Concordance between local, institutional, and central pathology review in glioblastoma: Implications for research and practice: A pilot study

Background: There is significant inter-observer variation amongst the neuro-pathologists in the typing, subtyping, and grading of glial neoplasms for diagnosis. Centralized pathology review has been proposed to minimize this inter-observer variation and is now almost mandatory for accrual into multicentric trials. We sought to assess the concordance between neuro-pathologists on histopathological diagnosis of glioblastoma. Materials and Methods: Comparison of local, institutional, and central neuro-oncopathology reporting in a cohort of 34 patients with newly diagnosed supratentorial glioblastoma accrued consecutively at a tertiary-care institution on a prospective trial testing the addition of a new agent to standard chemo-radiation regimen. Results: Concordance was sub-optimal between local histological diagnosis and central review, fair between local diagnosis and institutional review, and good between institutional and central review, with respect to histological typing/subtyping. Twelve (39) of 31 patients with local histological diagnosis had identical tumor type, subtype and grade on central review. Overall agreement was modestly better (52) between local diagnosis and institutional review. In contrast, 28 (83) of 34 patients had completely concordant histopathologic diagnosis between institutional and central review. The inter-observer reliability test showed poor agreement between local and central review (kappa statistic=0.12, 95 confidence interval (CI): -0.03-0.32, P=0.043), but moderate agreement between institutional and central review (kappa statistic=0.51, 95CI: 0.17-0.84, P=0.00003). Agreement between local diagnosis and institutional review was fair. Conclusions: There exists significant inter-observer variation regarding histopathological diagnosis of glioblastoma with significant implications for clinical research and practice. There is a need for more objective, quantitative, robust, and reproducible criteria for better subtyping for accurate diagnosis.