Phase II study of bevacizumab, temozolomide, and hypofractionated stereotactic radiotherapy for newly diagnosed glioblastoma.

PURPOSE: Bevacizumab is associated with decreased vascular permeability that allows for more aggressive radiotherapy schedules. We conducted a phase II trial in newly diagnosed glioblastoma utilizing a novel hypofractionated stereotactic radiotherapy (HFSRT) schedule combined with temozolomide and bevacizumab.

EXPERIMENTAL DESIGN: Patients with tumor volume ≤60 cc were treated with HFSRT (6 × 6 Gy to contrast enhancement and 6 × 4 Gy to FLAIR hyperintensity with dose painting) combined with concomitant/adjuvant temozolomide and bevacizumab at standard doses. Primary endpoint was 1-year overall survival (OS): promising = 70%; nonpromising = 50%; α = 0.1; β = 0.1.

RESULTS: Forty patients were enrolled (median age: 55 years; methylated MGMT promoter: 23%; unmethylated: 70%). The 1-year OS was 93% [95% confidence interval (CI), 84-100] and median OS was 19 months. The median PFS was 10 months, with no pseudo-progression observed. The objective response rate (ORR) was 57%. Analysis of The Cancer Genome Atlas glioblastoma transcriptional subclasses (Nanostring assay) suggested patients with a proneural phenotype (26%) fared worse (ORR = 14%, vs. 77% for other subclasses; P = 0.009). Dynamic susceptibility-contrast perfusion MRI showed marked decreases in relative cerebral blood volume over time (P < 0.0001) but had no prognostic value, whereas higher baseline apparent diffusion coefficient (ADC) ratios and persistent hypermetabolism at the 6-month FDG-PET predicted poor OS (P = 0.05 and 0.0001, respectively). Quality-of-life (FACT-BR-4) and neuropsychological test scores were stable over time, although some domains displayed transient decreases following HFSRT.

CONCLUSIONS: This aggressive radiotherapy schedule was safe and more convenient for patients, achieving an OS that is comparable with historical controls. Analysis of advanced neuroimaging parameters suggests ADC and FDG-PET as potentially useful biomarkers, whereas tissue correlatives uncovered the poor prognosis associated with the proneural signature in non-IDH-1-mutated glioblastoma.