Evaluation of Cerebrovascular Reactivity in Subjects with and without Obstructive Sleep Apnea

ARTICLE IN PRESS

Evaluation of Cerebrovascular Reactivity in Subjects with and without Obstructive Sleep Apnea Clodagh M. Ryan, MD,*,† Anne Battisti-Charbonney, MSc,‡ Olivia Sobczyk, MSc,§ David J. Mikulis, MD,‡,§ James Duffin, MD,‖ Joseph A. Fisher, PhD,§,¶ and Lashmi Venkatraghavan, MD¶

Background: Both obstructive sleep apnea (OSA) and altered cerebrovascular reactivity (CVR) are associated with increased stroke risk. Nevertheless, the incidence of abnormal CVR in patients with OSA is uncertain due to the high variability in the way CVR is measured both within and between studies. We hypothesized that a standardized CVR with a consistent vasoactive stimulus and cerebral blood flow (CBF) measure would be reduced in patients with severe OSA compared with healthy controls. Methods: This was a prospective study in which subjects with and without OSA were administered a standardized hypercapnic stimulus, and CBF was monitored by blood oxygen level-dependent magnetic resonance signal changes, a high space and time resolved surrogate for CBF. Results: Twentyfour subjects with OSA (mean age 45.9 years, apnea–hypopnea index [AHI] 26.8 per hour) and 6 control subjects (mean age 42.8 years, AHI 2.4 per hour) were included. Compared with controls, subjects with OSA had a significantly greater whole brain (.1565 versus .1094, P = .013), gray matter (.2077 versus .1423, P = .009), and white matter (.1109 versus .0768, P = .024) CVR, respectively. Conclusions: Contrary to expectations, subjects with OSA had greater CVR compared with control subjects. Key Words: Obstructive sleep apnea—cerebrovascular reactivity—cerebral blood flow—cerebrovascular disease. © 2017 National Stroke Association. Published by Elsevier Inc. All rights reserved.

Introduction From the *Toronto General Hospital, University Health Network, Toronto, Ontario, Canada; †Department of Medicine, University of Toronto, Toronto, Ontario, Canada; ‡Joint Department of Medical Imaging and the Functional Neuroimaging Laboratory, University Health Network, Toronto, Ontario, Canada; §Institute of Medical Sciences; ‖Departments of Physiology; and ¶Department of Anesthesia and Pain Management, University Health Network, University of Toronto, Toronto, Ontario, Canada. Received September 30, 2016; accepted August 13, 2017. Grant support: A grant-in-aid from the Ontario Thoracic Society supported this study. Notation of prior abstract publication and presentation: Some results of this study were presented in a form of poster presentation at the American Thoracic Society Conference, 2013 (Philadelphia, USA). Address correspondence to Clodagh M. Ryan, MD, 9N-967 Toronto General Hospital, 585 University Ave., Toronto, ON M5G 2N2. E-mail: [email protected]. 1052-3057/$ - see front matter © 2017 National Stroke Association. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2017.08.015

Obstructive sleep apnea (OSA) is a sleep breathing disorder that occurs in approximately 9%-24% of middleaged people. Patients with OSA have more than 3-fold increase in the risk of stroke and death independent of other known risk factors for stroke.1 Obstructive episodes are associated with severe hypoxia, hypertension, and hypercapnia, which may acutely disturb the relationship between brain oxygen requirements and supply, and in the long term may disturb normal brain blood flow regulation. However, the direct pathophysiological link between OSA and stroke remains unknown. Our aim was to study the brain blood flow regulation in patients with OSA. Brain blood flow regulation has been studied as the cerebral blood flow (CBF) response to a vasoactive stimulus, termed cerebrovascular reactivity (CVR). Altered CVR has been shown to be a strong predictor of stroke and transient ischemic attacks in patients with carotid artery

Journal of Stroke and Cerebrovascular Diseases, Vol. ■■, No. ■■ (■■), 2017: pp ■■–■■

1

ARTICLE IN PRESS C.M. RYAN ET AL.

