Intraobserver reproducibility in assigning brain tumors to classes in the World Health Organization diagnostic scheme. The Childhood Brain Tumor Consortium.

We studied how reproducibly neuropathologists assign childhood brain tumors to World Health Organization (WHO) diagnostic categories using a subsample from the Childhood Brain Tumor Consortium (CBTC) data. Among the 713 cases, 539 (76%) received the same diagnosis on two different readings by the same team of two neuropathologists. Among the 297 tumors in the supratentorial compartment, 201 (68%) received the same diagnosis on both reviews, whereas among the 392 tumors in the infratentorial compartment, 320 (82%) received the same diagnosis. The eight tumor diagnoses that were the most reliable were craniopharyngioma, germinoma, teratoma, medulloblastoma, pilocytic astrocytoma, ependymoma, subependymal giant cell astrocytoma, and choroid plexus papilloma, Px, the conditional probability of agreement given that one of the reviews resulted in specific diagnosis 'X', ranged from 0.97 to 0.75 for these diagnoses. In contrast, Px for 16 other childhood brain tumor diagnostic categories ranged from 0.60 to 0.00. Pilocytic astrocytoma, ependymoma, medulloblastoma, and astrocytoma not otherwise specified (nos) had significantly lower Px values (p less than 0.05) when these tumors were located in the supratentorial compartment than when below the tentorium. Px was not significantly lower for any category when tumors were located in the infratentorial compartment. The array of substituted diagnoses that lowered reproducibility was greater when tumors were located supratentorially. To improve the definitions, eliminate overlapping diagnostic categories, and sharpen the fuzzy boundaries that contribute substantially to limited reproducibility, we suggest: (1) the categories of astrocytoma nos, fibrillary astrocytoma, and protoplasmic astrocytoma be collapsed into a single category of astrocytoma; (2) the diagnostic category of desmoplastic medulloblastoma be combined with medulloblastoma; and (3) the criteria for anaplasia should be further refined to include quantification of critical histologic features, e.g., agreed upon operational definitions for amount of cell density, number of mitoses and pleomorphism for anaplastic astrocytoma and anaplastic ependymoma.