The influence of central review on outcome associations in childhood malignant gliomas: Results from the CCG-945 experience

To examine the influence of the pathology review mechanism on the results of analyses of therapeutic efficacy and biological prognostic correlates for pediatric high-grade gliomas, we evaluated the effects of using single-expert review or consensus review, as alternatives to institutional classification, in determining outcome results of a large randomized trial. The study group was the randomized cohort of Children's Cancer Group study 945, which compared efficacy of 2 chemotherapy regimens adjuvant to surgery and radiation. Trial eligibility required institutional histopathologic diagnosis of high-grade glioma. Sections of study tumors also were centrally reviewed, initially by a study review neuropathologist and subsequently by 5 neuropathologists, including the review pathologist. Reviews were independent, and reviewers were masked to clinical factors and outcomes, and consensus diagnoses of the panel were then established. Among 172 eligible patients, 42 tumors were classified as discordant on single-expert review and 51 on consensus review. Progression-free survival probabilities calculated for patients with tumors classified as high-grade gliomas by either single-expert or consensus review were inferior to those for the overall, institutionally diagnosed cohort. However, conclusions of the study regarding relative efficacy of treatment and clinical and molecular outcome correlates were unaffected by diagnosis method. Resection extent, proliferation index, and p53 expression were associated strongly with outcome, regardless of diagnosis method. However, comparisons between arms in which inclusion was determined by different review criteria for each arm caused spurious conclusions about efficacy differences between treatments. We conclude that the pathology review mechanism had little effect on within-trial comparisons of therapeutic effects or prognostic correlates in this randomized study, but strongly influenced survival distributions that were calculated for each treatment arm. These results support the implementation of expedited central review in therapeutic studies involving childhood malignant gliomas as a way to prospectively identify and exclude cases with discordant diagnoses and indicate the need for additional measures, such as molecular assessments, to increase the reproducibility of neuropathologic classification for these tumors.