Threshold models in radiation carcinogenesis.

Cancer incidence and mortality data from the atomic bomb survivors cohort has been analyzed to allow for the possibility of a threshold dose response. The same dose-response models as used in the original papers were fit to the data. The estimated cancer incidence from the fitted models over-predicted the observed cancer incidence in the lowest exposure group. This is consistent with a threshold or non-linear dose-response at low-doses. Thresholds were added to the dose-response models and the range of possible thresholds is shown for both solid tumor cancers as well as the different leukemia types. This analysis suggests that the A-bomb cancer incidence data agree more with a threshold or non-linear dose-response model than a purely linear model although the linear model is statistically equivalent. This observation is not found with the mortality data. For both the incidence data and the mortality data the addition of a threshold term significantly improves the fit to the linear or linear-quadratic dose response for both total leukemias and also for the leukemia subtypes of ALL, AML, and CML.