EEG features in late-onset epilepsy: possible correlation with cognitive impairment

Late-onset epilepsy (LOE) increases the risk of cognitive impairment and is associated with dementia in a bidirectional manner. Scalp EEG is a potential tool for assessing this relationship, but this has not been fully explored. This cross-sectional study reviewed people who had undergone 24-h EEG monitoring at West China Hospital. Individuals with epilepsy whose seizure onset ≥55 years and older healthy controls were included. Diagnosis of cognitive impairment was established at the time of EEG using standardized neuropsychological metrics. EEGs were reviewed and annotated independently by two neurophysiologists. EEG features were extracted according to the standardized computer-based organized reporting of EEG. Potential clinical, MRI, and EEG risk factors of cognitive impairment were analysed in ordinal and binomial regression models. Least absolute shrinkage and selection operator model was used to select the variables most discriminative of cognitive impairment. Among the screened individuals (n = 8318), eligible participants with LOE (n = 287) and healthy controls (n = 132) were included; median age 65 (55-85) years; female 38%. Epilepsy aetiology was unknown in 177 (62%) participants. Structural aetiology was the most common in epilepsies (93%) with a definite aetiology (n = 110). Interictal epileptiform discharges were detected in 56% of participants with LOE and were primarily left temporal. Bitemporal interictal epileptiform discharges were found predominantly (38%) in those of an unknown aetiology. Focal slowing was found in 56%, and generalized slowing in 13% of participants with LOE. EEG markers of temporal neural hyperexcitability, including temporal intermittent rhythmic delta activity (TIRDA) and anterior temporal epileptiform discharges, were associated with cognitive impairment. In LOE of unknown aetiology, bilateral anterior temporal epileptiform discharges were the most discriminative indicator of cognitive impairment [odds ratio 53.280, 95% confidence interval (CI) 11.359-249.917]. The least absolute shrinkage and selection operator model achieved an area under the receiver operating characteristic curve of 0.869 (95% CI 0.813-0.925). In LOE with a definite aetiology, TIRDA was associated with cognitive impairment. In this study, LOE was associated with specific EEG patterns. Signatures of temporal hyperexcitability on EEG might be related to cognitive impairment in LOE, especially when presented in both hemispheres. These results also suggested that LOE of an unknown aetiology might have a neurodegenerative origin, similar to Alzheimer's disease. Future longitudinal studies should explore the role of temporal hyperexcitability and other EEG features in the bidirectional link between LOE and dementia.