Successful continuous positive airway pressure treatment reduces skin sympathetic nerve activity in patients with obstructive sleep apnea

Abstract

Background

Obstructive sleep apnea (OSA) is associated with cardiovascular diseases and increased sympathetic tone. We previously demonstrated that OSA patients have increased skin sympathetic nerve activity (SKNA).

Objectives:

To test the hypothesis that continuous positive airway pressure (CPAP) treatment reduces SKNA.

Methods:

The ECG, SKNA, and polysomnography were recorded simultaneously in 9 OSA patients. After baseline recording, CPAP titration was performed and the pressure was adjusted gradually for the optimal treatment, defined by reducing the apnea-hypopnea index (AHI) to ≤ 5/h. Otherwise the treatment was considered suboptimal (AHI > 5/h). Fast Fourier transform (FFT) analyses were conducted to investigate the frequency spectrum of SKNA.

Results:

There were very low frequency (VLF), low frequency (LF) and high frequency (HF) oscillations in the SKNA. The HF oscillation matched the frequency of respiration. OSA episodes were more frequently associated with the VLF and LF than HF oscillations of the SKNA. Compared with baseline, CPAP significantly decreased the arousal index, AHI, and increased the minimal and average O₂ saturation. Optimal treatment significantly increased the dominant frequency (DF), and reduced the heart rate, average SKNA (aSKNA), SKNA burst duration and total burst area. The DF negatively correlated with aSKNA.

Conclusions:

VLF, LF and HF oscillations are observed in human SKNA recordings. Among them, VLF and LF are associated with OSA while HF is associated with normal breathing. CPAP therapy reduces the aSKNA and shifts the frequency of SKNA oscillation from VLF or LF to HF.

Introduction

Obstructive sleep apnea (OSA) is associated with increased risk of heart failure, hypertension, coronary artery disease, stroke, cardiac arrhythmia and sudden cardiac death (SCD).^1–3 OSA increases sympathetic tone, which is an important risk factor for cardiovascular diseases.^{4, 5} Continuous positive airway pressure (CPAP) is the standard therapy for OSA, but its efficacy in cardiovascular mortality has not been consistently demonstrated. It is possible that a reduction in sympathetic tone in patients with OSA will result in benefit. Identifying these patients in the clinical setting presents a challenge. Absence of a biomarker for measuring sympathetic tone in clinical practice limits the physicians’ ability to evaluate the sympathetic nerve activity (SNA) in the clinical setting. Microneurography studies showed a reduction of the muscle SNA during wakefulness in patients compliant with CPAP therapy.⁶ However, because of the technical difficulties, it is not practical to use microneurography to estimate the efficacy of CPAP therapy clinically. We invented a method (neuECG) to simultaneously record electrocardiogram (ECG) and skin sympathetic nerve activity (SKNA) using standard ECG patch electrodes and it has been verified in animal model and human subjects.^7–9 In addition to the amplitude modulation, the human SNA measured with microelectrodes show frequency modulation.^10–12 Consistent with those studies, we⁸ found similar frequency spectra in ambulatory canine SNA. The high frequency (HF) oscillations in blood pressure and heart rate (HR) correlate with the HF oscillations in the stellate ganglion nerve activity. These HF oscillations are background nerve activity and are present at all times. HF oscillations can be overshadowed by the much larger low frequency (LF) and very low frequency (VLF) burst activities. We recently performed neuECG recordings during sleep studies and found that SKNA was higher in OSA patients than normal controls⁹. However, whether successful CPAP therapy can reduce SKNA or change the frequency spectra of SKNA remains unclear. The purpose of the present study was to determine the immediate SKNA response to CPAP treatment in moderate to severe OSA patients. The results are used to test the hypothesis that CPAP therapy significantly reduces the average SKNA and alters the frequency spectrum of the SKNA.

