ObjectiveThe accurate prediction of seizure freedom after epilepsy surgery remains challenging. We investigated if (1) training more complex models, (2) recruiting larger sample sizes, or (3) using data‐driven selection of clinical predictors would improve our ability to predict postoperative seizure outcome using clinical features. We also conducted the first substantial external validation of a machine learning model trained to predict postoperative seizure outcome.MethodsWe performed a retrospective cohort study of 797 children who had undergone resective or disconnective epilepsy surgery at a tertiary center. We extracted patient information from medical records and trained three models—a logistic regression, a multilayer perceptron, and an XGBoost model—to predict 1‐year postoperative seizure outcome on our data set. We evaluated the performance of a recently published XGBoost model on the same patients. We further investigated the impact of sample size on model performance, using learning curve analysis to estimate performance at samples up to N = 2000. Finally, we examined the impact of predictor selection on model performance.ResultsOur logistic regression achieved an accuracy of 72% (95% confidence interval [CI] = 68%–75%, area under the curve [AUC] = .72), whereas our multilayer perceptron and XGBoost both achieved accuracies of 71% (95% CIMLP = 67%–74%, AUCMLP = .70; 95% CIXGBoost own = 68%–75%, AUCXGBoost own = .70). There was no significant difference in performance between our three models (all p > .4) and they all performed better than the external XGBoost, which achieved an accuracy of 63% (95% CI = 59%–67%, AUC = .62; pLR = .005, pMLP = .01, pXGBoost own = .01) on our data. All models showed improved performance with increasing sample size, but limited improvements beyond our current sample. The best model performance was achieved with data‐driven feature selection.SignificanceWe show that neither the deployment of complex machine learning models nor the assembly of thousands of patients alone is likely to generate significant improvements in our ability to predict postoperative seizure freedom. We instead propose that improved feature selection alongside collaboration, data standardization, and model sharing is required to advance the field.

To investigate the link between sleep disruption and cognitive impairment in childhood epilepsy by studying the effect of epilepsy on sleep homeostasis, as reflected in slow-wave activity (SWA). Method:We examined SWA from overnight EEG-polysomnography in 19 children with focal epilepsy (mean [SD] age 11 years 6 months [3 years], range 6 years 6 months-15 years 6 months; 6 females, 13 males) and 18 age-and sex-matched typically developing controls, correlating this with contemporaneous memory consolidation task scores, full-scale IQ, seizures, and focal interictal discharges.Results: Children with epilepsy did not differ significantly from controls in overnight SWA decline (p = 0.12) or gain in memory performance with sleep (p = 0.27). SWA was lower in patients compared to controls in the first hour of non-rapid eye movement sleep (p = 0.021), although not in those who remained seizure-free (p = 0.26). Full-scale IQ did not correlate with measures of SWA in patients or controls. There was no significant difference in SWA measures between focal and non-focal electrodes.Interpretation: Overnight SWA decline is conserved in children with focal epilepsy and may underpin the preservation of sleep-related memory consolidation in this patient group. Reduced early-night SWA may reflect impaired or immature sleep homeostasis in those with a higher seizure burden.

ObjectiveNeurosurgery is a safe and effective form of treatment for select children with drug‐resistant epilepsy. Still, there is concern that it remains underutilized, and that seizure freedom rates have not improved over time. We investigated referral and surgical practices, patient characteristics, and postoperative outcomes over the past two decades.MethodsWe performed a retrospective cohort study of children referred for epilepsy surgery at a tertiary center between 2000 and 2018. We extracted information from medical records and analyzed temporal trends using regression analyses.ResultsA total of 1443 children were evaluated for surgery. Of these, 859 (402 females) underwent surgical resection or disconnection at a median age of 8.5 years (interquartile range [IQR] = 4.6–13.4). Excluding palliative procedures, 67% of patients were seizure‐free and 15% were on no antiseizure medication (ASM) at 1‐year follow‐up. There was an annual increase in the number of referrals (7%, 95% confidence interval [CI] = 5.3–8.6; p < .001) and surgeries (4% [95% CI = 2.9–5.6], p < .001) over time. Duration of epilepsy and total number of different ASMs trialed from epilepsy onset to surgery were, however, unchanged, and continued to exceed guidelines. Seizure freedom rates were also unchanged overall but showed improvement (odds ratio [OR] 1.09, 95% CI = 1.01–1.18; p = .027) after adjustment for an observed increase in complex cases. Children who underwent surgery more recently were more likely to be off ASMs postoperatively (OR 1.04, 95% CI = 1.01–1.08; p = .013). There was a 17% annual increase (95% CI = 8.4–28.4, p < .001) in children identified to have a genetic cause of epilepsy, which was associated with poor outcome.SignificanceChildren with drug‐resistant epilepsy continue to be put forward for surgery late, despite national and international guidelines urging prompt referral. Seizure freedom rates have improved over the past decades, but only after adjustment for a concurrent increase in complex cases. Finally, genetic testing in epilepsy surgery patients has expanded considerably over time and shows promise in identifying patients in whom surgery is less likely to be successful.