Cerebral Oxygen Metabolic Stress in Children and Adults With Large Vessel Vasculopathy Due to Sickle Cell Disease

Abstract

Background and Objectives

Large vessel vasculopathy (LVV), or moyamoya syndrome, increases the risk of stroke in patients with sickle cell disease (SCD), yet effective treatments are lacking. In atherosclerotic carotid disease, previous studies demonstrated elevated oxygen extraction fraction (OEF) as a predictor of ipsilateral stroke. In a SCD cohort, we examined hemispheric hemodynamic and oxygen metabolic dysfunction as tissue-based biomarkers of cerebral ischemic risk in patients with LVV.

Methods

Children and adults with SCD were recruited from a SCD clinic associated with a tertiary medical center and underwent prospective brain MRI and MR angiography. LVV was defined as ≥75% stenosis in a major anterior circulation artery, excluding occlusion or previous revascularization surgery. Baseline characteristics, cerebral blood flow (CBF), normalized OEF (nOEF), infarct volume, white matter microstructure, and brain volume were compared in hemispheres with vs without LVV. In a cross-sectional analysis, mixed-effects linear multivariable models examined the effect of LVV on: (1) CBF and nOEF, as tissue markers of hemodynamic and oxygen metabolic stress, respectively, and (2) endpoints of cerebral ischemic injury including infarct volume, white matter microstructure, and brain volume.

Results

Of 155 patients (22 [12–31] years, 57% female), 33 (21%) had ≥25% stenosis, 22 (14%) had ≥50% stenosis, 14 (9%) had 75%–99% stenosis, and 5 (3%) had 100% occlusion. After excluding hemispheres with previous revascularization surgery, LVV was present in 16 hemispheres from 11 patients. Hemispheres with (N = 16) vs without (N = 283) LVV had lower CBF (25.2 vs 32.1 mL/100 g/min, p = 0.01) and higher nOEF (0.99 vs 0.95, p = 0.02). On multivariable analysis, CBF was nonsignificantly lower (β = −0.16, p = 0.07) while nOEF remained higher in hemispheres with LVV (β = 0.04, p = 0.03). Moreover, LVV was associated with greater hemispheric infarct volume, microstructural disruption, and atrophy.

Discussion

Beyond greater infarct burden, LVV was associated with hemispheric atrophy and white matter microstructural injury. As an indicator of active hypoxia, elevated nOEF likely represents a compensatory response to flow-limiting stenosis in hemispheres with LVV. The study is limited by a small number of patients with severe stenosis. Future studies are needed to evaluate the potential of tissue-based CBF and nOEF in assessing stroke risk and guide timely treatment of vasculopathy in SCD.

Introduction

Patients with sickle cell disease (SCD) are at risk of both overt strokes and silent cerebral infarcts (SCIs).¹ Patients with strokes have more cognitive disability, resulting in challenges with daily activities, academic performance, and employment.^2-4 Cerebral large vessel vasculopathy (LVV), such as moyamoya syndrome, substantially increases the risk of stroke in patients with SCD.^5-7 Patients with progressive arterial narrowing on MR angiography (MRA) were more likely to experience additional overt or silent strokes despite receiving chronic blood transfusion therapy.⁸

We and others have previously shown that cerebral blood flow (CBF) and oxygen extraction fraction (OEF) are globally elevated to compensate for decreased arterial oxygen content and anemia in children and young adults with SCD.^9-11 Elevated OEF, an indicator of oxygen metabolic stress, has been associated with cerebral infarcts and microstructural injury in normal-appearing white matter (NAWM).¹² While studies are limited in SCD, LVV may contribute to hemodynamic compromise, limiting the compensatory increase in CBF and resulting in additional oxygen metabolic stress. In non-SCD patients with atherosclerotic carotid disease^13,14 and moyamoya disease,¹⁵ vasculopathy was associated with increased OEF in territories distal to the steno-occlusive artery. Furthermore, increased hemispheric OEF predicted ipsilateral stroke risk in patients with symptomatic carotid disease and was used to select patients for inclusion into a clinical trial testing the benefit of revascularization surgery.^14,16

In this prospective, observational imaging study, we examined tissue-based CBF and OEF as a function of LVV severity, defined by the degree of stenosis. First, we hypothesized that hemispheres with flow-limiting LVV independently predicted lower regional CBF and higher regional OEF. Second, we hypothesized that LVV independently predicted greater infarct burden, microstructural disruption, and atrophy. If true, future longitudinal studies may focus on regional CBF and OEF as predictors of ischemic brain injury in the setting of SCD and LVV, aiding in clinical decision making.

Methods

Standard Protocol Approvals, Registrations, and Patient Consents

The Institutional Review Board at Washington University in St. Louis approved this study, and written informed consent was obtained from adult participants and parents or guardians for minor participants.

