Intra-individual stability of NREM sleep quantitative EEG measures in obstructive sleep apnea.

Abstract

Electroencephalography is collected routinely during clinical polysomnography, but is often utilised to simply determine sleep time to calculate apnea-hypopnea indices. Quantitative analysis of these data (quantitative electroencephalogram) may provide trait-like information to predict patient vulnerability to sleepiness. Measurements of trait-like characteristics need to have high test-retest reliability. We aimed to investigate the intra-individual stability of slow-wave (delta power) and spindle frequency (sigma power) activity during non-rapid eye movement sleep in patients with obstructive sleep apnea. We recorded sleep electroencephalograms during two overnight polysomnographic recordings in 61 patients with obstructive sleep apnea (median days between studies 47, inter-quartile range 53). Electroencephalograms recorded at C3-M2 derivation were quantitatively analysed using power spectral analysis following artefact removal. Relative delta (0.5-4.5 Hz) and sigma (12-15 Hz) power during non-rapid eye movement sleep were calculated. Intra-class correlation coefficients and Bland-Altman plots were used to assess agreement between nights. Intra-class correlation coefficients demonstrated good-to-excellent agreement in the delta and sigma frequencies between nights (intra-class correlation coefficients: 0.84, 0.89, respectively). Bland-Altman analysis of delta power showed a mean difference close to zero (-0.4, 95% limits of agreement -9.4, 8.7) and no heteroscedasticity with increasing power. Sigma power demonstrated heteroscedasticity, with reduced stability as sigma power increased. The mean difference of sigma power between nights was close to zero (0.1, 95% limits -1.6, 1.8). We have demonstrated the stability of slow-wave and spindle frequency electroencephalograms during non-rapid eye movement sleep within patients with obstructive sleep apnea. The electroencephalogram profile during non-rapid eye movement sleep may be a useful biomarker for predicting vulnerability to daytime impairment in obstructive sleep apnea and responsiveness to treatment.