Relationship between blood pressure and cerebral blood flow during supine cycling: influence of aging

Abstract

The cerebral pressure-flow relationship can be quantified as a high-pass filter, where slow oscillations are buffered (<0.20 Hz) and faster oscillations are passed through relatively unimpeded. During moderate intensity exercise, previous studies have reported paradoxical transfer function analysis (TFA) findings (altered phase or intact gain). This study aimed to determine whether these previous findings accurately represent this relationship. Both younger (20–30 yr; n = 10) and older (62–72 yr; n = 9) adults were examined. To enhance the signal-to-noise ratio, large oscillations in blood pressure (via oscillatory lower body negative pressure; OLBNP) were induced during steady-state moderate intensity supine exercise (∼45–50% of heart rate reserve). Beat-to-beat blood pressure, cerebral blood velocity, and end-tidal Pco₂ were monitored. Very low frequency (0.02–0.07 Hz) and low frequency (0.07–0.20 Hz) range spontaneous data were quantified. Driven OLBNP point estimates were sampled at 0.05 and 0.10 Hz. The OLBNP maneuvers augmented coherence to >0.97 at 0.05 Hz and >0.98 at 0.10 Hz in both age groups. The OLBNP protocol conclusively revealed the cerebrovascular system functions as a high-pass filter during exercise throughout aging. It was also discovered that the older adults had elevations (+71%) in normalized gain (+0.46 ± 0.36%/%: 0.05 Hz) and reductions (−34%) in phase (−0.24 ± 0.22 radian: 0.10 Hz). There were also age-related phase differences between resting and exercise conditions. It is speculated that these age-related changes in the TFA metrics are mediated by alterations in vasoactive factors, sympathetic tone, or the mechanical buffering of the compliance vessels.

maintenance of the blood supply to the heart and brain is of critical importance for the cardiovascular system, both at rest and during exercise. The brain is especially susceptible to alterations in blood flow, as it is unable to store large amounts of substrates and has a high metabolic rate (8). To maintain blood flow, the brain has developed the ability to regulate its blood flow somewhat independently of alterations to the rest of the body (1, 23, 25, 33, 36). An excellent example of this can be observed by examining the relationship between arterial blood pressure and cerebral blood flow during exercise (reviewed in Ref. 27). The global cerebral blood velocity (CBV) response during mild-moderate (40–60% of maximal intensity) bilateral cycling exercise has shown an increase in CBV (∼20–25%), while mean arterial pressure only increased ∼5–10% in both younger and older adults (40). As exercise intensity proceeds to exhaustion, there is a further elevation in mean arterial pressure (∼25–35%), while CBV can return to levels comparable to those at baseline, likely due to exercise-induced hypocapnia (40).

Over the course of normal healthy human aging, physiological changes occur that result in the structural and functional alterations of the cerebrovascular system such as loss of white matter (18), hormonal changes in the brain (26), loss of brain mass (18), and a progressive loss of cortical neurons (34, 35). As a result of these alterations, there is an observed reduction in resting cerebral blood flow and CBV (5). Cross-sectional studies have revealed that, with healthy aging, cerebral blood flow declines ∼25–40% between 30 and 89 yr of age (5, 11, 26). Similar findings were reported in a longitudinal study by Fotenos et al. (18), namely, there is a reduction in whole brain volume of 0.45% per yr after 30 yr of age.

The relationship between arterial blood pressure and cerebral blood flow is a frequency-dependent process that functions as a high-pass filter, where the cerebral arterioles are able to adjust blood pressure changes at frequencies below 0.20 Hz (complete oscillations ∼5 s in duration) and subsequently regulate the delivery of cerebral blood flow (1, 14, 47, 54). One common method to assess this dynamic relationship is through transfer function analysis (TFA) (14, 42, 54). Through TFA, the high-pass filter model has revealed that phase (timing buffer) decreases and gain (amplitude modulation buffer) increases from 0.02 to 0.20 Hz (14, 54) and, under resting conditions, has been shown to be unaltered in individuals up to ∼70 yr old (3, 13, 24, 27, 42, 45, 50). However, this notion has yet to be assessed in older adults, during exercise, with TFA.

Although conditions at rest are useful for understanding the intricacies of the frequency-dependent nature underlying this relationship, they might not inform how the brain deals with more profound changes in blood pressure, such as those encountered during everyday life and exercise. To date, there have been seven studies that have attempted to examine the cerebral pressure-flow relationship in the middle cerebral artery, during exercise, with TFA (Table 1) (4, 6, 10, 16, 28, 29, 31). All of these TFA studies were performed in young (mean age 25 yr) healthy individuals, and one study also included a middle-aged adult cohort (mean age 57 yr) (16). Two of these studies examined this relationship in both the very low frequency (VLF) and low frequency (LF) ranges, which revealed paradoxical findings (16, 29); namely, the TFA phase response does not follow the high-pass filter model (i.e., there is a greater buffering in the timing of the CBV and blood pressure, rather than the expected decrease, as frequency increases from 0.02 to 0.20 Hz), whereas gain does follow the expected trends (i.e., a greater extent of the amplitude of the blood pressure wave form is transmitted to the cerebrovasculature as frequency increases from 0.02 to 0.20 Hz). The findings from these studies are paradoxical in nature, as they imply that there is an enhanced timing buffer (greater CBV phase lead) present at the high-pass frequency cutoff point of 0.20 Hz, while the amplitude buffer is unchanged. This notion is inconsistent with the current understanding of high-pass filter nature of the cerebral pressure-flow relationship (1, 14, 47, 54). However, a major limitation of all of the previous TFA studies during moderate exercise is that they were assessed during spontaneous oscillations in blood pressure, which revealed a relatively low coherence (∼0.44) in both the VLF (0.02–0.07 Hz) and LF (0.07–0.20 Hz) bands during the moderate exercise intensity (∼50% heart rate reserve), a factor that makes the interpretation of the phase and gain metrics less reliable (42, 47, 54). Recently, it has been demonstrated that there is poor reproducibility for TFA metrics associated with spontaneous oscillations, and as such, findings from studies employing this method of cerebral pressure-flow dynamics assessment should be interpreted with caution (42). The gold standard method to counteract the poor reproducibility associated with spontaneous oscillations in blood pressure is to induce large oscillations in blood pressure (such as those attained with squat-stand maneuvers) (42). When oscillations in blood pressure are induced, they enhance the signal-to-noise ratio by increasing the power of the input signal (blood pressure) which is reflected in increased power of the output signal (CBV), thus increasing coherence and the mathematical interpretability of the phase and gain metrics (22, 41–44). Although this approach has been well defined at rest (9, 19, 42, 48), it has never been applied during exercise.

