The Assessment of Cerebrovascular Response to a Language Task from the Addenbrooke’s Cognitive Examination in Cognitive Impairment: A Feasibility Functional Transcranial Doppler Ultrasonography Study

Baseline demographics

55 participants were recruited to the study, of whom twenty-five were excluded due to: inadequate temporal windows (n = 13), visual impairment (n = 1), VaD diagnosis (n = 3), un-matched (n = 5) and misclassification (n = 3). There were no significant differences in age or sex of the excluded participants from those analyzed in the study. Mean age of the cohort was 73.2 (7.0) years, six (20%) were female, 29 (96.7%) Caucasian, and five (16.7%) were left-handed.

Baseline characteristics are listed for the total cohort and sub-groups in Table 1. Hypertension was more prevalent in the MCI group (60%, p = 0.036), and greater AChEI use was seen in the AD group (90%, p < 0.005). No other significant between-group differences were seen in demographic data (Table 1).

Study procedure

Volunteers who met the above criteria were invited to the Cerebral Haemodynamics in Ageing and Stroke Medicine Laboratory. Participants were requested to refrain from caffeine, alcohol, heavy meals, or strenuous exercise for at least 4 h prior to study commencement. All study procedures were undertaken in a quiet, temperature-controlled laboratory, free from distraction. Participants provided baseline information on age, sex, comorbidities, and medication use, or these were accessed from medical records. Handedness was assessed using the Edinburgh Handedness Inventory [20]. Right- and left-handed volunteers were suitable for study inclusion. Participants underwent continuous recording of bilateral CBFv (Viasys Companion III or DWL Doppler box) in the middle cerebral arteries (MCA), HR (3-lead ECG), continuous blood pressure (arterial volume clamping on non-dominant hand, Finometer, Finapres Medical Systems; Amsterdam, The Netherlands), and end-tidal CO₂ (ETCO₂, Capnography with nasal cannulae, Capnocheck Plus). TCD probes were secured in place using a headframe. Signals were sampled at 500 samples/s and stored in the data acquisition system (PHYSIDAS, Dept Medical Physics, University Hospitals of Leicester NHS Trust).

Initially, a 5-min baseline recording was taken at rest, followed by all tasks from the ACE-III, in the order that they would be undertaken in the clinical setting. There was a 1-min baseline between cognitive tasks to allow CBFv to return to baseline. The questions were divided into 3 sections and the equipment was paused between recordings to allow auto-calibration of the Finometer. The Finometer was calibrated prior to each recording using the brachial blood pressure. Question timings were marked manually with an event recorder. Each task was performed with one trial. The total assessment, including baseline data collection was approximately 2 h. A similar protocol has been published previously in a healthy cohort [21, 22].

Data were analyzed offline using software developed by this group previously [13, 23]. Data were visually inspected for large non-physiological spikes in CBFv (<100 ms), which were removed by linear interpolation. A median filter was applied to the CBFv signal to remove smaller spikes and all data were low-pass filtered at 20 Hz with a zero-phase Butterworth filter. Beat-to-beat mean CBFv, mean arterial pressure (MAP), ETCO₂, and HR were determined by the R-R interval (3-lead ECG). Data underwent standard polynomial interpolation and then re-sampling at 5 Hz in order to generate signals with a uniform time base [13, 23].

Signals were synchronized using the event marker, and population mean responses were calculated. Data are presented for the dominant and non-dominant hemispheres as peak percentage change in CBFv, MAP, HR, and ETCO₂ from a 20-s baseline prior to task initiation (T1). Two time points were selected to measure peak activation. Firstly, T2: the initial response to stimulation occurring 5 s after task initiation (mean peak percentage change in CBFv occurring at 25–30 s). Secondly, T3: the sustained response, occurring at 15 s after task initiation (mean peak percentage change in CBFv occurring at 30–40 s). This analysis has been published previously in a young, healthy cohort [21, 22]. In this communication of the initial results of the new approach, of coupling the traditional ACE-III assessment of patients with MCI and AD with NVC examination using TCD, we report only on one task: repeating words and phrases aloud. The reasons for this choice will be discussed below.

The dominant hemisphere was considered right side for left-handed individuals and left side for right-handed individuals, with the converse for the non-dominant hemispheres. The data were combined for the right and left hemispheres (left- and right-hand dominant, respectively) to produce a dataset for changes occurring in the dominant and non-dominant hemispheres as opposed to left- or right-hand dominance.

