Long-Term Blood Pressure Patterns in Midlife and Dementia in Later Life: Findings from the Framingham Heart Study

Abstract

INTRODUCTION:

Long-term blood pressure (BP) measures, such as visit-to-visit BP variability and cumulative BP, are strong indicators of cardiovascular risks. This study modeled up to 20 years of BP patterns representative of midlife by using BP variability and cumulative BP, then examined their associations with development of dementia in later life.

METHODS:

For 3,201 individuals from the Framingham Heart Study, multivariate logistic regression analyses were performed to examine the association between long-term BP patterns during midlife and the development of dementia (ages≥65).

RESULTS:

After adjusting for covariates, every quartile increase in midlife cumulative BP was associated with a sequential increase in the risk of developing dementia (e.g., highest quartile of cumulative SBP had approximately 2.5-fold increased risk of all-cause dementia). BP variability was not significantly associated with dementia.

DISCUSSION:

Findings suggest that cumulative BP over the course of midlife predicts risk of dementia in later life.

INTRODUCTION

Dementia is the most common neurodegenerative disorder leading to increased risk of disability, burden of illness, and costs. About 55 million older adults are impacted by dementia globally, and this number is projected to triple by 2050 with a rapid increase in the aging population, making prevention and treatment for dementia a public health priority [1–3].

While effective intervention for dementia is still lacking, research shows that elevated blood pressure (BP) is a modifiable risk factor for dementia[4, 5]. Nonetheless, various studies examining the effect of anti-hypertensive medication have produced inconsistent findings related to their protective effect on the development of dementia[6–8]. Such discrepancies may result from measurement challenges, including the lack of longitudinal intra-individual measurements over one’s life course. A single or an average measure of BP does not accurately capture long-term BP patterns due to their fluctuating nature[9–12] but they have been widely used to represent BP due to difficulties associated with longitudinal examination of BP. Measures that better take into account an individual’s long-term BP patterns have been attracting attention in investigating the effect of BP on various clinical and functional outcomes[13–16].

Fine grained (beat-to-beat or frequently sampled, e.g., every 10 minutes) long-term BP or ambulatory BP recordings offer a new richness of data but long-term impact will take time to ascertain. Though sparsely sampled, what is currently available is visit-to-visit BP variability (BPV), which has gained attention as a better integration of the total load of hemodynamic risk markers, compared to single BP measurements[16–18]. Both short-term and long-term variability have been associated with important prognostic implications, including cardiovascular health, mortality, depression, and cognitive disorders[11, 16, 19–21]. It is postulated that higher BPV could lead to cognitive dysfunction [17, 22, 23] and the development of AD[16, 24, 25] by contributing to arterial stiffness and altering cerebral blood flow[22, 26, 27].

Another novel approach to BP quantification includes cumulative exposure measures, which is a summary measure of cumulative burden of high BP that may capture the effects of these risk factors over several decades[28–30]. A critical advantage of a cumulative measure over a single- or an average measure of BP is that it may reflect the total duration of an individual’s history of high BP more accurately, therefore allowing for examination of the effect of chronic exposure to a cardiovascular risk factor[14]. Extensive studies on cumulative BP have been conducted from the CARDIA study, which found that high cumulative BP from young adulthood to middle age was associated with cardiovascular and cognitive risks in midlife[14, 29, 31], Other longitudinal studies in older adults demonstrated that a high cumulative SBP was associated with accelerated rates of brain atrophy and infarcts[32–35], as well as cognitive decline and elevated dementia risk[36].

There is considerable evidence to suggest that BPV and cumulative BP metrics could provide important clinical implications on dementia risks, but no study has applied these methods in 20+ years of long-term BP indices spanning midlife. Understanding the role of midlife BP patterns in the neurodegenerative process is especially important given that high BP in midlife is particularly a risk for dementia[37–39]. Additionally, given that neurodegenerative processes occur decades prior to symptom onset and that cognitive impairment resulting from dysregulated BP could begin as early as in midlife[32, 40], understanding the role of midlife BP patterns on brain health at late life is particularly important in understanding preventative methods for dementia.

