Prognostic biomarkers for progression-free survival in glioblastoma multiforme: quantitative imaging and gene expressions

Research output: Contribution to journalMeeting Abstract

Abstract

Prognostic biomarkers at initial diagnosis of untreated Glioblastoma Multiforme (GBM) are an important step towards development of predictive biomarkers in this patient pool for personalized medicine. We hypothesize that quantitative imaging biomarkers are independent to gene expression biomarkers in progression-free-survival (PFS) prognosis and overall survival (OS). It is known that patient with large strong enhancing tumor have a shorter PFS and OS. Weak enhancing tumor, which is believed to be more infiltrative in nature, has rarely been studied due to the difficulties in identifying them visually. Method: We developed an automated image analysis technique to quantify weak enhancing tumor, strongly enhancing tumor, edema and necrosis from a single source site (n = 25) of TCIA with gene expression from TCGA. We define tumor volume (in cm3) of Zone1A as the enhancing tumor plus necrosis; Zone1B as necrosis plus weak enhancing tumor; and Zone2 as total tumor load of necrosis, strong enhancing and weak enhancing tumor. Cox proportional hazard model was used to determine prognostic significance of imaging parameters to PFS and OS. Level 1 gene expression data from Affymetrix platform were processed in R. We filter the gene probe list to 7277 probes (31.9%) based on biological variability and take the probes that are >2-fold upregulated compared to normal control (n = 10) (t-test, p<0.05, multiple comparison corrected), resulted to 167 probes. Gene probes prognostic to PFS are selected based on unadjusted log-rank test (p<0.05). Result: There are 19 matched cases with imaging and gene expression with one living patient. MGMT is methylated in 4 patients and unmethylated in 10 patients. All cases are non-G-CIMP and IDH1 WT or unknown. The age and karnofsky score (kps) distribution are 55±15 yo and 85±11. Tumor size measurements are strongly prognostic for PFS adjusted by age and kps: Zone1A (HR = 1.04 [1.01, 1.07], p = 0.0137, log-rank p = 0.063), Zone1B (HR = 1.08 [1.03, 1.13], p = 0.0014, log-rank p = 0.00020), Zone 2 (HR = 1.037 [1.012 1.063], p = 0.0034, log-rank p = 0.011). Only 18 gene probes are prognostic to PFS, which are CASP4, CD44, COL5A1, COL6A2, HCP5, IL10RB, IL15, LOXL2, LY75, PTX3, SERPINEH1, SLC16A3, STC1, TNC. These genes are involved in ECM-receptor interaction, focal adhesion and cytokine mediated inflammation pathway. After adjusting for age and kps, only CD44, COL5A1, COL6A2, LOXL2, IL10RB, STC1 are significant based on log-rank test while the HR of age and kps are not significant. Adjusting for age, kps and Zone1B resulted to log-rank significance of CASP4, LOXL2, TNC, IL10RB, IL15, PTX3, STC1, HCP5, SLC16A3, with HR significance of Zone1B, gene, age and sometimes kps. The mean HR of Zone1B is 1.085±0.013 and the maximum log-rank p = 0.0002567 for all these genes. Conclusion: Total tumor load of necrosis and weak enhancing tumor is a strong independent prognostic biomarker of PFS in addition to gene, age and kps.
Original languageEnglish (US)
Article number5281
JournalCancer research
Volume75
DOIs
StatePublished - 2015

Fingerprint Dive into the research topics of 'Prognostic biomarkers for progression-free survival in glioblastoma multiforme: quantitative imaging and gene expressions'. Together they form a unique fingerprint.

Cite this