Cerebral metabolic rate of oxygen in obstructive sleep apnea at rest and in response to breath-hold challenge

Abstract

Obstructive sleep apnea (OSA) is associated with extensive neurologic comorbidities. It is hypothesized that the repeated nocturnal apneas experienced in patients with OSA may inhibit the normal apneic response, resulting in hypoxic brain injury and subsequent neurologic dysfunction. In this study, we applied the recently developed OxFlow MRI method for rapid quantification of cerebral metabolic rate of oxygen (CMRO₂) during a volitional apnea paradigm. MRI data were analyzed in 11 OSA subjects and 10 controls (mean ± SD apnea-hypopnea index (AHI): 43.9 ± 18.1 vs. 2.9 ± 1.6 events/hour, P < 0.0001; age: 53.8 ± 8.2 vs. 45.3 ± 8.5 years, P = 0.027; BMI: 36.6 ± 4.4 vs. 31.9 ± 2.2 kg/m², P = 0.0064). Although total cerebral blood flow and arteriovenous oxygen difference were not significantly different between apneics and controls (P > 0.05), apneics displayed reduced baseline CMRO₂ (117.4 ± 37.5 vs. 151.6 ± 29.4 µmol/100 g/min, P = 0.013). In response to apnea, CMRO₂ decreased more in apneics than controls (−10.9 ± 8.8 % vs. −4.0 ± 6.7 %, P = 0.036). In contrast, group differences in flow-based cerebrovascular reactivity were not significant. Results should be interpreted with caution given the small sample size, and future studies with larger independent samples should examine the observed associations, including potential independent effects of age or BMI. Overall, these data suggest that dysregulation of the apneic response may be a mechanism for OSA-associated neuropathology.

Keywords

Magnetic resonance imaging susceptometry-based oximetry cerebral oxygen metabolism breath-hold obstructive sleep apnea

Introduction

Obstructive sleep apnea (OSA) is defined by structural and functional failure of the upper airway to maintain patency during sleep, resulting in periodic cessations or reductions in breathing and subsequent arterial desaturations. One in five adults in the Western world is believed to have at least mild OSA,¹ a figure that is rapidly increasing as obesity, the main risk factor for OSA, becomes more prevalent. In addition to the typical symptoms of daytime sleepiness, snoring, and disturbed sleep, OSA is also associated with significant systemic comorbidities, including hypertension, myocardial infarction and congestive heart failure, stroke, and type 2 diabetes.² Of particular relevance to this study, patients with OSA have a high prevalence of central nervous system dysfunction, including depression, dementia, and diminished cognitive performance.³ Magnetic resonance imaging (MRI) studies of OSA patients have detected neurologic lesions suggestive of hypoxic damage, including focal loss of grey matter⁴ and white matter.⁵

The etiology of OSA-associated neurologic comorbidities is not well understood. Brain tissue is particularly sensitive to hypoxic damage and rapid reperfusion,⁶ and brain regions known to be more acutely affected by hypoxia, such as the hippocampus, are among those identified as having grey matter loss in OSA.⁴ In normal physiology, apnea-induced hypercapnia and hypoxia cause chemoreceptor-mediated central vasodilation and concurrent peripheral vasoconstriction, preferentially conserving oxygen delivery to the brain to prevent hypoxic brain injury.⁷ In fact, recent work by our group has demonstrated that the cerebral metabolic rate of oxygen (CMRO₂) is not just maintained, but slightly increased in young healthy subjects in response to 30-s volitional apnea.⁸ This may represent a mechanism for increasing energy stores in anticipation of prolonged apnea, and is consistent with gas-mixture breathing studies demonstrating increased CMRO₂ in response to steady-state hypoxia.⁹ However, it is possible that patients with OSA do not possess a normal apneic response, allowing hypoxic damage to occur during OSA-associated nocturnal apneas.

In support of this hypothesis are studies associating OSA with blunted cerebrovascular reactivity (CVR),^10–12 typically defined as the cerebral blood flow (CBF) change in response to a vasoactive stimulus, such as hypercapnia¹¹ or apnea,^10,12 the latter of which is particularly pertinent to OSA pathophysiology. In one study, CVR assessed in response to breath-hold by Doppler ultrasound was found to be significantly lower in OSA subjects, and more so in the morning, indicating that their diminished vasodilatory response is worsened by more recent exposure to nocturnal apneas.¹⁰ Recently, blood-oxygen-level dependent (BOLD) functional MRI (fMRI) detected reduced CVR in select brain regions of apneics, including the hippocampus.¹² Blunted cerebrovascular responses to autonomic challenges^13,14 (e.g., orthostatic hypotension, cold pressor challenge, etc.) suggest a mechanism of brain injury even during wakefulness, whereby day-to-day activities (e.g., standing) may not be met with an adequate cerebrovascular response to preserve central oxygen delivery. Finally, studies in both animals¹⁵ and healthy humans¹⁶ exposed to chronic intermittent hypoxia (CIH) paradigms indicate a causal link between exposure to cyclic hypoxia and impaired vascular reactivity. If initial hypoxic injury itself leads to further blunting of the apneic response, a negative feedback cycle of worsening hypoxic damage could ensue.

Though supporting a mechanism for hypoxic brain injury in OSA, these previous studies all measure surrogate markers of brain oxygen metabolism (i.e., blood flow, perfusion, or BOLD fMRI signal), reductions of which do not necessarily correlate with decreased oxygen delivery and consumption. Of more central interest is whether oxygen consumption itself is maintained. Direct quantification of CMRO₂ requires quantification of both cerebral blood flow and oxygen extraction, the latter posing the greater technical challenge. CMRO₂ is less variable than blood flow or oxygen extraction in healthy subjects at baseline,^17,18 and relatively conserved in response to physiologic challenges such as hypercapnia^19,20 and hypoxia,⁹ suggesting that CMRO₂ may be a more significant index for assessing neurovascular dysfunction than either blood flow or oxygenation alone. CMRO₂ reduction has been associated with many of the most common neurologic disorders, including Alzheimer’s disease,²¹ Parkinson’s disease,²² and multiple sclerosis.²³

Quantifying CMRO₂ in response to apnea requires temporal resolution on the order of several seconds. Although this temporal resolution can be achieved with BOLD fMRI, attempts to ‘calibrate’ the BOLD signal (i.e., convert fractional BOLD signal change to fractional change in CMRO₂) rely on models with many physiologic assumptions and complex experimental setups involving gas-mixture breathing.²⁴ Moreover, such calibration techniques still provide only relative changes in CMRO₂, with additional calibration needed to quantify baseline values in absolute physiologic units.