2 2

disease. However, for the most part, CVR has been performed using either uncontrolled hypercapnic stimuli such as infusion of carbon dioxide (CO2) into a mask3 or breath holding or administration of acetazolamide, all of which result in high variability in stimuli between studies and between patients (see Fierstra et al for discussion4). Furthermore, most studies have used transcranial Doppler (TCD)-measured middle cerebral artery blood velocity as a surrogate for CBF under the assumption that the middle cerebral artery diameter remains constant with hypercapnia. More recent studies have impugned this belief.5 Moreover, TCD provides good temporal resolution but poor spatial resolution limited to an entire vascular territory, leaving unanswered the relationship between flow velocity and actual flow, and between TCD and other surrogate flow measures such as xenon-enhanced computed tomography and positron-emission tomography imaging. Such differences in the methodology of CVR measures have resulted in an inconsistency in the findings6 between studies and even within subjects.7,8 In the present study, we measured CVR by applying a standardized hypercapnic stimulus and using the blood oxygen level-dependent (BOLD) magnetic resonance signal as a surrogate for CBF.9,10 Our hypothesis was that BOLD magnetic resonance imaging (MRI) CVR in patients with OSA would demonstrate a greater incidence of regional and/or global CVR abnormalities, compared with subjects without OSA. Some of the results have been previously reported in the form of an abstract.11

Methods Subjects Subjects between the ages of 40 and 60 years were recruited through local public advertisements. Exclusion criteria included those with (1) a known cardiac, respiratory, neurological, or major liver or kidney disease; (2) severe claustrophobia; (3) pregnancy; (4) resting SaO2 on room air of less than 95%; (5) diabetes; and (6) contraindications for MRI. All subjects were instructed to avoid smoking, heavy exercise, or caffeine on the day of the MRI. Informed written consent was obtained from all subjects, and the protocol was approved by the research ethics board of the University Health Network.

Protocol Following recruitment, the subjects had an overnight sleep study. The CVR measurements were made between 11 AM and 2 PM on the day after the study, or as soon as could be scheduled. Overnight Sleep Study The overnight sleep study was performed at the University Health Network sleep study laboratory using

standard techniques and criteria for scoring sleep stages, arousals, apneas, and hypopneas.12 Thoracoabdominal movements were monitored by a respiratory inductance plethysmograph (Respitrace; Ambulatory Monitoring Inc., White Plains, NY)13 and airflow was measured by a nasal pressure cannulae (BiNAPS; Salter Labs Inc., Arvin, CA). Arterial oxygen saturation (SaO2) was continuously monitored by a pulse oximeter (Nellcor; Sensormedics Corp., Anaheim, CA). An electrocardiogram was monitored. All the signals were recorded on a computerized sleep scoring system (Sandman; Nellcor Puritan Bennett Ltd., Ottawa, ON, Canada) Hypopneas were scored when there was a drop in the peak signal by 30% or higher of 10 seconds’ duration or longer, which was associated with a 3% or higher oxygen desaturation or arousal on the EEG. Hypopneas were scored as obstructive if there was thoracoabdominal paradoxical ribcage motion on the Respitrace, inspiratory flattening, or snoring. Otherwise, hypopneas were scored as central. The severity of sleep apnea was assessed by the number of apneas and hypopneas per hour of sleep (apnea– hypopnea index [AHI]). Subjects having an AHI of 5 per hour of sleep or higher were classified as having sleep apnea, and subjects with an AHI of less than 5 were classified as the nonsleep apnea (NSA) group.12 OSA was diagnosed when at least 85% of the respiratory events were of the obstructive type. The oxygen desaturation index (ODI) was defined as the number of oxygen desaturations per hour 3% or higher below baseline. The Epworth Sleepiness Scale, a subjective measure of daytime sleepiness, was completed on the night of the polysomnography. Imaging BOLD imaging was performed on a 3-T short-bore MRI system (GE Healthcare, Milwaukee, WI) using an 8-channel phased-array receiver coil. The BOLD acquisitions were obtained using an echo-planar gradient echo sequence (repetition time/echo time = 2000/30 milliseconds, 3.75 × 3.75 × 5.0 mm voxels, field of view 24 × 24 cm, 39 slices, slice thickness 5 mm, matrix size 64 × 64, number of frames = 254, flip angle = 85°).9 Each subject had an initial anatomical scan, following which the BOLD MRI scan was performed in conjunction with the dynamic hypercapnic stimulus.14-17 Hypercapnic Stimulus The hypercapnic stimulus was delivered via a closed face mask using skin tape to prevent gas leaks. All gas was supplied by a computer-controlled gas blender and sequential gas delivery circuit (Respiract; Thornhill Research Institute, Toronto, ON, Canada) using the prospective targeting algorithm of Slessarev et al.18 The repeatable stimulus took the form of 2 pseudo–square wave iso-oxic changes in the end-tidal carbon dioxide (PETCO2) of 10 mm Hg

