Data analysis: statistical analysis and use of historical control data.

Survival-adjusted methods for the statistical analysis of tumor data from long-term rodent carcinogenicity studies are described. Although most of these methods require knowledge of whether individual tumors are "fatal" or "incidental," such determinations may be difficult. Several methods for dealing with this and with other data analysis issues are discussed. Historical control tumor data may be useful in the interpretation of rodent carcinogenicity studies, particularly for rare tumors and for borderline effects. Although statistical methods are available for using historical control data in a formal testing framework, the primary difficulty is establishing a database that is truly comparable to the study under evaluation with respect to those factors known to influence tumor occurrence. Major sources of variability in tumor incidence include the animal room environment, dietary factors/body weight, gross necropsy and slide preparation procedures, and histopathology diagnosis. The National Toxicology Program's use of historical control data is briefly described and illustrated.