Diagnostic Yield vs Diagnostic Accuracy for Peripheral Lung Biopsy Evaluation: Evidence Supporting a Future Pragmatic End Point

Abstract

Diagnostic yield and accuracy endpoints have been used inconsistently in the evaluation of advanced diagnostic bronchoscopy devices and techniques, limiting between-study comparisons. In addition, diagnostic accuracy can only be adjudicated after prolonged clinical follow-up which delays reporting on the performance of novel devices. Will a conservative diagnostic yield definition would result in few false negative initial results to closely approximate diagnostic accuracy and represent a useful outcome for future studies of diagnostic utility? Commonly utilized definitions of diagnostic yield were applied to a prospective dataset of consecutive peripheral pulmonary lesions sampled by navigational bronchoscopy from 2017 to 2019. All consider malignancy to be diagnostic but differ in their classification of non-malignant biopsy findings, which were subcategorized as specific benign, nonspecific benign, or normal lung. Diagnostic yield calculations were also compared to diagnostic accuracy, defined as the proportion of biopsies deemed diagnostic by each definition that were confirmed accurate through two years of follow-up. A total of 450 biopsied lesions were analyzed. The prevalence of malignancy was 60.9% (274 of 450). On initial bronchoscopy pathology, there were 227 malignant diagnoses (50.4%), with a single false-positive (0.4%). Among 104 biopsies with specific benign findings, only two were false negative for malignancy (1.9%). There were 119 nonspecific benign biopsies with 46 false negatives for malignancy (38.7%). The discrepancy between diagnostic yield and accuracy was 0.7% for the conservative definition which only considered malignant or specific benign findings as diagnostic. A conservative diagnostic yield definition excluding nonspecific benign diagnoses closely approximated diagnostic accuracy through two years follow-up, with a less than 1% discrepancy. Using this conservative yield definition may allow for dissemination of reliable diagnostic utility data without protracted delays needed for follow-up data in this era of rapid technological change in advanced diagnostic bronchoscopy.