Methods

All subjects provided written informed consent. The study was approved by our institutional review board. Nonsmoking subjects aged 18 to 70 years with body mass indices ≥30 kg/m ² and without known cardiac, pulmonary, endocrine, or non-OSA sleep disorders were recruited from the community by advertisements posted in local newspapers and at local medical clinics. Subjects were ineligible if taking any oral contraceptives, hormone replacement therapy, sedatives, steroids, nonsteroidal anti-inflammatory agents, lipid-lowering medications, or antihypertensive medications. Potentially eligible subjects then underwent screening examinations by a licensed physician and were excluded from the study if blood counts, fasting glucose, thyroid-stimulating hormone, lipid panel, or blood pressure measurements were markedly abnormal. All subjects were also asked to keep a 2-week sleep diary to ensure regular sleep patterns prior to the study.

Subjects underwent standard in-laboratory polysomnography; recorded signals included electroencephalography (F4-A1, F3-A2, C4-A1, C3-A2, O2-A1, and O1-A2), left and right electrooculography, submental and bilateral tibial electromyography, surface electrocardiography, airflow, chest and abdominal excursion, oxyhemoglobin saturation, and body position. All data were collected using a digital polysomnographic system (Nihon Kohden, Foothill Ranch, California). Polysomnograms were scored by a blinded, registered sleep technician according to standard criteria. An apnea was scored if airflow was absent for 10 seconds, and a hypopnea was scored if there was a ≥50% reduction in airflow for 10 seconds or a discernable decrement in airflow for 10 seconds in association with either an oxyhemoglobin desaturation of ≥3% or an arousal.

Laboratory measurements were undertaken during the morning after polysomnography, while patients were fasting. Blood samples were obtained in sterile fashion from the antecubital vein. Total serum cholesterol and triglycerides were measured using the Synchoron CX analyzer (Beckman Systems, Fullerton, California). Serum high-density lipoprotein was measured directly (Sigma, St. Louis, Missouri), and low-density lipoprotein was calculated. Whole-blood glycosylated hemoglobin was measured using ion-exchange high-performance liquid chromatography.

Each subject had a 5- to 17-minute single-lead electrocardiogram recorded at 1,000 Hz while (1) supine while basal breathing, (2) supine while breathing was paced by an electronic tone at 12 breaths/min (0.2 Hz), and (3) standing while basal breathing. Continuous electrocardiographic data were recorded using Spike2 acquisition software (1401plus; Cambridge Electronic Design, Cambridge, United Kingdom), then extracted and converted to waveform database format. An automated QRS detector (ecgpuwave; Physionet, Cambridge, Massachusetts) was run on these rescaled electrocardiographic signals, and every beat was annotated as normal or ectopic. The annotation file with the fewest number of ectopic beats was selected for further analysis, and the time series of NN sinus intervals was extracted. The first 5 minutes of stable, stationary data were chosen for analysis from the 3 awake conditions. Outliers due to false or missed normal beat detections were removed using a sliding window average filter with a window of 41 data points and rejection of central points lying outside 20% of the window average. On average, the resulting annotation files contained 98.4% normal beats.

From the filtered NN interval files, time-domain statistics including the average of all NN intervals, the standard deviation of all NN intervals, the root mean square of successive NN intervals, and pNN50, pNN20, and pNN10 were calculated. Standard frequency-domain measures were also calculated, including low-frequency (LF) power (spectral power of all NN intervals between 0.04 and 0.15 Hz), high-frequency (HF) power (spectral power of all NN intervals between 0.15 and 0.4 Hz), the ratio of LF to HF power, and, for paced breathing, power at the paced breathing frequency (total spectral power of all NN intervals between 0.18 and 0.22 Hz). To eliminate the need for evenly sampled data, which are required by the standard fast-Fourier transform, frequency-domain measures were calculated using the Lomb periodogram for unevenly sampled data.

Approximately 45 minutes after patients awoke, simultaneous electrocardiographic and impedance cardiographic recording was also performed. After skin preparation, 4 impedance electrode strips were placed in a standard circumference tetrapolar band configuration, and 3 electrocardiographic leads were placed on the right and left upper chest and the left lower quadrant. Electrocardiographic and dZ/dt signals were sampled at 512 Hz and ensemble-averaged over 1-minute epochs. PEPs were recorded as five 1-minute averages during each of the awake conditions described previously. PEP was measured with the HIC 3000 Bioelectric Impedance Cardiograph System with Cop-Win/HRV version 6.0 data acquisition software (Bio-Impedance Technology, Inc., Chapel Hill, North Carolina).

Continuous data were inspected for normality of distribution and homogeneity of variance and compared between groups using independent Student’s t tests or Mann-Whitney U tests as appropriate. Categorical data were compared using chi-square tests. Missing data were excluded pairwise for all analyses. To determine independent predictors of various time-domain HRV indexes and PEP data, multiple linear regression forced-entry models were built. Multicolinearity between predictors was investigated using simple Pearson’s correlations and inspection of tolerance and variance inflation factor values. Finally, standardized residuals were inspected to identify potential outliers. All reported p values are 2 sided, and p values ≤0.05 were considered significant.