Predictors of preoperative and early postoperative seizures in patients with intra-axial primary and metastatic brain tumors: A retrospective observational single center study

Objective Antiepileptic treatment of brain tumor patients mainly depends on the individual physician's choice rather than on well‐defined predictive factors. We investigated the predictive value of defined clinical parameters to formulate a model of risk estimations for subpopulations of brain tumor patients. Methods We enclosed 650 patients > 18 years of age who underwent brain tumor surgery and included a number of clinical data. Logistic regressions were performed to determine the effect sizes of seizure‐related risk factors and to develop prognostic scores for the occurrence of preoperative and early postoperative seizures. Results A total of 492 patients (334 gliomas) were eligible for logistic regression for preoperative seizures, and 338 patients for early postoperative seizures. Age ≤ 60 years (odds ratio [OR] = 1.66, p = 0.020), grades I and II glioma (OR = 4.00, p = 0.0002), total tumor/edema volume ≤ 64cm3 (OR = 2.18, p = 0.0003), and frontal location (OR = 2.28, p = 0.034) demonstrated an increased risk for preoperative seizures. Isocitrate–dehydrogenase mutations (OR = 2.52, p = 0.026) were an independent risk factor in the glioma subgroup. Age ≥ 60 years (OR = 3.32, p = 0.041), total tumor/edema volume ≤ 64cm3 (OR = 3.17, p = 0.034), complete resection (OR = 15.50, p = 0.0009), diencephalic location (OR = 12.2, p = 0.013), and high‐grade tumors (OR = 5.67, p = 0.013) were significant risk factors for surgery‐related seizures. Antiepileptics (OR = 1.20, p = 0.60) did not affect seizure occurrence. For seizure occurrence, patients could be stratified into 3 prognostic preoperative and into 2 prognostic early postoperative groups. Interpretation Based on the developed prognostic scores, seizure prophylaxis should be considered in high‐risk patients and patient stratification for prospective studies may be feasible in the future. Ann Neurol 2015;78:917–928