Sleep-Mediated Heart Rate Variability after Bilateral Carotid Body Tumor Resection

Abstract

Study Objectives:

The carotid bodies are thought to play an important role in sleep-dependent autonomic changes. Patients who underwent resection of bilateral carotid body tumors have chronically attenuated baroreflex sensitivity. These subjects provide a unique opportunity to investigate the role of the baroreflex during sleep.

Design:

One-night ambulatory polysomnography (PSG) recording.

Setting:

Participants' homes.

Participants:

Nine patients with bilateral carotid body tumor resection (bCBR) (four women, mean age 50.4 ± 7.2 years) and nine controls matched for age, gender, and body mass index.

Interventions:

N/A.

Measurements:

Sleep parameters were obtained from PSG. Heart rate (HR) and its variability were calculated using 30-s epochs.

Results:

In bCBR patients, HR was slightly but not significantly increased during wake and all sleep stages. The effect of sleep on HR was similar for patients and controls. Low frequency (LF) power of the heart rate variability spectrum was significantly lower in bCBR patients in active wakefulness, sleep stage 1 and REM sleep. No differences were found between patients and controls for high frequency (HF) power and the LF/HF ratio.

Conclusions:

Bilateral carotid body tumor resection (bCBR) is associated with decreased low frequency power during sleep, suggesting impaired baroreflex function. Despite this, sleep-related heart rate changes were similar between bCBR patients and controls. These findings suggest that the effects of sleep on heart rate are predominantly generated through central, non-baroreflex mediated pathways.

Citation:

Niemeijer ND, Corssmit EP, Reijntjes RH, Lammers GJ, van Dijk JG, Thijs RD. Sleep-mediated heart rate variability after bilateral carotid body tumor resection. SLEEP 2015;38(4):633–639.

INTRODUCTION

Physiological sleep-dependent autonomic changes result from a complex interaction of peripheral cardiovascular reflexes and central modulation.^1,2 The baroreflex arc, with arterial baroreceptors mainly located in the carotid sinuses and aortic arch, is considered to be the critical relay in this complex integration.³ Baroreflex sensitivity is continuously modulated and differs markedly between behavioral and physiological conditions, including sleep.^1,3,4 During NREM sleep, a progressive decrease is seen in peripheral sympathetic nerve activity and blood pressure (BP), together with a decrease in heart rate (HR).^1,2,5,6 The latter sign suggests increased parasympathic vagal activity. Conversely, a net increase of HR and BP has been reported during REM sleep.^1,2,5,6 This increase is accompanied by irregular changes in autonomic activity.¹

Paragangliomas are rare neuroendocrine tumors of paraganglia, which are neural-crest derived chromaffin tissues associated with the autonomic nervous system.⁷ Paragangliomas in the head and neck region can arise from the carotid body, vagal body, or jugulotympanic tissue (i.e., paraganglioma of the temporal bone).^8,9 Due to their location in close proximity to important neurovascular structures, tumor growth may lead to serious morbidity and cranial nerve impairment. These tumors can be removed without recurrence.¹⁰ However, branches of the carotid sinus nerves may not be spared. Bilateral carotid body tumor resection (bCBR) may thus result in arterial baroreflex dysfunction.¹¹ Patients with bCBR are known to have significant lower baroreflex sensitivity compared with controls, i.e., a less marked heart rate response to a given rise or fall in blood pressure.¹¹ Baroreflex failure, whether from carotid endarterectomy, head and neck irradiation, mixed cranial nerve neuroma, neurosarcoidosis, or brain stem stroke, is associated with changes in heart rate variability (HRV).^12–17 Notably, these patients have little “low-frequency” (LF) power, an index of baroreflex-mediated HR control.^18,19 This parallels findings in mouse models, where carotid sinus denervation resulted in lower values of LF power and baroreflex sensitivity.²⁰ So far, no data are available on the effects of sleep on HRV following bCBR. These patients provide a unique opportunity to study the role of the baroreflex in sleep. We therefore monitored HR and HRV during nocturnal sleep in bCBR patients and compared them with controls matched for age, gender, and body mass index (BMI).

