On March 23, 2023, FDA released a Framework for the use of digital health technologies in drug and biological product development (the “DHT Framework”). This DHT Framework is on the heels of a Discussion Paper the Agency released earlier this month on the use of artificial intelligence (AI) in drug manufacturing to seek public input on issues of critical focus (the “AI Discussion Paper”). While both actions are significant, the AI Discussion Paper is one of CDER’s few policy statements related to the deployment of AI around regulated activities (though the Center did establish an AI steering committee in 2020). CDRH, on the other hand, has issued several policy documents around AI-based software potentially regulated as “software as a medical device” (SaMD), including through an April 2019 discussion paper that proposed a regulatory framework for modifications to AI-based SaMD, an AI “Action Plan” for SaMD in January 2021, and guiding principles to inform the development of Good Machine Learning Practice for AI-based medical devices in October 2021. CDER has requested public comment on the recent DHT Discussion Paper and AI Framework by May 1 and 23, respectively.
I. The AI Discussion Paper
The AI Discussion Paper recognizes the potential value that AI offers in the pharmaceutical industry, including optimizing process design and control, smart monitoring and maintenance of equipment, and trend monitoring of consumer complaints to drive continuous improvement. In order to advance the use of AI for such purposes, the Agency seeks public feedback on specific issues as the Agency considers “the application of its risk-based regulatory framework to the use of AI technologies in drug manufacturing.” The paper indicates that it is not exhaustive of all considerations that may be important in developing an AI policy and regulatory framework, for example, difficulties that could result from potential ambiguity on how to apply existing Current Good Manufacturing Practice (CGMP) regulations to AI. Rather, the scope of the paper is focused on five critical areas identified as follows:
- Cloud applications may affect oversight of pharmaceutical manufacturing data and records. The AI Discussion Paper notes that use of a third-party AI platform or service may lead to challenges in ensuring that the third-party creates and updates AI software with appropriate safeguards for data safety and security. Additionally, FDA posits that ongoing interactions between cloud applications and process controls could complicate the ability to establish data traceability and create potential cybersecurity vulnerabilities. The AI Discussion Paper does not opine on or ask specific questions regarding the level of diligence a drug sponsor should reasonably perform with third-party vendors.
- The Internet of Things (IoT) may increase the amount of data generated during pharmaceutical manufacturing, affecting existing data management practices. On the one hand, AI solutions provide tremendous value by allowing drug manufacturers to capture a broader array of data than is typically available through non-AI methods. But FDA acknowledges that if the raw data collected during the manufacturing process increases in both frequency and types of data recorded, there may be a need to balance data integrity and retention with the logistics of data management. For example, applicants could need further clarity from regulators on which data need to be stored and/or reviewed and how loss of these data would impact future quality decisions. FDA also flagged the potential challenges in storing the data in a structured manner that enables retrieval and analysis.
- Applicants may need clarity from FDA about whether and how the application of AI in pharmaceutical manufacturing is subject to regulatory oversight, especially as AI could be used in various manufacturing operations that are the focus of FDA oversight, such as monitoring and maintaining equipment, identifying areas for continuous improvement, scheduling and supply chain logistics, and characterizing raw materials.
- Applicants may need standards for developing and validating AI models used for process control and to support release testing. FDA has published guidance and spoken on the verification, validation, and usability of digital health technologies for remote data acquisition in clinical investigations, but the AI Discussion Paper recognizes the limited availability of industry standards and FDA guidance for the development and validation of AI models that impact product quality, particularly in the drug manufacturing sphere. This lack of guidance may create challenges in establishing the credibility of a model for a specific use, especially as AI methods for drug manufacturing become more complex. For example, the AI Discussion Paper notes that applicants may need clarity regarding how the potential to transfer learning from one AI model to another can be factored into model development and validation. This area for discussion suggests that FDA guidance on this issue may be forthcoming, but the AI Discussion Paper does not delve into details of the content of any future guidance or any positions the Agency may take (e.g., how the Agency would recommend allocating implementation responsibilities across AI developers and deployers).
- Continuously learning AI systems that adapt to real-time data may challenge regulatory assessment and oversight. On the device side, Congress recently amended the FDCA to create a more streamlined and efficient process for software-based medical device submissions that could undergo a number of foreseeable AI-informed modifications throughout the total product lifecycle by allowing for the submission of “predetermined change control plans,” and the Agency has issued a draft guidance on this topic. This area for consideration in the AI Discussion Paper signals that the Agency is now thinking of these issues on the drug side as well, recognizing that applicants may need clarity on the approach for FDA’s examination of continuously updated AI control models during a site inspection and for establishing product comparability after changes to manufacturing conditions introduced by the AI model.
Leveraging the above areas for consideration, the AI Discussion Paper asks the following questions:
- What types of AI applications do you envision being used in pharmaceutical manufacturing?
- Are there additional aspects of the current regulatory framework (e.g., aspects not listed in the “areas for consideration”) that may affect the implementation of AI in drug manufacturing and should be considered by FDA?
- Would guidance in the area of AI in drug manufacturing be beneficial? If so, what aspects of AI technology should be considered?
- What are the necessary elements for a manufacturer to implement AI-based models in a CGMP environment?
- What are common practices for validating and maintaining self-learning AI models and what steps need to be considered to establish best practices?
- What are the necessary mechanisms for managing the data used to generate AI models in pharmaceutical manufacturing?
- Are there other aspects of implementing models (including AI-based models) for pharmaceutical manufacturing where further guidance would be helpful?
