Cerebrovascular function and its association with systemic artery function and stiffness in older adults with and without mild cognitive impairment

This study included a comparison of adults with MCI (n = 41) and control participants (n = 33). Those with MCI were recruited through the NeuroExercise project (Devenney et al. 2017) at the German Sport University and were invited to undergo a series of additional cerebrovascular and cardiovascular assessments that form the basis of this cross-sectional investigation. Healthy adults of a similar age were recruited as control participants through local community advertisements. The study conformed to the Declaration of Helsinki (1975) and was approved by the institutional human research ethics committees. Participants provided written informed consent and were screened to ensure they met the inclusion criteria prior to participation.

MCI was classified according to established clinical criteria (Albert et al. 2011b), including the presence of memory decline without dementia (Clinical Dementia rating global score of 0.5), and a Montreal cognitive assessment (MoCA) score of 18–25 (Nasreddine et al. 2005). All MCI participants underwent cognitive assessment with a neuropsychology dementia specialist. Control participants did not report any memory impairments during an initial screening for subjective cognitive decline based on the framework by Jessen et al. (2014). Prospective participants were asked: (1) if they have any difficulty remembering routine tasks, (2) remembering things that happened recently, (3) whether others had expressed concerns regarding their memory, and (4) if they feel their memory is any worse than that of others of a similar age. Control participants achieved a score ≥ 26 on the MoCA during pre-study screening. MCI and control participants were aged between 65 and 85 years, had no history of prior cardiovascular events (i.e. no previous myocardial infarction or stroke), and prospective participants with hypertension (> 160/100 mmHg) that was not being treated were excluded.

After screening, all participants attended the laboratory on a single occasion to undergo a series of cerebrovascular and systemic vascular assessments. Participants refrained from alcohol and exercise for 24 h and caffeine for 12 h before the session and continued to take prescribed medication. Participants were fitted with instruments to measure middle cerebral artery blood flow velocity (MCAv), mean arterial blood pressure (MAP), end-tidal carbon dioxide (P_ETCO₂), and heart rate (HR) which were recorded throughout the study session. After resting supine measurements were obtained, cerebrovascular CO₂ reactivity was assessed using a breath-hold test. Participants then performed a repeated stand-to-sit test for the determination of cerebral pressure-flow responsiveness. Baroreflex sensitivity was also assessed during the repeated stand-to-sit procedure. Following this, participants underwent a further 15 min of supine rest prior to assessments of peripheral and central blood pressure, wave reflection characteristics, and carotid-femoral PWV. Finally, participants underwent a test of brachial artery FMD as a measure of systemic endothelial function. All assessments were conducted in a quiet, temperature controlled (23 ± 0.5 °C) laboratory.

Resting measurements were collected during a 5-min period of supine rest before the cerebrovascular function test. Cerebrovascular conductance was calculated as MCAv relative to MAP. Cerebrovascular pulsatility index (PI) was calculated using Gosling and King’s equation [PI = (V_s − V_d)/V_mean] (Gosling and King 1974), where V_s is the systolic velocity, V_d the diastolic velocity and V_mean the mean velocity.

Cerebrovascular reactivity was assessed as the MCAv response to changes in P_ETCO₂ that were induced with a repeated breath-hold test while in the supine position. The breath-hold test is a validated procedure to assess cerebrovascular function (Tancredi and Hoge 2013). After a period of paced breathing, participants held their breath for 20 s. A metronome was used as a guide for paced breathing at a rate of 16 breaths per min for 30 s before the next breath-hold started. Participants were instructed to give a small, forced exhalation at the end of each breath-hold. For each breath-hold manoeuvre, the increase in P_ETCO₂ (∆P_ETCO₂) was calculated by subtracting the average over the last two breaths before the breath-hold from the peak P_ETCO₂ response immediately after the breath-hold. The first two breath-hold manoeuvres were used to familiarise participants with the procedure (Tancredi and Hoge 2013). Breath-holds were analysed separately and a total of six breath-hold responses per person were averaged for comparison between groups (Tancredi and Hoge 2013; Murphy et al. 2012). Cerebrovascular reactivity was calculated as the increase in MCAv relative to the corresponding increase in P_ETCO₂ as absolute (∆MCAv/∆P_ETCO₂) and relative (%∆MCAv/∆P_ETCO₂) responses.

