|Home | About | Journals | Submit | Contact Us | Français|
To determine whether high blood pressure (BP) levels are associated with faster decline in specific cognitive domains.
Prospective longitudinal cohort.
Uniform Data Set of the National Institutes of Health, National Institute on Aging Alzheimer's Disease Centers.
One thousand three hundred eighty-five participants with a diagnosis of mild cognitive impairment (MCI) and measured BP values at baseline and two annual follow-up visits.
Neuropsychological test scores and Clinical Dementia Rating Sum of Boxes (CDR Sum) score.
Participants with MCI with two or three annual occasions of high BP values (systolic BP ≥ 140 mmHg or diastolic BP ≥ 90 mmHg) had significantly faster decline on neuropsychological measures of visuomotor sequencing, set shifting, and naming than those who were normotensive on all three occasions. High systolic BP values were associated as well with faster decline on the CDR Sum score.
Hypertension is associated with faster cognitive decline in persons at risk for dementia.
There is controversy as to whether hypertension is a risk factor for cognitive impairment and decline, with some studies finding a positive association1–7 and others not.8–12 recent evidence-based review13 of published studies conducted by an expert panel under the auspices of the National Institutes of Health (NIH) concluded that the evidence of such an association is weak, in part because of the heterogeneity in definitions of mild cognitive impairment (MCI) and hypertension and differences in hypertension ascertainment methods (e.g., measured blood pressure (BP) vs reliance on self-report). In the United States, nearly 70% of persons aged 60 and older have hypertension,14 and it is estimated that 15 million to 18 million persons in this age group will develop dementia by 2050.15 Therefore, determining whether an association exists between high BP and cognitive function is important for targeting potential neuroprotective strategies. The results of clinical drug trials with antihypertensive agents have been mixed concerning their efficacy in preventing cognitive decline and dementia onset,13,16 although encouraging findings are available from a recent clinical trial employing lifestyle changes. One study17 found that prehypertensive and hypertensive adults receiving treatment with diet and aerobic exercise over a 4-month period showed greater improvements in executive functioning, memory, and psychomotor speed than those exposed to a diet intervention alone or a placebo condition of standard care. Systolic and diastolic BP decreased significantly over the study period for the intervention group but not the control group.
The current study examined whether there is an association between high BP and decline in cognitive status in individuals with MCI over a 2-year period. The importance of the effect of adequate BP control on cognitive performance was demonstrated in a cross-sectional study of communityresiding older adults.18 Irrespective of a prior diagnosis of hypertension, persons with high BP (systolic ≥ 140 mmHg or diastolic ≥ 90 mmHg) at the time of neuropsychological testing performed worse than normotensive individuals on measures of visual memory, motor speed, and visuomotor integration. Persons with a prior diagnosis and high BP levels were most vulnerable to poor performance.
BP levels were examined at annual follow-up visits, rather than at a single baseline visit, to determine whether BP was routinely normotensive or high and whether this, in turn, affected the trajectory of cognitive changes. Rather than a single measure of overall cognitive status, participants in the current study underwent multiple tests examining attention, memory, language, and executive functioning. In older adults without dementia and those with MCI and Alzheimer's disease, hypertension is associated with a cognitive phenotype characterized by poorer attention and executive functioning and slower processing speed.10,19–25 Thus, it was expected that these same areas would be most sensitive to the chronic effects of high BP.
Information was collected as part of the Uniform Data Set (UDS), a standardized assessment and data protocol maintained by the National Alzheimer's Coordinating Center, with 31 participating NIH, National Institute on Aging (NIA) Alzheimer's Disease Centers (ADCs) nationwide.26,27 It was hypothesized that individuals with MCI with high BP readings, according to published guidelines for hypertension,28 on more than one occasion would exhibit faster overall cognitive decline than those with normotensive levels on all occasions, with greater vulnerability of attention, executive functioning, and speeded performance.
Information from the UDS as of January 2011 was used. Recruitment strategies vary across the ADCs, and participants may come from clinics or the community.27 Criteria used by all ADCs for diagnosing MCI follow guidelines set forth by an expert panel.29 These include clinical judgment that a person is not cognitively normal and does not meet diagnostic criteria for dementia, has preserved or only minimally impaired functional abilities, and has evidence of cognitive impairment or decline based on self- or informant report and objective cognitive tests.
