INTRODUCTION

Dementia impairs cognitive functions that severely interfere with daily activities. It is the most common neurodegenerative disease in older adults, with continuing growth due to the increased aging population with longer life expectancy [1]. Currently, there is no effective treatment for dementia. Thus, preventing dementia by identifying some modifiable risk factors in the early phase is an urgent need.

Hypertension is known as a modifiable risk factor for cognitive impairment and dementia [2]. However, there is controversy about the relationship between blood pressure (BP) and cognitive dysfunction. Although midlife hypertension was linked to an increased risk for dementia, late-life BP is more complicated, with both high SBP and low DBP associated with dementia [3]. Recently, emerging evidence indicated that BP variability (BPV), that is, fluctuation of BP, was related to cognitive impairment and dementia independent of mean BP levels [4–15]. Furthermore, the contribution of BPV to the risk of cognitive impairment and dementia may be beyond the BP value [13–15]. However, there are limitations in clinical studies regarding the relationship between BPV and dementia. Moreover, the mechanisms underlying this relationship are unclear.

The current review aims to summarize the latest evidence of BPV correlating with cognitive impairment and dementia in cognitively intact populations, patients with mild cognitive impairment (MCI), and different dementia types, introducing the new advances and critical confounding factors in the research. A second objective is to summarize the potential mechanisms underlying the relationship between BPV and cognitive impairment and dementia and briefly discuss sex differences in the relationship. At last, this review discusses the current challenges in the clinical trials and future directions to optimize BP management at an early stage to prevent cognitive impairment and dementia in later life.

MEASUREMENT OF BLOOD PRESSURE VARIABILITY

BP oscillates in response to external stimulation (environment), behavioral factors, and internal cardiovascular regulatory mechanisms [10,16]. BPV can be evaluated within different time frames: very short-term (beat-to-beat), short-term [within 24 h, usually measured by ambulatory BP (ABP) monitoring], mid-term (day-to-day, usually measured by home BP monitoring), or long-term (visit-to-visit, between clinic visits over weeks, months, and years). There are four types of indices for overall BPV measurement [16], including dispersion [such as SD, coefficient of variation (CV), and variability independent of the mean], sequence (such as average real variability), instability [such as range (maximum–minimum BP), peak size (maximum BP), and trough size (mean–minimum BP)], and frequency (such as residual variability). The most widely used indices are demonstrated and compared below (Table 1).

TABLE 1 - Blood pressure variability indices Index Equation Type of BPV measured Advantage Disadvantage Standard deviation (SD)

∑i=1n(xi−x¯)2(n−1)

Very short-term BPV
Short-term BPV
Mid-term BPV
Long-term BPV Most used; independent of measurement order Correlated with mean value; influenced by outliers and night-time BP fall Residual standard deviation (RSD)

∑i=1n(xi−x∘i)2(n−2)

Long-term BPV Less influenced by BP change over time compared with SD The assumption of a linear trend may not accurately reflect the nature of changes over time Coefficient of variance (CV) 100 × SD/

x¯

Short-term BPV
Mid-term BPV
Long-term BPV Weakly correlated with mean levels; correct correlations between SD and mean BP values Not sufficient for visit-to-visit BPV Average real variability (ARV)

1n−1 ∑i=1n−1|xi+1−xi|

Short-term BPV
Mid-term BPV
Long-term BPV Weight for the between-reading time intervals Partially dependent on the overall BP level and change in mean BP levels overtime; takes the order of the BP measurements into account Variation independent of mean (VIM) kxSD/

x¯

m Long-term BPV No correlation with mean BP level over visits Cannot be compared across populations Range Maximum–minimum Short-term BPV
Mid-term BPV
Long-term BPV Widely used Influenced by extreme values

x¯

, mean; ∑, sum of; N, number of values in the population; x, each value. ARV, average real variability; BP, blood pressure; BPV, blood pressure variability; CV, coefficient of variance; RSD, residual SD; VIM, variation independent of mean.
SD

SD is the most used index for measuring BPV across studies. SD is correlated with mean BP values independent of measurement order and affected by extreme BP values triggered by environmental stressors [17] as well as by day-night changes in BP (e.g. night-time BP fall) [18] (Table 1). The weighted SD (for short-term BPV measurement only) could remove the influence of night-time BP fall on 24-h SD by weighting daytime and night-time BP–SD, respectively, and by averaging them [18].