The repeated stand-to-sit transition test consisted of 13 stand-to-sit transitions in a 5-min period at a frequency of 0.05 Hz (10 s in the standing position and 10 s in the seated position). This same protocol has previously been used in older adults (van Beek et al. 2008; van Beek et al. 2010; Oudegeest-Sander et al. 2014; den Abeelen et al. 2014; Klein et al. 2020). The angle of the left knee was continuously measured with a bipolar sensor (Goniometer, MLTS700, ADInstrument, Bella Vista, NSW, Australia) to enable the alignment of data with each period of standing and sitting. During each repeated stand-to-sit transition, maximum values during the sit phase, and minimum (nadir) values during the stand phase, were identified for MCAv, MAP and P_ETCO₂. Whereas for HR responses, maximum values were identified during the stand phase and minimum values during each sit phase. For each transition, the response of each variable was calculated using the delta (sit − stand). The transition delta responses were also expressed as a percentage of the stand phase values (%∆), and the ratio of %∆MCAv/%∆MAP was calculated for each transition as previously reported (Brassard et al. 2017; Klein et al. 2020). Baroreflex sensitivity was calculated as the change from stand to sit in R-R interval relative to the change in systolic blood pressure (Xing et al. 2017). For each variable, responses during each of the 13 stand-to-sit transitions were averaged for further comparison between the groups.

MCAv was measured continuously using transcranial Doppler ultrasonography (TCD) where a 2 Hz probe was placed over the right temporal window and fixed at a constant angle with a customised headband (Multigon, Neurovision, Elmsford, N.Y., USA). The MCAv signal was identified according to standardized criteria guided by signal depth, velocity and wave characteristics (Aaslid et al. 1982; Willie et al. 2011), which remained constant for the testing session. If a right MCAv signal could not be clearly obtained, the left MCAv was used (n = 7). Heart rate was monitored by ECG and blood pressure was measured continuously at the left middle- or index-finger using photoplethysmography (Finometer MIDI, Finapres Medical Systems, Amsterdam, The Netherlands). Participants breathed through a leak-free respiratory mask (Hans-Rudolph, Kansas City, MO, USA), and expired air was continuously sampled for the determination of P_ETCO₂. All variables were sampled at 1000 Hz and stored (LabChart Pro v7.3.7 and PowerLab, ADInstruments, Bella Vista, NSW, Australia). Time-aligned signals were then resampled to second-by-second (1 Hz) for visual inspection and analysis.

Indices of peripheral and central blood pressure, wave reflection characteristics and pulse wave velocity were measured in the supine position using the SphygmoCor XCEL device (AtCor Medical, West Ryde, NSW, Australia). These tests have previously been described in detail (Perissiou et al. 2018) and were conducted in accordance with standardised guidelines (Townsend et al. 2015). Blood pressure was measured in triplicate on the right arm, with the first assessment discarded and the average of the last two used. Carotid-to-femoral PWV was obtained by measuring the right carotid pulse wave using applanation tonometry, and the femoral pulse wave using a pneumatic cuff (inflated to 80 mmHg) at the right mid-thigh. Once a regular and stable carotid pulse was detected, femoral pulse waves were collected simultaneously by inflation of the thigh cuff. An aortic pressure waveform was generated over 10 cardiac cycles, from which central systolic (cSBP) and diastolic pressures (cDBP), augmentation pressure (AP) and index (AIx) were calculated. Wave separation analysis was applied (SphygmoCor CVMS software, v.9) to determine the aortic forward (P_f) and backward (P_b) pressure waveforms, which were expressed as the wave reflection magnitude (RM; P_b/P_f × 100). This method assumes a triangular-shaped flow wave approximated from the estimated aortic pressure wave (Westerhof et al. 2006). The P_f and P_b pressure waves correspond to the peak and the end of the assumed flow wave, respectively. PWV was calculated as the distance (m) between the carotid and femoral sites multiplied by 0.8, divided by the time delay (s) between the pulse waves (Townsend et al. 2015). PWV was measured in duplicate, with an average recorded. If the measurement difference was > 0.5 m/s, a third measurement was performed and the median taken (Townsend et al. 2015; Huybrechts et al. 2011).

