Lowering mid-life levels of systolic blood pressure as a public health strategy to reduce late-life dementia

Abstract

To estimate the potential benefits towards preventing late-life dementia, of lowering systolic blood pressure [SBP] we estimated the population attributable risk (PAR) of elevated SBP on dementia. Analyses are based on the cohort of 8006 Japanese American men (b. 1900 – 1919) followed since 1965 as a part of the Honolulu Heart Program, continued as the Honolulu Asia Aging Study. Mid-life cardiovascular risk factors and late-life brain function are well described. We estimated the PAR of dementia cases attributed to mid-life SBP, grouped by JCN-7 criteria [<120, 120 – < 140, and ≥ 140 mmHG], taking into account treatment history, confounding factors, and competitive risk for death. The analysis is based on 7878 subjects, including 491 cases of dementia, with a mean interval of 25 years between measurement of BP and dementia diagnosis. Compared to those with SBP <120 mmHG, untreated and <50 years at baseline, 17.7% (95% CI 4.6% – 29.1%) of the cases are attributable to prehypertensive levels (SBP 120 – <140 mmHG) of SBP, translating into 11 excess cases per 1000. Among those who did not report taking anti-hypertensive medication in mid-life, 27% [95%CI 8.9%, 42.1%] of dementia cases can be attributed to systolic BP >120 mmHG, translating into 17 excess cases per 1000. Although PAR estimates for population sub –groups may differ by relative risk for dementia or prevalence of elevated levels of BP, these data suggest reducing mid-life systolic BP is an effective prevention strategy to reduce risk for late-life dementia.

Dementia causes devastation to the patient and family members, and costs millions of dollars in health care. As the numbers of persons with dementia increases with the aging of the population, preventing this trend is warranted in both human and economic terms. Currently, there are no proven strategies to reduce the occurrence of dementia.

High blood pressure [BP] has long been understood to cause stroke [1]. Still evolving, is our understanding of the role increased levels of BP plays in shaping the trajectory to dementia. In the past 10 years, several prospective cohort studies have provided compelling data suggesting increased levels of BP are associated with an increased risk for dementia [2 – 4]. Importantly, the several studies showing this association are based on measures of BP relatively distal to the time dementia is diagnosed [5]. These studies suggest high BP not only increases the risk for vascular dementia, but also Alzheimer’s disease (AD). Supporting data link high levels of BP to an increased risk for markers of the dementing processes, including cognitive impairment [6–8], global [9] and hippocampal atrophy [10,11], white matter lesions [12,13], neuritic plaques and cerebral vascular lesions [14]. Based on these studies the question arises: Will a shift in the population distribution of mid-life BP to lower levels be associated with lower rates of dementia in the population?

The population attributable risk (PAR) describes the proportion of cases that could be prevented if a specific exposure were eliminated in a target population (15,16). In the context of this analysis, the PAR measures the potential impact on dementia rates of reducing elevated levels of blood pressure in the population. As dementing processes can begin years before a clinical stage is reached, long periods of follow-up are needed to best identify and estimate the potential of changing risk factor levels to prevent the disease. Observational longitudinal cohort studies can provide such data, contributing importantly to estimates of the public health impact of interventions and to plans for clinical trials.

Here we present analyses based on the cohort of Japanese American men followed continuously since 1965 as a part of the Honolulu Heart Program (HHP) [17], and since 1991 as a part of the Honolulu Asia Aging Study (HAAS) (18). The long follow-up and serial, standardized assessment of blood pressure, dementia and other cardiovascular risk factors provides a unique opportunity to estimate the impact on dementia rates of reducing blood pressure levels. Specifically, we estimate the proportion of late-life dementia cases that are attributable to elevated systolic blood pressure and could be avoided if normal levels of systolic BP are maintained in mid-life. We account for several factors that contribute to PAR calculations for dementia attributed to elevated levels of SBP, including control for confounding due to other cardiovascular risk factors and disease; and control for the competitive risk for death associated with high levels of SBP (19). Not accounting for these factors may lead to an overestimate of the PAR for BP levels (20). Further, estimates are stratified by history of hypertension treatment, to follow published literature suggesting treatment history is an important modulator of risk for dementia [5,21], Here we focus on SBP; In 2003, the Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure concluded SBP it is the most potent risk factor for cardiovascular disease over 50 years of age [22].

