Evaluating diagnostic yield and accuracy as key performance metrics in pulmonary lung lesions

A conservative definition of diagnostic yields for assessing the performance of guided bronchoscopy has been proposed, but it has yet to be validated in practice.

Patients who underwent radial endobronchial ultrasound (R-EBUS) between April 2020 and April 2023 were included in the study. Diagnostic results were classified as malignant or non-malignant based on the post-lung-biopsy pathology. Non-malignant results were further categorized into specific benign (SB), nonspecific benign (NSB), atypical cells, and non-diagnostic (ND). All non-malignant lesions were confirmed using alternative biopsy methods or chest computed tomography (CT) during a follow-up of over 1 year. Diagnostic yield and accuracy were calculated using pre-defined methods (Box below). Predictors of sampling success were identified in a logistic regression analysis.

Among the 736 patients evaluated in this study, R-EBUS-guided TBLB revealed malignancy in 431 (58.6%) patients. The remaining 305 (41.4%) patients with non-malignant lesions were classified as SB (8.3%), NSB (21.3%), atypia (4.6%), and ND (7.2%). Diagnostic yield vs. accuracy values based on conservative, intermediate, and liberal definitions were 67% vs. 67, 88% vs. 77, and 100% vs. 79%, respectively. Thus, for the conservative definition, diagnostic accuracy and diagnostic yield were identical. Significant predictive factors for successful lung biopsy according to the conservative diagnostic yield included lesion size (> 20 mm), CT-bronchus subclassification (Ia, Ib), and radial probe position within the lesion.

Our study validated the use of the conservative definition of diagnostic yield as a reliable diagnostic endpoint for evaluating the performance of guided bronchoscopy. This definition could serve as a time-saving standard in prospective studies comparing the diagnostic effectiveness of various navigation devices.