Residual SD

A trend may exist when measuring BP over a long period (i.e. weeks or months). For example, the trend of BP will decrease over time with antihypertensive treatment in clinical trials. Under this situation, the extent of variability measured by SD may be exaggerated. If there is a linear relationship between time and BP, variability can be defined as the residual SD (RSD), which is calculated as the root-mean-square error of the differences between predicted BP and observed BP from the regression [19] (Table 1). RSD is less affected by changes in BP levels over time compared to SD [19]. The limitation of this measurement is the assumption of a linear trend, which may not apply to the nature of changes.

Coefficient of variation

As absolute values of BPV might be positively related to mean BP values, a mathematical correction made by the CV (CV = SD × 100/mean) is often applied to correct correlations between SD and mean BP values [16] (Table 1). However, in the case of visit-to-visit BPV, CV might still not be sufficient, as it was reported to be positively related to mean BP values in some cohorts [17,20].

Variation independent of mean

Variation independent of mean (VIM) is a transformation of SD [17,20], which removes the influence of mean BP on BPV with nonlinear regression analysis through a plot of SD (y-axis) against mean BP value (x-axis), for all individuals in the cohort. (i.e. proportional to SD/meanx, with x derived from curve fitting) [16] (Table 1). There is no agreement on the gold-standard approach to visit-to-visit BPV measurement. The goal of visit-to-visit BPV research is to determine the causal effects, and VIM is an ideal approach based on the previous studies. However, as parameter x is specific to each cohort, VIM cannot be compared across populations [19].

Average real variability

In addition to adjusting SD, indices unaffected by day–night changes or mean BP values are preferred. Average real variability (ARV) measures the overall variability based on the recording sequence, calculated as the mean absolute difference between consecutive measurements [19] (Table 1). It is an alternative for SD to measure variability in mean BP value [18]. ARV has been suggested to be more appropriate in measuring 24-h BPV and be more useful in predicting outcomes than SD. For example, evidence has shown that SD is similar in participants with different 24-h ABP profiles while ARV shows a difference [21]. Moreover, ARV offers a computationally simple way to assess variation with a trend [20]. For instance, ARV will be greater than SD if there is a tendency with an alternating pattern of increases and decreases in BP values between consecutive measurements [20].

Range

Range is calculated as maximum minus minimum BP values (Table 1), which is quantifying for short-term, mid-term, and long-term BPV [16,22]. Range is influenced by extreme BP values [22], and thus, is more varied. It is also highly related with SD and CV.

Some argue that BPV indices convey redundant and overlapping information [22]. For example, the most used indices in the BPV studies, such as SD, CV, and VIM, are correlated with a high agreement in measuring visit-to-visit BPV. Indeed, all three indices belong to the dispersion type, which assesses the variability around the mean BP value across visits. Thus, calculating one of these indices is sufficient, and repeating three does not add information but would give a false sense of pseudo-consistency [22], thereby should be avoided. Although in the stability type, indices, including range, maximum, peak size, and trough size are more varied and are influenced by extreme values. Recently, one study has introduced a new index, random slope, which measures BPV at unequal intervals between visits [8]. Random slope showed consistent correlations with traditional BPV indices in assessing associations between BPV and cognitive functions, validating it as an effective index of BPV. In addition, because traditional BPV indices are limited to BP measurements at specific predetermined intervals, the introduction of random slope could allow clinical studies assessing BPV utilizing data collected at irregular intervals and account for the varying intervals of measurements across patients [8]. Lacking consensus on optimal BPV indices and measurement methods, such as the number and the interval time of measurements, leads to great heterogeneity in exploring the relationship between BPV and cognitive impairment and dementia. Thus, it is critical to determine the most reliable indices and record BP values accurately and consistently among studies.

