Routine Clinical Evaluation of Cerebrovascular Reserve Capacity Using Carbogen in Patients With Intracranial Stenosis
Background and Purpose—A promising method for identifying hemodynamic impairment that may serve as a biomarker for stroke risk in patients with intracranial stenosis is cerebrovascular reactivity (CVR) mapping using noninvasive MRI. Here, abilities to measure CVR safely in the clinic using hypercarbic hyperoxic (carbogen) gas challenges, which increase oxygen delivery to tissue, are investigated.
Methods—In sequence with structural and angiographic imaging, blood oxygenation level–dependent carbogen-induced CVR scans were performed in patients with symptomatic intracranial stenosis (n=92) and control (n=10) volunteers, with a subgroup of patients (n=57) undergoing cerebral blood flow–weighted pseudocontinuous arterial spin labeling CVR. Subjects were stratified for 4 substudies to evaluate relationships between (1) carbogen and hypercarbic normoxic CVR in healthy tissue (n=10), (2) carbogen cerebral blood flow CVR and blood oxygenation level–dependent CVR in intracranial stenosis patients (n=57), (3) carbogen CVR and clinical measures of disease in patients with asymmetrical intracranial atherosclerotic (n=31) and moyamoya (n=29) disease, and (4) the CVR scan and immediate and longer-term complications (n=92).
Results—Noninvasive blood oxygenation level–dependent carbogen-induced CVR values correlate with (1) lobar hypercarbic normoxic gas stimuli in healthy tissue (R=0.92; P<0.001), (2) carbogen-induced cerebral blood flow CVR in patients with intracranial stenosis (R=0.30–0.33; P<0.012), and (3) angiographic measures of disease severity both in atherosclerotic and moyamoya patients after appropriate processing. No immediate stroke-related complications were reported in response to carbogen administration; longer-term neurological events fell within the range for expected events in this patient population.
Conclusions—Carbogen-induced CVR elicited no added adverse events and provided a surrogate marker of cerebrovascular reserve consistent with intracranial vasculopathy.
Recent studies have shown high 2-year ischemic stroke rates of ≈20% in patients with symptomatic intracranial stenosis.1 Despite this high stroke rate, risks related to endovascular revascularization, particularly in light of recent improvements in medical management, have led to recent halting of 2 major intracranial stent trials for symptomatic intracranial stenosis.2,3 Waning interest in intracranial stenting may ignore a subset of high-risk patients who have true hemodynamic instability, a finding associated with a 60.7% 3-year recurrent stroke rate in 1 prospective study.4 The critical barrier to better informing treatment decisions rests with a relative lack of abilities to measure the wide range of cerebral hemodynamic compromise possible and furthermore to use this information to evaluate how evolution of hemodynamic changes portends personalized stroke risk.
Reductions in cerebral perfusion pressure can be compensated for by increases in parenchymal cerebral blood volume and cerebral blood flow (CBF).5,6 Cerebrovascular reactivity (CVR), or the ability of vessels to regulate CBF and cerebral blood volume, provides an indicator of the cerebrovascular reserve and potentially how near tissue is to failing to meet hemodynamic demand. The magnitude and response time of CVR, assessed using exogenous vasodilator agents, have been demonstrated to correlate variably with stroke risk and symptomatology.7
CVR measurements induced by changes in blood oxygenation secondary to hypercarbic gas administration are increasingly used, primarily in research settings8,9 whereby elevated CO2 levels (4%–6%) are administered, most commonly with a balance of medical grade atmospheric air (21% O2/74% N2). This leads to increases in the arterial partial pressure of CO2 (Paco2), CBF, and cerebral blood volume and in turn an increase in blood oxygenation in capillaries and veins (Figure 1A). This small increase in oxyhemoglobin, relative to deoxyhemoglobin, manifests as an increase in T2*-weighted MRI signal (ie, the blood oxygenation level–dependent [BOLD] effect). Two factors have slowed BOLD CVR mapping in clinical stroke imaging. First, the safety of such hypercarbic gas mixtures is unclear in patients with ischemia, because challenging vasculature operating near reserve capacity may exacerbate symptoms or elicit new neurological events. Second, hypercarbic gas administration frequently involves unique setups that use mechanisms of end-tidal forcing or similar procedures.8,10 Although well-controlled, these systems may be impractical for routine clinical use.