The Framingham Heart Study (FHS) has tracked BP over several decades. The current study used data from the FHS to model BPV and cumulative BP over the course of up to 22 years of midlife, then investigated their association with the development of all-cause dementia and Alzheimer’s disease (AD) in later life. Based on previous studies linking high and irregular BP patterns to adverse brain pathology, we hypothesized that increased variability and chronic elevation of BP during midlife would be associated with an increased risk of developing dementia.

METHODS

Study Design and Participants

The FHS is a prospective epidemiological study of community-dwelling adults. The selection criteria and detailed study design have been described previously[41]. Briefly, 5,209 participants were recruited in 1948 into the Original cohort and underwent systematic biennial examinations. In 1971, children of the Original cohort and their spouses were enrolled into the Offspring cohort and underwent routine examinations on average about every 4 years. Individuals from both Original and Offspring cohorts were included in this study.

The conceptualization and terminology of longitudinal modeling have been presented in a previous study from the FHS[42]. Longitudinal patterns of BP were determined using participants’ BP measurement between ages 40–64 (“observation phase”) (median observation phase= 16.39 years) (Figure 1).

Figure 1. Figural Representation of Long-Term Blood Pressure Modeling.

Blood pressure (BP) modeling is based on the observation phase if participants attended all 5 examinations. For the Original Cohort, the observation phase began at the fourth examination, and for the Offspring Cohort, it began at the first examination. The hinge examination was the last examination within the observation phase. The follow-up phase began at the hinge examination and was used to determine the presence of dementia late life.

The baseline examination was defined as the first observation point. For the Original cohort, the earliest examination for the observation phase was set to the fourth examination cycle (1954–1958), when data on use of anti-hypertensive medication were first reliably collected. The first examination was used as the earliest BP observation examination on the Offspring cohort. Five BP measurements during the observation phase were used for BP modeling. The presence of incident all-cause dementia and Alzheimer’s disease was determined during the subsequent “follow-up phase”, in which the time origin for time-to-event analysis in this study was determined as participants reaching age 65 or more. The events of interest are all-cause dementia and Alzheimer’s disease, and the follow-up window was up to 42 years (median 17 years).

A total of 3,712 individuals had available data to model long-term BP in midlife (attended all observation examinations and were in ages 40- ≤65 during observation phase), but participants were excluded from analyses if they developed dementia (n=9), stroke (n=38), or brain tumor (n=1) prior to or during the observation phase (i.e., prevalent or interim onset between first and last BP observations). Individuals were also excluded if the last documented date of intact cognition preceded the last observation exam (n=30) or if they were missing information on education attainment (n=433). Finally, analyses were performed on 3,201 individuals (1,251 from Original and 1,950 from Offspring cohorts). The flow chart describing participant inclusion/exclusion is in Figure 2.

Figure 2. Flowchart of the study sample.

The process of participant selection and exclusions are presented.

All participants gave informed consent. The study protocol was approved by the Institutional Review Board of the Boston University Medical Center. This study is reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Modeling Long-Term Blood Pressure Patterns

Participants’ BP measurements were obtained during routine physical examinations, along with the medical history, at each FHS cycle examination. Two seated SBP and DBP were measured after a 5-minute rest period, and the average of each was taken as the BP measurement for the given examination. All BP measurements were conducted by a physician using a mercury column sphygmomanometer and a standard protocol.[43]

Blood Pressure Variability (BPV).

BPV was presented as standard deviations (SD) of BP measurements across those five examinations. SD’s of SBP and DBP were calculated separately and were presented as SBP variability (SBPV) or DBP variability (DBPV). SD was selected as a BPV metric for this study given its interpretability and generalizability to other samples (unlike variability independent of the mean (VIM) that is tied to the population being examined).[44] SBP- and DBP SDs were moderate-strongly associated with other BPV metrics commonly used in research, such as coefficient of variance and average real variability (rs~0.20–0.97, ps<0.0001), and weaker correlation was noted with VIM (e.g., for SBP SD, r=0.03, p=0.16).

Cumulative Blood Pressure.

