The change in excess risk of lung cancer attributable to smoking following smoking cessation: an examination of different analytic approaches using CPS-I data.

Background: Lung cancer risk is modified by smoking cessation. However, the inclusion in the group of former smokers of those who quit after developing symptoms or being diagnosed with lung cancer distorts the apparent risk in the first several years following cessation. This bias is termed the quitting ill effect.
Methods: Lung cancer mortality data from the American Cancer Society's CPS-I were used to calculate the excess mortality among white male former smokers compared to the predicted risk had those individuals continued to smoke. Alternate approaches to minimizing the quitting ill bias were investigated. Goodness-of-fit of the models was assessed graphically and formally.
Results: Poisson models were built for the absolute lung cancer risk for never smokers and the excess risk, over never smokers, for continuing smokers. The decrease in excess risk in former smokers was modeled by a negative exponential function. The models for the three smoker subgroups (continuing, never, and former), all fit the data well. Assuming that the fraction of excess risk remaining for former smokers does not decline for the first two years following cessation and that the quitting ill effect does not influence those who are five or more years post-cessation allowed a reasonable estimation of the change in risk of lung cancer with increasing duration of abstinence.
Conclusions: The reduction in the excess risk of lung cancer in former smokers can be estimated, and the quitting ill effect minimized, by the inclusion of a lag between cessation and onset of reduction in risk.