The Armitage and Doll version of the multistage model (see last slide) was later found to be quite useful as a quantitative tool. For example, a cohort analysis based on the multistage approach indicated that coke oven emissions do not appear to have a late stage effect of lung cancer in steelworkers (Dong, 1984; Dong et al., 1988). The number of carcinogenic stages involved was estimated to be 4 in that analysis, although there was only moderate to weak indication that the coal tar in the emissions acts as an initiator. These findings are compatible with those derived from other empirical data, such as that by Day and Brown (1980) showing that the benzo(a )pyrene fraction in the cigarette smoke condensate also affects predominately an early carcinogenic stage. One key substance contained in coal tar is also benzo(a )pyrene.

Like the single stage and multicell models described in the last slide, the multistage theory is not completely free of defects. For example, Moolgavkar (1978) has argued that the multistage theory can legitimately be applied only to cohort data for tumors that are not sensitive to environmental influences.

Nonetheless, it has been shown that mathematical models based on the general multistage theory are for the large part successful in describing many experimental and epidemiological data (Day and Brown, 1980; Whittemore and Keller, 1978). The multistage theory is gaining wide acceptance in part because there had been strong evidence that cancer is a single cell in origin (Fialkow, 1974; Gartler, 1974).