regression line in Region B; the TD50 based on the data in the CPDB falls above the upper confidence limit on the regression line in Region A.
The TD50s in the CPDB for the 12 potential supercarcinogens already included in Krewski's 191 chemicals are generally very close to those calculated by Krewski. There are, however, two notable differences: EDB and vinyl chloride. The TD50 for EDB in the CPDB is based on lifetable methods, and is thus somewhat different than that calculated by Krewski using summary tumor incidence data. The discrepency between the Gold and Krewski TD50s for vinyl chloride based on the data from the CPDB is apparently due to differences in the numerical procedures used in model fitting. (This difference is small in relation to the wide variation in TD50s in the CPDB based on different experiments with vinyl chloride.) Neither of these differences is particularly relevant to the search for supercarcinogens because the TD50s for these two compounds do not fall in Region C.
DISCUSSION
The results just discussed do not provide strong evidence of the existence of supercarcinogens. Of the 14 chemicals considered as potential supercarcinogens, only five fall inside Region C; even these five are only slightly beyond the boundary separating regions B and C.
These results are based on certain assumptions about the appropriate adjustments to be applied to dose (and hence to potency) in experiments that are terminated substantially earlier than the 2 year lifetime of rodents. Those assumptions are based on sparse empirical evidence and are somewhat arbitrary. A common generalization is that cancer incidence is proportional to fn where the exponent n may range from 2 to 6 (Armitage and Doll, 1961). The CPDB's procedures are equivalent to the assumption that n = 2, which gives relatively low estimates of carcinogenic potency. An assumption that n = 3 or higher would shift the estimates of TD50 for the chemicals under review still further into Region C.
In summary, the results of the committee's informal study suggest that supercarcinogens are rare. The best candidates for designation as supercarcinogens are a few agents that induce cancer in rodents unusually