The Need for Specificity in Quantifying Neurocirculatory vs. Respiratory Effects of Eucapnic Hypoxia and Transient Hyperoxia

Abstract

Hypersensitivity of the carotid chemoreceptor leading to sympathetic nervous system activation and ventilatory instability has been implicated in the pathogenesis and consequences of several common clinical conditions. A variety of treatment approaches aimed at lessening chemoreceptor-driven sympathetic overactivity is now under investigation; thus, the ability to quantify this outcome variable with specificity and precision is crucial. Accordingly, we measured ventilatory and neurocirculatory responses to chemoreflex inhibition elicited by transient hyperoxia and chemoreflex excitation produced by exposure to graded, steady-state eucapnic hypoxia in middle-aged men and women (n=82) with continuous positive airway pressure-treated obstructive sleep apnea. Progressive, eucapnic hypoxia produced robust and highly variable increases in ventilation (+83±59%) and muscle sympathetic nerve activity (MSNA) burst frequency (+55±31%), whereas transient hyperoxia caused marked reductions in these variables (−35±14% and −42±16%, respectively). Coefficients of variation for ventilatory and MSNA burst frequency responses, indicating test-retest reproducibility, were 9% and 24% for hyperoxia and 35% and 28% for hypoxia. Based on statistical measures of rank correlation or even comparisons across quartiles of corresponding ventilatory and MSNA responses, we found that the magnitudes of ventilatory inhibition with hyperoxia or excitation with eucapnic hypoxia were not correlated with corresponding MSNA responses within individuals. We conclude that, in conscious, behaving humans, ventilatory sensitivities to progressive, steady-state, eucapnic hypoxia and transient hyperoxia do not predict MSNA responsiveness. Our findings also support the use of transient hyperoxia as a reliable, sensitive, measure of the carotid chemoreceptor contribution to tonic sympathetic nervous system activity and respiratory drive.

Introduction

The carotid body chemoreceptors respond to alterations in arterial blood O₂, CO₂, and pH by eliciting reflex changes in ventilation and sympathetic nervous system activity. Assessment of carotid chemoreflex sensitivity has received recent attention in clinical conditions such as hypertension (Tafil-Klawe et al., 1985), heart failure (Schultz et al., 2007), and obstructive sleep apnea (OSA) (Narkiewicz et al., 1999), and also in the context of acclimatization to high altitude hypoxia (Dempsey et al., 2014). A variety of transient and steady-state stimulatory and inhibitory interventions (Narkiewicz et al., 1999; Leuenberger et al., 2001; Steinback et al., 2009; Sinski et al., 2014; Morgan et al., 2018; Phillips et al., 2018; Keir et al., 2019) has been employed to provoke measurable increases and decreases in ventilation and sympathetic outflow; however, limited information is available on whether ventilatory and neurocirculatory responses to carotid chemoreceptor stimulation and inhibition are associated in human subjects.

The purpose of this study was to evaluate the specificity and reliability of tests of peripheral chemoreflex excitation and inhibition in humans. In a large cohort of middle-aged men and women with continuous positive airway pressure (CPAP)-treated OSA, we measured ventilatory and neurocirculatory sensitivities to carotid chemoreflex excitation produced by exposure to graded, steady-state eucapnic hypoxia. We also assessed the effects of carotid chemoreflex inhibition on tonic sympathetic and respiratory drives using exposure to transient hyperoxia. The rationale was as follows: the carotid chemoreceptor is known to be the major mediator of both ventilatory and sympathetic responses to hypoxia and a major contributor to tonic sympathetic and respiratory drives. It is often presumed, but rarely demonstrated, that both excitatory and inhibitory tests of chemoreflex function show congruence in the end-organ responses. A recent study has shown that ventilatory responsiveness to variations in O₂ and CO₂ combinations does not predict muscle sympathetic nerve activity (MSNA) responsiveness (Keir et al., 2019). Finally, the reproducibility of these tests, while seldom evaluated, is an important consideration, given that in awake, naive humans the tests likely give rise to random influences on ventilatory and sympathetic responsivity, e.g. cortical influences such as anxiety, that are separate from the chemoreflex per se.

Methods

Ethical Approval.

This study conformed to the standards set by the latest version of the Declaration of Helsinki. The protocol was approved by the University of Wisconsin Health Sciences Institutional Review Board (HS IRB 2012–0036). Written informed consent was obtained from all subjects prior to participation.

The present analysis is based on data collected during a previously reported randomized clinical trial (NCT01637623) of the effects of allopurinol, losartan, and placebo on chemoreflex sensitivity in hypertensive patients with OSA (Morgan et al., 2018). The majority of data reported herein was obtained at the initial (pre-randomization, prior to drug treatment) visit to the laboratory. A detailed description of parent study methods has been published (Morgan et al., 2018).

Subjects.

Eighty-two men and women, aged 21–65, with an apnea-hypopnea index ≥25 events/hr and either a clinical diagnosis of hypertension or two recorded clinic blood pressure readings >140/90 in the previous 12 months served as subjects. Treatment with continuous positive airway pressure (CPAP) was allowed, provided that it was used for a minimum of 3 months prior to study participation. Stable CPAP usage for one month prior to enrollment was documented using the machines’ electronic data cards. Medications taken prior to enrollment were not altered during study participation. The most common medication types and percentages of subjects treated were: antihypertensives (46%), inhaled β-adrenergic agonists (29%), nasal, inhaled, or systemic corticosteroids (34%), sedatives (23%), opioids (29%) and acid-suppressing medications (44%). Potential subjects were excluded if they received angiotensin converting enzyme inhibitors, alpha-adrenergic and angiotensin receptor antagonists, potassium-sparing diuretics without accompanying loop/thiazide diuretics, allopurinol, oxypurinol, febuxostat, amoxicillin, ampicillin, azathioprine, or mercaptopurine. Subject characteristics are shown in Table 1.

