1. Pink Sheet
  2. >>Advanced Technologies
  3. >>AI

US FDA Opinions Vary On AI’s Black Box Issues. Will Uncertainty Follow?

 
• By Derrick Gingery

Chief Medical Officer Hilary Marston says a black box model could be a problem for reviewers, after CDER and CBER officials said unexplainable AI models could be acceptable but transparency is important.

Transparency is important with an AI black box, the FDA said. (Shutterstock)

Some US Food and Drug Administration officials may be willing to consider artificial intelligence models that are not fully explainable, but Chief Medical Officer Hilary Marston seems concerned about their use in applications, further illustrating the agency’s ongoing challenge regulating the emerging technology.

Key Takeaways
  • FDA CMO Hilary Marston said an AI black box could be problematic because reviewers like to redo analyses in applications.
  • Marston did not rule out using a black box model, but said sponsors should discuss their intentions with the agency early in product development

AI models sometimes produce the expected output, but their creators cannot fully explain how the model reached the answer or conclusion

Read the full article – start your free trial today!

Join thousands of industry professionals who rely on Pink Sheet for daily insights

  • Start your 7-day free trial
  • Explore trusted news, analysis, and insights
  • Access comprehensive global coverage
  • Enjoy instant access – no credit card required
Start Free Trial

Already a subscriber?

More from AI

US HHS Deputy Nominee Talks AI, Not FDA, In Confirmation Hearing

 
• By Ramsey Baghdadi

Silicon Valley investor Jim O’Neill’s Senate confirmation hearings showcased the unique background of someone with government and tech experience and avoided discussions of his past statements about lowering the bar for regulatory approval.

Canada’s HTA Agency Reveals How To Submit AI-Driven Evidence

 
• By Neena Brizmohun

Recognizing that the evidence it receives in applications for health technology assessments will increasingly be informed by artificial intelligence, the CDA-AMC has clarified its expectations for companies that use AI methods in the generation and/or reporting of evidence.

AI Could Be Used ‘In The Deliberation’ Of HTA Reviews In England, Says NICE

 
• By Eliza Slawther

England’s health technology assessment institute, NICE, is looking to “reimagine” its evaluation process with the help of AI, rather than just using this technology to speed up its existing processes.

EMA’s Newly Qualified AI Tool To Boost MASH Market Dynamics

 
• By Neena Brizmohun

The European Medicines Agency’s qualification of the AIM-NASH tool is said to signify a major advancement for clinical trials for metabolic dysfunction-associated steatohepatitis. The market size for MASH treatments is expected to grow substantially in the coming years.

More from Pathways & Standards

UroGen’s Mitomycin: Are Randomized Trials For Bladder Cancer Necessary?

 
• By Sue Sutter

The Oncologic Drugs Advisory Committee will vote on the risk-benefit of UGN-102 for low-grade, intermediate-risk, non-muscle invasive bladder cancer based on data from a single-arm trial, but the FDA also seeks input on requiring randomized studies for future treatments.

Life After Loper: Vaping Case Shows US FDA Retains Considerable Deference

 
• By Craig Bleifer, Royce DuBiner, Max Bogost, Michael Brody, and Allyson Maur

A US Supreme Court ruling in favor of the agency in an e-cigarette case has implications for drugs and other medical products as the justices decided not to disturb the FDA’s ‘change in position’ authority, while also creating a complex new landscape.

US FDA Legislative Lead Calls For User Fee Reforms As Renewal Talks Near

 
• By Sarah Karlin-Smith

Deputy FDA Commissioner Grace Graham acknowledged the importance of user fees, while also calling for restructuring. She also said part of MAHA's mission is to reexamine uses of drugs not supported by data.