NIMG-44. QUANTITATIVE MULTI-PARAMETRIC IMAGE PROFILING REVEALS REMARKABLE HETEROGENEITY WITHIN IDH-WILDTYPE GLIOBLASTOMA, OFFERING PROGNOSTIC STRATIFICATION BEYOND CURRENT WHO CLASSIFICATIONS. (5th November 2018)

• Record Type: Journal Article
• Authors: Rathore, Saima; Akbari, Hamed; Nasrallah, MacLean; Bakas, Spyridon; Binder, Zev; Rozycki, Martin; Min Ha, Sung; Mamourian, Elizabeth; Bilello, Michel; O'Rourke, Donald; Davatzikos, Christos
• Is Part Of: Neuro-oncology. Volume 20(2018)Supplement 6
• Page Start: vi186
• Page End: vi186
• Publication Date: 2018-11-05
• Subjects: Brain Neoplasms -- Periodicals; Brain -- Tumors -- Periodicals; Brain -- Cancer -- Periodicals; Nervous system -- Cancer -- Periodicals; 616.99481
• DOI: 10.1093/neuonc/noy148.770
• Languages: English
• ISSNs: 1522-8517

Abstract

Abstract: PURPOSE: The current WHO classifies astrocytomas by IDH mutational status given the significantly poorer prognosis of IDH-wildtype tumors, representing ~95% of de novo glioblastoma. Our previous studies revealed remarkable heterogeneity of these tumors, dividing them in three distinct radiographic subtypes (Rad-S). In this study, we hypothesize that this heterogeneity expands in Rad-S within IDH-wildtype glioblastoma, subdividing them further according to prognosis. METHODS: We analyzed pre-operative multi-parametric magnetic resonance imaging (mpMRI) data (T1, T1-Gd, T2, T2-FLAIR, DTI, DSC) of a retrospective cohort of pathology-proven de novo IDH-wildtype glioblastoma (n=76). Comprehensive quantitative imaging phenomic (QIP) features were extracted from distinct cancerous sub-regions (enhancing, non-enhancing, edematous), using the Cancer Imaging Phenomics Toolkit (CaPTk-www.cbica.upenn.edu/captk). QIP features comprised intensity histogram, volumetric, morphological, statistical, and textural descriptors. Unsupervised clustering of these features alone revealed tumor Rad-S, based on unambiguous clustering assignments across 1000 permutations, that were evaluated through survival and molecular characteristics. RESULTS: Three Rad-S were identified within IDH-wildtype glioblastoma, with statistically significant survival differences (long-:intermediate-:short-survival, median(months)=19.4:12.3:7.0, distribution=19.7%:34.2%:46.1%) measured by Kaplan-Meier analysis (P<0.001, log-rank) and Cox-Model (hazard-ratio=3.21, 95%CI:2.51–4.61). Rad-S correlate with survival independent of age, resection-status, post-surgical therapy, additional genetic alterations, and MGMT promoter methylation status. Importantly, long-survival Rad-S, compared to others, showed statistically significant (P<0.001, Kruskal-Wallis) central hypo-intense non-enhancing region surrounded by hyper-intense rim (T1-Gd), lower angiogenesis/neovascularization (DSC) and cell-density (DTI), and higher water concentration (T2). CONCLUSIONS: Quantitative analysis of mpMRI yields three distinct Rad-S within IDH-wildtype glioblastoma offering complementary stratification beyond current WHO classification, which is independent of any factor known to affect prognosis. These Rad-S provide an additional prognostic indicator as a component of precision diagnostics that may impact choice/timing of surgery, chemotherapy, bevacizumab and radiation, allowing personalized treatment. Further, our current understanding in clinical setting is insufficient to explain prognostic differences among the Rad-S. These results provide guidance for ongoing investigation to elucidate pathologic mechanism and consequently targeted therapeutic strategies.