Recently, we have introduced a MRI technique for rapid whole brain CMRO₂ quantification based on simultaneous susceptometry-based oximetry (SBO) and phase-contrast (PC)-MRI blood flow quantification – termed OxFlow.¹⁸ Subsequent iterations of the technique have employed view-sharing to achieve temporal resolution as low as 3 s for whole brain CMRO₂ quantification.^8,25 Unlike all previous CMRO₂ measurement techniques, this method has sufficient temporal resolution to detect CMRO₂ changes in response to apnea, allowing direct evaluation of the relationship between apnea, cerebrovascular reactivity, and brain oxygen delivery and consumption.

In this study, OxFlow was applied to compare the CMRO₂ response to apnea in OSA subjects and healthy controls. We hypothesized that OSA would be associated with reduced baseline CMRO₂, as well as a blunted CMRO₂ response to volitional apnea, and that this blunting would correlate with disease severity as measured by the apnea hypopnea index (AHI).

Materials and methods

Susceptometry-based quantification of Y_v

Susceptometry-based oximetry (SBO) exploits the relative paramagnetism of hemoglobin in the deoxygenated state, which results in a linear relationship between venous blood oxygen saturation (Y_v) and venous blood magnetic susceptibility relative to surrounding tissue (Δχ):

Δ χ = Hct \cdot (Δ χ_{do} (1 - Y_{v}) + Δ χ_{oxy})

(1)

where Hct is the hematocrit and Δχ_do and Δχ_oxy are the volume susceptibility differences between fully oxygenated and deoxygenated packed red blood cells and between fully oxygenated packed red blood cells and water, respectively. Values of 4π × 0.273 and 4π × 0.008 p.p.m. (SI units) are used for Δχ_do and Δχ_oxy, based on ex vivo calibration experiments.²⁶ The susceptibility offset (Δχ) induces a field shift (ΔB), which causes an increase in MR signal phase (Δφ) between blood and surrounding ‘reference’ tissue as a function of echo spacing (ΔTE) in a multi-echo gradient echo imaging sequence

Δ B = \frac{Δ φ}{γ Δ TE}

(2)

where γ is the proton gyromagnetic ratio.

Solving for Y_v thus hinges on determining Δχ from the measured ΔB, an inversion problem that is mathematically ill-posed in the general case. However, by modeling the blood vessel of interest as a long straight cylinder²⁷ with defined tilt angle (θ) relative to the main magnetic field (B₀), an expression relating the susceptibility and field offsets can be derived analytically

Δ B = \frac{1}{6} Δ χ B_{0} (3 \cos 2 θ - 1)

(3)

To quantify global Y_v, this long paramagnetic cylinder model is applied to the superior sagittal sinus (SSS), the largest cerebral vein, which drains about 50% of total cerebral outflow. Y_v in the SSS is nearly identical to global cerebral venous oxygenation measured in the internal jugular veins as shown by T₂-based oximetry methods.¹⁷ However, while trachea-induced susceptibility artifacts complicate SBO in the jugular veins, the field adjacent to the SSS is relatively homogeneous, making it the ideal candidate for global Y_v quantification via SBO.

Combined SBO and PC-MRI for CMRO₂ quantification (OxFlow)

SBO and PC-MRI can be readily combined as a single gradient echo sequence by applying flow encoding before a multi-echo gradient-recalled echo readout, achieving simultaneous quantification of blood oxygenation and flow (Figure 1(a) and (c)).^8,18 In this study, OxFlow was implemented with BRISK Cartesian view-sharing,²⁸ with one-quarter k-space acquired at each time point and a resulting temporal resolution of 2 s for each simultaneously acquired pair of field and velocity maps. BRISK is more robust against subject motion compared to previous implementations of OxFlow using Keyhole view-sharing.^29,30 Other OxFlow pulse sequence parameters used in this study were: reconstructed matrix = 208 × 208, resolution = 0.85 × 0.85 × 5.00 mm, TR/TE₁/ΔTE = 19.23/5.73/7.04 ms, bandwidth = 321 Hz/pixel, and VENC = 50 cm/s.

Figure 1.

MRI pulse sequences for high temporal resolution quantification of global CMRO₂: (a) Single-slice OxFlow pulse sequence with BRISK k-space sampling produces a velocity and field map at 2-s temporal resolution to quantify SSSBF and Y_v, respectively. (b) Dual-slice PC-MRI pulse sequence run prior to the OxFlow experiment allows quantification of the tCBF_cal/SSSBF_cal calibration factor used to upscale OxFlow-derived SSSBF to tCBF in equation (4). Arrows indicate the echoes used to generate each image. Circles indicate the vessels of interest (superior sagittal sinus in the head, internal carotid arteries (ICAs) and vertebral arteries (VAs) in the neck). (c) Sagittal maximum intensity projection showing the head (blue) and neck (red) slice locations and corresponding axial magnitude images. Boxes in the axial images correspond to the parametric map ROIs in (a) and (b). All images are from a representative OSA subject (male, 63 years old).

SSS blood flow (SSSBF) can be converted to total CBF (tCBF) via multiplication by a calibration factor determined through a separate two-slice interleaved PC-MRI acquisition toggled between the internal carotid and vertebral arteries in the neck (which comprise tCBF) and the SSS in the head (Figure 1(b) and (c)).⁸ This calibration step can be run before subsequent OxFlow experiments, allowing tCBF to be quantified as

tCBF = (\frac{{tCBF}_{cal}}{{SSSBF}_{cal}}) \cdot SSSBF

(4)

CMRO₂ can then be quantified via the Fick principle

{CMRO}_{2} = C_{a} \cdot tCBF \cdot (Y_{a} - Y_{v})

(5)

where C_a is the hematocrit-dependent arterial oxygen content of blood in µmol O₂/100 mL and Y_a is the arterial oxygen saturation in percent hemoglobin oxygen saturation (%HbO₂), which can be measured by digital pulse oximetry. The tCBF:SSSBF calibration pulse sequence parameters used in this study were: reconstructed matrix = 208 × 208, resolution = 0.85 × 0.85 × 5.00 mm, TR/TE = 12.02/5.73 ms, bandwidth = 321 Hz/pixel, VENC = 50 cm/s (head slice)/80 cm/s (neck slice), temporal resolution = 10 s, and averages = 4.