Methods

The HHP was started in 1965 as a response to the epidemic of cardiovascular disease in men. The study has been described in many publications over the past 40 years. Briefly, 8006 Japanese American men born between 1900 and 1919, living on Oahu in 1964, and registered on the selective service rolls, participated at baseline. The core HHP was based on three exams, carried out in 1965/68; 1967/70 and 1971/74 (Figure 1), with a mean interval of 6 years between the first and third exam. In this present analysis, these visits constitute the mid-life exams. At baseline participants were aged 45–68 (mean age, 54 yrs). From the beginning of the HHP, information on incidence of coronary artery disease, stroke and vital events has been obtained by monitoring local English and Japanese language newspapers, and by surveillance of hospital discharge records. A follow-up survey in the 1991–1993 examination found that only 5 men could not be traced for mortality information (23).

Figure 1.

Study design of the HHP/HAAS studies and analytical sample.

The HAAS began in 1991, on average, 25.1 years after the first HHP exam. The aim of HAAS is to assess diseases of old age, with an emphasis on brain aging and dementia [18.] When the HAAS started, there were 4768 survivors of the original 8006 and 3734 (80% of survivors) participated in the fourth examination (which constitutes the HAAS baseline). At exam 4, participants ranged in age from 71–93 years (mean 77.8 yrs). Dementia outcome data for these analyses were collected at exam 4 and three follow-up examinations carried out between 1994 and 2000. The follow-up exams had participation rates among survivors of 85.2%, 83.8%, and 82%, respectively. The Kuakini Medical Center Institutional Review Committee approved the study and all subjects, or their caretakers when subjects were demented, provided written informed consent.

Dementia-case finding

Case-finding was conducted according to a multi-step procedure previously described (18). At HAAS baseline (examination 4) and subsequent examinations, subjects were first screened for dementia using the 100-point Cognitive Abilities Screening Instrument (CASI), which is a combination of the Hasegawa Dementia Screening Scale, the Folstein Mini-Mental State Examination, and the Modified Mini-Mental State Test (24). Those scoring below a pre-defined cut-point on the CASI were further evaluated with a neurologic exam, neuropsychological testing, and an informant interview about changes in cognitive function and behavior. In subjects with dementia, a brain image was acquired and routine blood tests +conducted. Based on these data, a consensus diagnosis of dementia, Alzheimer’s disease [AD] and vascular dementia [VaD] following internationally accepted criteria [25–27], was made by the study neurologist and two physicians with expertise in dementia. This consensus conference also diagnosed other dementias, including those due to alcohol abuse, brain tumor, subdural hematoma, Parkinson’s disease, Lewy body disease, Pick’s disease, trauma, vitamin B₁₂ deficiency, hypothyroidism, progressive supranuclear palsy, and unknown cause. In this study we focus on total dementia, defined to include cases of AD, VaD (the most frequent subtypes of dementia) and mixed AD/VaD cases, and exclude other dementias. AD, VaD and mixed dementia have been shown to be associated with high BP in several observational studies (5), and neuropathologic studies show decedents have multiple lesions, with the combination of Alzheimer and vascular lesions being most prevalent (28). The other dementia subtypes are excluded from analyses since, as a group or individually, there is no evidence that risk for these diseases is modulated by high blood pressure.

Blood Pressure measurements

At each exam, blood pressure was measured three times on the left arm with subjects in a seated position. These measures were averaged to get a per visit mean blood pressure. To obtain a robust measure of exposure, we further averaged the means from exam 1 and exam 3. The mid-life SBP levels [mmHg] were categorized according to the JNC7 guidelines for systolic blood pressure [22]: normal (SBP <120); prehypertension (SBP 120 – <140); and combined Stage1 and Stage 2 hypertension (SBP ≥140). Subjects were further classified a priori as (n)ever treated with anti-hypertensive medication if they reported treatment when asked at the first three exams. Treatment has long been known to modulate risk for stroke and mortality. There is also a more recent body of literature suggesting treatment moderates the association of blood pressure to dementia [5] Studies based on the HAAS have consistently shown the risk for pathologic changes in brain function and structure are strongest in those never treated for hypertension [2,8,10,21].

