The Dynamics of Blood Pressure and Cognitive Functioning: Results From 6-Year Follow-Up of an Elderly Cohort


Ester Paran, MD, Hypertension Unit, Soroka University Medical Center, Beer-Sheva, Israel


J Clin Hypertens (Greenwich). 2011;13:813–817. ©2011 Wiley Periodicals, Inc.

The association between blood pressure (BP) and cognitive functioning in the elderly is still under debate. Theoretically, high BP could either prevent or enhance cognitive impairment. The authors assessed the changes that took place in BP and cognitive functioning over 6 years. A total of 318 noninstitutional elderly (81% of the survivors) were re-evaluated. BP was measured and a cognitive test was performed. Elderly patients who had higher systolic BP (SBP) and scored lower on the Mini-Mental State Examination (MMSE) at baseline were less likely to survive. At follow-up, the proportion of patients with normal or normalized BP by treatment increased. Considerable changes in SBP were observed. Most cognitive functions declined during follow-up; however, decline in SBP was associated with better verbal fluency and memory. Both an increase and a decline in SBP were associated with better MMSE scores. Changes in diastolic BP (DBP) were less evident and DBP was not related to cognitive functioning. The current study demonstrates the importance of studying the dynamics of both BP and cognition over time. It appears that transition from hypertension to normotension improves cognitive functions. Survival processes may restrict the evaluation of the BP-cognition interaction over time.

The mortality and morbidity consequences of hypertension in the elderly, and isolated systolic blood pressure (BP) in particular, have been well established since the early 1990s.1–5 Treating high BP in these patients can improve health and longevity. Meta-analyses of 9 large-scale, randomized, controlled studies on the effect of antihypertensive treatment in the elderly have indicated a decline of >30% in the incidence of stroke, 20% to 25% in coronary heart disease, and >10% in overall mortality.6–8

Chronically elevated BP concords with pathological changes observed in the brain tissue.9,10 Vascular remodeling, macroscopic cerebrovascular lesions, cerebral atrophy, ventricular enlargement, cerebral microbleeding, perfusional declines in the subcortex, and impaired cerebral autoregulation have all been observed more often among hypertensive patients. These observations have led to the assumption that controlling high BP among elderly patients will slow down these cerebral processes and thus decelerate cognitive impairment. However, low BP is also associated with cerebral changes that could enhance cognitive deterioration: cortical atrophy in frontal as well as in parietal areas, perfusion, and reduced blood flow have been described.11–13

The findings regarding the association between BP and cognitive functioning are far from conclusive.13–16 Recent reviews suggest that variability in research design and the different methodologies used hamper the understanding of the relationship between BP and cognition and restrict generalization.17 Both BP and cognitive ability change over time. Yet, while changes in the cognitive performance of the elderly have been quite intensively studied, the relationship between changes in BP, not introduced by a specific intervention, and cognitive ability have been relatively neglected.

The present study reports findings from a 6-year follow-up of a noninstitutionalized elderly population (aged 70–85 years at baseline) that originally demonstrated a positive association between BP and cognitive performance.18 The purpose of this study was to explore the changes that took place both in BP and in cognitive functioning during this period and the association between them.


In 1998–1999 (baseline survey, t1), a random sample of 750 noninstitutionalized elderly, aged 70 to 85 years, living in Beer-Sheva, Israel, was drawn from 3 sources.18 Inclusion criteria were age, willingness to participate, and cognitive ability to answer the sociodemographic questions in the study. Exclusion criteria were active cancer, severe cerebrovascular accident, and target organ or other serious organ damage. Data were collected from 517 patients. Interviews took place in patients’ homes and 495 had complete records. The age and sex distribution of respondents and nonrespondents to the survey was similar to those of the elderly population in the Beer-Sheva region.

A follow-up study was conducted in 2005–2006 (follow-up survey, t2). Of the original sample, 127 had died and 15 were lost to follow-up. Of the survivors, 18 were too ill to be interviewed, 18 refused to participate, and 21 could not be located at the addresses or the telephone numbers registered in different sources. The remainder of the survivors (318 [80.9%]) were successfully interviewed. The mean age at follow-up was 82.9±3.53 years. Men comprised 29.4% of the survivors, 36.8% were married, and 12.9% lived alone. A total of 36.2% had at least some college education. At both times the study was approved by the institutional ethics committee and the patients signed written consent forms.

The Instrument

A structured questionnaire included the following measures18:

• Sociodemographic characteristics: age, sex, educational attainment, marital status, country of birth, and age at immigration.

