The squares and bars represent the mean values and 95% CIs of the effect sizes and the area of the squares reflects the weight of the studies. Diamonds represent the combined effects and the vertical dotted line represents the line of no association.
aComposite of dementia and cognitive impairment.
The squares and bars represent the mean values and 95% CIs of the effect sizes and the area of the squares reflects the weight of the studies. The vertical dotted line represents the line of no association.
The squares and bars represent the mean values and 95% CIs of the effect sizes and the area of the squares reflects the weight of the studies. The diamond represents the combined effect and the vertical dotted line represents the line of no association.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Hughes D, Judge C, Murphy R, et al. Association of Blood Pressure Lowering With Incident Dementia or Cognitive Impairment: A Systematic Review and Meta-analysis. JAMA. 2020;323(19):1934–1944. doi:10.1001/jama.2020.4249
Is there an association between blood pressure lowering with antihypertensive therapy and the incidence of dementia or cognitive impairment?
In this meta-analysis that included 12 trials with 92 135 participants for the primary outcome measure, blood pressure lowering with antihypertensive agents, compared with control, was associated with the development of a composite dementia or cognitive impairment outcome in 7.0% vs 7.5% of patients over a mean trial follow-up of 4.1 years, a difference that was statistically significant.
Lowering blood pressure may be associated with a lower risk of dementia or cognitive impairment.
The benefit of blood pressure lowering for the prevention of dementia or cognitive impairment is unclear.
To determine the association of blood pressure lowering with dementia or cognitive impairment.
Data Sources and Study Selection
Search of PubMed, EMBASE, and CENTRAL for randomized clinical trials published from database inception through December 31, 2019, that evaluated the association of blood pressure lowering on cognitive outcomes. The control groups consisted of either placebo, alternative antihypertensive agents, or higher blood pressure targets.
Data Extraction and Synthesis
Data were screened and extracted independently by 2 authors. Random-effects meta-analysis models were used to report pooled treatment effects and CIs.
Main Outcomes and Measures
The primary outcome was dementia or cognitive impairment. The secondary outcomes were cognitive decline and changes in cognitive test scores.
Fourteen randomized clinical trials were eligible for inclusion (96 158 participants), of which 12 reported the incidence of dementia (or composite of dementia and cognitive impairment [3 trials]) on follow-up and were included in the primary meta-analysis, 8 reported cognitive decline, and 8 reported changes in cognitive test scores. The mean (SD) age of trial participants was 69 (5.4) years and 40 617 (42.2%) were women. The mean systolic baseline blood pressure was 154 (14.9) mm Hg and the mean diastolic blood pressure was 83.3 (9.9) mm Hg. The mean duration of follow-up was 49.2 months. Blood pressure lowering with antihypertensive agents compared with control was significantly associated with a reduced risk of dementia or cognitive impairment (12 trials; 92 135 participants) (7.0% vs 7.5% of patients over a mean trial follow-up of 4.1 years; odds ratio [OR], 0.93 [95% CI, 0.88-0.98]; absolute risk reduction, 0.39% [95% CI, 0.09%-0.68%]; I2 = 0.0%) and cognitive decline (8 trials) (20.2% vs 21.1% of participants over a mean trial follow-up of 4.1 years; OR, 0.93 [95% CI, 0.88-0.99]; absolute risk reduction, 0.71% [95% CI, 0.19%-1.2%]; I2 = 36.1%). Blood pressure lowering was not significantly associated with a change in cognitive test scores.
Conclusions and Relevance
In this meta-analysis of randomized clinical trials, blood pressure lowering with antihypertensive agents compared with control was significantly associated with a lower risk of incident dementia or cognitive impairment.
Hypertension, especially in midlife, is associated with dementia and cognitive impairment later in life.1-4 Some randomized clinical trials have reported a lower risk of dementia with blood pressure–lowering treatment.5-7 However, results of some previous meta-analyses of randomized clinical trials that have evaluated the association of antihypertensive therapy with the risk of neurocognitive syndromes, in either primary or secondary prevention populations, have been inconclusive.8-11 Two clinical trials have been recently published.12,13 The SPRINT MIND trial12 reported a lower risk of mild cognitive impairment in individuals randomized to an intensive blood pressure target group. Conversely, the HOPE-3 study13 reported no significant reduction in the risk of cognitive impairment or dementia with combination antihypertensive therapy compared with placebo. An updated meta-analysis was performed to determine whether blood pressure lowering was associated with a reduced risk of dementia or cognitive impairment.