Table 1.

Summary of previous literature assessing the cerebral pressure-flow relationship during moderate exercise (∼50% of heart rate reserve)

		Coherence		Phase, radians		Gain, cm·s−1·mmHg−1
	HR, bpm	VLF	LF	VLF	LF	VLF	LF
Brys et al. (10)
N = 40; 28 ± 5 yr	∼130		0.50 ± 0.10		0.78 ± 0.10		0.94 ± 0.07
Ogoh et al. (30)
N = 7; 26 ± 1 yr	∼115		0.70 ± 0.05		0.63 ± 0.13		0.82 ± 0.10
Ogoh et al. (29)
N = 7; 25 ± 1 yr	∼120		0.65 ± 0.03		0.20 ± 0.20		1.10 ± 0.20
Ogoh et al. (28)
N = 8; 22 ± 2 yr	∼130	0.38 ± 0.05	0.63 ± 0.04	0.31 ± 0.22	0.41 ± 0.05	0.62 ± 0.05	0.97 ± 0.08
Ainslie et al. (4)
N = 14; 25 ± 4 yr	∼130		0.60 ± 0.05		0.90 ± 0.50		0.85 ± 0.45
Ainslie et al. (6)
N = 28; 26 ± 5 yr	∼130		0.58 ± 0.06		0.83 ± 0.55		0.90 ± 0.50
Fisher et al. (16)
N = 9; 24 ± 3 yr	∼130	0.45 ± 0.15	0.52 ± 0.10	0.60 ± 0.52	0.85 ± 0.34	0.55 ± 0.25	0.85 ± 0.13
N = 10; 57 ± 7 yr	∼110	0.50 ± 0.12	0.56 ± 0.12	0.35 ± 0.44	0.75 ± 0.20	0.45 ± 0.20	0.75 ± 0.15

Values are means ± SD.

HR, heart rate; VLF, very low frequency (0.02–0.07 Hz); LF, low frequency (0.07–0.20 Hz).

This study was designed to examine the cerebral pressure-flow relationship during moderate intensity exercise to determine whether the previously reported paradoxical findings (altered phase or intact gain) accurately represent this relationship. For the first time, an experimental approach was devised using nonpharmacological manipulations of blood pressure variability (via oscillatory lower body negative pressure; OLBNP) that were induced during moderate intensity supine exercise and related frequencies of interest (0.05 and 0.10 Hz). This approach was designed to enhance the signal-to-noise ratio (to levels typically observed with squat-stand maneuvers) and thus provide a more accurate representation of the TFA metrics during exercise (42). The hypotheses were twofold: 1) OLBNP will elicit swings in blood pressure during exercise, which will enhance the coherence and thus interpretability of the phase and gain metrics; and 2) despite the decrease in CBV and alterations to the cerebrovasculature that occur within the older adult population (5, 16, 37, 39–41), the high-pass filter model of the cerebral pressure-flow relationship will be intact during moderate exercise in both younger and older adults.

METHODS

Ethical Approval

This study was approved by the clinical ethical committee of the University of British Columbia and adhered to the principles of the Declaration of Helsinki. All volunteers provided written informed consent, and procedures were followed in accordance with institutional guidelines.

Subjects

Nine healthy younger adult [10 males, 24.8 ± 2.7 yr, body mass index (BMI) 23.6 ± 2.4 kg/m²] and eight healthy older adult (8 males, 66.4 ± 3.7 yr, BMI 25.6 ± 2.7 kg/m²) subjects were recruited for this study. All of the subjects had a clear history of cardiorespiratory and cerebrovascular diseases and were not taking any form of medication. All older adult subjects were screened for any evidence of carotid stenosis, and after completion of the screening process one older male adult was excluded from the study; therefore, the older adults results are based on n = 9. All subjects abstained from exercise, caffeine, and alcoholic beverages for a period of 12 h prior to the study. Each subject underwent a familiarization of the laboratory and testing protocols before the initiation of the protocols.

Experimental Protocols

The subjects were required to visit the laboratory on two occasions, separated by at least 48 h. The first testing visitation determined the workload for the supine cycling exercise protocol (which was completed on the second visit). A maximal exercise test was implemented and consisted of participants performing a 5-min warm-up at a self-selected pace on the supine cycle ergometer. The initial workload was the equivalent of their warm-up pace and workload was increased by 25 W (20 W for older adults) every 2 min until a predetermined heart rate was achieved (130 beats/min for younger adults; 100 beats/min for older adults). After the predetermined heart rate level was achieved, the workloads were increased every minute until the subject was no longer able to continue. The end of the test was determined by the subject reaching an age-predicted maximal heart rate, a perceived exertion of 9–10 on a 1–10 Borg scale, volitional exhaustion had occurred, and/or the subject was no longer able to sustain a pedal cadence of 60 rpm. The peak workload obtained during this session was used to determine the workload for the moderate exercise challenge on the second testing day.

