Comment from IHI VICT3R project

The EU Innovative Health Initiative (IHI) VICT3R project (https://www.ihi.europa.eu/projects-results/project-factsheets/vict3r) is a groundbreaking public-private consortium leveraging Artificial Intelligence (AI) and legacy data from in vivo toxicity studies to develop Virtual Control Groups (VCGs), advancing the 3Rs principles (Replacement, Reduction, Refinement) in toxicology research. Given its focus on AI-driven regulatory-grade synthetic data generation, the consortium is keen to contribute insights to the FDA’s draft guidance, ensuring it accommodates emerging approaches like VCGs while upholding scientific and regulatory standards. The Comment highlights three key recommendations: 1. Expanding the scope of the AI definition to include VCGs The guidance acknowledges (lines 72-76) that “reducing the number of animal-based pharmacokinetic, pharmacodynamic, and toxicologic studies” is one of the frequent uses of AI in submissions to the FDA. Therefore, we recommend that the guidance should explicitly address (lines 24-32) the role of AI in generating VCGs, which can reduce reliance on traditional control groups in preclinical and clinical research. A discussion on feasibility, ethical considerations, and validation frameworks for AI-based VCGs would enhance the document’s applicability to cutting-edge methodologies. 2. Clarifying the definition and assessment of "Credibility" The current definition of "credibility" lacks specificity, making it difficult to operationalize. We recommend aligning it with established credibility frameworks (e.g., from regulatory bodies or peer-reviewed literature) and providing clear evaluation criteria to ensure consistency in AI model assessments. A non-exhaustive list of potential frameworks includes reporting guidelines from the EQUATOR network: TRIPOD+ AI [1], TRIPOD-LLM [2], CONSORT-AI [3], SPIRIT-AI [4], and others. 3. Enhancing practical guidance with more case studies and examples We appreciate that the guidance uses two use cases (one in clinical development and another in commercial manufacturing) that help demonstrate how an interested party should communicate with the FDA, however they don’t represent the whole spectrum of use cases that the guidance refers to. The document would benefit from additional illustrative examples, such as real-world applications or hypothetical scenarios, to clarify complex concepts. Concrete case studies, e.g. the use of AI for VCGs, would improve understanding of how AI can be effectively integrated into regulatory decision-making while addressing potential challenges. References: [1] Collins GS, et al TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ. 2024;385:e078378. [2] Gallifant J, et al The TRIPOD-LLM reporting guideline for studies using large language models. Nat Med. 2025. [3] Liu X, et al Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI Extension. Lancet Digital Health. 2020;2(10):e537-e548. [4] Rivera SC, et al Guidelines for clinical trial protocols for interventions involving artificial intelligence: the SPIRIT-AI Extension. Lancet Digital Health. 2020;2(10):e549-e560.