We performed a systematic review and meta-analysis, reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.14 The protocol was registered with PROSPERO (registration number CRD42019125088).
We developed the search strategy, without language restriction, for PubMed, EMBASE, and CENTRAL for articles published from database inception to December 31, 2019. The search terms included dementia, cognitive decline, cognitive impairment, blood pressure, hypertension, anti-hypertensive, and randomized controlled trials. The search strategy was peer-reviewed by a second information specialist. The full search strategy is included in the Supplement (eMethods 1). Three reviewers (D.H., C.J., and R.M.) independently screened titles and abstracts. Full texts were sourced for relevant articles. Inclusion criteria were assessed independently, and inconsistencies were resolved by consensus. The reference lists of included trials and other published meta-analyses were also reviewed for relevant articles.
Quiz Ref IDTrials were considered eligible if they were randomized clinical trials, compared blood pressure lowering with antihypertensive agents with a control, had at least 1 year of follow-up, included more than 1000 participants, and provided information on any of the prespecified outcomes. Control was defined as placebo, alternative antihypertensive agent, or higher blood pressure target (Table 1). Trials were required to report at least 1 of the following outcomes: dementia, cognitive impairment, cognitive decline, or change in cognitive test scores (Table 1). Trials that specifically recruited participants with known dementia or cognitive impairment at the start of the trial were excluded.
Data were extracted independently by 2 authors (D.H. and C.J.) using a standardized data extraction form. Extracted data were entered into a dedicated database and checked independently by 4 authors (R.M., M.C., E.L., and M.C.). We extracted the following data: study characteristics, baseline demographics of participants, description of the intervention, cumulative blood pressure changes, incidence of dementia and cognitive impairment, and cognitive test scores. The cumulative blood pressure change (net change in systolic blood pressure from baseline to longest follow-up between groups) was reported in 10 of the included trials and the difference between the systolic blood pressure of the groups at trial end was reported in the other trials. We reported outcomes at the point of longest follow-up.15 Majority primary prevention populations were defined as those in which greater than 50% of participants had no history of cardiovascular events. All other populations were considered majority secondary prevention populations.
The primary outcome of this meta-analysis was dementia or cognitive impairment. For our primary analysis, we used a hierarchical approach in which we included trials that reported incident dementia, or a composite of dementia or cognitive impairment (if dementia alone was not reported), on follow-up. We chose this approach to maximize the number of clinical trials included in our primary analysis, while also giving priority to the most clinically relevant neurocognitive outcome reported in trials. In addition, cognitive impairment and dementia represent a continuum of the same neurocognitive syndrome, and we expected blood pressure lowering using antihypertensives to have a consistent association with both. The definition of dementia was criterion referenced in 7 of the included trials (based on International Classification of Diseases criteria, the Diagnostic and Statistical Manual of Mental Disorders criteria, or criteria from an adjudicated panel), clinically based in 2 trials, and included in a composite in the remainder of the trials (Table 1).
The secondary outcomes were cognitive decline and mean change in cognitive test scores. The definition of cognitive decline varied among trials, and we used a definition of cognitive decline when the cognitive score decreased by an absolute value within the study period (eg, 3 points in the Mini-Mental State Examination [MMSE] score), alone or combined with below a cut point in cognitive score. All included studies reported a cognitive test score.
We used the Cochrane risk of bias tool16 to assess methodological quality of eligible trials. Trials were assessed on random sequence generation, allocation concealment, blinding of participants and health care personnel, blinded outcome assessment, completeness of outcome data, evidence of selective reporting, and other biases. Risk of bias assessments were performed independently by 2 reviewers (D.H. and R.M.), and disagreements were resolved by a third reviewer (C.J.). If 2 of the domains were rated as high, the study was considered to be at high risk of bias. A risk-of-bias summary table was created in Review Manager, version 5.3. Additional details are included in the Supplement (eTable 1, eFigure 1, and eFigure 2).
