The Cognition and Flow Study: A Feasibility Randomized Controlled Trial of the Effects of Cognitive Training on Cerebral Blood Flow

Abstract

Background:

Cognitive training (CT) has demonstrated benefits for healthy older adults (HG) and mild cognitive impairment (MCI), but the effects on vascular function are unknown.

Objective:

This is a feasibility trial investigating the effects of CT on cerebral blood flow velocity (CBFv).

Methods:

Twenty HG, 24 with Alzheimer’s disease (AD), and 12 with MCI were randomized to 12 weeks of multi-domain CT or control. Outcomes included: cognition (Addenbrooke’s Cognitive Examination III), mood, quality of life (QoL), physical, and neurovascular function (transcranial Doppler ultrasonography measured task activation of CBFv responses). Data are presented as mean difference (MD) and 95% confidence interval (CI).

Results:

47 participants completed the trial. There were three dropouts from the training arm in the AD group, and one in the HG group. The intervention was acceptable and feasible to the majority of participants with a high completion rate (89%). The dropout rate was higher among participants with dementia. Few changes were identified on secondary analyses, but QoL was significantly improved in HG post-training (MD: 4.83 [95% CI: 1.13, 8.54]). CBFv response rate was not significantly different in HG (MD: 1.84 [95% CI: –4.81, 1.12]), but a significant increase was seen in the patient group (MD: 1.79 [95% CI: 0.005, 3.58]), requiring sample sizes of 56 and 84 participants respectively for a fully-powered trial.

Conclusion:

A 12-week CT program was acceptable and feasible in HG, AD, and MCI. CT may be associated with alterations in vascular physiology which require further investigation in an appropriately powered randomized controlled trial.

Keywords

Cerebral blood flow cerebral hemodynamics cognition cognitive impairment dementia

INTRODUCTION

The number of people living with dementia is expected to double by 2050 [1], but strategies to prevent and treat dementia remain limited [2]. Mild cognitive impairment (MCI) is characterized by subjective and objective impairments in cognition, but without impacting functional status [3]. Approximately 10–20% of people living with MCI will convert to dementia over five years, representing a high-risk group for the development of dementia [3].

Cognitive training is a potential intervention which aims to improve cognitive function and processing by training individuals through a guided program of repeated cognitive exercises that target specific cognitive domains [4]. Cognitive training is attractive owing to fewer side effects than drug therapies, and the ability to be delivered flexibly at home [5]. In healthy older adults, and those with MCI, cognitive training has promising effects on the improvement and maintenance of cognitive function [5, 6], but few studies have investigated the mechanisms by which these effects are mediated [7, 8]. Recent systematic reviews of neuroimaging outcomes have identified cognitive training induced changes in brain structure and function in healthy older adults, MCI, and early dementia [7, 8].

Vascular dysfunction is increasingly recognized as a key pathological mechanism in the development of multiple dementia sub-types, including Alzheimer’s disease (AD) [9 –11]. The Rotterdam study demonstrated that lower CBF in healthy older adults was predictive of future dementia risk [12], and higher resting CBF was associated with better cognitive performance [13, 14]. Furthermore, hypoperfusion has been shown to accelerate and compound AD-related disease pathology, resulting in hyperphosphorylation of tau, increased amyloid deposition, and reduced amyloid clearance [9]. Given the importance of vascular pathology in the development and progression of cognitive disorders, understanding how vascular physiology and pathology is moderated by cognitive interventions is important. Cognitive processes are highly resource intensive for the brain and require a constant blood supply to meet these energy and oxygen demands [15]. Thus, in order to develop and sustain neuroplasticity, modifications to vascular structure and function will be a key process [16, 17]. This could be mediated through increasing the number of capillaries (angiogenesis), the release of vascular endothelial growth factor, and improvements in the efficiency of cerebral autoregulation and neurovascular coupling processes [17, 18]. However, few studies have specifically examined the effects of cognitive training on cerebrovascular function, and these are largely confined to healthy populations [19].

Transcranial Doppler ultrasonography (TCD) is a non-invasive, ultrasound-based technique to measure CBF velocity (CBFv) [20]. TCD has excellent temporal resolution, measuring rapid changes that occur in CBFv, and thus can be used to assess the integrity of cerebrovascular function [20]. No study has previously investigated the effects of cognitive training on TCD-measured changes in cerebrovascular function. Therefore, the primary aim of this study was to identify the feasibility of a randomized controlled trial of cognitive training in healthy older adults, MCI, and AD. Secondary objectives were to investigate the effects of cognitive training on clinical outcomes and vascular physiology using TCD-measured cerebral hemodynamic outcomes.

METHODS

Recruitment

The protocol for this study has been published previously [21]. In brief, 20 healthy older adults, 12 participants with MCI, and 24 with AD were recruited from poster advertisement, Join Dementia Research and outpatient memory clinics at the Leicestershire Partnership Trust in the United Kingdom. Recruitment and follow-up was conducted between January 2019 and April 2020. The trial closed early due to the COVID-19 pandemic. MCI and AD were defined by the Petersen and NIA/AA 2011 criteria, respectively [22, 23]. Participants with MCI and AD were screened using the Montreal Cognitive Assessment (MoCA) to ensure participants with severe cognitive impairment (MoCA ≤9) were not enrolled. Inclusion criteria were: age over 50 years, with access to the internet, and free of any medical co-morbidity or medication that could adversely affect cognition. Participants were suitable for inclusion irrespective of anti-dementia medication use. Volunteers with well-controlled co-morbidities (i.e., hypertension, diabetes) were considered for inclusion. Exclusion criteria were any medical condition or medication that could adversely affect cognition (healthy older adults), and poorly controlled medical conditions, specifically: severe heart failure (ejection fraction < 20%), significant carotid artery stenosis (> 50%), severe respiratory disease, major stroke, pregnancy or lactation, inadequate bilateral TCD windows, enrolled on an interventional study, and no access to the internet. The study was amended to provide the option of a laptop loan to improve access to the study for participants without suitable technology, and the upper limit of the MoCA was removed to enroll participants with mild to moderate dementia. All volunteers provided written, informed consent, or personal consultee declaration where participants lacked capacity. The study had ethics approval from the Bradford-Leeds Research Ethics Committee (ref: YH/18/0396), and all procedures were conducted in accordance with the Declaration of Helsinki 1975. The trial was registered prospectively on Clinicaltrials.gov, ref: NCT03656107.

