Quantifying the Effect of Seizure-Onset Zone Resection on Epilepsy Surgery Outcome: A Bayesian Causal Analysis

Background and Objectives – Identifying factors that determine success in epilepsy surgery is essential for improving patient selection and surgical planning. Previous studies often lack precise quantification of resection extent and rely on conventional statistics, which may be suboptimal for the small sample sizes typical of this field. We aimed to quantify how the resected fraction of the seizure-onset zone (SOZ), the clinical proxy of the epileptogenic zone, and the resection of non-SOZ tissue affect surgical outcome and to evaluate the contribution of clinical, imaging, electrophysiologic, and histopathologic factors.Methods – We retrospectively analyzed patients with drug-resistant epilepsy who underwent stereo-EEG (SEEG)-guided resective surgery at 3 centers in Canada and France, with >1 year of postoperative follow-up. Post-SEEG implantation and postoperative imaging were coregistered to preoperative imaging to compute SOZ-resected and non–SOZ-resected fractions in volumetric space. The overlap between the SOZ volume with the resection cavity determined the SOZ-resected fraction. A Bayesian causal inference workflow was applied to estimate the direct and indirect effects of these variables and other clinical factors on the probability of achieving a good surgical outcome (Engel I).Results – Two hundred patients (median age 30 years [6–57 years]; 52% female participants; 56% good outcome) were included. The SOZ-resected fraction had the strongest positive effect: each 10% increase raised the good outcome probability by 3.7% (median marginal effect 3.7% [95% CI 1.2%–6.1%], 99.7% certainty). Resection of non-SOZ tissue showed little effect. Using the 56% good outcome probability as reference, categorical determinant factors were concordance between multimodal presurgical and SEEG findings (66% [55%–78%], 92%), frontal lobe epilepsy (61% [50%–72%], 74%), focal SOZ (62% [54%–71%], 87%), and focal cortical dysplasia II (61% [51%–73%], 78%). Interestingly, posterior quadrant epilepsy (69% [55%–84%], 95%) and operculo-insular epilepsy (62% [46%–76%], 72%), often considered more challenging, increased the good outcome probability. Conversely, a widespread SOZ affected the good outcome probability negatively (33% [18%–48%], 99%).Discussion – Greater SOZ resection increases the probability of good outcome, but extending resection beyond this zone offers little additional benefit once the SOZ is removed. These findings emphasize precise SOZ targeting and the value of individualized surgical planning, particularly when multimodal data are concordant.

Duke Scholars

Author Birgit Frauscher Neurology, Epilepsy and Sleep