Table 1.

Subject characteristics. Data shown are mean±SD or n (%).

Age (yr)	48.4±9.6
Female Sex (n)	32 (39%)
Body mass index (kg/m2)	37.6±7.2
Antihypertensive medications (n)	40 (49%)
Apnea-hypopnea index (AHI) (events/hr)	40.9±26.1
Respiratory disturbance index (events/hr)	43.9±26.2
Time <88% SpO2 (% of total sleep time)	12.0±17.1
CPAP use (hr/night)	5.8±1.9
Residual AHI (events/hr)	2.5±2.7

Assessment of chemoreflex function.

Cardiorespiratory and MSNA data were acquired during wakefulness in the James B. Skatrud Pulmonary/Sleep Laboratory at the William S. Middleton Veterans Administration Hospital. Testing occurred with subjects in the supine position between 1100–1400 hours. Room temperature was maintained at 21–24° C. First, chemoreflex-mediated inhibition of MSNA was assessed during four 1-min exposures to hyperoxia (FIO₂, 1.0; modified Dejours test), which were separated by at least 3 min of room air breathing (see Figure 1). These hyperoxic exposures caused a rapid rise in end-tidal O₂ tension (P_ETO₂) to >250 mmHg, a level which silences nearly all carotid sinus nerve activity (Lahiri et al., 1987), within 2.3±0.7 breaths or 12.4±4.5 seconds. Beyond the 250 mmHg P_ETO₂ time point, 15-sec averages of heart rate, MSNA burst frequency, and MSNA burst incidence during each 1-min exposure were computed, and a nadir 15-sec value determined. For ventilatory measures, a single-breath nadir was noted during the one-minute period when PETO₂ was >250 mmHg. Nadir values for neurocirculatory and ventilatory variables in the 4 hyperoxia trials in each subject were then averaged. Second, chemoreflex-mediated activation of neurocirculatory and ventilatory variables was assessed during a progressive, steady-state eucapnic hypoxia exposure (see Figure 2). Baseline values for cardiorespiratory variables were obtained during 3 min of stable room air breathing measured after the approximately 45 min of supine rest required for instrumentation and completion of the hyperoxia trial. Then, data were collected during three continuous 3-min periods with arterial oxygen saturation (SpO₂) held constant at 90%, 85%, and 80%. Hypoxia was created by nitrogen supplementation of the inspired air. CO₂ was added, as needed, to maintain end-tidal CO₂ tension at the eupneic level. During normoxia and graded hypoxia, 3-min averages were computed for all respiratory and neurocirculatory variables. Chemoreflex responses to hypoxia were quantified by computing slopes of the linear relationships between ventilatory and neurocirculatory variables with SpO₂.

Figure 1.

Original record showing ventilatory and muscle sympathetic nerve activity (MSNA) responses to transient hyperoxia.

Figure 2.

Original record showing ventilatory and muscle sympathetic nerve activity (MSNA) responses to graded, eucapnic hypoxia.

Cardiorespiratory measurements.

During the chemoreflex assessments, we acquired heart rate (HR) from the electrocardiogram, SpO₂ by pulse oximetry (Biox 3740; Ohmeda, Madison, WI, USA), MSNA by direct intraneural recordings (see below). Blood pressure was measured by automated sphygmomanometry (Dinamap 1846SX/P; Critikon, Tampa, FL, USA) during the hypoxic exposure. Mean arterial pressure (MAP) was calculated as 1/3 pulse pressure + diastolic pressure.

Subjects breathed through a mouthpiece with the nose occluded. Airflow was measured with a heated pneumotachograph (#5719; Hans Rudolph, Kansas City, MO, USA), minute ventilation (V_E), tidal volume (VT) and breathing frequency (f_B) were calculated, and end-tidal PO₂ (P_ETO₂) and PCO₂ (P_ETCO₂) tensions were measured by expired air analysis (S-3A/I and CD-3A; Ametek, Pittsburgh, PA, USA). An increase of 3–7% in gas viscosity and pneumotachograph resistance with 40–70% O2:balance N₂ gas mixtures has been reported (Yeh et al., 1984). Given that these effects would be similar across subjects, we did not attempt to correct measurements of flow rate obtained during transitions from room air to transient hyperoxia. Central respiratory “drive” was estimated by calculating the tidal volume:inspiratory time ratio (VT:T_i), i.e. mean inspiratory flow rate.

Muscle sympathetic nerve activity.

Postganglionic MSNA was recorded from the fibular nerve using the technique of Vallbo et al. (Vallbo et al., 1979) as previously described (Khayat et al., 2004). Briefly, neural signals were passed to a differential preamplifier, an amplifier (total gain, 100,000), a band-pass filter (700–2,000 Hz), and an integrator (time constant, 100 msec). Placement of the recording electrode within a muscle fascicle was confirmed by: 1) presence of muscle twitches, but not paresthesias, in response to electrical stimulation; 2) pulse-synchronous nature of the nerve activity; 3) appearance of afferent activity in response to tapping or stretching of muscle, but not gentle stroking of skin, in the appropriate receptive fields; and 4) absence of neural activation in response to arousal stimuli. Once an acceptable nerve recording was obtained, the subject was instructed to maintain the leg in a relaxed position for the duration of the study. Acceptable neurograms were obtained in 79% of subjects. Sympathetic bursts were identified by computer-assisted inspection of the mean voltage neurogram. For purposes of quantification, MSNA was expressed as bursts per minute (burst frequency, BF) and bursts per 100 heart beats (burst incidence, BI). We employed only amplitude-free measures of MSNA for two reasons: 1) to guard against subtle, unrecognized shifts in electrode placement that could alter the neurogram’s baseline and cause spurious values for burst height and total activity, and 2) to assess the reproducibility of baseline MSNA by making measurements on 2 separate days with almost certain differences in electrode placement.