Methods

Participants

The present research was approved by the Institutional Review Board of the Indiana University School of Medicine. Nine patients undergoing split-night polysomnography at the Indiana University Health Sleep Disorders Center between 2017 and 2019 were included in this report. The data of the baseline sleep study (before CPAP) in 6 of the 9 patients were included in a previous report⁹.

Sleep study protocol

Sleep studies were conducted per American Academy Sleep Medicine (AASM) recommended protocol. The sleep study recording was described in a previous manuscript.⁹ The detailed methods of the sleep study and the SKNA analyses can be found in an Online Supplement. Briefly, the Alice® 6 Diagnostics Sleep System (Philips-Respironics, Amsterdam, Netherlands) was used for the polysomnography recording. For the patients with moderate to severe obstructive sleep apnea, when the apnea-hypopnea index (AHI) was more than 20/h in first few hours of sleep study, CPAP was applied the same night. CPAP was started at 4 cm H₂O and gradually raised to eliminate apneas, hypopneas and snoring. The treatment when AHI was less than 5/h was considered as optimal treatment. The remaining CPAP treatment (AHI>5/h) was defined as suboptimal treatment in this study.

SKNA analyses

We recorded SKNA from electrocardiographic patch electrodes in the chest to record ECG channel 1 with 10,000 samples/s. The data were then manually analyzed to exclude periods motion artifacts or with absent ECG signals (poor electrode contacts). The electrical recordings were band pass filtered between 0.05–150 Hz to display ECG and between 500–1000 Hz to display SKNA using computer software described in previously published methods.⁸ The average SKNA (aSKNA) is the average voltage of SKNA of each sample. The frequency spectra are analyzed using the Labchart 8 (ADInstruments Ltd, Bella Vista, Australia).⁸ More details are provided in the supplemental materials.

Statistics

The data with normal distribution were presented as mean ± standard deviation. Otherwise they were presented as median [interquartile range]. Statistics were performed using GraphPad Prism 8 (GraphPad Software, Inc., La Jolla, USA). Repeated measures ANONA was used to compare the values among 3 different groups (baseline, suboptimal and optimal treatment), with nonparametric test used to compare the non-normal distributed values. Spearman’s correlation was used to analyze the correlations between SKNA and dominant frequency (DF). P ≤ 0.05 was considered as statistically significant.

Averages of SKNA per 30 seconds were analyzed using a linear mixed-effects model (LMM), with subject as a random effect to account for correlations among repeated measurements of SKNA. Sleep stage (with levels wakefulness, N1-N3 and REM) and the optimality of CPAP treatment (baseline, suboptimal and optimal treatment) were included as the fixed effects. The maximum likelihood estimation was used the computation of the model. The LMM approach was also employed to assess CPAP’s effect on heart rate.

Results

The patient characteristics are shown in Table 1. The patient population included 2 women and 7 men, consistent with the clinical observations that moderate to severe obstructive sleep apnea is more prevalent in men than in women. Table 1 also shows racial diversity of the study population.

Table 1.

Demographics of patients.

	Gender	Age	BMI	Race	Diabetes	Insulin	Hyperten sion	Heart failure	coronary artery disease	β-blocker	Calcium channel blocker	ACEI/ARB
Patient 1	Female	52	36.9	Black	No	-	No	Yes	No	Yes	No	Yes
Patient 2	Male	69	30.7	white	Yes	No	Yes	No	No	No	No	No
Patient 3	Male	31	33.2	Black	Yes	No	Yes	No	No	No	No	Yes
Patient 4	Male	29	36.9	Black	No	-	No	No	No	No	No	No
Patient 5	Female	57	19.6	white	No	-	No	No	No	No	No	No
Patient 6	Male	53	42	white	No	-	Yes	No	Yes	Yes	No	Yes
Patient 7	Male	64	40.2	White	Yes	Yes	Yes	No	No	Yes	No	No
Patient 8	Male	52	34.1	Black	No	-	Yes	No	No	No	No	No
Patient 9	Male	52	35.4	Black	Yes	Yes	Yes	Yes	Yes	Yes	Yes	No