ARTICLE IN PRESS CEREBROVASCULAR REACTIVITY IN OSA

above the resting baseline lasting 45 and 120 seconds separated by 90-second return to baseline.18-20

3

Table 1. Patient demographics and polysomnography results

Data Processing: CVR All MRI images and PETCO2 data were imported into Analysis of Functional Neuroimages software.21 PETCO2 was time aligned with the whole-brain average time course (MATLAB 6.5; MathWorks, Natick, MA), and the raw BOLD signal was converted into percentage change from baseline values. A linear regression was performed between PETCO2 and the BOLD signal. The slope of the regression is defined as CVR. Therefore, CVR was the change in CBF per unit change in the vasodilatory stimulus (PETCO2). CVR values were color coded on a color scale and mapped onto the anatomical scan to form CVR maps.9,14 Technicians were blinded to the result of the polysomnogram. CVR were segmented into gray matter (GM) and white matter (WM) (SPM8; Wellcome Department of Imaging, Neuroscience, Institute of Neurology, University College London, London, UK). Statistical Analysis Data are expressed as the mean ± standard error of the mean and n for categorical data. Subject characteristics, sleep variables, and CVR values were compared between patients with and without OSA by independent t-tests for continuous variables and the Fisher exact test for categorical variables. One-way analysis of variance using the Bonferroni multiple comparison test or a Kruskal-Wallis with Dunn multiple comparison was performed, depending on the normal distribution, to compare the GM and the WM between the healthy control and the OSA groups. Only those subjects who attended for both the overnight sleep study and the BOLD MRI were included in the final analysis. A P value of .05 or lower defined statistical significance. All analyses were performed with SPSS 12.0 (SPSS Inc., Chicago, IL, USA).

Results Forty-six subjects were recruited of whom 30 completed the protocol. Of those subjects who did not complete the study, 10 subjects failed to attend both the sleep study and the MRI study. Six subjects attended the MRI study, but 4 of these withdrew following the trial of tolerance hypercapnia before entering the MRI scanner, and in 2, the scans were rejected for analysis due to movement artifacts. Of the 30 subjects who completed the full protocol, 24 subjects were categorized as OSA positive and 6 as nonOSA (NSA). There were no significant differences between groups for age, sex distribution, body mass index or Epworth Sleepiness Scale. Compared with the NSA group, the OSA group had more sleep fragmentation, a higher

Age (y) Sex, M : F BMI (kg/m2) ESS AHI, n (h) OI, n (h) TST (h) SE (%) N1 (%) N2 (%) N3 (%) Stage R (%) ARl I, n (h) PLMI, n(h) Mean SaO2, % Min SaO2, % ODI, n (h) T90 (%) Pre-MRI PetCO2 (mm Hg)

NSA (AHI <5) (n = 6)

OSA (AHI ≥5) (n = 24)

P Value

42.8 ± 4.1 6:0 26.5 ± 1.1 7.4 ± .2 2.4 ± .4 1.9 ± .5 5.1 ± .6 76.3 ± 9.1 5.9 ± 1.6 64.1 ± 6.2 15.2 ± 4.9 14.8 ± 3.7 13.2 ± 1.8 3.2 ± 2.9 94.7 ± .43 90.0 ± 1.0 1.8 ± .39 .08 ± .05 37.7 ± 1.0

45.9 ± 2.9 16:4 30.1 ± .9 9.7 ± .9 26.8 ± 3.3 23.1 ± 3.5 5.5 ± .24 78.9 ± 3.1 11.1 ± 1.7 60.4 ± 2.6 13.7 ± 1.9 14.9 ± 1.3 26.2 ± 2.9 7.4 ± 3.4 94.7 ± .42 85.2 ± 1.3 17.3 ± 4.2 6.8 ± 2.5 39.5 ± .6