METHODS

Subjects

We included 9 patients who had previously been treated with bCBR. These patients were recruited from the outpatient clinics of the departments of Endocrinology, Otorhinolaryngology and Surgery of the Leiden University Medical Center and through an advertisement on the website of the Dutch paraganglioma patient network. For each patient we recruited a healthy control subject matched for gender, age (± 5 years), and BMI (± 3 kg/m²).

Exclusion criteria were the presence of a pheochromocytoma, extra-adrenal paraganglioma, history of a psychiatric disorder, history of a diagnosed sleep disorder, or the use of sleep medication.

The Medical Ethics Committee of the Leiden University Medical Center approved the study protocol. All subjects provided written informed consent prior to the study.

Study Design

Polysomnography

Sleep was recorded at home using a portable polysomnography recorder (Somnologica Version 5.1.1, Embla, CO, USA). The measurement started at 16:00 and lasted for 24 hours. Duration of “active wakefulness,” sleep stages 1–3, and REM sleep were assessed. Active wakefulness was defined as all wake epochs occurring between the period from 16:00 until 20 min before onset of nocturnal sleep, and the period from 20 min after awakening in the morning until the end of the measurement.

Sleep stages and apnea-hypopnea events were manually scored in 30-s epochs by an experienced sleep technician, according to the guidelines of the American Academy of Sleep Medicine.²¹ The polysomnograms were analyzed by a technician who was blinded to the diagnosis of the subject. The autonomic parameters were analyzed automatically.

Respiration was monitored with a nasal pressure sensor and 2 elastic bands (thorax and abdomen). Oxygen saturation was assessed continuously with a pulse oximeter attached to the index finger. Apneas were defined as a drop in the peak thermal sensor excursion by ≥ 90% of baseline for > 10 s. Hypopneas were defined as a drop in the nasal pressure signal excursion by ≥ 30% of baseline for > 10 s, with a ≥ 4% desaturation from the pre-event baseline.

Electrocardiography and Respiration

A continuous wavelet transform was implemented in Matlab (Version 13.1, Mathworks, MA, USA) to detect R-peaks in the electrocardiogram.²² A filter was used to exclude outliers, with outliers defined as values that differed > 25 beats/min from the previous or next sample. The signal was resampled at 5 Hz.

R-peak detection resulted in a series of consecutive R-R intervals, split into consecutive 30-s epochs for analysis. For each epoch the mean HR was calculated and a frequency spectrum by creating time-frequency domains through fast Fourier transform. From the frequency spectrum, the LF (0.04–0.15 Hz) and high-frequency (HF) (0.15–0.4 Hz) power component were calculated. The HF component is considered to represent vagal activity, and the LF component to reflect baroreflex-mediated sympathetic activity.^18,23 LF/HF ratio was computed as a reflection of the sympathovagal balance.

Autonomic Parameters

LF power and HR were selected as main outcome parameters. Secondary outcome measures included HF and LF/HF ratio.

Selection of Epochs

All epochs of sleep following onset of nocturnal sleep and prior to awakening in the morning were included in our analysis. For active wakefulness, a selection of epochs was made, as the length of the active wakefulness period proved to exceed those of the sleep states considerably, which could affect the results. We therefore nullified this effect by limiting the number of epochs of active wakefulness to that of stage 2 epochs for each subject. The wake epochs were selected in a random fashion. To account for the effects of arousals, we labeled every epoch following a transition from sleep stage 3 to stage 2 or 1 and sleep stage 2 to stage 1. We defined these epochs as “arousal transitions.” In addition, we identified all epochs that coincided with either an apnea or a hypopnea to study the autonomic effects of respiratory arousals.