THE IMPACT OF BLOOD PRESSURE VARIABILITY ON COGNITION IN COGNITIVELY NORMAL PEOPLE

Mounting evidence has shown that elevated BPV (mostly systolic) is a risk factor for cognitive decline and dementia in cognitively normal people [4,6–7,9,11–12,23–41]. Although a few studies showed contradictory results [8,42–43]. For example, one retrograde study with 94 African-American reported that BPV did not correlate with global cognitive function, while higher DBP variability (DBPV) correlated with poorer verbal and incidental memory [8]. Another study reported that higher BPV contributed to better cognitive performance in older adults with cardiovascular disease (CVD) than those with stable BP [8]. However, these two studies have relatively small sample sizes of 94 and 97 participants. In addition, Tsang et al.[8] only measured the BPV with the three most recent clinic visits in African-American community-dwelling elderly. Another larger trial (Prevention of Dementia by Intensive Vascular Care, preDIVA trial with sample sizes of 2305 participants) found that high visit-to-visit BPV was associated with cognitive decline but not incident dementia among older people [43]. The inconsistent findings on the influence of BPV on cognitive impairment and dementia may be due to the insufficient attention to confounding factors, such as sex, age, race, and ethnicity, type of BPV, timeframes, and follow-up lengths. These abovementioned factors will be discussed below. Summary of the studies please see Table 2.

