Cerebrovascular Function is Altered in Hemodialysis Patients

Abstract

Key Points

• Hemodialysis patients have impaired cerebrovascular reactivity.

• Hemodialysis patients have cerebral structural deficits.

Background

Hemodialysis patients have declines in cerebral blood flow (CBF) and cerebral oxygenation during hemodialysis that may lead to ischemic brain injury. Cerebrovascular reactivity (CVR) may indicate which individuals are more susceptible to intradialytic hypoperfusion and ischemia. We hypothesized that hemodialysis patients would have decreased CVR and increased CBF relative to controls and deficits in CVR would be related to brain structural deficits.

Methods

We measured cortical thickness and white matter hyperintensity (WMH) volume from T1 and T2 fluid attenuation inversion recovery images, respectively; CVR from a breath hold blood oxygen level–dependent CVR functional magnetic resonance imaging (fMRI); and arterial transit time and CBF from arterial spin labeling. Cerebrovascular and structural deficits in gray matter and white matter (GM and WM) were tested by averaging across the tissue and with a pothole analysis. Finally, we correlated cortical thickness and WMH volume with GM and WM cerebrovascular variables to assess the relationship between brain structure and cerebrovascular health.

Results

In ten hemodialysis patients, cortical thickness was found to be decreased (P = 0.002), WMH volume increased (P = 0.004), and WM CBF increased (P = 0.02) relative to ten controls. Pothole analysis indicated a higher number of increased GM and WM CBF voxels (P = 0.03, P = 0.02) and a higher number of decreased GM and WM CVR voxels (P = 0.02, P = 0.01).

Conclusions

This pilot study demonstrates that hemodialysis patients have decreased CVR and increased CBF relative to controls, along with reduced brain integrity. Further investigation is required to fully understand whether these cerebrovascular deficits may lead to structural changes.

Introduction

Hemodialysis patients have high prevalence of cognitive impairments, vascular disease, and ischemic type brain lesions.^1–6 During the hemodialysis procedure, patients experience significant circulatory stress and have intravascular fluid removal that can result in drops in BP.^7,8 The vascular hypothesis of neurodegeneration in hemodialysis patients suggests that high rates of vascular disease predispose patients to cerebral ischemia during the drop in BP experiences during hemodialysis and leads to downstream brain injury, resulting in cognitive impairment over time.⁹

Impaired cerebrovascular reactivity (CVR) might contribute to brain pathology in hemodialysis patients. CVR is the cerebral small vessel response to changes in pCO₂, with vasodilation or vasoconstriction occurring to help maintain a steady cerebral blood flow (CBF). CVR can be assessed by measuring the vascular response to vasodilatory stimuli, such as an increase in CO₂, either with a breath hold or inspired CO₂.¹⁰ Although cerebral autoregulation is likely the key mechanism used to stabilize CBF during hemodialysis-induced drop in BP, both CVR and cerebral autoregulation rely on the same cerebrovascular capacity to dilate or constrict.¹¹ A sensitive method to measure CVR that allows for regional CVR measures uses a blood oxygen level–dependent (BOLD) fMRI acquisition. Assuming constant metabolism under hypercapnia, the BOLD signal is sensitive only to changes in CBF.¹² CVR research in hemodialysis patients is limited, but PET and transcranial Doppler ultrasound have detected deficits in CVR using CO₂ inspiration as a stimulus.^13,14

Baseline CBF is another potential marker of cerebrovascular health and can be measured using arterial spin labeling (ASL) magnetic resonance imaging (MRI).¹⁵ Patients with ESKD have increased baseline CBF relative to controls due to anemia secondary to kidney disease.^16,17 This upregulation of CBF requires vasodilation and could be a sufficient marker of exhausted vasodilatory reserve on its own. However, decreased hematocrit also changes the magnetic properties of blood, which requires correction to accurately estimate CBF, a correction that is not routinely done.¹⁸

In this pilot study, we characterized the cerebrovascular health of hemodialysis patients compared with healthy controls. We hypothesized that hemodialysis patients would have increased CBF and decreased CVR, indicating a baseline vasodilation and vasodilatory reserve deficits, relative to controls. In addition, we hypothesized that magnitude of structural and cerebrovascular deficits would be correlated in the hemodialysis patients.