.534 .302 .091 .198 <.0001 <.0001 .339 .978 .214 .589 .696 .897 .009 .918 .394 .079 .039 .035 .169

Abbreviations: AHI, apnea–hypopnea index; ARl I, arousal index; BMI, body mass index; ESS, Epworth Sleepiness Scale; F, female; M, male; min, minimum; MRI, magnetic resonance imaging; N1, stage 1 sleep; N2, stage 2 sleep; N3, stage 3 sleep; NSA, no sleep apnea; ODI, oxygen desaturation index; OI, obstructive index; OSA, obstructive sleep apnea; PetCO2, end-tidal carbon dioxide; PLMI, periodic leg movement index; SaO2, oxygen saturation; SE, sleep efficiency; Stage R, stage REM sleep; T90, percentage of night spent with a SaO2 less than 90%; TST, total sleep time. All values are mean ± standard error of the mean.

ODI, and a greater percentage of the night spent with an oxygen saturation of less than 90%. Resting baseline PETCO2 was similar in both the OSA and the control groups (Table 1). The GM and the WM were significantly greater in the OSA group than in NSA (Fig 1, Table 2). This robust response is also seen in Figure 2. No correlation was demonstrated between the GM and the WM and the ODI or the arousal index. The subgroup analysis of the male subjects demonstrated a persistent robust difference between those with and without OSA for GM (P = .017) and WM (P = .016). There was a nonsignificant trend toward increasing CVR between those with mild, moderate, and severe OSA (Table 3).

Discussion In the present study, we set out to better document the expected lower CVR in patients with OSA compared with subjects with NSA by using a standardized stimulus and a high time and space resolved measure

ARTICLE IN PRESS C.M. RYAN ET AL.

4

Model of Pathological Redistribution of Blood Flow in Hypercapnia

0.2

0.2 0.1

WM

WHOLEBRAIN

0.3

0.0

0.1 0.0

-0.1 -0.2

NSA

-0.1

OSA

NSA

OSA

0.4 0.3

GM

0.2 0.1 0.0

-0.1 -0.2 -0.3

NSA

OSA

Figure 1. Box-and-whisker plots illustrating the blood oxygen leveldependent magnetic resonance imaging cerebrovascular reactivity between subjects with and without obstructive sleep apnea. The whiskers represent the maximum and the minimum values excluding the outliers. The “o” and the “*” represent the outlier values. The outliers are those values less than 1.5 times the lower quartiles. Abbreviations: GM, gray matter; NSA, no sleep apnea; OSA, obstructive sleep apnea; WM, white matter.

of CBF. Instead, our data showed that CVR in patients with OSA is greater than those with NSA with a trend to an increase in CVR related to the severity of the OSA; indeed none of the patients with OSA that we studied had a detectable cerebrovascular pathology. Nevertheless, we propose that the finding of an increase in CVR better explains the heightened risk of stroke in patients with OSA than a reduction in CVR. This explanation is best understood using the recently published model accounting for CVR,10 which we briefly review further. This model also explains abnormal distributions of CBF in the presence of localized vascular pathology and would explain the development of ischemia in susceptible patients during obstructive episodes.

Table 2. Blood oxygen level-dependent magnetic resonance imaging CVR

CVR Whole brain GM WM

NSA (AHI <5) (n = 6)

OSA (AHI ≥5) (n = 24)

P Value

.1094 ± .0163 .1423 ± .0244 .0768 ± .0084

.1565 ± .0164 .2077 ± .0229 .1109 ± .0113

.013 .009 .024

The model holds that, with progressive global vasodilatory stimulation, blood flow increases in cerebral vascular beds.10 However, total cerebral inflow through the major arteries is limited below the flow capacity of the intracranial vasculature.22,23 When the extracranial arterial inflow capacity is reached from progressive intracranial hyperemia (e.g., due to graded hypercapnia), this results in a competition for flow between the cerebral vascular beds. Vascular steal, a reduction in flow with progressive hypercapnia, occurs when a pathological imbalance in the intracerebral redistribution of blood flow develops between vascular beds.22,24 We can call this “pathological” because this is generally not seen in healthy subjects25-27,a and is usually due to a narrowing of vessels proximal to the vascular bed. During nocturnal respiratory obstructive episodes, the PaCO2 rises to an extent where the stimulated intracranial flow demand may exceed inflow capacity. In addition apneas associated with severe hypoxia may cause increased cerebral vasodilation. The degree of steal is dependent on the relationship between (1) the intracranial flow demand and extracranial flow capacity, (2) the magnitude of the vasodilatory stimulus, and (3) the discrepancy in the flow reserve between the pathological and healthy vascular beds.