Experimental procedures

Participants were asked to refrain from caffeine, alcohol, heavy meals, or strenuous exercise for at least 4 h prior to the study. All TCD procedures were carried out in the Cerebral Haemodynamics in Ageing and Stroke Medicine research space, which is free from noise and distractions, and maintained at 24°C. Participants underwent continuous beat-to-beat monitoring of bilateral CBFv (DWL Doppler box) through insonation of the middle cerebral arteries, blood pressure (BP, Finometer, Finapres Medical Systems; Amsterdam, the Netherlands), heart rate (HR, three-lead ECG), and end-tidal CO₂ (ETCO₂, capnography, Capnocheck Plus). Initially, participants underwent a 5-min baseline recording at seated rest, followed by a task activation protocol lasting approximately 10–15 min. All measurements were made by the same investigator (LB) at baseline and follow-up. Participants were presented with five tasks in a standard order selected from the Addenbrooke’s Cognitive Examination III (ACE-III). Tasks covered five key cognitive domains: attention (serial subtraction), verbal fluency (naming words), language (repeating words and phrases), visuospatial (drawing a cube and infinity diagram), and memory (recalling three words learnt previously). The three words were presented at the beginning of the protocol and were recalled at the end of the tasks (approximately 10–15 min later).

There was 1 min rest between each task to allow CBFv levels to normalize prior to the next task. Each task was marked using an event recorder. A detailed description of the analysis and task activation protocol has been published previously [20]. Throughout, CBFv1 refers to the non-dominant hemisphere and CBFv2 to the dominant hemisphere. Hemispheric dominance was assessed by determining handedness using the Edinburgh Handedness Inventory [24]. Right-handed individuals were assigned as left hemisphere dominant, and left-handed individuals as right hemisphere dominant.

Outcome measures

In addition to an assessment of neurovascular function, participants also underwent assessments of cognitive function (full ACE-III), mood (15-item Geriatric Depression Scale [GDS]), quality of life (dementia quality of life [DEMQOL]), and activities of daily living (Lawton instrumental activities of daily living [IADL]). Data were collected on key baseline demographic variables, including medical and medication history. Following collection of baseline data, participants were randomized by Sealed Envelope© to either a 12-week cognitive training program or waiting list control. The randomization procedure was a stratified, permuted block randomization, with a block size of four. All participants were invited for a follow-up assessment at 12 weeks. The participants were randomized, enrolled, and assigned to the intervention by LB. Due to the nature of the intervention and outcome measures, participants and assessors could not be blinded to the intervention or outcomes. To mitigate this, the data collected on CBF was batch-blinded by an independent researcher prior to analysis.

Intervention: Cognitive training program

The intervention was a multi-domain, online cognitive training program provided by Lumosity©. The cognitive training program covered the five key cognitive domains assessed in the CBFv task-activation protocol (i.e., attention, language, fluency, visuospatial, and memory). Participants were requested to train for five, 30 min sessions per week and adherence was monitored by Lumosity© as the number of sessions trained per week. The frequency and number of sessions were determined from research undertaken previously by Lumosity© suggesting that a minimum of 20 h training time was required specifically for this program. Training for five 30 min sessions for 12 weeks would equate to a total of 30 h training, therefore mitigating any missed sessions. In addition, the training dose and duration for this program were informed by the results of a systematic review, which found that 30 min was determined to be the minimum effective duration for each training session, and that there is no benefit to training for more than three sessions per week [25]. We asked participants to train up to five times per week to mitigate for missed sessions and weeks where they would be unable to train (i.e., illness, holidays). Participants could train on additional sessions if they chose to. Given this is a feasibility study, understanding the adherence to the program by participants in each group was an important aspect of feasibility in the evaluation of this study. Participants in the intervention arm were offered a weekly telephone call or email contact for support during the program. The program consisted of a total of twenty games (5 attention, 5 language and verbal fluency, 5 memory, 5 visuospatial exercises), the order of which varied for each session. Each session included at least one attention, memory, visuospatial, and language task, and the remaining task cycled through an additional domain each session. Each exercise took approximately 5-6 min to complete. A full description of exercises included in this program is provided in Supplementary Table 1.

Sample size calculation

Given the primary aim of this study was to identify the feasibility of the protocol and intervention, sample size calculations were not performed a priori.

Statistical analysis

Data are presented as mean (standard deviation) for normally distributed data, and median [inter-quartile range] for non-normally distributed continuous data. The difference between training and control groups for continuous outcomes was tested by independent t-test (normally distributed) or Mann Whitney-U (non-normally distributed). Nominal data are presented as number (percentage), with statistical testing by Chi-square.

Task activation was measured by percentage change in CBFv from a 20 s baseline (T1) prior to task initiation. Two time points were taken to assess the change. Firstly, T2 which occurred at 5–10 s after task initiation, and T3 which occurred at 10–20 s after task initiation, where CBFv was averaged across these time intervals. These time intervals were determined by previous work from this group indicating that the peak response to task activation occurs at 5–10 s for the majority of tasks, indicating the initial response to stimulation [20, 26]. T3 therefore represents the sustained response after this initial peak, which was thought to be mediated more locally due to the metabolic gap [26].

In addition to the peak percentage change, a novel method to identify response to task activation was applied, which has been published recently [27]. In brief, response to task activation was classified as present or absent using two parameters: the cross correlation peak function (CCF) and the variance ratio (VR). The CCF is a measure of coherence between the individual response and the coherent average of the group response, and the VR is the ratio of the variance post-stimulus compared to the variance pre-stimulus. CCF values approaching one reflect a perfect response to task activation, and higher VR values correspond to larger task-activated response. Thresholds for the parameters are determined by calculating the upper 90th confidence limit based on randomly-marked baseline files, above which a response is determined as present. Disease-specific thresholds were determined to adjust for differences in baseline variability between healthy and disease groups which have been demonstrated in previous studies [28]. The group-specific thresholds are provided in Supplementary Table 2. Rather than a percent change from baseline, a response is considered present where either the CCF or VR are above the pre-determined threshold. This provides an objective measure of the presence or absence of a response to a given cognitive task. The cumulative response rate (CRR) was then calculated from the sum of responses to each task (n = 5) in each hemisphere (n = 2), to give a score out of 10. Increasing CRR therefore reflects an increasing number of responses to task activation.

Independent t-tests between intervention groups and within diagnostic group were used to determine estimates of intervention effect size for each diagnostic group in CBFv at and T3 and in CRR. Correlations between percentage changes in CBFv or CRR and clinical outcome measures were assessed using Pearson correlations for normally distributed and continuous variable comparisons, or Spearman correlations for non-normally distributed and continuous and ordinal comparisons. Cohen’s d was calculated in Excel for each outcome measure to determine effect size [29]. Effect sizes were considered small where d = 0.2, medium where d = 0.5, or large where d = 0.8 [29].

While we acknowledge this feasibility study was not powered to test for intervention effects and sub group analyses, we have provided p-values for such tests with the intent of better informing future research. However, these should be interpreted with caution.

