Interpreting the statistical analysis results of a virtual control group in toxicity studies and potential applicability of an artificial synthetic control group

• This study evaluated the reproducibility of statistical results when using VCGs • We assessed ASCGs as an alternative to reduce animal use in line with the 3Rs • We applied Open TG-GATEs data to compare VCGs, ASCGs, and CCGs • Low-dose significance had poor reproducibility, requiring cautious interpretation • VCGs and ASCGs are proposed as feasible tools for NAMs in non-clinical safety studies Reducing animal use in toxicity testing is a key principle of the 3Rs (Replacement, Reduction, and Refinement), driving interest in alternative approaches, such as virtual control groups (VCGs). This study evaluates the impact of introducing a VCG on statistical outcomes in non-clinical toxicity studies. Using clinical chemistry data from the Open Toxicogenomics Project Genomics Assisted Toxicity Evaluation System database, we retrospectively replaced the concurrent control group (CCG) with a VCG and conducted statistical analyses to assess the reproducibility of the statistical results. We performed similar analyses using an artificial synthetic control group (ASCG) and assessed the applicability of the ASCG in the toxicity study. Phospholipid measurement, which was normally distributed in the historical control data (HCD), was selected as the target parameter. A study showing statistically significant decreases in phospholipids from the low-dose group was chosen. One hundred VCG sets were created and used to replace the CCG, followed by 100 statistical comparisons with the treatment groups. The reproducibility of statistical significance in the low-dose group was low. Similar trends were observed in the analyses using the ASCG. These results suggest that both VCGs and ASCGs are implementable in toxicity studies, but replacing the control group may alter statistical outcomes. Particularly in studies with small sample sizes, statistical significance may be influenced by chance, indicating that such significance should be interpreted as a potential indication of toxicologically significant changes. Comprehensive evaluation of all data, including clinical pathology, general condition, and histopathology, is essential in toxicity assessment.