Statistical power analysis of standard cardiovascular safety pharmacology studies in telemetry implanted dogs and nonhuman primates

Abstract

Pivotal cardiovascular (CV) safety pharmacology studies using telemetered non-rodent (dog and nonhuman primate (NHP)) models provide key data that enable development of novel therapeutics. Statistical power calculations demonstrate the sensitivity of an experimental model as well as provide rationale for study design including sample size selection. The power of a statistical test is the probability of detecting a signal (e.g. a CV effect) when there truly is a signal. Robust understanding of statistical sensitivity also underpins the confidence in study results, yet systematic power analysis of standard CV studies is currently lacking. We analyzed pooled data from CV telemetry studies in standard cynomolgus monkeys (n = 21) and beagle dog (n = 27), separately, to determine the statistical power of these experimental models. Studies typically utilized a 4 × 4 (dog) or 8 × 4 (NHP) vehicle +3 dose level crossover paradigm. Data were collected for approximately 24 h, and derived results were binned into time intervals for statistical analysis using a linear ANOVA model. The minimum detectable differences (MDD) with 80 % statistical power were calculated for standard parameters (e.g. blood pressure (BP), heart rate (HR), ECG intervals etc). MDDs for dogs, using a N = 4 crossover design, were: BP (5–7 mmHg), HR (10 bpm), QT-interval (9 msec), and QTc-interval (6 msec). MDDs for NHP, using a N = 8 crossover design, were: BP (4–5 mmHg), HR (11 bpm), QT-interval (13 msec), and QTc-interval (9 msec). Additionally, we also report MDDs for alternate groups sizes (e.g. N = 4, 8 and 12) as well as reference intervals of root mean square error (RMSE) as a measure of variability in the studies. Using the 2.5th and 97.5th percentiles of the RMSE, we also report the lower and upper bounds of the MDDs for each parameter. Overall, our results indicate that the nonrodent CV model is a sensitive tool to detect CV risk in early safety studies. Furthermore, the results also demonstrate assay sensitivity of functional endpoints (e.g. QTc MDD <10 msec) and support use of data in the context of ICH E14/S7B Q&As. Lastly, these results will enable informed selection of appropriate models and study designs for CV studies.