AT-34CONSTRUCTION OF AN INTEGRATED DIAGNOSTIC ALGORITHM CONSISTING OF CONSENSUS HISTOLOGIC AND MOLECULAR PARAMETERS OF TWO EORTC TRIALS ON ANAPLASTIC GLIOMA

摘要

BACKGROUND: With the rapid discovery of prognostic and predictive molecular parameters for glioma, the status of histopathology in the diagnostic process should be scrutinized. The present project aimed at the construction of an evidence-based diagnostic algorithm for gliomas. To reach this goal, histological features were scrutinized for their prognostic relevance and acceptable inter-observer variability. In addition, molecular markers should be validated for their prognostic power and possible overlap when simultaneously applied. METHODS: The pathology slides of EORTC trials 26951 and 26882 were scanned for virtual microscopy, an electronic scoring form was created and a panel of six pathologists was formed to independently evaluate the slides. The molecular analysis for IDH1, 1p/19q loss; EGFR amplification; loss of chromosome 10 and 10q, gains of chromosome 7 and hypermethylation of the promoter of MGMT were available for appr. half of cases. The slides were divided in a test (discovery) set (n = 500) and a validation set (n = 300). All follow-up data were available at the EORTC data center in Brussels and the results were statistically tested with the aim to create and validate the diagnostic algorithm. RESULTS: In 66% of cases consensus of overall diagnosis was present. Cox regression analysis of consensus histological features and molecular markers revealed the prognostic significance of the parameters necrosis, cell density, mitotic index, loss 1p/19q and IDH1 mutation. A diagnostic algorithm consisting of relevant (consensus and prognostically significant) histological features and molecular markers was identified by CART analysis and consisted of 2 molecular markers and 1 histological feature. The order of prognostic power was: 1p/19q loss; EGFR amplification; astrocytic morphology, and 4 diagnostic nodes were identified. Validation of the nodes in the validation set yielded 4 significantly different survival curves (Overall Wald test p < 0.0001). CONCLUSION: we succeeded in the creation of a timely combined histologic and molecular diagnostic algorithm.