Statistical Analysis

Overall we included an average total number of 1,269 epochs per subject. For each epoch the autonomic parameters, sleep/wake stage, subject number, β-blocker use, presence of bCBR, arousal transitions, and apnea-hypopnea (Figure 1) were recorded and entered into our model. For each autonomic parameter a linear mixed effects regression model was formed to describe the effect of sleep stages and bCBR on the autonomic parameters. To obtain a normal distribution of the residuals of the models, a natural logarithm-transformation was applied. Sleep stage, β-blocker use, apneas-hypopneas, arousal transitions, and the interaction between bCBR and sleep stages were entered as fixed and subjects as random effects to the model. Because of expected differences in variability between sleep stages and between patients and controls, the random subject effects were stratified for sleep stage/wakefulness (5 levels) and patient status (2 levels).

Figure 1.

Schematic diagram illustrating our data analysis. For each subject, we analyzed the heart rate data during overnight sleep and a random selection of active wakefulness (lower part of the figure). Overall we included an average total number of 1,269 epochs per subject. The upper part of the figure zooms in on a representative 1-min segment illustrating how the heart rate was calculated from the one-channel ECG data. For each epoch the autonomic parameters, sleep/wake stage, subject number, β-blocker use, presence of bCBR, arousal transitions, and apnea-hypopnea were recorded and entered into our model.

Outliers with a standardized residual at a distance > 2.5 standard deviations from 0 were excluded from the linear mixed effects regression model.

P values below 0.05 were considered to be significant. All statistical analyses were performed using R (Version 3.0.0, R).

RESULTS

Participant Characteristics

The bCBR group comprised 6 patients with a mutation in subunit D of the SDH gene (SDHD), one obligate SDHD mutation carrier, one patient tested negative for germline mutations in SDH genes, and one patient who had not been genetically tested (Table 1). In the bCBR patients, resection of the first carotid body tumor (CBT) was performed 12.5 ± 7.6 years (range 2.2–25.2 years) and resection of the second CBT 8.9 ± 6.8 years (range 1.2–21.5 years) prior to current study. Three patients had additional head and neck paragangliomas. Two patients had a vagal body tumor and one patient had a jugular foramen tumor. Two patients used β-blockers (bisoprolol 2.5 mg once daily and propranolol, unknown dosage). None of the participants were shift-workers or were known to have arrhythmias.

Table 1.

Clinical characteristics of nine patients with bilateral carotid body tumor resection and their matched healthy controls.

Sleep Parameters

bCBR patients spent significantly more time in sleep stage 1 than controls (12.6% versus 7.5%; P < 0.05). Apart from this difference, polysomnography parameters were similar between groups. As reported previously, no significant differences in apnea-hypopnea index between patients and controls were found (Table 2).²⁴

Table 2.

Polysomnography results of nine patients with bilateral carotid body tumor resection and their matched healthy controls.

Autonomic Parameters

In bCBR patients, HR was slightly but not significantly increased during wake and all sleep stages (Table 3, Figure 2). The effect of sleep on HR was similar for patients and controls. LF power of the HRV spectrum was significantly lower in bCBR patients during active wakefulness (P < 0.05), sleep stage 1 (P < 0.01), and REM sleep (P < 0.05). (Table 3, Figure 2). No differences were found between bCBR patients and controls for HF power and the LF/HF ratio.

Table 3.

Regression coefficients derived from the linear mixed effects model with the effects of sleep on cardiovascular parameters.

Figure 2.

Effects of sleep on heart rate (HR) and low frequency (LF) power of heart rate variability in nine patients with bilateral carotid body tumor resection (dashed, blue lines) and their matched healthy controls (solid, red lines). *P < 0.05; **P < 0.01. Bars represent standard error of the mean. HR, heart rate; REM, rapid eye movement; ln, natural logarithm; LF, low frequency power; Active, active wakefulness.