Methods

Participants

We recruited patients with ESKD undergoing treatment with hemodialysis from Milwaukee, WI, area community dialysis units. We recruited the healthy controls by posting recruitment flyers on our institution's employee webpage. Race and ethnicity were self-reported by each participant. Each participant provided written informed consent to the protocol, which was approved by the Institutional Review Board at the Medical College of Wisconsin and adhered to the Declaration of Helsinki. Inclusion criteria were age 50 years or older and receiving thrice weekly conventional in-center hemodialysis. Exclusion criteria included a history of stroke, traumatic brain injury, brain tumor or surgery within the past year, non-English speaking, or diagnosis of dementia. In addition, age-matched and sex-matched controls were recruited using the same inclusion and exclusion criteria, the only exception being no evidence of kidney disease.

MRI

Participants with ESKD had MRI performed on off-hemodialysis days (Thursday for Monday–Wednesday–Friday patients and Friday for Tuesday–Thursday–Saturday patients) and healthy controls on any weekday. We used a 3T MRI (Signa Premier, GE Healthcare, Waukesha, WI) with a 32 channel Nova coil (Nova Medical, Wilmington, MA) to collect image data. No participant was given antianxiety or sedative medications for the scan, and all participants completed an MRI safety screen before the scan. T1-weighted anatomical images were acquired using a sagittal 3D T1 magnetized prepared rapid gradient echo sequence (eco time [TE]=2.8 ms, repetition time [TR]=2280 ms, inversion time [TI]=900 ms, voxel size=1×1×1 mm). T2 fluid attenuation inversion recovery (FLAIR) anatomical images were acquired using a sagittal T2 FLAIR Cube sequence (TE=105 ms, TR=6000 ms, TI=1734 ms, voxel size=0.5×0.47×0.47 mm). Multiband multiecho BOLD fMRI (TE=14, 33, 51 ms, TR=1000 ms, flip angle=60°, voxel size=3.0×3.0×3.0 mm, multiband factor=4) was collected over 5 minutes, 20 seconds starting with 66 seconds of paced breathing, followed by cycles of 24 seconds of paced breathing, 16 seconds of breath hold on expiration, and 16 seconds of self-paced recovery breathing. Respiratory band data were recorded simultaneously to confirm task compliance. Multidelay ASL was acquired using a Hadamard encoded sequence to simultaneously acquire seven post-labeling delays (PLD) (PLDs: 1, 1.4, 1.7, 2.1, 2.6, 3.1, 3.7 seconds, TR/TE/TI=7592/3.008/900 ms, voxel size=1.7×1.7×3.0 mm); an additional M0 image was acquired with a TR/TE=6780/13.1 ms for quantification of CBF.

Anatomical Processing

After inspecting T1 and T2 FLAIR-weighted images for artifacts, T1 images were bias-corrected using N4 bias correction, brain extracted using the Functional Magnetic Resonance Imaging of the Brain Software Library (FSL) Brain Extraction tool function. The Advanced Normalization Tools (ANTs) Atropos function was used for tissue segmentation, and the antsCorticalThickness pipeline was used to estimate cortical thickness.^19,20 Cortical thickness was averaged across the gray matter (GM) as a marker of GM structural health. T2 FLAIR images were segmented using Lesion Segmentation Toolbox's lesion prediction algorithm to obtain a mask of white matter hyperintensities (WMHs) in each participant to determine lesion load.²¹ WMH load was the total volume of WMH in milliliters. T1 images were registered into Montreal Neurological Institute (MNI) 152 T1 2-mm standard space using ANTs-deformation and affine registration, while T2 FLAIR images were registered into subject T1 space using affine registration.^22–24