A descriptive analysis of each individual trial is provided in Table 1. Baseline, follow-up, and mean difference in blood pressure for each trial are reported in Table 2. For dichotomous outcomes (dementia, cognitive impairment, and cognitive decline), odds ratios (ORs) and 95% CIs were estimated for each trial. Weighted pooled treatment effects were calculated using restricted maximum likelihood estimation to fit a random-effects meta-analysis model. For continuous outcomes (eg, MMSE score), the mean change from baseline to follow-up was analyzed. If this was not reported, the mean between-group difference reported at follow-up was used. Standard errors were calculated by converting 95% CIs using the following formula: SD = √N × (upper bound of the CI − lower bound of the CI)/3.92.17 The difference in MMSE score change between the intervention and control group was calculated when the difference was not reported in the trial. A pooled mean difference and 95% CI was calculated using a random-effects meta-analysis. A positive mean difference implies that the intervention, compared with the control, had a smaller magnitude of decrease in MMSE score between baseline and follow-up (ie, reduced cognitive decline on testing). For additional cognitive test scores, we calculated a pooled mean standardized difference (Cohen d) using a random-effects meta-analysis model. The variability across studies due to heterogeneity was investigated using forest plots and I2 statistics. Publication bias was assessed using a funnel plot (eFigure 3 in the Supplement). Two trials had 2 investigational treatment groups with a common control group.7,18 To prevent double counting and a unit-of-analysis error, we split the common control group into 2 equal-sized groups.17
A priori subgroup sensitivity analyses that assessed pooled estimates for trials that reported cumulative blood pressure change that was above and below the median cumulative blood pressure change and trials that reported years of follow-up above and below the median number of years of follow-up were performed. We tested for an interaction between subgroup relative risks by dividing the difference in log relative risk by its standard error.19 We completed meta-regression analyses to evaluate the association of select variables with treatment effect estimates, including baseline mean systolic blood pressure, years of follow-up, or cumulative systolic blood pressure change. In post hoc analyses, absolute risk reductions (ARRs) and 95% CIs were calculated for each study, the Mantel-Haenszel method was used to obtain a pooled estimate of the risk difference, and bootstrapping was used to estimate the ARRs for trials that reported dementia only. In addition, a sensitivity analysis that included only studies at low risk of bias was performed and the fragility index was calculated for the primary outcome. Statistical analyses were performed using the Metafor package20 in R statistical software, version 3.5.3. Comparisons were 2-tailed using a threshold of P ≤ .05 for significance for all analyses except for subgroup interactions, in which we used a threshold of P ≤ .10 for significance.21
The systematic search of articles published before December 31, 2019, identified 1543 results. After title and abstract screening, 163 articles were considered potentially relevant. Fourteen studies, available as 22 articles, were included after full-text review (eFigure 4 in the Supplement). Twelve studies reported the incidence of dementia (n = 9) or composite of dementia or cognitive impairment (n = 3) on follow-up and were included in the primary meta-analysis.5-7,12,13,22-27 Two studies were used for secondary outcomes only.18,28
In total, 96 158 participants were enrolled, comprising 394 558 participant-years of follow-up. The mean (SD) age of trial participants was 69 (5.4) years and 40 617 participants (42.2%) were women. The mean baseline systolic blood pressure among participants was 154 (14.9) mm Hg and the mean baseline diastolic blood pressure was 83.3 (9.9) mm Hg. The mean (range) duration of follow-up was 49.24 (26.4-68.4) months. The publication year ranged from 1994 to 2019 (Table 1). Nine trials were of a majority primary prevention population,5,6,12,13,18,22,24,25,28 3 trials were of a poststroke secondary prevention population,23,26,27 and 2 trials were of participants with cardiovascular disease.7 Ten trials were placebo controlled,5-7,13,18,22-26 3 trials compared different blood pressure targets,12,27,28 and 1 compared 2 antihypertensive agents alone or in combination (resulting in 2 comparisons).7
Risk of bias was assessed in all 14 trials (eTable 1, eFigure 1, and eFigure 2 in the Supplement). The overall risk of bias was deemed low in 11 trials, unclear in 1 trial, and high in 2 trials. The majority of trials (n = 11) were double-blind randomized clinical trials with prespecified outcomes and 1 was single-blinded.18 Two trials were open label,12,27 while 1 trial had a double-blind phase followed by open-label observational follow-up.5 Randomization sequence was adequately generated in 13 studies and 13 adequately concealed allocation. Reporting bias was noted in 1 trial.27 There was no evidence of publication bias for the primary outcome (Egger test, −0.53; P = .61).
