Change in blood pressure and incident dementia: a 32-year prospective study

Abstract

Studies of the association of high blood pressure [BP] to dementia are not consistent. Understanding long term trajectories in blood pressure of those who do and do not develop dementia can help clarify the issue. The Honolulu Heart Program/Honolulu-Asia Aging Study followed a cohort of Japanese American men for an average of 32 years, with systolic and diastolic BP measured at six examinations and dementia assessed at the final three. In an analysis of 1890 men who completed all six exams, 112 diagnosed with incident dementia at exam 6 were compared to the 1778 survivors without dementia. Trajectories in systolic and diastolic BP up to and including the sixth examination were estimated with a repeated measures analysis using 3 splines. From mid- to late-life, men who went on to develop dementia had an additional age-adjusted increase in systolic BP of 0.26 (95% CI 0.01-0.51) mmHg per year compared to survivors without dementia. Over the late-life examinations this group had an additional age-adjusted decline in systolic BP of 1.36 (0.64-2.07) mmHg per year. These associations were strongest for vascular dementia, and were reduced substantially in men who were previously taking antihypertensive medication. Similar changes in diastolic BP were observed but only for vascular dementia and the findings were not modified by antihypertensive treatment. Over a 32-year period, compared to men who did not, those who did develop dementia have had a greater increase, followed by a greater decrease in systolic BP. Both these trends are modified by antihypertensive therapy.

Introduction

The association between blood pressure and dementia is complex. On the one hand, raised mid-life blood pressure is associated with an increased risk of cognitive impairment, dementia and Alzheimer’s disease (AD) 10-20 years later.^{1; 2} On the other hand, by the time dementia has developed, blood pressure levels are relatively low compared to non-demented subjects,³ although this has not been a universal finding.⁴ This inverse association may reflect a decline in blood pressure that occurs in the prodromal stage of dementia,^5-7 possibly as a consequence of the weight loss or other metabolic changes that occur as a part of the dementia syndrome.⁸ Understanding the natural history of blood pressure over time and by dementia status can help to interpret observational research, design prevention strategies, and to make decisions about treating high blood pressure in old age.

The 32 year follow-up for the Japanese-American men who participated in the Honolulu Heart Program and later the Honolulu-Asia Aging Study, provided a unique opportunity to investigate long-term blood pressure trajectories in those who did or did not develop dementia. Based on this cohort, several reports have been published of the association between increased mid-life blood pressure levels and increased risk for adverse late-life cognitive and brain structure outcomes.^1;2 We have found these associations are moderated by use of antihypertensive medications and apolipoprotein E status (the e4 allele increasing the risk for AD).⁹ Therefore here we investigated the trajectories of blood pressure by dementia status and the extent to which they were modified by antihypertensive treatment and genetic susceptibility.

Methods

Participants

The study design of the Honolulu Heart Program and its follow-up, the Honolulu-Asia Aging Study (HAAS), has been previously described.⁸ Briefly, the cohort included Japanese-American men identified from selective service records, who were born between 1900 and 1919, and were living on the island of Oahu, Hawaii in 1965. Participants were examined on three occasions between 1965 and 1971. Of survivors, 80% participated in a fourth exam between 1991 and 1993, and were followed in two subsequent exams in 1994-96 and 1997-99 (with participation rates in survivors of 84% and 90% respectively). Dementia was ascertained at these three late-life exams. All participants gave written, informed consent at each exam. Proxies gave permission for cases of dementia. The protocol was approved by the Kuakini Medical Center Institutional Review Board.

Blood pressure measurement

Blood pressure was measured with the same standardized protocol at each exam. Briefly, after the participant had been seated for at least ten minutes, systolic and diastolic blood pressure were measured on three occasions five minutes apart on the left arm of a seated participant using a mercury sphygmomanometer with a standard cuff. Diastolic pressure was recorded as the fifth phase. Repeated readings were averaged for each examination.