Robust CVR Increases the Propensity for the Development of Brain Ischemia and Stroke According to the previously mentioned proposed model, the presence of vascular beds with supranormal robust responses to hypercapnea generates “steal” at lower PCO2 levels and in vessels with less compromise in reactivity, and would reduce the flow in a compromised vascular bed to a greater extent than do beds with normal reactivity. The other pathophysiological conditions related to OSA may contribute to the severity of blood flow disturbance and ischemia from apneic episodes. Endothelial dysfunction,28 carotid stenosis,29 and susceptibility to cardiac arrhythmias limit extracranial vessel blood supply. Hypercoagulability may predispose to thrombosis in vascular beds with reduced flow. All of these associations may predispose to compromised vascular beds and could predispose to greater vascular steal. However, these associations are observational and no data yet link them to vascular steal causally.

Previous Studies of CVR and OSA

Abbreviations: AHI, apnea–hypopnea index; CVR, cerebrovascular reactivity; GM, gray matter; NSA, no sleep apnea; OSA, obstructive sleep apnea; WM, white matter. All values are mean ± standard error of the mean.

Previous studies had found reductions30,31 or no change32 in CVR in OSA. Some authors have attributed the potential risk of stroke to the reduction in CVR31,33 as a

Although there are conditions where steal may be seen in certain locations in WM areas in healthy people, this is accounted for by the z scoring method we used.12

ARTICLE IN PRESS CEREBROVASCULAR REACTIVITY IN OSA

5

Figure 2. Cerebrovascular reactivity from a representative subject without (NSA) and with severe OSA. Note that, in the CVR map of the patient with severe OSA, the CVR ranges are shifted to the upper CVR values as represented by the warmer colors. Color scales are presented to the left of the rows. Abbreviations: CVR, cerebrovascular reactivity; NSA, no sleep apnea; OSA, obstructive sleep apnea. (Color version of figure is available online.)

supported by the empirical observations of the increased coincidence of stroke in both subjects with OSA1 and subjects with abnormal CVR.2 However, other studies found normal CVR in patients with OSA.32 In these studies, the CVR was performed using a range of unrepeatable hypercapnic stimuli such as breath holding and rebreathing from a bag; and used TCD flow velocity measures as surrogates for CBF. It has therefore been difficult to develop a consensus about the state of CVR in patients with OSA.

Where Our Study Extends the Current Understanding In our study, we used a standardized hypercapnic stimulus with BOLD MRI. This approach has not been previously used to evaluate CVR in subjects with OSA. Placidi et al and others31,34 used a “breath-hold index” where subjects were asked to hold their breath for 20 to 30 seconds after a “normal inhalation.” The PaCO2, which is the actual stimulus, is unknown during the breath hold. It is difficult to estimate beyond wide bounds as it varies, not only with duration of breath holding but also with the size of the “normal breath,” the body size, the metabolic

CO2 production, and recent activity, to name but a few. Breath holding also induces hypoxia, which is variable and also related not only to the duration but also to size of the breath, the residual volume of the lung, and the oxygen consumption, all of which are unmonitored. Droste et al32 induced hypercapnia by asking their subjects to rebreathe from a bag, permitting the end exhaled PCO2 to be measured after each breath and used to estimate the PaCO2. Otherwise, the same conditions affect the arterial PCO2 as with breath holding. Ameriso et al33 administered 6% CO2 in air via a nonrebreathing circuit. Urbano et al35 administered 5% CO2 in air. Although these studies seem to standardize the stimulus between subjects, the arterial PCO2 depends not just on the inspired PCO2 but also on the respiratory response (see Fierstra et al for discussion4). Most studies used TCD flow velocity as a surrogate of CBF. TCD has high temporal resolution but poor spatial resolution as it insonates only 1 major artery at a time. Of this sample, some found CVR in patients with OSA was reduced,8,31 and some found that CVRs were the same as those in the healthy cohort.32,35,36 The large variability in the stimulus