Brachial artery FMD was performed in line with technical recommendations described elsewhere (Thijssen et al. 2011; Black et al. 2008) and in accordance with previous methods described by our group (Bailey et al. 2018, 2017). FMD was performed with participants in the supine position, on the right arm with the cuff placed distal to the olecranon process. High-resolution duplex ultrasound with a 12-MHz multi-frequency linear array probe (T3000; Terason, Burlington, MA) was used to image the brachial artery in the distal third of the upper arm and to simultaneously record the longitudinal B-mode image and Doppler blood velocity trace. The Doppler angle of insonation was maintained at 60°. Images were optimised, and the settings (depth, focus position and gain) were maintained during each assessment. Following a 60-s recording period of diameter and velocity, the cuff was rapidly inflated (220 mmHg) and maintained for 5 min (D.E. Hokanson, Bellevue, WA). Diameter and velocity recordings resumed 30 s prior to rapid cuff deflation (< 2 s) and continued for 3 min thereafter. Analysis of brachial artery diameter was performed using custom-designed edge-detection and wall-tracking software, which is largely independent of investigator bias. Detailed descriptions of the analysis approach are described elsewhere (Thijssen et al. 2011; Black et al. 2008). FMD was calculated as [(peak diameter − baseline diameter)/baseline diameter] and expressed as a percent change in vessel diameter. FMD% was also corrected for variability in baseline diameter between groups, and expressed as scaled FMD (Atkinson et al. 2013). From synchronised diameter and velocity data, blood flow (the product of lumen cross-sectional area and Doppler velocity) was calculated at 30 Hz. Shear rate was calculated as 8 times mean blood velocity/vessel diameter (expressed as s⁻¹). Our laboratory has previously reported a between and within day coefficient of variation (CV) for FMD% of < 10% (Bailey et al. 2018). Consistent with recommendations for scaling FMD (Atkinson et al. 2013) and to account for the influence of baseline artery diameter (Atkinson 2013), FMD% was also calculated by allometric scaling of logarithmically transformed absolute diameter change [difference between peak artery diameter and baseline diameter (in mm)]. The logged absolute diameter changes were then back-transformed and interpreted in the conventional manner to obtain allometrically scaled FMD (percent diameter change).

Based on the mean effect extracted from a pooled dataset of > 500 MCI and control participants across multiple studies (Roher et al. 2011; Shim et al. 2015; Vinciguerra et al. 2019) this study was powered (alpha 0.05, beta 0.90) to detect a mean difference of 0.22 ± 0.20 in the primary variable of interest (cerebrovascular PI) between MCI and control. Gaussian distribution and homogeneity of variance of the data were confirmed using Shapiro–Wilk and Levene tests. Comparisons of cerebrovascular and systemic vascular function assessments were made between groups (MCI v CON) using independent t tests. To account for the potential influence of age on vascular outcomes, an ANCOVA was used to further analyse differences in key outcomes of interest between MCI and CON. Binary hierarchical logistic regression analysis was employed to examine whether the systemic (PWV and FMD) and cerebral (%MCA/%MAP and PI) measures of interest were able to discriminate between MCI and control. Age was entered at Step 1, followed by the systemic and cerebral variables of interest at Step 2. Large confidence intervals were observed for the odds ratio for PI. To address this, PI scores were subsequently categorised as high versus low (with values < 1 coded as low). However, treating PI as a categorical variable did not affect the results and thus raw values are reported.

Pearson’s correlation coefficients were used to examine the relationships between cerebrovascular and systemic function responses in the MCI and control groups. The effect size is reported as low if the value of r is around 0.1, medium if r ≥ 0.3, and large if r ≥ 0.5 (Cohen 2013). Statistical significance was set at p < 0.05. Statistical analyses were performed with Statistica 7.1. (StatSoft, Tulsa, USA).