Participants

Children and young adults with SCD were recruited from an outpatient SCD clinic associated with a tertiary medical center and prospectively enrolled in a longitudinal MRI study. Patients underwent MRI and time-of-flight (TOF) MRA at baseline and returned for follow-up imaging every 3 years after the baseline scan. For participants who did not return for follow-up research MRA, clinical TOF MRA performed as a part of medical care was used. Inclusion criteria were participants aged 5 years and older with HbSS, HbS beta⁰thal, HbSC, or HbS beta⁺thal who were able to tolerate unsedated MRI. Exclusion criteria were acute illness within 4 weeks before MRI, contraindication to MRI, pregnancy, history of stem cell transplantation, history of structural CNS abnormalities not caused by stroke, and illicit drug use or alcohol abuse. Hemispheres that underwent revascularization surgery before baseline or follow-up imaging were also excluded. Stroke history was classified as follows: (1) overt strokes, defined as a clinical history of ischemic or hemorrhagic stroke associated with clinical symptoms, or (2) SCIs, defined as nonovert strokes with cerebral lesions ≥3 mm in diameter on a 2D axial plane of fluid-attenuated inversion recovery (FLAIR) images.¹⁷

Laboratory test results, such as venous hemoglobin, venous co-oximetry, and capillary gel electrophoresis, and vitals were obtained before each scan visit. CaO₂ was calculated from 1.36 × hemoglobin (g/dL) × SpO₂. Hemoglobin used in CaO₂ calculation was adjusted according to the following: Hb = (total measured Hb) − (total Hb × % carboxyhemoglobin) − (total Hb × % methemoglobin). Carboxyhemoglobin and methemoglobin do not participate in oxygen exchange and were subtracted from total hemoglobin.¹⁸

Vasculopathy Definition

A board-certified neuroradiologist (M.R.), blinded to baseline characteristics and the participant cohort, reviewed TOF MRA of the head for arterial stenosis. A stenosis was defined as narrowing or focal signal loss but with normal signal intensity distally. An occlusion was defined as signal loss from the distal portion of an artery. Based on previous literature,⁸ the presence of vasculopathy involving the intracranial carotid artery and the first segments of anterior, middle, and posterior cerebral arteries, and the basilar artery were categorized into quartiles: 0%–24% (no stenosis), 25%–49% stenosis, 50%–74% stenosis, 75%–99% stenosis, or occlusion. In instances of a hemisphere with multifocal stenosis, the most proximal artery with the most severe stenosis was marked as the primary stenosis. For this study, we defined LVV as any presence of 75%–99% stenosis, excluding complete occlusion. Our rationale was supported by randomized controlled trials focusing on secondary stroke prevention in non-SCD patients with intracranial atherosclerotic stenosis.^19,20 These trials excluded patients with complete arterial occlusions before enrollment and found that the rate of stroke was higher in patients with ≥70% stenosis than in those with 50 to <70% stenosis.²⁰ Vasculopathy progression was defined as any increase in stenosis from one category to a higher category in any vessel segment at 3-year follow-up.

Imaging Protocol and Processing

MRI Acquisition, Segmentation, and Coregistration

All participants underwent research brain MRI and TOF MRA of the head on a Siemens 3 Tesla MR system. Standard 3D magnetization-prepared rapid gradient-echo (MPRAGE) T1-weighted (echo time [TE]/repetition time [TR] = 2.95/1,800 milliseconds, inversion time [TI] = 1,000 milliseconds, flip angle = 8°, resolution = 1.0 × 1.0 × 1.0 mm) and 2D T2-weighted FLAIR (TE/TR = 93/9,000 milliseconds, TI = 2,500 milliseconds, resolution = 1.0 × 0.9 × 5.0 mm for children, 1.0 × 0.9 × 3.0 mm for adults) sequences were acquired. 3D TOF MRA parameters were TE/TR = 3.59/21.0 milliseconds, TD = 0 millisecond, flip angle = 18°, and resolution = 0.6 × 0.6 × 0.7 mm.

MPRAGE T1 images were skull-stripped and segmented into gray matter and white matter with Statistical Parametric Mapping version 12 (Wellcome Institute of Neurology, London, United Kingdom).²¹ Coregistration aligned all images within a scan using the FMRIB Linear Image Registration Tool, with accuracy confirmed by visual inspection. White and gray matter volumes were normalized to total intracranial volume to adjust for head size. The total intracranial volume was defined as the sum of gray matter, white matter, and CSF volumes.

A board-certified neuroradiologist (M.R.), blinded to the participant cohort and baseline characteristics, reviewed all MRI scans. A SCI was defined as a lesion ≥3 mm in diameter on FLAIR images.¹¹ For each participant with SCIs identified by the neuroradiologist, lesions were manually delineated on the FLAIR map by a board-certified vascular neurologist (A.F.) who was blinded to the participant cohort and baseline characteristics. Voxels with lesions were excluded from the segmented white matter mask to generate a NAWM region. This allowed for regional measurements of CBF, OEF, and diffusion tensor imaging (DTI) metrics within NAWM, unaffected by infarcts.