Subjects

Subjects were recruited based on results of a clinically indicated sleep study (in-lab attended polysomnography) performed at the University of Pennsylvania Sleep Center. The AHI was calculated as the mean number of apnea and hypopnea events per hour of sleep. Obstructive apneas were defined as at least a 90% drop in the thermal sensor excursion of baseline lasting at least 10 s; hypopneas were defined as a 50% reduction in airflow for greater than 10 s and associated with greater than 3% decrement in oxyhemoglobin saturation and/or an arousal. Nasal pressure monitors were used in all subjects to measure airflow.

Thirteen newly diagnosed apneics (AHI > 15 events/h) and 10 non-apneic controls (AHI < 10 events/hour) were selected after screening for standard MRI exclusion criteria (claustrophobia, metal implants, pregnancy, etc.) and excluding diseases expected to affect cerebral metabolism and/or cerebrovascular reactivity, including congestive heart failure, chronic obstructive pulmonary disease, stroke, head trauma, and other significant neurological diseases. Cigarette smokers or users of other nicotine products were excluded as smoking can affect vasodilation. Subjects had no prior history of OSA diagnosis or continuous positive airway pressure (CPAP) use. Subject demographics are displayed in Table 1.

Table 1.

Group demographics and polysomnography data.

	OSA (N = 11)	Control (N = 10)	P-value
Age (years)	53.8 (8.2)	45.3 (8.5)	0.027*
Gender (male/female)	7/4	4/6	0.39
Race (CA/AA)	2/9	6/4	0.081
BMI (kg/m²)	36.6 (4.4)	31.9 (2.2)	0.0064*
AHI (events/h)	43.9 (18.1)	2.9 (1.6)	<0.0001*
Y_a Nadir (%HbO₂)	77.5 (8.5)	89.0 (3.7)	0.0001*

Note: Parameter values are reported as mean (SD) across subjects. P-values are based on Wilcoxon two-sample exact tests (age, BMI, AHI, Y_a Nadir) or Fisher’s exact tests (Gender, Race). CA, Caucasian; AA, African American; BMI, body mass index; AHI, apnea hypopnea index, Y_a, arterial oxygen saturation; %HbO₂, percent hemoglobin oxygen saturation.

P < 0.05.

Experimental procedures

All imaging protocols were approved by the Institutional Review Board of the University of Pennsylvania according to the ethical standards of the Belmont Report, and subjects provided written informed consent. Prior to scanning, a capillary blood sample was obtained and analyzed using an Hb 201⁺ (HemoCue, Brea, CA, USA) portable hemoglobin measurement device for determination of Hct in equation (1) and C_a in equation (5).

Volitional apnea paradigm

The apnea paradigm consisted of 30-s breath-holds at end-expiratory volume to mimic nocturnal apneas experienced in OSA. Coaching was used to maximize intra- and inter-subject repeatability and consistency of the apneas. Prior to scanning, breathing at normal end-expiratory volume was explained and demonstrated. During all breath holds, subjects were verbally instructed to “breathe in”, “breathe out”, and “stop breathing” at 6, 3, and 0 s, respectively, before the designated start of each apnea period, and instructed to “breathe normally” at the end of the apnea period. Each subject performed two practice apneas in the MRI scanner prior to OxFlow scanning and three during OxFlow scanning. Verbal instructions were given via MRI-compatible headphones. Breath-hold compliance was monitored by respiratory bellows.

MR imaging protocol

To minimize biological confounds and normal variations that might occur during the diurnal cycle, all subjects were scanned in the afternoon and instructed to abstain from caffeine (which promotes vasoconstriction) on the day of the study. All MR-imaging studies were performed on a 1.5 T wide-bore (70 cm) Siemens Espree system (Siemens Medical Solutions, Erlangen, Germany) using vendor-supplied 12-channel head and 2-channel neck receive coils. Subjects were fitted with pulse oximetry (Expression, Invivo Research Inc., Orlando, FL, USA) and respiratory bellows before performing the first practice breath-hold. A vendor-provided axial localizer scan was run for subsequent slice planning, followed by a second practice breath-hold. To allow tCBF normalization to brain mass, a 1-mm isotropic, 3D T₁-weighted MPRAGE data set was acquired. Next, the tCBF:SSSBF calibration scan was run, followed by second-order shimming over the brain volume. Finally, the OxFlow sequence was run continuously for 9 min, during which the subjects completed three coached 30-s apneas, each followed by 2 min of normal breathing recovery. The entire MRI protocol lasted approximately 20 min (Figure 2).

Figure 2.

MRI protocol for quantifying CMRO₂ at rest and in response to apnea: Red boxes indicate 30-s coached volitional apneas. Two practice apneas are performed during protocol setup. During continuous CMRO₂ quantification with OxFlow, three apneas are performed, each followed by 2 min of normal breathing recovery.

All subjects were able to successfully complete each breath-hold. However, two subjects (both apneics) failed to remain awake and experienced obstructive apneas during the recovery portions of the OxFlow acquisition, resulting in periodic desaturations throughout the paradigm. Their data were excluded from further analysis.