A complete dataset (MCI: n = 41, CON: n = 33) was available for the analysis of MoCA scores and other participant characteristics, as well as the assessment of central blood pressure and augmentation index. Due to technical difficulties in obtaining reliable MCAv or finger blood pressure signals during the sit-to-stand test, there were seven missing cases from the MCI group (MCI: n = 34, CON: n = 33); and an additional four missing MCI cases for the cerebrovascular reactivity dataset because of difficulties performing the breath hold manoeuvre (MCI: n = 30, CON: n = 33). For the assessment of pulse wave velocity there were four missing cases from the MCI group and three missing cases from the CON group (MCI: n = 37, CON: n = 30) because a reliable 10 s carotid artery pulse wave signal could not be obtained. Flow mediated dilation could not be assessed in two MCI cases due to poor ultrasound image quality (MCI: n = 39, CON: n = 33). Analyses were carried out with all available data without substitution or imputation for missing cases.

Participant characteristics are presented in Table 1. Mean age of participants with MCI (75 years) was slightly greater than the control participants (71 years, p = 0.001); and the MCI group also had a significantly lower MoCA score (p < 0.001). The significant difference in MoCA scores between groups remained when accounting for age (p < 0.001). Resting heart rate, peripheral blood pressures, and BMI were all similar between groups. There was a difference in aspirin medication use between MCI and controls (p < 0.001), and higher beta blocker use in MCI (p = 0.06); but the use of other medications was similar between groups.

Cerebrovascular outcomes are presented in Table 2. Cerebrovascular PI was significantly higher in MCI compared to control (p = 0.03), but this effect was lost when accounting for age as a covariate (p = 0.35). Cerebral pressure-flow responsiveness and cerebrovascular reactivity to CO₂ are presented in Tables 2 and 3, respectively. Resting MCAv, P_ETCO₂ and MAP were not different between MCI and control groups. The MCI group showed a lower %∆MCAv/%∆MAP response than the control group (p = 0.048). When accounting for age the difference in the %∆MCAv/%∆MAP was lost (p = 0.08).

The change in P_ETCO₂ was similar between groups during the breath-hold test. While the change in MCAv was significantly higher in control compared to MCI (p = 0.014), the peak MCAv also tended to be higher in control compared to MCI, but this did not reach statistical significance (p = 0.064). The cerebrovascular conductance (CVC) reactivity to CO₂ (p = 0.761) and relative cerebrovascular reactivity (p = 0.341) was similar between MCI and control groups. Findings for cerebrovascular reactivity were unaffected when accounting for age as a covariate (p = 0.75).

Indices of central blood pressure, wave reflection characteristics, pulse wave velocity and vascular function are presented in Table 4. Similar to the findings for peripheral blood pressure, there were no significant differences in central blood pressure or wave reflection characteristics between MCI and control groups. PWV was significantly higher in MCI compared with control (p = 0.002). Brachial artery FMD% was lower in MCI compared to control (p = 0.030). When accounting for age as a covariate, the significant differences in both PWV (p = 0.004) and FMD% (p = 0.02) remained. Across both cerebrovascular and systemic vascular outcomes, we observed similar differences between MCI and control groups when the male and female cohorts were considered separately (Refer to Online Resource: Supplementary Table 1). While some of the group differences were no longer statistically significant, the magnitude of the differences in means were similar to the effects reported in the full cohort.

Correlations between MoCA scores and the cerebral and systemic outcomes are presented as an Online Resource (Supplementary Table 2). MoCA score was associated with FMD in the pooled cohort (n = 71, r = 0.28, p = 0.017). Cerebrovascular PI was associated with systolic blood pressure in the pooled cohort (n = 67, r = 0.43, p < 0.001), and this association was stronger in MCI (n = 34, r = 0.46, p < 0.01) than in control (n = 33, r = 0.36, p < 0.05). Cerebrovascular PI was also negatively associated with diastolic MCAv at rest when examining the pooled cohort (n = 67, r = − 0.49, p < 0.01). These findings are shown as an Online Resource (Supplementary Fig. 1). When considered separately, this association remained significant in control (n = 33, r = − 0.53, p < 0.01) and in MCI (n = 34, r = − 0.39, p = 0.022). We found no significant associations between PWV and cerebrovascular parameters in MCI or control groups (p > 0.05). FMD% was positively associated with PI in the pooled cohort (n = 67, r = 0.28, p = 0.022), and this association was strong in MCI (n = 34, r = 0.56, p < 0.001), but not in control (n = 33, r = 0.23, p = 0.19).