Statistical analysis

Data were tested for normality prior to analysis using the Shapiro-Wilks test. Continuous, normally distributed data are presented as mean (standard deviation), and non-normally distributed as median [inter-quartile range]. Nominal data are presented as number (percentage), with significance testing by chi-square analysis. Testing for statistical significance was by one-way analysis of variance (continuous, normally distributed), or Kruskall-Wallis (continuous, non-parametric). For non-parametric data, correlation analysis was performed using Spearman correlation. To investigate for the effect of hemispheric dominance and for an interaction between patient group and dominance, a two-way mixed ANOVA was used. Where an interaction effect was present, a univariate analysis was undertaken at all levels of the dependent and independent variable to identify main effects. Where a main effect was present for dominance, this was further investigated using paired t-testing within patient groups. Post hoc testing was by Bonferroni correction or Tukey for multiple comparisons as appropriate. Analyses were considered significant where p < 0.05. Data analyses were performed using SPSS version 22 for Windows. Graphs were produced using GraphPad Prism Version 7.0 for Windows.

Baseline data

There were no significant differences in CBFv in the dominant and non-dominant hemispheres between groups at baseline (Table 2). Similarly, there were no differences in peripheral measurements of MAP, HR, or ETCO₂ at baseline (Table 2).

Changes in peripheral parameters during cognitive testing

Heart rate

A significantly lower peak percentage change in HR in both MCI and AD participants compared to HC at T2 (HC: 8.8 (8.2)%, MCI: 0.7 (4.3)%, AD: –0.5 (5.6)%; p = 0.005) (Table 3, Fig. 1). HR differences between groups were not seen at T3 (Table 3, Fig. 1).

Table 3

Parameters at T2 and T3. Data shown are peak percentage change in cerebral blood flow velocity (CBFv), mean arterial pressure (MAP), heart rate and end-tidal CO₂ (ETCO₂) from a 20-s baseline, in healthy control, mild cognitive impairment, and Alzheimer’s dementia groups. Statistical testing by one-way ANOVA with post-hoc testing by Tukey

Parameter	n	Healthy controls	Alzheimer’s disease	Mild cognitive impairment	p
n		10	10	10
CBFv Non-dominant
T2	30	15.9 (7.5)	0.1 (7.1)	6.7 (4.5)	<0.005
T3	30	9.4 (8.8)	1.9 (8.3)	9.4 (7.3)	0.08
CBFv Dominant
T2	30	16.0 (7.5)	–0.6 (8.4)	6.0 (4.4)	<0.005
T3	30	9.4 (10.0)	–1.5 (5.8)	9.0 (9.2)	0.012
Heart rate
T2	30	8.8 (8.2)	–0.5 (5.6)	0.7 (4.3)	0.005
T3	30	7.3 (5.8)	2.1 (7.2)	1.6 (4.8)	0.08
MAP
T2	30	7.9 (4.6)	0.9 (4.4)	–0.1 (2.9)	<0.005
T3	30	6.1 (4.0)	2.3 (5.1)	3.7 (4.0)	0.16
ETCO2
T2	30	–0.9 (3.2)	–5.2 (5.7)	0.9 (3.2)	0.006
T3	30	–3.0 (7.8)	–5.5 (6.5)	–1.9 (2.5)	0.39

Fig.1

The temporal changes in MAP (a), heart rate (b), and ETCO₂ (c), to the language task. All signals were normalized by baseline values. Continuous line = healthy control, dotted line = mild cognitive impairment, interrupted line = Alzheimer’s disease. a) non-dominant hemisphere; b) dominant hemisphere. Task initiation occurred at 20 s.

Changes in cerebral hemodynamic parameters during cognitive testing

Temporal changes

Temporal changes in CBFv occurring in response to a language task in both non-dominant and dominant hemispheres are shown in Fig. 2.

Fig.2

The temporal changes in cerebral blood flow velocity (normalized % CBFv) to the language task. All signals were normalized by baseline values. Continuous line = healthy control, dotted line = mild cognitive impairment. interrupted line = Alzheimer’s disease. a) non-dominant hemisphere; b) dominant hemisphere. Task initiation occurred at 20 s.

