Abstracts

Rationale: Resective epilepsy surgery is currently the treatment of choice for intractable focal epilepsy. Available literature has identified multiple isolated seizure outcome predictors, but there is no available method to systematically incorporate the constellation of a patient's complex clinical, imaging, and surgical characteristics into a unified model able to provide an individualized prediction of seizure outcome. Nomograms are well-studied and validated statistical tools often used in several disciplines of medicine, mainly oncology and cardiac surgery, to predict response to a given therapy. We aim to apply this knowledge to the study of epilepsy surgery outcomes and to develop a nomogram to predict seizure-freedom following resective surgery. Methods: A nomogram to predict seizure-freedom at 5 postoperative years based on gender, age at epilepsy onset (in years), age at surgery (in years), monthly pre-operative seizure-frequency, history of secondary generalization (present vs absent), side of surgery, and pathology (Mesial temporal sclerosis, malformations of cortical development, tumor, stroke, or other) was developed from patients treated with resective surgery at the Cleveland Clinic Epilepsy Center between 1996 and 2011. The nomogram was then tested in an external validation dataset of patients operated on over a similar timeframe in comprehensive epilepsy surgery centers within the USA, Europe, and Latin America (Niguarda Hospital, Milano, Italy; Mayo Clinic, Minnessota, USA; University of Campinas, Campinas, Brazil; INSERM, Marseille, France ). The nomogram was assessed by calculating concordance indices and testing calibration of predicted seizure-freedom with observed seizure-freedom. Results: The development cohort (DC) included 845 patients, while the validation cohort (VC) included 605 patients for a total patients analyzed of 1,450 cases. Patient characteristics were mostly similar among the DC and VC: half were male (50% in DC and 54% in VC); around half had left-sided surgery (52% in DC and 47% in VC); median monthly seizure-frequency was 8 in DC and 10 in VC; most common surgery was temporal lobectomy (68% in DC and 58% in VC); and mesial temporal sclerosis represented 30% of surgeries in both cohorts. The DC had less lesional patients (MRI abnormal in 80% in DC vs 89% in VC) and less malformations of cortical development (29% in DC vs 38% in VC). Using the DC, a predictive reduced model was created. Table 1 lists each variable along with its estimates, odd ratios and corresponding p-values. This model was then tested in the external "validation cohort". The corresponding nomogram calibration curve (Figure 1) revealed excellent concordance of predicted to observed seizure-freedom, throughout the range of seizure outcomes, with a concordance statistic of 0.65. Conclusions: This nomogram accurately predicts seizure-freedom at 5 years after resective surgery for intractable focal epilepsy. It will provide a new, valid and unique tool to individualize post-surgical outcome prediction in the setting of clinical practice.