Reproducibility testing.

Day-to-day variability of ventilatory and neurocirculatory responses to hypoxia and hyperoxia was evaluated in 17 subjects who had been randomized to the placebo group of the parent study. These subjects underwent hyperoxia and hypoxia testing at the same time of day on two occasions separated by 6 weeks. Body weight, apnea-hypopnea index, and use of CPAP and medications was stable over this period.

Statistics.

Spearman rank correlation matrices were established to examine the relationships among ventilatory and neurocirculatory responses to hyperoxia and hypoxia. Strength of correlation based on r values was interpreted as follows: 0.00–0.25, little or no relationship; 0.25–0.50, fair relationship; 0.5–0.75, moderate to good relationship; >0.75, good to excellent relationship (Portney & Watkins, 2009).

To evaluate day-to-day reproducibility of baseline values and responses to hyperoxia and hypoxia, we calculated method error (ME) for each variable by dividing the standard deviation of day 1 to day 2 difference scores of all subjects by the square root of 2 (Portney & Watkins, 2009). Coefficients of variation of the method error (CV_ME) for each variable, our measures of random variability, were calculated as follows:

$$\mathrm{C}{\mathrm{V}}_{\mathrm{M}\mathrm{E}}=\frac{2\text{ME}}{\text{X}1+\text{X}2}\text{*}100$$

where X1 and X2 represent the group mean scores for day 1 and day 2, respectively.

We interpreted CV_ME ≤24% as good reproducibility, 25–39% as moderate reproducibility, and ≥40% as poor reproducibility. These tests of random variability were accompanied by paired t-tests to assess systematic bias. To assess the importance of repeated measurements, we computed CV_ME based on means of 2, 3, and all 4 of the hyperoxia trials.

Data shown are means ± SD. P values <0.05 were considered statistically significant.

Results

Characteristics of the Test Responses (Tables 2 and 3).

Table 2.

Ventilatory and neurocirculatory responses to transient hyperoxia. Nadir timing refers to the number of seconds after P_ETO₂ reached 250 mmHg. Values shown are means ±SD. Note: Hyperoxia trials were performed only in those subjects with acceptable sympathetic nerve recordings (n=61).

	Baseline	Hyperoxic Nadir/Peak	Hyperoxic Delta	Nadir (% baseline)	Nadir/Peak Timing (sec)
VE (liters/min)	9.6±2.1	6.2±2.0	−3.3±1.5	65±14	34.4±9.7
VT (liters)	0.85±0.30	0.51±0.22	−0.33±0.17	61±15	32.4±9.4
fB (breaths/min)	12.8±4.1	9.1±2.9	−3.7±2.0	73±10	30.4±7.8
VT:Ti (liters/sec)	0.34±0.08	0.23±0.06	−0.11±0.06	69±13	33.5±8.2
PETCO2 (mmHg)	42.0±3.2	42.8±3.5	0.8±1.2	102±3	43.2±10.3
HR (beats/min)	69.4±9.4	65.9±9.3	−3.4±1.9	95±3	61–75†
MSNA (bursts/min)	25.3±10.9	15.3±10.1	−9.9±3.8	58±16	31–45†
MSNA bursts/100HB)	36.6±15.5	22.6±14.6	−14.1±5.5	58±16	31–45†

^†Nadir occurred during this interval (data were reported as 15-sec averages).

V_E, minute ventilation; VT, tidal volume; f_B, breathing frequency; VT:T_i, tidal volume:inspiratory time ratio; P_ETCO₂, end-tidal CO₂ tension; HR, heart rate; MSNA, muscle sympathetic nerve activity

Table 3.

Ventilatory and neurocirculatory responses to graded, steady-state hypoxia. Absolute and relative values (% baseline; in parentheses) are shown as means±SD (n=82 for all variables except MSNA, where n=61).

	Normoxia	90% SpO2	85% SpO2	80% SpO2
SpO2 (actual)	95.5±1.0	89.4±0.6	84.5±0.9	79.5±1.0
PETCO2 (mmHg)	42.0±3.2	41.3±3.0	41.1±3.0	41.1±3.0
VE (liters/min)	9.2±1.9	12.0±3.3 (132±27%)	14.3±4.7 (157±44%)	16.6±5.9 (183±59%)
VT (liters)	0.65±0.22	0.84±0.29 (130±30%)	0.99±0.40 (152±46%)	1.13±0.51 (173±56%)
fB (breaths/min)	15.4±4.5	15.7±4.3 (103±14%)	16.1±4.9 (106±20%)	16.2±4.8 (107±24%)
VT:Ti (liters/sec)	0.37±0.08	0.47±0.13 (127±29%)	0.55±0.17 (150±43%)	0.64±0.21 (175±53%)
HR (beats/min)	70.7±10.2	77.4±10.6 (110±6%)	81.4±11.3 (115±8%)	84.6±11.2 (120±8%)
MAP (mmHg)	94.5±9.8	95.6±10.4 (101±4%)	96.9±11.4 (101±13%)	98.2±11.8 (104±7%)
MSNA (bursts/min)	23.7±11.4	27.2±12.8 (118±26%)	30.9±14.4 (133±26%)	35.5±15.0 (155±31%)
MSNA (bursts/100HB)	34.6±16.9	35.7±16.5 (107±27%)	38.4±16.8 (115±23%)	42.5±17.0 (128±26%)