CVR Processing

Multiecho breath hold BOLD fMRI images were processed similarly to Cohen et al. and Moia et al.^25,26 We first removed the first 10 seconds of fMRI volumes and then ran motion correction with FSL's Motion correction using FMRIB's Linear Image Registration Tool (McFLIRT) McFLIRT function on the first echo, applying the linear transforms to the other two echoes. We then averaged the motion-corrected first echo and registered the image into T1 anatomical space.²⁴ We then applied FSL's bet function to the time averaged first echo to obtain a brain mask to apply to the second and third echoes.²⁷ Frame displacement was calculated from the coregistration transfers from McFLIRT. Tedana was applied to identify noise components using independent component analysis (ICA). This strategy takes advantage of linear dependence that the BOLD signal has on TE and identifies independent component time series unrelated to echo times as noise.²⁸ Finally, a block design text file of ones during the breath hold and zeros not during the breath hold was created. Sixteen additional text files were created by lagging the original text between −8 and +8 seconds.²⁶ Each of the 17 text files representing all the possible lags for the CVR task were convolved with the canonical hemodynamic response function.²⁹ The ICA-rejected components were orthogonalized to the breath hold task, ICA accepted components, head motion, and a fourth order Legendre polynomial in accordance with the conservative denoising model in Moia et al.²⁶

We calculated a CVR from the BOLD fMRI data for each participant. First, Analysis of Functional NeuroImages 3dDeconvolve and 3dREMLfit were run for each of the lagged task files, and the best fit lag was chosen by the highest R² of all the lag file fits for each voxel.³⁰ The timing file with the greatest R² value was used to calculate CVR as the beta weight of the timing file fit, indicating the amplitude of the CBF response to the breath hold. As the beta weight was used as the CVR measure, we designated this percent change in BOLD signal (%ΔBOLD). CVR maps were registered into standard space using the CVR to T1 and T1 to standard space transform files obtained from the previously described ANTs registration. CVR maps were reviewed qualitatively. High large-vein CVR and good GM white matter (WM) contrast with higher CVR in the GM were used as the indicators of quality data. CVR was averaged across the GM and WM as markers of cerebrovascular health.

ASL Processing

We calculated the CBF (ml/100 g/min) and arterial transit time (ATT) (in seconds) from ASL image data. The ASL raw images were decoded to produce seven PLDs and one M0 magnetization image. FSL's Bayesian Inference for Arterial Spin Labeling toolkit was used with a tissue T1 of 1.2 seconds and a blood T1 using the hematocrit correction proposed by Hales et al.¹⁸ Hemodialysis patient hematocrit was measured at their previous dialysis session and controls from any laboratory within the past 6 months or if not available they were assumed to have normal hematocrit, indicating a blood T1 of 1.6 seconds. A simple calibration model was used with variational Bayes inference to calculate CBF and ATT.^31,32 Partial volume correction was additionally used to correct for the decreased perfusion and lower signal to noise ratio in the WM.³³ M0 magnetization images were used to register ASL space to T1 anatomical space. CBF and ATT images were registered into standard space by using the ANTs transform files acquired previously. CBF and ATT were both averaged in the GM and WM as whole tissue cerebrovascular markers.

Pothole Analysis

A pothole analysis was conducted to identify subject-specific abnormalities in CVR and CBF between each hemodialysis participant and controls. First, we transformed all biomarker maps into standard MNI152 T1 2-mm space and calculated the mean and standard deviation of the healthy control group image data at every individual voxel. We then calculated the Z-score for each voxel in each participant in both groups relative to the control group mean and standard deviation. To identify abnormal voxels, a threshold of |Z_HD|=2.42 for the hemodialysis group and a threshold of |Z_HC|=1.82 for the healthy controls group was used. To correct for the differences in group distributions, the Disco-Z testing paradigm was used. The paradigm corrects for the size of each group and whether a voxel came from the control or comparison group in estimating the Z threshold for the abnormality of a given biomarker in the comparison group relative to a control group.³⁴ GM masks were applied to cortical thickness images, while cerebrovascular variables were considered in the GM and WM separately. Abnormal voxels were then counted within those masks. The percent of abnormal voxels (the number of abnormal voxels divided by the total number of voxels) was used as the outcome measures for comparison statistical testing.