Data processing

All image reconstruction was performed with in-house-written MATLAB (Mathworks, Natick, MA) scripts. BRISK-sampled raw OxFlow data were first reordered to create full k-space data sets corresponding to each echo at 2-s temporal resolution. Velocity maps were obtained from the phase difference between flow-encoded and flow-compensated images reconstructed from data acquired at TE₁. Field maps were generated from the phase difference between flow-compensated images reconstructed from data acquired at TE₁ and TE₂. Magnitude images at each time point were used to motion-correct the time series velocity and field maps using the StackReg plugin for ImageJ.³¹

OxFlow-derived SSSBF was determined by integration of the velocity map over an ROI fully containing the SSS. Data from the two-slice interleaved calibration sequence was processed analogously – with tCBF_cal quantified by integration over the internal carotid and vertebral arteries – to calculate the tCBF_cal/SSSBF_cal calibration factor to upscale OxFlow-derived SSSBF and determine tCBF in equation (4). Total brain volume was determined from the T₁-MPRAGE data using the BET tool in FSL,³² and converted to mass based on an average brain density of 1.05 g/mL.³³

For Y_v quantification, bulk susceptibility effects were removed from the field maps via second-order polynomial fitting of the field in brain tissue surrounding the SSS. Average phase was measured in two ROIs, one entirely within the SSS and another in a small reference region of brain tissue immediately surrounding the SSS approximately one vessel radius in width and located one vessel radius anterior to the SSS border. The difference in phase between these regions provides Δφ in equation (2).

Y_a values obtained via pulse oximetry were recorded at 2-s intervals matching each OxFlow time point. To correct for the temporal delay between central and peripheral blood arrival from the lungs, the pulse oximetry data was time-shifted for each subject such that the initial resaturation following apnea occurred 7 s after apnea cessation. This timing corresponds to the known circulatory transport delay between the lungs and brain³⁴ to within the temporal resolution of the pulse sequence. The arteriovenous oxygen difference (AVO₂D) was quantified as Y_a − Y_v. Combination of equations (1) to (5) was used to determine temporally resolved CMRO₂.

Statistical analysis

For each subject, time-course data were averaged over the three repeated blocks of the paradigm to improve signal-to-noise (SNR) and remove physiologic variation unrelated to apnea. For all parameters, average baseline values were quantified over the 24 s (12 data points) immediately preceding the “breathe in” command. For parameters that change monotonically in response to apnea, maximum (tCBF, Y_v) or minimum (Y_a, AVO₂D) percent changes relative to the average baseline values were quantified. To characterize the CMRO₂ apneic response, data were averaged over the second half (final 14 s, seven data points) of the apnea period to generate average end-apnea parameter values. The second half of the apnea period was used to eliminate residual breathing effects and because physiologic changes from apnea are not expected to occur instantaneously. The CMRO₂ apneic response was quantified as the percent change from the average baseline to the average end-apnea period.

Continuous outcomes were summarized using means and standard deviations (SDs) and categorical outcomes using frequencies and percentages. Given the relatively small number of apneics (N = 11) and controls (N = 10) in this study, summary measures were compared between groups using Wilcoxon two-sample exact tests (for continuous variables) and Fisher’s exact tests (for categorical variables). Baseline CMRO₂ and the CMRO₂ apneic response values were correlated with AHI using Spearman’s rank correlations. Statistical significance was defined as P < 0.05. Throughout the manuscript, parameter values are reported as mean [SD] and all P-values are two sided.

Given the limited overlap in age and BMI between OSA subjects and controls, it was not possible to model the independent effect of these variables on CMRO₂ or the CMRO₂ apneic response within the entire sample. Instead, the potential effect of between-group differences in these variables was assessed in a secondary analysis within a small sample of apneics and controls (N = four pairs), matched for age (within 2.5 years) and BMI (within 2.5 kg/m²). Differences in traits of interest were calculated within each pair, as the value in the apneic subject minus that in the control. Observed differences were tested for significance using an exact P-value from the non-parametric signed rank test on the difference. The calculated differences in the matched sample were compared to that in the overall population. If a similar magnitude was observed, it was concluded that results in the overall population were unlikely to be primarily driven by differences in age and BMI. As a further step toward understanding potential confounding effects, age and BMI were associated with CMRO₂ and the CMRO₂ apneic response using Spearman’s rank correlations. These tests were performed across all subjects (N = 21), apneics only (N = 11), and controls only (N = 10).

Results

Subject demographic and polysomnography group characteristics

AHI was significantly higher (43.9 [18.1] vs. 2.9 [1.6] events/hour, P < 0.0001) and Y_a nadir lower (77.5 [8.5] vs. 89.0 [3.7] %HbO₂, P = 0.0001) in apneics relative to controls. Subjects in the OSA group (AHI > 15 events/hour) were of greater age (53.8 [8.2] vs. 45.3 [8.5] years, P = 0.027) and BMI (36.6 [4.4] vs. 31.9 [2.2] kg/m², P = 0.0064), and had slightly though non-significantly larger brain mass (1437 [208] vs. 1376 [177] g, P = 0.39).

Baseline differences between OSA subjects and controls

Time-course plots of the MR- and pulse oximetry-measured parameters in a single OSA subject (male, 63 years old) (Figure 3a) demonstrate the expected increase in Y_v and tCBF and decrease in Y_a in response to apnea (red shading), as previously observed in young healthy subjects.³⁰ The resulting CMRO₂ time-course in the same OSA subject (Figure 3b) shows a CMRO₂ reduction from baseline (denoted ‘Base’) to end-apnea (denoted ‘EA’) of 12.1%.

Figure 3.

Apnea paradigm data in a representative OSA subject (male, 63 years old): (a) Single subject time-course plot of measured parameters (MRI-derived tCBF and Y_v and pulse-oximetry-derived Y_a) and (b) quantified CMRO₂ in absolute physiologic units. In all plots, data have been averaged over the three repeated blocks of the apnea paradigm and the gray shading indicates the apnea period. Black symbols correspond to the maximum (tCBF, Y_v) or minimum (Y_a) parameter values used to quantify the peak apneic response. ‘Base’ indicates the data averaged to quantify the baseline parameter values. ‘EA’ indicates the CMRO₂ data averaged to quantify the end-apnea CMRO₂ for determination of the CMRO₂ apneic response.

Group time-course plots of measured parameters (Figure 4a) demonstrate a lower tCBF and Y_a and higher Y_v in apneics versus controls, as well as a considerably lower CMRO₂ in apneics (Figure 4b) throughout the paradigm. Although tCBF, Y_a, and Y_v baseline-averaged parameters values were not statistically different between groups (Table 2), they synergistically resulted in a significantly lower CMRO₂ in apneics versus controls (117.4 [37.5] vs. 151.6 [29.4] µmol/100 g/min, P = 0.013). The initial rise and fall in tCBF and Y_v observed at the beginning of the apneic period is attributable to breath-hold-induced intrathoracic pressure changes causing modulations in cerebral venous return, as previously observed.⁸

Figure 4.

