Introduction
Systematic reviews and meta-analyses showed a 32%-57% prevalence of Depression or depression symptoms in patients with Narcolepsy disorders. MSLT protocols recommend a drug washout period of ≥14-days for antidepressants and other medications known to confound sleep architecture and REM sleep. Patient noncompliance with washout protocols for REM-suppressing antidepressants (REMS-AD) was found to reduce the incidence of SOREMP observations in MSLTs by ≥89% compared to washout-compliant patients. Prior studies demonstrated potential for improvements in Narcolepsy diagnostic sensitivity and reproducibility by exploiting more comprehensive analysis of full-night polysomnography (PSG) signals, through deep learning-based AI models (ROC-AUC 0.96) and algorithmic approaches. Prior research characterized Narcolepsy-specific microarchitectural sleep EEG abnormalities that manifest in Sleep Latency (SL), REM (Latency (REML), median/minimum duration, density, prevalence), N3 (time/prevalence), hypnodensity characteristics, and separately, Depression-specific state and trait EEG markers including low Delta Sleep Ratio (DSR), high/oscillatory Gamma, and high/asymmetric Alpha. Few works examine the extent of confounds to Narcolepsy diagnostic parameters from Depression comorbidity with and without the specific influence of REMS-ADs.
Methods
A nonrandomized, observational, de-identified cohort of N=282 PSGs were collected from subjects that underwent MSLT immediately following the PSG. Excluding subjects without UDS results, the final cohort (N=133) was composed of 59.6% Narcolepsy(+) subjects (N=35 NT1, N=44 NT2), and 40.4% Narcolepsy(-) Clinical Controls subjects. 47.9% of subjects had a documented Depression diagnosis, 52.1% with verified absence of Depression diagnostic history, and a 49.3% rate of comorbid Depression in Narcolepsy. The sample included N=39 subjects that were Narcolepsy(+)/Depression(+) for comparison to N=40 Narcolepsy(+)/Depression(-) subjects.
Among Narcolepsy(+)/Depression(+), N=29 were REMS-AD(+) and N=10 REMS-AD (-) confirmed by UDS. Mean latency, density, and prevalence were calculated for sleep stages (N1, N2, N3, REM, Wake, All-Stages), mean power spectral densities and ratios were calculated (Delta, Theta, Alpha, Beta, Gamma), then evaluated for differences and statistical significance in mean diagnostic parameter values based on two-sided 95% bootstrap median-percentile method Confidence Intervals (R=2,000).
Results
In NT1, NT2, and CC, Mean REML was shortest in MDD(+)/REM-AD(-) subjects (NT1/NT2: 67mins (54mins, 77mins), CC: 118mins (97mins, 125mins)), the longest in MDD(+)/REM-AD(+) subjects (NT1/NT2: 156mins (115mins, 203mins), CC: 190mins (133mins, 247mins)), with MDD(-) subjects falling between (NT1/NT2: 108mins (93mins, 125mins), CC: 120mins (94mins, 152mins)).
In both Narcolepsy and CC, Mean REM density was highest in MDD(+)/REM-AD(-) subjects (NT1/NT2: 111mins (88mins, 128mins), CC: 97mins (91mins, 105mins)), and substantially reduced in MDD(+)/REM-AD(+) subjects (NT1/NT2: 101mins (87mins, 113mins), CC: 82mins (62mins, 104mins)).
Mean DSR was lowest in REM-AD(-) Clinical Controls (CC: 0.96), and normalized in REM-AD(+) (CC: 1.45) relative to MDD(-) subjects (CC: 1.28). DSR was lower in REM-AD(+) and MDD(-) in Narcolepsy than Clinical Controls (NT1/NT1: REM-AD(+): 1.14, REM-AD(-): 1.19, MDD(-): 1.39).
Mean SOREMPs were lowest in REM-AD(+) both in Narcolepsy (NT1/NT2: REM-AD(+): 1.00, REM-AD(-) 1.71, MDD(-): 1.50) and clinical controls (CC: REM-AD(+): 0.21, REM-AD(-) 0.33, MDD(-): 0.42).
MSLT Mean Sleep Latency (MSL) was shorter in REM-AD(-) NT1/NT1 than REM-AD(+) NT1/NT2, shorter in REM-AD(-) Clinical Controls than MDD(-) or REM-AD(+).
Conclusion
Comorbid Depression patients with and without REM-ADs showed substantial alterations in key Narcolepsy disorder diagnostic and sleep architecture parameters. Given significant rates of comorbidity, these findings highlight needs to better understand the interactions and impact of Depression on Narcolepsy diagnosis. Future work involves further validation of AI-enabled identification and detection for NT1 and MDD respetively.
1 EnsoData Research, Ensodata, Madison, WI, USA | 2 Department of Neurology, University of Washington School of Medicine, Seattle, WA, USA | 3 Rush University Medical Center