Statistical Testing

Whole tissue averaged cortical thickness, WMH volume, CVR, CBF, and ATT were compared between age-matched controls and hemodialysis patients. Kolmogorov–Smirnov tests were first applied to test normality, after which it was determined that all comparative statistical testing was performed using a nonparametric Wilcoxon rank-sum test. Whole GM cerebrovascular variables were then correlated with cortical thickness using Spearman correlation in hemodialysis patients. The same process was repeated for WM cerebrovascular variables and WMH volume. Significance was considered at P < 0.05.

Results

Participants

Eighteen hemodialysis patients and 11 age-matched controls consented to participate. Subsequently, eight hemodialysis patients and one control were unable to get the MRI due to MRI screening failure or claustrophobia. Ten hemodialysis patients and ten age-matched controls were included in the final analysis. In one hemodialysis participant, the CVR task compliance looked insufficient and resulting maps appeared noisy, so that the participant was eliminated from CVR statistical testing. Demographic and comorbidity differences between the two groups as well as hemodialysis-related metrics are presented in Table 1.

Table 1.

Participant demographics

Variable	Hemodialysis (n=10)	Controls (n=10)	P Value
Age, yr	64.5±8.78	60.4±7.31	0.43
BMI	29.4±4.91	28.4±6.19	0.52
Sex	5/10 males	4/10 males	1
African Americana	60%	0%	0.015
Whitea	40%	100%	0.015
Diabetes	50%	10%	0.29
Hypertensiona	100%	40%	0.015
eGFR, ml/min per 1.73 mm2	n/a	64.43±13.2	n/a
Hemodialysis duration, yr	3.65±2.82	n/a	n/a
Hemoglobin, g/dlb	10.8±1.2	13.7±1.38	1.1×10−4

BMI, body mass index; n/a, not applicable.

^aP < 0.05.

^bP < 0.01.

Structural Deficits in Hemodialysis Patients

Using the whole brain averaged measures, structural deficits were observed in the hemodialysis patients relative the controls. On the basis of T1 and T2 FLAIR-weighted images, hemodialysis patients had significantly decreased cortical thickness (P = 0.002) and increased WMH volume (P = 0.004), as depicted in Figure 1. Similar differences were found when using the pothole analysis with an increase in the number of voxels with smaller GM thickness (HC: 1.9%±1.2%; hemodialysis: 4.3%±4.1%; P = 0.045).

Figure 1.

Whole tissue cortical thickness and WMH results. Hemodialysis participants are shown in grey, and age-matched controls are shown in white. Significant reductions in cortical thickness (left) and increases in WMHs (right) were observed in the hemodialysis group. The ** is in the WMH volume chart on the right and the *** is in the cortical thickness chart on the left. **P < 0.01, ***P < 0.001. CT, computed tomography; WMH, white matter hyperintensity.

