||Use of Tumour Lethality to Interpret Tumorigenicity Experiments Lacking Cause-of-Death Data.
Lagakos, S. W. ;
Louis., T. A. ;
||Harvard School of Public Health, Boston, MA. Dept. of Biostatistics. ;Minnesota Univ., Minneapolis.;Health Effects Research Lab., Research Triangle Park, NC.
Statistical analysis ;
Cause of death ;
Carcinogenicity tests ;
||Most EPA libraries have a fiche copy filed under the call number shown. Check with individual libraries about paper copy.
Several conditions complicate the comparison of control and exposed groups in tumorigenicity experiments. Time to onset is not observable, tumor types vary in their lethality, and risk of death from non-tumor causes can be quite different in the control and exposed groups. Two tests are commonly used for comparing control and exposed groups in such experiments, one of which assumes non-lethality of tumors and one of which assumes rapid lethality. When the risks of death from non-tumor causes in the groups differ, both tests are biased for tumors of intermediate lethality and can give very different indications about the tumorigenicity of the compound under test. A method is presented here to resolve some of the confusion that can arise in such cases, using an incomplete data analogue of a cause-of-death specific test for situations where the cause is not available, and examples are presented to illustrate its use. (Copyright (c) 1988 Royal Statistical Society.)