Wavelet decomposition analysis is a clinically relevant strategy to evaluate cerebrovascular buffering of blood pressure after spinal cord injury

Abstract

The capacity of the cerebrovasculature to buffer changes in blood pressure (BP) is crucial to prevent stroke, the incidence of which is three- to fourfold elevated after spinal cord injury (SCI). Disruption of descending sympathetic pathways within the spinal cord due to cervical SCI may result in impaired cerebrovascular buffering. Only linear analyses of cerebrovascular buffering of BP, such as transfer function, have been used in SCI research. This approach does not account for inherent nonlinearity and nonstationarity components of cerebrovascular regulation, often depends on perturbations of BP to increase the statistical power, and does not account for the influence of arterial CO₂ tension. Here, we used a nonlinear and nonstationary analysis approach termed wavelet decomposition analysis (WDA), which recently identified novel sympathetic influences on cerebrovascular buffering of BP occurring in the ultra-low-frequency range (ULF; 0.02–0.03Hz). WDA does not require BP perturbations and can account for influences of CO₂ tension. Supine resting beat-by-beat BP (Finometer), middle cerebral artery blood velocity (transcranial Doppler), and end-tidal CO₂ tension were recorded in cervical SCI (n = 14) and uninjured (n = 16) individuals. WDA revealed that cerebral blood flow more closely follows changes in BP in the ULF range (P = 0.0021, Cohen’s d = 0.89), which may be interpreted as an impairment in cerebrovascular buffering of BP. This persisted after accounting for CO₂. Transfer function metrics were not different in the ULF range, but phase was reduced at 0.07–0.2 Hz (P = 0.03, Cohen’s d = 0.31). Sympathetically mediated cerebrovascular buffering of BP is impaired after SCI, and WDA is a powerful strategy for evaluating cerebrovascular buffering in clinical populations.

INTRODUCTION

The capacity to maintain stable cerebral blood flow (CBF) in the face of changing blood pressure (BP; i.e., cerebrovascular buffering¹) is a crucial mechanism evolving with the teleological purpose of preventing cerebral hyper- and hypoperfusion. Appropriate CBF is crucial, as when the brain is overperfused there is risk of cerebral hemorrhage (43), posterior reversible encephalopathy syndrome (35), and deleterious cerebrovascular remodeling (28). On the opposite end of the spectrum, when the brain is underperfused, one risks ischemia, cognitive decline (5, 17, 27), and loss of consciousness (20, 22). The mechanism(s) responsible for cerebrovascular buffering of BP remain elusive but certainly require sophisticated and redundant control from the sympathetic nervous system as well as local myogenic regulation, in addition to the intrinsic compliant nature of the cerebrovasculature (46) and the overarching control of arterial CO₂ tension (1, 45).

Cerebrovascular health and function are impaired after spinal cord injury (SCI), which likely underlies the three- to fourfold increased risk/odds of stroke in the population (4, 48) as well as widespread orthostatic intolerance (22) and cognitive decline (27). Several factors may contribute to the reduced cerebrovascular health and function after SCI, which include high rates of diabetes, reduced ability to physical exercise, and repetitive cerebral hypoperfusion due to baroreflex dysfunction (3, 6, 20, 21, 23). Current strategies used to evaluate cerebrovascular buffering are not appropriate in many clinical populations, such as those with SCI, as they require the induction of large perturbations in BP (through either gravitational challenges or pharmacological interventions) in individuals who already suffer from life-threatening instability in BP (7, 40) and do not account for the potentially differing influences of arterial CO₂ tension on the cerebrovasculature after SCI (47). Furthermore, the previously applied approaches assume a linear relationship between BP and CBF using transfer function analysis (9) and/or that BP and CBF behave in a stationary manner and can be analyzed using nonvarying shaped ridge functions (39). These approaches are likely suboptimal. For example, it is known that the cerebrovascular buffering of BP is complex, with multiple regulatory mechanisms, and that these mechanisms behave in a nonlinear and nonstationary manner (17, 30, 41, 42) and are influenced by arterial CO₂ tension (1).

