Emphysema quantification compared with visual assessment on chest ct for predicting airflow obstruction in high-risk patients

SESSION TITLE: Late-Breaking Developments in the Management of Obstructive Lung Disease SESSION TYPE: Original Investigations PRESENTED ON: 10/10/2023 10:30 am - 11:30 am PURPOSE: Chronic Obstructive Pulmonary Disease (COPD) is a significant global health issue, and early detection is crucial. Despite recommendations, COPD remains undiagnosed in patients participating in lung cancer screening. While spirometry is the standard diagnostic tool, CT provides structural change of COPD. Therefore, this study aims to compare the performance of quantitative emphysema analysis and visual CT assessment in predicting airflow obstruction and investigate their association with clinical outcomes. METHODS: This retrospective study identified patients, who underwent non-thoracic surgery between January 2014 to December 2019 at a single tertiary institution. All patients over 60 years underwent spirometry and arterial blood gas analysis before general anesthesia. We included patients with (a) smoking history over 30 pack-years and (b) non-enhanced chest CT and spirometry (<3-month interval). CT scans were classified into seven subtypes by a thoracic radiologist, who was blinded to the results of spirometry: no abnormality, substantial paraseptal emphysema (PSE), bronchial wall thickening, mild centrilobular emphysema (CLE), and moderate to severe CLE (upper predominant, lower predominant, and diffuse), and dichotomized into non-moderate and moderate to severe. The percentage of low attenuated area under -950HU (LAA-950HU) was used for quantification. The primary endpoint was airflow obstruction, and the secondary endpoints were hypoxia (PaO2<80 mmHg) and postoperative respiratory complications. We compared LAA-950HU and CT subtypes using ROC curves and logistic regression analysis. RESULTS: The study included 565 patients (mean age, 70.4±7.0 years, 556 men). Approximately 41.5% (235/565) of patients exhibited airflow obstruction, with half of them being of moderate to severe degree; however, only 11.2% of patients were diagnosed. The most common CT subtype was no abnormality (n=221) followed by bronchial wall thickening (n=130), mild CLE (n=77), substantial PSE (n=73), and moderate to severe CLE, regardless of distribution (n=64). The mean LAA-950HU was 1.5±3.9. The AUCs for predicting airflow obstruction were 0.749 (95% confidence interval, 0.710-0.786) for LAA-950HU, and 0.708 (0.668-0.745) for visual CT subtypes, significantly exceeding the value for dichotomization (AUC=0.609 [0.568-0.650], P<0.001). The results were consistent for predicting moderate to severe airflow obstruction (FEV1<80%), with LAA-950HU exhibiting the highest performance and an AUC value of 0.735 (95% CI:0.696-0.772). Among the radiologic and clincial parameters, LAA-950HU was associated with hypoxia (OR=1.054 [1.007-1.103], P=0.025) and postoperative complications (OR=1.075 [1.020-1.134], P=0.007). CONCLUSIONS: Emphysema quantification can enhance the performance of CT in predicting airflow obstruction in high-risk patients when compared to visual assessment. CLINICAL IMPLICATIONS: Our results emphasize the potential benefits of incorporating the percentage of low attenuated area under -950 Hounsfield units into clinical practice, particularly in the context of lung cancer screening. Emphysema quantification in LDCT was challenging, but deep learning or other kernel conversion methods have enabled emphysema quantification in LDCT comparable to that from standard-dose chest CT. Further studies are necessary to determine the optimal cutoff and validate its use in low-dose CT. DISCLOSURES: No relevant relationships by Min Jae Cha No relevant relationships by Hyewon Choi No relevant relationships by Kang-Mo Gu No relevant relationships by Kang-Mo Gu