Does high blood pressure increase the risk of getting Alzheimer’s disease? Although several studies have highlighted hypertension as a potential AD risk factor, getting prominent play in the press, the epidemiologic evidence to date remains surprisingly weak. This is the conclusion of researchers led by Deborah Blacker at Massachusetts General Hospital, Boston, in a paper published online today in the journal Epidemiology. While their meta-analysis yielded inconclusive data, the authors point out that this may be due to the small number of studies that met inclusion criteria, as well as bias and other problems with the original data. They recommend further research in this area to make the data stronger and more comparable. The results also debut today in AlzRisk, the open online database hosted by Alzforum that curates and analyzes non-genetic risk factors for AD. The AlzRisk analysis is very similar but not identical to the Epidemiology paper. Hypertension is the fifth factor to be entered into AlzRisk, and marks a major expansion of the database. AlzRisk’s curators, led by Blacker and Jennifer Weuve at Rush University, Chicago, Illinois, plan to complete additional factors before the end of the year, and expect that scientists, journalists, and others interested in AD will find the database a useful resource.

Hypertension is of interest because it is treatable, many epidemiological studies have linked cardiovascular disease to dementia, and two previous reviews have found an association between high blood pressure and dementia (see Qiu et al., 2005; Kennelly et al., 2009). However, much of this link may be due not to blood pressure per se but to cerebrovascular pathology, which often coexists with AD and may contribute to cognitive impairment. First author Melinda Power and colleagues wanted to understand risk factors specifically for AD. To this end, they included only studies that reported results for AD dementia, and required that AD be diagnosed through clinical examination. This meant that a significant fraction of the literature on blood pressure and dementia had to be left out. Power and colleagues used stringent criteria for selecting the 18 studies they included in their meta-analysis. For example, data had to be from prospective cohort studies or nested case-control studies, include blood pressure as an exposure variable, provide confidence intervals, and adjust for age and sex. Several well-known stalwarts, such as the Atherosclerosis in Communities Study and the Cardiovascular Health Study, were excluded because they did not meet all the pre-specified inclusion criteria for the AlzRisk systematic reviews. Some researchers interviewed by ARF thought these omissions substantially weakened the review.

Counterintuitively, the meta-analysis showed that high blood pressure in late life was linked to a slightly lower risk of AD. This may be due to selection bias and other problems with the original data, the authors note. For example, people with either uncontrolled hypertension and/or AD are more likely to die or be too ill to participate in a study. People with both disorders may, therefore, drop out. Alternatively, the Alzheimer’s disease process itself might lower blood pressure, muddying the results. On the other hand, high blood pressure in middle age did seem to link to a higher risk of developing AD in late life. This finding was based on only four studies, however, and the data were not always consistent. The results highlight the need for further research in middle-aged populations, Power told ARF. An advantage of focusing on middle-aged populations is that they are likely to be free of the selection bias and disease effects that may confound results in the older groups. “I think it’s going to be surprising to the Alzheimer’s community that the evidence is as weak as it is,” Power noted.

The authors also recommend making improvements in data collection. Currently, different studies report blood pressure in varied ways, making it difficult to combine the data. The authors suggest reporting estimates of risk per 10-mm Hg increments in systolic and diastolic blood pressure, as well as using standard categories of blood pressure recommended by the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure (see Chobanian et al., 2003). This would create standardized measures across studies, avoiding apples-to-oranges comparisons and allowing data to be more easily combined. Also, most studies do not show whether hypertension is being treated. This data should be collected and analyzed, Power told ARF. Controlling hypertension may alter AD risk, and this knowledge would have public health implications.

These types of inconsistencies are common in epidemiological research, creating difficulties for researchers who want to analyze studies and spot overall trends. The goal of the AlzRisk database is to make this job easier. AlzRisk aims to combine all available data on a given non-genetic risk factor into one searchable database and provide meta-analyses where feasible. By pulling out the same data elements from each study on a given risk factor, the reader can compare apples and apples in a way that is not feasible with, say, a PubMed search (see hypertension table). AlzRisk complements the AlzGene database, which collates all genetic risk factors for AD. Weuve manages the AlzRisk project and supervises its team of curators, who are mostly doctoral students and postdoctoral fellows at the Harvard School of Public Health.