Cumulative BP was calculated based on five examinations spanning up to a 22-year observation period, and was defined as the summed average BP for each pair of consecutive examinations multiplied by the time between these two consecutive visits in years:

$$[\frac{{\mathrm{B}\mathrm{P}}_1+{\mathrm{B}\mathrm{P}}_2}{2}\times time1-2]+[\frac{{\mathrm{B}\mathrm{P}}_2+{\mathrm{B}\mathrm{P}}_3}{2}\times time2-3]+[\frac{{\mathrm{B}\mathrm{P}}_3+{\mathrm{B}\mathrm{P}}_4}{2}\times time3-4]+[\frac{{\mathrm{B}\mathrm{P}}_4+{\mathrm{B}\mathrm{P}}_5}{2}\times time4-5]$$

${\mathrm{B}\mathrm{P}}_1,{\mathrm{B}\mathrm{P}}_2,{\mathrm{B}\mathrm{P}}_3,{\mathrm{B}\mathrm{P}}_4$, and ${\mathrm{B}\mathrm{P}}_5$ indicate $\mathrm{B}\mathrm{P}$ measurements at examinations 1 (baseline), 2, 3, 4, and 5. $\mathrm{T}\mathrm{i}\mathrm{m}\mathrm{e}1-2,\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}2-3,\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}3-4$, and $\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}4-5$ indicate the participant-specific time interval between consecutive examinations 1–5 in years.

Screening and Diagnosis of Dementia

Dementia diagnosis consensus meetings were held regularly, with a panel of at least one neurologist and one neuropsychologist reviewing each case of possible dementia. When available, data from FHS neuropsychological test performance, neurologist’s examination, as well as hospital records and information from primary care physicians, family interview were used to determine clinical diagnosis. Criteria from the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (American Psychiatric Association, 1994) were used to define dementia. Severity was defined by a Clinical Dementia Rating scale equivalent of one or greater, as well as presence of symptoms of dementia for a period of at least 6 months. Presence of AD was determined using the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association (NINCDS-ADRDA) criteria for definite, probable, or possible AD[45].

Covariates

Demographic characteristics, lifestyle factors, and medical histories collected during the baseline examination and the hinge examination were used as covariates. Education was grouped into <high school degree, any high school through college degree, or >college degree. Height and weight were measured during the routine physical examination, and body mass index (BMI)(kg/m2) was calculated. However, because height was measured irregularly (e.g., exams 5, 10, 11) in the Original Cohort, average height computed for the Observation Phase was used to calculate baseline BMI when baseline height was missing. Coronary heart disease (CHD) was determined by medical history, medical record review, and physical examination, and the presence or the development of CHD during the observation phase was used as a covariate as a time-dependent variable. Cumulative percentage of time on anti-hypertensive medication was calculated to reflect cumulative exposure to hypertension treatment ((# of exams with medication use / # of attended observation examination) x 100). ApoE genotype was analyzed as a dichotomous variable to represent the presence or absence of an ApoE ε4 allele (ε2/ε4, ε3/ε4, ε4/ε4).

Statistical Analysis

Multivariate logistic regression was performed to examine the association between BP variables and dementia outcomes (binary variables). The risk of all-cause dementia and Alzheimer’s disease (AD) in relation to BP indices was evaluated using multivariable logistic regression analyses, adjusting for the effect of the potential confounding variables (age, self-reported sex, education, CHD, ApoE ε4, and anti-hypertensive medication use. For analyses with BPV, a linear temporal trend in BPV was calculated with BP values regressed across the visits (BPreg), and these variables were included as a covariate to control for the direction of change in BP, along with the average BP values (BPmean). Time-dependent variables, including SD, BPreg, BPmean, presence/development of CHD, and cumulative percentage of time on anti-hypertensive medication use was based on measures through the observation phase, while a participant remained at risk. Findings were presented for every SD increase in BPV and Cumulative BP to facilitate interpretation. Results reflect odds ratio (OR) and 95% confidence interval (CI).