SpO₂, arterial oxygen saturation; P_ETCO₂, end-tidal CO₂ tension; V_E, minute ventilation; VT, tidal volume; f_B, breathing frequency; VT:T_i, tidal volume:inspiratory time ratio; HR, heart rate; MAP, mean arterial pressure; MSNA, muscle sympathetic nerve activity

Exposure to transient hyperoxia (FIO₂, 1.0) produced decreases in V_E and its components, with single-breath nadirs reaching 30–40% below control values. These nadirs occurred 30–35 seconds (7.4±2.9 breaths) after P_ETO₂ reached 250 mmHg. Following the drop in ventilation, a peak in P_ETCO₂ of 0.8 mmHg occurred at 42±10 seconds (9.1±3.3 breaths) after attainment of the target P_ETO₂. MSNA decreased by ~40% with approximately the same time course as ventilation (MSNA nadir occurred between 31–45 sec; V_E nadir occurred at 34±9 seconds). HR decreased by 5%, with the nadir occurring later than either the ventilatory or MSNA response (5^th 15-sec interval after P_ETO₂ reached 250 mmHg). Individual subject and group mean V_E and BF responses to hyperoxia are shown in Figure 3.

Figure 3.

Individual subject ventilatory (upper panel) and muscle sympathetic nerve activity (MSNA; lower panel) responses to hyperoxia. These plots illustrate the magnitude and wide range of ventilatory and sympathetic responses that occurred as P_ETO₂ was raised from baseline to hyperoxic levels (106±7 to 543±70 mmHg). Group mean values are shown in the heavy dashed line.

Steady-state exposure to progressive eucapnic hypoxia caused peak increases in V_E, VT, and VT:T_i of +70–80% above baseline at 80±1% SpO2, with very little change in f_B. Peak increases in HR and MAP were +20±8% and +4±7% above baseline. MSNA burst frequency rose by +55±31% and burst incidence rose by +28±26%. Plots of individual subject and group mean V_E and BF responses to hypoxia are shown in Figure 4. Slopes of the relationships between SpO₂ and ventilatory variables, expressed as percentage change in ventilation per percentage change in SpO₂, were −5.2±3.6 (range, −20 to 0), −4.9±3.7 (−20 to 2), and −4.7±3.1 (−19 to 0) for V_E, VT, and VT:T_i, respectively. Slopes of the relationships between SpO₂ and MSNA, expressed as percentage change in MSNA per percentage change in SpO₂ were −3.4±1.8 (−10 to 1) and −1.7±1.6 (−6 to 2) for BF and BI, respectively.

Figure 4.

Individual subject responses ventilatory (upper panel) and muscle sympathetic nerve activity (MSNA; lower panel) to progressive, eucapnic hypoxia. This figure shows the magnitude and range of slopes in both variables. Group mean values are shown in the heavy dashed line.

Congruence of end-organ responses to hyperoxia (Table 4).

Table 4.

Correlation of ventilatory and neurocirculatory responses to hyperoxia. Values represent the differences between baselines and nadirs (means of the 4 hyperoxia trials). In this two-part table, Spearman correlation coefficients in the upper rows were based on absolute values and those in the lower rows were based on percent of baseline (n=61). No correlation was significantly significant.

	ΔVE (liters/min)	ΔVT (liters)	ΔfB (breaths/min)	ΔVT:Ti (liters/sec)
ΔHR (beats/min)	−0.04	0.05	−0.02	0.006
ΔMSNA (bursts/min)	0.03	0.007	0.05	−0.11
ΔMSNA (bursts/100 HB)	0.05	0.08	−0.02	−0.16
	ΔVE (%)	ΔVT (%)	ΔfB (%)	ΔVT:Ti (%)
HR (% baseline)	−0.07	−0.06	0.15	−0.01
MSNA BF (% baseline)	−0.13	−0.12	−0.04	−0.13
MSNA BI (% baseline)	−0.10	−0.11	−0.07	−0.11

There were no statistically significant relationships among ventilatory and neurocirculatory responses to hyperoxia. (Note that arterial pressure was not measured during hyperoxia trials.) This lack of congruence was also apparent when ventilatory and MSNA responses to hyperoxia were compared on a quartile-by-quartile basis, a less rigorous method for assessing congruence (see Figures 5 and 6). The magnitude of hyperoxic BF inhibition was relatively constant across the four quartiles of ventilatory responses, which varied by nearly three-fold (Figure 5, upper panel). Figure 6 (upper panel) shows an analogous lack of congruence when subjects were grouped according to their BF responses--V_E inhibition remained constant over the four quartiles of BF responses while MSNA decreased by three-fold.

Figure 5.

Individual and mean±SD values for ventilation (VE) and MSNA burst frequency (BF) in the 4 quartiles of ventilatory response to hyperoxia (A and B) and hypoxia (C and D). Hyperoxic responses are shown as percentage decreases from pre-exposure baselines. Hypoxic ventilatory responses are represented by slopes of the linear relationship between SpO₂ and V_E and BF. Note that the hypoxic slopes, which have negative values, have been converted to positive slopes for illustration purposes.

Figure 6.