Group-averaged apnea paradigm data in OSA subjects and controls: (a) Time-course plots of OSA subject (solid lines) and control subject (dotted lines) measured parameters (MRI-derived tCBF and Y_v and pulse-oximetry-derived Y_a) and (b) quantified CMRO₂ in absolute physiologic units. (c) Baseline normalized tCBF, AVO₂D and (d) CMRO₂ illustrate the change in parameter values in response to apnea in each group. Error bars represent standard errors within each group and correspond to individually sampled time points. In all plots, data have been averaged over the three repeated blocks of the apnea paradigm and the gray shading indicates the apnea period. ‘Base’ indicates the data averaged to quantify the baseline parameter values. ‘EA’ indicates the data averaged to quantify the end-apnea CMRO₂ for determination of the CMRO₂ apneic response.

Table 2.

Summary of baseline and apneic response parameters in OSA subjects and controls.

		OSA (N = 11)	Control (N = 10)	P-value
tCBF	Ave. Baseline (mL/100 g/min)	42.4 (11.3)	45.4 (7.1)	0.20
	Peak apneic response (%)	35.6 (17.1)	35.7 (8.8)	0.56
Y_a	Ave. baseline (%HbO₂)	95.6 (1.8)	96.9 (1.9)	0.13
	Peak apneic response (%)	−6.8 (0.5)	−6.1 (1.4)	0.036*
Y_v	Ave. baseline (%HbO₂)	62.2 (6.2)	58.7 (7.2)	0.25
	Peak apneic response (%)	15.5 (10.6)	13.4 (4.7)	0.86
AVO₂D	Ave. baseline (%HbO₂)	33.4 (6.5)	38.2 (7.5)	0.17
	Peak apneic response (%)	−40.9 (14.0)	−32.0 (6.4)	0.099
CMRO₂	Ave. baseline (µmol/100 g/min)	117.4 (37.5)	151.6 (29.4)	0.013*
	Ave. baseline to end-apnea change (%)	−10.9 (8.8), P = 0.0049*	−4.0 (6.7), P = 0.13	0.036*

Note: Parameter values are reported as mean (SD) across subjects. P-values are based on Wilcoxon two-sample exact tests. tCBF, total cerebral blood flow; Y_a, arterial oxygen saturation; Y_v, venous oxygen saturation; AVO₂D, arteriovenous oxygen saturation difference; CMRO₂, cerebral metabolic rate of oxygen.

P < 0.05.

Apneic response in OSA subjects and controls

To better illustrate the apneic response, measured parameters (Figure 4c) and CMRO₂ (Figure 4d) are displayed in terms of percent changes relative to average baseline values. CVR – the change in tCBF in response to apnea – was not different between groups. However, there was a trend toward a larger decrease in oxygen extraction (AVO₂D) in apneics (−40.9 [14.0] vs. 32.0 [6.4]%, P = 0.099). CMRO₂ decreased significantly in apneics (−10.9 [8.8]%, P = 0.0049) but not in controls (−4.0 [6.7] %, P = 0.13), with a significant group difference (P = 0.036). Y_a reduction in response to apnea was greater in apneics (−6.8 [0.5] vs. −6.1 [1.4] %, P = 0.036), although it should be noted that the magnitude of this difference was quite small, and thus had little impact on the observed differences in the CMRO₂ apneic response.

Relationship between CMRO₂ and AHI

To examine the sensitivity of CMRO₂ to OSA disease severity, AHI was correlated with both baseline CMRO₂ (Figure 5a) and the CMRO₂ apneic response (Figure 5b). When including all subjects (N = 21), AHI correlated significantly with baseline CMRO₂ (Spearman’s ρ = −0.65, P = 0.0014) and the CMRO₂ apneic response (Spearman’s ρ = −0.53, P = 0.013). When restricted to apneics only, AHI correlation with baseline CMRO₂ was only marginally significant (Spearman’s ρ = −0.61, P = 0.047), and AHI correlation with the CMRO₂ apneic response (Pearson’s ρ = −0.47, P = 0.14) was only a trend.

Figure 5.

Relationship between CMRO₂ and AHI: (a) Correlation plots of baseline CMRO₂ vs. AHI and (b) CMRO₂ apneic response vs. AHI. OSA subjects (solid diamonds) and controls (empty triangles) are clearly separated by AHI. Least squares regression lines are plotted for OSA-subjects only (solid lines) and for all subjects (dotted lines) with corresponding Spearman’s rank correlations and P-values indicated.

Age and BMI effects analysis

Examination of a sub-sample of age- and BMI-matched subjects provided insight into possible confounding effects of the slight mismatch in these group characteristics. As expected given matching, pairs were similar with respect to age (mean [SD] difference: −0.25 [1.26] years, P > 0.99) and BMI (1.2 [1.5] kg/m², P = 0.375). Within the matched sample, apneics had a baseline CMRO₂ 45.9 µmol/100 g/min lower on average compared to matched controls (P = 0.125) and a CMRO₂ apneic response 10.4% lower (P = 0.375). These differences are greater than those observed between apneics and controls in the overall sample, supporting the effect size seen in that population. This suggests that although statistical significance was not achieved in this small matched sample, associations in the overall sample were not completely driven by imbalances in age and BMI.

Though not reaching significance, the data were suggestive of a negative correlation between CMRO₂ and BMI (ρ = −0.40, P = 0.070) across the entire sample. In contrast, correlation between CMRO₂ and BMI in apneics only (ρ = −0.38, P = 0.25) or controls only (ρ = 0.25, P = 0.49) was not significant, and correlations between BMI and the CMRO₂ apneic response were not significant in any group. Across all subjects, correlations between age and both baseline CMRO₂ and CMRO₂ apneic response were small and non-significant. Age correlated significantly only with the CMRO₂ apneic response in control subjects (ρ = 0.73, P = 0.016), and approached significance when correlated with baseline CMRO₂ in control subjects (ρ = 0.54, P = 0.105).

