Clinical and analytical validation of a targeted gene expression biomarker predicting meningioma outcomes and radiotherapy responses

2009 Background: Surgery is the mainstay of treatment for meningioma, the most common primary intracranial tumor, but improvements in meningioma risk stratification are needed and current indications for postoperative radiotherapy are controversial. Here we report multicenter clinical and analytical validation of a gene expression biomarker using 1856 retrospective and prospective meningiomas from 12 institutions across 3 continents. Our results reveal the gene expression biomarker provides additional information for meningioma outcomes compared to other classification systems, including prediction of radiotherapy responses. Methods: Gene expression profiling and regularized Cox regression was performed on a discovery cohort of 173 meningiomas to generate a 34-gene expression biomarker and continuous risk score for local recurrence. The model and thresholds for low, intermediate, and high-risk scores were locked and applied to 3 validation cohorts: a multicenter retrospective clinical validation cohort (N = 6 centers, N = 866 meningiomas, median follow-up 5.2 years), a prospective investigator-blinded clinical validation cohort from RTOG 0539 (N = 103 meningiomas, median follow-up 8.4 years), and an analytical validation set for test/re-test and FFPE/frozen analyses, as well as comparison across different platforms for gene expression quantification (N = 8 centers, N = 1219 meningiomas). Gene expression biomarker performance was compared to 9 contemporary molecular classification systems. Favorable versus unfavorable meningiomas were defined as gene expression low risk with any resection, or gene expression intermediate risk with gross-total resection (favorable), versus gene expression intermediate risk with subtotal resection, or gene expression high risk with any resection (unfavorable). Results: The biomarker outperformed WHO grade (5-year local freedom from recurrence [LFFR] delta-AUC 0.11, 95% CI 0.07-0.17, p < 0.001) and all other classification systems for LFFR and OS on multivariate analysis, achieving a negative predictive value for recurrence at 5-years of 91.9%. The biomarker was predictive for LFFR after postoperative radiotherapy, with a hazard ratio of 0.33 for unfavorable primary WHO grade 2 meningiomas (95% CI 0.14-0.76, P = 0.009), and 0.54 for unfavorable propensity-matched meningiomas across all WHO grades (95% CI 0.37-0.78, p = 0.0001). The biomarker reclassified 39.8% of prospectively collected meningiomas from RTOG 0539, including downstaging 30.5% of patients who received postoperative radiotherapy, and was independently prognostic for overall survival on multivariate analysis. Conclusions: A targeted gene expression biomarker improves discrimination of meningioma outcomes compared to recent classification systems and predicts postoperative radiotherapy responses.