Inter-observer variation between pathologists in diffuse parenchymal lung disease

BACKGROUND: There have been few inter-observer studies of diffuse parenchymal lung disease (DPLD), but the recent ATS/ERS consensus classification provides a basis for such a study. METHODS: A method for categorising numerically the percentage likelihood of these differential diagnoses was developed, and the diagnostic confidence of pathologists using this classification and the reproducibility of their diagnoses were assessed. RESULTS: The overall kappa coefficient of agreement for the first choice diagnosis was 0.38 (n = 133 biopsies), increasing to 0.43 for patients (n = 83) with multiple biopsies. Weighted kappa coefficients of agreement, quantifying the level of probability of individual diagnoses, were moderate to good (mean 0.58, range 0.40-0.75). However, in 18% of biopsy specimens the diagnosis was given with low confidence. Over 50% of inter-observer variation related to the diagnosis of non-specific interstitial pneumonia and, in particular, its distinction from usual interstitial pneumonia. CONCLUSION: These results show that the ATS/ERS classification can be applied reproducibly by pathologists who evaluate DPLD routinely, and support the practice of taking multiple biopsy specimens.