As a result of the loss to follow-up, data on AD and MCI participants were analyzed as one cohort (cognitively impaired), and sub-group analyses were undertaken where possible.

Sample size calculation methodology for future studies

Sample size calculations were performed using Statulator© [30], an online sample size calculator to determine the sample size required to detect the expected mean difference between training and control groups at follow-up (intervention effect size) that we found in this study. Sample sizes were calculated within diagnostic groups (healthy, control, AD and MCI combined, AD and MCI separately for CRR). This is a basic sample size calculation using a t-test. Given the sample size is powered to detect a difference within and not between groups using composite outcome measures, correction for multiple comparisons was not performed.

Data availability

Additional data are provided in the Supplementary Material, including: CCF and VR thresholds, resting CBFv and peripheral parameters at baseline and follow-up, peak percentage change in CBFv and peripheral parameters in training and control groups at baseline and follow-up, significant correlations with cognitive test scores, and sample size calculations. Additional data are available on request from the corresponding author.

RESULTS

Baseline demographics

Table 1 summarizes the key baseline characteristics of the total sample, and between training and control groups. The mean age±SD of participants in this sample was 65.2±8.4 years (healthy cohort) and 70.2±8.6 years (cognitively impaired cohort). In healthy older adults, 50% were female, compared to 30.6% in the cognitively impaired group. The majority of healthy older adults and cognitively impaired cohort were Caucasian (95% and 100%, respectively) and right handed (90% and 91.7%, respectively). At baseline, there were no differences in the age of training or control groups, or the proportion of females in each group. At baseline, body mass index was significantly higher in the healthy control group (27.4 versus 24.2, p = 0.001), and scores were lower on the ACE-III memory domain (25.5 versus 23.5, p = 0.032) in the healthy control group. In the cognitively impaired cohort, antidepressant use was significantly higher in the control compared to training group (44.4 versus 11.1%, p = 0.03). Median time to follow-up was longer in the healthy training group (98 versus 86 days, p = 0.002), but comparable in the cognitively impaired cohort (90 versus 84.5 days, p = 0.093). There were no other significant differences between control and training groups for either the healthy or cognitively impaired cohorts at baseline (Table 1).

Table 1

Baseline demographics of the healthy participants and cognitively impaired cohort (Alzheimer’s disease and mild cognitive impairment) by training or control. Data are number (%) for nominal data, mean (standard deviation) for continuous parametric data, and median [inter-quartile range] for continuous non-parametric data. Significance testing was by Chi-square for nominal data, independent t-test for continuous parametric data, and Mann-Whitney-U for non-parametric, continuous data. Significance level was set at p < 0.05. BMI, body mass index; SBP, systolic blood pressure; DBP, diastolic blood pressure; AChEI, acetylcholinesterase inhibitors; NMDA, N-Methyl-D-aspartic; ACE-III, Addenbrooke’s Cognitive Examination; DEMQOL, Dementia Quality of Life; GDS, Geriatric Depression Scale; IADL, Instrumental Activities of Daily Living

Healthy older adults				Cognitively impaired cohort
Demographic	All	Training	Control	p	All	Training	Control	p
	(n = 20)	(n = 10)	(n = 10)		(n = 36)	(n = 18)	(n = 18)
Age (y)	65.2 (8.4)	63.0 (7.1)	67.4 (9.3)	0.25	70.19 (8.6)	71.8 (9.0)	70.3 (8.2)	0.72
Sex (n, % female)	10 (50)	6 (60)	4 (40)	0.37	11 (30.6)	5 (27.8)	6 (33.3)	0.72
Ethnicity (n, % Caucasian)	19 (95)	10 (100)	9 (90)	0.31	36 (100)	18 (100)	18 (100)	N/a
Diagnosis (n, %)	-	-	-	-
mild cognitive impairment					24 (66.7)	6 (33.3)	6 (33.3)	1.00
Alzheimer’s disease					12 (33.3)	12 (66.7)	12 (66.7)
Education (y)	17.1 (3.7)	18.5 (3.6)	15.6 (3.3)	0.078	12 [11–17]	15.5 [12–17]	11 [11–12.75]	0.10
Handedness (n, %)
Left	2 (10)	2 (20)	0 (0)	0.14	1 (2.8)	0 (0)	1 (5.6)	0.60
Right	18 (90)	8 (80)	10 (100)		33 (91.7)	17 (94.4)	16 (88.9)
				2 (5.6)	1 (5.6)	1 (5.6)
Smoker (n,%)
Current	0 (0)	0 (0)	0 (0)	0.37	1 (2.8)	0 (0)	1 (5.6)	0.53
Ex	10 (50)	4 (40)	6 (60)		17 (47.2)	8 (44.4)	9 (50)
Never	10 (5)	6 (60)	4 (40)		18 (50.0)	10 (55.6)	8 (44.4)
Alcohol (units/week)	2 [0–10]	0 [0–4.25]	5 [0.5–13]	0.19	0 [0–6.5]	0 [0–7.5]	0 [0–5]	0.82
BMI (kg/m²)	25.8 (2.5)	24.2 (1.8)	27.4 (2.0)	0.001	25.1 (3.8)	24.3 (3.8)	25.8 (3.8)	0.23
SBP (mmHg)	139.3 (14.0)	137.9 (15.4)	140.6 (13.0)	0.68	135.1 (18.5)	129.8 (14.0)	140.2 (21.0)	0.35
DBP (mmHg)	81.0 (10.3)	78.8 (10.2)	83.1 (10.5)	0.37	76.6 (10.9)	75.8 (11.6)	77.4 (10.5)	0.86
Diabetes (n,%)	1 (5)	1 (10)	0 (0)	0.31	5 (13.9)	3 (16.7)	2 (11.1)	0.63
Hypertension (n,%)	8 (4)	3 (30)	5 (50)	0.36	13 (36.1)	6 (33.3)	7 (38.9)	0.73
Hypercholesterolemia (n, %)	6 (30)	3 (30)	3 (30)	1.00	19 (52.8)	10 (55.6)	9 (50)	0.74
Heart disease (n, %)	1 (5)	1 (10)	0 (0)	0.31	4 (11.1)	1 (5.6)	3 (16.7)	0.29
Previous stroke (n, %)	0 (0)	0 (0)	0 (0)	-	3 (8.3)	1 (5.6)	2 (11.1)	0.55
Depression (n, %)	2 (10)	0 (0)	2 (20)	0.14	7 (19.4)	2 (11.1)	5 (27.8)	0.21
Heart failure (n, %)	1 (5)	0 (0)	1 (10)	0.31	0 (0)	0 (0)	0 (0)	N/A
Total number of medications	2 [0.75–3.5]	2 (2.1)	2.5 (2.1)	0.60	5 [3–6]	3 [1.25–5.75]	5 [4–7]	0.053
AChEI (n, %)	0 (0)	0 (0)	0 (0)	-	16 (44.4)	8 (44.4)	8 (44.4)	1.00
NMDA (N, %)	0 (0)	0 (0)	0 (0)	-	7 (19.4)	3 (16.7)	4 (22.2)	0.67
Antidepressant (n, %)	2 (10)	0 (0)	2 (20)	0.14	10 (27.8)	2 (11.1)	8 (44.4)	0.03
Antihypertensive (n, %)	7 (35)	3 (30)	4 (40)	0.64	15 (41.7)	6 (33.3)	9 (50)	0.31
ACE-III
Total	97.5 [94.8–99]	97.5 [95.5–99.8]	97.5 [93.5–98]	0.48	82 [74–92]	84 [74–86.8]	82 [75–93]	0.58
Attention	18 [17.5–18]	18 [18-18]	18 [17.3–18]	0.74	16 [12–17]	16 [14.3–17.8]	15 [12–17]	0.17
Fluency	13.5 [11.8–14]	13 [12–14]	14 [11.3–14]	0.91	10 [8–12]	9 [7.3–11]	10 [9.3–12]	0.29
Memory	25 [23–26]	25.5 [25–26]	23.5 [22.3–25]	0.035	18 [12.75–23]	18 [12.3–23]	18 [13.5–19]	0.67
Language	26 [26-26]	26 [26-26]	26 [26-26]	0.48	25 [24–26]	25 [24.3–26]	25 [24–26]	0.74
Visuospatial	16 [15.8–16]	16 [16-16]	16 [15.3–16]	0.74	15 [14.8–16]	16 [15-16]	15 [14–16]	0.21
DEMQOL	105 [100.3–109.3]	106.4 (5.9)	101.7 (9.2)	0.19	93 [85–102]	96 [86.8–101.5]	91 [85–101.3]	0.54
GDS	1 [0.75–2]	1 [0.25–1]	1.5 [1–2.8]	0.14	2 [1–4]	2.5 [1–3]	2 [2–4]	0.72
Lawton IADL	8 [8-8]	8 [8-8]	8 [8-8]	1	6 [5–8]	7 [5–8]	5 [5–7.8]	0.52
Time to follow up (days)	89 [85.5–98]	98 [90.8–117.5]	86 [84–89]	0.002	86 [84–91]	90 [83.5–101]	84.5 [84–88]	0.093