Confounders and covariates

To estimate the PAR we controlled for variables shown in previous studies to be significant correlates of dementia (2,3,4,7,8,9): age; years of education; mid-life smoking history; mid-life body mass index [BMI; calculated as measured weight (kg) divided by height (m) squared]; and presence of diabetes, defined as a history of diabetes diagnosed by a physician, taking diabetic medications or insulin, or glucose intolerance based on a non-fasting 1-hr glucose level [29]. Prevalent CHD [angina pectoris, myocardial infarction, coronary insufficiency]. and CVA were ascertained by questionnaire at baseline and updated thereafter through a previously described continuous surveillance of hospital discharge and death records [30]. A positive history referred to either a stroke or coronary heart disease before examination 4.

Analytical sample

From the original 8006 cohort members, we excluded 42 with no information after exam 1. Cases (n=86) diagnosed with sub-types of dementia other than AD and VaD were also excluded, as described above. Hence, we had an analytical data set of 7878 subjects, including 439 dementia cases, monitored through 10-03-2000, the date of the last exam included in these analyses (Figure 1).

The 7878 included 346 with no exams after exam 3, and therefore no cognitive data and no notification of death. These 346 men were on average younger, had less diabetes and coronary disease, but had similar blood pressure levels compared to those who had at least one cognitive exam.

There were an additional 728 men who had at least one CASI test that was screen negative, who did not participate in any subsequent exam and there was no record of death before the end of follow-up. For the main analyses, we imputed case status in these 728 men using age and CASI scores, the main factors associated with risk for dementia. The C index, a measure of how well the independent variables together classify subjects according to outcome, was good (0.81) and did not substantially change with the addition of more variables [31]] Imputation was done by calculating the observed proportion of dementia cases in cells defined by CASI score [<60, 60–79, 79 and above] [24], and 5-year age groups, separately for exam 4, 5, and 6. Using those percentages we then randomly selected subjects within the CASI/age cells in the 728. This resulted in 52 cases who, compared to examined cases, were on average younger (56 (SD 5.4) vs. 54 (SD 4.2 years old) but otherwise similar in mid-life blood pressure levels and other risk factors associated with high blood pressure. There was no change in the proportion of cases across the blood pressure groups, but among the imputed cases there was a higher proportion of those who were treated with anti-hypertensive medications.

Time to event

Censoring events included dementia diagnosis, death if it occurred before a dementia diagnosis, or no dementia or death by the end of follow-up date. Time to event was defined as follows: 1) For the cases identified at exam 4, we assumed a time of onset two years prior to the exam 4 date, which is approximately mid-way between time of dementia diagnosis and institutionalization or death [32]; 2) for the incident measured and imputed cases, we assumed a time of onset in the mid-point between the exam at which the diagnosis was made and the previous exam; 3) for those who died, the censor date was date of death; 4) for those with unknown status after exam 3 (n=346), the date of exam 3 was used for censoring; 5) all others were censored at October 3, 2000.

Analytical Strategy

Because high blood pressure is a strong risk factor for death [19], we calculated the relative risk with a proportional hazard competing risk model that accounts for the possibility that someone dies before reaching the clinical dementia end-point, or the end of the follow-up. Not taking competing risk into account may lead to an overestimate of the risk for disease, specifically when the second outcome, in this case death, is associated with the main exposure of interest (20). The model is based on the method of Fine and Gray [33] and was implemented in the program function crr in the R package cmprsk [34].

Stratification of the cohort

We a priori stratified the sample by treated vs. not treated in mid-life, as described above. Because the R program does not allow delayed entry, the cohort was further stratified by baseline age [≤50 and ≥50 yrs old], giving 9 BP/age/treatment cells. The statistical package also requires other confounders be entered into the model as dichotomous variables. This was done as follows: education [<12, ≥12 yrs); smoking [ever/never]; BMI [<25 vs. ≥25] (35); and presence/absence of diabetes, CHD and stroke.

