To determine the influence of interobserver variability in histopathologic reporting of bronchial biopsy specimens on the results of autofluorescence bronchoscopy in detection of preinvasive bronchial neoplasia, we compared original histopathologic reports with reports of a review panel. For both observers, the results of the histologic examination were compared with the bronchoscopic findings. Sensitivity, specificity, and predictive values for detection of preinvasive lesions were calculated. Kappa statistics were used to assess interobserver agreement. A total of 343 biopsies were performed during 74 bronchoscopies in 59 patients. The original observer reported 45 preinvasive lesions, but the review observer reported only 22 of them. Kappa values for moderate dysplasia, severe dysplasia, and carcinoma in situ were, respectively, 0.02, 0.46, and 0.57. Using the original histopathologic reports or the review reports, the results of autofluorescence bronchoscopy combined with conventional bronchoscopy for detection of preinvasive lesions were, respectively, as follows: sensitivity 78% and 95%; specificity 59% and 60%; positive predictive values of 25% and 18%; and negative predictive values of 94% and 99%. There is a considerable interobserver variability in the histopathologic reporting of preinvasive bronchial neoplasia that may have important implications for any study or treatment that relies on the histopathologic diagnosis of these lesions.
Journal of Bronchology7:210-214, 2000.