Discussion

Interpretation of apnea paradigm data

While a range of technologies have been used to study the pathophysiology of OSA, to the best of our knowledge, this is the first study to directly measure CMRO₂ and its change in response to apnea in OSA subjects. We highlight two main findings: (1) baseline CMRO₂ is lower in OSA subjects relative to controls and (2) there is a larger CMRO₂ decrease in response to apnea in OSA subjects. Given the small sample size of the study, and that confounding by age and/or BMI cannot entirely be excluded given that groups were not fully matched, these preliminary findings should be interpreted with caution and replicated in larger samples. However, the results are consistent with growing evidence that blunted autoregulatory mechanisms in OSA may contribute to OSA-associated neuropathology,^10–14,16 and suggest a potential role for CMRO₂ in studying these mechanisms.

The observed reduction in baseline CMRO₂ in OSA subjects is a consequence of both oxygen delivery (tCBF) and oxygen extraction (AVO₂D) reduction. Although OSA subjects had lower values in both tCBF and AVO₂D on average, results did not reach statistical significance. As mentioned, these negative results must be interpreted with caution given the limited sample size. Nevertheless, they suggest that CMRO₂ may provide a more sensitive marker of baseline metabolic dysfunction than blood flow or oxygen extraction alone. Recently, a similar study of OSA subjects and controls also found no group differences in baseline blood flow and no change in baseline blood flow in apneics treated with CPAP.¹² We emphasize that our observed CMRO₂ reduction in OSA subjects cannot be explained by brain atrophy, as the CMRO₂ is normalized to brain volume, and furthermore, brain mass was not significantly different between groups.

In additional to lower baseline CMRO₂, OSA subjects had a significantly larger decrease in CMRO₂ in response to apnea compared to controls. In contrast, the increase in flow in response to apnea (CVR) was nearly identical between groups, with the reduced CMRO₂ apneic response largely driven by a greater reduction in AVO₂D in apneics. Just as CMRO₂ has been proposed as a better measure of baseline neuronal function than blood flow, our results suggest that the CMRO₂ response to vasoactive challenges may provide a more sensitive marker of regulatory dysfunction than flow-based CVR. While some studies have associated OSA with reduced CVR,^10–12 others, including the present study, have not.^13,35 More so than CVR reduction alone, inability to maintain CMRO₂ during apnea provides a potential mechanism to explain the development of hypoxic brain damage in OSA.

Although not reaching statistical significance, the control group in this study also displayed a negative CMRO₂ response to apnea. This contrasts with previous data in 10 young, non-obese healthy subjects with no major underlying medical conditions, where a small but significant (6.0 [3.5] %, P = 0.0004) increase in CMRO₂ was observed during apnea using the same imaging protocol.³⁰ One likely explanation for the different CMRO₂ apneic response in these two control groups is the differences in their clinical characteristics. Controls in the present study were recruited among older, relatively obese subjects from a Sleep Clinic. These subjects likely have a higher prevalence of obesity, high blood pressure, and metabolic syndrome compared to the previously studied young healthy cohort.

AHI correlated significantly with baseline CMRO₂ across all subjects as well as in apneics only, demonstrating that baseline CMRO₂ may be sensitive not just to apnea status but also apnea disease severity. In contrast, AHI correlation with the CMRO₂ apneic response was significant among all subjects, but only a trend when restricted to apneics. Thus, studies with more subjects are needed to determine the relative sensitivity of baseline vs. apneic response CMRO₂ to OSA disease state.

Study limitations and future directions

Our study has several limitations. First, it was performed in a relatively small sample of apneics and controls. A larger sample size would increase power to detect group differences, some of which were marginally significant. Thus, negative results (P > 0.05) should be interpreted with caution. However, our study did observe significant differences in our primary outcomes of interest, and, in general, power was approximately 75% to observe a mean difference between apneics and controls of 1.25 standard deviations.

In this study, OSA subjects were significantly older and more obese than controls. While these differences reflect differences between OSA and non-OSA patients in the general population, the small sample and limited covariate overlap restricted our ability to control for potential confounders which could affect CMRO₂ – such as age and BMI – in statistical models.

While some recent work suggests that baseline CMRO₂ may increase slightly with age,³⁶ such a trend would have biased our results towards the null. In our data, there was a lack of association between CMRO₂ measures and age in the overall sample, further suggesting that age did not confound the relationship between OSA and CMRO₂ measures in our study. While a significant positive correlation was observed between age and the CMRO₂ apneic response in controls only (ρ = 0.73, P = 0.016), it is unclear why this effect would exist only in controls, and, therefore, this result should be interpreted with caution.

The relationship between BMI and CMRO₂ has not been specifically investigated previously, and presents an added complication in our study, as obesity may affect the subjects’ ability to hold their breath. For instance, obesity lowers functional residual capacity, which could potentially cause more significant hypoxia or hypercapnia to develop during volitional apnea, impacting the apneic response. While we observed a negative correlation between CMRO₂ and BMI across the entire sample (ρ = −0.40, P = 0.070), this does not necessarily demonstrate confounding by BMI, as such a trend would in fact be expected if CMRO₂ were independently associated with OSA, given the BMI mismatch between apneics and controls. Furthermore, the fact that the same trend was not observed among control subjects only (ρ = 0.25, P = 0.49) suggests that in the absence of OSA pathology, CMRO₂ is not independently associated with lower BMI, thus arguing against BMI effects driving our observed group differences. Finally, the observed correlation between BMI and CMRO₂ was only 62% as strong as the correlation between AHI and CMRO₂ (ρ = −0.65, P = 0.0014). Thus, it is possible that the observed CMRO₂ correlation with BMI is due to an independent CMRO₂ association with AHI, rather than BMI. Future, larger studies including subjects and controls with a similar range of BMIs could better assess whether BMI independently affects CMRO₂, and whether any such effects exist across the general population or are restricted to apneics alone.

CMRO₂ group differences observed in the overall study sample were similar to those in the subset of age and BMI matched subjects, again arguing in favor of OSA independently lowering baseline CMRO₂ and the CMRO₂ apneic response. However, it will be essential for larger independent studies to confirm the observed associations. Furthermore, future studies should examine the possible effects of additional OSA-associated co-morbidities, including type 2 diabetes.