Estimation of parameters

For the estimation of relative risk we chose as our reference group the cell at the lowest risk for death or dementia associated with SBP level – those aged <50 yrs at baseline who had normal blood pressure and did not report taking anti-hypertensive medication (see supplementary table 1 on http://hyper.ahajournals.org). The relative risk of developing dementia was calculated within each of the 8 other BP level/age/treatment cells [see appendix 1, http://hyper.ahajournals.org for details of the statistical methods] so all comparisons are made to the same reference group.

Within each age/treatment strata, the attributable fraction (AF) was obtained for each level of BP by multiplying the cell specific RR by the total number dementia cases. Individual cell AF were added up for estimates of PAR by BP level. The excess/prevented number of dementia cases was calculated by multiplying the number of dementia cases within each age/treatment strata by the attributable fraction for those in the normotensive, prehypertensive and hypertensive systolic BP groups, separately.

Confidence levels were estimated for each RR and PAR estimates by bootstrapping with 1,000 bootstraps on 80% subsets. SAS v9.1.3 [36] was used to calculate the PAR and bootstrapped 95% confidence intervals (95%CI).

Results

The total cohort mean age at baseline was 54 yrs [SD 5.6 yrs]; the mean age of the ≤50 group was 48 yrs and of the >50 group was 56.5 yrs. There were 41.3% of men with systolic mid-life prehypertension, and 37.4% with systolic hypertension. The prevalence of all cardiovascular risk factors and disease (BMI, diabetes, coronary heart disease, and stroke) increased with increasing SBP group (Table 1). There were however proportionately more dementia cases in the systolic prehypertension group.

Table 1.

Description of the cohort by mid-life systolic blood pressure group: HHP/HAAS

Characteristic	Systolic BP (mmHG)			Total Sample
	Normotensive (< 120)	Prehypertension (120 – <140)	Hypertension (≥ 140)
	(n=1709)	(n=3221)	(n=2948)	(n=7878)
Percent of the sample	21.7	40.9	37.4	100
Age (yrs) (mean ± SD)	53.2 (±5.2)	53.9 (±5.4)	55.7 (±5.8)	54.5 (±5.6)
Mid-life Systolic BP (mmHG) (mean ± SD)	112.1 (±6.0)	129.9 (±5.7)	156.1 (±14.0)	135.8 (±19.6)
Mid-life Diastolic BP (mmHG) (mean ± SD)	72.4 (±6.2)	81.6 (±6.5)	91.4 (±9.4)	83.3 (±10.5)
Trt in mid-life with anti-hypertensives [n, %]	63 (3.7]	794 (24.7]	2376 (80.6]	1709 (21.7]
Non smokers [n, %]	464 (27.2]	871 (27.0]	776 (26.3]	2111 (26.8]
Education [n, % with 12 yrs]	859 (50.3]	1598 (49.6]	1298 (44.0]	3755 (47.7]
BMI ≥25 [n, %]	358 (21.0]	1111 (34.5]	1237 (42.0]	2706 (34.4]
Diabetes [n, %]	454 (26.6]	947 (29.4]	1118 (37.9]	2519 (32.0]
History of coronary heart disease [n, %]	91 (5.3]	231 (7.2]	311 (10.6]	633 (8.0]
History of stroke [n, %]	18 (1.1]	46 (1.4]	117 (4.0]	181 (2.3]
Event Type
Death [n, %]	923 (54.0]	1862 (57.8]	2184 (74.1]	4969 (63.1]
Dementia [n, %]	102 (6.0]	235 (7.3]	154 (5.2]	491 (6.2]
Non-demented [n, %]	684 (40.0]	1124 (34.9]	610 (20.7]	2418 (30.7]
Number of follow-up years
Dead (mean ± SD)	22.0 (±8.8)	21.8 (±8.3)	19.9 (±8.4)	21.0 (±8.5)
Demented (mean ± SD)	27.1 (±3.7)	27.2 (±3.2)	26.4 (±3.1)	27.0 (±3.3)
Not demented (mean ± SD)	28.3 (±9.5)	27.3 (±10.2)	25.8 (±11.1)	27.2 (±10.3)
Person years of follow-up
Dead	20320.5	40491.2	43379.2	104190.9
Demented	2767.7	6393.3	4068.7	13229.7
Not demented	19357.7	30660.3	15717.6	65735.6

