Use of historical control data in carcinogenicity studies in rodents.

This paper considers the use of historical control data in the evaluation of tumor incidences from carcinogenicity studies in rodents. Although the most appropriate control group for interpretative purposes is always the concurrent control, there are instances in which the use of historical control information can aid an investigator in the overall evaluation of tumor incidence data. One example is rare tumors; another is a tumor that shows a marginally significant result relative to concurrent controls. However, before historical control data can be used in a formal testing framework, a number of important issues must first be considered. The nomenclature conventions and diagnostic criteria for each study should be identical to insure unambiguous identification of all relevant tumors in the historical control database. Criteria should be established that will aid in determining whether a particular study should be included in the database. This will assure a homogeneous set of studies upon which to base statistical comparisons. Since study-to-study variability in tumor rates may exceed what would be expected by chance alone, these sources of variability should be identified and controlled. Finally, statistical procedures should be employed that adjust for extra-binomial variability. This paper also summarizes tumor incidence data from untreated Fischer 344 rats and B6C3F1 mice in the National Toxicology Program (NTP) historical control database. All studies in the database are of two years duration, and all neoplasms occurring with a frequency of 0.5% or more are reported.