• Health status: self-reported chronic conditions and cardiovascular risk factors (diabetes, hypercholesterolemia, the consumption of tobacco and alcohol). Disability was measured by the physical performance subscale taken from the 36-Item Short-Form Health Survey (SF-36).19

• Cognitive functioning was measured by the Mini-Mental State Examination (MMSE),20 which is the most commonly used instrument for comprehensive assessment of cognitive functioning. In addition, Verbal Fluency Tests for Letters,21 Buschke Selective Reminding Test for long-term memory,22 and the Trail-Making Test for concentration23 were administered.


In both waves, the first contact with the respondent was a short telephone conversation performed by a trained interviewer, inviting each patient to participate in the study and to set a time for a home visit. On average, each interview lasted 40 minutes. During the interview, BP was measured 3 times in a sitting position (1 minute apart) and once after 60 seconds of standing. All 4 measurements were taken before the cognitive tests were administered. The mean of the first 3 readings was calculated to establish the mean systolic BP (SBP) and diastolic BP (DBP) value of each respondent. Patients were asked to present all medications taken to the interviewer and report all medically significant events (eg, cardiovascular events, onset of chronic diseases, hospitalizations) that had occurred since the last interview. All information was recorded and included in the database. In a randomly selected 10% of the cases, these data were compared with the electronic medical files of the participant. A correlation of 0.91 was observed between the self-reported and the documented morbidity.

Statistical Analysis

Logistic regression was employed to detect the variables that were associated with death. Patients were categorized into 4 groups according to the changes observed in BP: large increase, moderate increase, no change, and decline. The rationale behind this classification was that both increase and decline in BP increases the risk for undesirable outcomes.9,11,12,24,25 Multiple regression (least squares) was used to assess the changes in cognitive functioning during the 6-year follow-up in relation to the changes in BP. In order to detect those changes, baseline cognitive scores were included in the equations. All analyses were performed with SPSS version 17 (SPSS Inc, Chicago, IL).


During the 6 years, 25% (127) of the patients died, which is a somewhat lower rate than expected in a random population of Israelis that age. The lower death rate can be explained by the exclusion criteria used in the first wave of this study, as stated in the methods section.

As presented in Table I, women had greater odds to survive: older age, heart disease, and disability. Moreover, death was related to the two main variables under study, BP and cognitive functioning. Higher SBP and lower MMSE score increased the risk of death. DBP was not related to survival, nor any of the other cognitive faculties studied. Surprisingly, the number of chronic conditions was positively associated with survival, probably the result of the exclusion criteria. Pulse pressure was not related to survival.

Table I.   Odds of Survival After 6 Years
Confounding VariablesOdds RatioaConfidence Interval
  1. Abbreviation: MMSE, Mini-Mental State Examination. aOdds ratio by logistic regression analysis. bP<.01. cP<.05. dTaken from the 36-Item Short Form Health Survey (SF-36): range from 0–30. eP<.001. fP=.051.

Male sex0.48b0.30–0.77
Number of chronic conditionsd1.29c1.05–1.59
Heart disease0.51e0.28–0.90
Systolic blood pressure0.97e0.96–0.98
Diastolic blood pressure1.02f1.00–1.04

During the same time, considerable changes took place both in cognitive functioning of the survivors and their BP. The correlation between SBP and DBP at baseline and at follow-up was just 0.30 and 0.24, respectively, which means that BP at baseline explains no more than 9% of the variance in BP at follow-up, which is statistically significant but of little predictive value. The correlation between the MMSE scores obtained at t1 and t2 was 0.43 and 0.32, respectively, for verbal fluency and 0.08 for memory (between 1% and 16% overlap).

The mean SBP increased from 140.6 mm Hg at baseline to 148.7 mm Hg at t2 (t=5.9, P<.001). About half of the patients (47.8%) remained in the same treatment/hypertension category (Table II). Two thirds of those who had normal BP at baseline (58 of 91) became hypertensive. Of these, one half received treatment but nonetheless had a higher BP than recommended. Of those who were normalized at baseline, BP had become uncontrolled in 46.3% of patients despite treatment. In the 59 hypertensive patients who were not treated for the condition at baseline, about one quarter (14, or 24.1%) were still not treated 6 years later. Yet, despite the increase in the number of hypertensive patients treated, the majority had higher BP levels than recommended (77.8%, 29 of the 37 treated). Most persistent were those patients who were treated at baseline although not adequately normalized: about three quarters of them (74 of 101) remained in the same category at follow-up. The changes in DBP were smaller, showing a decline from 79.0 mm Hg at t1 to 78.1 mm Hg at t2 (t=1.0, P=.3).