On the second testing visitation, the subjects were examined with OLBNP maneuvers both at rest and during moderate exercise. In both cases, the subject was placed in a supine position in an industrially designed pressure chamber that sealed at their waist and were held in place via a mounted bicycle seat to prevent movement into the OLBNP chamber. The straps used to seal the participants in the OLBNP chamber did not interfere with the subjects' breathing patterns. For the resting data, a 5-min period at 0 mmHg was performed to collect baseline data. The pressure in the box was then oscillated between −50 and 0 mmHg at 0.10 Hz (5 s at −50 mmHg followed by 5 s at 0 mmHg). This process was then repeated at a frequency of 0.05 Hz (10 s at −50 mmHg followed by 10 s at 0 mmHg). The −50 mmHg pressure change occurred within 1.5 s. The −50 mmHg pressure was selected to match the pressure generated in the steady-state findings of Formes et al. (17), who tested a similarly aged population (68 ± 1 yr) at a sustained negative pressure of −50 mmHg. The results from this study revealed that this level of negative pressure is well tolerated in older individuals. For the exercise data, the subject workload was set at 46 ± 4% of their previous day's peak wattage and was performed until steady-state was achieved. The moderate exercise intensity was selected based on previous research (40), which has demonstrated that this intensity is associated with the greatest elevation in CBV and coincides with the minimal change to end-tidal CO₂ (Pet_CO2). After the subject was at steady state, the testing protocol of the cycling exercise was initiated in the OLBNP chamber. The first trial consisted of 5 min at 0 Hz of steady-state exercise with the vacuum turned on to simulate the experience of the negative pressure but provide a comparable environment to the previously performed studies. The second and third trials consisted of steady-state cycling bouts while the OLBNP chamber was cycled between 0 and −50 mmHg, with pressure changes occurring within 1.5 s (typical trace presented in Fig. 1.). A greater than 5-min rest break was provided between frequencies to allow for the subject to partially recover from the previous exercise bout. The order of the 0.05- and 0.10-Hz trials was randomly selected for OLBNP maneuvers, both at rest and during exercise.

Fig. 1.

Typical trace for blood pressure (BP), cerebral blood velocity (CBV), and end tidal CO₂ (Pet_CO2) during spontaneous (top), 0.05 Hz (middle), and 0.10 Hz (bottom) oscillatory negative pressure trials during a moderate supine cycling intensity.

Instrumentation

Subjects first received a three-lead electrocardiogram (ECG) for measurement of the R-R interval. Blood pressure was measured in the finger by photoplethysmography (Finometer; Finapres Medical Systems, Amsterdam, The Netherlands). This method has been shown to reliably assess the dynamic changes in beat-to-beat blood pressure that correlate well with the intra-arterial recordings and can be used to characterize the dynamic relationship between blood pressure and cerebral blood flow (32, 38).

Bilateral middle cerebral arteries were insonated by placing 2-MHz Doppler probes (Spencer Technologies, Seattle, WA) to obtain CBV. The middle cerebral arteries were identified and optimized according to their signal depth, wave form, and velocities (2, 52). Once the cerebral arteries were identified, the probes were secured and locked in place with a headband (Spencer Technologies). Cerebrovascular resistance index (CVRi) was calculated from mean arterial pressure divided by mean CBV. Pet_CO2 was monitored using an online gas analyzer (ML206; AD Instruments, Colorado Springs, CO) and was calibrated with a known gas concentration prior to each subject. Heart rate was calculated from the ECG. All data were recorded and stored for subsequent analysis using commercially available software (LabChart version 7.1; AD Instruments).

Data Processing

All data were simultaneously sampled at 1,000 Hz via an analog-to-digital converter (Powerlab 16/30 ML880; AD Instruments). Real time beat-to-beat mean values of blood pressure and middle cerebral artery velocity were determined from each R-R interval. All data were processed and analyzed with custom-designed software in LabView 2011 (National Instruments, Austin, TX).

Power Spectrum and Transfer Function Analysis

Beat-to-beat blood pressure and CBV signals were spline interpolated and resampled at 4 Hz for spectral and transfer function analyses based on the Welch algorithm. Each 5-min recording was first subdivided into five successive windows that overlapped by 50%. Data within each window were linearly detrended and passed through a Hanning window prior to discreet Fourier transform analysis. For transfer function analysis, the cross-spectrum between blood pressure and middle cerebral artery velocity was determined and divided by the mean arterial pressure autospectrum to derive the transfer function coherence, absolute gain (cm·s⁻¹·mmHg⁻¹), normalized gain (%/%), and phase (rad).

The transfer function coherence, gain, and phase of the driven blood pressure oscillations were sampled at the point estimate of the driven frequency (0.05 or 0.10 Hz). These points are the frequency at which the OLBNP was driven and have the greatest coherence and thus are most appropriate for the linear interpretability of the associated TFA metrics (42). These point estimates were selected as they are in the very low (0.02–0.07 Hz) and low (0.07–0.20 Hz), frequency ranges where cerebral autoregulation is thought to be operant (47, 54).

Statistical Analysis

Statistical analyses were performed using SPSS version 22.0. A three-way multivariate ANOVA was conducted to examine the effect of modality (resting or moderate exercise), age (younger vs. older adults), and frequency of oscillation (VLF, 0.05 Hz, LF, and 0.10 Hz) on the hemodynamic and cerebrovascular responses (mean arterial pressure, CBV, heart rate, and Pet_CO2). A three-way multivariate ANOVA was also conducted to examine the effects of the aforementioned parameters on the Fourier transform (mean arterial pressure and middle cerebral artery power spectra) and transfer function metrics (coherence, phase, absolute gain, and normalized gain). A Kolmogorov-Smirnov test for normal distribution was run on the data and found the phase, absolute gain, and normalized gain all had normal distributions (P > 0.05). Significance between frequency comparisons were determined with a Tukey post hoc comparison. Data are presented as means ± SD. Significance was set a priori at P < 0.05.

RESULTS

Respiratory, Cardiovascular, and Cerebrovascular Responses

Under both resting and exercise conditions there was no main effect between the different oscillation frequencies (spontaneous, 0.05, or 0.10 Hz) in either age group (P > 0.183) for any of the respiratory, cardiovascular, or cerebrovascular responses. There were no significant two- or three-way interactions with Pet_CO2 (P > 0.713; Tables 2 and 3). Mean arterial pressure was elevated ∼15–25% in the older adults during both resting and exercise conditions (P < 0.001). The CBV for older adults was reduced by ∼20% compared with younger adults throughout the interventions, which resulted in the older adult CVRi being elevated ∼45% (P < 0.001). By study design, there were main effects for both age group and modality on heart rate (P < 0.001). There was a significant interaction effect on heart rate (age group by modality; P < 0.001) in the current study. Overall, this resulted in the younger adults having an elevated heart rate (123.6 ± 13.1 beats/min; 47.4 ± 7.9% heart rate reserve) compared with the older adults (103.0 ± 12.4 beats/min; 47.3 ± 4.8% heart rate reserve; P < 0.001) averaged across all exercise trials. This was performed to ensure that, although both groups were exercising at different absolute heart rates, they were at the same relative moderate exercise intensity based on their heart rate reserves.