Feasibility assessment

Both the protocol and the intervention were well tolerated by healthy participants and those with MCI; there were no withdrawals from the study as a result of the study procedures, or TCD protocol. There were three dropouts (25%) from the AD group as a result of the intervention, suggesting it was less well tolerated than in the healthy and MCI groups. Data on hemodynamic outcomes were lost on five participants in the AD group due to lack of TCD window (n = 1), and the COVID-19 pandemic precluding face-to-face assessment. In healthy older adults, there was one loss to follow-up in the control group, but this was unrelated to the study. There were no losses to follow-up in the MCI group. All participants were willing to be randomized to either intervention or waiting-list control.

In terms of adherence to the protocol, the median number of sessions trained by healthy older adults was 86.5 (Range: 51–228 sessions, IQR: 75–102), but this was lower in the cognitively impaired cohort at 70 sessions (Range: 28–168, IQR: 47–83). Figure 1 summarizes the recruitment process for the study.

Fig. 1

Recruitment flow chart.

Secondary outcome measures

Table 2 summarizes the clinical outcomes by training and control groups. There were no differences between groups in cognitive function (ACE-III), mood (GDS), or function (Lawton IADL). There were no differences in cognitive sub-domains of the ACE-III between groups. Quality of life was significantly improved in healthy training compared to healthy control (MD: 4.83 [95% CI: 1.13, 8.54]), but not cognitively impaired, cohort.

Table 2

Secondary outcome measures. Data are mean Δ baseline to follow-up (standard deviation), and mean difference (95% confidence interval). Significance testing was by independent t-test and the significance level was set at p < 0.05. ACE-III, Addenbrooke’s Cognitive Examination-III; DEMQOL, Dementia Quality of Life; GDS, Geriatric Depression Scale; IADL, Instrumental Activities of Daily Living

Healthy older adults				Cognitively impaired cohort
Outcome	Δ Training	Δ Control	MD	p	Δ Training	Δ Control	MD	p
	(n = 10)	(n = 9)	(95% CI)		(n = 15)	(n = 18)	(95% CI)
ACE-III total	0.20 (1.61)	–0.11 (3.95)	0.31 (–2.55, 3.18)	0.83	1.33 (4.45)	–0.53 (4.00)	1.86 (–1.19, 4.91)	0.22
Attention	–0.1 (0.88)	–0.56 (1.01)	0.46 (–0.46, 1.37)	0.46	–0.13 (1.30)	0.00 (2.20)	–0.13 (–1.45, 1.18)	0.84
Fluency	0.40 (1.26)	0.56 (1.13)	–0.16 (–1.32, 1.01)	0.78	0.40 (2.03)	–0.06 (0.94)	0.46 (–0.64, 1.55)	0.43
Visuospatial	0.20 (0.42)	0.00 (0.50)	0.20 (–0.25, 0.65)	0.36	–0.20 (0.94)	0.18 (1.33)	–0.38 (–1.22, 0.47)	0.37
Language	–0.1 (0.32)	–0.22 (0.67)	0.12 (–0.37, 0.62)	0.12	0.27 (1.49)	–0.24 (1.30)	0.50 (–0.50, 1.51)	0.32
Memory	–0.2 (1.13)	0.78 (2.28)	–0.98 (–2.82, 0.86)	0.25	0.93 (2.89)	–0.22 (2.32)	1.16 (–0.69, 3.00)	0.21
DEMQOL	1.50 (2.72)	–3.33 (4.77)	4.83 (1.13, 8.54)	0.014	2.53 (6.97)	–0.94 (10.10)	3.48 (–2.81, 9.77)	0.27
GDS	–0.1 (0.74)	0.33 (1.00)	–0.43 (–1.28, 0.41)	0.29	–0.20 (1.70)	0.44 (2.62)	–0.64 (–2.25, 0.96)	0.42
Lawton IADL	0.00 (0.00)	–0.11 (0.33)	0.11 (–0.15, 0.37)	0.35	0.07 (1.10)	0.11 (0.90)	–0.04 (–0.75, 0.67)	0.90

Resting cerebral and peripheral physiological data

There were no differences in resting cerebral and physiological parameters at baseline or follow-up between control and training groups in the healthy or cognitively impaired cohorts. Within group, there were no differences in resting values in the cognitively impaired or healthy training cohorts. In the healthy control group, resting values in CBFv1 and 2 were lower at follow-up. Supplementary Tables 3 and 4 summarize the resting cerebral and peripheral physiological data for baseline and follow-up visits by training and population group.

