Martin Tammemagi

DVM, MSc, PhD

Brock University

Professor

NLST (Learn more about this study)

NLST-25

Jul 22, 2013

Do repeat normal lung screens lower model estimated baseline estimates of lung cancer risk?

It has been shown that enrolment into lung cancer screening programs using lung cancer risk based on accurate prediction models is more efficient than using National Lung Screening Trial entry criteria, and it is likely that if lung cancer screening programs are widely adopted in high-risk individuals in the future, enrolment will be based on model predicted elevated risk. It has been proposed that several repeat normal lung screens may suggest reduced lung cancer risk and may indicate that less than annual screening might be required for such individuals. Lung cancer risk prediction models, such as the Tammemagi PLCOm2012 model (NEJM 2013), provide lung cancer risk estimates at baseline excluding any subsequent lung cancer screening outcome data. Let's consider individuals with similar baseline lung cancer risks, but the one group has three consecutive negative screens and at the other extreme are individuals with three abnormal suspicious for lung cancer screens which turn out not diagnosed to be lung cancer within one year of the abnormal screen. In subsequent follow-up from T3 to T9 is lung cancer risk significantly lower in individuals with normal screens compared to those with abnormal suspicious for lung cancer screens. In this example, we contrasted clear cut opposites. If there is a real relationship, it is more likely to exist in a dose-response relationship. Our analytic plan would assess dose-response relationships.