Table 2.

Hemodynamic and cerebrovascular responses in younger adults at rest and during exercise with spontaneous and driven blood pressure (0.05 and 0.10 Hz) oscillations

	Rest	Exercise
Spontaneous Supine
Mean arterial pressure, mmHg	82.5 ± 9.9	97.7 ± 17.6*
Left CBV, cm/s	66.6 ± 4.7	72.2 ± 7.3*
Right CBV, cm/s	63.2 ± 9.2	69.8 ± 10.1*
Cerebrovascular resistance index, mmHg·cm−1·s−1	1.3 ± 0.2	1.4 ± 0.3
Heart rate, bpm	59.2 ± 11.5	120.6 ± 11.6*
End tidal CO2, mmHg	39.5 ± 2.6	39.8 ± 1.9
0.05 Hz LBNP
Mean arterial pressure, mmHg	82.4 ± 12.0	93.2 ± 16.8*
Left CBV, cm/s	64.5 ± 4.5	66.6 ± 6.8*
Right CBV, cm/s	60.3 ± 7.8	65.4 ± 9.8*
Cerebrovascular resistance index, mmHg·cm−1·s−1	1.3 ± 0.3	1.4 ± 0.3
Heart rate, bpm	60.0 ± 12.2	125.9 ± 14.4*
End tidal CO2, mmHg	38.4 ± 2.1	38.6 ± 2.4
0.10 Hz LBNP
Mean arterial pressure, mmHg	83.4 ± 13.3	92.8 ± 16.8*
Left CBV, cm/s	62.7 ± 4.7	68.3 ± 7.7*
Right CBV, cm/s	58.9 ± 7.2	66.8 ± 10.1*
Cerebrovascular resistance index, mmHg·cm−1·s−1	1.4 ± 0.3	1.4 ± 0.4
Heart rate, bpm	60.8 ± 12.1	124.4 ± 13.5*
End tidal CO2, mmHg	37.9 ± 2.4	38.6 ± 2.6

Values are means ± SD.

LBNP, lower body negative pressure; CBV, cerebral blood velocity. Note: there was a main effect for modality but there were no main effect sbetween frequencies for younger adults.

^*Significantly different from rest, P < 0.05.

Table 3.

Hemodynamic and cerebrovascular responses in older adults at rest and during exercise with spontaneous and driven blood pressure (0.05 and 0.10 Hz) oscillations

	Rest	Exercise
Spontaneous Supine
Mean arterial pressure, mmHg	103.5 ± 11.2	113.1 ± 13.3*
Left CBV, cm/s	53.6 ± 9.7	57.9 ± 7.4
Right CBV, cm/s	53.1 ± 6.8	59.0 ± 8.4
Cerebrovascular resistance index, mmHg·cm−1·s−1	1.9 ± 0.4	2.0 ± 0.4
Heart rate (bpm)	63.8 ± 14.5	101.7 ± 13.6*
End tidal CO2, mmHg	37.5 ± 3.1	39.3 ± 3.5
0.05 Hz LBNP
Mean arterial pressure, mmHg	102.4 ± 12.1	108.3 ± 14.1*
Left CBV, cm/s	54.0 ± 9.8	56.5 ± 6.4
Right CBV, cm/s	53.3 ± 6.4	56.0 ± 6.7
Cerebrovascular resistance index, mmHg·cm−1·s−1	1.9 ± 0.4	2.1 ± 0.3
Heart rate, bpm	63.6 ± 13.0	103.3 ± 13.1*
End tidal CO2, mmHg	37.5 ± 2.7	37.8 ± 3.2
0.10 Hz LBNP
Mean arterial pressure, mmHg	102.3 ± 12.7	109.6 ± 17.0*
Left CBV, cm/s	53.4 ± 11.8	57.0 ± 6.9
Right CBV, cm/s	53.9 ± 8.2	56.3 ± 5.9
Cerebrovascular resistance index, mmHg·cm−1·s−1	1.9 ± 0.4	2.0 ± 0.3
Heart rate, bpm	63.5 ± 14.5	103.8 ± 10.9*
End tidal CO2, mmHg	37.8 ± 2.6	38.3 ± 3.0

Values are means ± SD.

LBNP, lower body negative pressure; CBV, cerebral blood velocity.

Note: there was a main effect for modality but there were no main effect between frequencies for older adults.

^*Significantly different from rest, P < 0.05.

TFA

Power spectrum and coherence.

There were main effects for both mean arterial pressure and CBV autospectra power for both modality and frequency with a significant three-way interaction for age group times modality times frequency (P = 0.004). The post hoc Tukey analysis revealed significant differences between spontaneous (VLF and LF) vs. driven metrics (0.05 Hz vs. 0.10 Hz; P < 0.002). There were no differences for the autospectra when assessed with the same condition (VLF vs. LF; P = 1.000) and 0.05 Hz vs. 0.10 Hz (P = 1.000; Tables 4 and 5).

Table 4.

Transfer function analysis metrics in younger adults for the relationship between mean arterial pressure and cerebral blood flow at rest and during moderate exercise