AlzRisk fills a similar need to AlzGene, Weuve said, but pulling together epidemiologic studies presents more of a challenge than summarizing genetic data. “We are concerned about sources of bias, and we feel that those things require discussion. We are essentially writing a publishable paper with each risk factor,” Weuve noted. The strength of the database is its depth of analysis and documentation, Weuve told ARF. The other major advantage is that the database will be regularly updated to keep up with new findings in the literature. AlzRisk brings together and distills study findings on a particular risk factor to estimate the strength of its association with AD. Weuve suggested these summaries (see Current Understanding and Discussion) could be useful to scientists who are writing grants and need to cite research, as well as to epidemiologists who are designing new studies. Blacker, who is the principal investigator of the AlzRisk project, added, “We are trying to write the commentary at the Scientific American level, so that it will be useful to journalists, too.”

The database currently includes four factors that may modify AD risk: diabetes, physical activity, nutritional antioxidants, and inflammatory biomarkers. In addition to hypertension, the team expects to add several more factors this year and next, such as hormone replacement therapy, obesity, and head injury. “We select risk factors based on the volume of available study results, and also the clinical importance,” Weuve told ARF. Studies must meet strict criteria to be included. The database emphasizes large-scale prospective cohort studies, excluding cross-sectional research that simply compares characteristics of groups of people who have AD to those of groups who do not. “That design has so many problems that we eliminated it from consideration,” Weuve said. This limits the number of studies that can be considered.

As might be expected with such a complex topic, scientists in the field do not agree on the best approach for selecting studies. Benjamin Wolozin at Boston University, Massachusetts, would like AlzRisk to include more large, population-based datasets, such as those from Finland, Belgium, and Canada, as well as the Kaiser Permanente and Veterans Affairs databases in the U.S., because these datasets provide great statistical power. Edo Richard at the University of Amsterdam, The Netherlands, wrote to ARF that vascular dementias should be included in the hypertension analysis. “The vast majority of patients suffer from mixed dementia. Excluding patients with a vascular component to the dementia probably contributed to the disappointing results of this meta-analysis,” he said (see full comment below). However, scientists contacted for this article agreed on the need for a centralized compendium of epidemiologic studies on AD. “In general, the approach of doing meta-analyses has been incredibly important,” Wolozin said. “As AlzRisk generates more data, I think it will be a very valuable resource.”—Madolyn Bowman Rogers


  1. Power et al. have done a thorough, systematic review and meta-analysis, according to the highest standards, on the association between hypertension and incident Alzheimer’s disease (AD). The results, however, are disappointing, since a clear association between hypertension and incident AD was not found. In the well-written discussion they address all potential sources of bias that could have contributed to these results.

    They briefly address the possibility of misclassification of AD if, for example, hypertension leads to a greater likelihood of diagnosing vascular dementia or mixed dementia. Before concluding that there is no association between hypertension and dementia, the more fundamental problem of classification of old-age dementia, which is crucial for the interpretation of these epidemiological data, should be discussed. The distinction between AD and vascular dementia in old-age dementia is somewhat artificial, and the vast majority of patients suffer from mixed dementia. Excluding patients with a vascular component to the dementia probably contributed to the disappointing results of this meta-analysis, whereas from a pathophysiological point of view, these patients should have been included. In spite of the hypotheses underlying a potential link between hypertension and AD pathology (plaques and tangles), the association of hypertension with cerebrovascular disease and resulting cognitive decline is currently stronger.

    Several randomized, controlled trials of hypertension treatment using the more pragmatic outcomes of cognitive decline or incident dementia instead of AD suggest a potential beneficial effect of hypertension treatment on cognitive decline, even in the very elderly over 80 years of age (Forette et al., 2002; Peters et al., 2008; reviewed in Ligthart et al., 2010). These results contrast with the current findings from epidemiological studies.