Cerebrovascular Deficits in Hemodialysis Patients

When looking at the whole brain averaged difference in hemodialysis patient's versus controls, we found significantly increased WM CBF (P = 0.017), as shown in Figure 2. No group differences were found in whole brain averaged ATT or CVR in GM or WM (ATT GM: P = 0.79, ATT WM: 0.19; CVR GM: P = 0.65, CVR WM: P = 0.39). The pothole analyses did demonstrate more voxels with higher CBF in GM and WM for hemodialysis versus controls (GM: HC: 0.3±2.1% versus hemodialysis: 6.3%±7.8%; P = 0.026, WM: HC: 5.4%±1.5% versus hemodialysis: 21.9%±20.0%, P = 0.021, respectively) and more voxels with lower CVR in GM and WM (GM: HC: 1.1%±0.6% versus hemodialysis: 4.4%±5.8%; P = 0.037, WM: HC: 1.9%±1.0% versus hemodialysis: 6.2%±5.0%, P = 0.02), as presented in Table 2 and Figure 3. Distributions of potholes in MNI T1 2-mm standard space are shown in an example control and hemodialysis patient in Figure 4. Visual inspection of pothole maps noted that increased CBF voxels were distributed throughout the brain, whereas decreased CVR voxels appeared to be most common in the periventricular watershed areas between cerebral artery blood flow territories.

Figure 2.

Whole tissue CBF results: hemodialysis participants are shown in blue and age-matched controls are shown in green. In the whole brain group analysis, a significant increase in CBF was observed in WM tissue (right), but not in GM. *P < 0.05. CBF, cerebral blood flow; GM, gray matter; WM, white matter.

Table 2.

Pothole analysis results comparison between hemodialysis and controls, Z>0 indicate more voxels in hemodialysis patients than controls

Tissue	Measure	Relative to Controls	Z	P Value
GM	ATT, s	Slower	0.64	0.52
	CBFa, ml/100 g/min	Increased	2.23	0.03
	CVRa, %ΔBOLD	Decreased	2.08	0.04
WM	ATT, s	Slower	1.1	0.27
	CBFa, ml/100 g/min	Increased	2.31	0.02
	CVRa, %ΔBOLD	Decreased	2.33	0.02

ATT, arterial transit time; BOLD, blood oxygen level dependent; CBF, cerebral blood flow; CVR, cerebrovascular reactivity; GM, gray matter; WM, white matter.

^aP<0.05.

Figure 3.

Pothole results for GM and WM for CBF voxels above control threshold (left), ATT voxels below control threshold (middle), and CVR voxels below control threshold (right). Significantly higher numbers of CBF voxels above threshold were observed in the hemodialysis participants in both GM and WM, compared with controls (left). By contrast, significantly higher numbers of CVR voxels below threshold were observed in hemodialysis participants in both GM and WM, compared with controls (right). No significant differences in the number of ATT voxels above threshold were observed. *P < 0.05. ATT, arterial transit time; CVR, cerebrovascular reactivity.

Figure 4.

Graphic of pothole Z images, CBF Z-scores above threshold in an example healthy control participant (Z>1.82; upper left) and hemodialysis patient (Z>2.42; bottom left), as well as CVR potholes below threshold in the same example healthy controls (−Z<1.82; upper right) and hemodialysis patients (−Z<2.42; bottom right). Bright yellow is a larger Z-score pothole in that voxel, and dark red is a smaller Z-score pothole in that voxel.

Whole Tissue Cerebrovascular and Structural Correlation

There were no significant whole tissue groupwise correlations between cerebrovascular (CVR. CBF, and ATT) and structural variables, as presented in Table 3 (GM) and Table 4 (WM).

Table 3.

P and Rho values from the Spearman correlation between gray matter cerebrovascular variables and cortical thickness in hemodialysis patients

Gray Matter Cerebrovascular Measure	Cortical Thickness, mm
	Rho	P Value
GM ATT, s	−0.43	0.22
GM CBF, ml/100 g/min	0.27	0.45
GM CVR, ΔBOLD	−0.26	0.49

ATT, arterial transit time; BOLD, blood oxygen level dependent; CBF, cerebral blood flow; CVR, cerebrovascular reactivity; GM, gray matter.

Table 4.

P and Rho values of the Spearman correlation between white matter cerebrovascular variables white matter hyperintensity volume in hemodialysis patients

White Matter Cerebrovascular Measure	WMH Volume, ml
	Rho	P Value
WM ATT, s	0.13	0.73
WM CBF, ml/100 g/min	−0.17	0.65
WM CVR	0.38	0.32

ATT, arterial transit time; CBF, cerebral blood flow; CVR, cerebrovascular reactivity; WM, white matter; WMH, white matter hyperintensity.