Overall, 41.3% of the subjects reported taking anti-hypertensive medication including 24.6% in the prehypertension, and 80% in the hypertension group. Compared to those who were treated, those who were not treated had lower SBP and DBP blood pressures and BMI, were less likely to have cardiovascular disease and a lower proportion of the group died [see supplementary table 1 on http://hyper.ahajournals.org]. Compared to the rest of the cohort, the reference group [those who were untreated normotensive and <50 yrs old at baseline] had lower systolic and diastolic blood pressure, and less cardiovascular disease. The highest proportion of deaths occurred in the hypertensive group (72.8%), followed by those in the prehypertension group (57.0%); mortality was lowest in those with normal levels of SBP (53.6%). Compared to the men in the reference group, those with hypertensive levels of systolic blood pressure had a significantly higher risk for death; Mortality in the prehypertension group did not differ significantly from the normotensive group [Table 2; see supplementary table 2 on http://hyper.ahajournals.org for an analysis by blood pressure/treatment/age group].

Table 2.

Risk^* of death by mid-life systolic blood pressure group: HHP/HAAS

Systolic BP (mmHG)
Normotensive (<120) HR (95% CI)	Prehypertension (120 – <140) HR (95% CI)	Hypertension (≥ 140) HR (95% CI)
reference†	reference†	reference†
0.84 (0.70– 1.00)	0.89 (0.76– 1.05)	1.23 (1.03–1.47)

^*Hazard Ratio (95% CI)

^† Reference group (n = 493) includes untreated men with normal blood pressure who are < 50 years of age at baseline in 1965

Accounting for the competitive risk for death, and compared to the reference group, the <50 year olds had a moderately elevated risk for dementia, which was significant in the untreated group, [Table 3; see supplementary table 3 on http://hyper.ahajournals.org for a sensitivity analysis that excludes the 52 imputed dementia cases].

Table 3.

Proportional hazard competing risk^* of dementia by mid-life systolic blood pressure, treatment status, and age strata: HHP/HAAS

Systolic Blood Pressure (mmHG)	Treated with anti-hypertensive meds				Untreated
	<50 yrs old (n=631)		≥50 yrs old (n=2602)		<50 yrs old (n=688)		≥50 yrs old (n=3464)
	RR* (95% CI)†		RR* (95% CI)†		RR* (95% CI)†		RR* (95% CI)†
	reference‡		reference‡		reference‡		reference‡
Normotensive (<120)	1.04	(0.96–1.13)	0.92	(0.88–0.98)			1.02	(0.95–1.08)
Prehypertension (120 – <140)	1.12	(0.90–1.40)	0.84	(0.75–0.95)	1.54	(1.20–1.96)	1.03	(0.90–1.18)
Hypertension (≥ 140)	1.19	(0.85–1.66)	0.76	(0.65–0.93)	2.29	(1.41–3.61)	1.05	(0.86–1.27)

^*Adjusted for competing risk of death

^†1000 bootstrapped 95% CI for all competing risk RR

^‡Reference group (n = 493) includes untreated men with normal blood pressure who are < 50 years of age at baseline in 1965

Based on the combination of risk and number of cases, 27% [95%CI 8.9%, 42.1%] of dementia cases can be attributed to untreated mid-life levels of systolic BP >120 mmHG; this translates into 17 excess cases per 1000. There are 17.7% (95% CI 4.6% – 29.1%) of cases that are attributable to prehypertensive levels (SBP 120 – <140 mmHG) of BP, regardless of treatment status, which translates into 11 excess cases per 1000 (Table 4).

Table 4.