Alternatively, carbogen, consisting of hypercarbia with a balance of oxygen (ie, 5% CO2/95% O2: hypercarbic hyperoxia) can also induce changes in CVR, while also increasing oxygen transport to tissue.11 However, transient hyperoxia may influence metabolism and will increase the partial pressure of arterial oxygen (Pao2) and arterial (Ya) and venous (Yv) oxygen saturation, increasing blood oxygenation unrelated to CVR. It remains unclear how carbogen-induced CVR measurements correlate with angiographic and clinical impairment in patients with intracranial stenosis. This issue is fundamental, because carbogen administration is in principle a more clinically feasible and safer challenge than hypoxia or hypercarbic normoxia.
We enrolled patients with intracranial stenosis for a 3-year window and recorded carbogen CVR measurements obtained in a clinical radiological unit using standard gas delivery equipment available at most hospitals with 4 main aims: (1) to characterize differences in hypercarbic normoxic and carbogen CVR in healthy parenchyma, (2) to assess the relationship between carbogen-induced CBF CVR and BOLD CVR in ischemic and healthy tissue, (3) to correlate carbogen CVR with standard clinical measures of disease, and (4) to provide preliminary data on intrascan and longer-term (6–12 months) complications resulting from carbogen CVR imaging.
All study components were compliant with Health Insurance Portability and Accountability Act and the Declaration of Helsinki and approved by the local Institutional Review Board. Subjects were enrolled between December 2011 and March 2014 and included healthy volunteers (n=10), patients with confirmed-or-suspected intracranial stenosis (n=92), and follow-up scans in a subgroup of patients (n=33). Patients were enrolled as part of the Vanderbilt Assessment of Multimodal MRI in Patients at-Risk for stroke with Intracranial Stenosis (VAMMPRIS) trial and provided informed, written consent for this prospective study. This study included 4 components with differing inclusion/exclusion criteria (additional details provided as online-only Data Supplement).
Hypercarbic Gas Comparison
The purpose of this study was to contrast hypercarbic normoxic and carbogen-induced BOLD CVR in healthy volunteers (Table 1; n=10).
All MRI measurements were performed at 3.0 T (Philips). CVR was assessed in response to delivery (12 L/min) of carbogen gas (5% CO2/95% O2) administered through a nonrebreathing oxygen facemask (Salter Labs, Ref: 8130). Physiological monitoring was achieved using an In Vivo Research Inc (3150 MRI) device and Millenia Vital Systems Monitoring System (3155MVS). Monitored parameters included heart rate, blood pressure, Ya, and end-tidal CO2 (EtCO2) response (ΔEtCO2; Salter Labs; Ref: 400F). The stimulus paradigm consisted of 2 blocks of 3-minute carbogen administration and hypercarbic normoxia (5% CO2/21% O2/74% N2) interleaved with 3 minutes of medical grade air (≈21% O2/≈79% N2: normocarbic normoxia). Stimulus order was randomized between subjects. Whole-brain BOLD images (echo time=35 ms) were acquired with spatial resolution of 3.5 mm isotropic.
All data were corrected for motion and baseline drift using standard affine-correction algorithms and spatially smoothed (full-width-half-maximum=3 mm).12 CVR was calculated according to relative signal changes (ΔS/S0), defined as the mean difference in the signal during the final 90 s of each stimulus block and baseline block, divided by the mean signal during the final 90 s of the baseline block. To allow for comparison between subjects, all data were coregistered to a standard T1-weighted atlas and lobar CVR was recorded (Figure I in the online-only Data Supplement). Pearson correlation and linear regression were performed for CVR measures derived from the 2 stimuli.
CBF Versus BOLD CVR
The purpose of this study was to assess the extent to which carbogen-induced BOLD CVR measures correlated with carbogen-induced CBF CVR derived from independent CBF measures obtained from arterial spin labeling (ASL).
A subgroup (Table 1; n=57) of patient scans had BOLD and ASL CVR measurements recorded in the same scan session. The BOLD protocol was as outlined above except without the hypercarbic normoxia challenge. For ASL, a pseudocontinuous sequence was used with 1.6 s labeling pulse train, followed by postlabeling delay of 1.525 s. The gas paradigm consisted of ≈5-minute room air breathing and 5-minute carbogen.
