Session Item

Local treatment modalities, i.e. surgery and radiotherapy (RT), are the cornerstone in the treatment for glioblastoma (GBM). Radiation dose-escalation is a logical next step to improve local control and survival, but results are so far unsatisfactory in unselected patients. We hypothesise that patients with a high risk of local-only disease, may potentially benefit most from intensified local treatment.

The aim of this retrospective single-centre longitudinal MRI study was therefore, to evaluate and predict the occurrence of local-only disease.

All adults with newly diagnosed, histology confirmed GBM between 2014 and 2019 were identified via the national registry. Only patients who were planned for postoperative, conventional chemoradiotherapy (≥54 Gy) at our institute were selected. RT was given according to ESTRO-ACROP guidelines, including 3-monthly MRI-scan at follow-up. Treatment response was assessed using RANO criteria.

To assess local-only disease, patients were scored at first progression as: local-only (overlap with GTV), non-local (new contrast-enhancing lesions without overlap with GTV), or combined progression.

Time to progression and OS were estimated using the Kaplan–Meier method. Multivariable logistic regression was used to analyse the association of tumour focality, MGMT-status, and extent of surgery (EOS), with type of progression. This was furthermore used to develop a prediction model for type of progression, which was validated using a random split-sample (0.75:0.25).

We identified 256 eligible patients, and excluded 41 (no planning MRI) and 20 (no follow-up MRI), resulting in a study population of 195 patients (table). Of those 171 patients had radiologic progression. Progression was local-only in 56%, non-local in 15%, and combined in 16% of patients. Beyond first progression, combined progression increased to 31%, non-local remained stable (15%), and local-only decreased to 42%.

Median (95% CI) time to progression/OS was 7.0 (5.8-8.2)/18.0 (15.1-20.9) months in the local-only, 12.0 (10.9-13.2)/24.0 (16.3-31.7) in the non-local, and 7.0 (4.3-9.8)/12.0 (9.8-14.2) in the combined group, respectively.

In multivariable analysis, MGMT-status and EOS were significantly associated with type of progression (p-value 0.023 and 0.024, respectively). The obtained prediction model (AUC 0.71) contained these variables and tumour focality as this was regarded as an important potential confounder (figure). The probability of having local-only progression varied between 20 and 89% for the respective subpopulations. All patients with unifocal disease, and those with multifocal disease who had undergone a partial or complete resection, had the highest probability of local-only progression, regardless of MGMT-status.

Our results suggest that all patients with unifocal disease, and those with multifocal disease who have undergone a partial or complete resection, are candidates for radiation dose-escalation.