Malignant mesothelioma of the pleura: interobserver variability.

AIMS To assess the consistency of histopathological diagnosis of pleural malignant mesothelioma by estimating interobserver agreement between five pathologists. METHODS Eighty eight histological sets pertaining to original diagnoses of pleural malignant mesothelioma were reviewed separately by each pathologist. Diagnostic likelihood was graded as definite (A), probable (B), possible (C), improbable (D), and definitely not (E) malignant mesothelioma. The following indexes were estimated: observed proportion of agreement (Po), kappa statistics and proportion of agreement for "positive" (Ppos) and "negative" (Pneg) diagnoses. RESULTS Sixty cases (68.2%) were rated by at least three reviewers as A or B and by none of the others as D or E. Five (5.7%) were rated by at least two reviewers as D or E and by none of the others as A or B. Nine (10.2%) showed a serious disagreement, rated A or B and D or E. Agreement for sets obtained at necropsy/surgery (median kappa w = 0.57) was similar to that for endoscopic material (median kappa w = 0.54). Agreement was poor on material obtained by needle biopsy (median kappa w = 0.21). The median value of Ppos varied between 0.94 (necropsy/surgery) and 0.67 (needle biopsy) and that of Pneg between 0.78 (necropsy/surgery) and 0.34 (unspecified biopsy). Interobserver agreement on histological typing was good overall (median kappa = 0.59). CONCLUSIONS Of the original histological diagnoses, 70% were consistently reproduced through panel review. Most indexes of agreement between pathologists ranged from poor (needle biopsy) to moderate (necropsy/surgery). Agreement in confirming malignant mesothelioma was greater than that regarding exclusion of this diagnosis. Of the cases finally considered to have malignant mesothelioma, the reproducibility of histological typing was relatively satisfactory.