By Anna Bratulic
The FDA has kicked off the new year with a significant draft guidance on the use of AI to produce information or data on the safety, effectiveness or quality of drugs that would then be used to support regulatory decision-making for those medicines.
The agency says there has been an uptick in AI-powered submissions for drug approvals, reflecting the growing role AI is playing in the drug development process.
The draft guidance released on Monday provides a framework for evaluating the credibility of AI models used in these situations. The US regulator notes that while AI, including machine learning, can help streamline clinical trials, improve manufacturing processes, and enhance post-marketing surveillance, issues like data quality, algorithmic bias, and the complexity of AI models pose challenges.
To address these, the guidance introduced a seven-step plan to ensure AI systems remain reliable in pharmaceutical development. The framework requires companies to clearly establish how their AI will be used and assess its potential risks, while implementing continuous monitoring throughout the model’s lifecycle in order to ensure it performs as intended. The latter would help avoid issues like data drift, for example, where performance deteriorates over time or in different settings because new data inputs can differ from the data used to train the model.
“AI-based models may be highly sensitive to variations or changes in model inputs,” the FDA said. “Due to the evolving nature of AI models, sponsors should anticipate inherent, model-directed changes and the need to identify and evaluate those changes, as well as any intentional changes to the model over the drug product life cycle.”
