A clinical nomogram for predicting drug-resistant epilepsy and its long-term persistence in children: a cohort study.

PURPOSE: To develop and internally validate a clinical prediction nomogram for drug-resistant epilepsy (DRE) in children and to identify predictors of long-term persistence. METHODS: We retrospectively analyzed the data of 4208 pediatric patients with epilepsy (aged 0-18 years) diagnosed between 1995 and 2017 at a tertiary pediatric center. DRE was defined using the adapted ILAE criteria. The cohort was randomly split into training (70%) and validation (30%) sets for analysis. Predictors were selected using LASSO regression and incorporated into a multivariable logistic regression model presented as a nomogram. The model performance was assessed using discrimination, calibration, and decision curve analyses. Among patients with DRE who were followed up for ≥ 5 years without surgical intervention, predictors of persistent versus resolved DRE were evaluated. RESULTS: Of the 4208 patients, 1419 (33.7%) developed DRE. The nomogram demonstrated good discrimination (validation AUC 0.778; 95% CI 0.751-0.805) and adequate calibration. Among 1000 patients with DRE and extended follow-up, 652 (65.2%) had persistent DRE at the last follow-up. Independent predictors of persistence were psychiatric disorders (OR 2.82), daily seizure frequency (OR 1.64), seizure onset before 1 year (OR 1.46), and perinatal hypoxic-ischemic encephalopathy (OR 0.54). CONCLUSION: This nomogram enables early risk stratification of pediatric patients with epilepsy using routine clinical variables. Distinct predictors of persistent versus resolved DRE highlight the heterogeneous nature of drug resistance and support individualized management strategies.