Recently, a new approach to evaluate cerebrovascular buffering has emerged based on wavelet decomposition analysis (WDA) of spontaneous BP and CBF fluctuations (30). This approach is designed to examine cerebrovascular buffering while identifying the nonlinear and nonstationary relationships between BP and CBF (19, 31). An additional benefit of this approach is that it can account for influences from arterial CO₂ tension (1, 47). Furthermore, this approach is also arguably more clinically relevant, in that it does not involve BP perturbations for improved accuracy/statistical power but instead requires only resting steady-state hemodynamic recordings (30). It has been recently demonstrated that WDA is capable of detecting alterations in cerebrovascular buffering due to pharmacological blockade of the sympathetic nervous system in healthy individuals, when standard linear and stationary approaches were not (30). This study demonstrated that the sympathetic nervous system impacts cerebrovascular buffering of BP in the ultra-low-frequency (ULF) range of 0.02–0.03 Hz (30, 31).

To explore the hypothesis that SCI results in impaired cerebrovascular buffering, we implemented WDA on BP and CBF in individuals with cervical SCI and uninjured control subjects. As disrupted supraspinal sympathetic control of the cerebrovasculature occurs after high-level SCI (26), we reasoned that WDA would reveal impaired cerebrovascular buffering in the ultra-low 0.02- to 0.03-Hz range in individuals with SCI, and this would persist after accounting for the influence of CO₂ tension. We also hypothesized that the standard linear transfer function analysis between BP and CBF would not detect differences between SCI and uninjured groups in the ULF range of 0.02–0.03 Hz.

MATERIALS AND METHODS

All participants with SCI had sustained a cervical SCI [all participants were American Spinal Injury Association Impairment Scale (AIS) A or B, with the exception of three AIS C participants and one AIS D participants]. All injuries took place at least 1 yr before the testing (range: 1.6–44 yr since injury; see Table 1). Participants with SCI (n = 14, 5 women and 9 men) and uninjured individuals (n = 16, 7 women and 9 men) were age matched and were instructed to abstain from exercise and alcohol for 24 h before testing and caffeine on the day of testing and to only have a small meal (e.g., a small yogurt) ∼1 h before the testing. Nine participants with SCI were drug free, whereas two participants were taking Baclofen (10 mg/day) and three participants were prescribed Lyrica (2 × 300 mg/day), Ventolin, or Alvesco for asthma (one inhalation in morning) for each. Similarly, 13 uninjured participants were drug free, whereas three participants were taking Alvesco (one inhalation in the morning), acetylsalicylic acid (100 mg/day), or Euthyrox (100 $\mathrm{\mu }$g/day), respectively. All participants provided written informed consent. This study was approved by the clinical research ethics board of School of Medicine in Split, University of Split (Split, Croatia). Participants were tested over a 1-h protocol that consisted of at least 15 min resting in the supine position while being outfitted with the assessment equipment. Next, 10 min of resting data were collected for analysis. Two subjects had motion arteficts over the course of data collections and only 6 min of their clean data were used for analyses. For each participant, brachial BP was measured (Dinamap, General Electric Pro 300V2, Tampa, FL) on the left brachial artery each minute throughout the protocol. The following measurements were sampled at 1 kHz using an analog-to-digital converter (Powerlab/16SP ML 795, AD Instruments, Colorado Springs, CO) interfaced with data-acquisition software on a laptop computer (LabChart 8, AD Instruments): beat-by-beat BP via finger photoplethysmography (Finometer PRO, Finapres Medical Systems, Amsterdam, The Netherlands) and the electrocardiogram. Middle cerebral artery (MCA) blood velocity was recorded in the right MCA (ST3 Transcranial Doppler, Spencer Technologies, Redmond, WA) using a 2-MHz probe mounted on the temporal bone and a fitted head strap, according to detailed instructions published elsewhere by our group (22, 27). Systolic BP (SBP) and diastolic BP (DBP) as well as peak MCA blood velocity (MCAv_max) and minimum MCAv (MCAv_min) were used to calculate mean arterial BP as follows: (2 × DBP + SBP)/3 and mean MCA velocity (MCAv_mean) = (2 × MCAv_min + MCAv_max)/3, respectively. End-tidal Pco₂ ($\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$) was sampled from a leak-free mouth piece and measured by a gas analyzer (ML206, AD Instruments). Consistent with our previous study (30), the 1-kHz continuous recordings of BP, MCA velocity, and $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ were transformed to beat-to-beat mean arterial pressure (MAP) and MCAv_mean values and breath-to-breath $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ followed by cubic-spline interpolation and decimated to 2 Hz. A high-pass fifth-order Butterworth filter with a cutoff frequency of 0.005 Hz was used to remove very slow variations and baseline drift. The occasional artifacts were removed using Grubb’s test. Analyses were done using signal processing software Ensemble (version 1, Elucimed, Wellington, New Zealand) and MATLAB (version R2014b, MATHWORKS, Natick, MA).