Twelve trials reported dementia or cognitive impairment on follow-up (92 135 participants).5-7,12,13,22-27 Dementia or cognitive impairment was diagnosed in 2992 participants in the intervention group and 2558 participants in the control group. Blood pressure lowering with antihypertensive agents compared with control was significantly associated with a reduction in dementia or cognitive impairment (7.0% vs 7.5% over a mean trial follow-up of 4.1 years; OR, 0.93 [95% CI, 0.88-0.98]; ARR, 0.39% [95% CI, 0.09%-0.68%]) (Figure 1). Heterogeneity was low (I2 = 0.0%). A sensitivity analysis that divided trials by cumulative blood pressure change above and below the median cumulative blood pressure change did not reveal a significant difference between subgroups (P value for interaction = .13) (Figure 2; eFigure 5 in the Supplement). A sensitivity analysis that divided trials by baseline blood pressure above and below the median baseline blood pressure did not reveal a significant difference between subgroups (P value for interaction = .36) (Figure 2; eFigure 6 in the Supplement). Meta-regression analysis showed no significant association of select variables with treatment effect estimates, including age, baseline systolic blood pressure, years of follow-up, and cumulative systolic blood pressure change (eFigure 7 in the Supplement).
In post hoc analyses, for trials that employed a criterion-referenced definition for diagnosis of dementia (7 trials; 41 719 participants), blood pressure lowering was significantly associated with a reduction in dementia (OR, 0.87 [95% CI, 0.78-0.97]; ARR, 0.20% [95% CI, 0.05%-0.70%]). A sensitivity analysis that only included studies at low risk of bias did not materially alter the findings (OR, 0.94 [95% CI, 0.877-0.997]) (eFigure 8 in the Supplement). The fragility index for a meta-analysis of the primary outcome was 9.29
Eight trials reported cognitive decline on follow-up (67 476 participants).6,7,12,13,23,24,26 Cognitive decline was reported in 5513 participants in the intervention group and 4468 participants in the control group. Blood pressure lowering with antihypertensive agents compared with control was significantly associated with a reduction in cognitive decline (20.2% vs 21.1% over a mean trial follow-up of 4.1 years; OR, 0.93 [95% CI, 0.88-0.99]; ARR, 0.71% [95% CI, 0.19%-1.2%]) (Figure 3). Heterogeneity was low to moderate (I2 = 36.1%). Sensitivity analysis by cumulative change in blood pressure (above and below median) showed a significant association for cumulative blood pressure change above the median (OR, 0.89 [95% CI, 0.82-0.96]), but there was no statistically significant association for cumulative blood pressure change below the median (OR, 0.98 [95% CI, 0.92-1.05]) (P value for interaction = .07) (Figure 2; eFigure 9 in the Supplement). Sensitivity analysis by baseline blood pressure above and below the median reported no significant subgroup interaction (P value for interaction = .74) (Figure 2; eFigure 10 in Supplement). Meta-regression analysis showed no significant association of select variables with treatment effect estimates, including age, baseline systolic blood pressure, years of follow-up, and cumulative systolic blood pressure change (eFigure 11 in Supplement).