- Are there aspects of the application of AI in pharmaceutical manufacturing not covered in this document that FDA should consider?
II. The DHT Framework
In connection with the commitments FDA made as part of user fee reauthorizations, FDA established a framework to guide the use of data derived from digital health technologies (DHTs) in regulatory decision-making for drug and biological products. The Agency has emphasized the importance of DHTs in drug development and made advancing the use of DHTs a front burner issue, also releasing draft guidance in December 2021 encouraging drug sponsors to make use of voluntary qualification programs to allow for reliance on DHTs in multiple clinical investigations for different premarket submissions. The DHT Framework focuses on both internal programs to support DHT-related activities within FDA, and external programs to engage industry and other stakeholders. Specifically, with respect to internal programs:
- DHT Steering Committee. FDA has established a DHT Steering Committee to oversee implementation of Agency commitments related to evaluating DHT-based measurements in regulatory submissions. We recommend monitoring this Committee’s activities and watching for developments, as the Committee may make policy recommendations that impact the use and evaluation of DHT-based measurements in drug and biological product development.
- Technical Expertise and Training. FDA stated it will develop training within the drugs and biological products programs to enhance internal knowledge regarding the use of DHTs in drug development. This will be an area to watch, as the Agency may issue guidance on topics such as verification and validation of DHTs, technical and performance specifications associated with using a study participant’s own DHTs or general-purpose computing platforms in clinical investigations, how to incorporate upgrades and updates of DHTs in drug development, and incorporating DHTs using AI algorithms into drug development.
- Consistency of Evaluations Across Review Divisions. The DHT Steering Committee will help to facilitate consistent approaches to the review and evaluation of submissions that use a single DHT measurement for studies of different diseases and different drugs. This program suggests that the Agency is open to the possibility of using a single measurement for multiple purposes, which could increase efficiency and streamline submission processes.
- Statistical Considerations in the Analysis of DHT-Derived Data. FDA plans to address novel considerations for endpoints derived from DHT data and technical data specifications to facilitate submission of readily analyzable DHT-derived data supporting drug development. Sponsors should watch for Agency recommendations here, as adhering to FDA’s suggestions may facilitate the Agency’s acceptance of DHT-derived data used in submissions.
- IT Capabilities. FDA plans to enhance its IT capabilities to support the review of DHT-generated data, including by establishing cloud technology to review, aggregate, store, and process large volumes of data and implementing standards to reduce the handling necessary to make DHT data analyzable. Sponsors should consider adhering to any future data standards for DHT-generated datasets as a best practice and ensure that they have systems in place to effectively use and interact with any new cloud technologies.
With respect to external programs and engagement:
- FDA Meetings With Sponsors. The DHT Framework notes how engagements between sponsors and FDA – including the DHT Steering Committee – may occur at different stages of drug development. If deciding to use DHTs in the drug development process, sponsors should be prepared to discuss the regulatory status of DHTs, development of trial endpoints, selection of DHTs for clinical investigations, and verification and validation of DHTs.
- Drug Development Tool Qualification Program. As noted above and in our previous blog, FDA has qualification programs that are intended to support the development of DHTs for use in assessing medical products. Sponsors should consider whether to pursue qualification of DHTs as drug development tools for a specific context of use, as the DHT Framework highlights that the a qualified DHT may be relied upon in multiple clinical investigations to support premarket submissions for drugs where the context of use is the same.
- Guidance. FDA plans to issue draft guidances that reflect FDA’s current thinking on DHT topics, including on decentralized clinical trials for drugs, biological products, and devices; and regulatory considerations for prescription drug use-related software. The DHT Framework notes FDA plans to publish such draft guidances in 2023, and may publish additional draft guidances in identified areas of need informed by stakeholder engagement in 2024. Sponsors should watch for the publication of these draft guidances and consider whether to provide public comment to the Agency to inform final guidance.
- Public Meetings. FDA plans to host a series of five public meetings or workshops to gather input on issues related to the use of DHTs in regulatory decision-making related to drug and biological product development, including priorities for the development of DHTs to support clinical investigations, approaches to DHT verification and validation, DHT data processing and analysis, regulatory acceptance of safety monitoring tools that use AI-/ML-based algorithms for pharmacovigilance purposes, and emerging challenges. Sponsors should consider attending these public meetings and workshops to monitor Agency thinking on these key topics and ensure they have a seat at the table.
- Demonstration Projects. The DHT Framework states that FDA will identify at least three demonstration projects to inform methodologies for efficient DHT evaluation in drug development, which may include validation methods for DHTs, endpoint development, analytic approaches to missing data, and use of DHTs in decentralized clinical trials. These projects may inform policy development, so monitoring the projects and project feedback will be important.
- External Organizations. In addition to collaborating with sponsors and hosting public meetings, FDA also plans to engage with external organizations and participate in forums organized by professional bodies. Sponsors should monitor FDA participation in such meetings and take note of any positions the Agency takes, as these meetings could inform actions the Agency takes to meet the objectives outlined in the DHT Framework.
Companies in this space should strongly consider submitting comments on the AI Discussion Paper and DHT Framework, as industry feedback could inform the Agency’s thinking on future guidance documents or frameworks on AI/use of DHTs in drug manufacturing going forward. Indeed, at a February 17, 2023 workshop, FDA officials noted that the questions outlined in the AI Discussion Paper are “very important for [the Agency] as [they] think in terms of providing regulatory clarity for the use of AI in drug development,” with a goal of adopting “a flexible, risk-based regulatory framework that promotes innovation but also protects patient safety.”