To examine potential threshold effects, BP indices were categorized into quartiles. SBPV groups were created by determining quartiles (SBP SD <6.76, 6.76- <9.48, 9.48-<13.01, and ≥13.01). DBPV was defined as the SD of the DBP measurements across the observation examinations, and DBPV quartile groups were defined as DBP SD<4.34, 4.34-<5.94, 5.94-<7.98, and ≥7.98). Similarly, the systolic and diastolic counterparts for cumulative BP were stratified by quartiles. CSBP was categorized into Q1 (<1872.01 mmHg x Year), Q2 (1872.01 - <2083.26 mmHg x Year), Q3 (2083.26 - <2306.13 mmHg x Year), and Q4 (≥2306.13 mmHg x Year). CDBP groups were also created by quartiles: Q1 (<1158.18 mmHg x Year), Q2 (1158.18 - <1281.36 mmHg x Year), Q3 (1281.36 - <1423.64 mmHg x Year), and Q4 (≥1423.64 mmHg x Year). CSBP was categorized into Q1 (<1872.01 mmHg x Year), Q2 (1872.01 - <2083.26 mmHg x Year), Q3 (2083.26 - <2306.13 mmHg x Year), and Q4 (≥2306.13 mmHg x Year). CDBP groups were also created by quartiles: Q1 (<1158.18 mmHg x Year), Q2 (1158.18 - <1281.36 mmHg x Year), Q3 (1281.36 - <1423.64 mmHg x Year), and Q4 (≥1423.64 mmHg x Year). The associations between the quartiles of BP indices and the risk of dementia were assessed using the multivariate logistic regressions with abovementioned covariates. The Q1 group was used as a reference group, and results were presented for Q2-Q4 as OR and 95% CI relative to the Q1.

P<0.05 was considered statistically significant. All analyses were performed using SAS software version 9.4 (SAS Institute Inc, Cary, NC).

RESULTS

Demographic characteristics of the study sample are presented in Table 1. All participants from Original and Offspring cohorts were Non-Hispanic White[46], and self-reported sex was evenly distributed (55.23% women) in this sample. All-cause dementia was present in 406 (12.68%) of our sample, and among them, 283 (69.7%) had Alzheimer’s disease.

Table 1.

Demographic/Clinical Characteristics of the Study Sample at Baseline and Hinge Examinations

	Baseline Exam	Hinge Exam
Age (mean±SD)	45.91±2.37	62.39±1.75
Sex (Female)	1768 (55.23%)
BMI (mean±SD)	26.14±4.36	27.70±4.97
Education (n, %)
<High school degree	452 (14.12%)
High School Graduate	1096 (34.24%)
College Degree	664 (20.74%)
>College graduate	989 (30.90%)
Race (n, %)
Non-Hispanic Whites	3201 (100%)
Current Smoking (n, %)	1028 (35.01%)	483 (16.47%)
Hypertension (n, %)	561 (17.60%)	1399 (43.73%)
Anti-Hypertensive Medication (n, %)	194 (6.10%)	1091 (34.20%)
Cumulative Use of Anti-Hypertensive Medication (%, mean±SD)	17.43 ±27.47
Diabetes medication (n, %)	16 (0.50%)	179 (5.64%)
Total cholesterol (mean±SD)	216.15±41.99	207.78±41.28
ApoE ε4 carrier (n=2,762)(n, %)	734 (22.93%)
SBP, mmHg (mean±SD)	123.74±16.66	130.47±18.27
DBP, mmHg (mean±SD)	79.92±10.51	76.55±9.58

Note: Values are presented as mean±SD for continuous variables and n (%) for categorical variables. Categorical variables are presented as n,% and continuous variables are presented as mean±SD.

SD=standard deviation; BMI=body mass index; SBP=systolic blood pressure; DBP=diastolic blood pressure

BPV and Dementia Outcomes

BPV variables were examined as continuous variables, and their univariate and multivariate associations with dementia outcomes were assessed along with their covariates (Table 2). Among SBP variables, SBPmean, but not SBPV or SBPreg, was significantly associated with all-cause dementia (SBPmean OR=1.02, P<.0001). Only DBPmean was associated with all-cause dementia, among DBP variables (DBPmean OR=1.03, p<.0001). Similar results were observed for the development of AD: Higher SBPmean and DBPmean were associated with greater risk of AD (OR=1.02, p<.0001 for SBPmean and OR=1.04, p<.0001 for DBPmean), but other BP variables were not significantly associated with AD.

Table 2.