HF, LF and VLF oscillatory patterns of SKNA

Our previous study⁸ in canine models indicate the SNA has a regular background HF oscillation which could be overshadowed by LF and VLF oscillations, most of which are burst activities. Similar patterns were observed in the present study. Fig. 1A showed the typical SKNA recording in OSA patients, during which we can see different oscillatory patterns of HF (0.15–0.40 Hz), LF (0.03–0.15 Hz) and VLF (0–0.03 Hz). In Fig. 1B (zoomed in from the red dotted box region of Fig.1A), a HF oscillatory pattern (about 0.19 Hz) was present and not overshadowed by the larger VLF and LF components. The frequency spectrograms of integrated SKNA (iSKNA) in Fig.1C show the power density of different frequency components from the recording window shown in Fig.1A. The highest power was around 0.19 Hz in the beginning of the recording. The VLF and LF bands were present in later portion of the recording. In Fig.1D, the peak power density frequency is at about 0.19 Hz, in the HF range. However, there are also peaks in VLF and LF ranges.

Figure 1.

SKNA oscillatory patterns detected in OSA patients. A. Raw tracing of SKNA signals along with iSKNA, ECG, and heart rate. The HF, LF, and VLF oscillatory patterns could be observed in SKNA and iSKNA channels. The regular HF component served as a background which could be overshadowed by larger LF and VLF bursts; B. Signals zoomed in from the red dotted box region from panel A. These HF bursts of SKNA with a frequency at about 0.19 Hz were not overshadowed by VLF or LF bursts at night, and were clearly visible in the SKNA and iSKNA channels; C. The frequency spectrograms of iSKNA based on panel A; D. Frequency distribution of signals from panel A shows the VLF, LF and HF ranges and the dominant frequency (DF) is 0.19 Hz (HF) during this episode. VLF: Very low frequency (0–0.03Hz); LF: Low frequency (0.03–0.15Hz); HF: High frequency (0.15–0.40Hz).

OSA increases aSKNA and reduces the dominant frequencies of SKNA

We synchronized the polysomnography recording with the SKNA recording in Fig. 2A. The air flow measurement shows regular stable respiration with breathing rate of 16.2/min (0.27 Hz) in the first half of the recording. At the same time, there was oscillation of the SKNA at the same frequency. The red triangle marks the time of a large SKNA burst that occurred coincidentally with an arousal in the electroencephalogram (EEG) channel, followed by the typical breathing pauses/sleep apnea events (black arrows) in the second half of the recording with a frequency of about 0.015 Hz on the flow channel. The SKNA and iSKNA channels also show oscillatory patterns at the same frequency. Panel B shows the frequency spectrum of the iSKNA in the entire recording. The first half (during stable respiration) showed HF in synchrony with respiration while the second half shows increased power of VLF oscillations which is associated with sleep apnea events. Panels C and D were based on the black and red dotted boxes in panel A respectively, showing DF at 0.27 Hz during normal respiration and 0.015 Hz during apneic events.

Figure 2. Effects of sleep apnea on the SKNA frequency spectra.

A. Tracings of the sleep study and simultaneously recorded SKNA. The first half shows stable normal respiration while the second half shows apnea episodes; B. The frequency spectrograms of iSKNA based on panel A showing HF band was dominant on the left while VLF band was dominant on right half of the figure; C and D. The frequency distribution and power density of signals from black and red dotted boxes, respectively in Panel A.