Eight trials reported a change in cognitive score as an outcome.5,6,13,18,23,24,27,28 For the meta-analyses, change in MMSE score in 5 trials,5,6,23,24,28 change in Trail Making Test score in 2 trials,13,18 and change in Cognitive Abilities Screening Instrument z score in 1 trial27 were used. Three studies reported baseline cognitive scores but not follow-up scores, and these data were insufficient to include in the meta-analysis.7,26 Blood pressure lowering with antihypertensive agents compared with control was not significantly associated with a difference in the standardized mean cognitive score (standardized mean difference, 0.25 [95% CI, −0.10 to 0.61]; P value for heterogeneity < .01; I2 = 99.5%; Q = 853.24) (eFigure 12 in the Supplement). For trials that reported change in MMSE score, blood pressure lowering with antihypertensive agents compared with control was not significantly associated with a difference in mean MMSE score (mean difference, 0.44 [95% CI, −0.22 to 1.10]; I2 = 98.5%; Q = 143.17) (eFigure 13 in Supplement).
Quiz Ref IDThis meta-analysis, which included 12 trials with 92 135 participants for the primary outcome analysis, found that blood pressure lowering with antihypertensive agents compared with control was significantly associated with a lower risk of dementia or cognitive impairment. This study builds on previous meta-analyses that have examined the association of blood pressure lowering and dementia and includes the largest number of randomized clinical trials to date. A 2013 pooled analysis that combined randomized clinical trials and observational studies reported a similar risk reduction with treatment of hypertension to the current analysis, but no significant association in an analysis of trials alone.10 A meta-analysis by van Middelar et al30 reported a similar, but nonsignificant, magnitude of association of blood pressure lowering, and included 2 trials that evaluated multicomponent lifestyle interventions rather than blood pressure lowering alone. Both these meta-analyses, and Cochrane reviews, were published before the SPRINT MIND and HOPE-3 trials.9,10,30 The most recent meta-analysis, by Peters et al,11 which included the SPRINT MIND trial, reported an association of blood pressure lowering with reduced risk of dementia (OR, 0.93 [95% CI, 0.86-1.00]) and included fewer trials than the current meta-analysis (8 trials) due to different selection criteria. In an analysis confined to trials that reported a between-group blood pressure difference of greater than 10 mm Hg (4 trials), Peters et al11 reported an OR of 0.88 (95% CI, 0.78-0.98), but did not report a P value for interaction. The approach taken in the current meta-analysis resulted in the inclusion of a larger number of clinical trials and a more extensive panel of reported outcome measures (eg, cognitive decline and mean change of cognitive test scores) and in the completion of a meta-regression for preselected variables. Although the increased number of clinical trials resulted in a statistically significant summary estimate, the upper bound of the CI was close to 1.0, which should prompt some caution in interpreting the findings as definitive evidence of an association of blood pressure lowering with dementia or cognitive impairment.
Quiz Ref IDAlthough observational studies report hypertension to be an important risk factor for dementia,1,3,4,31 the benefit of blood pressure lowering on dementia or cognitive impairment in clinical trials is modest (OR, 0.93 [95% CI, 0.88-0.98]) and lower than the risk reduction for stroke.5,6,18,22-24 The causes of neurocognitive syndromes are more heterogenous than stroke, including Alzheimer disease and other causes, and the population-attributable fraction of hypertension for dementia is lower than that reported for stroke based on indirect comparison of studies.32,33 In addition, the association of hypertension with neurocognitive syndromes, mediated through chronic covert vascular damage (ischemia, microhemorrhage, or atrophy34), appears to have an extended time lag between cause and clinical consequence, although dementia may be a complication of acute stroke. Observational studies relating blood pressure to neurocognitive outcomes have often required extended follow-up (eg, >10 y). Therefore, large sample sizes, with extended follow-up, are required to identify an effect of antihypertensive treatment on neurocognitive outcomes. These considerations may explain why many individual randomized clinical trials have failed to demonstrate a treatment effect.
Quiz Ref IDEpidemiologic studies have reported a stronger association of hypertension in midlife with neurocognitive outcomes in late-life than hypertension in late-life, during which a null or inverse association has been reported in some studies.35,36 These findings have led some investigators to speculate that populations included in some blood pressure trials may have been in an older age group that may not benefit from blood pressure lowering to prevent cognitive outcomes. Findings from the current meta-regression analyses would not fully support this contention because baseline age was not a determinate of treatment effect and the mean age of participants in included trials was 69 years at baseline.
