Profiles of survival prediction in glioblastoma

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.

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

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.

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.

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.