Individual and mean±SD values for MSNA burst frequency (BF) and ventilation (VE) in the 4 quartiles of BF response to hyperoxia (A and B) and hypoxia (C and D). Hyperoxic responses are shown as percentage decreases from pre-exposure baselines. Hypoxic ventilatory responses are represented by slopes of the linear relationship between SpO₂ and BF and V_E. Note that the hypoxic slopes, which have negative values, have been converted to positive slopes for illustration purposes.

Congruence of end-organ responses to hypoxia (Table 5).

Table 5.

Correlation of ventilatory and neurocirculatory responses to hypoxia. In this two-part table, Spearman correlation coefficients in the upper rows were based on absolute values and those in the lower rows were based on percent of baseline (n=82 for MAP and HR slopes; n=61 for BF and BI slopes).

	VE slope	VT slope	fB slope	VTTi slope
MAP slope	0.50**	0.61**	−0.04	0.54**
HR slope	0.26*	0.40**	−0.07	0.25*
BF slope	0.08	0.16	−0.04	0.06
BI slope	−0.02	0.05	−0.05	−0.01
	VE slope (%)	VT slope (%)	fB slope (%)	VTTi slope (%)
MAP slope (%)	0.46**	0.55**	−0.06	0.49**
HR slope (%)	0.29**	0.35**	−0.05	0.27**
BF slope (%)	−0.10	−0.10	−0.02	−0.09
BI slope (%)	−0.13	−0.15	−0.00	−0.12

^*p<0.05

^**p<0.01

Statistically significant moderate-to-good correlations were noted between MAP slope and the V_E, VT, and VT:T_i responses. Fair relationships were observed between HR slope and the V_E, VT, and VT:T_i responses. R values for these relationships were comparable when the data were expressed relative to baseline rather than in absolute terms. There were no statistically significant relationships between ventilatory and MSNA responses to eucapnic hypoxia. When congruence of ventilatory and MSNA responses to hypoxia was examined on a quartile-by-quartile basis, no association was noted (Figure 5, lower panel). Figure 6 (lower panel) shows an analogous lack of congruence—increases in V_E remained relatively constant over the four quartiles of BF responses.

Day-to-day variation in responses to hyperoxia and hypoxia (Figure 7).

Figure 7.

Coefficients of variation of ` error (CV_ME) (n=17) showing day-to-day reproducibility of respiratory and neurocirculatory variables at baseline (upper panel) and responses (percent baseline) to hyperoxia (middle panel) and hypoxia (lower panel). Accompanying paired t-tests revealed no systematic differences between means of day 1 and day 2 values for eupneic PETCO₂ or any of the other neurocirculatory variables (p=0.17–0.91) with the exception of hyperoxia-induced decreases in BF and BI, which were smaller by 3.8 bursts/min and 4.2 bursts/100 heart beats on day 2 vs. day 1 (p=0.02 and 0.01, respectively). Note: blood pressure not measured during transient hyperoxia tests and PCO₂ during hypoxia was held constant by design.

VE, minute ventilation; VT, tidal volume; fB, breathing frequency; VTTi, tidal volume:inspiratory time ratio; HR, heart rate; MAP, mean arterial pressure; BF, MSNA burst frequency; BI, MSNA burst incidence; PCO₂, end-tidal CO₂ tension

Values for eupneic PETCO₂ prior to hyperoxia testing were nearly identical on day 1 vs. day 2 (42.5±2.9 vs. 42.6±3.6 mmHg). Likewise, values for eupneic PETCO₂ prior to hypoxia testing were similar on day 1 vs. day 2 (42.2±2.4 vs. 43.1±3.5 mmHg). The reproducibility of all baseline ventilatory and neurocirculatory measures obtained prior to chemoreflex testing on two occasions separated by 6 weeks is shown in the top panel of Figure 7. Ventilation and its components, HR, MAP, and P_ETCO₂ demonstrated good reproducibility (coefficients of variation of the method error (CV_ME), 4–22%). Baseline values for BF and BI were moderately reproducible (CV_ME, 32 and 39%).

During hyperoxia, decreases in ventilatory variables and HR and increases in P_ETCO₂ demonstrated excellent reproducibility (CV_ME, 2–10%; Figure 7, middle panel). Hyperoxia-induced decreases in BF and BI demonstrated good reproducibility (CV_ME, 24 and 22%). These computations of CVME were based on means of all 4 hyperoxia trials on the day 1 vs. day 2. Computations based on 2- and 3-trial means yielded poorer reproducibility (CV_ME, 2–5 fold higher).

The slopes of V_E, VT, and VT:T_i responses to eucapnic hypoxia were moderately reproducible (CV_ME, 34–35%; Figure 7, lower panel). Hypoxic f_B responses were not at all reproducible, probably because f_B changed very little (<2 breaths/min) during the hypoxic exposure. The HR response to hypoxia demonstrated good reproducibility (CV_ME, 24%), whereas MAP reproducibility was poor (CV_ME, 62%). Reproducibility of MSNA responses to hypoxia was moderate (CV_ME, 28 and 38% for BF and BI, respectively).

All estimates of day-to-day reproducibility shown above are indicative of random variability. No indications of systematic variability between day 1 and day 2 were noted (paired t-test p values=0.17–0.91) with the exception of hyperoxia-induced decreases in BF and BI, which were smaller by 3.8 bursts/min and 4.2 bursts/100 heart beats on day 2 vs. day 1 (p=0.02 and 0.01, respectively).

