Introduction: The Health and Environmental Sciences Institute of the International Life Sciences Institute (ILSI/HESI) Cardiovascular Safety Subcommittee outlined a set of in vivo telemetry studies to determine how well this preclinical model identified compounds known to cause torsades de pointes (TdP) and prolong QT interval in humans. In the original analysis of these data, QT, QTcB (Bazett model), QTcF (Fridericia model), and QTcQ (animal-specific model) were evaluated. We further evaluate the statistical properties of these measurements, using a method that can properly account for the sources of variability in the dataset.
Methods: The ILSI/HESI telemetry studies were conducted as a double Latin square design where eight dogs each received a vehicle control and three dose levels of a compound on four separate dosing days. We statistically analyzed the QT/QTc intervals using a repeated measures analysis of covariance and evaluate the powers for QT, QTcF and QTcQ based on simulations.
Results: The analyses for QTcF and QTcB intervals show that all six compounds which were known to cause TdP in humans were identified as positive and all six compounds known to be free of TdP events in their clinical use had no statistically significant treatment-related effects, while the analyses for QTcQ identified all positive compounds except pimozide. The power analysis shows that the method can detect a 7% increment of QT, a 5% increment of QTcF, and a 4% increment of QTcQ, with greater than 80% of power when n=8.
Discussion: We describe a repeated measures procedure to perform statistical analysis of covariance on Latin square designs and show that it can be used to detect meaningful changes in the analysis of QT/QTc intervals.