Dementia ascertainment

The three-stage procedure for dementia case-finding, described previously, included a cognitive pre-screening, neuropsychological testing, a proxy interview, a neurological examination and neuroimaging.^{2; 10} Consensus diagnoses were made by a neurologist and two physicians. Diagnosis was made according to the following internationally accepted criteria: dementia according to DSM-III-R, AD according to NINCDS-ADRDA and vascular dementia (VaD) according to the California Alzheimer’s Disease Diagnostic and Treatment Centers critera.^11-13 The criteria for probable vascular dementia require dementia, CT/MRI evidence of at least one infarct outside the cerebellum and then either clinical/imaging evidence of two or more ischaemic stokes or a single stroke with a clear temporal relationship to the onset of dementia. Further support is allowed if there is evidence of multiple infarcts in brain regions known to affect cognition, multiple transient ischaemic attacks, a history of vascular risk factors and an elevated Hachinski Ischemic Scale score. Characteristics of participants with and without dementia have been described previously for an identical sample: those with dementia were older, and were more likely to have had a stroke, previous hypertension, previously impaired physical function and carry the APOE ε4 allele.⁸

Covariates

The following factors, ascertained at the fourth exam unless stated otherwise, were also included in the analysis: age (at entry to the study), years of formal education, history of stroke (obtained through surveillance of hospital records), previous hypertension (previous treatment, systolic blood pressure >160 mmHg or diastolic blood pressure >95 mmHg), diabetes (WHO criteria), smoking status (current, previous or never), impaired physical function (inability to rise from a chair or walking speed 0.4 m/s or below), and depressive symptoms (CES-D¹⁴). Associations of interest were also adjusted for weight change as a time-dependent covariate. Participants had been weighed at all six exams, and at the first exam had also been asked about weight at the time of military service (early adulthood). Two variables were generated to investigate confounding in this respect. One was the weight recorded at the 1^st exam (i.e. cross-sectional between-person effect) and the other was within-person weight change (kg/year) from the first exam at each blood pressure recording (i.e. longitudinal effect). Based on previous studies in this cohort,^{2; 15} apolipoprotein E (APOE) genotype (presence or absence of the ε4 allele) and antihypertensive use (up to and including exam 4) were considered as potential modifying factors.

Statistical analysis

Here, we investigated trajectories in the 1890 men who participated at the sixth exam and were not previously diagnosed with dementia. As described previously,⁸ this approach was taken so as to maximize the number of previous examination points for estimation of trajectories. Of the 3734 HAAS participants at the fourth exam, 226 had dementia at that time, 135 had dementia by the fifth exam, 850 died prior to the sixth exam and 633 did not participate for other reasons. Those present at the fourth exam but absent at the sixth were older (mean difference in age at the fourth exam 2.8 (2.5-3.1) years, had a higher mid-life SBP (mean difference at the first exam 4.8 (3.6-6.0) mmHg), and had a higher late-life SBP (mean difference at the fourth exam 1.5 (0.0-3.0) mmHg). Individual blood pressure changes across the six exams were analyzed using random effects models (with random intercept and slope functions) to account for between-participant heterogeneity and unequal time intervals between visits. Analyses were carried out separately for systolic and diastolic blood pressure (SBP, DBP). Blood pressure was entered as the dependent variable with dementia, time and a dementia*time interaction entered as independent variables. Time intervals were calculated individually from exam dates. In order to model changes in trajectories flexibly, a three-piece linear spline was used with two knots fixed at times when the mean ages of the sample were 61 years (consistent with the mean time-gap between exams 1 and 3) and 78 years (consistent with the mean time-gap between exams 1 and 4). Individual interaction terms were summed to estimate differences in slope between stratification variables (e.g. dementia status) and standard errors calculated from estimated covariance matrices via the Delta method.¹⁶ Other independent variables were entered into the model to investigate confounding effects. The fully adjusted model was then repeated for dementia subtypes and was stratified for any previous antihypertensive medication use and presence/absence of the APOE ε4 allele. SAS (version 8) and STATA (version 6) statistical software were used. The authors had full access to the data and take responsibility for its integrity. All authors have read and agree with the manuscript as written.

Results

At the first exam in 1965, the age range was 46-68 years; at the sixth exam participants were between 77-98 years old. At the sixth exam, there were 112 cases with dementia: 74 with AD and 15 with VaD. Characteristics of the sample are summarized in Table 1. The sample as a whole had experienced an increase in SBP and DBP from mid- to late-life and a decrease in SBP and DBP in late-life (Figures 1-4). Compared to those who did not, in men who developed dementia, SBP increased an additional 0.26 mmHg/year from mid- to late-life and decreased an additional 1.36 mmHg/year in late-life (Table 2, Figure 1a). These changes in SBP over these two time periods were greater in VaD cases than in AD cases (Table 2; Figure 2). For the whole sample, adjustment for education and previous health status at exam 4 did not have a substantial impact on these differences in slope. The difference in late-life SBP change between those who did and did not develop dementia was reduced modestly after adjustment for baseline weight and weight change (Table 2). In general, there was little evidence for significant differences in blood pressure level at baseline, apart from a negative coefficient for systolic blood pressure in the fully adjusted model.