CPAP treatment significantly improved sleep structure, increased DF, decreased aSKNA and HR

Comparing to baseline sleep in moderate to severe obstructive sleep apnea patients, CPAP treatment reduced arousals, improved sleep structure and increased minimal and mean oxyhemoglobin levels (Fig 3). Fig. 4 demonstrates the relationship of SKNA burst, obstructive sleep apnea and CPAP treatment. Optimal CPAP therapy was associated with reduction of aSKNA and HR (Fig 3, Table S1 and S2), as well as shifting the SKNA frequency from VLF/VF range to HF range (Fig 4–5). The actual recordings at baseline and after optimal therapy are shown in B and E, respectively. Both channel 1 and channel 2 (on the left arm) recorded LF and VLF SKNA oscillations at baseline and the oscillation frequencies were in LF and VLF range. The SKNA bursts on the chest wall corresponded to the intermittent respiratory frequencies during sleep apnea but the SKNA bursts on the arm has lower number of bursts. The dominant frequencies on both channels were in the VLF/LF range. After optimal CPAP therapy, the SKNA oscillates in HF range in channel 1 (F), but these changes were less apparent in channel 2 (G).

Figure 3. CPAP improved the sleep structure, increased the SpO₂, decreased arousal index and AHI.

A. The sleep structure alteration of each stage before and after CPAP treatment; B-C. The average SpO₂ and minimal SpO₂ levels before and after CPAP treatment; D-E. The arousal index and AHI alteration during different sleep stages before and after CPAP treatment. BS: baseline; Sub: Suboptimal CPAP treatment; Opt: Optimal CPAP treatment. REM: rapid eye movement sleep stage; N1-N3: non-REM sleep stage 1–3. *, P<0.05, and **, P<0.01 compared with baseline; #, P<0.05 compared with suboptimal CPAP treatment.

Figure 4. The effects of CPAP treatment on aSKNA, heart rate, and dominant frequency.

A. Representative tracing showing the effects of CPAP therapy on various parameters. The temporal changes of aSKNA, heart rate, and dominant frequency (DF) were calculated every 3-min. Horizontal dotted line (0.15Hz) marks the boundary between the HF and LF. B shows tracings zoomed in from red box in A. Intermittent bursts of SKNA were observed in both channels 1 and 2. The power densities of iSKNA in channel 1 (C) and 2 (D) show that VLF and LF are the dominant frequencies. E shows the signal from the black box in A. The power densities of iSKNA in channel 1 (F) shows that the optimal CPAP therapy moved the dominant frequency from VLF/LF to HF range. The same change was not observed in channel 2.

Figure 5.

CPAP treatment increased dominant frequency (DF), decreased apnea events, heart rate, and aSKNA. A. The dominant frequency before and after CPAP treatment; B-D. The percentage of VLF, LF, and HF dominant episodes before and after CPAP treatment; E-F. The aSKNA and heart rate during different sleep stages before and after CPAP treatment; G-H. The burst duration and burst area of SKNA before and after CPAP treatment. BS: baseline; Sub: Suboptimal CPAP treatment; Opt: Optimal CPAP treatment. *, P<0.05, and **, P<0.01 compared with baseline; #, P<0.05, and ##, P<0.01 compared with suboptimal CPAP treatment.

We also calculated the incidence of sleep apnea every 3-min and found that the proportion of apnea free periods during VLF and LF dominant episodes were 48.9% (14.7% - 69.6%) and 52.9% (21.4% - 72.7%) respectively, but significantly increased to 79.3% (56.8% - 95.1%) during HF dominant episodes (Table 2).

Table 2.

The percentage of apnea-free periods during VLF, LF, and HF dominant episodes.

