Next Generation Sequencing Mutational Profiles of Gliomas Reveal Clinical and Genetic Factors Predicting Refractory Postoperative Seizures

INTRODUCTION: Postoperative seizures are a common complication of gliomas following resection, but little is known about their course and the role of tumor biology in their development. METHODS: Single-center retrospective review of adults who underwent surgical management of glioma with targeted next generation sequencing of tumor samples obtained 2017-2020. Univariate analysis for relevant cohort characteristics and adjustment with multivariate regression was performed. RESULTS: Among 341 patients with at least 3 months of follow-up, 174 (51.0%) experienced postoperative seizures and 89 (26.1%) had refractory seizures, defined as seizures occurring monthly or more frequently. Patients with refractory seizures were younger, presented with seizure, had IDH-mutated tumors, and had recurrent tumors. Postoperative refractory seizures had a median length of 9.4 months. In a multivariate analysis adjusting for age, seizures at presentation, newly diagnosed or recurrent status, WHO grade, and IDH status, PIK3R1 mutation emerged as an independent predictor of increased odds of developing refractory seizures (log(OR) 0.89 [0, 1.8], p = 0.044) with PIK3CA mutations trending towards significance (log(OR) 0.66 [-0.12, 1.4], p = 0.09). In another stratified multivariate analysis, CKD4 (log(OR) 1.9 [0.42, 3.4], p = 0.014), PIK3CA (log(OR) 1.3 [0.15, 2.4], p = 0.026) emerged as predictors of refractory seizure among recurrent tumors while presenting with seizure (log(OR) 0.81, [0.08, 1.5], p = 0.03) was the only significant predictor among newly diagnosed tumors. CONCLUSIONS: PIK3R1 and PIK3CA mutations were independent predictors of increased odds of postoperative refractory seizures, with CDK4 mutations as an additional predictor among recurrent tumors. Seizures at presentation predicted refractory postoperative seizures in newly diagnosed tumors, suggesting the need for close clinical follow-up in these patients. Future analysis of this cohort will enable better characterization of the influence of tumor mutational profile on postoperative seizure activity.