Table 3. Blood oxygen level-dependent magnetic resonance imaging CVR in OSA groups

CVR Whole brain GM WM

Mild OSA (AHI ≥5.0-14.9) n=4

Moderate OSA (AHI ≥15.0-29.9) n = 15

Severe OSA (AHI ≥30) n=5

P Value

.0939 ± .085 .1251 ± .120 .0663 ± .055

.1599 ± .012 .2096 ± .0175 .1161 ± .0009

.1962 ± .0166 .2679 ± .0192 .1310 ± .0158

.232 .169 .464

Abbreviations: AHI, apnea–hypopnea index; CVR, cerebrovascular reactivity; GM, gray matter; OSA, obstructive sleep apnea; WM, white matter.

ARTICLE IN PRESS C.M. RYAN ET AL.

6

and the difficulty in normalizing the response for it make the findings regarding CVR uncertain. In contrast, our study applied a standardized stimulus of 10 ± 2 mm Hg in a standardized pattern. The flow surrogate, BOLD MRI, provided good spatial resolution throughout the brain. In the present study, we used BOLD MRI as a surrogate for CBF. The signal relies on a difference of magnetic susceptibility between oxyhemoglobin and deoxyhemoglobin representing the net balance between oxygen delivery and consumption. The BOLD MRI has high time and spatial resolution of CBF, in contrast to TCD, which insonates only a single vessel.

Neurophysiology We propose that the elevated CVR in our group of subjects with OSA may be a result of either (1) a primary mechanism designed to promote ventilatory stability, or (2) a compensatory mechanism developed by subjects to induce metabolite clearance,37 or (3) a compensatory mechanism to maintain normal pH in the brain and brain oxygen delivery during the hypercapnic hypoxia associated with the multiple prolonged apneic episodes. If the elevated CVR is a primary mechanism, it may not improve with treatment; however, if it is compensatory, the treatment of OSA may cause the CVR to revert to normal levels.

Limitations The present study has a number of limitations. Our subjects with OSA were of similar age profiles, and it is uncertain whether or not CVR alters over time and with duration of disease. Although the groups have been matched for age and sex, those in the OSA group were more obese. We do not know whether or not obesity, independent of OSA, has an impact on CVR38 secondary to the downstream effects of obesity on oxidative stress and inflammation. Furthermore, we did not perform vascular imaging to rule out silent cardiovascular pathology. Follow-up BOLD MRI scans were not performed on this population to evaluate whether or not the treatment of OSA reduced CVR. As the CVR studies were performed during the daytime, we did not assess for variability in CVR and cannot comment on diurnal variation. Finally, the effect of hypercapnia on arterial blood pressure and mean arterial pressure was not evaluated in the present study, nor was sympathetic nervous activity. Both blood pressure responses and baseline sympathetic nervous activity may impact the responses of the cerebral vasculature.36 In particular, elevation in peripheral blood pressure demonstrated with hypercapnia36 has an unknown effect on the cerebral perfusion pressure, which in turn impacts cerebral vessel resistance and CBF.

Summary and Conclusion In summary, contrary to our initial hypothesis, the present study demonstrates increased BOLD MRI CVR in subjects with OSA without known comorbid illnesses. We propose that this increased daytime hypercapnic CVR is a spillover effect, reflecting the ongoing nocturnal hemodynamic oscillations due to repetitive hypoxia and hypercapnia in OSA. This robust response to CO2 increases the potential for steal when there is a reduced regional vascular reactivity from steno-occlusive arteriopathies. Over time, the development of a steal phenomenon secondary to vascular narrowing or simply a dysfunction such as vasculitis in subjects with OSA may portend an increased risk of stroke. To further investigate this suggested synergy of CVR and OSA in the cause of stroke, we suggest longitudinal studies evaluating BOLD MRI CVR on those subjects with OSA without cerebrovascular steno-occlusive disease to detect the development of steal and to evaluate the risk of stroke. Certainly, in those with stroke, both OSA39 and abnormal CVR40 are highly prevalent. Finally, determining whether CVR changes after the treatment of OSA would establish whether or not the robust response is reactive or reversible, and what effect treatment has on the incidence of stroke.

