Diagnostic performance of low-dose computed tomography screening for lung cancer over five years

Low-dose computed tomography (LD-CT) screening can reduce lung cancer mortality; however, it is essential to improve nodule management protocols. We analyze the performance of the diagnostic protocol of the Continuous Observation of SMOking Subjects single-center screening study, after long-term follow-up.

Between 2004 and 2005, 5203 asymptomatic high-risk individuals (≥20 pack-years, aged 50 years or older) were enrolled to undergo annual LD-CT for 5 years. Nodules 5 mm or smaller underwent repeat LD-CT a year later. Nodules larger than 5.0 mm and 8.0 mm or smaller received LD-CT 3 to 6 months later. Nodules larger than 8.0 mm or growing underwent CT-positron emission tomography. True positives were any stage prevalent lung cancer, progressing nodules diagnosed at stage 1, localized multifocal cancer, or new nodules diagnosed at any stage. False negatives were progressing nodules diagnosed at stage >1. False positives were benign nodules resected surgically.

Compliance was 79% over 5 years; 175 primary lung cancers were detected (0.76% per year), 136 (77.7%) were N0M0 and three were interval cancers. Eleven second primary lung cancers were diagnosed. Resectability was 87.4%; postoperative mortality 0.6%. Recall was 6.4% overall, 10.1% at baseline. False negatives were 14 of 175 (8%). Protocol sensitivity was 158 of 175 (90%); specificity 4994 of 5028 (99.4%); positive predictive value was 158 of 187 (84.5%); and negative predictive value was 4994 of 5016 (99.7%). Twenty-nine of 204 (14.2%) benign lesions were diagnosed surgically. Five-year overall and cancer-specific survival were 78% (95% confidence interval, 72-84) and 82% (95% confidence interval, 76%-88%) respectively.

The performance of the CT protocol was satisfactory with an acceptable number of benign lesions biopsied surgically, low recall rate, and good oncological outcomes. However, interval and advanced cancers, and misdiagnoses, need to be reduced, perhaps by risk modeling and use of serum markers.