Task activated cerebral and peripheral physiological data

In healthy older adults, peak percentage change in CBFv was significantly decreased at follow-up in the visuospatial task at T2 (right MD: –10.22% [95% CI: –19.87, –0.57]; left MD: –10.69% [95% CI: –20.93, –0.44]) in healthy training relative to healthy control (Fig. 2). There were no significant differences in the task-activated response at T2 or T3 for either healthy or cognitively impaired cohorts for the other tasks. Data are provided for all task-activated responses in cerebral and peripheral parameters in Supplementary Tables 5 and 6. Figure 2 provides a representative trace of the temporal CBFv change to the visuospatial task in the control and training groups at follow-up in the healthy cohort.

Fig. 2

Changes in the visuospatial task in the healthy cohort. a) Δ Percentage change in CBFv with mean and 95% confidence intervals at T2 (left) and T3 (right) from baseline to follow-up in control, and training groups for CBFv1 (black circles) and CBFv2 (clear circles). b) Change in peak percentage at T2 for CBFv1 (left), and CBFv2 (right) in control and training groups from baseline to follow-up. c) Δ Percentage change in MAP (left) at T2 (black circles) and T3 (clear circles), and ETCO₂ (right) at T2 (black circles), and T3 (clear circles). d) Representative trace of the temporal CBFv response (solid line) with standard deviation (dashed line) to the visuospatial task at follow-up in the training (grey) and control (black) groups for CBFv1 (right) and CBFv2 (left) hemispheres. The arrow denotes task initiation. ^*Significantly different between training and control (p < 0.05).

At T2 the estimated difference between the training and control intervention for the healthy cohort was –5.35 (95% CI: –12.82, 2.11), and for the cognitively impaired cohort was –1.86 (95% CI: –5.60, 0.87) (Table 3). At T3, the estimated difference between the training and control intervention for the healthy cohort was 3.81 (95% CI: –10.03, 2.42), and for the cognitively impaired cohort was –0.27 (95% CI: –3.27, 2.73) (Table 3). Effect sizes were medium to large for the difference in peak percentage change between training and control groups at T2 for both cohorts (Cohen’s d = 0.59–0.69), and for the healthy cohort at T3 (Cohen’s d = 0.53), but small for the cognitively impaired cohort at T3 (Cohen’s d = 0.07) (Table 5).

Table 3

Mean and 95% confidence intervals for the average peak % change in CBFv response to 5 tasks by training group, diagnosis, and time-point. Statistical testing by independent t-test

			T2				T3
Diagnosis	Δ Training	Δ Control	Mean difference (95% CI)	p	Δ Training	Δ Control	Mean difference (95% CI)	p
Healthy	–2.87 (7.57)	2.48 (7.85)	–5.35 (–12.82, 2.11)	0.15	–3.18 (6.06)	0.62 (6.81)	–3.81 (–10.03, 2.42)	0.22
Cognitively impaired	–0.73 (4.01)	1.13 (3.04)	–1.86 (–5.60, 0.87)	0.17	–0.01 (3.12)	0.25 (4.26)	–0.27 (–3.27, 2.73)	0.86
Mild cognitive impairment	–3.39 (1.67)	0.40 (1.64)	–3.79 (–5.92, –1.65)	0.003	–0.34 (1.51)	1.32 (3.50)	–1.66 (–5.12, 1.60)	0.31
Alzheimer’s disease	1.92 (3.95)	1.57 (3.64)	0.35 (–3.81, 4.51)	0.86	0.31 (4.35)	–0.39 (4.72)	0.70 (–4.38, 5.78)	0.77

CRR analysis

The estimated difference in CRR between the training and control intervention for the healthy cohort was –1.84 (95% CI: –4.81, 1.12), and for the cognitively impaired group was 1.79 (95% CI: 0.005, 3.58) (Table 4, Fig. 3). Effect sizes were medium to large for both cohorts in CRR difference between training and control (Cohen’s d = 0.41–1.11) (Table 5).

Table 4

Mean (standard deviation), mean difference and 95% confidence interval for change in cumulative response rate in training and control groups. Significance testing by independent t-test

Diagnosis	Δ Training	Δ Control	Mean difference	p
Healthy control	–2.40 (3.57)	–0.56 (2.35)	–1.84 (–4.81, 1.12)	0.21
Alzheimer’s disease and mild cognitive impairment	0.92 (1.73)	–0.88 (2.60)	1.79 (0.005, 3.58)	0.049
Mild cognitive impairment	–0.17 (0.75)	–1.00 (2.37)	0.83 (–1.43, 3.09)	0.43
Alzheimer’s disease	2.00 (1.79)	–0.80 (2.86)	2.80 (–0.00, 5.60)	0.050

Fig. 3

Δ CRR from baseline to follow-up in control and training groups. Left panel = mean and 95% confidence interval for ΔCRR in control (black circles), and training (clear circles). Right panel = change per participant in baseline CRR (back circles), to follow-up CRR (clear circles) by training and control groups. a) Healthy older adults, b) Alzheimer’s disease, c) mild cognitive impairment, and d) Alzheimer’s disease and mild cognitive impairment combined. ^*Significantly different between training and control (p < 0.05).

Table 5

Sample sizes required to detect the mean difference in percentage change in CBFv or CRR between training and control groups, with standard deviation pooled from the follow-up data. Cohen’s d provided as a measure of effect size. Alpha was set at 0.05 and power at 80%

Diagnosis	Expected mean difference	Cohen’s d	Standard deviation	Sample size
Δ Peak % change in CBFv
Healthy control
T2	5.35	0.69	9.59	102
T3	3.81	0.59	9.59	200
AD and MCI
T2	1.86	0.53	6.49	384
T3	0.27	0.07	6.49	18,140
ΔCRR
Healthy control	–1.84	0.60	3.00	84
AD and MCI	1.79	0.79	2.39	56
MCI	0.83	0.47	2.66	334
AD	2.80	1.11	2.14	20

Correlations between clinical and CBFv outcomes

There were a number of significant correlations in both the control and training groups between clinical outcome measures and change in task activated CBFv in healthy older adults. These are summarized in Supplementary Table 7.

In the cognitively impaired cohort, there were no significant correlations between Δ peak percentage change in CBFv and change in clinical outcome measures (DEMQOL, GDS, IADL) or cognitive test scores for any tasks at T2 or T3.

Sample sizes for a future study

Table 5 provides estimates of sample sizes needed for future studies based on the estimates observed in this study. To detect the observed difference in CBFv at T2 and T3 would require a sample size of 102–200 participants in the healthy cohort, and 384–18140 in the cognitively impaired cohort. CRR required fewer participants to detect the observed changes: 84 participants in the healthy cohort, and 56 participants in the cognitively impaired cohort.

