make it difficult to interpret the biological significance of the increase in the incidence of MCL observed in the treatment groups. Indeed, NTP has decided to stop using its F344/N rat colony in its bioassays for reasons that include the high background rate of MCL (King-Herbert and Thayer 2006). A supplemental analysis performed by NTP considered disease progression, latency, and various statistical treatments. The analysis suggested an increase in tumor incidence over controls at both test concentrations despite the high spontaneous tumor incidence in the controls.
The significance of MCL findings in multiple NTP bioassays that used the F344 rat was the subject of a recent reanalysis by Thomas et al. (2007), which EPA should reference in the draft IRIS assessment. They examined the incidence of leukemia in 2-year bioassays that included untreated male and female F344 rats from 1971 to 1998. They found that background tumor incidence increased substantially, from 7.9% to 52.5% in males and from 2.1% to 24.2% in females, over that period. The analysis also found that MCL responses are highly variable and subject to substantial modulation by dietary factors.
Thomas et al. (2007) also evaluated MCL incidence in male and female rats exposed to 500 chemicals. On the basis of 34 NTP studies that yielded evidence of a chemically related increase in the incidence of leukemia, which included the 1986 NTP study of tetrachloroethylene, the authors conducted a reanalysis of dose-response data by comparing results with four statistical methods: Fisher’s test for pair-wise comparison of leukemia incidence between a dose group and a control group, the Cochran-Armitage test for incidence trend, logistic regression for incidence, and life tables for survival-adjusted incidence. Tetrachloroethylene was one of five chemicals shown by the authors to produce leukemia in both sexes of rats. They used the rigid Food and Drug Administration (FDA) statistical criteria for testing dose-related cancer incidences (p < 0.01 for pairwise comparison; p <0.005 for trend test). The results in male rats in the 1986 NTP study revealed a significant dose-response trend when analyzed with a life table (p = 0.004) assuming that MCL is lethal but a nonsignificant trend with logistic regression (p = 0.097) assuming the MCL is nonlethal. Pairwise comparisons revealed dose-related incidences (p = 0.046) for both dose groups, and the trend test yielded a p value of 0.034; neither met the FDA criteria for statistical significance. The borderline significance of the trend test and nonsignificance of logistic regression for the latter two comparisons could be explained in part by the fact that the incidences did not follow an incrementally increasing relationship with dose. In female rats in the NTP study, use of a life table (p = 0.053), logistic regression (p = 0.012), a trend test (p = 0.018), and Fisher’s test (p = 0.014 and 0.022, respectively, for two doses) all revealed a borderline significant dose-related incidence. However, there is inconsistency in statistical significance between the sexes and uncertainty about the shape of the dose-response curve, especially in the lower range of the study. The authors recommended the use of life-table analysis for survival-adjusted leukemia incidence, noting that it is “closer to reflecting the true statistical significance of the carcinogenic effect” than logistic-regression treating dose as linear. Life-table