Development and validation of an individualized predictor of meningioma recurrence: A multicenter retrospective cohort study
Nassiri F, Mamatjan Y, Suppiah S, Badhiwala J, Mansouri S, Karimi S, Poisson L, Noushmehr H, Harter P, Baumgarten P, Weller M, Preusser M, Herold-mende C, Sahm F, Von Deimling A, Aldape K, and Zadeh G. Development and validation of an individualized predictor of meningioma recurrence: A multicenter retrospective cohort study. J Neurol Surg Part B Skull Base 2019; 80.
J Neurol Surg Part B Skull Base
Background: An important clinical problem in the management of patients with meningiomas is the difficulty in predicting recurrence at the individual patient level which limits the appropriate selection of patients who would benefit from adjuvant therapy to delay recurrence. The current WHO grading of meningiomas is based entirely on histology, without molecular considerations, but recent data suggests the utility of tumor DNA methylation as a clinically relevant biomarker. We therefore aimed to develop and validate a combined molecular and clinical predictor of meningioma recurrence in individual patients to help personalize decision making regarding the need for adjuvant treatment in meningiomas. Methods: In this multicenter cohort study, we utilized global DNA methylation profiles from human tumor samples to generate a methylome-based predictor of recurrence-free survival (RFS) in meningioma. Cox modeling of individual probes was used to select features for model generation in a training cohort (N = 228 patients) which was then applied to two independent validation cohorts (N = 54; N = 140 patients). Frequency and burden of total DNA copy number alterations was computed per sample and correlated with distinct risk groups defined by the methylome-based predictor. Gene-expression analysis was correlated to DNA methylation profiles using two published microarray datasets (GSE16581; GSE9438). Finally, penalized Cox modeling was used to generate a 5-year meningioma recurrence score based on a nomogram that integrated our validated methylome-based predictor with established prognostic clinical factors. Results: Using a 5-year RFS metric, the methylome-based predictor performed favorably compared with a grade-based predictor in both validation cohorts (delta AUC = 16%; delta AUC = 12%) and was independently associated with RFS after adjusting for tumor grade, extent of resection and burden of copy number alterations (EOR; HR 3.7, 95% CI: 1.6-9.0, p < 0.001). The majority (>86%) of probes included in the methylome-based predictor demonstrated hypermethylation in recurrence-prone tumors relative to non-recurrence-prone tumors and functional annotation of the included probes implicated the homeobox gene family in tumor recurrence. A nomogram constructed using the validated methylomepredictor with WHO grade and EOR demonstrated greater predictive performance than a nomogram using clinical factors alone (delta AUC = 7-7%) and resulted in two different risk groups with distinct recurrence patterns (p < 0-001). Conclusion: Current diagnostic modalities do not sufficiently enrich patients into clinical groups for precise stratification of individual patient management in meningioma. The methylome-based predictor and meningioma recurrence score developed and validated in this study provide important prognostic information not captured by previously established clinical and molecular factors that could be used to individualize decisions regarding postoperative therapeutic interventions, in particular, whether to treat patients with adjuvant radiation therapy versus observation alone.