BOLD data were postprocessed as described above. Pseudocontinuous ASL data were surround-subtracted and normalized by M0 to generate CBF-weighted maps, spatially smoothed to match the BOLD postprocessing, and CBF was quantified on application of the solution of the flow-modified Bloch equation assuming a blood water T1 reduction from 1.6 to 1.4 s and bolus arrival time reduction of 5% for the transition from room air to carbogen. CBF reactivity was defined as CBF change normalized by baseline CBF (ΔCBF/CBF0).
BOLD CVR and Lateralizing Disease
The purpose of this study was to contrast carbogen-induced BOLD CVR with angiographic and clinical measures of disease severity in patients with atherosclerotic and moyamoya intracranial stenosis.
We focused on lateralizing intracranial stenotic disease. Inclusion criterion for patients with atherosclerosis was ipsilateral intracranial stenosis ≥50%. Exclusion criterion was cervical stenosis >70%. In moyamoya subjects, inclusion criterion was ≥1 hemisphere with modified Suzuki Score (mSS) >0. These criteria led to 29 moyamoya and 31 atherosclerotic patients meeting inclusion criteria. Stenosis degree of major intracranial and cervical vessels was classified by a board-certified neuroradiologist (M.K.S.; experience=13 years). In patients with clinically confirmed idiopathic moyamoya disease, mSS was calculated separately in right and left hemispheres.13 Additional stroke risk factors including smoking status, diabetes mellitus, prior infarct, body mass index, cardiovascular events, and age were recorded, along with antiplatelet and anticoagulant medications. The BOLD imaging protocol was identical as outlined in CBF Versus BOLD CVR section.
CVR was calculated in 2 different ways. First, relative signal changes, as outlined in the Hypercarbic Gas Comparison section, were calculated. Second, Z-statistics (Z-stat) were calculated. Here, the carbogen waveform was used in the FMRIB Software Library12 design matrix, and a corresponding parameter estimate image was calculated, which corresponded to how strongly that stimulus fits the data. The parameter estimate map was converted to a T-statistic image by normalizing by the SE. The T-image was then transformed into a Z-stat. Signal change and Z-stat maps were coregistered to a standard T1-weighted brain atlas. For patients with intracranial atherosclerosis, affected hemispheres were defined as hemispheres with intracranial stenosis ≥50%, and contralateral hemispheres were defined as hemispheres with no intracranial stenosis ≥50%. In patients with moyamoya disease, hemispheres were oriented by higher mSS (affected hemispheres) versus lower mSS (contralateral hemispheres). The coregistered CVR maps were oriented: radiological right=affected and radiological left=contralateral. CVR was calculated in 11 brain regions. A Student t test assessed significance between affected and contralateral hemispheres. A general linear model, including major stroke risk factors as defined by the Framingham study,14 was also applied to understand the extent to which different risk factors contributed to lateralizing CVR.
The purpose of this study was to record any immediate (during the scan) and longer-term (≈1-year follow-up) complications that arose in patients participating in this study.
All patient scans were analyzed, which included 92 separate patient scans plus 33 follow-up scans. Of these 125 total scans, immediate follow-up was assessed immediately after the patient completed the MRI scan and additional longer-term follow-up was performed after ≈1 year. For short- and long-term follow-up, a stroke research nurse contacted the patient by phone and conducted a questionnaire. Findings were interpreted by a board-certified stroke neurologist (L.C.J.). Complications were classified as either (1) new stroke-like symptoms, (2) stroke-like symptoms that began before the CVR scan, or (3) symptoms of nonstroke origin.
Results of the safety questionnaire were recorded, and long-term complication rates were contrasted with expected neurovascular event rates in this patient population.1
Hypercarbic Gas Comparison
Figure 1A shows representations of fractional changes in oxyhemoglobin relative to deoxyhemoglobin at baseline and for different stimuli. Figure 1B shows that BOLD ΔS/S0, normalized by ΔEtCO2, provides similar contrast for the 2 gas stimuli, albeit with different intensity. Figure 1C demonstrates that the 2 measures correlate tightly (P<0.05) across all brain lobes. Figure 1D demonstrates that CVR trends between brain regions are similar for the 2 gas stimuli. These data are consistent with carbogen BOLD CVR providing similar information to hypercarbic normoxic BOLD CVR, however with an additional contribution from increased oxygen saturation secondary to hyperoxia. The best fit line to these data was found to provide a slope of 2.53 (intercept=0.00).