	Rest	Exercise
Spontaneous Oscillations
VLF MAP power, mmHg2	5.90 ± 3.00	7.47 ± 7.50
LF MAP power, mmHg2	2.93 ± 1.35	5.27 ± 3.28
VLF MCAv power, (cm/s)2	7.43 ± 4.15	9.52 ± 2.81
LF MCAv power, (cm/s)2	4.30 ± 1.98	6.36 ± 2.68
VLF coherence	0.35 ± 0.16	0.28 ± 0.11
LF coherence	0.65 ± 0.14	0.48 ± 0.16
VLF phase, radians	0.76 ± 0.60	1.56 ± 0.61‡
LF phase, radians	0.54 ± 0.29	1.01 ± 0.54‡
VLF gain, cm·s−1·mmHg−1	1.14 ± 0.36	0.95 ± 0.32
LF gain, cm·s−1·mmHg−1	1.32 ± 0.32	1.09 ± 0.18
VLF gain, %/%	1.45 ± 0.78	1.34 ± 0.55
LF gain, %/%	1.67 ± 0.54	1.50 ± 0.42
Driven Oscillations
0.05 Hz MAP power, mmHg2/Hz	944 ± 649*	15,799 ± 7441*‡
0.10 Hz MAP power, mmHg2/Hz	727 ± 480†	8,406 ± 3287†‡
0.05 Hz MCAv power, (cm/s)2/Hz	412 ± 247*	3,552 ± 2492*‡
0.10 Hz MCAv power (cm/s)2/Hz	501 ± 332†	9,001 ± 4649†‡
0.05 Hz coherence	0.84 ± 0.11*	0.97 ± 0.02*‡
0.10 Hz coherence	0.96 ± 0.04†	1.00 ± 0.00†‡
0.05 Hz phase, radians	1.61 ± 0.52	1.23 ± 0.49‡
0.10 Hz phase, radians	0.61 ± 0.25	0.73 ± 0.18†‡
0.05 Hz gain, cm·s−1·mmHg−1	0.67 ± 0.29*	0.48 ± 0.17*
0.10 Hz gain, cm·s−1·mmHg−1	0.82 ± 0.22†	1.02 ± 0.16
0.05 Hz gain, %/%	0.69 ± 0.21*	0.65 ± 0.28*
0.10 Hz gain, %/%	1.10 ± 0.34	1.34 ± 0.37

Values are means ± SD.

MAP, mean arterial pressure; MCAv, middle cerebral artery velocity; LF, low frequency; VLF, very low frequency.

Statistical significance was set at P < 0.05.

^*Significantly different from VLF;

^†significantly different from LF;

^‡significantly different from rest.

Table 5.

Transfer function analysis metrics in older adults for the relationship between mean arterial pressure and cerebral blood flow at rest and during moderate exercise

	Rest	Exercise
Spontaneous Oscillations
VLF MAP power, mmHg2	4.86 ± 3.16	8.58 ± 17.30
LF MAP power, mmHg2	1.98 ± 1.04	2.31 ± 1.76
VLF MCAv power, (cm/s)2	7.95 ± 6.36	4.54 ± 2.76
LF MCAv power, (cm/s)2	1.83 ± 0.72	2.11 ± 1.53
VLF coherence	0.37 ± 0.13	0.40 ± 0.11
LF coherence	0.61 ± 0.18	0.48 ± 0.19
VLF phase, radians	1.03 ± 0.45	1.47 ± 0.48
LF phase, radians	0.62 ± 0.24	0.70 ± 0.19
VLF gain, cm·s−1·mmHg−1	1.05 ± 0.54	0.95 ± 0.54
LF gain, cm·s−1·mmHg−1	1.08 ± 0.35	0.99 ± 0.47
VLF gain, %/%	1.90 ± 1.08	1.89 ± 1.44
LF gain, %/%	1.92 ± 0.49	1.74 ± 0.74
Driven Oscillations
0.05 Hz MAP power, mmHg2/Hz	1,845 ± 2,178*	7,685 ± 7,247*‡
0.10 Hz MAP power, mmHg2/Hz	1,311 ± 2,087†	4,616 ± 3,169†‡
0.05 Hz MCAv power, (cm/s)2/Hz	722 ± 1,070*	1,668 ± 1,114*‡
0.10 Hz MCAv power, (cm/s)2/Hz	786 ± 1,370†	3,665 ± 2,501†‡
0.05 Hz coherence	0.90 ± 0.10*	0.98 ± 0.02*‡
0.10 Hz coherence	0.89 ± 0.13†	0.98 ± 0.02†‡
0.05 Hz phase, radians	0.99 ± 0.35	1.35 ± 0.32‡
0.10 Hz phase, radians	0.68 ± 0.41	0.46 ± 0.27†‡
0.05 Hz gain, cm·s−1·mmHg−1	0.61 ± 0.15*	0.59 ± 0.24*
0.10 Hz gain, cm·s−1·mmHg−1	0.76 ± 0.45	0.91 ± 0.27
0.05 Hz gain, %/%	1.07 ± 0.32*	1.11 ± 0.44*
0.10 Hz gain %/%	1.32 ± 0.75	1.64 ± 0.54

Values are means ± SD.

MAP, mean arterial pressure; MCAv, middle cerebral artery velocity; LF, low frequency; VLF, very low frequency.

Statistical significance was set at P < 0.05.

^*Significantly different from VLF;

^†significantly different from LF;

^‡significantly different from rest.

Compared with the oscillations present in the spontaneous supine data, the enhanced blood pressure variability as a result of the OLBNP maneuvers during moderate exercise resulted in a greater than 300-fold increase in the mean arterial pressure power and a greater than 100-fold increase in the middle cerebral artery velocity power in both the younger and older adults, with no differences present in the coherence levels in the driven oscillations (P = 0.432; Fig. 2). The enhanced input and output signals at both rest and during moderate exercise were reflected in a main effect of frequency on coherence (rest: ∼0.75–0.90, P < 0.001; exercise: >0.97, P < 0.001; Fig. 3). There was also a significant modality times frequency interaction effect for coherence (P < 0.001), which was reflected in an average within-subject coherence elevation from ∼0.30–0.65 to >0.97 during exercise (compared with the same frequency at rest) in both younger (Table 4) and older adults (Table 5).

Fig. 2.

Absolute values of the power spectrum densities (PSD) for the mean arterial pressure [MAP; younger adults (A), older adults (C)]; and cerebral blood velocity [CBV; younger adults (B), older adults (D)] with spontaneous and driven oscillations in blood pressure during moderate exercise. The frequency where the PSD reached peak amplitude (either 0.05 or 0.10 Hz) was used as a basis for sampling the point estimates for coherence, phase and gain values. Note the minimum >100-fold increase in PSD for the driven (0.05 and 0.10 Hz) compared with the spontaneous oscillations.

Fig. 3.

The absolute (left) and relative (right) delta changes from resting to exercise conditions in coherence, phase, absolute gain, and normalized gain during driven oscillations in younger and older adults. Note that there was a consistent elevation in coherence and 0.10 Hz gain metrics in all individuals with driven oscillations during exercise. There were age-related differences in the expression of the phase metric at both 0.05 and 0.10 Hz. *Differences between younger and older adults.

