Development and temporal validation of a postoperative nomogram based on clinical, surgical, histopathological, and semi-quantitative EEG features for predicting seizure-free survival in children after focal epilepsy surgery

Vollständiger Abstract

Worum geht es in dieser Arbeit?

Objective Accurate estimation of seizure recurrence risk after pediatric epilepsy surgery may support individualized postoperative surveillance and management. This study aimed to develop and temporally validate a prognostic nomogram for estimating 1-year and 2-year seizure-free survival after focal or lobar resection in children with drug-resistant epilepsy. Methods This single-center retrospective cohort study included 256 children treated between 2015 and 2024. Patients treated during 2015–2021 constituted the training cohort ( n = 181), whereas those treated during 2022–2024 constituted the temporal validation cohort ( n = 75). Eighteen prespecified clinical, neuroimaging, scalp EEG, surgical, and histopathological candidate predictors were evaluated using least absolute shrinkage and selection operator penalized Cox regression with 10-fold cross-validation. Selected variables were entered into a multivariable Cox proportional hazards model. Internal validation repeated the complete model-development procedure in 1,000 bootstrap resamples. Model performance was assessed using Harrell's C-index, time-dependent area under the curve, calibration, Brier scores, and decision curve analysis. Results Five predictors were retained: longer epilepsy duration, absence of a definitive structural MRI lesion, histopathological diagnosis, subtotal resection, and a high non-rapid eye movement interictal spike index (>38 spikes/h). The optimism-corrected C-index was 0.81 in the training cohort, and the C-index was 0.79 in the temporal validation cohort. The 1-year and 2-year time-dependent AUCs were 0.82 and 0.81 in the training cohort and 0.80 and 0.78 in the temporal validation cohort, respectively. Calibration was generally acceptable, and decision curve analysis suggested potential net benefit across the evaluated threshold range. Conclusions The nomogram showed useful performance for estimating postoperative seizure-free survival in this single-center cohort. Because histopathological diagnosis and resection completeness are required, the model is intended for early postoperative risk stratification rather than preoperative candidate selection or presurgical counseling. Independent multicenter validation and prospective evaluation of clinical impact are required before routine implementation.