CBF Versus BOLD CVR
Figure 2A shows baseline and CBF reactivity (upper) and BOLD reactivity (Z-stat/ΔEtCO2 and ΔS/S0/ΔEtCO2) for the same patients (Table 1; Table I in the online-only Data Supplement; n=57). In Figure 2B, the relationships between BOLD and CBF CVR for all gray matter in affected and contralateral brain hemispheres are shown. Correlations are found between both measures of CVR for all scenarios except BOLD Z-stat versus CBF CVR in the contralateral hemisphere.
BOLD CVR and Lateralizing Disease
CVR maps calculated according to signal changes (ΔS/S0), T-statistic, and Z-stat are shown in Figure 3. Figure 3A and 3B shows data for atherosclerotic (Table 1; Table II in the online-only Data Supplement; n=31) patients, and Figure 3C and 3D shows data for moyamoya (Table 1; Table III in the online-only Data Supplement; n=29) patients. These data demonstrate that in patients with atherosclerosis, CVR is consistently reduced, with the exception of the occipital lobes and cerebellum, in the hemisphere of the stenotic vessel when BOLD Z-stat are used for CVR determination. Sparing of occipital lobes and cerebellum is consistent with the predominantly anterior circulation stenosis in our patients with intracranial atherosclerosis (19% had posterior circulation stenosis and 13% had only posterior circulation involvement). Lateralizing CVR is less apparent when ΔS/S0 is considered, or when moyamoya patients are considered, consistent with the bilateral nature of moyamoya disease. A multivariate analysis was performed to understand the extent that additional stroke risk factors contributed to lateralizing CVR (Table 2). It was found that age, but not smoking or diabetes mellitus, contributed significantly to lateralizing disease in patients with atherosclerosis. In patients with moyamoya disease, it was found that mSS, but not age, smoking, or diabetes mellitus, was a significant additional predictor of lateralizing CVR. Figure 4 shows multimodal data from an (Figure 4A) atherosclerotic and (Figure 4B) moyamoya patient, demonstrating how CVR maps can complement standard clinical imaging.
Table 3 shows safety results. No patients reported stroke-like symptoms in the immediate follow-up; 2 of 125 (1.6%) experienced claustrophobia during the scan. For the longer-term follow-up (median=372 days), 5.0% of patients had stroke-like symptoms as determined by a neurologist (L.C.J.), which is within the range of expected neurovascular events for this population.1 All symptoms are documented in Table IV in the online-only Data Supplement.
The major findings from this study are that noninvasive CVR values derived from carbogen stimuli correlate (1) with hypercarbic normoxic gas stimuli in healthy tissue, (2) with CBF CVR to carbogen in affected and contralateral brain hemispheres of patients with intracranial stenosis, and (3) with angiographic measures of disease severity both in atherosclerotic and moyamoya patients after appropriate processing (Z-stat). Finally, (4) we provide preliminary information showing no immediate or short-term stroke-related complications from this protocol and longer-term complications that fall within the expected range of neurovascular events for this patient population.
These findings provide the foundation for noninvasively measuring cerebral hemodynamic compromise in patients with intracranial stenosis, which may provide a marker for stroke risk. Following the Stenting and Aggressive Medical Management for Preventing Recurrent Stroke in Intracranial Stenosis (SAMMPRIS) trial,1 management for patients with intracranial stenosis remains unclear. Although the risks of endovascular stenting exceeded the risks for aggressive medical management in this trial, other trials suggest that a subset of subjects with true hemodynamic instability have much high rates of stroke.4 MRI measurements of CVR represent a promising class of screening procedures to identify hemodynamic instability. Traditional vasodilatory agents, including acetazolamide and remifentanil, have been used to elicit changes in CVR and monitored using BOLD and CBF-weighted techniques; however, these agents are not universally safe. By contrast, the carbogen stimulation increases oxygen delivery to tissue and can be applied without the same concerns related to dose restrictions. Additionally, unlike positron emission tomography and single-photon emission computed tomography, MRI has no ionizing radiation exposure.