Table 1.

Characteristics of the Honolulu-Asia Aging Study cohort at previous and current examination points

Characteristic		Status at Examination 6 (1997-99)
		All participants (n=1890)	No dementia (n=1778)	Incident cases
				Dementia (n=112)	AD (n=74)	Vasc dementia (n=15)
Mean (SD) age at Exam 6 (yrs)		83 (3.8)	82 (3.7)	85 (5.0)	85 (4.9)	84 (5.3)
Mean (SD) education (yrs)		11 (3.2)	11 (3.2)	11 (3.4)	10 (3.4)	11 (4.2)
Previous stroke (%)*		2	2	5	1	40
Hypertension (%)*		46	46	49	43	80
Diabetes (%)*		32	32	33	35	27
Smoking (%)*	Never	36	36	38	36	50
	Past	58	58	59	60	0
	Current	6	6	3	4	50
Impaired physical function (%)*†		2	2	9	4	13
Median (interquartile range)
CES-D score*		3 (1-5)	3 (1-5)	3 (0-6)	3 (0-6)	3 (0-10)
APOE e4 (%)		18	18	21	23	7
Antihypertensive use (%)*		40	40	41	37	73
Mean systolic / diastolic blood pressure (mmHg)
Examination 1 (1965-68)		128 / 81	128 / 81	127 / 80	126 / 80	132 / 82
Examination 2 (1967-70)		128 / 83	128 / 83	128 / 83	126 / 80	136 / 85
Examination 3 (1971-74)		131 / 84	131 / 84	131 / 84	132 / 85	134 / 83
Examination 4 (1991-93)		149 / 76	148 / 81	151 / 81	148 / 80	166 / 86
Examination 5 (1994-96)		143 / 76	143 / 76	145 / 77	145 / 77	157 / 82
Examination 6 (1997-99)		143 / 73	143 / 73	137 / 72	138 / 73	137 / 70

^*Classified/measured at Examination 4 (1991-93)

^†Inability to rise from a chair or walking speed 0.4 m/s or below

Figure 1.

Graphs plotting previous systolic blood pressure adjusted for age according to whether men did or did not develop dementia between exams 5 and 6: (a) Total sample [n=1890]; (b) No previous antihypertensive treatment [n=1126]; (c) Previous antihypertensive treatment [n=765].

Figure 4.

Graphs plotting previous diastolic blood pressure adjusted for age according to whether men did or did not develop dementia subtypes between exams 5 and 6: (a) Alzheimer’s disease; (b) Vascular dementia.

Table 2.

Previous systolic blood pressure change associated with incident dementia at the sixth examination (1997-99) estimated from the random effects model.

Model (all age-adjusted)	Difference at baseline	p-value	Additional change in rate of change in systolic blood pressure associated with dementia (mmHg/year)			p-value*
			Mean age 54-60	Mean age 61-78	Mean age >78
All dementia (n=1890)	-2.00 (-5.11, 1.11)	0.21	0.17 (-0.29, 0.62)	0.26 (0.01, 0.51)	-1.36 (-2.07, -0.64)	<0.001
Stratified analysis
…if no previous antihypertensive treatment (n=1126)	-0.98 (-4.15, 2.19)	0.55	0.55 (0.02, 1.07)	0.28 (-0.01, 0.57)	-1.92 (-2.70, -1.14)	<0.001
…if any previous antihypertensive treatment (n=765)	-3.93 (-8.89, 1.04)	0.12	-0.39 (-1.17, 0.40)	0.24 (-0.18, 0.66)	-0.54 (-1.85, 0.77)	0.615
Dementia sub-types
Alzheimer’s disease (n=1852)	-3.42 (-6.91, 0.06)	0.05	0.45 (-0.09, 0.98)	0.03 (-0.25, 0.31)	-0.76 (-1.55, 0.03)	0.090
Vascular dementia (n=1793)	4.82 (-3.22, 12.86)	0.24	-0.29 (-1.54, 0.96)	1.18 (0.77, 1.59)	-3.75 (-5.61, -1.90)	<0.001
Adjusted analyses for all dementia
1. Adjusted for education	-2.01 (-5.12, 1.10)	0.20	0.17 (-0.29, 0.62)	0.26 (0.01, 0.51)	-1.36 (-2.07, -0.64)	<0.001
2. Model 1 plus vascular disease†‡	-2.48 (-5.49, 0.52)	0.11	0.16 (-0.30, 0.62)	0.26 (0.01, 0.51)	-1.36 (-2.07, -0.65)	<0.001
3. Model 2 plus disability†§ and depression†	-2.30 (-5.28, 0.68)	0.13	0.16 (-0.29, 0.62)	0.26 (0.01, 0.51)	-1.36 (-2.07, -0.65)	<0.001
4. Model 3 plus baseline weight and weight change	-3.03 (-5.90, -0.15)	0.04	0.22 (-0.24, 0.67)	0.29 (0.04, 0.54)	-1.04 (-1.76, -0.32)	0.002