Cognitive test score correlation analysis

Significant, positive correlations were present between peak percentage CBFv change and total ACE-III score for the whole cohort (HC, MCI, and AD combined) for non-dominant hemisphere and language task response at T2 (Table 4, Fig. 3). There were no significant correlations between total ACE-III score and the language task with peak percentage CBFv change at T3 (Table 4, Fig. 2). There were no significant correlations between language task score and peak percentage change in CBFv response (Table 4, Fig. 3).

Fig.3

Positive correlations between normalized percentage change in cerebral blood flow velocity (CBFv) and Addenbrooke’s Cognitive Examination III (ACE-III) score, a = non-dominant hemisphere, b = dominant hemisphere. Correlations by Spearman testing.

Hemispheric dominance analysis

There were no statistically significant effects of dominance or interaction between patient group and dominance for CBFv change to the language task on 2-way mixed ANOVA. In a further analysis, left-handed participants were removed, and data were re-analyzed just for right-handed participants (n = 25). There were no differences in baseline CBFv or peripheral parameter data, nor in peak percentage change. The only difference in statistical significance was that the correlation of non-dominant CBFv response at T2 with total ACE-III score became non-significant (p = 0.07).

Main findings

In summary, the use of a language task from the ACE-III to stimulate CBFv responses was both feasible and acceptable in healthy older adults, MCI, and AD. The protocol was well tolerated and acceptable, and good quality bilateral data were obtained for central and peripheral parameters in 30 participants. Although a power calculation was not performed, peak CBFv response to a language task showed significant differences in peak percentage change in CBFv between groups, however, this varied between hemispheres and between time points (i.e., T2 or T3). In addition, where these differences were present, they were also accompanied by changes in the peripheral parameters (MAP, HR, or ETCO₂). At T2, it is notable that the larger rise in CBFv was accompanied by a larger rise in both HR and MAP in the HC group, and a larger negative change in ETCO₂ in the AD group, which could account for the differences in CBFv change seen across the groups. Furthermore, there was a significantly lower stimulation in CBFv with the language task in the dominant hemisphere in the AD group at T3, which was not accompanied by any changes in HR, MAP, or ETCO₂. Therefore, the relationships between stimulation, changes in MAP, HR, and ETCO₂ are complex in interpreting their relative contributions to the CBFv change seen with task activation. There was significant correlation between CBFv change to the language task and total ACE-III score, but not within the language domain score, and only at T2. Finally, there was no effect of hemispheric dominance on the CBFv response to the language task.

Task selection

In previous communications, we described CBFv responses to the complete set of 20 ACE-III tasks in healthy controls [22, 24]. When examining the feasibility and sensitivity of this approach in patient populations however, we elected to focus on a single paradigm (repeating words and phrases aloud). There were several reasons for this. Firstly, it would be extremely difficult to report on all 20 paradigms with the same level of detail as in this paper, specifically describing the concomitant changes in systemic variables, including MAP, HR, and ETCO_2, that could influence the amplitude and temporal pattern of the CBFv response. Secondly, we purposely selected one of the tasks where the ACE-III score alone was not able to discriminate among the three groups of subjects (Table 1). The finding that significant differences in the CBFv response were obtained for this task (Fig. 2, Table 3) is encouraging and highlights the potential of this novel approach. Finally, we focused on a task involving speech as this presents particular feasibility challenges for TCD-based NVC studies, given the potential interference of speech with the Doppler signal quality and parallel changes in ETCO₂, as reflected by the temporal patterns of ETCO₂ in Fig. 1.

Despite the very promising results obtained for feasibility and sensitivity with the selected task, caution is needed at this stage regarding the potential performance for the other 19 paradigms. Indeed, further studies, and detailed data analyses, will be needed to identify those that would optimize the clinical benefits of adding TCD as an adjunct to the standard ACE-III examination.

Neurovascular coupling response

Despite demonstrating a difference in stimulation between groups, in this population, language was largely well preserved in the AD and MCI groups, with no differences on ACE-III score, Table 1. This may suggest that the differences in CBFv change to the language task across groups are more likely due to the changes in peripheral parameters, rather than stimulation. However, these results may also indicate that reduced CBFv response is a more sensitive marker than clinical cognitive testing and that subtle changes in CBFv may be apparent before clinical deficits are detectable. This is supported by the sustained differences in CBFv response at T3, despite a lack of change in the peripheral parameters at this time point.

