Using AI And ML In Drug Development: Five Takeaways From The FDA

The Food and Drug Administration (FDA) revealed the release of a new discussion paper on Wednesday, May 10, 2023, with the title “Using Artificial Intelligence and Machine Learning in the Development of Drug and Biological Products.” The purpose of the discussion paper is to encourage a dialogue with stakeholders about the use of artificial intelligence and machine learning (AI and ML) in the creation of pharmaceuticals as well as medical devices that are meant to be used in conjunction with them. Three primary areas are covered in the discussion paper: the landscape of existing and potential uses of AI and ML, factors to take into account when using AI and ML, and next steps and stakeholder engagement.

Here are the top five conclusions:

1. The FDA recognizes the existence of AI and ML applications at every stage of drug development: All phases of drug research, from drug discovery to pharmaceutical production, may benefit from the use of AI and ML. In order to enhance drug development, AI and ML have been applied to both real-world data (RWD) and data from digital health technologies (DHTs). The discussion paper’s first section provides a summary of the various applications of AI and ML in pharmaceutical manufacturing, clinical and nonclinical research, post-marketing surveillance, and drug discovery.

2. The FDA is expanding its own expertise in using AI and ML for drug development: The FDA has noticed an increase in medication and biological product filings that mention AI and ML in recent years. In response, the FDA has taken a variety of steps, including creating the CDER AI Steering Committee, the Model-Informed Drug Development (MIDD) Pilot Programme, and the Innovative Science and Technology Approaches for New Drugs (ISTAND) Pilot Programme. The CDER Sentinel System, CBER Biologics Effectiveness and Safety (BEST) system, and CDRH National Evaluation System for Health Technology (NEST) projects are investigating AI and ML ways to improve current systems for postmarket safety surveillance.

3. The FDA is aware It is essential to develop standards and best practises for using AI and ML: The American government and the global community have shown a stronger commitment to promoting AI innovation and uptake. To help the development of moral AI, regulators and standards bodies have created and released standards. In August 2019, for instance, the National Institute for Standards and Technology (NIST) published “U.S. Leadership in AI: A Plan for Federal Engagement in Developing Technical Standards and Related Tools.” The FDA, Health Canada, and the UK’s Medicines and Healthcare Products Regulatory Agency (MHRA) also produced a list of 10 guiding principles in October 2021 to help with the creation of Good Machine Learning Practises (GMLP) for medical devices employing AI or ML.

4. Key Questions on AI and ML in Drug Development Have Been Identified by the FDA: The FDA wants to start a conversation with stakeholders about the three following important topics and offers specific questions to get input.

Human-led accountability, transparency, and governance

What particular use cases or applications of AI or ML in drug development call for more regulatory clarity the most?
What, in your opinion, are the primary impediments to and enablers of transparency with respect to AI and ML employed in the drug development process (and in what context)?
How are pre-specification activities controlled, changes recorded, and monitoring procedures carried out to guarantee the safe and efficient application of AI and ML in drug development?
Data’s accuracy, dependability, and representativeness

What other data factors should be taken into account for AI and ML during the medication development process?
What are some of the most important techniques used by interested parties to support data security and privacy?
What procedures are developers using to manage and identify bias?
Model creation, execution, evaluation, and validation

What procedures and records are being utilised to guide the selection of data sources and the inclusion or exclusion standards?
How have you balanced performance and explainability factors? How are stakeholders addressing explainability in the context of use?
What are some examples of the tools, procedures, methods, and best practises currently being used by stakeholders for things like choosing the right model types and algorithms for a particular application, figuring out when to use particular techniques for validating models and gauging performance in a particular setting, assessing transparency and explainability, boosting model transparency, etc.?
5. FDA Requests Your Comments: The FDA is looking for input on the advantages and disadvantages of employing AI and ML in the development of pharmaceuticals and medical devices. In the discussion paper, the FDA asks a number of questions for feedback, and a workshop with stakeholders is scheduled to offer a chance for more interaction.