Authors :
Presenting Author: Eliana Li, – UTHealth Houston
Oman Magana-Tellez, PhD – UTHealth Houston
Luo Xi, PhD – UTHealth Houston
Norma Hupp, BS – UTHealth Houston
Sandhya Rani, PhD – UTHealth Houston
Johnson Hampson, MBBE – UTHealth Houston
Nuria Lacuey, MD, PhD – UTHealth Houston
Samden Lhatoo, MD – UTHealth Houston
Rationale:
Slow-wave activity (SWA) during non-rapid eye movement (NREM) sleep is a key indicator of sleep homeostasis. SWA power is highest in the early portion of the night and declines across subsequent NREM cycles, reflecting successful synaptic downscaling. However, emerging evidence suggests that normal SWA trends may be disrupted in individuals with epilepsy, with some patients exhibiting abnormally increasing SWA power throughout the night. Specifically, generalized tonic-clonic seizures (GTCS) have been associated with altered SWA trends. Additional clinical factors may also influence SWA patterns, though the extent to which these variables contribute to abnormalities remain at question. In this study, we aimed to identify clinical or electrophysiological correlates that differentiate patients with abnormal SWA trends from those exhibiting normal SWA trends by analyzing electroencephalography (EEG) recordings across the night and collecting clinical variables from patient reports.
Methods:
We collected 78 overnight seizure-free EEG recordings from 65 patients with a history of GTCS. Clinical data was obtained from patient medical reports. EEG processing was performed in the Brainstorm suite in MATLAB, and an automatic sleep staging algorithm, SleepCRNN, was used to classify sleep. SWA power was calculated using the bandpower function in MATLAB and fitted with linear regression model to determine their overnight SWA progression. Patients were classified into 2 groups: abnormal SWA (positive slope) and normal SWA (negative slope). Finally, a receiving operating curve (ROC) analysis was performed to determine which of the clinical or electrophysiological descriptors can be used to differentiate between the patient cohorts. A random forests algorithm was trained for predicting future slope progression in patients with multiple visits.
Results:
From the 65 patients in our database, 32 patients had a normal SWA slope while 33 had an abnormal SWA slope. A total of 53 patients had a single admission and 12 patients had more than one. The ROC analysis dividing patients between abnormal and normal SWA showed that significant descriptors were the average band power across the night from the Theta (p=0.008), Beta (p=0.023), and Delta (p=0.041) bands. Specifically, abnormal SWA slope exhibited more average Theta and Beta band power, but less average Delta band power than normal SWA slope. Additionally, our algorithm trained on the prediction of future slope progression showed that time spent in N2 sleep stage was the most significant continuous predictor (AUC 0.85, p< 0.001) of SWA slope, with greater time spent in N2 sleep resulting in a more downwards SWA trend.
Conclusions:
Our results suggest that half of all patients with GTCS have abnormal sleep homeostasis and spend less time in deep sleep than patients with normal SWA, thus resulting in impaired restorative sleep processes. Interestingly, GTCS frequency and other clinical characteristics were not significantly different between those with normal and those with abnormal sleep homeostasis, suggesting that other, as yet unidentified factors may be responsible for these findings.
Funding: NIH-NINDS U01 NS090407