Abstract
PURPOSE: Response assessment in brain tumors is currently done with the 2D Response Assessment in Neuro-Oncology (RANO) criteria. However, bevacizumab-treated patients may show pseudoresponse on contrast enhanced T1w- (T1w+C), and more infiltrative growth on T2w/FLAIR images at recurrence. We assessed whether in recurrent glioblastoma volumetric methods improve assessment of progressive disease (PD) over the RANO criteria, by correlating PD with overall survival (OS), the golden endpoint in oncology studies. METHODS: Patients with MRI-data (T1w, FLAIR) from the Dutch multicenter BELOB-trial (n=148), treated with lomustine (CCNU) and/or bevacizumab, were included. Patients underwent standardized MRI at 6-week intervals. Presence of PD was determined with RANO and 4 volumetric methods: 1) RANO, 2) enhancing tumor volume (T1w+C), 3) subtracted enhancing volume (T1w+C minus pre-contrast T1w), 4) enhancing (T1w+C) + non-enhancing (FLAIR) volume, and 5) enhancing subtraction + FLAIR volume. Volumetric PD was defined as ≥40% increase in enhancing/subtraction volume, ≥25% increase in FLAIR volume, or new lesions. Central RANO assessment was performed by a neuro-oncologist and 2 neuroradiologists; volumetric analysis by a single reviewer using a semi-automatic method in I-Plan Cranial and manual delineation in MRIcron for subtraction. At 6 and 12 weeks follow-up (FU1, FU2), differences in OS between PD and nonPD per method were assessed with the log-rank test and cox-regression analysis. Hazard ratios (HR) and their 95% confidence intervals (CI) were analyzed. Analyses were performed in all patients together and per treatment group (CCNU or bevacizumab +/- CCNU). Results were considered valid if posthoc power was >0.80. RESULTS: For all patients together at FU1 and FU2 all methods, except method 3 at FU1, showed significant differences in OS between PD and nonPD (p<.001). The largest risk-increase for death in case of PD at FU1 and FU2 was found with method 1: resp. HR 2.81 (95% CI 1.92–4.10) and 2.80 (95% CI 1.75–4.49). There were no significant differences between methods. The CCNU-only group was excluded from further analysis (power <0.80). In the bevacizumab-group, analyses were valid at FU1 for methods 1, 2, 4 and 5, and at FU2 for methods 1 and 4. Both at FU1 and FU2, significant differences in OS were found for all methods. At FU1 the largest HR was found with method 2: 7.21 (95% CI 3.20–16.2). At FU2 the HR for methods 1 and 4 were resp. 2.44 (95% CI 1.46–4.08) and 2.35 (95% CI 1.32–4.19). Again, no significant differences between methods were found. CONCLUSION: Our results indicate that volumetric and subtraction methods are not superior for response evaluation of recurrent glioblastoma compared to RANO, even in patients treated with bevacizumab.