Table 1.

Selected demographic information and supine resting physiological data for uninjured individuals and participants with spinal cord injury

Variable	Uninjured	Spinal Cord Injury	P Value
Age, yr	43 ± 3	43 ± 4	0.89
Height, m	1.8 ± 0.04	1.8 ± 0.02	0.92
Mass, kg	81 ± 5	71 ± 4	0.04
Body mass index, kg/m2	25 ± 0.7	22 ± 0.8	0.007
Years since injury	NA	20 ± 3	NA
Heart rate, beats/min	57 ± 2.4	59 ± 3.0	0.52
Mean arterial pressure, mmHg	84 ± 2.1	77 ± 2.8	0.21
Systolic blood pressure, mmHg	116 ± 3.6	105 ± 3.6	0.05
Diastolic blood pressure, mmHg	69 ± 1.7	63 ± 2.8	0.08
Mean MCA blood velocity, cm/s	57 ± 3.3	54 ± 2.8	0.52
Maximum MCA blood velocity, cm/s	89 ± 4.9	85 ± 4.3	0.51
Minimum MCA blood velocity, cm/s	41 ± 2.7	39 ± 2.3	0.50
End-tidal Po2, mmHg	39 ± 1.1	40 ± 1.3	0.42
Breathing rate, breaths/min	14 ± 0.6	15 ± 0.4	0.55

Values are means ± SE. MCA, middle cerebral artery; NA, not applicable. P values were not corrected for multiple comparisons.

Wavelet decomposition analysis.

As described in Ref. 30, the wavelet transform of a signal x(t) is given by the following equation:

$$W_x\left(a,t_0\right)=\frac{1}{\sqrt{a}}{\displaystyle \underset{t=-\infty }{\overset{\infty }{\int }}x\left(t\right)\psi *\left(\frac{t-t_0}{a}\right)}\text{d}t$$ (1)

where a is the scale creating translation and dilation of the complex mother wavelet ψ (Morlet wavelet in this study), which is given by the following equation:

$$\psi \left(t\right)=\frac{1}{\sqrt{\pi f_{\text{b}}}}{e}^{j2\pi f_{\text{c}}t}{e}^{-t^2/f_{\text{b}}}$$

where f_b and f_c are the bandwidth parameter and center frequency, respectively, which are both set to 1 in this study. The time index t₀ moving along a time series extracts local features of the signal x(t)$.$ The scale parameter a is related to f_c as follows:

$$f_{\text{a}}=\frac{f_{\text{c}}}{a\delta t^{\prime }}$$

where f_a is a pseudo-frequency and δt is the sampling period. The wavelet transform W_x(a,t₀) can be used to derive the instantaneous phase of signal x(t) as follows:

$${\varphi }_x\left(a,t_0\right)=-i\log \left[\frac{{\text{W}}_x\left(a,t_0\right)}{\left|{\text{W}}_x\left(a,t_0\right)\right|}\right]$$

The phase difference between MAP and MCAv_mean is given by the following equation:

$$\Delta {\varphi }_{\text{BM}}={\varphi }_{\text{B}}\left(a,t_0\right)-{\varphi }_{\text{M}}\left(a,t_0\right)$$

where ϕ_B and ϕ_M are wavelet-based instantaneous phases of MAP and MCAv_mean, respectively. The phase synchrony index (PSI) is given by the following equation:

$$\gamma \left(a\right)=\frac{1}{N}\left({\left\{{\displaystyle \sum _{t_0}\cos \left[\Delta {\text{Ø}}_{\text{BM}}\left(a,t_0\right)\right]}\right\}}^2+{\left\{{\displaystyle \sum _{t_0}\sin \left[\Delta {\text{Ø}}_{\text{BM}}\left(a,t_0\right)\right]}\right\}}^2\right)$$ (2)

where N is the data length. The PSI values range in the interval 0 < γ < 1. The smallest value of PSI (γ = 0) represents a uniform distribution of wavelet phase differences between MAP and MCAv_mean (i.e., MAP and MCAv_mean are not synchronized), which may indicate that cerebrovascular buffering of BP is pronounced. A large PSI value (γ = 1) denotes perfect synchronization (i.e., changes in MAP and MCAv_mean are occurring at an identical phase angle), indicating MCAv_mean closely follows MAP, and that cerebrovascular buffering of BP is not occurring.