Discussion

We found that hemodialysis patients had decreased brain structural integrity relative to controls with decreased cortical thickness and increased WMHs, with confirmation using our pothole analysis techniques on these same variables. With our cerebrovascular measurements, WM CBF was significantly increased in hemodialysis patients when looking at whole brain averages. By using pothole analysis, which avoids a wash-out effect when looking at whole brain averaged GM and WM, we found an increase in the number of WM and GM CBF voxels above the control group threshold and we found an increase in the number of WM and GM CVR voxels below the control group threshold. CBF voxels above threshold indicate a relative increase of CBF in hemodialysis patients, indicating potential upregulation of CBF, and CVR voxels below threshold reflect areas of impaired CVR.

Cerebral Structural Findings

Consistent with the current literature, we found the evidence of reduced cerebral structural integrity in both GM and WM. The hemodialysis patients demonstrated reduced cortical thickness when evaluated by both the whole brain averaging and by the number of voxels with reduced cortical thickness. The hemodialysis also had increased volume of WMHs compared with controls. This demonstrates that our hemodialysis cohort did have the cerebral disease that is commonly seen in hemodialysis patients, particularly hemodialysis patients with hypertension who have significant cerebral atrophy and WM disease.^1,5,35

Cerebrovascular Findings

The pothole analysis demonstrates that the regions of impaired CBF and CVR may vary between individuals. In the whole tissue analysis, only WM CBF was significantly different between groups. By contrast, the pothole analysis showed a significant number of increased GM and WM CBF voxels in the hemodialysis group and a significant number of decreased GM and WM CVR voxels in the hemodialysis group. These results indicate that the cerebral regions of increased CBF and decreased CVR were not consistent across participants. Thus, whole tissue averaging likely washed out both the increased GM CBF and the decreased GM and WM CVR. This is expected because vascular disease may not be uniform throughout the brain and is likely to vary among patients. In participants where the potholes did overlap, they qualitatively appeared to favor the cerebral watershed areas, as shown in Figure 4, consistent with the expectation that CVR would be more likely in regions where the vasculature is less robust and more susceptible to infarcts.³⁶

Our findings of reduced CVR in hemodialysis patients adds new information to the results from the few published small studies particularly on the differences in methodology used to measure CVR and potential patient characteristics that may be important. Consistent with a recent study of CVR (measured with transcranial doppler) in hemodialysis patients, we also found that hemodialysis patients had lower CVR than healthy participants.¹⁴ Impaired CVR may be most prominent in dialysis patients as a study comparing patients with CKD (eGFR of 39±13 ml/min per 1.73  m²) and healthy controls did not find a difference in CVR between the groups, although the patients with CKD had greater cerebral arterial stiffness.³⁷ It may be that patients requiring dialysis have even greater arterial stiffness leading to worse CVR. The patient population and characteristics are also important to consider. In a study that did not find a difference in CVR between dialysis patients and controls, the dialysis patients all had glomerulonephritis as cause of ESKD and no diabetes, uncontrolled hypertension, or peripheral vascular disease.³⁸ That cohort is very different from our more typical patient cohort of those with diabetes and hypertension. This may indicate that vasculopathy associated with underlying comorbidities is an important factor for impaired CVR in hemodialysis patients. The worse CVR noted in hemodialysis patients compared with patients with CKD may also indicate that in patients with vasculopathy, undergoing the hemodyanmic stress of hemodialysis may further worsen CVR. In addition, the method of measuring CVR may also contribute to the differences in results, specifically if impairment in CVR is more regional rather than widespread. In our study, we found significant difference in the percent of impaired CVR voxels between hemodialysis patients and controls but did not note differences when looking at the whole brain averaged CVR. In the previous study, they used transcranial Doppler to measure change in middle cerebral artery velocity in response to hypercapnia to measure CVR,³⁷ which looks more globally at small vessel reactivity compared with the individual voxel changes that we measured with the BOLD MRI. If impaired CVR occurs more regionally, differences may be washed out when looking at whole brain averages but are detected when looking at the number of affected individual voxels or specific regions. It is important to also consider the stimulus and the actual CVR outcome, as study in kidney transplant patient used exercise as the stimulus had different outcomes of change in velocity of middle cerebral artery and the kinetics of the change in velocity as the cerebrovascular response.³⁹ The differences in outcomes may lead to different inconsistent results if not interpreted carefully.