CBF and OEF Imaging Parameters

Pseudocontinuous arterial spin labeling (pCASL, TE/TR = 12/3,780 milliseconds, resolution = 3.0 × 3.0 × 5.0 mm, labeling duration = 1,500 milliseconds, postlabeling delay = 1,500 milliseconds for adults) with 2D echo planar imaging readouts was used to measure CBF.²² For children, there was a protocol change for the postlabeling delay from 1,000 to 1,500 milliseconds, with a large subset having obtained both 1,000 and 1,500 milliseconds. We performed a linear regression analysis using 53 scans that used both postlabeling delays (left hemisphere β = 0.84, SE = 0.057, p < 0.001, R² = 0.81; right hemisphere β = 0.83, SE = 0.066, p < 0.001, R² = 0.75) and imputed CBF for the 47 scans with a postlabel delay of 1,000 milliseconds using the equations left hemisphere: CBF_{1,500 msec} = CBF_{1,000 msec} × 0.84 + 3.83 and right hemisphere CBF_{1,500 msec} = CBF_{1,000 msec} × 0.83 + 2.99. To quantify CBF, blood T1 was individually measured per participant to improve CBF reproducibility because T1 varies with age and hematocrit.²³ Blood T1 was measured in the superior sagittal sinus and estimated using a 4-parameter model as previously described.²⁴ Asymmetric spin echo (ASE), a sequence measuring tissue deoxyhemoglobin (TE/TR = 64.0/4,400 milliseconds, field of view = 220 mm, resolution = 1.7 × 1.7 × 3.0 mm, acquisition time = 7:16 minutes), was used to quantify OEF.^25,26 Partial volume correction was performed for both pCASL and ASE.²⁷ CBF and OEF maps were transformed to Montreal Neurological Institute (MNI) space for hemispheric analysis.

Diffusion Tensor Imaging Parameters and Processing

DTI scans (TE/TR = 89/10,100 milliseconds, resolution = 2.0 × 2.0 × 2.0 mm, 25 directions, b = 0–1,400 s/mm²) were processed using the FMRIB Software Library.²⁸ In brief, for each scan, diffusion images were corrected for eddy current distortion and head motion using the b = 0 image as a reference. A binary brain mask in DTI space was then calculated using FSL's Brain Extraction Tool on the same b = 0 image. Dtifit²⁹ was then applied to fit the diffusion tensor model, generating fractional anisotropy (FA) and mean diffusivity (MD) at each voxel. DTI maps were transformed to MNI space for hemispheric analysis.

Hemispheric Analysis

All LVVs in this study were in the anterior circulation (internal carotid artery, anterior cerebral artery, middle cerebral artery), and thus, we restricted our CBF and OEF quantification to NAWM supplied by the anterior circulation, to minimize contributions from the posterior circulation. This mask of the anterior circulation territory (eFigure 1A), subsequently referred to as “hemispheric,” was derived from a publicly available 3D brain arterial territory atlas in MNI space created from diffusion-weighted MRI scans of 1,298 patients with acute and subacute ischemic stroke territory.³⁰ For each hemisphere, hemispheric OEF of the anterior circulation was divided by whole-brain OEF to create normalized hemispheric OEF (nOEF), accounting for intersubject variation and permitting better quantification of hemometabolic changes within hemispheres with vs without LVV. The mean hemispheric cerebral metabolic rate of oxygen utilization (CMRO₂) was calculated from mean hemispheric CBF × mean hemispheric OEF × CaO₂, an equation derived from the Fick law. Similarly, we created a supratentorial hemispheric mask (eFigure 1B) to quantify infarct volume within gray and white matter, mean FA and mean MD of NAWM, normalized white matter volume, and normalized gray matter volume of individual hemispheres.

Statistical Analysis

Data are presented as median (interquartile range [IQR]). As our planned analysis comparing patients with and without LVV (defined as 75% stenosis or greater, given in Methods “Vasculopathy definition”), we compared their baseline clinical characteristics and laboratory parameters. Baseline imaging parameters were compared between hemispheres with and without LVV. The Mann-Whitney U and Fisher exact test were used for continuous and categorical variables, respectively. To account for multiple testing, significance was adjusted to maintain a familywise error rate <5% using the Benjamini-Hochberg procedure.

To evaluate whether LVV was an independent predictor of (1) hemodynamic and oxygen metabolic stress (CBF and OEF, respectively) and (2) ischemic brain injury endpoints hemispheres from baseline and 3-year follow-up MRI scans were included in a mixed-effect linear multivariable model with residual maximum likelihood estimation. Repeated measurements within participants, from contribution of bilateral hemispheres and repeated scans over time, were included as a random effect. Endpoints of ischemic brain injury included (1) infarct volume; (2) white matter microstructure, defined as FA and MD in NAWM; and (3) atrophy defined as normalized gray matter volume and normalized white matter volume against total intracranial volume. Age and hemoglobin were entered as covariates because of their known association with the dependent variables.^31-34 For prediction of DTI metrics and normalized brain volumes, infarct volume was included as an additional covariate because of the association of infarct volume with microstructural disruption and atrophy.^33,35 Only covariates with a p value <0.20 were retained in the final multivariable model. Based on residuals that were not normally distributed when analyzed as a continuous variable, infarct volume was analyzed as an ordinal outcome with 3 levels (0, 0.001–0.7, and >0.7 mL) reflecting increasing category of stroke lesion volume. A threshold of 0.7 mL was set based on 25% of the hemispheres having an infarct volume greater than 0.7 mL. Infarct volume was analyzed using multinomial generalized estimating equations with a cumulative logit link function predicting the probability of a larger infarct volume category. Unstandardized parameter estimates with 95% CIs were reported. The fit of the multivariable models was confirmed by assessing the distribution of the regression residuals. Lack of collinearity among covariates was verified with a variance inflation factor of 2 or less.³⁶

To describe the progression of vasculopathy, we examined the subgroup of participants with 2 MRA scans. We were able to use clinical brain MRA performed as a part of medical care for 20 participants who did not undergo any follow-up research MRA. Clinical MRA scans performed 2–3 years after the baseline scan were selected, in accordance with our research scan interval. In the sensitivity analysis (eTable 1), younger patients and those on blood transfusion therapy were more likely to have follow-up MRA. However, there was no difference in the proportion of patients with severe stenosis/occlusion or SCI between those with and without follow-up MRA. All statistical analyses were conducted using SAS software, version 9.4 of the SAS System for Windows (SAS Institute Inc., Cary, NC).

