Abstracts

Rationale: Presurgical evaluation for temporal lobectomy includes MRI, FDG-PET, routine interictal surface EEG (rEEG), ictal surface video EEG (vEEG), ictal invasive EEG (iEEG), and interictal intraoperative monitoring (IO). Some of these carry significant cost and risk to patients. It is currently unknown how much predictive power each independently or in combination have for estimating long-term seizure and aura freedom after surgery. Methods: All 151 epilepsy surgeries at the National Institutes of Health from 1984-2011 were reviewed. Excluding repeat surgeries, extratemporal cases, and lack of follow up, 116 cases remained. Presurgical test results were coded relative to the location of resection as: concordant, discordant, partially concordant, or normal/diffuse. A Kaplan-Meier survival curve estimated overall seizure/aura freedom for all patients. Survival curves estimated the seizure/aura freedom for those codes which had sufficient number of patients in each modality, and a log-rank test evaluated for statistical difference between codings. Additionally, linear models combining the results from different modalities (with and without interaction terms) were also evaluated for positive predictive values (PPV) and negative predictive values (NPV). Results: Overall seizure-freedom rates were 58%, 53%, 49% and 47% at 0.5, 1, 2 and 5 years respectively. MRI concordance (71%, 64%, 60% and 58% at at 0.5, 1, 2 and 5 years) vs normal/diffuse (43%, 37%, 34% and 31% at at 0.5, 1, 2 and 5 years) was the only modality/code that achieved significance (p=0.025), while vEEG (p=0.063), rEEG (p=0.517), PET (p=0.875), iEEG (p=0.846), and IO (p=0.257) did not. More advanced models did not achieve PPV higher than 73% and NPV higher than 54% at 1 year. Conclusions: The low PPV and NPV would preclude clinical utility of advanced models from these data. Larger datasets may permit more comprehensive modeling of prediction. Nevertheless, complete resection of MRI positive lesions confer a clear seizure and aura freedom advantage which is sustained over years post-operatively. Interestingly, other modalities (PET, rEEG, vEEG, iEEG and IO) which would be expected to have greater independent predictive power did not achieve statistical significance. Decisions regarding epilepsy surgical planning should consider complete resection of MRI lesions a primary objective for achieving significantly better long term outcomes.