Multi-Center Evaluation of Epilepsy Surgery Prediction Models: Comparative Performance of the Epilepsy Surgery Nomogram and the Epilepsy Surgery Grading Scale
Abstract number :
1.317
Submission category :
9. Surgery / 9A. Adult
Year :
2021
Submission ID :
1826184
Source :
www.aesnet.org
Presentation date :
12/4/2021 12:00:00 PM
Published date :
Nov 22, 2021, 06:52 AM
Authors :
Amanda Zhao, BA - NYU Grossman School of Medicine; Christina Boada, MD – University of Pennsylvania; Scott Grossman, MD – New York University; Seyda Erdogan, MD – Cleveland Clinic; Zhou Quin, MD – Cleveland Clinic; Billakota Santoshi, MD – New York University; Vineet Punia, MD – Cleveland Clinic; Patricia Dugan, MD – New York University; Jacqueline French, MD – New York University; Lara Jehi, MD – Cleveland Clinic
Rationale: While the field of epilepsy surgery has advanced substantially over the past few decades, clinical decisions on surgical candidacy still occur by consensus agreement following a multidisciplinary evaluation. Given the subjective nature of the evaluation, numerous models have been developed to determine which patients are likely to achieve seizure freedom from epilepsy surgery. Here we evaluate two models, the Epilepsy Surgery Nomogram (ESN) developed at the Cleveland Clinic, and the Epilepsy Surgery Grading Scale (ESGS) at New York University to assess their comparative accuracy across both patient populations and determine if a superior model based on preoperative data can be derived.
Methods: This is a multicenter, retrospective study of all patients who underwent resective epilepsy surgery at either the Cleveland Clinic or the NYU Comprehensive Epilepsy Center between 2007-2011. Inclusion criteria included age thirteen and older and a diagnosis of focal epilepsy. Patients were excluded if they received prior resective neurosurgery, multilobar resections, callosotomy, hemispherectomy, or had generalized epilepsy, psychogenic nonepileptic seizures, a progressive neurodegenerative disorder or less than one year of follow-up available. Basic demographic and clinical data were collected for all patients ranging from age at presentation to lobe of resection. For each patient, both predictive models were run. The ESN input variables were epilepsy duration, seizure frequency, history of generalized tonic-clonic seizure (GTC), cause of epilepsy and lobe of resection, while the ESGS relied on IQ, aura semiology, MRI, EEG and concordance between the MRI and EEG. For the ESN, data was entered to calculate the 2- and 5-year probability of complete seizure freedom and freedom from impaired awareness seizures. The ESGS Scale was then applied to each patient, with patients placed in one of three grades, each with decreasing probability of seizure freedom from surgery. Finally, a new model was derived based on the database. Engel outcomes at 2 and 5 years were then determined by chart review. The models were compared using Receiver Operating Curve (ROC) analysis to determine their ability to predict seizure freedom following surgery.
Results: 302 patients met inclusion criteria (Table 1). The ESN and ESGS models did not differ significantly in performance, with an area under the curve (AUC) of 0.595 (95% CI 0.532-0.659) and 0.562 (95% CI 0.498-0.625), respectively (Table 2). From this set, a new nomogram was created based on gender, history of generalized tonic-clonic seizures, and epilepsy duration with an AUC of 0.613 (95% CI 0.549-0.678) (Table 2).
Conclusions: Despite the different input variables used for calculating the ESN and ESGS predictions, the models did not differ significantly in predictive accuracy. The newly developed model relied on history of generalized tonic clonic seizure, gender and epilepsy duration. Given this model relies on purely clinical data available in the primary care and general neurologist’s office, it holds potential as a screening tool for determining epilepsy surgery candidacy early on.
Funding: Please list any funding that was received in support of this abstract.: None.
Surgery