Univariate and Multivariate Associations between BPV as continuous variable and Dementia

	All-Cause Dementia					Alzheimer’s Disease
	Unadjusted		Adjusted a			Unadjusted		Adjusted a
	OR	95% CI	OR	95% CI		OR	95% CI	OR	95% CI
SBP SD	1.06	(0.96–1.17)	0.98	(0.86–1.12)		0.99	(0.86–1.12)	0.95	(0.81–1.11)
SBPmean	1.01	(1.007–1.02)	1.02	(1.01–1.03)		1.007	(0.99–1.02)	1.02	(1.006–1.03)
SBPreg	1.02	(0.79–1.35)	0.71	(0.52–0.98)		1.08	(0.78–1.48)	0.75	(0.51–1.11)
	All-Cause Dementia					Alzheimer’s Disease
	Unadjusted		Adjusted a			Unadjusted		Adjusted a
	OR	95% CI	OR	95% CI		OR	95% CI	OR	95% CI
DBP SD	0.95	(0.85–1.05)	0.97	(0.86–1.10)		0.90	(0.79–1.02)	0.95	(0.82–1.10)
DBPmean	1.02	(1.003–1.03)	1.03	(1.02–1.05)		1.01	(0.99–1.03)	1.04	(1.01–1.05)
DBPreg	1.06	(0.74–1.53)	0.95	(0.59–1.52)		1.28	(0.84–1.97)	0.97	(0.55–1.68)

When BPV groups were divided into quartiles to examine any threshold effects, no group differences were observed in the development of dementia or AD across SBPV and DBPV groups (Supplemental Table 1).

Cumulative BP Measures and Dementia Outcomes

Every SD increase in CSBP was associated with 50% increased odds of developing all-cause dementia and 45% increased odds of developing AD (Ps=<.0001) (multivariate ORs 1.50 and 1.45, respectively) (Supplemental Table 2). Similarly, every SD increase in CDBP was associated with 58% increase in developing all-cause dementia and more than 71% increase in developing AD (Ps<.0001).

When the proportion of dementia was compared across the quartiles of CSBP, increased prevalence of dementia was observed with increasing quartiles (CSBP Q1<Q2<Q3<Q4) (Figure 3A). In particular, the highest quartile (Q4) had the highest odds of developing dementia, compared to Q1 (OR=2.58, P<.0001). CDBP showed a similar relationship, where Q4 was associated with the highest prevalence of dementia (OR=2.72, p=<.0001), followed by Q3 and Q2, compared to Q1 (Figure 3B). Similar trend was observed when AD was examined as an outcome, with highest quartiles of CSBP and CDBP having greatest odds of developing AD (OR=2.45, P<.0001 and OR=3.42, P<.0001, respectively).

Figure 3. Association Between Dementia and Cumulative SBP (Figure 3A) and DBP (Figure 3B) groups.

The x-axis represents different quartiles of cumulative systolic blood pressure (CSBP)(Figure 3A) and cumulative diastolic blood pressure (CDBP)(Figure 3B). The y-axis represents odds ratio of developing all-cause dementia (front row) and Alzheimer’s disease (back row).

Sensitivity Analyses

Additionally, we repeated analyses by self-reported sex, and findings were comparable in males and females (e.g., OR of developing all-cause dementia was 2.65 in Q4 vs. Q1 for men and 2.30 for women, ps<0.0001). BPV was not significantly associated with dementia outcomes (Ps≥0.28). We also repeated analyses by the presence of hypertension at baseline. In individuals who had hypertension at baseline, being placed in the highest quartile of CSBP predicted a four-folds risk of developing all-cause dementia compared with the lowest quartile (OR=3.85, P=0.03). However, the 2^nd or 3^rd quartile groups were not significantly different compared with the lowest quartile. In those without baseline hypertension, findings were comparable with those from the entire sample (e.g., OR 2.54, P<.0001) in reference to Q1). Dementia outcomes were not significantly different by BPV metrics in separate analyses by hypertension groups (Ps≥0.19).

We have also performed additional analyses with additional adjustment of baseline smoking status (i.e., current smoking) and baseline cholesterol level in a subsample of 2,919 individuals who had complete data. These analyses yielded consistent findings, indicating that higher CSBP and CDBP were significantly associated with increased risk of all-cause dementia (highest quartile vs. lowest quartile OR=2.53 (95% CI 1.66–3.85) for CSBP and OR =.85 (1.85–4.40) for CDBP) and AD (highest quartile vs. lowest quartile OR=2.85 (1.85–4.40) and OR=3.64 (2.18–6.09)). BPV was consistently not associated with dementia outcomes even after adjusting for smoking status and cholesterol.