Table II.   Changes in Hypertensive Status During 6 Years
Follow-Up CategoryBaseline Category
NormotensiveaNormalized HypertensiveUntreated HypertensiveTreated but not NormalizedTotal
  1. aBlood pressure in the normal range without antihypertensive treatment. bData on treatment were missing from 12 cases. The numbers in italics emphasize the proportion of respondents for whom there was no change in hypertension status. The change from baseline to follow up is statistically significant, as indicated by the chi-square value, χ2=70.9, d.f.=9, P<.001.

Normotensivea37.9 (n=33)11.1 (n=6)13.8 (n=8)8.1 (n=8)18.5 (n=55)
Normalized hypertensive11.5 (n=11)38.9 (n=22)13.8 (n=8)14.1 (n=15)17.8 (n=56)
Untreated hypertensive17.2 (n=17)3.7 (n=2)24.1 (n=14)4.0 (n=4)11.7 (n=37)
Treated but not normalized33.3 (n=30)46.3 (n=25)48.3 (n=29)73.7 (n=74)52.0 (n=158)
Total100 (n=91)100 (n=55)100 (n=59)100 (n=101)100 (n=306)b

Since grouping the sample according to this 4×4 typology would have resulted in too many small cells for meaningful analysis, patients were classified into quartiles, which reflect changes in SBP over time: large increase (≥20 mm Hg), moderate increase (9–19 mm Hg), no change (between −8 mm Hg and +8 mm Hg), and decline (SBP decreased by ≥10 mm Hg). Most patients whose SBP increased by ≥20 mm Hg were normotensive (SBP <140 mm Hg) in t1 (61 of 88), and 87% (53 of 61) became hypertensive during follow-up. In the group with the moderate increase in SBP, 29 patients were normotensive and 31 were hypertensive. During follow-up, another 13 patients became hypertensive. Most patients whose SBP declined were hypertensive at baseline (63 of 72) and became normotensive at follow-up (increase from 9 patients to 43).

The cognitive performance of the majority of the patients declined during the follow-up period. The MMSE scores of the whole group of survivors declined, on average, by 2.64±5.19 scores (tpaired=8.94, P<.01). Verbal fluency declined by 3.5±7.5 words for the letter b and by 1.7±4.4 for letter z (tpaired=7.99 and 6.59, P<.001, respectively). The change in memory was not statistically significant.

Table III presents the degree to which changes in BP are related to cognitive functioning. After controlling for the effects of education and economic state, baseline BP, and respective baseline cognitive scores, an improvement in verbal fluency and memory was observed among those whose SBP had declined during the 6 years of follow-up.

Table III.   SBP Change and Cognitive Functioning at Follow-Up (Standardized β Coefficients)
DependentMMSE at Follow-UpVerbal Fluency at Follow-UpMemory at Follow-Up
PredictorsLetter bLetter z
  1. Abbreviations: MMSE, Mini-Mental State Examination; SBP, systolic blood pressure. aP<.001. bP<.01. cP<.05. Models are linear least-squares regression models. Outcome variables are cognitive functioning at follow-up. Cognitive score at baseline are the respective scores attained at the first evaluation.

Economic state0.21b0.16b0.16b0.13c
Cognitive score at baseline0.25a0.080.22a0.05
SBP change
 Large increase0.14c0.080.020.04
 Mild increase0.
 No changeReferenceReferenceReferenceReference

The association between MMSE scores and the SBP change observed is more complex. Both a decline of >10 mm Hg and an increase of ≥20 mm Hg in SBP were associated with higher MMSE scores than the scores obtained by patients whose SBP did not change during follow-up (β=0.13 and β=0.14, P<.05, respectively). In other words, the general cognitive performance, as measured by the MMSE, improved in two groups: among patients whose SBP increased the most and among patients whose SBP declined during the follow-up period. The level of DBP and the change in DBP that took place during the follow-up period were not statistically significant in relation to any of the cognitive tests scores obtained at follow-up.


The relationship between BP and cognitive functioning is far from being resolved. Pathological changes in the brain associated with hypertension lead to conflicting hypotheses and past research has indeed yielded inconsistent results. The aim of the study reported here was to contribute to the understanding of the association between BP and cognitive functioning by following-up elderly community residents over 6 years.