DISCUSSION

We found that the LF component of HRV was significantly lower during active wakefulness, sleep stage 1, and REM sleep in bCBR patients compared to controls, reflecting baroreflex dysfunction. Interestingly, in spite of these signs of baroreflex dysfunction, the effect of sleep on HR was similar in bCBR patients and their matched controls. These findings suggest that the sleep-related HR decrease primarily results from nonbaroreflex mediated pathways.

Sleep Studies

Patients with bCBR spent significantly more time than controls in sleep stage 1. This increase in light sleep could not be explained by an increased prevalence of sleep disordered breathing.²⁴ Whether these findings are of clinical importance is disputable, as no differences were found in measures of daytime sleepiness between bCBR patients and controls.²⁴

Baroreflex and Sleep

As in previous daytime studies, bCBR patients had a lower LF indicating baroreflex failure.^12–17 Differences in LF power during sleep between bCBR patients and controls were most apparent in sleep stage 1 and during REM sleep. During deeper sleep (sleep stages 2 and 3), this difference was less marked. This contrast could not be attributed to an increase in LF during deeper sleep in the bCBR patients: only minimal LF changes were seen throughout sleep stages (Figure 2). Instead, the healthy controls appeared to have a marked decrease in LF activity during sleep stages 2 and 3 compared to sleep stage 1 and REM sleep. Accordingly, previous work indicated that sleep stages 2 and 3 are associated with the lowest values of sympathetic outflow in healthy controls.^1,5,6,25 The absence of significant differences in LF power between patients and controls during sleep stages 2 and 3 is thus explained by the transient suppression of baroreflex function during normal deep sleep. Baroreflex function is thus state-dependent, meaning that it is differently modulated by central influences in the different sleep phases and by wake adaptive behaviors.^1,3,26

In spite of the marked contrasts in baroreflex function during sleep stage 1 and REM sleep in bCBR patients, sleep-related HR changes were similar for bCBR patients and controls. Notably, the relative higher LF values of the controls seen during sleep stage 1 and REM compared to the bCBR patients did not translate to more marked HR contrasts in these sleep stages (Figure 2). Also, within the healthy controls we found that sleep stage 3 was associated with lowest LF values, while HR was similar between both sleep stages 2 and 3. This confirms previous work on sleep-related sympathetic outflow: sleep stage 3 was associated with a consistently lower value in muscle sympathetic nerve activity, whereas HR remained stable between both sleep stages 2 and 3.^1,5,6,25 Taken together, these findings suggest that the sleep-related HR changes primarily result from non-baroreflex mediated pathways.

Which alternative pathways should be considered? It could be argued that the HR decrease during sleep results from inactivity. However, the gradual decrease of HR seen in different NREM sleep stages and the contrasting effects in REM sleep favor central modulation. Accordingly, overnight infusion of vasopressive drugs (phenylephrine) in healthy subjects results in a sustained decrease in blood pressure the following morning, thus suggesting that overnight blood pressure increases are counteracted by central mechanisms.^27,28 Thus, while inactivity may, of course, in part contribute, this cannot explain the complex dynamics between sleep stages. Diurnal contrasts in autonomic control could also result from neuroendocrine changes, e.g., circadian rhythms in adrenocorticotropic activity and the renin-angiotensin-aldosterone system (RAAS). Clear circadian patterns have been identified for HR and its variability.²⁹ The overall effects of these circadian effects, however, appear to be modest and can only partly explain the sleep-related HR changes. Supporting this view, the blood pressure dipping pattern has shown to be primarily related to sleep-wake phases rather than endogenous circadian oscillators.²⁶ The close correlation between HR and sleep stages argues for direct effects of the sleep-wake cycle on the central autonomic network. Sympathetic outflow decreases and baroreflex sensitivity increases along with the depth of NREM sleep.^1,5,6,25 Sleep-related autonomic alterations may thus well fit in the general concept of sleep as a state of adaptive inactivity.³⁰ The central autonomic network during NREM sleep may involve the hypothalamic ventro-lateral preoptic area, central thermoregulatory and central baroreflex pathways, and command neurons in the pons and midbrain.^26,31 The intact sleep-related HR decrease in bCBR patients suggest that the peripheral baroreceptors play a minor role in the cardiovagal modulation during sleep. We speculate that the balance of neuronal autonomic control changes throughout the sleep-wake cycle. While awake, baroreceptors have an important role in buffering circulatory oscillations induced by activity and mental stress. During NREM sleep, these oscillations decrease. Consequently, the influence of the baroreceptors gradually declines and autonomic outflow is predominantly driven by the central autonomic network (Figure 3).^26,32 During REM sleep, both pathways are likely equally important: centrally induced transient augmentations of sympathetic outflow cause an increase in baro-receptor activity.^1,2,5,6