Lastly, we have further examined BPV with tertiles of SD and did not find any significant differences in outcomes (e.g., the risk of developing all-cause dementia in the 3^rd tertile of SBP SD vs. 1^st tertile was not statistically significant, with OR of 0.85 and 95% CI 0.75–1.29).

DISCUSSION

The current study was conducted with the aims of examining the association between midlife BP and the risk of dementia using both BPV and cumulative BP to model long-term BP patterns. Our results indicate that while higher cumulative SBP and DBP were strong predictors of developing all-cause dementia and AD in late life, BPV was not associated with increased dementia risk. We also noted that those who were already hypertensive, and in the highest quartile of CSBP, had the greatest risk, which may reflect the effect of a longer (prior to sampling) exposure to high BP. The lack of impact at the 2^nd and 3^rd quartile level in those with prior hypertension may reflect successful autoregulatory compensation[47].

The role of cumulative BP on brain health has been studied extensively in the CARDIA study. Their findings consistently demonstrate that elevated CBP from early adulthood to middle age was associated with poor cognitive function[48], increased white matter hyperintensity[49], along with negative cardiovascular health outcomes[14, 50]. Separate studies also demonstrated that elevated BP exposure in midlife[29, 30] and across the lifespan (ages 5–95)[31] were associated with faster cognitive decline. While these findings provide strong evidence to suggest that elevated CBP in older age may predict increased risks for dementia, this hypothesis was confirmed through our findings that revealed significant associations between high midlife CBP and increased dementia and AD risks in later life. While our findings are novel, they are in line with previous literatures on midlife BP that highlighted the negative impact of elevated BP on dementia risks[39, 51, 52].

The mechanism linking cumulative exposure to elevated BP and cognitive function can be postulated from longitudinal neuroimaging studies that demonstrated the impact of high BP on accelerating the risk of structural brain changes, including both atrophy and infarcts that lead to cognitive impairment[32, 33]. Chronically elevated BP leads to blood vessel wall thickening and reduced luminal diameter in microvessels and large cerebral arteries through plaque build-up.[14] Consequently, there could be a reduction in cerebral blood flow, resulting in infarction of the cerebral tissues, which in imaging, present lower total brain volumes, increased WMH and abnormal white matter diffusion signals, and enlarged ventricles[53, 54]. In particular, increased WMH may mediate cognitive impairment associated with high cumulative BP in middle-aged adults, as suggested by a recent study[55]. Lower functional connectivity in the hippocampus, which is a well-established region of AD pathology, was also associated with elevated cumulative BP in middle-aged adults[56].

Hypertension also impairs cerebrovascular reactivity[57–59]. Flow of cerebrospinal fluid through perivascular spaces in the brain is important for clearance of metabolic waste, and this is reduced in hypertension[60]. The glymphatic system aids in removal of waste products including amyloid β from the brain. Using a rat model of hypertension, cerebrospinal flow and glymphatic influx and efflux rates tracked with dynamic contrast-enhanced MRI showed compromised transport in both early and advanced stages of hypertension[61]. An experiment using the same model and a fluorescent tracer showed consistent results[62]. Vascular dysfunction, characterized by endothelial dysfunction, low-grade inflammation and structural remodeling, plays an important role in the initiation and maintenance of essential hypertension, and can all impact brain function[63]. Contrary to our initial hypothesis, higher BPV across midlife was not associated with the development of dementia in later life. Previous studies in older adults found that higher BPV was associated with clinical and biomarker correlates of dementia, including poorer cognitive performance [18, 26, 64], medial temporal volume loss[16], and CSF biomarkers (phosphorylated tau, total tau, and amyloid)[16]. Our results also more directly contradict studies that found a significant role of BPV in the development of dementia, including AD and vascular dementia[16, 18, 65–68].