^*P-value based on likelihood ratio testing of the overall significance of the interaction terms between dementia status and time slopes

^†All ascertained at the 4^th exam, 1991-93

^‡Ischemic heart disease, diabetes, stroke, smoking, hypertension

^§Inability to rise from a chair or a walking speed of 0.4m/s or slower

Figure 2.

Graphs plotting previous systolic blood pressure adjusted for age according to whether men did or did not develop dementia subtypes between exams 5 and 6: (a) Alzheimer’s disease; (b) Vascular dementia.

Antihypertensive treatment modified these observed differences in SBP trajectories. Among those who had not previously reported antihypertensive treatment, those who developed dementia had a significantly greater increase in SBP in mid-life, and a significantly greater decrease in late-life (Figure 1b, 1c), compared to those who did not develop dementia. Among those who did receive anti-hypertensive treatment, SBP trajectories did not differ between those who did and did not develop dementia. Carrying an APOE ε4 did not modify the trajectories of those who did and did not develop dementia (data not shown). Differences in DBP trajectories were weaker but in the same direction as those for SBP change (Table 3, Figure 3), and were only significant for VaD (Figure 4).

Table 3.

Previous diastolic blood pressure change associated with incident dementia at the sixth exam (1997-99) estimated from the random effects model.

Model (all age-adjusted)	Difference at baseline	p-value	Additional change in rate of change in diastolic blood pressure associated with dementia (mmHg/year)			p-value*
			Age 54-60	Age 61-78	Age >78
All dementia (n=1890)	-0.03 (-1.99, 1.93)	0.98	0.16 (-0.16, 0.48)	0.01 (-0.12, 0.14)	-0.18 (-0.54, 0.18)	0.532
Stratified analysis
…if no previous antihypertensive treatment (n=1126)	0.99 (-1.28, 3.25)	0.39	0.18 (-0.18, 0.54)	-0.03 (-0.19, 0.12)	-0.21 (-0.66, 0.24)	0.494
…if any previous antihypertensive treatment (n=765)	-1.92 (-4.97, 1.12)	0.21	0.13 (-0.44, 0.71)	0.08 (-0.13, 0.28)	-0.13 (-0.72, 0.47)	0.753
Dementia sub-types
Alzheimer’s disease (n=1852)	-0.77 (-3.03, 1.50)	0.51	0.44 (0.07, 0.81)	-0.09 (-0.24, 0.06)	-0.06 (-0.48, 0.36)	0.158
Vascular dementia (n=1793)	1.73 (-4.02, 7.48)	0.55	-0.24 (-1.19, 0.71)	0.47 (0.16, 0.78)	-1.28 (-1.93, -0.64)	0.013

^*P-value based on likelihood ratio testing of the overall significance of the interaction terms between dementia status and time slopes

Figure 3.

Graphs plotting previous diastolic blood pressure adjusted for age according to whether men did or did not develop dementia between exams 5 and 6: (a) Total sample [n=1890]; (b) No previous antihypertensive treatment [n=1126]; (c) Previous antihypertensive treatment [n=765].

Of the 109 participants who were diagnosed with dementia at the sixth exam, 58% had experienced a fall in SBP of at least 10 mmHg over the previous six years (since exam 4) compared to 39% of those without dementia (odds ratio 2.10, 95% CI 1.42-3.12). Thirty nine percent of those with dementia had experienced a SBP fall of greater than 20 mmHg compared to 24% of those without dementia (odds ratio 2.09, 95% CI 1.40-3.11).

Discussion

Here, in an analysis of data collected over three decades, we found that men who developed dementia had a greater rise in SBP from mid- to late-life, and a greater decrease in blood pressure over an approximately six-year period before the dementia diagnosis. For both SBP and DBP, these associations were strongest for VaD. Further, anti-hypertensive medication modulated the trajectories, such that differences in change in SBP over time were greatly diminished and levels remained relatively stable.

