Data collection and statistical analysis

These studies were conducted in compliance with Nonclinical Laboratory Studies Good Laboratory Practice Regulations issued by the U.S. Food and Drug Administration (Title 21 of the Code of Federal Regulations, Part 58). Data were collected into Provantis v9.2.3 (Instem, Philadelphia, PA) and calculation of endpoints was performed within this validated electronic data collection and management system.

Results are presented as mean ± SEM. Jonckheere’s test was used to test for dose-related trends (Jonckheere 1954). Body weight and organ weight data, which typically exhibit a normal distribution, were analyzed using a parametric multiple comparison procedure. If a significant trend was detected at p ≤ 0.01, a Williams’ test was used (Williams 1986); if the trend was not significant, a Dunnett’s test was used (Dunnett 1955). Data for other endpoints were analyzed using a non-parametric multiple comparison procedure. If a significant trend was observed Shirley’s test was used (Shirley 1977); if the trend was not significant Dunn’s test was used (Dunn 1964). Positive control group data was compared to the vehicle control group using the Kruskal-Wallis test. Data that were different from control at p ≤ 0.05 were considered significant. Extreme values were identified by the outlier test of Dixon and Massey (Dixon and Massey 1957). All flagged outliers were examined by NTP personnel, and implausible values were eliminated from the final analyses.