References 1. Arzt M, Young T, Finn L, et al. Association of sleepdisordered breathing and the occurrence of stroke. Am J Respir Crit Care Med 2005;172:1447-1451. 2. Markus H, Cullinane M. Severely impaired cerebrovascular reactivity predicts stroke and TIA risk in patients with carotid artery stenosis and occlusion. Brain 2001;124:457-467. 3. Mark CI, Fisher JA, Pike GB. Improved fMRI calibration: precisely controlled hyperoxic versus hypercapnic stimuli. Neuroimage 2011;54:1102-1111. 4. Fierstra J, Sobczyk O, Battisti-Charbonney A, et al. Measuring cerebrovascular reactivity: what stimulus to use? J Physiol 2013;591:5809-5821. 5. Verbree J, Bronzwaer AS, Ghariq E, et al. Assessment of middle cerebral artery diameter during hypocapnia and hypercapnia in humans using ultra-high-field MRI. J Appl Physiol 2014;117:1084-1089. 6. Totaro R, Marini C, Baldassarre M, et al. Cerebrovascular reactivity evaluated by transcranial Doppler: reproducibility of different methods. Cerebrovasc Dis 1999;9:142-145. 7. Foster GE, Hanly PJ, Ostrowski M, et al. Effects of continuous positive airway pressure on cerebral vascular response to hypoxia in patients with obstructive sleep apnea. Am J Respir Crit Care Med 2007;175:720725. 8. Morgan BJ, Reichmuth KJ, Peppard PE, et al. Effects of sleep disordered breathing on cerebrovascular regulation: a population-based study. Am J Respir Crit Care Med 2010;182:1445-1452. 9. Prisman E, Slessarev M, Han J, et al. Comparison of the effects of independently-controlled end-tidal PCO(2) and PO(2) on blood oxygen level-dependent (BOLD) MRI. J Magn Reson Imaging 2008;27:185-191.

ARTICLE IN PRESS CEREBROVASCULAR REACTIVITY IN OSA 10. Sobczyk O, Battisti-Charbonney A, Fierstra J, et al. A conceptual model for CO-induced redistribution of cerebral blood flow with experimental confirmation using BOLD MRI. Neuroimage 2014;92:56-68. 11. Ryan CM, Battisti A, Duffin J, et al. Spatial and temporal patterns of cerebrovascular response to hypercapnic stimuli in subjects with, and without, obstructive sleep apnea. Am J Respir Crit Care Med 2013;187:A2316. 12. Berry RB, Brooks R, Gamaldo CE, et al. The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specifications. Darien, Illinois: American Academy of Sleep Medicine, 2012. 13. Chadha TS, Watson H, Birch S, et al. Validation of respiratory inductive plethysmography using different calibration procedures. Am Rev Respir Dis 1982;125:644649. 14. Mandell DM, Han JS, Poublanc J, et al. Selective reduction of blood flow to white matter during hypercapnia corresponds with leukoaraiosis. Stroke 2008;39:1993-1998. 15. Mandell DM, Han JS, Poublanc J, et al. Mapping cerebrovascular reactivity using blood oxygen leveldependent MRI in Patients with arterial steno-occlusive disease: comparison with arterial spin labeling MRI. Stroke 2008;39:2021-2028. 16. Spano VR, Mandell DM, Poublanc J, et al. CO2 blood oxygen level-dependent MR mapping of cerebrovascular reserve in a clinical population: safety, tolerability, and technical feasibility. Radiology 2013;266:592-598. 17. Mutch WA, Mandell DM, Fisher JA, et al. Approaches to brain stress testing: BOLD magnetic resonance imaging with computer-controlled delivery of carbon dioxide. PLoS ONE 2012;7:e47443. 18. Slessarev M, Han J, Mardimae A, et al. Prospective targeting and control of end-tidal CO 2 and O 2 concentrations. J Physiol 2007;581:1207-1219. 19. Ito S, Mardimae A, Han J, et al. Non-invasive prospective targeting of arterial P(CO2) in subjects at rest. J Physiol 2008;586:3675-3682. 20. Somogyi RB, Vesely AE, Preiss D, et al. Precise control of end-tidal carbon dioxide levels using sequential rebreathing circuits. Anaesth Intensive Care 2005;33:726732. 21. Cox RW. AFNI: software for analysis and visualization of functional magnetic resonance neuroimages. Comput Biomed Res 1996;29:162-173. 22. Faraci FM, Baumbach GL, Heistad DD. Cerebral circulation: humoral regulation and effects of chronic hypertension. J Am Soc Nephrol 1990;1:53-57. 23. Iadecola C, Nedergaard M. Glial regulation of the cerebral microvasculature. Nat Neurosci 2007;10:1369-1376. 24. Brawley BW. The pathophysiology of intracerebral steal following carbon dioxide inhalation, an experimental study. Scand J Clin Lab Invest Suppl 1968;102:XIII:B. 25. Bhogal AA, Philippens ME, Siero JC, et al. Examining the regional and cerebral depth-dependent BOLD