Data Availability

Anonymized data not published within this article will be made available by request from any qualified investigator.

Results

Characteristics of Patients and Hemispheres With MRA-Defined LVV

One hundred and fifty-five participants with SCD (HbSS, N = 123; HbS beta-thal⁰, N = 13; HbSC, N = 16; and HbS beta-thal⁺, N = 3) underwent a total of 241 MRI scans and MRA scans. Seventy-eight patients had 1 follow-up MRA scan, and 4 had 2. The median time between baseline and follow-up MRA scans was 3.0 years (IQR 2.1–3.2). Division of the cohort into the presence of LVV by patients and by hemispheres is shown in Figure 1. At baseline, 33 patients (21%) (N = 46 hemispheres) and 22 patients (14%) (N = 33 hemispheres) had arterial stenosis ≥25% and ≥50%, respectively. Nineteen patients (12%) had at least 1 hemisphere with 75%–99% arterial stenosis (N = 14) or complete occlusion (N = 5, Figure 1). Eight patients had bilateral hemispheres with ≥75% stenosis or occlusion while 11 patients had unilateral hemispheric involvement. Compared to patients who had both hemispheres with <75% stenosis, patients who had any hemisphere with ≥75% hemispheric stenosis or occlusion were similar in age (median age: 29 vs 21.5 years, p = 0.08, Table 1) and received more exchange transfusions (63% vs 17%, p < 0.001) and thus had higher HbA and lower HbS percentages. Hemoglobin was not different between the 2 groups. Additional clinical characteristics and laboratory values are summarized in Table 1.

Figure 1. Flow Diagram of the Study Cohort.

Baseline enrollment of children and young adults with SCD organized by vasculopathy status. Vasculopathy status grouped by patients (green box) and by hemispheres (blue box). LVV = large vessel vasculopathy; SCD = sickle cell disease.

Table 1.

Baseline Patient Characteristics

	Patients with <75% stenosis (N = 136) Median (IQR)	Patients with ≥75% stenosis or occlusion (N = 19) Median (IQR)	p Value
Age, y	21.5 (12–29)	29 (16–34)	0.08
Black/African American, n (%)	136 (100)	19 (100)	1.00
Female, n (%)	78 (57)	10 (53)	0.70
Hemoglobin, g/dL	8.9 (7.8–9.9)	8.5 (7.4–9.7)	0.45
Hemoglobin A, %	0 (0–18.3)	51.4 (15.8–58.2)	<0.001a
Hemoglobin S, %	74.2 (51.6–82.2)	45.2 (34.4–73.3)	0.002a
SpO2	98 (95–99)	97 (94–100)	0.90
Methemoglobin, %	2 (1.7–2.3)	2.2 (1.9–2.4)	0.19
Carboxyhemoglobin, %	2.5 (1.8–3.1)	2.7 (2.2–3.4)	0.13
Hydroxyurea, n (%)	63 (46)	9 (47)	0.95
Exchange transfusion, n (%)	23 (17)	12 (63)	<0.001a
Revascularization, n (%)	0 (0)	4 (21)	<0.001a
SCI presence, n (%)	75 (55.2)	17 (100)	<0.001a
Overt stroke, n (%)b	13 (10)	11 (58)	<0.001a
Smoker, n (%)c	13 (10)	1 (7)	0.62

Abbreviations: SCI = silent cerebral infarct; SpO₂ = pulse oximetry.

^aRaw p values are reported. Cohort differences remained significant after correcting for multiple comparisons using the Benjamini-Hochberg procedure.

^bOf the patients with <75% stenosis, 11 patients had overt ischemic stroke, 1 patient had hemorrhagic stroke, and 1 patient had both hemorrhagic and overt ischemic strokes. In patients with ≥75% stenosis or occlusion, all 11 cases of overt strokes were ischemic stroke.

^cMissing data on 15 and 4 patients without and with ≥75% stenosis or occlusion, respectively.

