Association between gastrointestinal effects in rodent toxicity studies and decreased locomotor activity in rodent safety pharmacology studies

Abstract

Safety pharmacology (SP) evaluations are required for small molecule drug candidates (ICH-S7A) to identify effects on the cardiovascular, central nervous (CNS), and respiratory systems. Gastrointestinal (GI) effects are common in drug development, manifesting preclinically as emesis (large animals), nonspecific clinical observations, and/or histopathological abnormalities in GI tissues (toxicology studies). Decreases in locomotor activity (LA) are a common finding in rodent CNS SP studies; however, it is often not possible to differentiate primary CNS effects from secondary GI effects, particularly given that rodents are non-emetic. The relationships between GI-related clinical signs and GI pathology in rodent toxicity studies and decreased quantitative LA in rodent SP studies was quantified via contingency tables using Chi-Squared tests. A binary logistic regression was computed to describe the log odds of a decrease in LA when certain GI predictors are present. Data evaluated across similar dose levels for each compound (n = 65) showed a correlation between the presence of GI clinical signs and GI pathology (X^2 = 5.454, p-value = 0.01952. Sensitivity = 50 %, Specificity = 76.23 %), as well as the presence of GI clinical signs and decreased LA (X^2 = 23.265, p-value = 1.411e-06. Sensitivity = 45.45 %, Specificity = 83.06 %). Although the GI pathology and decreased LA correlation was not significant, the specificity for this association was high at 94.54 %, indicating that the there is a strong correlation between negative LA and negative pathology findings (X^2 = 0.50847, p-value = 0.4758. Sensitivity = 7.79 %, Specificity = 94.54 %). A final regression model (reduced to the lowest Akaike Information Criterion [AIC]) identified food consumption and distended abdomen as predictors. Food consumption was a significant predictor of decreased LA (p = 2.16e-05), suggesting that compounds that impact food consumption could also impact animal behavior, including activity levels. The final regression model has an AUC of 0.621 (95 % CI: 0.5533–0.6797); while not exceptionally strong, it is slightly better than random at distinguishing occurrences of decreased LA. When present, the potential contribution of GI effects to decreased LA in SP assessments should be considered in addition to direct effects on the CNS.