After $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ correction, the phase difference becomes as follows:

$$\Delta {\varphi }_{BM}^c\left(a,t_0\right)=\Delta {\varphi }_{BM}\left(a,t_0\right)-{\tan }^{-1}\left[\frac{\left|{\text{W}}_{\text{C}}{H}_{\text{CM}}\right|\sin \left({\varphi }_{\text{BC}}+{\psi }_{\text{CM}}-{\psi }_{\text{BM}}\right)}{\left|{\text{W}}_{\text{B}}{H}_{\text{BM}}\right|+\left|{\text{W}}_{\text{C}}{H}_{\text{CM}}\right|\cos \left({\varphi }_{\text{BC}}+{\psi }_{\text{CM}}-{\psi }_{\text{BM}}\right)}\right]$$ (3)

where ϕ_BC is the wavelet-based phase difference between MAP and $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ and ψ_CM and ψ_BM are transfer function-based phase differences of $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$-MCAv_mean and MAP-MCAv_mean, respectively, i.e., ${\psi }_{\text{CM}}=\angle H_{\text{CM}}\left(e^{jw}\right)$ and ${\psi }_{\text{BM}}=\angle H_{\text{BM}}\left(e^{jw}\right)$. H_CM and H_BM are the transfer functions of $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$-MCAv_mean and MAP-MCAv_mean, respectively. W_C and W_B are the wavelet coefficients using Eq. 1 for $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ and MAP, respectively. In Eq. 3, the second term is the identified phase error caused by $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$.

Transfer function analysis.

As previously described (30), the transfer function analysis metrics (i.e., coherence, gain and phase) were derived using the following equation:

$$H_{\text{BM}}=\frac{S_{\text{BM}}}{S_{\text{BB}}}$$

where S_BM is the cross-spectrum between MAP and MCAv_mean and S_BB is the auto-spectrum of MAP. The power spectral density (PSD) and mean values of transfer function metrics were calculated in traditional very-low-frequency (VLF; 0.02–0.07 Hz), low-frequency (LF; 0.07–0.2 Hz), and high-frequency (HF; 0.2–0.4 Hz) ranges. We did not adjust or control for coherence while averaging metrics across these frequency ranges.

Statistics.

The normality of data was verified using a Shapiro-Wilk test. Transfer function coherence and phase were r-to-z and arcsine transformed, respectively, to obtain asymptotic distributions. For ease of interpretation and analysis, all values are presented here in standard units. Following the assessment of variance homogeneity, participants with SCI and uninjured individuals were compared using parametric (i.e., independent-samples t-tests) or nonparametric comparisons (i.e., related-samples Wilcoxon signed-rank test or independent-samples Mann-Whitney U-test) with Bonferroni corrections. Significance was set a priori at P < 0.05 unless otherwise stated. Values are represented as means ± SE unless otherwise stated.

RESULTS

Physiological characteristics of the two groups are shown in Table 1. There were no significant differences between groups for maximum, minimum, or mean blood velocity in the MCA, heart rate, $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$, and breathing rate. In participants with SCI, MAP tended to be reduced (Table 1).

There was greater PSI between MAP and MCAv_mean in the ULF ranges (0.02–0.025 and 0.025–0.03 Hz) in the SCI group, indicating impaired capacity for cerebrovascular buffering of MAP (Fig. 1, A and B) (16). Furthermore, participants with SCI had greater PSI between $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ and MCAv_mean at 0.02–0.025 Hz, indicating a more direct influence of $\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$ on MCAv_mean after SCI. None of the PSD or transfer function analysis metrics were different between groups in the ULF ranges (0.02–0.025 and 0.025–0.03 Hz; Fig. 2). In terms of traditional frequency ranges, the PSD for MAP or MCAv_mean was not different between groups across VLF or HF ranges; however, MCAv_mean PSD was reduced in participants with SCI for the LF range (Table 2). Coherence and gain were not different between groups for any frequency range;, however, the LF phase was reduced in the SCI group (Table 2). The effect size (Cohen’s d) for PSI at 0.02–0.03 Hz was 0.89, and LF transfer function phase effect size was 0.31. As subanalyses, we also compared 1) only participants with motor complete injury (AIS A or B) with control subjects as well as 2) participants with motor complete (AIS A or B) versus incomplete (AIS C or D) injuries. With either of these analyses, the primary outcomes (e.g., PSI or transfer function analysis metrics) did not change.

