A Prevalence Analysis that Adjusts for Survival and Tumour Lethality

If animals with or without the tumor of interest die at a common rate, the tumor is typically referred to as non-lethal and survival-adjusted prevalence comparisons can be based on mortality and tumor response data alone. If these death rates differ, however, most analyses also require cause-of-death data or sacrifice data. This paper describes a regression analysis that adjusts for survival and allows different conditional death rates but does not require any data on cause of death and often is practical with few sacrifices. The interval analysis of Hoel and Walburg and the logistic regression analysis of Dinse and Lagakos, both of which are appropriate for non-lethal tumours, are special cases of the analysis presented here. The methods proposed provide a framework for incorporating covariates, as well as for estimating the tumour's relative risk, and are illustrated with liver tumour data from the ED01 study.