Advantages of this study include the large sample size, the six blood pressure measures acquired with a standardized protocol, the long duration of follow-up and structured screening and diagnostic procedures for dementia on three occasions. Further, any measurement error in BP is unlikely to have been different between outcome groups, and will therefore have obscured rather than exaggerated any group differences in trajectories – for example, if systolic blood pressure is easier to estimate than diastolic then this might account for more marked differences in one compared to the other . It should be noted that these results pertain to men who survived to exam 6 of the HHP/HAAS. An advantage of this approach is that it reduces the effect of survival bias on the comparisons between demented and non-demented subjects and permits the calculation of trajectories based on complete data. However, the sample clearly limits the generalisability of findings to women and other population sub-groups. Also the analysis is based on survivors to age 77-98 years, i.e. a relatively elderly cohort and findings cannot necessarily be generalized to dementia of earlier onset.

These longitudinal trajectories provide important insight into research on the association of blood pressure with risk for dementia. A large number of studies have investigated blood pressure level in relation to dementia with the most consistent findings for higher blood pressure as a risk factor coming from those with longest follow-up periods. These have been comprehensively reviewed.¹⁷ As far as we are aware, ours is the first report on change in blood pressure from mid- to late-life in relation to dementia, which can help with interpretation of extant studies. The greater rise in SBP from mid- to late-life in men who later developed dementia is consistent with previous findings of associations between raised mid-life blood pressure levels and dementia, which have also been found to be strongest for untreated hypertension.^{1; 2} Similarly, modification by previous antihypertensive treatment was most evident for the dementia-associated difference in mid-life SBP trajectories rather than in mid- to late-life slopes. Although SBP of those who developed dementia and were taking antihypertensives was lower at all time points compared to those who did not develop dementia, these differences were marginal.

The nature of the relation of blood pressure to dementia remains controversial. Blood pressure decline may be a consequence of neurodegeneration. Several studies have found that people with dementia have relatively low resting,^{3; 5} and orthostatic¹⁸ blood pressure. Other studies found that blood pressure is lowest in people with most advanced dementia.^{3; 19} Dementia has also been found to be associated with other physical changes such as weight loss and a fall in total cholesterol.^{8; 20-22} In this cohort, exaggerated weight loss was detectable in men with incident dementia prior to the clinical onset,⁸ and was detectable over a similar period to SBP decline. The exaggerated SBP decline in men with incident dementia was partly but not wholly explained by previous weight change (the late-life dementia*time coefficient for SBP in Table 2 being reduced by about 25% following adjustment). SBP decline may therefore be one feature of a more general metabolic change associated with neurodegeneration – possibly due to effects on brainstem regulating centres, and particularly if cerebrovascular disease is an important component (discussed below).

However, if blood pressure decline was purely secondary to dementia then associations between the two would not be expected to be modified by previous antihypertensive treatment. Instead, we found a much weaker association between dementia and late-life SBP change in men on antihypertensive treatment. Hypertension in mid-life, particularly when untreated, may give rise to increased arterial stiffness (measured using proxy markers such as increased pulse pressure and pulse wave pressure) which is itself associated with cognitive decline.²³ Disturbances in cerebral autoregulation in people with hypertension has been found to be associated with white matter damage,²⁴ and neuropathological studies have found that watershed infarctions and arterial stenoses are associated with Alzheimer pathology.^{25; 26} Furthermore, worse cerebral autoregulation has been found to be associated with over-expression of the amyloid precursor protein in mice.²⁷ When untreated, these changes may all confer vulnerability of the brain to hypotensive episodes. Our findings therefore raise the possibility that treatment of hypertension not only reduces dementia through prevention of blood pressure elevation 10-20 years earlier but might also reduce cerebral damage due to hypotension later in life. This conclusion is further supported by the stronger associations in men with VaD: while there is strong overlap between AD and VaD,²⁸ this group is will have had more severe cerebrovascular pathology suggesting a vascular origin for both the exaggerated increase and decrease in BP (both systolic and diastolic), and its possible modification (for systolic BP) by previous antihypertensive treatment. However, the VaD group was small in our cohort and two other studies have found stronger associations between low blood pressure and dementia in people taking antihypertensive agents;^{29; 30} therefore further research is required.

Perspectives

While there is growing evidence that hypertension is a modifiable risk factor for VaD and AD, it is important for clinicians to be aware that a person’s blood pressure may have begun to decline by the time they develop dementia. Early antihypertensive use may reduce the extent of this decline but further research is required into the risks or benefits of blood pressure lowering when neurodegenerative processes are at a more advanced stage.