Inclusion criteria required that participants have a diagnosis of MCI from the clinicians at each center, no reported history of a remote or recent stroke or transient ischemic attack, and BP values available at baseline and at least two annual follow-ups. All participants signed consent forms that the institutional review boards at their study sites approved. The outcome measures used in the analyses are listed below.
Clinical dementia rating sum of boxes (CDR Sum).30 The CDR was administered using a structured interview format with the participant and his or her informant to assess the participant's current cognitive and functional status. Memory, orientation, judgment and problem solving, community affairs, home and hobbies, and personal care are each rated for their level of impairment. The CDR Sum provides a composite of the overall level of impairment.
Cognitive test scores were based on the core battery of measures collected by the ADCs.27 Four cognitive domains were evaluated: attention, language, memory, and executive functioning. Attention was assessed according to the maximum number of correct trials for Digits Forward31 and the number of seconds needed to sequence numbers using a pencil (Trail-Making Test (TMT) Part A).32 Language was examined according to the 30-item version of the Boston Naming Test.33 The evaluation of memory included verbal episodic memory (immediate and delayed story recall)31 and semantic memory (timed generation of animal names in 60 s).34 Finally, executive functioning was measured using set shifting tasks involving mental manipulation of digits31 and rapid alternation of numbers/letters and symbols (TMT B and Digit Symbol).32,35
Systolic and diastolic BP readings were obtained at all three yearly visits. The UDS guidelines require that BP be measured with the individual seated.
Longitudinal linear regression analyses were conducted to determine whether there was a difference in cognitive change over time between those with high and normotensive BP readings (PROC MIXED, SAS, SAS Institute, Inc., Cary, NC). All participants had three observations, spaced approximately 1 year apart. Three years was chosen as the follow-up period because this was the largest number of visits that most participants had in common. Only 52% of participants with three visits had cognitive testing or measured BP values at the fourth visit, because of pending visits or incomplete data. It was therefore reasoned that a more-homogenous analytical data set with all members having the same number of visits was optimal. A repeated-measures analysis with a compound symmetry correlation matrix was used, equivalent to an analysis in which subjects are included as a random effect.
Separate models were first run to evaluate the main effects of time (a continuous variable coded 1, 2, and 3) and BP (normotensive or high, see below) on cognitive tests. The primary goal was to determine whether individuals with MCI with high BP worsened more rapidly than those with normal BP. To test this, an interaction term between time as a continuous variable coded as occasion 1, 2, and 3 and a dichotomous variable for high versus normotensive BP was next included in the model. Normotensive BP was defined as readings on all three occasions of systolic BP less than 140 mmHg and diastolic BP less than 90 mmHg.28 High BP was defined as Stage 1 hypertension and higher (systolic BP ≥ 140 mmHg or diastolic BP ≥ 90 mmHg) on two or three occasions.28 Analyses were also conducted examining the effect of one occasion of high BP on cognitive decline.
There were 11 dependent variables: CDR Sum, MMSE36 total score, number o seconds to complete TMT A and B, maximum number of correct trials for Digits Forward and Digits Backward, number of story units recalled immediately after hearing a story and after a delay, number of completed pairings within 90 seconds on the Digit Symbol subtest, number of correct responses on the Boston Naming Test, and number of animal names generated in 60 seconds. For TMT A and B, analyses were conducted of the original variable and the log-transformed variable, because the latter satisfied the normality assumption, and the former did not. Results were concordant, and the untransformed results are reported for ease of interpretation.
Variables that could confound the relationship between BP and cognitive function were entered a priori in all of the statistical models. These covariates were age (continuous); race (white; nonwhite); sex; education (no high school degree vs ≥ high school); and presence vs absence of a self-reported history of baseline diabetes mellitus, heart disease (e.g., myocardial infarction, atrial fibrillation, or congestive heart failure), or depression within the last 2 years. A self-reported history of hypertension within the last 2 years was also included in the models to assess the influence of actual BP readings on performance, irrespective of whether there was a report of hypertension.
Data from 1,385 participants with a baseline diagnosis of MCI were available for analysis. Two hundred four (15%) participants had high readings on all three occasions, and 323 (23%) had high readings on two of the three occasions involving Time 1 and Time 2 or Time 1 and Time 3. Preliminary analyses revealed that the individuals with three high readings had patterns of change over time on the cognitive tests similar to the patterns of those with two high readings, so these two groups were combined for analysis. Three hundred seventy-three participants (27%) had one high reading, and 485 (35%) had no high readings. The latter participants served as the reference group for those with two or more high readings and for those with one high reading.