To better assess the effect of arterial stenosis on regional imaging physiology and endpoints of ischemic brain injury, we performed a hemispheric analysis instead of patient-based analysis. The primary stenosis of each hemisphere on baseline MRA, categorized by severity and location, is shown in eTable 2. Among the 27 hemispheres with ≥75% stenosis or occlusion, vasculopathy was evenly distributed across the left and right hemispheres (left hemisphere 63%, p = 0.18) and most frequently involved internal carotid arteries (82%, p < 0.001). Four patients, accounting for 6 of the 22 hemispheres with 75%–99% stenosis, had previous revascularization surgery. Revascularized hemispheres were excluded from the analysis because the surgery promotes direct and indirect collateralization to circumvent the high-grade stenosis. Similarly, hemispheres with 100% occlusion were excluded from the analysis because of small sample size (N = 5 hemispheres) and exhibited CBF similar to nonstenotic hemispheres (100% occlusion: 42.4 [40.2–46.7] vs 0%–74% stenosis: 32.1 [24.6–40.4] mL/100 g/min, p = 0.16), suggesting chronic collateralization and hemodynamic compensation in the setting of complete occlusion. The final cohort for our analysis of hemispheres with LVV (defined as 75%–99% stenosis, 16 hemispheres from 11 patients) and without LVV is shown in Figure 1.

Hemodynamic and Oxygen Metabolic Dysfunction in Hemispheres With LVV

To determine the impact of LVV on hemodynamic and cerebral oxygen metabolism, we compared NAWM CBF, nOEF, and CMRO₂ of hemispheres with LVV to those without LVV. An example of a patient in early 30s with HbSS and left internal carotid artery LVV is shown in Figure 2. On baseline MRI, hemispheres with LVV had lower CBF (25.2 vs 32.1 mL/100 g/min, p = 0.02), higher nOEF (0.99 vs 0.95, p = 0.02), and lower CMRO₂ (1.08 vs 1.16 mL/100 g/min, p = 0.02) compared to hemispheres without LVV (Table 2, Figure 3). A broad range of nOEF values in the LVV group (Figure 3) were observed, suggesting large individual variability in tissue hypoxia in the setting of LVV. On mixed-effect linear multivariable stepwise regression, by accounting for multiple scans per patient and adjusting for age and hemoglobin, CBF remained nonsignificantly lower in hemispheres with LVV while nOEF was elevated in hemispheres with LVV (Table 3). Finally, CMRO₂ was nonsignificantly lower in hemispheres with LVV compared to hemispheres without LVV after adjusting for age and hemoglobin.

Figure 2. Representative Maps Indicating Regional Hemodynamic and Oxygen Metabolic Stress in a Hemisphere Distal to Severe Arterial Stenosis.

Time-of-flight MRA of a participant in early 30s with SCD with a left-side severe ICA stenosis at the terminus and a lesser degree of stenosis at the first segment of the right anterior circulation artery (A). Lower white matter CBF (B), higher white matter OEF (C), and smaller white matter and gray matter volumes (D) are seen in the left hemisphere with severe internal carotid artery stenosis. Values displayed are those of the right hemisphere vs left hemisphere. CBF = cerebral blood flow; MRA = MR angiography; ICA = internal carotid artery; OEF = oxygen extraction fraction; SCD = sickle cell disease.

Table 2.

Baseline Hemodynamic, Oxygen Metabolic, and Structural Metrics of Hemispheres With and Without LVV

	Hemispheres without LVV (N = 283) Median (IQR)	Hemispheres with LVV (N = 16) Median (IQR)	p Value
Hemodynamics and oxygen metabolism of NAWM
CBF, mL/100 g/min	32.07 (24.55–40.40)	25.22 (19.46–33.58)	0.01a
Normalized OEF	0.95 (0.91–0.98)	0.99 (0.94–1.02)	0.02a
CMRO2, mL/100 g/min	1.16 (0.92–1.48)	1.08 (0.78–1.18)	0.02a
Structural and microstructural imaging endpoints
Infarct volume, mL	0.02 (0–0.49)	11.17 (3.15–21.04)	<0.001a
NAWM FA	0.40 (0.39–0.42)	0.36 (0.32–0.39)	<0.001a
NAWM MD, 10−3 mm2 s−1	0.74 (0.72–0.78)	0.78 (0.76–0.90)	<0.001a
Normalized white matter volume	0.33 (0.30–0.35)	0.30 (0.27–0.33)	0.005a
Normalized gray matter volume	0.56 (0.53–0.59)	0.50 (0.48–0.52)	<0.001a

Abbreviations: CBF = cerebral blood flow; CMRO₂ = cerebral metabolic rate of oxygen utilization; FA = fractional anisotropy; IQR = interquartile range; LVV = large vessel vasculopathy, defined as 75%–99% stenosis without revascularization surgery; MD = mean diffusivity; NAWM = normal-appearing white matter; OEF = oxygen extraction fraction.

^aRaw p values are reported. Cohort differences remained significant after correcting for multiple comparisons using the Benjamini-Hochberg procedure.

Figure 3. Hemodynamic and Oxygen Metabolic Dysfunction in Hemispheres With Large Vessel Vasculopathy.

On baseline MRI, hemispheres with LVV had lower NAWM CBF (A), higher nOEF (B), and lower CMRO₂ (C) compared with hemispheres without LVV. LVV was defined as 75%–99% stenosis without revascularization surgery. The red line denotes group median. Raw p values are shown. Cohort differences remained significant after correcting for multiple comparisons using the Benjamini-Hochberg procedure. CBF = cerebral blood flow; CMRO₂ = cerebral metabolic rate of oxygen utilization; MRA = MR angiography; NAWM = normal-appearing white matter; nOEF = normalized OEF; OEF = oxygen extraction fraction.

Table 3.

