Use of dual control groups to estimate false positive rates in laboratory animal carcinogenicity studies

Tumor incidence data from 18 recently completed carcinogenicity studies utilizing male and female mice and rats were examined to determine if the frequency of significant (p less than 0.05) pairwise differences between the two concurrent control groups employed in these experiments exceeded chance expectation. Although marked study-to-study variability was observed for some tumors, no evidence of extra-binomial within-study variability between the two concurrent control groups was found. The total number of observed significant (p less than 0.05) paired-control differences was virtually identical to what would be expected from the usual binomial model assumptions; the corresponding overall observed (44%) and expected (47-50%) false positive rates were essentially the same. While one should not overgeneralize the implications of these findings, these results should lessen concerns that elevated false positive rates resulting from extra-binomial within-study variability might be adversely affecting the interpretation of long-term laboratory animal carcinogenicity studies. On the other hand, these results reaffirm the conclusions of other investigators that (particularly for commonly occurring tumors) more stringent evidence than an isolated p less than 0.05 effect should be required before an increased tumor incidence is regarded as biologically significant; otherwise, the study may have an unacceptably high false positive rate.