Harnessing Artificial Intelligence for Regulatory Excellence in Pharmaceuticals and Biologics
In January 2025, the FDA issued a draft guidance to industry and stakeholders on the use of artificial intelligence (AI) to generate information or data supporting regulatory decisions about drug safety, efficacy, and quality. This marks a significant step in incorporating AI into the regulatory landscape, offering a clear framework to assess AI model credibility.
The Seven-Step Risk-Based Process
The guidance outlines a structured, risk-based seven-step process to establish and evaluate the credibility of an AI model:
Define the Question of Interest
Clearly articulate the topics the AI model will address.
Define the Context of Use (COU)
Detail the AI model's role and scope, including the processes that will be modeled and how its outputs will be applied.
Assess AI Model Risk
Evaluate risk based on: - Criticality: The consequence of the decisions made using the model. - Model Influence: The extent to which the model’s outputs are relied upon considering possible additional controls. Risk assessment determines the level of detail required in subsequent activities.
Develop a Credibility Assessment Plan
The plan should address: - Model development: Data used, training methods, and development processes. - Model evaluation: Data segregation, applicability, results assessment, and quality controls.
Execute the Plan
Implement the credibility assessment activities as outlined.
Prepare a Credibility Assessment Report
Document results, discuss the effects of any deviations, and provide insights into the model’s performance.
Determine Model Adequacy
Use the Credibility Assessment Report to evaluate if the AI model is suitable for its intended COU. For models that fall short, the guidance outlines possible corrective actions.
Maintenance and Monitoring of AI Models
AI models require ongoing monitoring due to their dynamic nature. Maintenance activities must align with model risk levels and be clearly detailed in plans.
FDA’s Collaborative Approach
The FDA encourages early and interactive engagement with developers during the planning and execution phases. This collaborative approach aims to set clear expectations and to address challenges proactively.
About the Author
This blog was authored by Sergi Arcas, Pharmaceutical Consultant at Rephine. For more insights on regulatory compliance and AI integration in the pharmaceutical industry, contact us at Rephine.