TABLE 2 - Summary of the relationship of blood pressure variability and cognitive function in cognitive normal individuals. Author year Age, year avg ± SD sample size Male% Region Ethnics Health status (database) HT treatment % BP index BP time point Follow-up duration Cognition Methods Cognition time point Adjustment comorbidity factors Results Alpérovitch 2014 [6] 73.7 ± 5.2 6506 38 France NA Well functioning volunteers, the three-city study No dementia group: 52.8; Incident dementia group: 47.1 CV 2 years, 4 years Median 6.8 years; 8 years (1999–2001–2008) MMSE, IST, DSM-IV 2,4 and 7–8 years Sex, study center, education, diabetes, history of vascular diseases, antihypertensive drug use, mean BP BPV was associated with an increased risk of incident dementia, whereas mean BP was not Böhm 2015 [7] ≥55 24 593 NA Worldwide NA W.o. demntia and CVD. ONTARGET and TRANSCEND trials NA CV 6 weeks, 6 months, and every 6 months thereafter 56 months MMSE Baseline, after 2 years, and 3–5 years Baseline MMSE, DBPV, age, BMI, estimated glomerular filtration rate, sex, ethnicity, physical activity, formal education, alcohol consumption, stroke, diabetes mellitus, and concomitant medications SBP-CV and mean HR are independent predictors of cognitive decline and cognitive dysfunction in patients at high CV risk Cho 2018 [23] 77.7 ± 8.3 232 33.6 Japan Asian Ambulatory patients with one or more CV risk factors 85.3 wSD Every 30 min for 24 h 24 h MoCA-J NA Age and 24-h SBP In elderly patients with well ambulatory BP control, higher BP variability but not average ambulatory BP level was associated with cognitive impairment de Havenon 2021 [58] 67.9 ± 9.3 8379 64.9 US White 58.6%, Black 29.3%, Hispanic 10.4% SPRINT MIND trial NA SD First 600 d, 7.8 measurements 3.2 ± 1.4 years NA 2 years, 4 years, then 1 year thereafter Age, sex, race/ethnicity, history of CVD, hypertension, education, level of physical activity, smoking, SPRINT randomization arm, No. of BP measurements, and mean SBP Higher BPV was associated with the development of probable dementia with excellent BP control Dore 2018 [24] 62.0 ± 12.8 980 41.4 US White, black Community-dwelling individuals, the Maine-Syearacuse Longitudinal Study 51.2 SD 15 times (5 times each sitting, reclining, and standing) in 2001–2006 5 years Neuropsychological test battery Following BP assessment Age, sex, education, ethnicity, mean BP, diabetes, BMI, TC, smoking, and alcohol consumption BPV is an important predictor of cognition with advancing age Ernst 2021 [25] ≥65 19 114 NA Australia and US White, black, Hispanic ASPREE trial (community-dwelling free of dementia and CVD) NA SD, CV, ARV Annually 4.7 years MMSE Annually Age, SBP, AHT medications, education, diabetes, depression, BMI, and statin medications. High BPV in older adults without major cognitive impairment, particularly men, is associated with increased risks of dementia and cognitive decline Fujiwara 2018 [26] 83.2 ± 3.2 497 44.3 Japan Asian W.o. CVD Working memory impairment: 60.6; without working memory impairment: 66.8 Short-term: SD, CV, wSD of 24-h SBP and DBP; long-term: SD, CV, MMD Office: baseline and each office visits for 1 year; Ambulatory: every 30 min for 24 h 1 year MMSE NA Age Exaggerated BPV was significantly associated with working memory impairment in very elderly individuals Godai 2020 [27] 85–87 111 47.7 Japan Asian SONIC study, community-dwelling oldest-old population 64 CV HBP every morning 30 days MoCA-J Entry and 30 d Sex, the corresponding mean HBP, antihypertensive medication, diabetes mellitus, history of arrhythmia, WHO-5, and gait speed In the community-dwelling oldest-old population, higher day-to-day HBPV, but not the value of HBP, was associated with cognitive impairment Keary 2007 [42] 69.8 ± 7.5 97 57.7 US NA Nondemented older adults with CVD NA SD NA NA MMSE NA Age, education Greater BPV was associated with better, not poorer, cognitive test performance Li 2021 [28] HRS cohort: 66.3 ± 8.0; ELSA cohort: 62.4 ± 8.9 12 298 HRS cohort: 40.4; ELSA cohort: 43.8 NA White Dementia free, HRS & ELSA study NA SD, CV, VIM Wave 1 (2002–2003), to wave 9 (2018–2019), at least 3 waves HRS cohort: 8–12 years; ELSA cohort: 10–14 years Standardized Z score of cognitive function NA Age, sex, race, education, cohabitation status, smoking, alcohol consumption, physical activity, BMI, mean BP, depressive symptom, history or presence of CVD, diabetes, lung disease and cancer, AHT medication Higher long-term BPV was associated with accelerated cognitive decline among general adults aged ≥50 years, with nonlinear dose-response relationship Ma 2019 [9] 67.6 ± 8 5273 41.9 Netherland NA Dementia-free 29.3 CV, SD, ARV Every 4 years Median 14.6 years MMSE, geriatric mental schedule, DSM-III-R, ADRDA, AIREN 5, 10, or 15 years Age, sex, education, APOE genotype, vascular risk factors, and history of CVD A large BPV over a period of years was associated with an increased long-term risk of dementia. The association between BPV and dementia appears most pronounced when this variation occurred long before the diagnosis. An elevated long-term risk of dementia was observed with both a large rise and fall in BP Ma 2021 [29] 71.4 ± 6.4 1835 37 Netherland NA Adults from The Rotterdam Study 31.2 BP complexity: sample entropy; BPV: CV of beat-to-beat BP NA 20 years MMSE, GMS, DSM-III-R, ADRDA NA Age, sex, APOE genotype, mean SBP, and other confounding factors. Lower complexity and higher beat-to-beat SBPV are potential novel risk factors or biomarkers for