Table 1 shows the baseline demographic and clinical features of the participants broken down according to the number of occasions with high BP readings. Participants with two or three high readings were significantly older than those with no high readings or one high reading. These latter two groups, in turn, did not significantly differ in age. Mean systolic and diastolic BP readings differed significantly between all three groups, as did the percentage of persons with a self-reported history of hypertension at baseline. Thirty-nine percent of those with no high readings, 53% of those with one high reading, and 65% of those with two or three high readings had a history of hypertension within the past 2 years. There were no significant differences between the groups in distribution of education, sex, or race, or in self-reported history of diabetes mellitus, heart disease, or depression at baseline.
Table 2 shows the unadjusted mean raw scores for the CDR Sum and the neuropsychological measures broken down according to number of occasions with high BP readings. The results of linear regression analyses, adjusted for potential confounders, are shown in Table 3. There were significant main effects of high BP on the number of seconds needed to complete TMT A and B. Those with high readings on two or three occasions were significantly slower than those with normotensive readings. As expected, performance worsened significantly over the 2 years for all outcome measures (time variable), with the exception of delayed paragraph recall.
Significant interactions were observed between number of high BP readings and time for three cognitive outcomes. Those with high readings on two or three occasions had a greater decline than those with normotensive readings on all occasions in number of seconds to complete TMT A and B and ability to name pictured items. There was a trend (P = .08) as well for a faster decline on the Digits Backward subtest for those with high readings on two or three occasions. Figure 1(A–D) depicts the time trends for individuals with high BP and those with normotensive values for these four tests (using observed baseline values and model-predicted values for years 2 and 3).
Less than 2% of the total sample reported experiencing an intervening stroke between the first evaluation and the follow-up occasions. Exclusion of these individuals did not change the results.
The above analyses were also performed comparing those with one occasion of high BP with those who were normotensive on all three occasions. There were no significant interactions for any of the outcomes, indicating comparable changes over time for the groups.
Most high readings were from high systolic rather than diastolic pressure. Sixty-three percent of participants had at least one systolic reading of 140 mmHg or greater, compared with 19% with at least one diastolic reading of 90 mmHg or higher. The percentages of individuals with two or more high readings were 38% and 4%, for systolic and diastolic BP, respectively. Thus, supplemental analyses were conducted focusing on the effect of systolic BP on performance over time. First, the effect of two or more high (≥ 140 mmHg) readings of systolic BP versus never having had high systolic BP was examined. After adjusting for confounders as before, significantly greater worsening over time was observed for those with two or more high readings than for those with no high readings on CDR Sum (P = .03), TMT A (P = .02), TMT B (P = .007), and the Boston Naming Test (P = .005). Average systolic BP readings across the three visits were also examined and compared with average systolic BP of 140 mmHg or greater versus less than 140 mmHg over three visits. Adjusting for potential confounders, significant interactions were found between group and time for TMT B (P = .04) and the Boston Naming Test (P = .02), whereas the interaction with time for TMT A was not significant (P = .06).
The results support an association between high BP and risk of cognitive decline in MCI. Individuals with high BP readings on two or three annual assessments experienced greater slowing on TMT A and B than those with no high readings on any occasion, as well as a trend for greater worsening on Digit Span Backward. Intervening strokes did not explain these findings. The observation of a faster decline in naming ability was unexpected but could reflect the time demands of this measure, because individuals are penalized if their responses exceed 20 seconds. High systolic BP appeared to account for the results based on the higher percentage of persons with high systolic (63%) versus diastolic (20%) readings, but the importance of diastolic BP should not be dismissed, because its role could not be adequately evaluated due to its low frequency of occurrence. Examination of relationships with actual systolic BP levels, based on the number of occasions with clinically significant high readings or averaged values, also demonstrated vulnerability on tasks requiring rapid performance and set shifting (TMT A and B) and expressive language (naming). Moreover, CDR Sum declined more precipitously in those with two or three occasions of high systolic BP.
