Introduction
The FDA is set to implement AI-assisted scientific review processes across its centers, with a full deployment target by June 30, 2025. This initiative [1] [2], led by Commissioner Makary, aims to enhance the efficiency and consistency of premarket application evaluations for drugs, biologics [1] [2], and medical devices.
Description
Commissioner Makary has mandated the FDA to implement AI-assisted scientific review across its centers [1], aiming for full deployment by June 30, 2025 [1]. This initiative will be utilized in the Center for Drug Evaluation and Research (CDER) for new drug applications [1], in the Center for Biologics Evaluation and Research (CBER) for biologics license applications [1], and in the Center for Devices and Radiological Health (CDRH) for device marketing submissions [1]. The deployment is overseen by Chief AI Officer Jeremy Walsh and Sridhar Mantha [1], who previously directed the FDA’s Office of Strategic Policy [1].
The primary objective of AI-assisted review is to enhance the efficiency of premarket application evaluations [1]. However, the FDA’s review process is influenced by user-fee driven goals [1], raising questions about whether sponsors will experience improved review timelines or if existing timelines will remain unchanged [1]. Concerns have been raised regarding potential layoffs at the FDA [1], which could impact review timelines [1], although AI-assisted review may help mitigate these delays [1].
The involvement of AI in the scientific review process may become a focal point in future appeals or requests for supervisory review following adverse decisions on premarket applications [1]. This raises critical questions about the role of AI as a decision support tool versus its potential undue influence on FDA reviews [1], paralleling ongoing inquiries into clinical decision support software [1]. The deployment of AI could introduce new legal challenges for courts if applicants pursue judicial review [2].
Generative AI operates on a foundation of existing data [1], which may include FDA review documents from other sponsors [1] [2]. The extent to which trade secrets and confidential commercial information from one application could inform the evaluation of another remains uncertain [1]. While the FDA is expected to utilize a closed system to safeguard reviewed information from public disclosure [1], specifics are not provided [1].
AI-assisted review could enhance the FDA’s ability to conduct holistic evaluations of premarket applications [2], potentially providing a broader context for data comparison across similar drugs [2], biologics [1] [2], and devices [2]. This could lead to more consistent premarket reviews [2], addressing challenges sponsors face with detailed critiques of trial data [2]. However, it is unclear how the AI tools will adapt to evolving product categories and whether they will be designed to exclude prior applications that could bias future evaluations [1].
The FDA plans to release further details on its AI initiative [2], which may include guidance documents or public webinars [2]. While there is currently no solicitation for feedback from industry stakeholders [2], future updates may address these outstanding questions [2].
Conclusion
The implementation of AI-assisted review processes at the FDA is poised to significantly impact the efficiency and consistency of premarket evaluations. While it promises to streamline reviews and potentially reduce delays, it also raises important questions about data security, legal implications, and the adaptability of AI tools to new product categories. The FDA’s forthcoming guidance will be crucial in addressing these concerns and ensuring the successful integration of AI into its review processes.
References
[1] https://www.jdsupra.com/legalnews/fda-announces-completion-of-ai-assisted-6550328/
[2] https://www.kslaw.com/news-and-insights/fda-announces-completion-of-ai-assisted-scientific-review-pilot-and-deployment-of-agency-wide-ai-assisted-review