Mixed-Effect Linear Multivariable Modeling of Hemispheres With LVV After Adjusting for Covariates and Repeated Measurements

Dependent variable	No. of hemispheres	Independent variable		Covariatesa
		LVV		Age	Hemoglobin	Infarct volume, ordinal
		Estimate (95% CI)b	p Value	p Value	p Value	p Value
NAWM CBFc	383	−0.161 (−0.340 to 0.017)c	0.068	<0.001	<0.001	—
Normalized NAWM OEF	377	0.040 (0.006 to 0.074)	0.03	NS	<0.001	—
NAWM CMRO2c	365	−0.146 (−0.349 to 0.057)c	0.12	<0.001	<0.001	—
Infarct volume, ordinald	419	2.73 (1.70 to 3.75)d	0.001	<0.001	0.14	—
NAWM FA	374	−0.039 (−0.055 to −0.024)	0.001	<0.001	<0.001	0.07
NAWM MD	374	0.071 (0.047 to 0.094)	<0.001	<0.001	<0.001	0.09
Normalized white matter volume	415	−0.031 (−0.048 to −0.015)	0.003	<0.001	0.004	0.002
Normalized gray mater volume	415	−0.057 (−0.077 to −0.038)	<0.001	<0.001	NS	0.17

Abbreviations: CBF = cerebral blood flow, unit of mL/100 g/min; CMRO₂ = cerebral metabolic rate of oxygen utilization, unit of mL/100 g/min; FA = fractional anisotropy; LVV = large vessel vasculopathy, defined as 75%–99% stenosis without revascularization surgery; MD = mean diffusivity, unit of 10⁻³ mm² s⁻¹; NAWM = normal-appearing white matter; NS = variable not significant at α less than 0.20 and was removed from the final model; OEF = oxygen extraction fraction.

^aEstimates for covariates are provided in eTable 3.

^bEffect of LVV was adjusted for all covariates in the model and for repeated measurements among patients contributing 2 or more hemispheres. Data are unstandardized parameter estimates with corresponding 95% CIs to compare the hemispheres with LVV with the hemispheres without LVV as the reference.

^cDependent variable log-transformed before analysis; parameter estimates must be interpreted on the log scale.

^dInfarct volume was analyzed as an ordinal outcome: 0, 0.001–0.7, and >0.7 mL, using multinomial generalized estimating equations with a cumulative logit link function predicting the probability of a larger infarct volume category.

Structural and Microstructural Disruption in Hemispheres With LVV

To assess whether alterations in cerebral perfusion and oxygen metabolism were associated with imaging endpoints of cerebral ischemic injury, we compared infarct volume, microstructural injury, and brain volume of hemispheres with LVV vs without LVV. On baseline MRI, infarcts were present in 100% and 48% of hemispheres with and without LVV (p < 0.001), respectively. The median infarct volume was greater in hemispheres with LVV (11.17 vs 0.02, p < 0.001, Table 2). On multivariable analysis, after including imaging data from all time points and adjusting for age and hemoglobin, LVV remained associated with larger infarct volume (Table 3).

Similarly, hemispheres with LVV exhibited greater NAWM microstructural disruption. Compared to hemispheres without LVV, NAWM FA was lower and NAWM MD was higher in hemispheres with vs without LVV, respectively (Table 2). On multivariable analysis, FA remained decreased and MD remained elevated (Table 3) in hemispheres with LVV, after adjusting for age, hemoglobin, and infarct volume, suggesting microstructural disruption in hemispheres with LVV independent of infarct volume.

Finally, we examined the effect of LVV on white matter and gray matter brain volumes. On baseline MRI, hemispheres with LVV had smaller normalized white matter and gray matter volumes compared to hemispheres without LVV (Table 2). On multivariate analysis, we included age, hemoglobin, and infarct volume as potential covariates. Normalized white matter volume and normalized gray matter volume remained smaller in hemispheres with LVV (Table 3), suggesting that LVV status is independently associated with smaller brain volumes even after adjusting for the effects of previous infarction.

Vasculopathy Progression

To assess the stability of vasculopathy over time in SCD, those patients who had longitudinal MRA scans were examined for progression. A subgroup of 82 patients (53% of the study cohort) had follow-up MRA scans (median interval 3.0 years [IQR 2.1–3.2]). Over a 3-year period, most patients remained free from vasculopathy or without progression. No stenosis had improved. 20% (16 of 82 patients) had evidence of progressive stenosis, including 9 patients without stenosis on baseline MRA. Four of the 6 patients with 75%–99% stenosis at baseline developed more extensive stenosis with moyamoya features but did not demonstrate complete occlusion on follow-up. Progression of vasculopathy by hemisphere is shown in Figure 4. Compared to those without progression, patients with progression were more likely to have 75%–99% stenosis and receive chronic transfusion therapy and less likely to receive hydroxyurea at baseline (eTable 4), suggesting patients with more severe disease at study onset and requiring chronic transfusion therapy were at risk of LVV progression.

Figure 4. Longitudinal Assessment of Hemispheric Arterial Stenosis in a Subgroup of 82 Patients With Follow-Up MRA.

