Profiles of survival prediction in glioblastoma

Abstract

Introduction: Glioblastomas are aggressive brain tumors. While the extent of resection of contrast-enhancing-tumor on MRI correlates with improved survival, the prognostic value of other tumor compartments remains unclear.

Research question: We aimed to find neuroimaging predictive profiles related to postoperative outcomes.Clinical and volumetric data were collected from 214 glioblastoma patients.

Material and methods: Quantitative volumetric measurements were performed on pre-/postoperative MRI using a 3D-semi-automatic-segmentation-software. Volumetric parameters were correlated with clinical outcomes using non-parametric tests and principal component analyses.

Results: Mean overall survival was 16.51 months, mean progression-free survival 12.53 months. Mean contrast-enhancing tumor was 16.60 cm³. Mean total tumor volume was 33.62 cm³, mean FLAIR-volume was 97.88 cm³. Older age was significantly associated with poorer OS and PFS (p = 0.024). MGMT-methylation and chemotherapy were significantly correlated with better outcome (p = 0.003 (OS); p = 0.000 (PFS)/ p = 0.000 (OS); p = 0.002 (PFS)); radiotherapy improved OS (p = 0.013), but not PFS (p = 0.291). None of the volumetric parameters showed a linear correlation with overall or progression-free survival. Classification regression trees were constructed to model OS-subgroups.

Discussion and conclusion: Our regression trees show that clinical factors and volumetric compartments play a significant role in predicting survival in glioblastoma patients. While linear correlations of volumetric parameters and survival may not clearly be identified, a multifactorial individualized approach to surgical management has the potential to benefit patients and facilitate individual treatment decisions.

Keywords: FLAIR; Glioblastoma; IDH-Wildtype; Overall survival; Progression-free survival; Tumor segmentation; Tumor volumetry.

Fig. 1

Depiction of a Spearman correlation for age and survival, showing a significant association of age and logarithmic OS (LOS) as well as logarithmic PFS (LPFS).

Fig. 2

a) Independent samples Mann-Whitney-U-test to test the difference in OS between patients with methylated and unmethylated MGMT status, b) Independent samples Mann-Whitney-U-test to test the influence of radiotherapy on OS.

Fig. 3

Group comparison analysis showing the associations between of OS (in months), methylation status (no MGMT methylation vs. MGMT methylation) and treatments received (chemotherapy: CTx, radiotherapy: RTx) – the lowest survival can be found in the non-methylated group with no treatments received, whereas the highest survival is found in the patient group that is methylated and recived both treatments.

Fig. 4

Principal Component Analysis reveals OS associated with EOR_ET, EOR_Vt, CTx, RTx and MGMT and negatively associated with age and ET_post.

Fig. 5a

CRT OS groups based on image parameters total tumor including T2 FLAIR hyperintensity (VtFLAIR), total tumor (Vt) and EOR of enhancing disease (EOR_ET): less FLAIR volume results in better OS, but outcome can still be improved in large tumors if EOR remains high.

Fig. 5b

Influence of EOR of the initial tumor including FLAIR, initial tumor size and EOR of enhancing tumor on predicted OS: less initial FLAIR results in better outcome, but outcome can be improved even if there is substantial FLAIR, if the residual enhancing disease after resection remains small.

Fig. 5c

CRT including EOR of the tumor volume, as well as pre-op tumor volume and EOR of enhancing disease: in subtotally resected tumors, outcome can be substantially improved when the tumor volume is smaller and a gross total resection is achieved.

Fig. 5d

Influence of EOR of the enhancing disease and initial tumor size on predicted OS: in tumors resected with less than 94.7 %, OS can be significantly better in tumors smaller than 26.1 cm³.