DISCUSSION

Main findings

In summary, the study protocol was feasible and acceptable to both healthy older adults, and participants with a diagnosis of MCI or AD. Adherence to the training program was better in the healthy compared to cognitively impaired cohort, with more training-related dropouts in the AD group. Given that 89% of participants were able to complete the trial, and hemodynamic assessment was well tolerated, this warrants follow-up investigation in a larger trial. Drop-outs were more likely to occur in the AD group, indicating this group may need additional support, or alterations to the training program to support completion. On secondary outcome analysis, preliminary results suggest there may be some evidence of training-related effects on cerebrovascular function, but we are unable to draw definitive conclusions from this study. Below we outline considerations for a future trial investigating the effects of CT on cerebral hemodynamics in the context of the current literature.

Context and possible mechanisms of vascular plasticity

Resting CBF

A recent systematic review by Ten Brinke et al. identified nine studies which examined functional imaging outcomes following cognitive training in healthy older adults and MCI [7]. The results of these studies were mixed, demonstrating both increased and decreased task-based fMRI responses [7]. The majority of studies have thus far investigated the effects of cognitive training on resting CBF in healthy populations rather than task activated responses [19, 31]. Mozolic et al. found resting CBF increased in the prefrontal cortex following an eight-week attention and distractibility program in 66 healthy older adults, which correlated with improvements in cognitive outcome measures [19]. Similarly, Chapman et al. investigated resting CBFv changes in 37 healthy older adults randomized to cognitive training or control for 12 weeks [31]. They also demonstrated training related rises in resting CBF which mirrored increases in functional connectivity in the default mode and central executive networks, preceded structural increases and were sustained across all time points of the study [31]. Given that both healthy aging and dementia are associated with consistent declines in resting CBF [32 –37], and that resting CBF can correlate with cognitive performance [13 , 38] and predict cognitive decline [12], increasing resting levels as seen in these studies may represent restitution of CBF to levels that are seen in younger adults or pre-pathological states [19, 31].

Task-activated CBF

In a recent systematic review, the majority of studies in MCI demonstrated an increase in task-activated responses after training [8]. There was only one study that investigated participants with established dementia and demonstrated a reduction in task-activated responses [39]. A meta-analysis of 2097 healthy older adults demonstrated both region-dependent increases and decreases in task-activated responses post-cognitive training [40]. Thus, it re-mains unclear whether training related increases or decreases in task-activation are beneficial. Certainly, this may depend on whether the focus is prevention of cognitive decline in healthy older adults, or restitution of normal brain function in those with MCI or dementia. Healthy aging is associated with a number of key physiological changes when compared to younger adults, with lower resting CBF, loss of hemispheric lateralization (HAROLD theory) [41, 42], prefrontal hyperactivation [16, 42], and increased cortical responses to task activation [34]. These changes could be compensatory [42], as the recruitment of additional neural circuits (CRUNCH theory) maintained cognitive performance in line with that of younger adults, but with a significant increase in resource requirement [43]. Thus, returning patterns of brain activity and activation to those seen in younger adults may indicate improvements in neural processing efficiency and reduction in resource requirement and compensation. In those with cognitive disorders, such as MCI and dementia, studies have demonstrated inconsistent findings in hemodynamic changes with both increases and decreases in resting and task-activated CBF [44 –46]. One theory is the notion of a compensatory “hyperperfused” state in early MCI [46], progressing to hypoperfusion and reduced task activation in established dementia [36 , 48]. Thus, in the early stages of cognitive decline, the goal could be similar to that in healthy older adults, but in later cognitive decline, this may shift to increasing resting CBF and task activated response. In this context, the trend towards reduced CBFv responses in the healthy cohort could represent a reduction in resource requirement by healthy older adults to execute a given task, which may be achieved through structural or functional vascular and neural remodeling [16]. In contrast, the increased response rate seen in participants with AD and MCI here, may represent an improvement in the blood flow response to cognitive stimulation, providing better coupling between brain activity and blood supply. However, these are cautious interpretations given the small sample size and need for further investigation.

Considerations for a future trial, limitations, and future directions

Only quality of life was significantly better in the healthy training cohort. Although the study was not powered to detect changes in secondary outcome measures, healthy older adults and participants with MCI performed well on the ACE-III and Lawton IADLs, suggesting participants were either close to or already at the ceiling of these scales. Thus, more sensitive scales may be needed for healthy adults and participants with MCI to detect clinically meaningful changes. In the absence of sensitive scales, physiological or functional brain imaging outcomes may provide a useful surrogate marker of cognitive training effects in studies of healthy older adults and participants with MCI.

The current study is limited by a small sample size. While it was large enough to establish feasibility of the protocol and intervention, the sample size was not powered to detect changes in clinical and CBF outcome measures. In particular, the AD and MCI sub-groups were combined due to small sample sizes to improve the power of the analyses. AD and MCI may have different hemodynamic profiles depending on the stage of cognitive decline, and mechanisms at play (e.g., predominantly amnestic or not). However, in recent analyses [49], the hemodynamic profiles of these participants were found to be similar, which may be due to the relatively milder stage at which the AD participants were enrolled. However, the data from this study have provided the information to calculate the sample size needed for a future appropriately powered trial to investigate these findings further. Importantly, estimates of sample size varied considerably depending on the hemodynamic outcome measure used (between 20–18000 participants). In particular, CRR and peak percentage change at T2 provided much smaller samples for future trials and this is an important consideration in the choice of outcome measure for future studies. The lack of blinding was a limitation in this study, as both patient and investigator could have introduced bias. Blinding of participants was problematic due to the lack of active control condition, and participants in the training arm may be more likely to report improvements or benefits, particularly on self-reported outcomes (quality of life, mood, and function), although few benefits were identified in these measures. Blinding of the investigator was, similarly, problematic due to limited operators with sufficient skills to undertake the physiological protocol. To mitigate this, hemodynamic and peripheral data were batch-blinded for analysis, but changes in operator technique cannot be excluded. In a future trial, blinding of both investigator and participants is important to minimize the effects of bias. In this study, the control condition was standard or usual care, with no active comparator. In studies investigating cognitive training, an active control or treatment comparator are preferred to eliminate other potential effects (i.e., computer use), and thus results may not be all directly attributable to the intervention. In addition, only participants with internet access were recruited, which may have entered selection bias by recruiting those with regular internet usage. The education level was lower among the cognitively impaired cohort compared to healthy older adults, although not significantly different between training and control groups. There were differences in some baseline co-variates between the intervention groups (e.g., BMI, antidepressant use). This may account for some of the differences in response to cognitive training between or within population groups. In future, a fully powered linear mixed effect model, taking into account significant co-variates, should be used to identify training related effects on cerebral hemodynamics. The use of TCD to measure changes in CBFv relies on the assumption that the vessel diameter remains constant despite fluctuations in BP and ETCO₂. TCD measured-task activated has limited reproducibility [50] and this may affect the reliability of estimates of effect sizes. The tasks used in this protocol were not all selective for the MCA territory, but have been shown previously to produce robust responses in this region in healthy and cognitive impaired cohorts [20, 49]. Furthermore, TCD is limited by a lack of spatial resolution, even with selective tasks, and studies which use multiple neuroimaging techniques would be able to provide complementary information on spatial as well as temporal characteristics of the hemodynamic changes post-training. Peak percentage change in CBFv is susceptible to differences in resting CBFv baseline seen in cognitive impairment, and between baseline and follow-up measurements within groups as identified in the healthy control group in this study. However, CRR is susceptible to differences in resting variability of the baseline, which were corrected for using disease-specific thresholds from a larger dataset. CRR therefore may provide a better outcome measure due to reduced susceptibility to baseline fluctuations, effects of CO₂ and blood pressure, and smaller sample sizes to power a future trial. Studies may also consider investigating the additive effects of multi-component interventions by investigating the effects of physical exercise, diet, education, and cognitive training on cerebral hemodynamic parameters. Future studies should investigate longer term outcomes of cognitive training and whether “booster” sessions are required to maintain functional brain changes. Furthermore, the optimal training dose (i.e., duration of program and sessions, number and frequency of sessions) has not been fully determined in cognitively impaired populations. The training program for this study was based on data from trials conducted with healthy participants [25], and further work is required to identify if a greater or lesser training dose is required to be effective in cognitively impaired populations.

