Interobserver reliability in the histopathologic diagnosis of oral pre-malignant and malignant lesions

BACKGROUND: The histologic classification of pre-cancerous and cancerous oral lesions has generally shown poor agreement between pathologists, but lesion and patient characteristics that may affect diagnostic reliability have not been explored. METHODS: Eighty-seven clinically suspicious oral lesions biopsied from 81 patients with previous upper aerodigestive tract cancer were independently classified by their local pathologist and a central pathology committee. Interobserver reliability between the local pathologist and the central pathology committee was measured with weighted kappa (kappa(w)) statistics and corresponding 95% confidence intervals (CI). RESULT: The kappa(w) for pathologic diagnosis was 0.59 (95% CI: 0.45, 0.72), and was higher for lesions without inflammation (0.67 (95% CI: 0.53, 0.80) than inflamed lesions (-0.10 (95% CI: -0.27, 0.07)). Greatest agreement was seen for lesions located in the buccal mucosa/vestibule (kappa(w) = 0.68 (95% CI: 0.46, 0.91)) and tongue (kappa(w) = 0.62 (95% CI: 0.40, 0.84)). Least agreement was found for lip/labial mucosa lesions (kappa(w) = -0.04 (95% CI: -0.34, 0.27)). Punch biopsies (kappa(w) = 0.67 (95% CI: 0.54, 0.80)) had greater interobserver reliability than wedge biopsies (kappa(w) = 0.38 (95% CI: 0.12, 0.64)). CONCLUSIONS: These data suggest that the presence of inflammation, lesion site, and biopsy technique modifies the reliability of oral lesion histologic diagnoses.