dementia Matsumoto 2014 [30] 63.3 ± 4.7 485 28 Japan Asian Ohasama study w.o. cognitive decline at the enrollment 30 SD HBP, every morning for 4 weeks 7.8 years MMSE 4 years apart Sex, age, history of CVD, low level of education, baseline MMSE score <27, and follow-up duration A significant association between day-to-day HBPV and cognitive decline independent of SBP McDonald 2017 [31] 72 353 58 England NA Community-dwelling cohort NA CV ABP 5 years MMSE, Cambridge Cognitive Examination Baseline and at 5 years Age, sex, and education Greater daytime BPV was associated with poorer cognitive function while night-time BPV was not associated with cognitive function at baseline or cognitive decline Nagai 2012 [4] 79.9 ± 6.4 201 25 Japan Asian Elderly patients at high risk of CVD 71.2 SD, CV, max BP, min BP, delta BP 12 visits once a month 12 months MMSE, GDS 3 months after Apr 2007 Age, calcium channel blockade use, low density lipoprotein, SBP In the high-risk elderly, exaggerated visit-to-visit BPV was significant indicators for cognitive impairment Nagai 2014 [32] 79.9 ± 6.4 201 24 Japan Asian 3SCO study, high risk of CVD 71.2 CV, delta BP Monthly 12 months MMSE, GDS 3 months after Apr 2007 Age, calcium channel blockade use, low density lipoprotein, average HR, and average SBP Exaggerated visit-to-visit SBPV and advanced carotid artery remodeling (high IMT and high stiffness parameter b) have a synergetic association with cognitive dysfunction Ogliari 2016 [33] 75.2 ± 3.3 4745 47.4 Netherland NA High CV risk, In PROspective Study of Pravastatin in the Elderly at Risk 40.2 SD Every 3 months during the first 18 month 3.2 years The Lawton-Brody activities of daily living scale, instrumental activities of daily living First at 18 months and then during follow-up until 48 months Age, sex, country, education, CV risk factors (smoking, BMI, hypertension and diabetes), CV morbidities (myocardial infarction, stroke/transient ischemic attack, claudication and glomerular filtration rate), use of AHT, statin treatment, mean SBP or DBP, No. of measurements of BP Higher visit-to-visit SBPV but not DBPV was associated with steeper functional decline in older adults at high cardiovascular risk Oishi 2017 [34] 71 ± 7 1674 44.1 Japan Asian Community-dwelling elderly w.o. dementia, Hisayama study 43.3 CV HBP, every morning for 4 weeks 5.3 years MMSE, HDS, HDS-R 2005–2006, 2012–2013 Age, sex, low education, use of AHT agents, ECG abnormalities, diabetes, serum total cholesterol, BMI, history of CVD, smoking, alcohol, and regular exercise Increased day-to-day BPV is an independent significant risk factor for the development of all-cause dementia, vascular dementia, and AD in the general elderly Japanese population Qin 2016 [35] 64 ± 6 976 48 China Asian Community-dwelling older adults, prospective cohort study, China Health and Nutrition Survey 12 SD, CV, VIM 3 or 4 visits 3.2 years Cognitive screening test ≥2 visits in 1997, 2000, or 2004 Age, sex, education, time, smoking physical activity, ever used antihypertensive treatment, mean SBP Higher long-term visit-to-visit BPV is associated with a faster rate of cognitive decline among older adults Rouch 2020 [11] 77.7 ± 6.2 3319 43.5 France NA Noninstitutionalized patients from S.AGES cohort 70.8 SD, CV, ARV, VIM, RSD Every 6 months 3 years MMSE, DSMMD Every 6 months Age, sex, educational level, SBP or DBP or MAP or pulse pressure, AHT drug use, coronary artery disease, type 2 diabetes, chronic heart failure, AF, transient ischemic attack or stroke, smoking and dyslipidemia at baseline Higher BPV is associated with poorer cognitive function and incident dementia, independent of mean BP Sabayan 2013 [36] 75.3 ± 3.3 5461 48.3 Netherland NA At risk of CVD NA SD Every 3 months 3.2 years Selective attention, processing speed, immediate and delayed memory Every 3 months Age, sex, BMI, statin treatment, smoking, cholesterol level, history of vascular diseases, history of hypertension, history of diabetes mellitus, and average BP measures Higher visit-to-visit BPV independent of average BP was associated with impaired cognitive function in old age Sakakura 2007 [37] Younger elderly: 71.9 ± 4.5; very elderly: 84 ± 3.9 202 Younger elderly: 22.8; very elderly: 18.8 Japan Asian 101 very elderly (≥80) & 101 younger elderly (61–79) Younger Elderly: 78.2; Very Elderly: 73.3 SD 24-h ABP 24 h MMSE, short-form 36 items health survey NA NA Very elderly had larger BPV than younger elderly. Exaggerated ABPV was related to cognitive dysfunction in the elderly, especially in the very elderly, and was related to lower QOL in the younger elderly Tadic 2019 [38] 63 ± 5.7 471 53 Italy NA PAMELA study 31.4 SD ABP: every 20 min for 24 h; Long-term: NA 10 years MMSE Entry and at the end NA Individual residual SBPV and DBPV gradually decreased with the increase in MMSE score Tsang 2017 [8] 69.2 ± 6.8 94 62.7 US Black Normal African Americans NA Range, SD, CV, random slope 3 most recent clinic visits NA MMSE, computer assessment of MCI NA Age, sex, education, medical conditions (diabetes, hypercholesterolemia, obesity, and stroke) In a sample of cognitively intact older African American adults, BP variability did not correlate with global cognitive function, as measured by the MMSE. However, higher diastolic BP variability correlated with poorer verbal and incidental memory Van Middelaar 2018 [43] 74.2 ± 2.5 2305 44.8 Netherland White Community preDIVA trial 54.1 CV, SD, ARV, delta BP Every 2 years 6.4 years MMSE, DSM-IV Every 2 years