Abstracts

Rationale: Intracranial electrodes are indicated when there is a reasonable assumption that most or all seizures originate from a single region and that a surgical procedure can be performed, with some additional localizing information. The recent emergence of responsive neurostimulation (RNS) and deep brain stimulation (DBS) as viable options for seizure control has changed the landscape of epilepsy surgery but removal of seizure onset zone still offers the best chance for seizure freedom. Prior to outcome prediction, it is critical that we improve our ability to predict the possibility of a resective surgery. Patient heterogeneity, limited sampling and tailored electrode implantation are inherent limitations of any intracranial study design. However, even with these limitations, the key question remains: how can we better patient selection? This question has never been more relevant than in the age of DBS, which does not require accurate seizure localization prior to implantation of thalamic electrodes.  Methods: Retrospective chart review of all intracranial studies performed for localization of seizure onset in patients with medically refractory epilepsy at Yale-New Haven Hospital from 2002 to 2016. We identified a total of 16 variables of interest: duration of epilepsy, seizure risk factors, aura, MRI, PET, subtraction SPECT, interictal scalp EEG, ictal scalp EEG, correlation of interictal v ictal scalp EEG (concordant or discordant), correlation of scalp ictal EEG v MRI (concordant or discordant), pre-implant hypothesis by lateralization, pre-implant hypothesis by lobe, total number of electrode contacts, side of intracranial study (bilateral/dominant/non-dominant), duration of intracranial EEG recording, and intracranial EEG seizure onset (localized vs not localized). We hypothesize that these variables may inform the likelihood of predicting the possibility of a resective surgery following an intracranial implant, which is the primary outcome measure. Variables were analyzed using the Fisher's exact test or the Mann-Whitney test, as appropriate. P values < 0.05 were considered significant.  Results: There were 178 patients in the study group. 116 (65%) were in the resection group (purely resective surgery=93, resection and multiple subpial transections (MST) = 19, resection and RNS = 4). 62 were in the non resective surgery group (No surgery=42, MST only=5, RNS only=14, laser ablation=1).  Presence and type of aura, seizure risk factors, PET, subtraction SPECT, interictal scalp EEG findings and number of electrode contacts were not significantly different between the two groups. Shorter duration of epilepsy, well-lateralized scalp ictal EEG onsets, concordant findings on interictal v ictal scalp EEG, and ictal scalp EEG v MRI, lesional MRI, well lateralized pre-implant hypothesis, non-dominant hemispheric implants, and well localized intracranial EEG onsets were all associated with a higher probability of having a resective surgery (p< 0.05).   Conclusions: 65% (116/178) patients in this cohort underwent resective surgery following an intracranial implant. A well-lateralized, pre-implant hypothesis requiring a non-dominant hemispheric implant was most likely to result in a resective surgery (35/37, p<0.0001); this had a specificity of 96.7% (95% CI 89-99%), positive predictive value of 94.6% (95% CI 82-99%) and false positive rate of 3.3% (95% CI 1-11%). Other combination of variables (poorly lateralized dominant or non-dominant and well-lateralized dominant) had a yield that was only marginally better than chance level outcome for predicting the probability of resective surgery.   Funding: No funding