Population Attributable Risk^* for dementia attributed to mid-life systolic blood pressure: HHP/HAAS

Systolic Blood Pressure (mmHG)	Treated with anti-hypertensive meds		Untreated		Total
	PAR	(95% CI)†	PAR	(95% CI)†	PAR	(95% CI)†
Prehypertension (120 – <140)	−1.41%	(−6.46%,3.17%)	19.14 %	(6.54%, 28.81%)	17.73	(4.65%, 29.15%)
Hypertension (≥ 140)	−11.00%	(−27.70%, 5.46%)	7.94%	(2.40%, 13.23%)	−3.05	(−20.52%, 14.53%)
Combined 120 and higher	−12.40%	(−34.68%, 8.57%)	27.08 %	(8.95%, 42.14%)	14.68	(−13.61%, 41.14%)

^*Risk estimates are adjusted for confounding factors and competitive risk for death.

^†1000 bootstrapped 95% CI for all PAR

Discussion

Our PAR analyses suggest 17% of late-life dementia cases are attributable to mid-life SBP levels between 120 and 140 mmHG. Among those who did not report taking anti-hypertensive medication in mid-life, 27% of dementia cases can be attributed to systolic blood pressure levels of 120 mmHG and higher. These estimates are based on more than 25 years of follow-up, take into account the presence of other correlated factors suspected to increase the risk for dementia, and accounts for the competing risk of mortality associated with high blood pressure.

These are the first data to provide adjusted estimates of the potential impact of reducing blood pressure on rates of dementia based on a long period of observation of a population-based cohort. There are several factors related to the internal and external validity of the finding, however, that need to be considered when evaluating these PAR estimates and generalizing them to other samples and studies.

Internal validity

This study is based on a well-described large population-based cohort that has been followed for over 25 years. Standardized assessment of blood pressure, continual surveillance of mortality, and well-characterized dementia cases make the HHP/HAAS a unique cohort to base estimates of the population attributable risk of dementia due to high blood. However several assumptions had to be made in the analysis, which need to be noted. First, the 346 for whom we had no data after exam 3 were censored at exam 3, essentially assigning them the status of non-demented. These men were healthier than those who had at least one cognitive exam; the bias introduced depends on the extent to which these men were at even higher risk for dementia or death than those in the sample; given their favorable health status, the error introduced is likely to be small. Similarly we do not know the error introduced by imputing dementia status in the men with incomplete follow-up after exam 4. However, we did have a cognitive score from a prior exam, and we based the imputation on age and CASI score, which are the most important predictors of whether or not someone will develop dementia. Including this group results in a decrease in the PAR so our estimates are conservative.

Second, treatment is defined as ever having used an anti-hypertensive medication. The study did not collect data on the indication for taking the medication, the dosage of prescribed medication or whether the medications were taken continuously. However, we did find expected differences in the cardiovascular risk factor profile between the treated and untreated groups.

Several factors can account for the difference in risk and benefit between the treated and untreated groups; it may reflect differences in SBP level and variability; the effect of a particular medication; or some third factor, yet unknown that was correlated with answering positively to question about treatment. Clearly additional research is needed to answer this question.

External validity

Data are based on men born 1900–1919, which defines the extent to which these analyses are generalizable to others. First, the cohort was middle age in the 1960’s, and treatment patterns reflect the state-of-the-art at that time. At that time, the main goal was to treat diastolic blood pressure [37]. Isolated systolic hypertension was not a strong indication for treatment, and levels of blood pressure indicating treatment were higher than the current standards.

Second, these estimates are based on Japanese-American men. In this cohort, the incidence of dementia is similar to that reported for cohorts of mainly Caucasian origin [38,39], the prevalence of, and mortality associated with prehypertension is similar to that reported in NHANES [40], and the relationship of mid-life BP and late-life dementia has been replicated in cohorts of men and women from different race/ethnic groups [5]. Nevertheless, the population distributions of SBP do vary by sex, race/ethnic group, socio-economic status, diet, the force of mortality associated with elevated BP, as well as other factors [41]. Therefore, as for any PAR, separate calculations would have to made for significant sub-groups in the population.

