Analysis of Animal Carcinogenicity Experiments with Multiple Tumor Types


Full text of this article may be found in Biometrics, 51(10), 73-86, 1995.


In animal carcinogenicity experiments, 50 to 100 animals per dose are treated with a potential carcinogen (3 to 4 doses, including controls), and followed over time to observe the development of several types of tumors. Common practice for analysis of such data involves testing the effect of the carcinogen on each type of tumor separately as if that is the only tumor of interest. This raises the problem of multiple comparison. Moreover, due to the occult nature of these tumors, the time to tumor is not observable. Therefore, information about the tumor incidence rate is confounded by the mortality rate and the lethality of tumor.

We will address the question of identifiably for the multiple tumor development model and use the model to develop tests for comparing two or more treatments with respect to marginal tumor incidence rates of several types of tumors simultaneously. Proposed methods are applied to data from the ED01 study, designed to describe the carcinogenic effect of the compound 2-AAF at lower doses in about 24,000 animals and in 15 different sites. Some simulation results, based on the ED01 study, are presented to discuss the use of these procedures with smaller experiments.


Disease Modeling | Epidemiology

This document is currently not available here.