A Novel Bayesian Method for Efficacy Assessment in Animal Oncology Studies

In pre-clinical oncology studies, the investigator often compares efficacy of two or more treatments on tumor-bearing animals over a period of time. This is accomplished either through comparison of tumor volumes at various time points or overall survival. Due to ethical concerns, animals are euthanized before the end of the study whenever their tumor volumes reach a prespecified limit, the loss of bodyweight is beyond 20% or severely ulcerated tumors are observed. The traditional cross time-point tumor-size comparisons should not be performed when animals leave the study in a nonrandom nature. Survival analysis may alternatively be carried out to compare the time-to-euthanasia or natural death across treatment groups. As a result of government regulations, however, animal survival studies are hampered by small-sample sizes, often making statistical inference of survival data difficult. For example, it may be impossible to estimate median survival for an efficacious treatment group in which the majority of animals survive to the end of the study. Such cases create challenges to investigators who wish to rank-order the treatment effects, in effect picking the “winners” for further investigation. In this aricle, we list the benefits and shortcomings of popular methods and then propose a novel hypothesis test based directly on the survival probability to rank order the efficacy of treatments. The performance of the method is illustrated using a real-life example.