Reduced REM sleep: a potential biomarker for epilepsy - a retrospective case-control study.

Objective: Establishing the diagnosis of epilepsy can be challenging if interictal epileptic discharges (IEDs) or seizures are undetectable. Many individuals with epilepsy experience sleep disturbances. A reduced percentage of REM sleep (REM%) has been observed following seizures. We aimed to assess differences of REM% in individuals with epilepsy in comparison with differential diagnoses.

Methods: We performed a retrospective, monocentric, two-armed case-control study with 128 age-matched individuals who underwent ≥72 hours of continuous video-EEG monitoring at our epilepsy monitoring unit (EMU) for diagnostic evaluation. We assessed REM% on the first and last night of EMU admission. Logistic regressions models were used to evaluate the predictive value of REM%.

Results: We included 64 individuals diagnosed with epilepsy and 64 with a differential diagnosis. REM% in the epilepsy group was significantly lower [12.2% (±4.7) vs. 17.2% (±5.2), p<0.001]. We found no significant influence of sex, age, anti-seizure, or other medications. A REM%-based and an IED and seizure-based regression model were not significantly different [area under the curve (AUC) 0.791 (95% confidence interval (CI): 0.713-0.870) vs. 0.853 (95% CI: 0.788-0.919), p=0.23]. A combined model, based on IEDs, seizures, and REM%, was superior to the IED model alone [0.933 (0.891-0.975), p<0.01].

Interpretation: Our study shows significantly reduced REM% in individuals with epilepsy. REM%-based models show a good predictive performance. REM% assessment could improve diagnostic accuracy - especially for challenging cases, e.g., when IEDs or seizures are absent and patient history and semiology appear ambiguous. REM% as a biomarker should be evaluated in prospective, multicentric trials.