Observer variation in the pathologic diagnosis of malignant lymphoma in Canada

Biopsies from "poor prognosis" non-Hodgkin's lymphomas were studied to assess diagnostic variations among pathologists. The original diagnosis and the diagnoses of a nationwide panel of local reference pathologists (LRPs) from 104 cases were reviewed and the technical quality of each biopsy evaluated. Using a modified Rappaport classification, there was agreement in 57% of cases among the panel. This increased to 84% when diagnoses were grouped according to Working Formulation grades. Of the 90 cases in which a comparison of the contributing pathologist's diagnosis and the most common panel diagnosis was possible, there was exact agreement in 51% (kappa = 0.311) of cases and agreement regarding WF grade in 68% (kappa = 0.318). Major panel disagreements occurred in 26 (25%) cases, involving the distinction of metastatic anaplastic carcinoma or Hodgkin's disease from non-Hodgkin's lymphoma in 17 cases and follicular versus diffuse architecture in 16 (overlap present). The contributing pathologist had a major disagreement with the panel in 18 cases (20%), of type and frequency similar to those within the panel itself. Poor fixation (complete or partial) was present in 49 cases (47%); this resulted in significantly fewer majority agreements than in the well-fixed group (25% versus 60%, P = 0.020) and also increased rates of disagreement in the diagnosis of lymphoma versus anaplastic carcinoma. Thus improvements in the technical quality of biopsy material could further improve the diagnostic accuracy of lymphoma diagnosis.