Site of first treatment failure after standard-of-care for newly diagnosed glioblastoma in a Danish cohort

Background: The aim of this nationwide study was to identify predictive factors for isolated local, non-local, and combined treatment failure in newly diagnosed glioblastoma patients. 

Methods: All adults with newly diagnosed glioblastoma in Denmark between 2014 and 2019 who were planned for long-course (chemo)radiotherapy were included. The MRI scan of first progression following RANO criteria was used to assess the site of failure. Multivariable multinomial logistic regression was applied to analyze the association between potential risk factors and the site of failure. Prediction models were derived. 

Results: We included 939 patients, of whom 758 had radiographic progression. Of those, 525 patients (69%) had isolated local failure, 114 (15%) non-local, and 119 (16%) combined. In patients with an MGMT promotor methylated (300) versus unmethylated (398) tumor, local failure was 66 versus 73%, non-local 21 versus 10%, and combined 12 versus 17%, respectively (P-value <.001). In multivariable analysis, MGMT promotor methylation status and tumor location were significantly associated with the failure site. In a model using MGMT promotor methylation status, tumor location, focality, extent of resection, and WHO performance status (poorly calibrated, C-index 0.63), the predicted probability (95% CI) of local failure varied from 0.47 (0.24 to 0.70) to 0.85 (0.74 to 0.97) between respective subgroups. That of non-local varied from 0.03 (0.00 to 0.07) to 0.36 (0.15 to 0.57) and combined from 0.05 (0.03 to 0.14) to 0.35 (0.05 to 0.65). 

Conclusions: It was not feasible to identify strong predictors for the site of first treatment failure, and hence, to derive an accurate prediction model. Isolated local failure was dominant in both patients with an MGMT promotor methylated and unmethylated tumor.