Third, these are data that show the predicted mid-life attributable risk for dementia, without taking into account subsequent changes in BP levels or treatment. However, the finding that high blood pressure increases the risk for dementia, has been most consistent in studies where the study interval was long and BP was measured at mid-life [5]. There is increasing awareness about the difficulties in interpreting relative risks for dementia associated with blood pressure measured close to the time dementia is diagnosed. The issues that plague interpretation include the lowering of measured blood pressure that is the consequence of dementing processes, and the increasing prevalence of pathologic hypotension in older populations. Further, the strength of this analysis is the demonstration that regardless of what happens in the intervening period, from a public health prevention perspective, treatment of blood pressure in mid-life is indicated in programs aimed to reduce the burden of dementia in late-life dementia. Whether or not reducing SBP in late-age reduces the risk for dementia, or conversion from MCI to dementia, can be addressed in clinical trials such as Hyvet [42].

Finally, a randomized trial to test the efficacy of BP control to prevent dementia is needed. There are few trials examining the efficacy of treating BP with particular drugs and the risk for dementia [43]. However, follow-up was short, they achieved very small differences in BP between placebo and control, and they were not powered for dementia outcomes. The design of the study should take into account findings from observational studies that suggest longer treatment and follow-up may be important factors for prevention success.

In this cohort, the PAR for hypertension is lower than for prehypertension. This group is the smallest proportion of the sample, and is at significantly increased risk for death compared to others. It may not be surprising that hypertension is associated with risk for dementia. What this study shows is that those with prehypertension in mid-life have a higher attributable risk for dementia than either the normotensive or hypertensive men. This reflects the combination of a moderately raised risk for dementia, a similar risk for death compared to normotensives, and a higher number of prehypertensive men and within that group of cases.

Efforts to control hypertension have been well on their way for decades. Risk factors and outcomes in persons with hypertension are the most likely to change with changes in detection and treatment practices. It is only recently that the JNC-7 has modified its classification to include the term prehypertension. The new designation is intended to identify those individuals that would benefit from early intervention to reduce BP, slow down the rate of progression of BP to hypertensive levels, or prevent hypertension entirely. Our findings suggest there may also be a benefit for prevention of dementia.

Implications

Although secular trends documented through the NHANES studies show the prevalence of hypertension has decreased and the use of anti-hypertensive medication has increased over time, recent studies suggest some worrying trends. For instance, a large proportion of those with hypertension are either not aware of their condition, are aware but have not sought treatment, or have been treated, but unsuccessfully [44]. Further the rates of obesity and diabetes, both important co-morbidities with hypertension, are increasing, particularly in children. With these trends, the number of people with prehypertension in mid-life will inevitably increase. The changes in medical practice, coupled with the changes in life style that occur with time argue for a broad surveillance system that can monitor the end impact of population changes in BP levels and changes in the incidence of dementia. Although the PAR estimates may differ among subgroups in the population, these data are clearly suggestive of the need to consider mid-life SBP level as a modifiable risk factor for dementia.

Perspectives.

Dementia has high individual, societal and economic costs. It is an age-related disease and as the proportion of older persons in the population increases, proportionately more cases will develop. Prevention is a priority, but to date there is no accepted prevention strategy. In this regard, two concepts are becoming crystallized in experimental, clinical and epidemiologic studies: 1) Clinical dementia is the end of processes that begin many years before a diagnosis is made, and, 2) cardiovascular risk factors are associated with an increased risk for late-life dementia. Together, these concepts suggest early intervention on cardiovascular risk factors may be one approach to prevent late-life dementia. Use of population based longitudinal studies, such as the HHP/HAAS, can contribute importantly to identifying the most relevant CV risk factors, the long term effect on the brain of these risk factors, and the possible effects of intervening early on late-life dementia outcomes. This current study shows roughly 20 % of dementia cases are attributed to higher levels of SBP. BP control is an intervention that is known to benefit for many other reasons and this study shows it may also help reduce dementia in late-life The study also shows, similar to cardiovascular outcomes the risk for dementia is raised even in persons with SBP levels in the prehypertension range. These data, in combination with results from clinical trials, should provide the basis for developing monitoring and evaluation tools to track trends in dementia as it relates to changes in the population distribution of BP.