Phase and gain.

There were main effects for frequency on the phase, absolute gain, and normalized gain metrics (P < 0.001). During spontaneous oscillations where coherence was generally low (<0.50), the phase and gain metrics in the younger adults followed the expected high-pass filter model of the cerebrovasculature (Fig. 4). Although the older adults had a paradoxical response, i.e., the phase response followed the expected trend of a decreasing lead as frequency approached 0.20 Hz, the absolute gain and normalized gain were maintained (instead of increasing as expected) with increasing frequency (post hoc Tukey results: absolute gain VLF vs. LF, P = 0.634; normalized gain VLF vs. LF, P = 0.991). These findings were consistent under both resting and exercise conditions as assessed with spontaneous oscillations.

Fig. 4.

The phase [younger adults (A); older adults (B)] and normalized gain [younger adults (C); older adults (D)] during moderate exercise. Frequency band sampling was used for spontaneous oscillation data (●), and point estimates were sampled for the driven data (0.05 and 0.10 Hz). Note the high-pass filter nature of the cerebrovasculature under all frequencies for phase (regardless of age) and the younger adults for normalized gain. *Differences between younger and older adults. The mean spontaneous oscillation normalized gain data for the older adults is relatively unchanged throughout the 0.00–0.20 Hz range, whereas the driven data demonstrates the expected increases with increasing frequency at the point estimates. VLF, very low frequency; LF, low frequency.

The increased coherence values for the driven oscillations (vs. spontaneous oscillations) during exercise (Fig. 5) reflect an increase in the signal-to-noise ratio of the system and thus result in the phase and gain values being more mathematically interpretable (22, 41, 43, 44). With the improved signal-to-noise ratio of the cerebral pressure-flow relationship during the OLBNP maneuvers during the exercise trials, the point estimates for the phase and gain metrics were revealed to follow the expected trends of the high-pass filter model in both younger and older adults (P < 0.001; Fig. 6).

Fig. 5.

Augmentation in population mean coherence to ∼1.00 during the driven oscillations in blood pressure during moderate exercise in younger (A) and older (B) adults. Note the relatively low coherence (<0.60) in the spontaneous data (●) for both the younger and older adults. Regardless of age, the coherence point estimate of the driven oscillations (0.05 or 0.10 Hz) was ∼1.00, indicating that the cerebral pressure-flow relationship is able to be approximated in a linear fashion at the frequencies of interest, which greatly enhances the interpretation of the associated transfer function metrics. For standard deviation information please refer to Tables 4 and 5.

Fig. 6.

The transfer function analysis of coherence (top), phase, absolute gain, and normalized gain at the driven frequencies of 0.05 Hz (left) and 0.10 Hz (right) for the younger and older adults during moderate exercise. Note the high-pass filter nature of the phase and gain responses for the younger and older adults. *Differences between younger and older adults.

Compared with the driven oscillations at rest, the exercise data during the OLBNP maneuvers revealed within-subject changes to the TFA phase metric that appear to be age related (significant 3-way interaction effect; P = 0.028; Fig. 3). During exercise the younger adults phase was lower −26.5 ± 24.1% (∼0.40 radian) than resting data at 0.05 Hz, whereas the older adults had an elevated phase by 27.6 ± 36.1% (∼0.35 radian). In contrast at 0.10 Hz, the younger adults phase was higher 35.1 ± 49.7% (∼0.15 radian) during exercise, while the older adults had a reduced phase of −20.7 ± 42.9% (∼0.25 radian). Despite the absolute and normalized gain metrics being elevated in all subjects at 0.10 Hz during exercise, there were no within-subject age-related changes in the absolute or normalized gain metrics between rest and exercise OLBNP trials (Fig. 3).

Effects of Aging on TFA Metrics

There was a significant main effect for age group on the normalized gain metric (P = 0.001). The augmented signal-to-noise ratio during the OLBNP protocol revealed that the older adults had a normalized gain +71% elevated at 0.05 Hz (+0.46 ± 0.36%/mmHg) compared with the younger adults. There was also a significant three-way interaction for age group times modality times frequency for the phase metric (P = 0.028). It was noted that phase was reduced −31% at 0.10 Hz (−0.24 ± 0.22 radian: Fig. 3).

DISCUSSION

Using the novel approach to drive blood pressure during a supine cycling intervention, the main findings of the study were: 1) oscillating blood pressure during moderate exercise resulted in coherence values approaching 1.00; 2) the paradoxical nature of the cerebral pressure-flow response in older adults under spontaneous oscillations was normalized with improvements to coherence; 3) during the driven blood pressure oscillations, the older adults had slightly elevated normalized gain at 0.05 Hz and reduced phase at 0.10 Hz, compared with the younger adults; and 4) the comparison between resting and exercise OLBNP maneuvers revealed subtle changes to the TFA phase metric. Collectively, these findings support our hypothesis and for the first time conclusively demonstrate that, despite the decrease in CBV within the older adult population, the high-pass filter model of the cerebrovasculature is present during moderate supine cycling exercise. Despite the intact nature of the high-pass filter model, these findings also reveal subtle age-related changes in the TFA phase metric and how it is expressed during moderate exercise.

