Multistage carcinogenesis: population-based model for colon cancer.

BACKGROUND Recent laboratory work and previous models suggest that multiple mutations are associated with the development of colon cancer. PURPOSE AND METHODS To estimate the number of mutations required for colon carcinogenesis, we used likelihood analysis to obtain the best fit between mathematical models postulating different numbers of rate-limiting events and data on incidence rates of colon cancer in two human populations--the general population of Birmingham, England, from 1968 to 1972 and patients with familial adenomatous polyposis (FAP) diagnosed at Saint Mark's Hospital, London, England, between 1925 and 1965. The Armitage-Doll model and two- and three-mutation models that explicitly incorporated cell proliferation kinetics were used. RESULTS When cell kinetics were considered, models postulating either two or three rate-limiting events described the incidence rates of colon cancer equally well in both human populations. A comparison of the incidence rate of colon cancer in the general population with that in patients with polyposis suggests that a mutation at the FAP gene locus is not one of the rate-limiting events in colon carcinogenesis. CONCLUSIONS Compared with the two-mutation model, the three-mutation model was more consistent with the number of mutations reported to date in colon cancer and with mutation rates measured in the laboratory. IMPLICATIONS A mutation at the FAP locus may not be a rate-limiting step but may act indirectly by stimulating the proliferation of stem cells. Thus, the development of colon cancer in patients with FAP does not appear to fit the retinoblastoma paradigm.