Summary

In summary, a protocol investigating the effects of cognitive training on cerebral hemodynamics in healthy older adults, and participants with cognitive impairment was feasible and acceptable. Training related changes were seen in cerebral hemodynamics, but the interpretations are limited by the small sample size. These results require further investigation in a larger appropriately powered trial of cognitive training with hemodynamic outcome measures to understand the role of cognitive training in modulating vascular plasticity and cognitive function in healthy older adults and those with dementia.

Footnotes

ACKNOWLEDGMENTS

We would like to acknowledge the input of the members of the patient and public involvement group and trial steering committee in the developments and oversight of this study. We would like to acknowledge Lumosity© for providing the cognitive training program used in this study.

LB is a Dunhill clinical research training fellow (RTF1806∖27). TGR is an NIHR Senior Investigator. Lumosity© provided the cognitive training program but did not provide any financial support and were not involved in the design or implementation of the study.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Prince

, Wimo

, Guerchet

, Ali

, Wu

, Prina

(2015) World Alzheimer Report 2015. The Global Impact of Dementia. Alzheimer’s Disease International, London.

World Health Organization (2017) Global action plan on the public health response to dementia 2017 - 2025. World Health Organization, Geneva, pp. 18-22.

Petersen

, Caracciolo

, Brayne

, Gauthier

, Jelic

, Fratiglioni

(2014) Mild cognitive impairment: A concept in evolution. J Intern Med 275, 214–228.

Gates

, Sachdev

, Fiatarone Singh

, Valenzuela

(2011) Cognitive and memory training in adults at risk of dementia: A systematic review. BMC Geriatr 11, 55.

Hill

, Mowszowski

, Naismith

, Chadwick

, Valenzuela

, Lampit

(2017) Computerized cognitive training in older adults with mild cognitive impairment or dementia: A systematic review and meta-analysis. Am J Psychiatry 174, 329–340.

Lampit

, Hallock

, Suo

, Naismith

, Valenzuela

(2015) Cognitive training-induced short-term functional and long-term structural plastic change is related to gains in global cognition in healthy older adults: A pilot study. Front Aging Neurosci 7, 14.

ten Brinke

, Davis

, Barha

, Liu-Ambrose

(2017) Effects of computerized cognitive training on neuroimaging outcomes in older adults: A systematic review. BMC Geriatr 17, 139.

Beishon

, Intharakham

, Swienton

, Panerai

, Robinson

, Haunton

(2020) Neuroimaging outcomes in studies of cognitive training in mild cognitive impairment and early Alzheimer’s disease: A systematic review. Curr Alzheimer Res 17, 472–486.

Zlokovic

(2011) Neurovascular pathways to neurodegeneration in Alzheimer’s disease and other disorders. Nat Rev Neurosci 12, 723–738.

10.

Attems

, Jellinger

(2014) The overlap between vascular disease and Alzheimer’s disease–lessons from pathology. BMC Med 12, 206.

11.

Strickland

(2018) Blood will out: Vascular contributions to Alzheimer’s disease. J Clin Invest 128, 556–563.

12.

Ruitenberg

, den Heijer

, Bakker

SLM

, van Swieten

, Koudstaal

, Hofman

, Breteler

MMB

(2005) Cerebral hypoperfusion and clinical onset of dementia: The Rotterdam Study. Ann Neurol 57, 789–794.

13.

Poels

, Ikram

, Vernooij

, Krestin

, Hofman

, Messen

, van der Lugt

, Breteler

(2008) Total cerebral blood flow in relation to cognitive function: The Rotterdam Scan Study. J Cereb Blood Flow Metab 28, 1652–1655.

14.

Leeuwis

, Benedictus

, Kuijer

JPA

, Binnewijzend

MAA

, Hooghiemstra

, Verfaillie

SCJ

, Koene

, Scheltens

, Barkhof

, Prins

, van der Flier

(2017) Lower cerebral blood flow is associated with impairment in multiple cognitive domains in Alzheimer’s disease. Alzheimers Dement 13, 531–540.

15.

Iadecola

(2017) The neurovascular unit coming of age: A journey through neurovascular coupling in health and disease. Neuron 96, 17–42.

16.

Park

, Bischof

(2013) The aging mind: Neuroplasticity in response to cognitive training. Dialogues Clin Neurosci 15, 109–119.

17.

Zatorre

, Fields

, Johansen-Berg

(2012) Plasticity in gray and white: Neuroimaging changes in brain structure during learning. Nat Neurosci 15, 528–536.

18.

Vermeij

, Kessels

RPC

, Claassen

JAHR

, Heskamp

, Simons

EMF

, Dautzenberg

PLJ

(2017) Prefrontal activation may predict working-memory training gain in normal aging and mild cognitive impairment. Brain Imaging Behav 11, 141–154.

19.

Mozolic

, Hayasaka

, Laurienti

(2010) A cognitive training intervention increases resting cerebral blood flow in healthy older adults. Front Hum Neurosci 4, 16.

20.