Changes in peripheral parameters have previously been reported in a number of task activation studies, where sympathetic activation can result in rises in HR and MAP, and hyperventilation or breath holding can result in changes in ETCO₂ [14, 25]. Cardio-inhibitory effects, due to parasympathetic activation, of AChEIs have been documented previously [26, 27]. In this cohort, AChEI use was significantly higher among the AD group, and could account for the blunted HR response to the language task. A number of recent studies have demonstrated that changes in CBFv are likely to be as a result of changes occurring at the metabolic or myogenic level in response to cognitive [28–30], or motor [31] stimulation. This challenges the interpretation that CBFv changes are purely a reflection of changes in MAP [23].

Cognitive impairment and vessel function

Impaired cerebral blood flow occurs early in patients with both MCI and dementia [5, 6]. The two-hit hypothesis describes a cascade of pathological events, initiated and propagated by impaired CBF, resulting in progressive amyloid accumulation and tau hyperphosphorylation [5, 32]. Amyloid accumulation not only damages neurons directly, but also damages the structural integrity of blood vessels, and induces vasoconstriction, thus worsening cerebral hypoperfusion [5, 32]. Furthermore, amyloid accumulation results in blood-brain barrier dysfunction, protein leakage, and further neuronal toxicity [5, 32, 33]. Atherosclerosis in the circle of Willis is found more commonly among those with AD dementia, than non-AD dementia [32]. Despite these findings at a microvascular level, studies of vessel function in dementia and MCI remain mixed [6, 7, 9]. In a recent study, despite lower resting CBFv in dementia, cerebral autoregulation was better in dementia compared to healthy controls [34]. However, while dynamic cerebral autoregulation appears to remain intact, a number of meta-analyses have consistently shown lower resting flow velocities, higher vessel resistance, and lower vasoreactivity in cognitive impairment [6, 7, 9]. Fewer studies have examined NVC in dementia, and these are outlined in more detailed below. Thus, NVC may be a more useful maker of cognitive impairment than dynamic CBF autoregulation.

Results in context of existing literature

Few studies have employed TCD as a method by which to investigate NVC in dementia [6, 7, 9, 35]. Certainly, recent reviews and meta-analyses have consistently demonstrated lower resting CBF in both AD and VaD sub-types of dementia [6, 7, 9, 35, 36]. Furthermore, markers of vessel resistance (pulsatility index, cerebrovascular resistance index) are higher among those with both dementia sub-types and the ability of the cerebrovasculature to respond to changes in CO₂ is lost (vasoreactivity) [6, 7, 9, 36]. Furthermore, VaD and AD can be discriminated from one another on the basis of pulsatility index and vasoreactivity [7].

No previous studies have measured NVC in response to a language paradigm in cognitive impairment from the ACE-III by TCD. Two TCD studies examined CBFv responses in the posterior cerebral artery during visual stimulation, where those with VaD could be discriminated from AD and healthy controls [16, 37]. These findings suggest intact occipital function in early AD [16, 37]. The results presented here, suggest that CBFv responses to language stimulation may be detected at an earlier stage than visual stimulation, and therefore may be a better discriminator of cognitive impairment from healthy aging. In a study by Matteis et al., a verbal fluency task produced distinctly lateralized rises in CBFv in the healthy control group; however, bilateral rises were seen in both patient groups, with loss of hemisphere selectively in the CBFv response [17]. The bilateral rise in CBFv is likely to represent a compensatory mechanism to cope with rising cognitive demands resulting from declining cognitive ability [38, 39]. This phenomenon has also been demonstrated in older adults [38, 39], and those with MCI and early dementia [40]. In keeping with findings of Matteis et al. [17], Yeung et al. also found loss of hemispheric specialization, using a category fluency task [41]. Interestingly, this was detectable as early as the MCI stage, and was more pronounced in those with amnestic sub group MCI [41]. In contrast to these studies, we did not find a significant difference in hemispheric lateralization between or within groups. Although left handers might have a tendency to less lateralization in response to cognitive testing [14], the fact that our population included only two left handed participants in each group is unlikely to explain the lack of a significant hemispheric lateralization within groups. On analyzing data for right-handed participants only (n = 25), there were no differences in baseline CBFv or peripheral parameter data, nor in peak percentage change, indicating that left-handed individuals did not significantly affect the results of this feasibility study. In keeping with our findings here, a number of MRI, PET, and SPECT studies have demonstrated correlation between language deficits in MCI and AD, and regional hypoperfusion or hypometabolism [42–45]. In addition, recent fMRI studies report reduced activation during language processing in MCI and AD, as demonstrated here [44–47], but also compensatory hyperactivation in MCI groups [45, 48]. In an fMRI study by Vandenbulcke et al., amnestic MCI patients with preserved language function, demonstrated reduced activation in the posterior temporal lobes compared to controls during a word reading task [46]. Taken together, these results suggest that CBF responses to language tasks can be detected prior to the appearance of clinical deficits, and thus could provide an early imaging biomarker of cognitive decline.