Comparison to Previous Studies

The previous studies that have examined the relationship between arterial blood pressure and CBV speculated that their findings were representative of an intact response (4, 6, 10, 16, 29, 31). Only two of these studies reported both the VLF and LF data for phase and gain (16, 29), whereas the other studies based their interpretation on phase and gain values strictly in the LF range (4, 6, 10, 31). The two studies that did report both frequency ranges of interest noted paradoxical findings in both younger and middle-aged adults (Table 1). For example, the phase and gain values both responded in a similar direction with increasing frequency, which is counter to the high-pass filter nature of the cerebrovasculature. The collective findings from these studies are similar to the spontaneous data of the older adults within the current study (Table 3). These supposedly impaired cerebral autoregulatory results during moderate exercise (elevated LF phase) were speculated, in an earlier study by Ogoh et al. (30), to occur as a result of an increase in systemic metabolites and/or excess ammonia released from the skeletal muscles crossing the blood-brain barrier and disrupting the regulatory properties of the cerebrovasculature. A second explanation was presented by Ainslie et al. (6), who noted that an elevated LF gain response may be due to capillary overperfusion occurring during exercise. A final explanation for these findings was presented by Fisher et al. (16), who noted that the low coherence values (as highlighted in Table 1) associated with the cerebral pressure-flow response may impact the interpretation of these findings. This final notion has recently been confirmed (under resting conditions) as TFA metrics associated with spontaneous oscillations have been demonstrated to have poor reproducibility, and thus findings associated with this method should be interpreted with caution (42). Our current study largely validates this latter concept (16), as the spontaneous coherence values associated with moderate steady-state exercise were on average below 0.50 in both the VLF and LF ranges regardless of age. When the OLBNP was applied during exercise, the associated oscillations in blood pressure were more similar to those typically observed during squat-stand maneuvers (which has recently been demonstrated to be the gold standard for assessing the cerebral pressure-flow relationship with TFA), than OLBNP at rest (42). This is likely a result of the elevated cardiac output during exercise enabling a greater volume of blood to be pooled in the venous system of the lower body during the OLBNP maneuvers. With these enhanced oscillations in blood pressure, it was revealed that in the relationship between arterial blood pressure and cerebral blood flow the signal-to-nose ratio was augmented (coherence >0.97 for all ages and frequencies; Tables 4 and 5), thus increasing the interpretability of the associated phase and gain metrics. However, these large transient swings in blood pressure induced with OLBNP did not alter the 5-min averages for mean arterial pressure, CBV, heart rate, or Pet_CO2 in either the younger or older adults at rest or during exercise (Tables 2 and 3). This finding likely indicates that these maneuvers did not have a major impact on the physiological interactions and responses to exercise, and the oscillations were merely used to enhance the TFA coherence and thus the interpretability of the phase and gain metrics (42). With this novel methodological approach, examination of the cerebral pressure-flow relationship during moderate exercise revealed that the cerebrovascular system is indeed functioning as a high-pass filter with gain increasing and phase decreasing as frequency increases from 0.02 to 0.20 Hz. Thus, although some of the previous research in this field speculated that the regulatory properties of the cerebrovascular were intact during moderate exercise, this is the first study to conclusively demonstrate this notion.

Effect of Aging on the Cerebrovasculature During Exercise

There is a general reduction in cerebral blood flow of ∼0.8%/yr after the age of 20 (5). It has been proposed that this reduction occurs due to increases in CVRi (39), which could be a result of reductions in brain mass (37). The current study observed similar alterations to CVRi and cerebral blood flow as noted elsewhere (5, 41). There was an age range of ∼35–40 yr between the groups in the current study, with a 20–25% reduction in CBV in the middle cerebral artery and ∼45% increase in CVRi noted in the older adults (Tables 2 and 3). When OLBNP maneuvers were performed during moderate exercise, it was revealed that, while the cerebrovasculature of both younger and older adults responded as a high-pass filter during moderate exercise, there were some phase and gain differences with aging. Namely, it was revealed that the older adults' normalized gain at 0.05 Hz was elevated (indicating a greater transfer of the blood pressure oscillation to the cerebrovasculature) and phase was reduced at 0.10 Hz (indicating a reduced timing buffering capacity), compared with the younger adults (Fig. 3). These findings are in contrast to the spontaneous data within the current study and those reported by Fisher et al. (16). Previous research has shown that, under resting conditions (3, 24, 45, 50) and during moderate exercise (16) with spontaneous oscillations, it appeared as though there were no differences in TFA metrics across the aging spectrum in individuals up to ∼70 yr old.

However, the enhanced coherence in the current study has revealed that the cerebrovasculature of younger and older adults may have slight differences in the TFA patterns, despite maintaining the high-pass filter nature of cerebral blood flow regulation. The findings from the current study demonstrated subtle age-related changes in the expression of the TFA phase metric between resting and moderate exercise conditions. In the exercise trials, the younger adults had a reduced phase at 0.05 Hz and elevated phase at 0.10 Hz compared with the same frequency at rest. In contrast, the older adults had an augmented phase at 0.05 Hz and a decreased phase at 0.10 Hz (Fig. 3). It is unlikely the cardiac baroreflex is responsible for the difference in findings and expression of the TFA metrics between younger and older adults, brcause research under resting conditions has demonstrated that the baroreflex plays a minimal role in the cerebral pressure-flow response (41, 53). Instead, it is more likely that the increase in CVRi with aging (Tables 2 and 3) leads to alterations in either arteriolar tone (44), sympathetic tone (10, 16), and/or changes within the mechanical buffering of the compliance vessels (48). Any of these mechanisms could affect the cerebral pressure-flow response during moderate exercise (exacerbating the age-related differences). However, more research in this area is needed to confirm these possibilities.

It is also highly improbable that just a single mechanism (arteriolar tone, sympathetic tone, downstream capacitance) is responsible for the age-related differences that were observed during exercise (Figs. 3 and 6). However, related to vascular compliance, it is possible that the previously mentioned mechanisms could be mediated by nitric oxide, because it has been shown to have differentiated effects across the aging spectrum (21). For instance, in younger adults (mean age 25 yr), infusion of a nitric oxide inhibitor (l-NMMA) has little to no impact on cerebral blood flow (but does increase mean arterial pressure), indicating that the nitric oxide pathway is not vital in controlling cerebral blood flow in young adults (21). Contrasting this finding are results from an older population (mean age 78 yr) where the same nitric oxide blockade resulted in a decrease in cerebral blood flow by ∼22% and elevations in cerebrovascular resistance of ∼40% (21). The authors speculated that these findings confirm that the nitric oxide pathway is important in the control of cerebral blood flow, but only in an elderly population where endothelial dysfunction and/or atherosclerosis may be present (21). Although animal studies indicate that exercise induces increases in nitric oxide levels (30–40%) in the medial prefrontal cortex and stria terminalis (12), it is unknown whether this occurs in humans or is altered with aging.

