Noninvasive Predictors of Subdural Grid Seizure Localization in Children With Nonlesional Focal Epilepsy

Abstract

Subdural grid evaluation (SDE) in refractory focal epilepsy aims to precisely define the ictal onset zone and map eloquent cortex. In a small but significant proportion of children, SDE shows multifocal or diffuse, rather than focal, seizure onset. Resective epilepsy surgery is denied, or is unsuccessful, in the majority of such patients. The authors investigated whether the noninvasive data could be abstracted to predict subsequent SDE electrographic outcome (focal vs. multifocal/diffuse ictal onset). The authors retrospectively reviewed charts of 66 children with refractory focal epilepsy undergoing SDE at Cleveland Clinic over a 7-year period, studied previously by Pestana Knight et al. A semiquantitative "score" summarizing the localizing value and concordance between selected noninvasive investigations (interictal and ictal EEG; positron emission tomography [PET], and/or single-photon emission computed tomography [SPECT]), as well as Bayesian predictors of individual investigations and their combinations, were adapted from the study of Kalamangalam et al to the subset of patients with nonlesional cranial MRI. Forty (60.6%) patients had a single MRI brain lesion, 7 (10.6%) had bilateral or diffuse MRI changes, and 19 (28.8%) were nonlesional. Subdural grid evaluation ictal onset was nonfocal in four patients in the first group (10%) and in two patients (28.5%) in the second group. One patient in the third (nonlesional) group was excluded because of incomplete data. In the remainder (n = 18), SDE ictal onset was multifocal or diffuse in 5 (27.8%) and focal in 13 (72.2%). Focality on SDE was positively correlated with higher noninvasive scores in the nonlesional patient group (χ test, P < 0.025). Bayesian predictors in this group were highest for concordance between the interictal and ictal scalp EEG (likelihood ratio = 3.85). Considered separately, interictal and ictal EEG were of equivalent predictive value (likelihood ratio = 2.3 and 2.1, respectively). Metabolic imaging was the least useful modality. (1) Diffuse or multifocal ictal onsets on SDE are almost three times as likely in nonlesional patients as in those with a single definite MRI brain lesion. (ii) The noninvasive data of children with nonlesional brain MRI may be summarized by a score that rewards localizing information and intermodality concordance: low-scoring patients are more likely to exhibit diffuse or multifocal ictal onset on subsequent SDE. (iii) Bayesian likelihood ratios predictive of ictal focality on SDE are highly favorable for concordant scalp interictal-ictal EEG combinations. (iv) Decision-theoretic methods of this type may find use in the selection of nonlesional pediatric presurgical candidates offered SDE.