The first important observation in this longitudinal study of seniors aged 70 to 85 years was the mortality pattern. A total of 75.4% of the patients survived 6 years, although Israeli life tables for this age predicted 66% survival. The higher rate of survival in our study can be explained by the inclusion-exclusion criteria defined at baseline, aimed at ensuring patients’ capability to provide the data and perform the cognitive tests.18 These criteria were determined on the basis of previous studies in the field and resulted in the inclusion of relatively healthy elderly. Similar to most of these studies, our patients did not have health conditions that threatened their life in the short-term.26

As expected, the risk of death was higher among men, disabled patients, and those with heart conditions. Survivors reported more chronic conditions at baseline, a surprising finding that is probably the result of the same inclusion-exclusion criteria described above. Thus, the selected patients had nonfatal conditions such as low back pain, digestion problems, and skin disorders. Moreover, the number of chronic conditions was positively related to normalized BP, which suggest that these individuals were well treated for their complaints. Additionally, this finding may reflect differences in awareness to health issues and differences in health behavior.

Second, and more importantly, death was not random with regard to the two main variables of interest in this body of research. Those who passed away were more likely to have higher SBP and to perform poorly on the MMSE at baseline. In other words, like all follow-up studies, our current analysis is based on a sample that is curtailed on the two main variables.

Neither BP nor cognitive functioning were stable during the 6 years of follow-up. As expected, SBP increased significantly during follow-up, and many patients who were normotensive at baseline became hypertensive, while hypertensive patients who were not being treated at baseline started treatment.27 These findings are in agreement with the report of the Framingham Heart study, which reported up to a 90% lifetime risk of hypertension.28 As a result of these dynamics, just half of the survivors remained in the same treated-normal-normalized BP category.

Moreover, at follow-up, higher values of SBP were more often measured among patients whose SBP values had been the lowest at baseline, and patients with high SBP values at baseline had now, on average, the lowest SBP, most of them due to treatment. These findings could be the result of changes in medical practice that took place during the follow-up period. Health care organizations (the sick-funds) held several campaigns that encouraged both care providers and the general population to pay more attention to SBP.

The cognitive performance of the patients declined during the 6 years of follow-up except for memory. These findings are similar to those observed in previous studies.29 In this study, better verbal fluency and memory were associated with a decline in SBP but were not significantly related to DBP. Yet, the association between BP and the performance on the MMSE, the comprehensive cognitive functioning test most commonly employed in this field of research, was intriguing: the best performance was observed in patients who experienced a large increase in SBP (>20 mm Hg) and in those in which it declined, compared with patients whose SBP remained relatively unchanged throughout the follow-up.

The current study allowed us to explore the dynamics of both BP and cognition over time, ie, to observe the interaction between the changes that took place in BP and cognitive abilities. At baseline,18 verbal fluency was not significantly related to BP, while the association between BP and memory took on a J-shaped curve and MMSE scores were linearly and positively related to SBP. During the 6 years of follow-up, verbal fluency and memory improved for patients whose SBP declined. Note that two thirds of the patients whose SBP declined changed from a status of hypertension to normotension.

As for the comprehensive MMSE, patients whose SBP declined the most during follow-up performed better on the MMSE, a pattern similar to the more specific tests for memory and verbal fluency. At the same time, it seems that some increase in SBP in elderly with low SBP could also improve their general cognitive performance. It would appear that transition from hypertension to normotension and increase from low SBP have comprehensive beneficial effects on cognitive functioning. Parallel changes in BP and cognitive abilities have been reported in previous studies (eg, the Göteborg Population Study30 and in the Honolulu-Asia Aging Study31). Yet, while these studies did not explore the effect of the magnitude of change in BP, they do point to the importance of such an approach for understanding the association between SBP and cognitive faculties.


Two limitations should be considered. First, all medical information and prescribed medications were self-reported. Only in a sample of 10% were these data validated. This could have biased the study, although, in the random 10% of cases, a more than satisfactory correlation was observed.

Second, our observations regarding the change in BP raise the question of regression to the mean. Nevertheless, we would like to note that a higher proportion of the patients were taking antihypertension treatment at follow-up than at baseline and that increase in SBP is a well-documented phenomenon in the process of aging.


Future research should thus further develop this line of research. At the same time, attention should be paid to the process of selective mortality: the longer the follow-up, the greater the influence of survival processes on the dependent variables on BP and cognitive functioning. These processes could hamper the possibility of generating sound conclusions on the relationship between the dynamics of BP and cognitive functioning along the life cycle.


Acknowledgments and disclosures:  This study was partly supported by the Israel National Institute for Health Policy and Health Services Research, grant #2003/151/a. The authors have no conflicts of interest to disclose.