Of 164 hemispheres, 21 hemispheres had worsening arterial stenosis on follow-up imaging. Seventeen hemispheres had stenosis that progressed at least 1 quartile in severity. *Four hemispheres with 75%–99% stenosis on baseline imaging developed more extensive stenosis with moyamoya features but without complete occlusion. No hemispheres had improved stenosis. White represents improvement, light gray represents no change, and dark gray represents worsening in arterial stenosis. MRA = MR angiography.

Discussion

LVV, or moyamoya syndrome, in SCD is associated with a 3-fold increase in the risk of stroke while its progression increases stroke risk by 12-fold.^6,8,37 Our study provides a putative mechanism for this increased risk by demonstrating regional hemodynamic and oxygen metabolic dysfunction in hemispheres with LVV in children and young adults with SCD. Flow-limiting vasculopathy ≥75% was associated with reduced hemispheric CBF and elevated hemispheric nOEF compared to unaffected hemispheres. Tissue-based OEF, an indicator of active tissue hypoxia, may aid in identifying hemispheres with insufficient hemodynamic compensation and those that are highly susceptible to ischemic injury.

Beyond infarction, hemispheres with LVV demonstrated white matter microstructural disruption and both gray and white matter atrophy even after controlling for infarct volume. These findings suggest that chronic blood flow restriction may restrict brain development in children and accelerate brain tissue loss in adults, beyond the expected effect of Wallerian degeneration from previous infarction.⁷ While the impact of vasculopathy on white matter microstructure in SCD has not been previously studied, patients with atherosclerotic carotid disease showed improvements in both microstructural disruption and cognitive function after undergoing carotid revascularization.^38,39 This suggests that timely correction of flow-limiting vasculopathy could potentially mitigate microstructural disruption and associated cognitive dysfunction.⁴⁰ Future studies are warranted to investigate the impact of vasculopathy on cognitive performance in SCD.

Treatment options for vasculopathy are limited. Chronic exchange transfusion therapy, although effective in primary and secondary stroke prevention in patients with elevated transcranial Doppler velocities, does not reverse arterial narrowing.^41,42 Vasculopathy can progress and pose a stroke risk despite regular transfusions.⁸ In our data, we observed higher nOEF in hemispheres with LVV even before adjusting for hemoglobin, despite more patients with LVV receiving regular blood transfusions. Although chronic exchange transfusion has been shown to reduce OEF in SCD,³⁴ this effect may be less apparent in patients with severe vasculopathy because blood flow is restricted by stenosis and patients may experience persistent hypoxia. Similarly, while hematopoietic stem cell transplant is curative of SCD, high-grade arterial stenosis persists after curative stem cell transplant.⁴³

Revascularization surgery, providing alternative collateral blood flow, may be an effective treatment option, especially if patients continue to experience stroke or cognitive manifestations of ischemic brain injury, despite adherence to disease-modifying therapy.^44,45 Clinical trials are needed, however, especially in the adult SCD population, to address patient selection, surgical approach, and long-term outcomes regarding infarct recurrence and cognitive function after revascularization.^44-46 Our data (eFigure 2) offer potential preliminary mechanistic evidence behind the effectiveness of revascularization surgery in SCD through improvement of blood flow and alleviation of oxygen metabolic stress. Although limited by sample size within the 75%–99% stenosis group, most hemispheres with previous revascularization (blue dots in eFigure 2) had CBF above the cohort mean and nOEF below the cohort mean.

In a research setting, tissue-based imaging biomarkers, such as regional CBF and nOEF, may help assess stroke risk from vasculopathy and aid in patient selection for a larger clinical trial on revascularization surgery. While studies are limited in SCD literature, elevated ipsilateral OEF predicted future strokes in patients with symptomatic carotid artery occlusion, which was used as an imaging selection criterion in a clinical trial testing surgical bypass.^14,16 Currently, MRI-based OEF sequences vary and are limited to a few academic centers.^9,47 We have examined the correlation between different MRI-based OEF sequences within our institution⁴⁷ and are studying the intercenter correlation of ASE-OEF on various MRI platforms to improve generalizability.

In a clinical setting, our results suggest that routine MRA screening for vasculopathy may be warranted because 20% of participants developed new or progressive arterial stenosis over 3 years, similar to previous literature.⁸ While the current guideline already recommends at least a 1-time MRI screening for SCI, it recommends MRA only for children on chronic transfusions wishing to discontinue therapy, potentially underdiagnosing flow-limiting stenosis.⁴⁶ In addition, there is no consensus on the definition of SCD-related vasculopathy or the best detection tool (e.g., MRA vs transcranial Doppler), leading to prevalence estimates ranging from 10% to 60%.^5,48,49 Timely diagnosis and management of vasculopathy may mitigate risk of stroke, white matter microstructural impairment, and brain atrophy.³⁹

