The use of virtual control groups (VCGs) has the potential to reduce animal usage by up to 25%, however, the application of VCGs to preclinical regulatory toxicity testing has not occurred primarily due to the hitherto unquantified variability of control animal data in relation to experimental conditions (sex, strain, administration route, etc.). The main barrier to such analysis has been the lack of access to sufficiently large datasets. Here we aim to address this through a retrospective analysis of preclinical histopathology data in the eTOX database. Firstly, we curated the original 6,111 studies which, owing to their origination within 13 distinct eTOX partner organisations, consisted of a heterogeneous range of experimental conditions and histopathological findings. The curated dataset contained studies conducted across 8 species, comprising between one and eight different strains each, making up 800,000 individual data points of observations across 28 selected histopathological findings. We then quantified the variability in histopathological finding prevalence with respect to experimental conditions. Changes in experimental conditions were related to significant differences (chi-squared test FDR p-value ≤ 0.05) in the prevalence of a subset of control animal histopathological findings which varied according to species, organ class, and finding type. For example, Sprague-Dawley rats were observed to have significantly higher prevalence rates across a range of liver histopathologies compared to Wistar Han rats in short-term studies (time point of observation ≤ 28 days). Finally, we analysed the impact of pooling control animal data investigated under different experimental conditions (pseudo-VCGs) on a data-driven treatmentrelatedness relabeling of histopathological findings and compared these to the labels assigned by the original study pathologist. We observed an increased propensity to assign treatment-related labels (leading to a concurrent increase in recall, but a decrease in specificity compared to the study pathologist labels) as the VCG became larger and less similar in terms of experimental conditions to the dose group animals, with the derived label also becoming increasingly dissimilar to the label assigned when only the concurrent control group was used. The results highlight the effect experimental conditions can have on control animal histopathology prevalence rates and how pooling control data across such conditions can lead to a divergence in preclinical finding labeling compared to when only the concurrent control group is used.