Two subjects (both apneics) were excluded from data analysis due to inability to stay awake between breath-holds. Compliance could be improved in future studies via modification of the breath-hold protocol to include more frequent, shorter apneas, or by introducing visual cuing of the breath-holds, which, in addition to sustaining subject attention, has been shown to improve breath-hold reproducibility.³⁷ However, even with perfect subject compliance, volitional apnea may not elicit the same neurovascular response as true sleep-associated nocturnal apneas. To address this, our methodology could be applied during sleep to capture true apneic events in OSA subjects. The feasibility of high temporal resolution MR imaging during sleep has been demonstrated in previous studies examining airway closure in OSA.³⁸

A limitation of the OxFlow technique is that it is confined to quantification of global blood flow, oxygenation, and CMRO₂. Though apnea, like hypercapnia and hypoxia, can be thought of as a global cerebrovascular challenge, OSA has been associated with focal brain lesions,^4,5 and a recent study employing BOLD fMRI detected regional differences in CVR in OSA subjects.¹² Unfortunately, CMRO₂ mapping techniques are still at a developmental stage, with temporal resolutions on the order of many minutes and image noise levels often requiring whole-brain averaging to achieve physiologically plausible parameter values. While mapping of BOLD fMRI signal is possible at high temporal resolution, it does not provide a direct measure of either brain oxygen metabolism or blood flow, but rather reflects a complex interplay between blood flow, tissue properties, and CMRO₂. A combination of both quantitative global measures, such as OxFlow, and qualitative but spatially resolved measures such as BOLD fMRI, could offer an ideal approach for future studies of vascular and metabolic dysfunction in OSA.

Interpreting changes in baseline and apneic tCBF and CMRO₂ is challenging, as such alterations could be viewed as either a cause or effect of underlying neuropathology. In a study of baseline CMRO₂ and hypercapnic CVR in subjects with mild cognitive impairment (MCI),³⁹ reduced CMRO₂ in the setting of maintained CVR was interpreted as suggesting less demand for oxygen due to primary neural dysfunction, as opposed to failure to meet demand due to vascular dysfunction. In our study, reduced baseline CMRO₂ and maintained CVR were also observed, though the CMRO₂ apneic response was reduced, suggesting that there may be a component of supply-side deficiency in OSA not accounted for by CVR. Longitudinal monitoring of CMRO₂ in OSA could help to discriminate between these various interpretations, for instance, by determining whether changes in baseline CMRO₂ and CMRO₂ apneic response occur concurrently or serially in OSA disease progression, and the extent to which CPAP therapy and resulting neurocognitive improvements are reflected by CMRO₂ metrics. The simplicity, speed, and robustness of the OxFlow technique make it well suited for such applications.

Conclusion

In summary, our results suggest that baseline CMRO₂ is reduced in OSA subjects relative to controls, and, furthermore, that OSA subjects may fail to maintain normal CMRO₂ during apnea. These findings add to the growing evidence that OSA-associated neuropathology is a consequence of autoregulatory dysfunction. MRI-based quantification of CMRO₂ may offer a new method for better understanding the mechanisms of neurologic impairment in patients with sleep apnea.

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by NIH T32-EB000814, R21-HD069390, R01-HL109545, and R01-HL122754, and by the Institute of Translational Medicine and Therapeutics of the University of Pennsylvania (grant number UL1RR024134 from the National Center for Research Resources).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

ZBR designed the study and study methods, performed the study experiments, analyzed the data, and wrote the manuscript; SEL designed the study and study methods, performed the study experiments, and analyzed the data; BTK analyzed the data and edited and reviewed the manuscript; LGK designed the study methods; RJS and FWW designed the study and study methods and edited and reviewed the manuscript.

References

Somers

White

Amin

. Sleep apnea and cardiovascular disease: an American Heart Association/American College of Cardiology Foundation Scientific Statement from the American Heart Association Council for High Blood Pressure Research Professional Education Committee, Council on Clinical Cardiology, Stroke Council, and Council on Cardiovascular Nursing. J Am Coll Cardiol 2008; 52: 686–717.

Pack

Gislason

. Obstructive sleep apnea and cardiovascular disease: a perspective and future directions. Prog Cardiovasc Dis 2009; 51: 434–451.

Kushida

Nichols

Quan

. The Apnea Positive Pressure Long-term Efficacy Study (APPLES): rationale, design, methods, and procedures. J Clin Sleep Med 2006; 2: 288–300.

Macey

Henderson

Macey

. Brain morphology associated with obstructive sleep apnea. Am J Respir Crit Care Med 2002; 166: 1382–1387.

Macey

Kumar

Woo

. Brain structural changes in obstructive sleep apnea. Sleep 2008; 31: 967–977.

Globus

Busto

Martinez

. Comparative effect of transient global ischemia on extracellular levels of glutamate, glycine, and gamma-aminobutyric acid in vulnerable and nonvulnerable brain regions in the rat. J Neurochem 1991; 57: 470–478.

Gooden

. Mechanism of the human diving response. Integr Physiol Behav Sci 1994; 29: 6–16.

Rodgers

Jain

Englund

. High temporal resolution MRI quantification of global cerebral metabolic rate of oxygen consumption in response to apneic challenge. J Cereb Blood Flow Metab 2013; 33: 1514–1522.

Liu

Pascual

. Effect of hypoxia and hyperoxia on cerebral blood flow, blood oxygenation, and oxidative metabolism. J Cerebral Blood Flow Metab 2012; 32: 1909–1918.

10.

Placidi

Diomedi

Cupini

. Impairment of daytime cerebrovascular reactivity in patients with obstructive sleep apnoea syndrome. J Sleep Res 1998; 7: 288–292.

11.

Reichmuth

Dopp

Barczi

. Impaired vascular regulation in patients with obstructive sleep apnea: effects of continuous positive airway pressure treatment. Am J Respir Crit Care Med 2009; 180: 1143–1150.

12.

Prilipko

Huynh

Thomason

. An fMRI study of cerebrovascular reactivity and perfusion in obstructive sleep apnea patients before and after CPAP treatment. Sleep Med 2014; 15: 892–898.

13.