Quiz Ref IDThese findings have the potential to inform public health strategies to reduce the burden of dementia globally. Effective screening for and management of hypertension is essential for reducing premature dependence from dementia. Although the lower risk associated with blood pressure treatment is modest for an individual, the effect at a population level, given the incidence of dementia in an aging population, may be considerable. Rates of blood pressure control are low, even in high-income countries, but especially in middle- and low-income countries, which carry the largest burden of dementia.37 The World Health Organization’s global action plan on the public health response to dementia recommends management of hypertension in midlife to reduce the risk of dementia, a recommendation supported by the results of the current study.38
Although there was a significant reduction of clinically important neurocognitive syndromes, there was no significant difference in mean change in cognitive testing, contrasting from the clinical outcomes. This finding supports the need for large, simple trials with clinically important outcomes to evaluate preventative interventions in populations.39 None of the included clinical trials reported dementia as their primary outcome measure in the original trial. When dementia was reported, it was as a secondary outcome with differences in outcome definition. In post hoc analyses confined to include clinical trials that reported criterion-referenced dementia, the association of blood pressure lowering and dementia was most evident (Figure 1).
This study has several limitations. First, there are inherent challenges in performing and interpreting a meta-analysis with heterogenous populations, interventions, and definitions of the outcomes of dementia, cognitive impairment, and cognitive decline. Second, the low incidence of dementia in all clinical trials, despite the large number of participants, reduced power to detect differences in treatment effect and limited exploration of subgroups or meta-regression. Third, underdetection of dementia in clinical trials due to preferential loss to follow-up of participants with dementia and the potential effect of survival bias (participants with blood pressure reductions are more likely to be alive) are unmeasured sources of potential error. Fourth, we are unable to identify the optimal blood pressure range for dementia prevention.
In this meta-analysis of randomized clinical trials, blood pressure lowering with antihypertensive agents compared with control was significantly associated with a lower risk of incident dementia or cognitive impairment.
Corresponding Author: Michelle Canavan, PhD, HRB Clinical Research Facility, Galway University Hospital, Newcastle Road, Galway, Ireland, H91YR7 (firstname.lastname@example.org).
Accepted for Publication: March 13, 2020.
Author Contributions: Drs Hughes and Judge had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis as co–first authors.
Concept and design: Hughes, Judge, Murphy, Loughlin, O'Donnell, Canavan.
Acquisition, analysis, or interpretation of data: Hughes, Judge, Murphy, Costello, Whiteley, Bosch, O'Donnell, Canavan.
Drafting of the manuscript: Hughes, Judge, Murphy, Loughlin, Costello, O'Donnell, Canavan.
Critical revision of the manuscript for important intellectual content: Hughes, Judge, Murphy, Whiteley, Bosch, O'Donnell, Canavan.
Statistical analysis: Hughes, Judge, Costello, Canavan.
Obtained funding: Judge.
Administrative, technical, or material support: Murphy, Loughlin, Canavan.
Supervision: Murphy, Loughlin, O'Donnell, Canavan.
Conflict of Interest Disclosures: Dr Judge reported receiving grants from the Wellcome Trust and the Health Research Board during the conduct of the study. Dr Whiteley reported receiving grants from the Chief Scientist Office outside the submitted work. Dr Bosch reported receiving personal fees from Bayer AG outside the submitted work. No other disclosures were reported.
Funding/Support: Dr Judge was supported by the Irish Clinical Academic Training Programme, the Wellcome Trust, the Health Research Board (grant number 203930/B/16/Z), the Health Service Executive, National Doctors Training and Planning, and the Health and Social Care Research and Development Division Northern Ireland. Dr O’Donnell was supported by the European Research Council (COSIP grant 640580).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional Contributions: John Ferguson, PhD (HRB Clinical Research Facility, Galway), contributed to the updated analysis (bootstrapping method for applying relative risk reduction to baseline risk of dementia). He did not receive compensation for his contribution.