These results are consistent with the findings of a previous study that recently demonstrated an association between hypertension and risk for cognitive decline in a Chinese population,3 but unlike that study, which examined the risk of a change in diagnosis from MCI to Alzheimer's disease, the current study examined specific cognitive domains that are especially vulnerable. In addition, unlike the previous study, the participants in the current study were excluded if they had a previous history of stroke or transient ischemic attack. Thus, the results demonstrate the influence of a vascular risk factor on cognitive changes in the absence of overt clinical disease.
The results of studies examining hypertension as a risk factor for cognitive decline and Alzheimer's disease have been mixed, with two recent reviews13,37 concluding that the overall evidence is weak. The current study circumvented some of the methodological problems mentioned in these recent reviews that can hamper interpretation of many studies, including reliance on self-reported hypertension at baseline and reliance on a single BP reading at baseline. Studies that rely solely on self-reports of hypertension may incorrectly classify some participants. Findings from the National Health and Nutrition Examination Surveys, which indicate that, between 2005 and 2006, 7% of the U.S. population had systolic BP readings of 140 mmHg or diastolic BP readings of 90 mmHg or higher. Yet, no healthcare provider had told these individuals that they had high BP. These results highlight the unreliability of self-report data. Overall, only 78% of hypertensive adults were aware that they had this condi-tion.38 Moreover, studies that rely on a single reading at one time point do not take into account whether the status of individuals changes. For example, in the data from the current study, 43% of those who were normotensive at baseline had one or two high BP readings at year 2 and 3 visits. Such changes would result in the misclassification of individuals, weakening associations with cognitive performance and functional outcome.
A limitation of this study is that the follow-up period was short (2 years). Only 52% of the sample had complete cognitive data and BP readings at 3 years. Given the desire to study a homogenous population with equal length of follow-up, using data for the third year would have halved the sample size. An alternative approach of imputing missing data for the third year was deemed inadvisable given the amount of missing data that would need to be imputed. Further follow-up would allow clearer determination of whether the differences between the hypertensive and normotensive individuals persist over time. Two aspects of the findings for the cognitive test results increase confidence in their validity. First, participants with two or three high readings had greater worsening of cognition over time than those with no high readings, whereas those with only one high reading did not. This suggests a doseresponse relationship whereby only those with more-sustained uncontrolled BP showed an effect. Another source of confidence is that the pattern of impairments is consistent with the clinical phenotype of slow processing speed and set shifting difficulty reported in the literature.19–25
Another limitation of this study is that an established database was relied on, as opposed to the study being designed and therefore allowing a mediation model for the association between hypertension and cognitive decline to be tested. There are potentially many mechanisms through which high BP exerts a deleterious effect on the course of cognitive function, including disruption of cerebral white matter integrity and the greater risk for white matter infarcts, greater brain atrophy, and silent strokes.39–42 Brain imaging is not a required data element of the ADCs, but findings of greater cerebrovascular changes in those classified with high BP would clarify the mechanism of the observed relationships. Neuropathological studies have also demonstrated an association between uncontrolled hypertension and neurofibrillary tangles and neuritic plaques.43–45 Studies using in vivo amyloid imaging techniques and studies examining relationships between hypertension and associated measures of chronic inflammation, including serum interleukin-6 and C-reactive protein, would be of interest. Other possible mechanisms, including treatment with antihypertensive medications and cholinesterase inhibitors, may have been operative as well. These potential associations will await longitudinal studies to further elucidate mechanisms for cognitive decline.
We thank the ADC participants for their willingness to devote their time to research and the staff members who work tirelessly to make the research possible.
Supported by the Emory Alzheimer's Disease Research Center (NIH-NIA 5 P50 AG025688) (AIL).
Sponsor's Role: This project was accomplished through the auspices of the National Alzheimer's Coordinating Center (NACC), a NIH-NIA funded Center that facilitates collaborative research. NACC provided the data used in this study under cooperative agreement number U01 AG016976.
Conflict of Interest: The editor in chief has reviewed the conflict of interest checklist provided by the authors and has determined that the authors have no financial or any other kind of personal conflicts with this paper.
Author Contributions: Conception and design: Goldstein, Levey, and Steenland. Analysis and interpretation of data: Steenland. Drafting of the manuscript: Goldstein and Steenland. Critical revision of the manuscript for important intellectual content: Goldstein, Levey, and Steenland. Obtaining funding: Levey. Administrative, technical, or material support: Levey. Supervision: Goldstein.