Besides the sparse sampling inherent in the pattern of BP measurements in all the noted studies including ours, we believe that this inconsistency may be due to methodological discrepancies. Currently, there are over 36 different ways to measure BPV and there is a lack of consensus in measuring BPV, including the BP metrics (class, type, timing), the number of measurements, and examining BPV independent of mean BP[12, 69–71]. We defined BPV using SD of 5 visits spanning up to 22 years of midlife, and while this reflects variability over a longer observation period, previous studies have used data from shorter measure-to-measure intervals (e.g., 2 years). While our Offspring cohort was followed up every 4 years and shorter intervals could not be obtained with these data, we believe that shorter intervals and more visits may reveal a stronger association between BPV and dementia, as it did for findings from the Alzheimer’s Disease Neuroimaging Initiative, which used 3–4 BP measurements over the course of 12 months[16, 72].

Another potential explanation behind our null BPV findings include the age-related difference in the role of BPV. Previous studies that found significant relationship between BPV and dementia have focused on BPV in older adults; therefore, discrepancies in our findings could reflect different pathological processes and mechanisms in midlife BPV and dementia in late life. These different mechanisms may include age-related changes in BP and its synergistic effect with neurodegenerative processes. BP changes are age-dependent, and across the life span, SBP increases and DBP decreases toward late life, with advancing age[11]. BPV also increases with age and may represent more prominent clinical implications in later life[73]. In older adults, more advanced neurodegenerative process could also impact autoregulatory dysfunction, causing changes in BP[28, 65]. Therefore, the association between BPV and cerebrovascular dysfunction is suggested to be age-dependent, and BPV in midlife may be more subtle in its impacts compared to BPV in late life[13]. As such, current literatures highlight elevated BP in midlife to be specifically of risk to the aging brain[37, 39].

This study adds important information to the literature on midlife BP patterns and brain aging outcomes. It is the first to model BPV and Cumulative BP across multiple visits spanning decades of midlife and to examine their associations with the development of dementia. Along with these notable strengths of the current study, there are also limitations. First, the complexity of the current analyses made it difficult to thoroughly examine the role of anti-hypertensive medication in the link between BPV and brain outcome measures. In general, a large proportion of individuals diagnosed with hypertension exhibit poor adherence to anti-hypertensive medication, which could result in prolonged elevation of BP and fluctuations in visit-to-visit BP measurements[74]. Self-report data on anti-hypertensive medication and missing information on various classes of medications also limit our findings to examine differential effects on BPV[8]. Additionally, the current sample had lower overall variability of BP and lower overall mean BP, which may have obscured the examination of potential link between BPV and dementia outcomes. The study participants are all white, highly educated, and comparatively healthy with a low prevalence of cardiovascular risk at midlife. As a final point, the measured “variability” is based on a tiny fraction of blood pressure cycles, nor are important sleep and circadian effects considered.[75–77]

Conclusion

In summary, findings from the current study provide insight into the elevated risk of all-cause dementia and Alzheimer’s disease with prolonged and cumulative exposure to high BP in midlife. These findings suggest that controlling elevated BP in midlife may be an actionable step to reduce risk of dementia in later life. While BPV was not a significant predictor of dementia, further efforts to investigate this topic is warranted given a lack of consensus in methodology. Further research using more rigorous medication data may also elucidate the role of anti-hypertensive treatment in the association between cumulative BP and the brain functioning.

Table 3.

Association between Dementia and Cumulative BP as Continuous Variables

	CSBP (per SD)				CDBP (per SD)
	Model 1		Model 2		Model 1		Model 2
	OR	95% CI	OR	95% CI	OR	95% CI	OR	95% CI
All-Cause Dementia	1.27	(1.13–1.43)	1.50	(1.32–1.72)	1.35	(1.18–1.54)	1.58	(1.36–1.83)
Alzheimer’s Disease	1.18	(1.03–1.35)	1.45	(1.24–1.69)	1.37	(1.17–1.59)	1.71	(1.44–2.03)

Model 1: Adjusted for age, sex, and education; Model 2: Model 1 + additional adjustment of body mass index, coronary heart disease, ApoE e4 allele, and % duration of anti-hypertensive medication during observation phase (time-dependent variable)

BP=blood pressure; CSBP=cumulative systolic BP; CDBP= cumulative diastolic BP; OR=odds ratio, CI=confidence interval

Highlights.

• Long-term blood pressure (BP) patterns are strong indicators of vascular risks.

• Cumulative BP and BP variability were used to reflect BP patterns across midlife.

• High cumulative BP in midlife is associated with increased dementia risk.

• Visit-to-visit BP variability was not associated with the onset of dementia.