Abstract PO-060: Individualized prediction of meningioma recurrence risk over prolonged time periods

Abstract

Abstract Rationale and purpose: Meningiomas are benign tumors, but significant number of meningiomas display risk of early tumor recurrence. The inter-observer variability among pathologists for grading and some indistinguishable features of meningioma under the microscope prevent accurate prediction of recurrence risk that critically limits appropriate treatment and management of patients who may benefit from adjuvant radiation therapy. We aim to develop individualized prediction of meningioma recurrence risk over prolonged time periods using DNA methylation signatures and other associated clinical factors. Method: The recurrence risk predictor is based on continuous survival random-forest modelling to calculate the probabilities of a recurrence for each meningioma patient over various prolonged time frames. We processed over 500 meningioma samples using our machine-learning model that predicts probability of tumor recurrence based on methylation data. We used Random Forest based Rangers package as the survival model that returns the probabilities of a recurrence not happening over various prolonged time frames. Result: Methylation-based predictor distinguishes high-low risk groups over prolonged time periods. The predictor was validated using external validation cohort based on the selected probes and original training model. We found that methylation based predictor has potential to identify in patients with grade II meningiomas especially, while it also can identify grade I patients with higher recurrence risk over prolonged time periods. Conclusion: There were limitations of standard of care classifications in meningioma especially for Grade 2 patients due to significant intra- and inter-observer variability for grading and selection of patients for treatment. The individualized risk predictor can help determining the decisions regarding adjuvant radiation treatment versus observation along for each meningioma patient in the clinic. Citation Format: Yasin Mamatjan, Farshad Nassiri, Mira Salih, Kenneth Aldape, Gelareh Zadeh. Individualized prediction of meningioma recurrence risk over prolonged time periods [abstract]. In: Proceedings of the AACR Virtual Special Conference on Artificial Intelligence, Diagnosis, and Imaging; 2021 Jan 13-14. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(5_Suppl):Abstract nr PO-060.