Our study has several strengths. The study is unique and strengthened by conducting a hemispheric analysis that focused on tissue-based CBF and OEF measurements within the vascular territory supplied by the anterior circulation. Because LVV is mostly confined to the anterior circulation in SCD, this study design enhanced the examination of LVV's impact on hemodynamic and oxygen metabolic stress while minimizing the contribution from the posterior circulation. The hemispheric analysis also permitted examination of imaging endpoints distal to the intracranial stenosis. Our study has several limitations. First, we cannot examine causality between vasculopathy, CBF, OEF, and ischemic brain injury in this cross-sectional analysis. We are performing longitudinal analysis to determine whether vasculopathy, with OEF as a mediator, predicts ischemic brain injury on follow-up MRI. Second, given only half of the study population received follow-up vessel imaging, our vasculopathy progression results have inherent biases and should be interpreted with caution. Third, to expand our subgroup of participants with longitudinal vessel imaging when evaluating vasculopathy progression, we used MRA scans performed for clinical care, which may have undergone the scanning protocol changes. Finally, the interpretation of LVV's effect on CBF is limited by our small sample size and single postlabeling delay pCASL methodology. pCASL may underestimate CBF in participants with moyamoya disease because of their extended arterial transit time through collateral vessels.⁵⁰ In our larger study, we used the same postlabeling delay time for patients with SCD and healthy controls, who have a longer arterial transit time compared to those with SCD. The potential effect of prolonged transit time superimposed on a baseline shortened transit time in SCD with moyamoya features may negatively affect our CBF accuracy. We did not see a difference in CBF between hemispheres of 75%–99% stenosis with and without moyamoya features (data not presented). In the future, we are transitioning to a multidelay ASL protocol to measure CBF in participants with and without SCD.

In a large cohort of children and young adults with SCD, we found increased oxygen metabolic stress, defined as elevated nOEF, in hemispheres with 75%–99% stenosis of a major intracranial artery. In addition, LVV was associated with greater hemispheric microstructural disruption and cerebral atrophy, independent of infarct volume. These findings suggest that tissue-based imaging biomarkers such as regional CBF and nOEF may aid in the assessment of stroke risk, as a guide to clinical decision making.

Glossary

ASE

asymmetric spin echo

CBF

cerebral blood flow

CMRO₂

cerebral metabolic rate of oxygen utilization

DTI

diffusion tensor imaging

FA

fractional anisotropy

FLAIR

fluid-attenuated inversion recovery

IQR

interquartile range

LVV

large vessel vasculopathy

MD

mean diffusivity

MNI

Montreal Neurological Institute

MPRAGE

magnetization-prepared rapid gradient-echo

MRA

MR angiography

NAWM

normal-appearing white matter

nOEF

normalized OEF

OEF

oxygen extraction fraction

pCASL

pseudocontinuous arterial spin labeling

SCD

sickle cell disease

SCI

silent cerebral infarct

TE

echo time

TI

inversion time

TOF

time-of-flight

TR

repetition time

Appendix. Authors

Name	Location	Contribution
Yan Wang, MD, MSCI	Department of Neurology, Washington University School of Medicine, St. Louis, MO	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Slim Fellah, PhD	Department of Neurology, Washington University School of Medicine, St. Louis, MO	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Martin Reis, MD	Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data; analysis or interpretation of data
Kristin P. Guilliams, MD	Department of Neurology, Mallinckrodt Institute of Radiology, and Division of Pediatrics, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data; study concept or design; analysis or interpretation of data
Melanie E. Fields, MD, MSCI	Department of Neurology, and Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data; analysis or interpretation of data
Karen Steger-May, MA	Center for Biostatistics and Data Science, Washington University School of Medicine, St. Louis, MO	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Amy E. Mirro, BS, MEng	Division of Pediatrics, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data
Josiah B. Lewis, PhD	Department of Neurology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data
Chunwei Ying, PhD	Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data; analysis or interpretation of data
Rachel A. Cohen, BS	Washington University in St. Louis, St. Louis, MO	Major role in the acquisition of data
Monica L. Hulbert, MD	Division of Pediatrics, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data
Allison A. King, MD, MPH, PhD	Division of Hematology/Oncology, Department of Medicine, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data
Yasheng Chen, DSc	Department of Neurology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data
Jin-Moo Lee, MD, PhD	Department of Neurology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data; analysis or interpretation of data
Hongyu An, DSc	Department of Neurology, and Mallinckrodt Institute of Radiology, Washington University School of Medicine, St. Louis, MO	Major role in the acquisition of data; analysis or interpretation of data
Andria L. Ford, MD, MSCI	Division of Hematology/Oncology, Department of Medicine, Washington University School of Medicine, St. Louis, MO	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data

Study Funding

This research was supported by the National Center for Advancing Translational Sciences (5KL2TR002346 [Y.W.]), the National Institute of Neurological Disorders and Stroke (K23NS099472 [K.P.G.], RF1NS116565 [H.A., A.L.F.]), and the National Heart, Lung, and Blood Institute (K23HL136904 and R01HL157188 [M.E.F.], R01HL129241 [H.A., A.L.F.]).

Disclosure

M.E. Fields reports equity ownership in Proclara Biosciences and compensation from Global Blood Therapeutics and Pfizer Inc. for consultant services. M.L. Hulbert reports compensation from Pfizer Inc., Bluebird bio, and Novo Nordisk for consulting services, and spouse employment at Pfizer. H. An reports compensation from Pfizer Inc. for consultant services. Y. Wang, S. Fellah, M.N. Reis, K.P. Guilliams, K. Steger-May, A.E. Marro, J.B. Lewis, C. Ying, R. Cohen, A.A. King, Y. Chen, J.-M. Lee, and A.L. Ford report no disclosures relevant to the manuscript. Go to Neurology.org/N for full disclosures.