医学
气体分析呼吸
间质性肺病
呼气
队列
内科学
肺
生物标志物
曲线下面积
呼出气冷凝液
接收机工作特性
胃肠病学
队列研究
放射科
病理
成像生物标志物
诊断生物标志物
探索性分析
呼出的空气
高分辨率计算机断层扫描
回顾性队列研究
曲线下面积
呼吸试验
线性判别分析
计算机断层摄影术
呼出的空气
作者
Iris G. van der Sar,Iris A. Simons,R. Düren,Esther J. Nossent,Marlies Wijsenbeek,JanWillem Duitman,Arlette E. Odink,Lilian J. Meijboom,Catharina C. Moor,Paul Brinkman
标识
DOI:10.1088/1752-7163/ae2ef4
摘要
Interstitial lung disease (ILD) encompasses pulmonary disorders characterised by varying degrees of inflammation and/or fibrosis. The presence and extent of these pulmonary abnormalities on (high-resolution) computed tomography (CT) have consequences for diagnosis and treatment; however, inter-observer assessment varies. Analysis of exhaled volatile organic compounds (VOCs) through gas chromatography-mass spectrometry (GC-MS) offers a noninvasive approach to biomarker discovery and pathophysiology understanding. Our study aims to explore the ability of GC-MS-driven exhaled breath analysis to differentiate ILD patients with predominant fibrotic, inflammatory, or a combination of fibrotic and inflammatory pulmonary abnormalities in a training and an external validation cohort. In a multicentre cross-sectional study, patients diagnosed with ILD were recruited. After central review of chest CT scans by independent radiologists, patients were categorised as fibrotic, inflammatory or mixed phenotype group based on the percentage of chest CT scan abnormalities. Breath samples were collected and analysed via GC-MS. Significantly different VOC fragments between groups were selected and used to differentiate groups in the training cohort with sparse partial least squares discriminant analysis. Analyses were validated with patients from an external cohort. 53 patients were included, 21 patients in the fibrotic, 14 in the inflammatory and 18 in the mixed phenotype group. Area under the curve (AUCs) for discrimination between groups ranged from 0.89-1.00 in training cohorts. An attempt to confirm these findings in our external validation cohort resulted in AUCs of 0.63-0.84. Re-evaluation of the training model led to an AUC of 0.78-0.83. This study shows that GC-MS driven exhaled breath analysis towards differentiation of ILD phenotypes is challenging. Current findings emphasise the importance of predefined validation steps during the process of biomarker discovery.
科研通智能强力驱动
Strongly Powered by AbleSci AI