Evaluating Korean Lung-RADS Category 2b: Diagnostic Performance and Interreader Agreement in Lung Cancer Screening.

BACKGROUND: Korean Lung Computed Tomography Screening Reporting and Data System (Lung-RADS) category 2b was introduced to reduce false positives in a tuberculosis-endemic setting. Although its clinical validity has been partly demonstrated, interreader agreement in applying category 2b remains unclear. The aim of this study was to evaluate the diagnostic performance and interreader agreement of the modified classification. METHODS: This retrospective study included 1,371 participants who underwent low-dose chest computed tomography (CT) scans as part of the Korean national lung cancer screening program between September 2019 and January 2024 at a single tertiary hospital. Korean Lung-RADS incorporates category 2b for nodules that are likely benign-such as tuberculous granulomas or intrapulmonary lymph nodes-even if their size corresponds to category 3 or 4. All CT scans were interpreted using Korean Lung-RADS (version 1.0 until 2021, version 1.1 from 2022 through 2023, and version 2022 implemented in 2024). To compare diagnostic performance, all category 2b nodules were reclassified into original Lung-RADS categories based on nodule size and type, and diagnostic metrics were evaluated for both systems. To assess interreader agreement, three thoracic radiologists independently reviewed nodules initially categorized as 2b to determine whether they would also assign category 2b. Diagnostic performance was compared using McNemar and Park's permutation tests, whereas interreader agreement was analyzed using the Fleiss κ statistic. RESULTS: Category 2b nodules were found in 29 of the 1,371 participants (2.1%). Among them, one (3.5%) was diagnosed with lung cancer, whereas the remaining 28 (96.6%) were benign. Interreader agreement for 2b categorization was fair (Fleiss κ = 0.35; 95% confidence interval, 0.17-0.56). In terms of diagnostic performance, Korean Lung-RADS demonstrated significantly higher specificity (92.5% vs. 90.4%, P < 0.001) and comparable sensitivity (89.3% vs. 92.9%, P = 1.00) compared with the original Lung-RADS. CONCLUSION: Korean Lung-RADS category 2b contributes to reducing false positives in lung cancer screening. However, its subjective nature results in fair interreader agreement, underscoring the need for greater standardization to improve consistency in its application.