Comment from Lamia Anwar

Dear Sir/Madam, Thank you for the opportunity to comment on the draft guidance Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products (January 2025). The document represents an important step in supporting the responsible and effective use of AI in drug and biological product development. I respectfully submit the following comments for your consideration: 1. Scope Clarification While the exclusion of drug discovery and operational efficiencies is noted, these areas increasingly influence regulatory submissions (e.g., protocol optimization, patient recruitment). Greater clarity on how FDA distinguishes between “operational” versus “regulatory” use cases would support consistent interpretation. 2. Risk-Based Credibility Framework The proposed 7-step framework is valuable, but the model risk matrix lacks objective scoring criteria for categorizing low, medium, or high risk. Introducing clear thresholds or illustrative examples would improve consistency in implementation across sponsors. 3. Data Quality and Bias The emphasis on dataset representativeness is welcome. However, explicit guidance on assessing and mitigating algorithmic bias, particularly for underrepresented populations (e.g., pediatrics, rare diseases), would strengthen regulatory confidence in AI outputs. 4. Model Transparency and Explainability While sponsors are asked to describe model architecture and training, more emphasis on explainability methods (e.g., SHAP, LIME) would enhance trust when AI models significantly influence regulatory decisions. 5. Lifecycle Maintenance and Data Drift The discussion on lifecycle maintenance appropriately highlights data drift, but does not provide quantitative thresholds for when retraining or revalidation is required. Including examples of acceptable performance degradation ranges would support effective risk-based change control. 6. International Harmonization Alignment with ICH guidances (e.g., Q9(R1), Q10, Q12) should be more explicitly discussed to facilitate global regulatory convergence and reduce variability across submissions to multiple agencies. In summary, the draft provides an important foundation, but further clarity on scope, model risk categorization, bias assessment, lifecycle management, and international harmonization would significantly enhance its utility for sponsors and regulators alike. Thank you for considering these comments. I look forward to the finalized guidance. Sincerely, Dr Lamia Anwar Egyptian Drug Authority Lamia.Anwar@edaegypt.gov.eg lamiaanwar85@gmail.com +2001007832111