Mechanisms for Maintaining the Cerebrovasculature High-Pass Filter During Exercise

During moderate exercise there is a significant increase in mean arterial pressure that challenges the cerebrovasculature to effectively regulate against overperfusion and increases to intracranial pressure. Because cerebral hemorrhages do not occur during routine exercise, the cerebrovasculature must have several regulatory protective mechanisms. In baboons, it has been shown that there is a smaller redistribution of cardiac output to the brain compared with the skeletal muscle, (20) which helps partially protect the brain. Hohimer et al. (20) concluded that the relatively constant level of cerebral blood flow at rest and during moderate exercise was due to the CVRi increasing ∼20% during exercise. This increase in the CVRi has been proposed to be due to enhanced sympathetic tone, which could partially explain the observed TFA metric changes with aging (10, 16). It has also been speculated that the cardiac baroreceptor-mediated control of blood pressure variability may play a significant role in the regulation of arterial blood pressure during exercise and thus by extension regulate cerebral blood flow (31). However, in more recent publications, it seems that the cardiac baroreceptors play a minimal role in the regulation of the cerebral pressure-flow relationship during healthy aging (53) or following long-term heart transplantation (41). The roles of myogenic tone (46, 49), increased capacitance (via Windkessel modeling; Ref. 48), and enhanced sympathetic activity are likely the main mechanisms responsible for the regulation of cerebral blood flow during exercise. These potential mechanisms could possibly be mediated by the exercise-induced release of nitric oxide, which may explain the phase differences observed between the younger and older adults in the current study (Figs. 3 and 6). More research in this area is needed to confirm these speculations.

Implications for the Assessment and Interpretation of Pressure-Flow Relationships

The previous findings on the relationship between mean arterial pressure and cerebral blood flow have highlighted the issue with using a linear assessment tool (TFA) to measure a system that is reliant on coherence for interpretation (22, 41–44). A low coherence value can mean a number of things: 1) there is a poor signal-to-noise ratio, 2) there are multiple inputs to the system, 3) the system is inherently nonlinear, and 4) there is simply no relationship between the input and output variables (54). This study addressed the concern noted by Fisher et al. (16) that the low coherence values during moderate exercise with spontaneous oscillations in blood pressure make interpretation of the findings difficult. Oscillating blood pressure during exercise, the novel design of this study, enhanced the Fourier transform power spectrums of both mean arterial pressure (∼2,000 mmHg²/Hz) and middle cerebral artery velocity [∼500 (cm/s)²/Hz] above that which is typically observed during OLBNP (9, 48) to levels typically observed during squat-stand maneuvers (41–44). The enhanced autospectra input to the system is reflected by a greatly increased output autospectra (Fig. 2), which maximizes the signal-to-noise ratio and results in a coherence approaching 1.00 (Fig. 5), producing much more interpretable phase and gain metrics in both younger and older adults alike (Figs. 4 and 6) (42). Thus future studies assessing the cerebral pressure-flow response during exercise should employ a methodology that enhances the coherence through oscillations in blood pressure (42).

Methodological Considerations

Flow vs. velocity.

Cerebral blood flow was indexed in this study through the application of transcranial Doppler ultrasound. This method utilizes the assumption that, when the diameter of the insonated vessel is constant, the velocity of the red blood cells within the vessel approximates the flow within that vessel. This concept has been recently reviewed by Ainslie and Hoiland (7), where they summarized the findings of two recent high-resolution magnetic resonance imaging studies of the cerebrovasculature (15, 51). It was shown that the diameter of the middle cerebral artery is relatively constant when Pet_CO2 values are held within 8 mmHg of eucapnia. In the current study, Pet_CO2 had a range of 2–3 mmHg around eucapnia, which indicates that the velocity data are representative of cerebral blood flow and can be utilized to assess the cerebral pressure-flow relationship.

Cerebral autoregulation.

Due to the myriad of unknown variables that are suspected to be responsible for cerebral autoregulation, the focus of this paper was not to discuss the black box concept of cerebral autoregulation but to present the findings as they concern the relationship between mean arterial pressure and CBV. The complexities associated with the all-encompassing concept of cerebral autoregulation are likely (at least to an extent) nonlinear in nature. However, most nonlinear systems can still be approximated within a linear model (42), as is the case with the findings of the current study. The coherence within the current study (mean range of 0.97–1.00) clearly indicates that, when the appropriate methodology is applied during exercise (OLBNP), the interpretability of the associated transfer function metrics is greatly enhanced (42).

In conclusion, this is the first study to conclusively demonstrate that the high-pass filter regulatory properties of the cerebrovascular are intact during moderate exercise in healthy aging. This was demonstrated through the novel approach of oscillating blood pressure during a moderate steady-state exercise with lower body negative pressure. The cerebral pressure-flow system was nearly linear in nature as indicated by a coherence of ∼1.00. The increased signal-to-noise ratio resulted in an enhanced interpretability of the phase and gain metrics. With this novel methodology, it was also revealed that, despite maintaining the high-pass filter nature of the cerebrovasculature, the older adults had an elevated normalized gain at 0.05 Hz and reduced phase at 0.10 Hz when compared with the younger adults during exercise. A within-subject comparison from rest revealed age-related differences in the expression of the phase metric at both 0.05 and 0.10 Hz. It is speculated that these age-related changes to TFA metrics are due to either alterations in arteriolar tone or sympathetic tone, or to the mechanical buffering of the compliance vessels that are possibly influenced by the exercise-induced release of nitric oxide. However, more research in this area is required to confirm these notions.

GRANTS

P. N. Ainslie and the work conducted in this project were supported by a Canada Research Chair in Cerebrovascular Physiology and an Natural Sciences and Engineering Research Council of Canada ( NSERC) Discovery Grant. J. D. Smirl was supported by NSERC and Killam Predoctoral Fellowships.

DISCLOSURES

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

AUTHOR CONTRIBUTIONS

Author contributions: J.D.S. and P.N.A. conception and design of research; J.D.S., K.H., A.H., and P.N.A. performed experiments; J.D.S., Y.-C.T., and P.N.A. analyzed data; J.D.S., K.H., Y.-C.T., A.H., and P.N.A. interpreted results of experiments; J.D.S. prepared figures; J.D.S. and P.N.A. drafted manuscript; J.D.S., K.H., Y.-C.T., A.H., and P.N.A. edited and revised manuscript; J.D.S., K.H., Y.-C.T., A.H., and P.N.A. approved final version of manuscript.