Statistical applications of virtual control groups to nonrodent animal toxicity studies: An initial evaluation

Area of Science:

• Toxicology
• Pharmacology
• Animal Research Ethics

Background:

• Virtual control groups (VCGs) are proposed to support the 3Rs (reduce, refine, replace) principle in animal research.
• Systematic comparisons of VCGs versus concurrent control groups (CCGs) in nonrodent safety studies are lacking.
• Optimal methods for VCG generation and integration with CCGs require clarification.

Purpose of the Study:

• To evaluate the statistical performance of VCGs in detecting test article effects in nonrodent systemic toxicity studies.
• To compare the sensitivity of VCGs against CCGs for body weight and clinical pathology endpoints.
• To identify optimal criteria for generating VCGs.

Main Methods:

• Retrospective analysis of 22 nonrodent (dog and nonhuman primate) 1-month oral gavage toxicity studies.
• Evaluation of three VCG generation methods, including propensity score matching.
• Comparison of VCG statistical sensitivity against CCGs using fold change from baseline.

Main Results:

• VCGs showed similar or higher statistical sensitivity than CCGs in detecting test article effects.
• Propensity score matching (by body weight, age, study initiation year) yielded higher VCG sensitivity compared to random sampling or fixed criteria.
• The findings support VCGs as a viable tool in nonrodent toxicity assessments.

Conclusions:

• Virtual control groups (VCGs) are a statistically sound and ethically advantageous alternative to concurrent control groups (CCGs) in nonrodent safety studies.
• Propensity score matching enhances the sensitivity of VCGs.
• VCGs can effectively support the 3Rs principles in toxicological research.