Discussion

We sought to determine the specificity of methods used to quantify peripheral chemoreflex control of ventilation and sympathetic nervous system activity in a large sample of human subjects with a wide range of sensitivities in both ventilatory and MSNA responses. Our major finding is that ventilatory responses to eucapnic, steady-state hypoxia and to transient hyperoxia are not congruent with MSNA responses to the same interventions. We conclude that, in conscious humans, sympathetic nervous system responses to eucapnic, steady-state hypoxia and transient hyperoxia cannot be predicted from ventilatory ones. Our finding of insignificant correspondences in carotid chemoreceptor-driven sympathetic and ventilatory responses is consistent with those of a previous study of chemoreceptor stimulation via O₂:CO₂ combinations (Keir et al., 2019). We also extend their findings by studying responses to carotid chemoreflex inhibition as well as excitation, by including an ample number of subjects for probing correlations, and by evaluating test-retest reproducibility.

Strength of ventilatory:sympathetic associations.

It is well established that cardiovascular and ventilatory responses to carotid chemoreceptor inhibition (and stimulation) share common sensory pathways (Guyenet, 2000). Consistent with this concept, our findings showed: a) that average magnitudes of inhibition of ventilatory and sympathetic responses to transient hyperoxia (−34% and −42% for V_E and BF, respectively) were robust and similar, with similar within-group variations (CV, 14% and 16%; see Figure 3); b) that the timing of nadirs of ventilatory and MSNA inhibition (30–40 seconds) with hyperoxia were very similar; and c) that average magnitudes of eucapnic hypoxic excitation of both ventilatory and MSNA measures (slopes of relationships with SpO₂ expressed as percent of normoxic control) were robust (−5.2±3.6% for V_E and −3.4±1.8% for BF) with substantial within-group variation (CV, 71% and 55%; see Figure 4). In addition, our relatively large sample was likely adequate to uncover meaningful associations. Accordingly, we expected that the responses would have yielded strong intraindividual correlations between tests; however, that did not occur as r² values were very low across the entire sample. Even when our subjects were divided into quartiles based on their ventilatory responses, a substantially less rigorous method for assessing association, MSNA and ventilatory responses were not associated.

Why were the sympathetic and ventilatory responses dissociated?

We have three potential explanations for the low correspondences. First among these are dissimilar influences on MSNA vs. ventilation of the secondary feedback mechanisms accompanying acute hypoxia and hyperoxia. Most prominent of these would be the hypoxia-induced increase in blood pressure (mean increase in MAP, +5±7 mmHg; range −8 to +23 mmHg) which would have greater inhibitory effects on MSNA vs. ventilation (Wallin et al., 1973; Saupe et al., 1995). Steady-state hypoxia causes variable increases in cerebral blood flow and therefore reductions in the arterial--to-brain PCO₂ difference which would likely influence the “eucapnic” hypoxic ventilatory response more than the MSNA response (Xie et al., 2006; Hoiland et al., 2016). In addition, lung stretch via increased tidal volume may inhibit within-breath MSNA more than respiratory drive (Seals et al., 1990; Iber et al., 1995); although in the steady-state, augmented tidal volume or changes in breathing pattern have shown no significant effect on MSNA burst frequency (Seals et al., 1990; Fatouleh & Macefield, 2011; Limberg et al., 2013). Finally, hypoxia-induced “mental stress” would be expected to have a systematic stimulatory effect on V_E but a variable effect on MSNA (Wallin et al., 1973; Carter & Ray, 2009).

Secondly, we found quite different magnitudes of within-subject reproducibility, i.e. random variability, between ventilatory and MSNA responses, which likely served to prevent stronger correlations between responses. As we determined in a subset of subjects retested after 6 weeks, our reproducibility coefficients of variation generally agreed with most, but not all, of the reports in the literaturè(Chua & Coats, 1995; Jensen et al., 2010), but varied between tests as follows: a) day-to-day reproducibility of MSNA and V_E responses to hypoxia (at unchanged levels of eucapnia on both days) were roughly comparable, but only moderate (CV_ME, >20%); and b) our transient hyperoxia responses, with each trial representing the mean of 4 repeated measurements, showed excellent day-to-day reproducibility for ventilatory responses and only moderate reproducibility for MSNA. Hyperoxic decreases in BF and BI had CV_ME >20%, which was more than twice as large as the CV_ME for ventilatory variables. Therefore, differences between MSNA and ventilatory responses in the amount of random variation with repeat testing within individuals likely contributed significantly to our difficulty in finding strong correlations between these response variables.

Thirdly, the integrative physiology literature contains multiple examples of chemoreceptor interventions in which ventilation and MSNA do not respond in lock-step fashion. Short-term exposure to asphyxia causes persistent increases in MSNA, but not ventilation, after return to room air breathing (Morgan et al., 1995). Central nervous system hypoxia, in the absence of carotid chemoreceptor input, causes simultaneous sympathetic activation and phrenic depression (Wasicko et al., 1990; Mitra et al., 1992). Experimental manipulation of central respiratory motor output in humans does not affect sympathetic outflow (St Croix et al., 1999). Stimulation of the carotid body with intravenous cyanide causes large, sustained increases in ventilation, but only very transient sympathetic activation, presumably because a brisk blood pressure rise causes feedback sympathoinhibition (Marcus et al., 2010). Following a hypocapnia-induced apnea, sympathetic reactivation precedes resumption of breathing (Trzebski & Kubin, 1981; St Croix et al., 1999).