Although the CVR response to hypercapnia is not the same as the cerebral autoregulatory response to hypotension that occurs during hemodialysis, they both rely on the same vasodilatory reserve,¹¹ and we note that previous studies indicate that measuring CVR is well tolerated by patients, particularly when using transcranial Doppler. By contrast, measuring dynamic cerebral autoregulation using an induced a drop in systemic BP¹¹ would be less tolerable in our hemodialysis patients who commonly have vascular disease and are more susceptible to symptoms from hypotension. There are also measurements of cerebral autoregulation that rely on smaller spontaneous variations in BP, but these require more complex transformations of data and are hampered by low signal-to-noise ratio.^40,41 Our results and small previous studies on CVR measures in patients on dialysis suggests that exploring options for a clinically relevant and tolerated CVR methodology, such as transcranial Doppler, could be applied to a wider range of patients, particularly those at high risk of vascular disease, and useful in determining which hemodialysis patients have the vasodilatory capacity to respond to hypotensive episodes during hemodialysis.

Cerebrovascular and Structural Correlations

We had hypothesized that patients with reduced CVR may be at higher risk for ischemic injury and would demonstrate decreased cortical thickness and increased WM hypersensitizes. However, we observed no correlations between whole tissue averaged structural variables and cerebrovascular variables in hemodialysis patients. Cerebral injury in hemodialysis patients is likely multifactorial with uremic neurotoxic injury playing a role in addition to ischemic vascular type injury. Furthermore, damage may have occurred previously and may not be reflective of the current cross-sectional CVR that we collected in this study. Given our small sample of patients and large heterogeneity in hemodialysis patients, further study should explore these relationships longitudinally to understand their progression in hemodialysis patients.

Limitations

A significant limitation of this work was the small sample size, which limited the statistical power of group comparisons and correlations. We did find that a substantial number of hemodialysis patients who we recruited were hesitant to undergo an MRI for a research study. Second, although a breath hold is capable of inducing a drop in CO₂, we did not actually measure the change in CO₂ that was induced by the breath hold. Thus, we could not calculate a %ΔBOLD per change in CO₂ that would have made our CVR measure more accurate. The lack of a measurement of change in signal per change in CO₂ makes comparison across participants and between groups difficult. Furthermore, one study showed that PaCO₂ does not reach steady state until approximately 30 seconds into a breath hold event, meaning that the PaCO₂ during our 16-second breath hold may have still been changing and the rate of change is likely influenced by participant characteristics and disease states.⁴² Further interpretation of CVR measures is more complex in hemodialysis patients due to their underlying acidosis. Interdialytic acidosis likely leads to PaCO₂ increases, which could complicate the CVR stimulus, particularly when end-tidal CO₂ is not measured. The only known study comparing CVR in controls and hemodialysis patients found end-tidal CO₂ to be significantly increased in the hemodialysis cohort compared with controls.¹⁴ Finally, BOLD signal response during visual stimulation have been found to correlate with hematocrit measured in individuals, implying that anemic hemodialysis patients could have a lower BOLD-CVR response due to decrease in oxygen carrying capacity not decreased CVR.⁴³ Further study is required to rule out decreased BOLD signal-to-noise ratio due to anemia as a cause of measured differences. For ASL data collection, Hadamard encoded multi-PLD ASL is particularly susceptible to motion artifacts. Because all PLDs are collected simultaneously, head motion during one volume will contaminate all PLDs. Furthermore, given the documented low signal to noise ratio of ASL in the WM,⁴⁴ WM CBF results are somewhat limited in their interpretation. In addition, hematocrit was collected from standard-of-care medical records for hemodialysis patients, so the delay between hematocrit and scanning could have affected the hematocrit-based correction of the T1 of blood to calculate CBF in hemodialysis participants.¹⁸ However, we are one of few studies that corrected ASL CBF estimation for hematocrit and the only study that we know of that used multidelay ASL to correct for transit delays.