Figure 3.

Simplified schematic diagram illustrating the major changes in neuronal autonomic control throughout the sleep/wake cycle. While awake, cardiovagal outflow is predominantly controlled by peripheral baroreceptors, whereas during NREM sleep the central autonomic network becomes the driving force. During REM sleep, both pathways are likely equally important: central induced transient increases of sympathetic outflow cause an increase in baroreceptor activity.

Limitations

We did not quantify baroreflex function as we lacked continuous blood pressure measurements and did not perform daytime standardized baroreflex tests. The low LF values in our bCBR patients are however a clear indication of baroreflex dysfunction.^18,19 Accordingly, a previous small study in eight bCBR subjects confirmed that bCBR causes chronic impairment of baroreflex control of both heart rate and sympathetic nerve activity.¹¹ Also, ideally given the complex nature of the autonomic nervous system measurements should be multi-modal (e.g., including muscle sympathetic nervous activity, sympathetic skin response, pulse arterial tonometry [PAT]) to account for regional differences.

The small sample size of our study was inevitable in view of the extremely low prevalence of paragangliomas. The study did not have enough power to detect small differences. Therefore we were not willing to correct for multiple testing; this may have caused a type I error. However, we believe that our conclusions are valid, as the directions of the results were consistent and in line with previous daytime studies.^12–17 Again, because of the small number of subjects we included two patients who were using β-blockers. To overcome this limitation, we corrected for β-blocker use in our mixed effects model, but no significant effects were observed. Another limitation is the lack of measurement of leg movements. We are not aware of an increased prevalence of periodic leg movements in bCBR patients. Even if such a difference would have been the case, the effects on our outcome parameters would be likely minimal, as leg movements only have short-term effects: autonomic parameters did not differ between patients with periodic leg movements and controls if the periodic leg movements epochs were excluded.³³

CONCLUSION

In conclusion, the arterial baroreceptors are a critical relay in the autonomic network modulating both the peripheral and the central autonomic outflow. Our small study in patients with probable baroreflex failure, however, indicates that the sleep-related HR decrease predominantly results from non-baroreflex mediated, central mechanisms.

DISCLOSURE STATEMENT

This was not an industry supported study. Dr. Lammers has received an unrestricted research grant from UCB Pharma and is member of the Medical Advisory Board on Xyrem of UCB Pharma. Dr. Thijs has received research support from Dutch Epilepsy Foundation, Christelijke Vereniging voor de Verpleging van Lijders aan Epilepsie (Nederland), NUTS Ohra Fund, Medtronic and AC Thomson Foundation, and has received fees for lectures from Medtronic, UCB Pharma and GSK. The other authors have indicated no financial conflicts of interest.

ABBREVIATIONS

bCBR

bilateral carotid body tumor resection

BMI

body mass index

BP

blood pressure

BPM

beats per minute

EEG

electroencephalographic

HF

high frequency

HR

heart rate

HRV

heart rate variability

LF

low frequency

NREM

non-rapid eye movement

PLMS

periodic leg movements

REM

rapid eye movement

SDHD

succinate dehydrogenase subunit D