	VLF		LF		HF
	Apnea	Apnea free	Apnea	Apnea free	Apnea	Apnea free
Patient 1	73.3%	26.7%	27.3%	72.7%	4.9%	95.1%
Patient 2	39.2%	(60.8%)	43.1%	56.9%	15.4%	84.6%
Patient 3	-	-	47.1%	52.9%	20.7%	79.3%
Patient 4	85.3%	14.7%	78.6%	21.4%	5.5%	94.5%
Patient 5	34.4%	65.6%	60.7%	39.3%	43.2%	56.8%
Patient 6	35.3%	64.7%	42.6%	57.4%	41.3%	58.7%
Patient 7	30.4%	69.6%	57.1%	42.9%	29.0%	71.0%
Patient 8	82.4%	17.6%	57.1%	42.9%	27.3%	72.7%
Patient 9	63.0%	37.0%	41.8%	58.2%	16.7%	83.3%
Summary	51.1% (30.4% – 85.3%)	48.9% (14.7% – 69.6%)	47.1% (27.3% – 78.6%)	52.9% (21.4% – 72.7%)	20.7% (4.9% – 43.2%)	79.3% (56.8% – 95.1%)

VLF: Very low frequency; LF: low frequency; HF: high frequency.

The correlation between DF and aSKNA

The statistical analyses indicate optimal but not the suboptimal CPAP treatment significantly increased the frequency of DF of iSKNA (Fig. 5A). The percentage of HF dominant episodes were also significantly increased during optimal CPAP treatment (Fig. 5D). Fig. 5E–F indicate the aSKNA and heart were both significantly decreased during every sleep stage except N3 or rapid eye movement (REM) because not all patients had N3 or REM sleep stages. It should be noted that more patients had N3 sleep after CPAP treatment. Fig. 5G–H showed that the burst duration and burst area were both significantly reduced by the optimal but not by the suboptimal CPAP treatment. Significant negative correlations between aSKNA and DF with the r value as low as −0.779 (Fig. 6) was found in 7 patients.

Figure 6.

Correlation between aSKNA and DF. A-I. The correlation between aSKNA and DF of all 9 patients: there is a significant negative correlation between aSKNA and DF in 7 of 9 patients.

Discussion

Major finding

Similar to previous findings in ambulatory canines,⁸ we found that the patterns of VLF, LF, and HF could be detected in the human SKNA. The HF oscillation in SKNA matched with the frequencies of normal respiration. These findings are consistent with the notion that loading/unloading of the carotid baroreceptors are associated SNA variations.¹¹ We noted that most of the apneic events happened during LF and VLF dominant periods, while very few apnea events happened when HF was the DF. In addition, optimal CPAP therapy shifts the DF from LF/VLF to HF range, along with a significant reduction of arousal index and AHI, as well as the increase of SpO₂. We also found that SKNA burst duration and area, aSKNA, and HR were significantly reduced by optimal CPAP treatment. There is a negative correlation between DF and aSKNA values.

Optimal CPAP treatment significantly improved sleep structure and altered the spectra of SKNA

Multiple studies have shown that moderate to severe obstructive sleep apnea increases cardiovascular risk. The outcomes of positive airway treatment in reducing cardiac events or all-cause mortality were inconsistent. The difference in sleep apnea severity, treatment modalities (such as CPAP, BiLevel Positive airway therapy, adaptive servo ventilation) and treatment adherence (ranged from 1.4 h/day to 6.6 h/ day) may all contribute to the different prognosis.¹³ Our study found effective CPAP treatment immediately reduces SKNA and changes its DF to HF range. Bigger et al found that VLF analysis was strongly associated with cardiac arrhythmia and all cause mortalities.¹⁴ We found VLF was significant increased in moderate to severe OSA patients. CPAP treatment not only reduced the total aSKNA level but also shifted the DF from VLF to HF, which was the DF associated with the baseline normal respiratory pattern.

Spectral analyses of SKNA in OSA patients

Spectral analysis of heart rate variability (HRV) has been used as indirect indices of autonomic functions for many years. However, the value of HRV analysis on SNA has been challenged. For example, Cottin et al found the LF/HF of R-R interval is not a good indicator of autonomic modulation during exercise.¹⁵ Rahman et al showed the LF power of HRV was unrelated to cardiac sympathetic innervation thus could not be an index of cardiac sympathetic tone.¹⁶ The power of VLF, accounting for most of the total power sometimes, were usually ignored. With neuECG, SNA is directly recorded. The frequency analyses further revealed that the DF of SKNA shifts to VLF or LF range during sleep apneas and to HF range with optimal CPAP treatment. Investigation of the roles of these signals might provide new insights about the neurophysiology associated with OSA.