Fig. 1.

Wavelet decomposition analysis of mean arterial pressure (MAP) and middle cerebral artery (MCA) velocity (MCAv), with and without end-tidal Pco₂ ($\mathrm{P}\mathrm{E}{\mathrm{T}}_{\mathrm{C}{\mathrm{O}}_{\mathrm{2}}}$) correction, in uninjured individuals and participants with a spinal cord injury (SCI). The group with SCI presented with impaired capacity for cerebrovascular buffering, as indicated by the increased phase synchrony index [PSI; in arbitrary units (au)], in the MCA in the ultra-low-frequency ranges (0.02–0.025 and 0.025–0.03 Hz, A and D). This difference persisted (B and E) after correction for the direct influence that arterial CO₂ tension exerted on MCAv (C and F). *Significance at P < 0.05.

Fig. 2.

Power spectrum and transfer function analysis of mean arterial pressure (MAP) and middle cerebral artery velocity (MCAv) in uninjured individuals and participants with a spinal cord injury (SCI). No significant changes were found across ultra-low-frequency ranges (0.02–0.025 and 0.025–0.03 Hz). PSD, power spectral density, AU, arbitrary units.

Table 2.

Power spectrum and transfer function analysis across traditional frequency ranges

Variable	Uninjured	Spinal Cord Injury	P Value
VLF MAP PSD, mmHg2/Hz	2.4 ± 0.5	2.1 ± 0.5	0.67
LF MAP PSD, mmHg2/Hz	2.5 ± 1.1	1.3 ± 0.4	0.38
HF MAP PSD, mmHg2/Hz	0.4 ± 0.2	0.9 ± 0.3	0.29
VLF MCAvmean PSD, cm2·s−2·Hz−1	3.0 ± 0.6	2.9 ± 0.6	0.51
LF MCAvmean PSD, cm2·s−2·Hz−1	2.0 ± 0.3	1.1 ± 0.3	0.01
HF MCAvmean PSD, cm2·s−2·Hz−1	0.4 ± 0.1	1.0 ± 0.5	0.79
VLF coherence, AU	0.4 ± 0.04	0.3 ± 0.1	0.41
LF coherence, AU	0.5 ± 0.04	0.4 ± 0.03	0.47
HF coherence, AU	0.6 ± 0.06	0.5 ± 0.07	0.76
VLF gain, cm·s−1·mmHg−1	0.8 ± 0.1	0.9 ± 0.1	0.45
LF gain, cm·s−1·mmHg−1	1.1 ± 0.2	0.8 ± 0.1	0.25
HF gain, cm·s−1·mmHg−1	1.1 ± 0.2	0.9 ± 0.1	0.27
VLF phase, radians	0.9 ± 0.1	1.0 ± 0.1	0.21
LF phase, radians	0.7 ± 0.6	0.2 ± 0.2	0.03
HF phase, radians	0.2 ± 0.6	0.3 ± 0.3	0.46

Values are means ± SE. VLF, very low frequency (0.02–0.07 Hz); MAP, mean arterial blood pressure; PSD, power spectral density; LF, low frequency (0.07–0.2 Hz); HF, high frequency (0.2–0.4 Hz); MCAv_mean, middle cerebral artery mean blood velocity; AU, arbitrary units.

DISCUSSION

Nonlinear and nonstationary cerebrovascular buffering of BP is impaired after sustaining a cervical SCI, which persists after accounting for the cerebrovascular influence of arterial CO₂ tension. This supports our hypothesis. That this impairment occurred in the ULF range recently identified to be influenced by the sympathetic nervous system indicates that the impaired capacity for cerebrovascular buffering of BP is likely the result of disrupted supraspinal sympathetic control over the cerebrovasculature (30).

