Clinical and Molecular Recursive Partitioning Analysis of High-grade Glioma Treated With IMRT

Despite multimodal treatment for high-grade gliomas, prognosis remains grim. Prior Radiation Therapy Oncology Group-Recursive Partitioning Analysis (RTOG-RPA) reports indicate based on pretreatment and treatment-related factors, a subset of patients experience a significantly improved survival. Since the development of the RTOG-RPA, high-grade gliomas have seen the widespread introduction of temozolomide and tumor oncogenetics. Here we aimed to determine whether the RTOG-RPA retained prognostic significance in the context of modern treatment, as well as generate an updated RPA incorporating both clinical and genetic variables.

Patients with histologically proven glioblastoma, gliosarcoma, anaplastic astrocytoma, and anaplastic oligodendroglioma treated with intensity-modulated radiation therapy (IMRT) between 2004 and 2017 were reviewed. The primary endpoint was overall survival from date of diagnosis. Primary analysis compared actual survival rates to that expected of corresponding RTOG-RPA class. Secondary analysis utilized the rpart function to recursively partition overall survival by numerous clinical and genetic pretreatment and treatment-related variables. A tertiary analysis recursively partitioned a subset of patients in which the status of all genetic markers were known.

We identified 878 patients with histologically proven high-grade glioma treated with IMRT and 291 patients in our genetic subset. Median overall survival for the entire cohort was 14.2 months (95% confidence interval, 13.1-15.3). Applying the RTOG-RPA to our cohort validated the relative prognostic ordering of the survival classes except class II. Generating our new RPA created 7 significantly different survival classes (P<0.001, χ=584) with median survival ranging from 96.4 to 2.9 months based on age, histology, O6-methylguanine-DNA methyltransferase methylation status, radiation fractions, tumor location, radiation dose, temozolomide status, and resection status. Our second RPA of our genetic subset generated 5 significantly different survival classes (P<0.001, χ=166) with survival ranging from 65.3 to 5.6 months based on age, isocitrate dehydrogenase 1 mutation status, O6-methylguanine-DNA methyltransferase methylation status, neurological functional classification, hospitalization during IMRT, temozolomide status, and Karnofsky performance status.

The RTOG-RPA retains partial prognostic significance, however, should be updated to reflect recent advancements. This series represents a large RPA analyzing both clinical and genetic factors and generated 7 distinct survival classes. Further assessment of patients with fully available genetic markers generated 5 distinct survival classes. These survival classifications need to be validated by a prospective data set and compared against the RTOG-RPA to determine whether they provide improved prognostic power.