Urbano

Roux

Schindler

. Impaired cerebral autoregulation in obstructive sleep apnea. J Appl Physiol 2008; 105: 1852–1857.

14.

Macey

Kumar

Ogren

. Global brain blood-oxygen level responses to autonomic challenges in obstructive sleep apnea. PloS ONE 2014; 9: e105261.

15.

Lefebvre

Godin-Ribuot

Joyeux-Faure

. Functional assessment of vascular reactivity after chronic intermittent hypoxia in the rat. Respir Physiol Neurobiol 2006; 150: 278–286.

16.

Foster

Brugniaux

Pialoux

. Cardiovascular and cerebrovascular responses to acute hypoxia following exposure to intermittent hypoxia in healthy humans. J Physiol 2009; 587: 3287–3299.

17.

. Noninvasive quantification of whole-brain cerebral metabolic rate of oxygen (CMRO2) by MRI. Magnet Resonance Med 2009; 62: 141–148.

18.

Jain

Langham

Wehrli

. MRI estimation of global brain oxygen consumption rate. J Cerebral Blood Flow Metab 2010; 30: 1598–1607.

19.

Brier

. The influence of carbon dioxide on brain activity and metabolism in conscious humans. J Cerebral Blood Flow Metab 2011; 31: 58–67.

20.

Jain

Langham

Floyd

. Rapid magnetic resonance measurement of global cerebral metabolic rate of oxygen consumption in humans during rest and hypercapnia. J Cerebral Blood Flow Metab 2011; 31: 1504–1512.

21.

Ishii

Kitagaki

Kono

. Decreased medial temporal oxygen metabolism in Alzheimer's disease shown by PET. J Nuclear Med 1996; 37: 1159–1165.

22.

Borghammer

Vafaee

Ostergaard

. Effect of memantine on CBF and CMRO2 in patients with early Parkinsons disease. Acta Neurol Scandinavica 2008; 117: 317–323.

23.

Zhang

. Characterizing brain oxygen metabolism in patients with multiple sclerosis with T2-relaxation-under-spin-tagging MRI. J Cerebral Blood Flow Metab 2012; 32: 403–412.

24.

Blockley

Griffeth

VEM

Simon

. A review of calibrated blood oxygenation level-dependent (BOLD) methods for the measurement of task-induced changes in brain oxygen metabolism. NMR Biomed 2012; 26: 987–1003.

25.

Rodgers

Englund

Langham

. Rapid T2- and susceptometry-based CMRO2 quantification with interleaved TRUST (iTRUST). Neuroimage 2015; 106: 441–450.

26.

Jain

Abdulmalik

Propert

. Investigating the magnetic susceptibility properties of fresh human blood for noninvasive oxygen saturation quantification. Magnet Resonance Med 2012; 68: 863–867.

27.

Fernandez-Seara

Techawiboonwong

Detre

. MR susceptometry for measuring global brain oxygen extraction. Magn Reson Med 2006; 55: 967–973.

28.

Doyle

Walsh

Blackwell

. Block regional interpolation scheme for k-space (BRISK): a rapid cardiac imaging technique. Magnet Resonance Med 1995; 33: 163–170.

29.

Van Vaals

Brummer

Dixon

. “Keyhole” method for accelerating imaging of contrast agent uptake. J Magnet Resonance Imag 1993; 3: 671–675.

30.

Rodgers ZB, Kini L, Jain V, et al. High temporal resolution, simultaneous quantification of intravascular blood flow and oxygen saturation with BRISK k-space sampling. In: Proceedings of the 21st annual meeting of ISMRM 20–26 April 2013, Salt Lake City, Utah, USA, p. 1351.

31.

Thevenaz

Ruttimann

Unser

. A pyramid approach to subpixel registration based on intensity. IEEE Trans Image Process 1998; 7: 27–41.

32.

Smith

. Fast robust automated brain extraction. Hum Brain Mapp 2002; 17: 143–155.

33.

Herscovitch

Raichle

. What is the correct value for the brain–blood partition coefficient for water? J Cereb Blood Flow Metab 1985; 5: 65–69.

34.

Batzel

. Cardiovascular and respiratory systems: Modeling, analysis, and control, Philadelphia, PA: Society for Industrial and Applied Mathematics, 2007.

35.

Foster

Hanly

Ostrowski

. Ventilatory and cerebrovascular responses to hypercapnia in patients with obstructive sleep apnoea: effect of CPAP therapy. Respir Physiol Neurobiol 2009; 165: 73–81.

36.

Peng

Dumas

Park

. Age-related increase of resting metabolic rate in the human brain. Neuroimage 2014; 98: 176–183.

37.

Thomason

Glover

. Controlled inspiration depth reduces variance in breath-holding-induced BOLD signal. Neuroimage 2008; 39: 206–214.

38.

Shin

Holbrook

Capasso

. Improved sleep MRI at 3 tesla in patients with obstructive sleep apnea. J Magnet Resonance Imag 2013; 38: 1261–1266.

39.

Binu PT, Sheng M, Tseng B, et al. Characterization of CMRO2, resting CBF, and cerebrovascular reactivity in patients with very early stage of Alzheimer’s Disease. In: Proceedings of the 21st annual ISMRM scientific meeting 20–26 April 2013, Salt Lake City, Utah, USA, p. 619.

Cerebral metabolic rate of oxygen in obstructive sleep apnea at rest and in response to breath-hold challenge

Abstract

Keywords

Introduction

Materials and methods

Susceptometry-based quantification of Yv

Combined SBO and PC-MRI for CMRO2 quantification (OxFlow)

Subjects

Experimental procedures

Volitional apnea paradigm

MR imaging protocol

Data processing

Statistical analysis

Results

Subject demographic and polysomnography group characteristics

Baseline differences between OSA subjects and controls

Apneic response in OSA subjects and controls

Relationship between CMRO2 and AHI

Age and BMI effects analysis

Discussion

Interpretation of apnea paradigm data

Study limitations and future directions

Conclusion

Footnotes

Funding

Declaration of conflicting interests

Authors’ contributions

References

Susceptometry-based quantification of Y_v

Combined SBO and PC-MRI for CMRO₂ quantification (OxFlow)

Relationship between CMRO₂ and AHI