Limitations

Measurement of resting flow velocities relies on the assumption that the vessel diameter remains constant and therefore CBFv can be measured to approximate for CBF [10]. This study reports a high window failure rate (∼24%), however this is consistent with other studies of cognitive impairment and older adults [7, 49]. The sample size studied here was relatively small, and results need to be interpreted in this context. In addition, diagnostic classification was found to be complex, we therefore employed a consensus approach to classifying participants in this study. Due to the window failure rate, a number of participants were not able to be matched on age and sex, and were therefore excluded from this analysis.

The MCA supplies approximately 80% of the cerebral cortex [14] and, while we can achieve high temporal resolution by TCD measured CBFv change [10, 14], we cannot map this regionally to determine specifically where these changes are occurring.

In terms of the population studied, there was a relatively higher window failure rate among females, and they are therefore under-represented in this study. As expected, the mean age was relatively high in this study, with more volunteers recruited with older onset cognitive impairment. The majority of the cohort were Caucasian, and few were left handed. The results may therefore not be generalizable to the wider population. In addition, for this study it was assumed that left handed individuals would lateralize to the right hemisphere in terms of dominance. However, this is not always the cases, with approximately one quarter of left handed individuals demonstrating atypical lateralization patterns [50]. A more complex classification of hemispheric dominance can instead be achieved through calculation of the laterality index based on relative hemispheric response to cognitive testing [51].

In previous studies of task activation, multiple cycles of activation have been utilized to improve signal-to-noise ratios [36], but we have previously demonstrated that peak changes in CBFv are obtainable with one cycle of task activation in healthy subjects, thus limiting accommodation of responses through repeated trials [22]. One important finding of the current study though, is the demonstration that a single cycle of activation is also feasible in patients with MCI and AD and leads to good quality measured responses. In this study, TCD alone was used, with no method of measuring neuronal activation directly, such as electroencephalography (EEG). Future studies can consider the combination of these techniques to measure flow and neuronal activity simultaneously.

Implications of the findings and further work

This study has demonstrated that TCD-measured task activation to a language task taken from a clinical assessment tool (ACE-III) is feasible in a cognitively impaired population. The protocol was acceptable to patients and we were able to achieve measures of change in CBFv, MAP, HR, and, ETCO₂. This study has highlighted a number of key areas that require further consideration for future studies of TCD in dementia.

Specifically, further studies with larger sample sizes, examining a range of paradigms within the sub-domains of the ACE-III are required to identify if these trends are seen across the different cognitive domains, or isolated to individual tasks. Future studies should ideally measure concomitant changes in MAP, HR, and ETCO₂ alongside CBFv response. The use of more complex dynamic multivariate modelling, examining the relative contributions of MAP, HR, ETCO₂, and stimulation to the total CBFv response, would allow for a more accurate and comprehensive assessment of NVC in TCD studies of task activation in dementia. Studies with a broader population, to examine changes in female and ethnic minority groups would allow results to be more generalizable to the wider population. The inclusion of participants with VaD would improve our understanding of the changes in NVC occurring comparatively in dementia sub-types.

Conclusions

There is a growing body of evidence that vascular dysfunction is an important pathological mechanism implicated in the development of AD, in addition to VaD. This warrants further investigation both as a biomarker of early disease, but also for novel therapeutic approaches. TCD measured CBFv changes are feasible in a cognitively impaired population and remains a promising approach to studying NVC changes associated with aging and dementia.