The respiration-SKNA coupling in OSA patients

The respiratory modulation of SNA has drawn the attention of researchers for years¹⁷: a respiratory rhythm has been found in the blood pressure, R-R interval, muscle SNA and skin SNA spectra.^{18, 19} The phenomenon of respiratory sinus arrhythmia and a within-breath sympathetic variance with a peak at the end-expiration and a valley at the end-inspiration/early expiration presents in all different types of ventilation.¹¹ The postganglionic sympathetic nerve activity could be affected both by the tidal volume and the respiratory rate,^19–21 and nocturnal respiratory rate is an independent predictor for the outcomes in patients with acute coronary syndrome.^{22, 23} Consistent with these results, the SKNA bursts changed along with sleep apneic events and a strong SKNA-respiration coupling was found in the present study. Under stable normal respiratory pattern, the DF of SKNA matches with the frequency of respiration. When the sleep apnea events occurred, the DF shifted to LF or VLF, which were also consistent with ventilation periodicities, accompanied by the increase of AHI and decrease of SpO₂. Compared with the HF, more sleep apnea events occurred during VLF and LF dominant episodes. Therefore, the highly peak to peak respiration-SKNA coupling indicates the spectral analysis of SKNA might reflect the change of the respiratory patterns and indicate the sleep apnea incidence in OSA patients.

Potential mechanism of respiration-SKNA coupling

The underlying mechanism of the respiration-SKNA coupling is complicated. Studies show that a common central origin of respiration and SNA may exist. Shigeo et al found that muscle and skin SNA are usually accompanied by K-complexes on EEG induced by spontaneous or arousal stimulus during sleep stages.²⁴ Sound stimuli during different respiratory phases induced different sympathetic responses.^{25, 26} Svein et al found general anesthesia could alter the oscillatory components of human microcirculation related to the SKNA or MSNA, as well as the respiration related oscillatory components.²⁷ Anatomical studies revealed the neural network connecting the respiratory system and sympathetic nervous system from the medulla, especially the rostral ventrolateral medulla.^{28, 29} The mechanism of the respiration-SKNA coupling was not involved in the present study, but we observed large bursts from the EEG channel accompanied by SKNA bursts and apnea episodes, which was consistent with previous findings from others. These findings indicate that the respiration-SKNA coupling may be regulated by a common central command.

Clinical implications

neuECG is a convenient and noninvasive method to record the SKNA signals, making it feasible for large scale clinical studies to assess SNA and cardiovascular diseases. The strong respiration-SKNA coupling and oscillatory changes of VLF signal found in the present study indicates the spectra of SKNA could correlate with respiratory events. These findings suggest that SKNA might be used as a potential biomarker for obstructive sleep apnea. Considering the rapid development of smart wearable medical equipment and the computing power of chips, the SKNA may provide a more feasible choice as a new home sleep study device and improve the ease of diagnosis of OSA.

Study limitations

We did not record the parasympathetic nerve activity in this study. Therefore, we are not able to evaluate the changes of parasympathetic nerve activity and correlate that activity with the HF oscillations. It remains possible that HF oscillations are coupled with increased parasympathetic activity. The number of patients is too small to evaluate the effects of diabetic neuropathy on the SKNA. Further studies is needed to determine if diabetes affects the SKNA activities in obstructive sleep apnea patients.

Conclusion

VLF, LF and HF oscillations are observed in human SKNA recordings. Among them, VLF and LF are associated with OSA while HF is associated with normal breathing. Optimal CPAP therapy reduces the average SKNA and shifts the frequency of SKNA oscillation from VLF/LF to HF.