From a clinical perspective, the present data demonstrate that resting BP and CBF recordings can be used to detect ULF cerebrovascular buffering of BP in clinical populations where pharmacological or gravitational manipulations of BP may not be induced. From a mechanistic perspective, these data identify that sympathetic pathways responsible for cerebrovascular control are disrupted by injury to the spinal cord. From a technical perspective, our data indicate that PSI may be a statistically more powerful metric to detect differences in cerebrovascular buffering of BP between uninjured and SCI groups and can account for important direct influences of arterial CO₂ tension.

The results indicate that impaired cerebrovascular buffering after SCI may be due to the loss of supraspinal sympathetic control over the cerebrovasculature, which results from the disruption of descending sympathetic pathways traversing the spinal cord (20, 36). This frequently occurs after cervical SCI as the sympathetic preganglionic neurons contributing to the superior cervical ganglia that are responsible for cerebrovascular tone [located primarily at the T1−T3 spinal segments (20, 36)] are disconnected from supraspinal sympathetic tonic support (20). The increase in phase synchrony in cervical SCI may contribute to the highly prevalent orthostatic intolerance in this population, as elevated phase synchrony is associated with fainting (16).

In this study, WDA clearly demonstrated impaired cerebrovascular buffering after SCI in the identical frequency range (0.02–0.03 Hz) to that shown by our group previously to be impaired by pharmacological blockade of sympathetic cerebrovascular control in uninjured individuals (30). The anatomic disruption of descending sympathoexcitatory pathways is similar to pharmacological α₁-receptor antagonism such that in both situations effects of norepinephrine on the vasculature are largely inhibited. Cervical SCI results in blunted sympathetic preganglionic excitation, and therefore very little norepinephrine release occurs by ganglionic neurons innervating the vasculature. In contrast, α₁-receptor antagonism leads to the reduced capacity for norepinephrine to influence downstream cascades within the smooth muscle responsible for altering vascular tone. With cervical SCI, there is additional disruption of sympathetic influences on β-receptors, which may additionally impact cerebrovascular regulation (8, 10). In both scenarios, any parasympathetic influence on cerebrovascular tone is intact and functional (37, 38). In short, our findings provide supporting evidence using known sympathetic neuroanatomic changes occurring after SCI to support previous pharmacological sympathetic α₁-blockade data from our group, indicating that adrenergic pathways exert influences on the cerebrovasculature in the ULF range (30).

WDA detected a greater influence of arterial CO₂ tension on CBF in participants with SCI, as indicated by the increased PSI between CO₂ and CBF. Our group has previously examined the CBF responsiveness to induced hypercapnia and hypocapnia (47); however, the present data are the first to show that at rest in participants with SCI the relationship between arterial CO₂ and CBF is altered. These relationships certainly deserve to be explored further, as alterations in cerebrovascular sensitivity to arterial CO₂ tension are associated with a number of prevalent secondary health issues after SCI, such as cognitive dysfunction and sleep disordered breathing (1, 2, 27, 32–34, 44). Even after statistically accounting for the resting CO₂ tension, the significant impairment of cerebrovascular buffering of BP in SCI persisted. As such, the utilization of WDA for cerebrovascular buffering is a robust research tool to evaluate the cerebrovascular buffering of BP while accounting for the influence of arterial CO₂ tension.

This study corroborates work demonstrating a previously unknown frequency (0.02–0.03 Hz) in which the sympathetic nervous system appears to be modulating cerebrovascular tone (11–13, 30, 31). Although the systemic vasculature has been well established to have supraspinal sympathetic modulation occurring at a frequency of 0.07–0.2 Hz, there is now mounting evidence that the cerebrovasculature is modulated at VLF or ULF ranges (30, 31). The teleological rationale for the brain operating as a high-pass filter, where rapid changes in BP are relatively less buffered compared with slower changes may be to allow for increased perfusion during periods of mental activation, an organization that would theoretically not interfere with other more rapid cerebrovascular regulatory mechanisms such as neurovascular coupling (17, 27, 40).