7

26.

27.

28.

29.

30.

31.

32.

33.

34.

35.

36.

37.

38.

39. 40.

cerebrovascular reactivity response at 7T. Neuroimage 2015;114:239-248. Bhogal AA, Siero JC, Fisher JA, et al. Investigating the non-linearity of the BOLD cerebrovascular reactivity response to targeted hypo/hypercapnia at 7T. Neuroimage 2014;98:296-305. Sobczyk O, Battisti-Charbonney A, Poublanc J, et al. Assessing cerebrovascular reactivity abnormality by comparison to a reference atlas. J Cereb Blood Flow Metab 2015;35:213-220. Ip MS, Tse HF, Lam B, et al. Endothelial function in obstructive sleep apnea and response to treatment. Am J Respir Crit Care Med 2004;169:348-353. Drager LF, Bortolotto LA, Lorenzi MC, et al. Early signs of atherosclerosis in obstructive sleep apnea. Am J Respir Crit Care Med 2005;172:613-618. Loeppky JA, Miranda FG, Eldridge MW. Abnormal cerebrovascular responses to CO2 in sleep apnea patients. Sleep 1984;7:97-109. Placidi F, Diomedi M, Cupini LM, et al. Impairment of daytime cerebrovascular reactivity in patients with obstructive sleep apnoea syndrome. J Sleep Res 1998; 7:288-292. Droste DW, Ludemann P, Anders F, et al. Middle cerebral artery blood flow velocity, end-tidal pCO2 and blood pressure in patients with obstructive sleep apnea and in healthy subjects during continuous positive airway pressure breathing. Neurol Res 1999;21:737-741. Ameriso SF, Mohler JG, Suarez M, et al. Morning reduction of cerebral vasomotor reactivity. Neurology 1994;44:1907-1909. Buterbaugh J, Wynstra C, Provencio N, et al. Cerebrovascular reactivity in young subjects with sleep apnea. Sleep 2015;38:241-250. Urbano F, Roux F, Schindler J, et al. Impaired cerebral autoregulation in obstructive sleep apnea. J Appl Physiol 2008;105:1852-1857. Ryan CM, Battisti-Charbonney A, Sobczyk O, et al. Normal hypercapnic cerebrovascular conductance in obstructive sleep apnea. Respir Physiol Neurobiol 2014; 190:47-53. Xie L, Kang H, Xu Q, et al. Sleep drives metabolite clearance from the adult brain. Science 2013;342:373377. Rodriguez-Flores M, Garcia-Garcia E, Cano-Nigenda CV, et al. Relationship of obesity and insulin resistance with the cerebrovascular reactivity: a case control study. Cardiovasc Diabetol 2014;13:2. Dyken ME, Im KB. Obstructive sleep apnea and stroke. Chest 2009;136:1668-1677. Geranmayeh F, Wise RJ, Leech R, et al. Measuring vascular reactivity with breath-holds after stroke: a method to aid interpretation of group-level BOLD signal changes in longitudinal fMRI studies. Hum Brain Mapp 2015; 36:1755-1771.