The confound of using carbogen in conjunction with noninvasive MRI measures is that carbogen increases blood oxygenation level in a manner that is nonspecific to CVR, and also hyperoxia reduces blood water T1. A recent study found no correlation between BOLD and CBF CVR in response to carbogen in a smaller (n=16) population of healthy adults.15 Several differences between this study and ours should be noted. First, we used an ASL postlabeling delay of 1.525 s, approximately consistent with recent recommendations from the International-Society-For-Magnetic-Resonance-In-Medicine ASL study section.16 In the above study,15 a much shorter postlabeling delay of 800 ms was used, which drastically increases sensitivity to intravascular blood water relative to CBF. Second, the small sample size and focus on healthy tissue only in the above study15 limit the range of flow–volume coupling scenarios, thereby limiting the power available to find significant correlations. This is consistent with Figure 2, whereby stronger relationships between BOLD and CBF CVR are observed in affected regions, where a broader range of flow–volume coupling is present. We do agree that hypercarbic normoxia is a simpler challenge that is more straightforward to interpret, yet this method has met resistance for clinical use because of concerns related to hypercarbic administration in patients with cerebrovascular impairment.
There are several limitations. First, it is unclear whether hyperoxia from carbogen influences metabolism and perfusion. Small vasoconstrictive effects are likely present in hyperoxia; however, these effects have been measured to be small relative to the vasodilatory effect from hypercarbia.11 Additionally, although carbogen is thought to be primarily isometabolic in healthy adults, it is unclear whether ischemic tissue may upregulate cerebral metabolic rate of oxygen consumption during periods of increased oxygen delivery. Second, multiple factors contribute to stroke risk, and given the sample size of the clinical components of this study, n=57 (CBF versus BOLD CVR study), n=60 (BOLD CVR and lateralizing disease study), and n=125 (safety study), not all factors could be controlled. We chose to restrict inclusion criteria to a relatively homogeneous population of patients with asymmetrical intracranial stenosis with no cervical stenosis >70%, and we performed multivariate analysis of the major stroke risk factors only (online-only Data Supplement). We did not control for dolichoectasia, although patients with dolichoectasia and related cerebrovascular morphological abnormalities of cervical vessels may have increased stroke risk.17 Unfortunately, our patients do not routinely undergo carotid echo-color Doppler and only a subset of our patients had cerebral angiography; larger studies that assess the relationship of CVR with other stroke risk factors may benefit by incorporating separate measurements of dolichoectasia. Finally, the purpose of this study was to outline relationships between carbogen-induced BOLD CVR and more standard imaging measures including CBF, vascular stenosis, and hypercarbic normoxic CVR in a relatively controlled cohort (asymmetrical intracranial stenosis) of patients, as well as to provide preliminary safety information on this protocol. Ongoing work in which more patients are enrolled and monitored longitudinally should provide information as to whether characteristics of such contrast has prognostic use.
Results provide an exemplar for how CVR changes in response to carbogen stimulation in patients with cerebrovascular disease and could provide an avenue for using this stimulus in clinical scenarios.
Sources of Funding
This work was supported by National Institutes of Health/National Institute of Neurological Disorders and Stroke grant 5R01NS078828.
Dr Shyr is a consultant for GlaxoSmithKline, AduroBiotech Inc, and Janssen. The other authors report no conflicts.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.005975/-/DC1.
- Received April 28, 2014.
- Accepted May 20, 2014.
- © 2014 American Heart Association, Inc.
- Derdeyn CP,
- Chimowitz MI,
- Lynn MJ,
- Fiorella D,
- Turan TN,
- Janis LS,
- et al
- Zaidat OO,
- Castonguay AC,
- Fitzsimmons BF,
- Woodward BK,
- Wang Z,
- Killer-Oberpfalzer M,
- et al
- Derdeyn CP,
- Videen TO,
- Yundt KD,
- Fritsch SM,
- Carpenter DA,
- Grubb RL,
- et al
- Donahue MJ,
- Strother MK,
- Hendrikse J
- Mandell DM,
- Han JS,
- Poublanc J,
- Crawley AP,
- Stainsby JA,
- Fisher JA,
- et al
- Ashkanian M,
- Gjedde A,
- Mouridsen K,
- Vafaee M,
- Hansen KV,
- Ostergaard L,
- et al
- D’Agostino RB,
- Wolf PA,
- Belanger AJ,
- Kannel WB
- Alsop DC,
- Detre JA,
- Golay X,
- Gunther M,
- Hendrikse J,
- Hernandez-Garcia L,
- et al