Guest Column | October 6, 2025

SCDM 2025: Practical Advice From Global Regulators — Including The FDA

By Abby Proch, executive editor, Clinical Leader

SCDM2025

SCDM 2025 opened its second day with a regulatory townhall featuring representatives from around the world to discuss the impact of ICH E6(R3) and how the FDA evaluates the use of AI in scientific reviews. 

Echoing sentiment from another panel discussion, Cheryl Grandinetti, associate director for clinical policy, Division of Clinical Compliance Evaluation, Office of Scientific Investigations, Office of Compliance, CDER, FDA, reminded attendees that “perfect trial data is unrealistic” and that, instead, the FDA is looking for data that’s fit-for-purpose.

On the topic of integrating AI into research, Annie Saha, associate director for strategic initiatives, digital health center of excellence, CDRH, and associate director for data science and artificial intelligence policy (acting), Office of Medical Policy, CDER, FDA, reminded attendees that AI only needs to be mentioned to the FDA if it involves safety, efficacy, and quality. Out-of-scope uses, and therefore those of no interest to the FDA, include operational efficiencies. She also noted a sharply increasing trend of submissions with AI, with just one submission in 2016 and more than 300 in 2024. Oncology has the highest number of submissions thus far, and the most common submission type using AI is the IND, with 248, followed distantly by NDAs, BLAs, and others. Use cases include patient selection, outcome prediction, confounding adjustment, pharmacometrics modeling, digital twins/external control arms/synthetic controls, and post-market safety monitoring.

Following the presentations and subsequent discussions, the global regulators were asked to impart one bit of practical advice — related to E6(R3), AI, or otherwise — and here’s  a summation of what they said:

  • If you’re using quality by design (QbD), critical to quality factors (CTQ), and risk-based quality management (RBQM), you’re on the right path. — Cheryl Grandinetti, FDA
  • If you’re planning to use AI in your clinical research — whether for trial design, dosing, manufacturing, or something else — talk to the FDA early. — Annie Saha, FDA
  • There is no expectation from the FDA to achieve 100% source data verification. Do not obsess about trying to achieve it. — Kassa Ayalew, division director, Division of Clinical Compliance Evaluation (DCCE), FDA
  • Ahead of an inspection, have critical vendors on standby, because they’ll need to talk through documentation details with inspectors. — Daniel Bjermo, pharmaceutical inspector GCP, Swedish Medical Products Agency
  • Take care to develop blinding and unblinding plans, and document protocol deviations thoroughly. — Lisbeth Bregnhøj, GCP inspector, The Danish Medicines Agency (DKMA)
  • Qualify your clinical trial partners; do not just go with “trust.” Also, objectify risk. Don’t go with gut feelings like, “It doesn’t feel like a big problem.” — Torsten Stemmler, head of GCP Inspection Unit, Federal Institute for Drugs and Medical Devices Germany (BfArM)
  • Regulators want to work with you, not against you. No reverence is needed; they’re people and patients just like you. — Rachel Mead, senior GCP inspector, MHRA, U.K.
  • Do not work in silos. — Myriam Salem, national compliance and enforcement supervisor, Clinical Trial Compliance Program (CTCP), Health Canada
  • If running trials in Canada, you have until April 1 for ICH E6 (R3) to take effect. Use the next six months to enroll in training and ask questions. — Alicja Kasina, senior compliance and enforcement advisor, Clinical Trial Compliance Program (CTCP), Health Canada