Inter- and intraobserver reliability and validation of a new method for determination of eosinophil counts in patients with esophageal eosinophilia.

BACKGROUND: Diagnosis of eosinophilic esophagitis (EoE) requires quantification of esophageal eosinophilia. AIMS: The aims of this study were to assess inter- and intraobserver reliability for measuring esophageal eosinophil counts and to validate a novel method of determining tissue eosinophil density using digitized histopathology slides.
METHODS: Patients were selected from the University of North Carolina EoE clinicopathologic database. Glass slides were de-identified and scanned to create digitized slides. Using a set protocol, 40 slides were read by each of three pathologists for interobserver measures, and were also reread by one pathologist as traditional glass slides. Different sets of 20 unique slides were read twice by each pathologist for intraobserver measures. Correlation and agreement were calculated with Pearson's rho and the kappa statistic.
RESULTS: There was excellent correction between digitized images and glass slides (r = 0.91-0.95, P < 0.001). For maximum eosinophil densities, interobserver correlations were 0.91, 0.76, and 0.79. For mean densities, interobserver correlations were 0.90, 0.89, and 0.85. Intraobserver correlations for maximum densities were 0.99, 0.94, and 0.96, and for mean densities were 0.97, 0.87, and 0.89 (P < 0.001 for all correlations). Agreement was in the "substantial" to "near-perfect" range for pathologists using several diagnostic cut-points for EoE.
CONCLUSIONS: Both inter- and intraobserver correlations were excellent for determining eosinophil densities and counts. A method of using digitized slides was valid when compared with traditional glass slides. This protocol could be adopted for research and clinical purposes to further standardize the diagnostic process for EoE.