In this study, we found evidence of impaired CVR and increased CBF in hemodialysis patients compared with indicating baseline vasodilation and impaired vasodilatory reserve in the hemodialysis cohort. These findings may have important clinical ramifications as hemodialysis patients have less ability to properly modulate CBF during hemodialysis, when changes in systemic pressure require a cerebral vasculature response to avoid ischemic injury. We did not find evidence that these changes lead to the cerebral structural damage but are limited but our pilot study of 20 participants. Further research evaluating whether impaired CVR leads to increased risk of cerebral ischemia during hemodialysis, identifying patient factors that increase risk, and determining targets that may reduce this risk is needed to improve the cerebral health of our hemodialysis patients.

Disclosures

M.J. Goodman reports the following: Research Funding: Nonin Medical, Inc.; Advisory or Leadership Role: Medical College of Wisconsin Class of 2024 Student Assembly Research Representative, Medical College of Wisconsin International Medical Student Pen-Pal Club Co-President, and Medical College of Wisconsin Radiology Interest Group Co-President; and Other Interests or Relationships: ASN Kidney TREKS 2022 Scholar. T.B. Meier reports the following: Employer: Amgen (spouse); Ownership Interest: Amgen; and Advisory or Leadership Role: Clinical and Scientific Advisory Board member (self) and Quadrant Biosciences Inc. B.D. Schmit reports the following: Patents or Royalties: Case Western Reserve University. Y. Wang reports the following: Research Funding: Ongoing research funded by GE Healthcare. All remaining authors have nothing to disclose.

Funding

D.F. Wolfgram: National Institute of Diabetes and Digestive and Kidney Diseases (5K23DK113119, R03DK132441). W.T. Richerson: National Center for Advancing Translational Sciences (TL1TR001437). Daniel M Soref Charitable Trust (n/a).

Author Contributions

Conceptualization: Wesley T. Richerson, Brian D. Schmit, Dawn F. Wolfgram.

Formal analysis: Wesley T. Richerson.

Funding acquisition: Wesley T. Richerson, Dawn F. Wolfgram.

Investigation: Max J. Goodman, Wesley T. Richerson, Brian D. Schmit, Dawn F. Wolfgram.

Methodology: Alexander D. Cohen, Timothy B. Meier, Wesley T. Richerson, Brian D. Schmit, Yang Wang.

Project administration: Max J. Goodman.

Resources: Timothy B. Meier, Brian D. Schmit, Yang Wang.

Software: Timothy B. Meier, Brian D. Schmit, Yang Wang.

Supervision: Alexander D. Cohen, Timothy B. Meier, Brian D. Schmit, Yang Wang, Dawn F. Wolfgram.

Validation: Wesley T. Richerson.

Visualization: Wesley T. Richerson.

Writing – original draft: Wesley T. Richerson, Brian D. Schmit, Dawn F. Wolfgram.

Writing – review & editing: Alexander D. Cohen, Max J. Goodman, Timothy B. Meier, Wesley T. Richerson, Yang Wang, Dawn F. Wolfgram.

Data Sharing Statement

Anonymized data created for the study are or will be available in a persistent repository upon publication. Aggregated Data; Encoded Data. NIDDK Repository. Not yet submitted.