Modified Lung-RADS Improves Performance of Screening LDCT in a Population with High Prevalence of Non–smoking-related Lung Cancer

We proposed a modification of the ACR Lung Imaging Reporting and Data System (Lung-RADS) to clarify the characteristics of subsolid nodules with categories 1–11, and to compare the diagnostic accuracy with Lung-RADS and National Lung Screening Trial criteria in an Asian population with high prevalence of adenocarcinoma. We analyzed a retrospective cohort of 1978 consecutive healthy subjects (72.8% nonsmoker) who underwent low-dose computed tomography from August 2013 to October 2014 (1084 men, 894 women). Lung-RADS categories 2 and 3 were modified to include subcategories of 2A/2B/2C and 3A/3B/3C, respectively. Clinical information and nodule characteristics were recorded. Receiver operating characteristic curves were used to compare diagnostic accuracy at different cutoffs. Thirty-two subjects (30 nonsmokers) had pathology-proven adenocarcinoma spectrum lesions in the follow-up period (1.6 ± 0.5 years). Modified Lung-RADS, using modified Lung-RADS category 2C as cutoff, had an area under the curve (AUC) of 0.973 in predicting adenocarcinoma spectrum lesions (sensitivity of 100%, specificity of 89.3%), which was significantly higher than that of Lung-RADS (AUC = 0.815, P