The notion that carotid chemoreceptor hypersensitivity results in parallel augmentations of ventilatory and sympathetic responsiveness is supported by animal models of heart failure that have consistently demonstrated increased carotid sinus nerve activity, chemoreflex hypersensitivity manifest as periodic breathing during sleep and exaggerated ventilatory responses to exercise, increased tonic levels of renal sympathetic nerve activity, and enhanced sympathetic responses to acute hypoxia and to exercise (Stickland et al., 2007; Marcus et al., 2014). In contrast, studies of acute hypoxic responses in human heart failure have not revealed a coexistence between enhanced sympathoexcitation and enhanced ventilatory stimulation (Narkiewicz et al., 1999; Di Vanna et al., 2007), whereas both ventilatory and MSNA responses to acute hypoxia have been shown to be exaggerated in heart transplant recipients (Ciarka et al., 2006) and in patients with OSA and metabolic syndrome (Trombetta et al., 2013). Moreover, tests of chemoreceptor inhibition in several clinical populations, e.g. hypertension, chronic obstructive pulmonary disease, OSA, and heart transplantation, have revealed that sympathetic and ventilatory hypersensitivities to hyperoxia are sometimes in parallel (Phillips et al. 2018), but not always (Tafil-Klawe et al. 1985; Narkiewicz et al. 1998; Ciarka et al. 2006; Siński et al. 2012; Edgell et al. 2015). Thus, it is likely that a complexity of factors related to regulation of ventilation and sympathetic outflow, as well as secondary feedback mechanisms and methodological considerations, affect the congruence of MSNA and V_E during chemoreflex activation and deactivation in health and disease.

Associations between ventilatory and MAP and HR responses to hypoxia.

Even though ventilatory and MSNA responses to eucapnic hypoxia were not correlated, we observed moderate-to-good statistically significant correlations between MAP slope and the V_E, VT, and VT:T_i responses. This finding is somewhat surprising, given that hypoxia-induced increases in blood pressure are thought to be caused, at least in part, by increases in sympathetic vasoconstrictor outflow. Nevertheless, it must be borne in mind that systemic blood pressure is determined not only by peripheral vascular resistance, but also by cardiac output. In experimental animals, the asphyxia caused by imposed airway obstructions elicits blood pressure rises that are caused by increased cardiac output and HR (Schneider et al., 1997). Based on our HR data, we suspect that cardiac output rose during exposure to eucapnic hypoxia, and that this increase likely contributed to the MAP response. We observed fair relationships between V_E, VT, and VT:T_i responses and HR slope.

Limitations of the present study.

All subjects in this investigation had diagnoses of hypertension and OSA, and most were adequately treated via CPAP; however, that does not preclude possible residual effects of their sleep-disordered breathing on cardiorespiratory regulation that existed prior to treatment. In addition, our subjects received a variety of medications. Statin medications, which were taken by 12 of 82 of the current subjects, have recently been shown to reduce basal MSNA in humans with hyperlipidemia, hypertension, and heart failure (Deo et al., 2012; McGowan et al., 2013; Lewandowski et al., 2015) and to alter carotid chemoreceptor responses to hypoxia in a rat model of heart failure (Haack et al., 2014). Nevertheless, we could find no indication in the literature that statins would have differential effects on MSNA and ventilatory responses to hypoxia and hyperoxia, and therefore consider it unlikely that statins were responsible for the incongruent MSNA and V_E responses we observed. When we reconstructed the correlation matrices for hyperoxia and hypoxia without the subjects who received statins, there were minor variations in r values, but no change in the number of statistically significant correlations. It is unlikely that our reproducibility analysis was influenced by statins, because only one of the 17 subjects received at statin medication, and we documented stable use over the 6-week follow-up period.

We assume that the 17 subjects in our reproducibility analysis were representative of the entire sample. They did not differ in mean age, apnea-hypopnea index, or concomitant pharmacotherapy. However, they had somewhat higher blood pressures at baseline and during hypoxia and their ventilatory responses to hyperoxia were blunted relative to the group as a whole. In addition, there was a difference in gender distribution (28% female vs. 40% female in the entire sample). We did not have the scheduling flexibility to test female subjects in a standardized phase of the menstrual cycle. Ventilation (Macnutt et al., 2012), MSNA (Minson et al., 2000; Carter et al., 2013), and ventilatory and sympathetic chemoreflex responsiveness (Schoene et al., 1981; Macnutt et al., 2012) vary across the menstrual cycle; thus, hormone fluctuations may have decreased reproducibility of measurements in pre-menopausal subjects (20% of female participants). CPAP treatment was stable and effective in the 17 reproducibility subjects over the 6-week period, as judged by hours of usage (5–6 hours per night) and insignificant residual apnea-hypopnea index, and concomitant pharmacotherapy was also stable as judged by self-report. Nevertheless, both forms of treatment, i.e. CPAP plus drugs, render this subject group less than ideal for testing day-to-day reproducibility, although we have no evidence that this subgroup was not appropriate for comparison of reproducibility between the ventilatory and cardiovascular outcome variables.

In the present study, sympathetic outflow to vascular structures in the leg was recorded. This neural discharge is representative of sympathetic outflow to skeletal muscle vascular beds throughout the body (Rea & Wallin, 1989); however, our findings cannot be extrapolated to other organs or vascular beds. Finally, we did not perform familiarization trials prior to data collection, which would have reduced the novelty of sensations experienced during chemoreflex testing and perhaps lessened behavioral responses and improved reproducibility.

Quantifying the Tonic Carotid Chemoreceptor Drive.