Beishon

, Williams

CAL

, Panerai

, Robinson

, Haunton

(2018) The assessment of neurovascular coupling with the Addenbrooke’s Cognitive Examination: A functional transcranial Doppler ultrasonographic study. J Neurophysiol 119, 1084–1094.

21.

Beishon

, Evley

, Panerai

, Subramaniam

, Mukaetova-Ladinska

, Robinson

, Haunton

(2019) Effects of brain training on brain blood flow (The Cognition and Flow Study-CogFlowS): Protocol for a feasibility randomised controlled trial of cognitive training in dementia. BMJ Open 9, e027817.

22.

Petersen

(2004) Mild cognitive impairment as a diagnostic entity. J Intern Med 256, 183–194.

23.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr. , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

24.

Oldfield

(1971) The assessment and analysis of handedness: The Edinburgh inventory. Neuropsychologia 9, 97–113.

25.

Lampit

, Hallock

, Valenzuela

(2014) Computerized cognitive training in cognitively healthy older adults: A systematic review and meta-analysis of effect modifiers. PLoS Med 11, e1001756.

26.

Beishon

, Williams

CAL

, Robinson

, Haunton

, Panerai

(2018) Neurovascular coupling response to cognitive examination in healthy controls: A multivariate analysis. Physiol Rep 6, e13803.

27.

Beishon

, Williams

, Intharakham

, Batterham

, Barnes

, Haunton

, Robinson

, Panerai

(2020) An objective method to identify non-responders in neurovascular coupling testing. J Neurosci Methods 341, 108779.

28.

Lam

, Haunton

, Robinson

, Panerai

(2019) Dynamic cerebral autoregulation measurement using rapid changes in head positioning: Experiences in acute ischemic stroke and healthy control populations. Am J Physiol Heart Circ Physiol 316, H673–h683.

29.

Salkind

(2010) Encyclopedia of Research Design. SAGE Publications, Inc., Thousand Oaks, CA, pp. 181–185.

30.

Dhand

, Khatkar

, Statulator: An online statistical calculator. Sample Size Calculator for Comparing Two Independent Means, http://statulator.com/SampleSize/ss2M.html,

31.

Chapman

, Aslan

, Spence

, Hart

Jr. , Bartz

, Didehbani

, Keebler

, Gardner

, Strain

, DeFina

, Lu

(2015) Neural mechanisms of brain plasticity with complex cognitive training in healthy seniors. Cereb Cortex 25, 396–405.

32.

Oudegeest-Sander

, van Beek

, Abbink

, Olde Rikkert

, Hopman

, Claassen

(2014) Assessment of dynamic cerebral autoregulation and cerebrovascular CO2 reactivity in ageing by measurements of cerebral blood flow and cortical oxygenation. Exp Physiol 99, 586–598.

33.

Dineen

, Panerai

, Brodie

, Robinson

(2011) Effects of ageing on cerebral haemodynamics assessed during respiratory manoeuvres. Age Ageing 40, 199–204.

34.

Beishon

, Minhas

, Patrick

, Shanmugam

, Williams

CAL

, Panerai

, Robinson

, Haunton

(2018) The effects of healthy ageing on cerebral blood flow responses to cognitive testing. Curr Aging Sci 11, 226–235.

35.

Catchlove

, Macpherson

, Hughes

, Chen

, Parrish

, Pipingas

(2018) An investigation of cerebral oxygen utilization, blood flow and cognition in healthy aging. PLoS One 13, e0197055.

36.

Sabayan

, Jansen

, Oleksik

, van Osch

, van Buchem

, van Vliet

, de Craen

, Westendorp

(2012) Cerebrovascular hemodynamics in Alzheimer’s disease and vascular dementia: A meta-analysis of transcranial Doppler studies. Ageing Res Rev 11, 271–277.

37.

Keage

, Churches

, Kohler

, Pomeroy

, Luppino

, Bartolo

, Elliott

(2012) Cerebrovascular function in aging and dementia: A systematic review of transcranial Doppler studies.. Dement Geriatr Cogn Dis Extra 2, 258–270.

38.

Tarumi

, Dunsky

, Khan

, Liu

, Hill

, Armstrong

, Martin-Cook

, Cullum

, Zhang

(2014) Dynamic cerebral autoregulation and tissue oxygenation in amnestic mild cognitive impairment. J Alzheimers Dis 41, 765–778.

39.

Huntley

, Hampshire

, Bor

, Owen

, Howard

(2017) Adaptive working memory strategy training in early Alzheimer’s disease: Randomised controlled trial. Br J Psychiatry 210, 61–66.

40.

Duda

, Sweet

(2020) Functional brain changes associated with cognitive training in healthy older adults: A preliminary ALE meta-analysis. Brain Imaging Behav 14, 1247–1262.

41.

Cabeza

(2002) Hemispheric asymmetry reduction in older adults: The HAROLD model. Psychol Aging 17, 85–100.

42.

Goh

, Park

(2009) Neuroplasticity and cognitive aging: The scaffolding theory of aging and cognition. Restor Neurol Neurosci 27, 391–403.

43.

Berlingeri

, Danelli

, Bottini

, Sberna

, Paulesu

(2013) Reassessing the HAROLD model: Is the hemispheric asymmetry reduction in older adults a special case of compensatory-related utilisation of neural circuits? Exp Brain Res 224, 393–410.

44.

Hays

, Zlatar

, Wierenga

(2016) The utility of cerebral blood flow as a biomarker of preclinical Alzheimer’s disease. Cell Mol Neurobiol 36, 167–179.

45.

Wierenga

, Hays

, Zlatar

(2014) Cerebral blood flow measured by arterial spin labeling MRI as a preclinical marker of Alzheimer’s disease. S. J Alzheimers Dis 42, 411–419.

46.

Sierra-Marcos

(2017) Regional cerebral blood flow in mild cognitive impairment and Alzheimer’s disease measured with arterial spin labeling magnetic resonance imaging. Int J Alzheimers Dis 2017, 5479597.

47.

Asil

, Uzuner

(2005) Differentiation of vascular dementia and Alzheimer disease: A functional transcranial Doppler ultrasonographic study. J Ultrasound Med 24, 1065–1070.

48.

Matteis

, Silvestrini

, Troisi

, Bragoni

, Vernieri

, Caltagirone

(1998) Cerebral hemodynamic patterns during stimuli tasks in multi-infarct and Alzheimer types of dementia. Acta Neurol Scand 97, 374–380.

49.

Beishon

, Panerai

, Robinson

, Subramaniam

, Haunton

(2018) 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. J Alzheimers Dis Rep 2, 153–164.

50.

Beishon

, Williams

CAL

, Panerai

, Robinson

, Haunton

(2017) Reproducibility of task activation using the Addenbrooke’s cognitive examination in healthy controls: A functional Transcranial Doppler ultrasonography study. J Neurosci Methods 291, 131–140.