Abstract The study aimed to explore whether Prolonged Focal Impaired Consciousness seizures (pFIC) predict drug resistance in patients with epilepsy and to identify the optimal seizure duration threshold for predicting this outcome. We registered all patients with epilepsy who were consecutively admitted to the Comprehensive Epilepsy Unit for long-term video-electroencephalogram (VEEG) monitoring at the First Affiliated Hospital of Sun Yat-sen University from January 2013 to December 2023. The database consisted of demographic information, clinical characteristics and electroencephalogram records. We evaluated the predictive performance of the recorded longest duration of Focal Impaired Consciousness seizures (FIC) through Receiver Operating Characteristic (ROC) analysis in patients with drug-resistant epilepsy (DRE). We developed and validated the optimal prognostic cut point for FIC longest duration for predicting DRE outcome using the Youden index. We compared the clinical and electroencephalographic characteristics between those with seizure duration shorter or longer than the cut point and developed a DRE prediction model. Within 18924 patients admitted in the registration database, 490 patients who experienced FIC during VEEG monitoring were enrolled in this study, among whom 122 (24.9%) were diagnosed with DRE. The optimal cutoff point for the longest duration of FIC was 3 minutes (Youden index of 0.111, sensitivity of 47.5%, and specificity of 63.6%). 192 (39.2%) patients’ longest FIC duration was greater than or equal to 3 minutes, and 298 (60.8%) patients’ longest FIC duration for less than 3 minutes. There were more DRE patients in pFIC duration group than in shorter FIC group (21.5% vs. 30.2%, P 0.05). In multivariate analysis we found that pFIC, subclinical seizures, multiple seizure types, frontal seizure onset and known etiology were associated with increased risk of DRE (P 0.05). The FIC duration did not show multicollinearity (tolerance 0.988, variance inflation factor 1.012). We incorporated these indicators to multicollinearity analyses and established a DRE risk prediction model using five indicators: pFIC, frontal lobe origin, known etiology, multiple seizure types, and subclinical seizures. The area under the ROC curve was 0.700. Our findings demonstrated that FIC duration≥3 minutes was an independent predictor for DRE. The integrative prediction model incorporating pFIC, frontal lobe origin, known etiology, multiple seizure types, and subclinical seizures facilitated early identification of high-risk patients.
Liu et al. (Thu,) studied this question.