The prevalence of patients with bronchiectasis (BE) has been rising in recent years, which increases the substantial burden on the family and society. Exploring a convenient, effective, and low-cost screening tool for the diagnosis of BE is urgent. We expect to identify the accuracy (ACC) of breath biomarkers (BBs) for the diagnosis of BE through breathomics testing and explore the association between BBs and clinical features of BE. Exhaled breath samples were collected and detected by high-pressure photon ionization time-of-flight mass spectrometry in a cross-sectional study. Exhaled breath samples were from 215 patients with BE and 295 control individuals. The potential BBs were selected via the machine learning (ML) method. The overall performance was assessed for the BBs-based BE detection model. The significant BBs between different subgroups such as the severity of BE, acute or stable stage, combined with hemoptysis or not, with or without nontuberculous mycobacterium (NTM),P. aeruginosa(P.a) isolation or not, and the BBs related to the number of involved lung lobes and lung function were discovered and analyzed. The top ten BBs based ML model achieved an area under the curve of 0.940, sensitivity of 90.7%, specificity of 85%, and ACC of 87.4% in BE diagnosis. Except for the top ten BBs, other BBs were found also related to the severity, acute/stable status, hemoptysis or not, NTM infection,P.aisolation, the number of involved lobes, and three lung functional parameters in BE patients. BBs-based BE detection model showed good ACC for diagnosis. BBs have a close relationship with the clinical features of BE. The breath test method may provide a new strategy for BE screening and personalized management.
Read full abstract