A Targeted Gene Expression Risk Score Predicts Meningioma Outcomes and Responses to Radiotherapy

Purpose/Objective(s) Indications for postoperative radiotherapy after meningioma resection are controversial. DNA methylation profiling identifies meningiomas at risk for recurrence, but logistic and technical barriers have encumbered clinical translation of this approach. The aim of this study was to optimize and validate a clinically-tractable targeted gene expression biomarker to predict meningioma outcomes and responses to postoperative radiotherapy. Materials/Methods Targeted gene expression and Illumina 850k DNA methylation profiling were performed on a discovery cohort of 173 meningiomas (median follow-up 7.8 years) and an external validation cohort of 331 meningiomas (median follow-up 5.8 years) from patients treated with surgery (n = 504) and postoperative radiotherapy (n = 73) at independent, international institutions (70% WHO grade 1, 24% WHO grade 2, 6% WHO grade 3). RNA sequencing was performed on the discovery cohort, and 125 genes were selected for targeted gene expression profiling based on associations with meningioma recurrence. Regularized Cox regression was used to develop continuous gene expression risk score for local freedom from recurrence (LFFR). The model (34 meningioma genes and 7 housekeeping genes) and risk quartiles (low, low-intermediate, high-intermediate, and high) were locked and validated on the external cohort. Multivariate regressions (MVR) incorporating WHO grade, DNA methylation grouping, extent of resection, primary versus recurrent presentation, and postoperative radiotherapy were used to assess the risk score across clinical contexts. Results The gene expression risk score (concordance-index 0.78 ± 0.03) outperformed DNA methylation grouping (0.71 ± 0.03) and WHO grade (0.65 ± 0.03) in stratifying meningioma LFFR in the validation cohort (n = 331). The risk score classified 43% of WHO grade 1 meningiomas as high-intermediate (n = 82) or high risk (n = 34), with 5-year LFFR and overall survival (OS) of 83% and 89% for high-intermediate risk WHO grade 1 meningiomas, and 41% and 73% for high risk WHO grade 1 meningiomas, respectively. Low risk WHO grade 1 meningiomas (n = 50) had 5-year LFFR and OS of 92% and 88%, respectively. High and high-intermediate risk scores were independently prognostic for LFFR (HR 2.9, 95% CI 1.2-4.4) and OS (HR 2.7, 95% CI 1.3-5.6) on MVR. An interaction term between postoperative radiotherapy and risk score was predictive for LFFR (P = 0.046) and OS (P = 0.001) on MVR in the validation cohort, suggesting meningiomas with higher risk scores derived greater benefit from postoperative radiotherapy. Conclusion Here we use targeted gene expression profiling to develop a risk score predicting meningioma outcomes and responses to postoperative radiotherapy. This cost-effective assay outperforms DNA methylation grouping and WHO grade in discriminating meningioma outcomes, and may be useful for guiding clinical trial design.