Patient decision-making, safety, and trust in AI/ML enabled medical devices hinge on successful communication of how AI tools are developed, their performance, and how they impact care.
Project Description and Goals
The goal of this proposed research is to evaluate the use of AI to facilitate clinical trial designs for clinical trials, which will inform methods and establish standards for future regulatory submissions that contain data or study results informed by AI. This will address a key gap between the development of AI methods and the applications of these in adaptive enrichment clinical trials. Therefore, this research will systemically assess whether the application of AI is feasible in adaptive enrichment designs for clinical trials and the impact of such AI-based enrichment approaches in clinical trial operating characteristics.
Goal 1. To evaluate the applicability of AI-identified treatment outcomes during the trial as data accumulate, determining whether we can predict the HTE identified at the end of the trial, based on case studies from completed randomized controlled trials (RCTs).
Goal 2. To evaluate the applicability of AI-based adaptive strategies that allow mid-course enrichment, determining the extent to which entry criteria can be modified for later stages of a trial if factors can be identified that increase event rate or treatment response.
Upon completion, the data generated will establish standards for a data-driven approach to identify subgroups for adaptive enrichment, which will help increase the efficiency of RCTs. The findings can help the FDA better evaluate RCTs influenced by AI, determine whether the methods are trustworthy, and provide guidance to the medical community and the industry for using AI in adaptive trials.