Our analysis indicates that WDA holds greater statistical power to detect differences in cerebrovascular buffering of BP. Transfer function analysis parameters were not different in the ULF range between groups; however, we confirmed our previous findings that in the LF range, phase was reduced between BP and CBF (22), implying impaired cerebrovascular buffering in SCI. Our group’s work has demonstrated that the LF phase closely correlates with Windkessel compliance (42), indicating that viscoelastic properties of individuals with SCI may be compromised. This finding corroborates our preclinical and clinical work showing that SCI is associated with stiffening of central and cerebral arteries (18, 24, 25). The effect size for the ULF PSI was threefold greater than the LF transfer function phase between uninjured and SCI groups, suggesting that WDA may be a more statistically powerful approach for detecting impairments in cerebrovascular buffering of BP.

The present data support the contention that nonlinear and nonstationary approaches are able to detect differences in the ULF cerebrovascular buffering from supine resting hemodynamic data, which is complementary to that detected in standard frequency ranges using traditional linear approaches such as transfer function analysis. The concept that nonlinear and nonstationary relationships underlie hemodynamic interactions supports recent work showing that baroreflex regulation of the systemic vasculature exhibits nonlinear and dynamic behavior (14, 15).

The primary cardiovascular consequence after SCI is the loss of supraspinal tonic control of the vasculature, which is widely thought to operate in the LF range (roughly 10 s in humans), at least in the systemic vasculature, and certainly at the level of sympathetic neurons (29). We cannot be certain, as is the case in most study designs, that what we have measured reflects solely the loss of sympathetic control after SCI. However, our data, based off what we know about descending sympathetic pathways after SCI, combined with our previous work using pharmacological blockade (30), strongly suggests this is the case. Further to this, although difficult to conclusively show, our data suggest that sympathetic medially cerebrovascular buffering is occurring at a frequency previously unknown to be sympathetically mediated. Specifically, two studies, using either pharmacological [i.e., Prazosin in Saleem et al. (30)] or neuroanatomic [i.e., SCI) in the present study] interventions, have demonstrated that there appears to be a loss of regulation at a much different frequency (i.e., 0.025–0.03 Hz). This unique aspect of the present study remains to be conclusively demonstrated.

In the present study, WDA detected significant impairments in similar frequency ranges to those previously shown to be impacted by sympathetic vascular blockade (30). These changes persisted after correction for the direct influence that arterial blood gases exert on CBF. As WDA and transfer function analysis both detected changes without potentially dangerous perturbations in BP, these approaches both appear to be clinically relevant for detecting altered cerebrovascular buffering, although WDA was more statistically powerful. Future studies should explore strategies for preventing/rehabilitating cerebrovascular buffering of BP. Reducing the amount of disruption of descending sympathetic pathways due to SCI may be a viable target for restoring cerebrovascular function and health and mitigating the drastically elevated risk of stroke (4, 48), vascular-cognitive impairment (22, 26, 44), and orthostatic intolerance in this population.

GRANTS

This work was supported by funding to the laboratory of A. A. Phillips from the Libin Cardiovascular Institute of Alberta, Hotchkiss Brain Institute, Rick Hansen Institute, and Michael Smith Foundation for Health Research as well as an ICORD seed grant supported by the Blusson Integrated Cures Partnership (to P. N. Ainslie and A. A. Phillips). P. N. Ainslie and G. B. Coombs were supported by the Natural Sciences and Engineering Research Council, and P. N. Ainslie is a Canada Research Chair.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).

AUTHOR CONTRIBUTIONS

S.S., D.V., Z.S., A.H.X.L., O.F.B., G.B.C., T.M., Z.D., Y.-C.T., and A.A.P. conceived and designed research; S.S., A.H.X.L., G.B.C., T.M., and A.A.P. analyzed data; S.S., D.V., A.H.X.L., J.W.S., A.V.K., P.N.A., Y.-C.T., and A.A.P. interpreted results of experiments; S.S., J.W.S., and A.A.P. prepared figures; S.S., Z.S., J.W.S., A.V.K., P.N.A., Y.-C.T., and A.A.P. drafted manuscript; S.S., D.V., Z.S., A.H.X.L., J.W.S., O.F.B., G.B.C., T.M., A.V.K., P.N.A., Z.D., Y.-C.T., and A.A.P. edited and revised manuscript; S.S., D.V., Z.S., A.H.X.L., J.W.S., O.F.B., G.B.C., T.M., A.V.K., P.N.A., Z.D., Y.-C.T., and A.A.P. approved final version of manuscript; D.V., Z.S., O.F.B., T.M., A.V.K., Z.D., and A.A.P. performed experiments.