A Prognostic Meningioma Gene Expression Signature and Risk Score

Abstract

Meningiomas are the most common primary intracranial tumors and are typically well controlled with surgery and adjuvant radiotherapy. However, a subset of meningiomas recur despite local therapy, leading to significant morbidity and mortality. Clinically tractable biomarkers of meningioma outcomes are lacking. The aim of this study was to identify a prognostic gene expression signature and risk score for meningioma. A discovery cohort of 96 meningiomas that were suitable for targeted gene expression analysis were retrospectively identified from an institutional biorepository. With a median follow-up of 6.4 years, our discovery cohort was enriched for WHO grade II meningiomas (76%) and clinical endpoints of local recurrence (58%), mortality (42%), and disease-specific mortality (49% of deaths). Recurrences were dichotomized based on median time to recurrence (TTR). Targeted expression analysis was performed using a 266 gene expression panel. Prediction analysis for microarrays (PAM) and elastic net regression were used to develop a prognostic gene signature and risk score. Results were validated using gene expression data (GSE58037) from a cohort of 56 meningiomas treated at an independent institution more representative of typical meningioma patients (29% WHO grade II, 20% local recurrence, 18% mortality, median follow-up 5.4 years). Targeted gene expression analysis identified a 36-gene signature achieving an AUC of 0.86 in delineating patients with TTR faster than the median. Consistently, unsupervised hierarchical clustering based on the gene signature demonstrated two robust subgroups of meningiomas with differences in local freedom from recurrence (LFFR, median 0.9 vs 7.8 years, P<0.0001) and overall survival (OS, 4.0 vs 14.4 years, P = 0.0003). A meningioma gene signature risk score between 0 and 1 was generated that strongly correlated with shorter TTR (F-test, P<0.0001) and outperformed WHO grade in stratifying cases by LFFR (concordance-index [c-index] 0.75±0.03 vs 0.57±0.04) and OS (c-index 0.72±0.04 vs 0.59±0.05). On multivariate Cox regression (MVA), the gene risk score, but not WHO grade, was significantly associated with worse LFFR (RR 1.56 per 0.1 increase, 95% CI 1.30-1.90, P<0.0001) and OS (RR 1.32 per 0.1 increase, 1.07-1.64, P = 0.01). Our meningioma risk score was also prognostic for LFFR among WHO grade II meningiomas treated with gross total resection (N = 26, RR 1.72 per 0.1 increase, 1.08-2.86, P = 0.03). In the validation cohort, the gene risk score was correlated with shorter TTR (P = 0.002), outperformed WHO grade in stratifying OS (c-index 0.76±0.11 vs 0.67±0.10), and was associated with worse OS on MVA (RR 1.86 per 0.1 increase, 1.19-2.88, P = 0.005). The prognostic meningioma gene expression risk score presented here may be useful for guiding recommendations for postoperative surveillance or adjuvant therapy after resection.