In the present study, cognitive status of healthy older adults and individuals with MCI was confirmed by using predefined MoCA cut-off scores (Nasreddine et al. 2005). Furthermore, individuals with MCI were recruited through the NeuroExercise project and had an established diagnosis of MCI according to the Albert et al. criteria (Devenney et al. 2017; Albert et al. 2011b). Even though the MoCA has high sensitivity (84%) and specificity (79%) to discriminate between healthy older adults and individuals with MCI (Roalf et al. 2013), it cannot fully exclude false positives and cannot be used to assess the severity of cognitive impairment. Therefore, future studies might apply a more extensive neuropsychological test battery, which would not only clarify the severity of cognitive impairment, but also allow for comparisons between different cognitive domains. Based on the screening procedures used it is unlikely that any of the control participants had dementia, although this possibility cannot be excluded as a clinical dementia rating assessment was not undertaken by the control group. Despite our attempts to recruit a healthy control group of similar age, we did have a small but significant difference (~ 3 years) between MCI and control in this study. However, both groups were over 65 years of age and all are considered “older adults”. Cerebrovascular outcomes have previously been shown to be similar between elderly (65–69 years) and older-elderly adults (74–86 years) (Oudegeest-Sander et al. 2014). There is evidence that cognitive impairment is more prevalent, and that it progresses at a faster rate in women compared with men (Lin et al. 2015; Sohn et al. 2018). The higher prevalence of cognitive impairment in women may be explained in-part by sex-related differences in cardiovascular risk factors and associated impairments in systemic and cerebrovascular function (Volgman et al. 2019). Our findings (see Online Resources, Supplementary Table 1) show that the differences in vascular function between people with and without MCI are likely to be generalisable to both males and females. We also found no differences between males and females for any of the measures of vascular and cerebrovascular function. However, this study was not powered for comparisons within or between the separate male and female cohorts and these comparisons should be verified in future studies with a larger sample size.

We did not measure physical activity levels or exercise capacity in all the individuals included in this study, and we cannot discount the possibility that potential differences in fitness levels between groups account for the differences in vascular outcomes we report. We have previously shown a positive association between cardiorespiratory fitness and MoCA score in adults with MCI (Stuckenschneider et al. 2018), and future research should aim to understand how cardiorespiratory fitness modifies cerebrovascular and systemic vascular function in adults with early cognitive decline. We deliberately excluded participants with prior cardiovascular events and untreated hypertension to maintain the homogeneity of the groups. It is recognised that cardiovascular disease exacerbates the risk of MCI and vascular dysfunction (Stefanidis et al. 2019, 2017). Based on the age and prescription-medication profile of the participants included in the present study, it is likely that some participants may have or are at risk of underlying cardiovascular disease. It was also apparent from some of the measures of spontaneous blood pressure, e.g. during the cerebrovascular assessments (see Supplementary Fig. 1), that blood pressure may have been poorly controlled in some participants, although mean blood pressure was not different between the MCI and CON groups. Future studies should confirm the present findings in those with and without a history of cardiovascular disease. To date, it is unclear which people with MCI progress to dementia, remain stable, or reverse to normal (Albert et al. 2011a; Roberts et al. 2014; Ganguli et al. 2011). Longitudinal prognostic studies are warranted to establish the role of cerebrovascular changes in the course of MCI and may help to identify which individuals are at the highest risk to progress to dementia. Consideration of intracranial cerebral blood flow dynamics is important for understanding cerebrovascular disease development. TCD is non-invasive and has high-temporal resolution, which allows for dynamic stimulus response measurement, and is often used in clinical populations. However, TCD does not measure changes in intracranial artery diameter, which have recently been shown to change during assessments where there are large changes arterial blood gases and blood pressure (Hoiland et al. 2019). Despite these limitations, TCD remains a very useful and reliable tool in the assessment of intracranial cerebral blood flow velocity.