Evidence accumulated over the past two decades supports a significant chemoreceptor-driven tonic drive to eupneic ventilation and to baseline sympathetic nervous system activity in health and especially in chronic diseases such as heart failure, hypertension, chronic obstructive pulmonary disease, and OSA. The following evidence supports a key role for a carotid chemoreceptor contribution to the enhanced drives. In humans and other mammals, measurements made before and in the first few days following carotid chemoreceptor denervation demonstrated substantial hypoventilation, i.e. an increase of 8–10 mmHg PaCO₂ or a decrease of 20–25% in alveolar ventilation (Bisgard et al., 1976; Forster et al., 2000; Rodman et al., 2001; Dahan et al., 2007). In the awake canine, maximum inhibition of the intact, extracorporeally perfused carotid chemoreceptor with hyperoxic-hypocapnic blood reduced V_E and VT:T_i (i.e. inspiratory drive) to nadir values 60% below eupnea that were sustained over time at 40% below control despite a 10 mmHg increase in PaCO₂ (Blain et al., 2009). This sustained inhibition in the face of systemic hypercapnia likely reflects the additional inhibitory effects of reduced carotid chemoreceptor input on central CO₂ sensitivity (Blain et al., 2010; Smith et al., 2015). In healthy humans, inhalation of 100% O₂ produces variable inhibitory effects. Previous investigators observed 10–15% decrements in 15-sec averages of V_E which occurred 30–60 secs after the onset of hyperoxia (DRIPPS & COMROE, 1947; DEJOURS, 1963; Tafil-Klawe M et al. 1985). In the present study, our large sample of CPAP-adherent OSA patients showed mean reductions of −35 and −31% in single-breath nadirs of V_E and VT:T_i and −40% reductions in 15 s average nadirs of MSNA, all reached within the first 30–40 s following the rise in P_ETO₂ >250 mmHg. Intra-individual reproducibility measures were excellent for hyperoxia-induced ventilatory changes (CV_ME, <10%) and moderate for MSNA (CV_ME, ~20%).

We recommend the use of transient nadir responses to hyperoxia as a means of quantifying prevailing tonic drives to breathing and to sympathetic vasoconstrictor outflow. This transient test might also be applied using assessment of vascular resistance as a surrogate for MSNA, e.g. see (Tafil-Klawe et al., 1985). We used the average of four repeat trials of transient hyperoxia per subject, which appeared to be sufficient to produce reproducible single-breath ventilatory nadirs. In contrast, the reproducibility of our MSNA measurements, which were made with traditional analysis techniques over a longer time period (15-sec averages of BF and BI), was only moderate and perhaps could have been improved with >4 repeat trials. For future investigations, we propose that MSNA quantification procedures with greater temporal sensitivity, e.g. wavelet analysis, may be better suited for capturing the hyperoxic nadir (Zhang et al. 2007; Salmanpour et al. 2010; Cairo et al. 2019). We do not recommend the common practice of utilizing hyperoxic exposures with P_ETO₂ >250mmHg for longer than one minute to quantify tonic chemoreceptor drives because prolonged hyperoxia has repeatedly been shown (in both intact humans and following carotid chemodenervation) to be stimulatory to both V_E and sympathetic outflow (Becker et al., 1996; Sinski et al., 2014). These substantial excitatory effects of sustained hyperoxia have been attributed to central chemoreceptor acidification secondary to reduced buffering of oxygenated hemoglobin (i.e. Haldane effect) and/or to “direct” effects of central nervous system hyperoxia (Dean et al., 2004; Gourine & Funk, 2017).

Summary and implications.

Sympathoexcitation and increased ventilatory drive and instability resulting from hypersensitivity and/or increased activity of the carotid chemoreceptor has been implicated in the pathogenesis and clinical consequences of conditions such as OSA, chronic pulmonary disease, hypertension, and heart failure (Hedner et al., 1988; Heindl et al., 2001; Parati & Esler, 2012; Niewinski & Ponikowski, 2017; Phillips et al., 2018). As a result, there is much current interest in quantifying the carotid chemoreceptor’s contribution to sympathetic nervous system activity and to ventilation. Based on the current data we conclude: a) that sympathetic nervous system responses to progressive, steady-state, eucapnic hypoxia and transient hyperoxia cannot be predicted from ventilatory responses, and vice versa. Thus, tests that examine the specific outcome variable of interest, i.e. MSNA or ventilation, are required; b) potential factors contributing to this lack of correlation include confounding influences of secondary feedback mechanisms (e.g. the inhibitory effect of hypoxia-induced blood pressure rise on MSNA) and also random differences between V_E and MSNA responses to hypoxia and hyperoxia in their degrees of day-to-day variability; and c) transient hyperoxia provides a reliable, sensitive means of quantifying tonic carotid chemoreflex-mediated drive to breathe and to MSNA.

Key Point Summary.

• The carotid chemoreceptor mediates the ventilatory and muscle sympathetic nerve activity (MSNA) responses to hypoxia and contributes to tonic sympathetic and respiratory drives. It is often presumed that both excitatory and inhibitory tests of chemoreflex function show congruence in the end-organ responses.

• We measured ventilatory and neurocirculatory (MSNA, blood pressure, and heart rate) responses to chemoreflex inhibition elicited by transient hyperoxia and to chemoreflex excitation produced by steady-state eucapnic hypoxia in a cohort of 82 middle-aged individuals,

• We found that ventilatory and MSNA responsiveness to hyperoxia and hypoxia were not significantly correlated within individuals.

• We conclude that ventilatory responses to hypoxia and hyperoxia do not predict MSNA responses and recommend that tests using the specific outcome of interest, i.e. MSNA or ventilation, are required.

• We recommend the use of transient hyperoxia as a sensitive and reliable means of quantifying tonic chemoreceptor-driven levels of sympathetic nervous system activity and respiratory drive.