A retrospective analysis of the good, the bad, and the ugly of virtual control groups in rat studies

ACT 2025 -- Nonclinical safety assessment requires animal models for endpoint risk assessment and dose setting to move into a clinical setting. There is regulatory- and industry-wide support toward the development of new approach methodologies (NAMs) and efforts supporting animal reduction and refinement strategies. One strategy is the replacement or reduction of control animals with virtual controls selected from the collective historical control data (HCD) available. The purpose of this work was to determine appropriate methodology to generate a working HCD pool, filter for study-specific inclusion and exclusion criteria, and randomly generate 10,000 virtual control groups (VCGs) based on the remaining virtual control pool.