    It would be of great clinical relevance if the current approach would be repeated with the more pragmatic outcome "dementia" instead of the specific nosological entity of AD. The new AlzRisk database could potentially be a good forum to mediate in this process. The results of Power et al. should be interpreted with the greatest caution, as the authors stress in their discussion. More epidemiological research, but certainly also more randomized, controlled trials of hypertension treatment to prevent cognitive decline and dementia in various populations, and with the more pragmatic outcome "incident dementia," are warranted and underway (Richard et al., 2009).


    . The prevention of dementia with antihypertensive treatment: new evidence from the Systolic Hypertension in Europe (Syst-Eur) study. Arch Intern Med. 2002 Oct 14;162(18):2046-52. PubMed.

    . Incident dementia and blood pressure lowering in the Hypertension in the Very Elderly Trial cognitive function assessment (HYVET-COG): a double-blind, placebo controlled trial. Lancet Neurol. 2008 Aug;7(8):683-9. PubMed.

    . Treatment of cardiovascular risk factors to prevent cognitive decline and dementia: a systematic review. Vasc Health Risk Manag. 2010;6:775-85. PubMed.

    . Prevention of dementia by intensive vascular care (PreDIVA): a cluster-randomized trial in progress. Alzheimer Dis Assoc Disord. 2009 Jul-Sep;23(3):198-204. PubMed.

  2. Power et al. have conducted a systematic review of studies of the associations of hypertension and incident Alzheimer's disease. Their bottom line is that their analysis does not offer either convincing support or refutation of the association. I applaud them for the huge amount of work that went into the review that itself will be of great aid to future researchers. My view is definite: that hypertension might be related to AD pathology or maybe not. The problem is that hypertension is a ubiquitous diagnosis these days. In our Mayo Clinic Study of Aging, a majority of subjects carry the diagnosis in their medical record. Undoubtedly, the chronicity and the severity make a big difference, but those features are very hard to quantify. Furthermore, there is convincing evidence that risks for AD due to hypertension are higher in middle age, or decades before dementia onset, and then slowly attenuate to the point that higher blood pressure actually looks protective in the older age group. This occurs for at least two reasons. First, orthostatic hypotension becomes more common with advancing age and is associated with a poor prognosis and with underlying degenerative disease of the brain. Second, cognitive impairment itself is associated with weight loss and other factors that might also cause blood pressure to drop as a person becomes symptomatic. Therefore, the timing of the baseline measurement of blood pressure relative to the onset of dementia is critical. It is certainly likely that many studies simply do not have the extended time of observation needed to demonstrate an association between hypertension and AD dementia.

    View all comments by David Knopman
  3. Reply to comments by Edo Richard, David Knopman, and Ben Wolozin
    We very much appreciate the careful attention to and appraisal of our systematic review and meta-analysis of prospective epidemiologic studies of blood pressure and risk of Alzheimer's disease (AD). Drs. Wolozin, Richard, and others raised concerns about the selection of studies for our meta-analysis and the appropriateness of evaluating AD rather than total dementia or cognitive decline as an outcome.

    Dr. Wolozin and others maintained that the AlzRisk database should include data from additional cohorts. We’d like to clarify that we pre-specified the inclusion criteria for AlzRisk for all risk factors on the site, specifically, 1) the risk factor is studied in prospective relation to AD; 2) the report contains sufficient information to determine what covariates were included in analyses; 3) the analysis adjusts for at least age, gender, and education; 4) the report contains the effect estimates' 95 percent confidence intervals or corresponding P values, or sufficient information for us to compute them; and 5) the report includes results specific to AD (not just dementia). In addition, in order to perform meta-analysis, we require that there are at least four available studies where the exposure is defined sufficiently similarly to conduct a meta-analysis. If qualifying reports on blood pressure and AD from these sources appear in the future, we will most definitely add them to our database and update our meta-analyses.

    Dr. Richard expressed the concern that limiting the meta-analysis to studies of AD may have excluded many of the cases with mixed or vascular dementias that could have been attributed in part to elevated blood pressure. Further, Dr. Richard proposed a review of “more pragmatic” outcomes such as total dementia and cognitive decline. While we agree that such an analysis would have been interesting, our goal was to evaluate blood pressure in relation to AD. This question was of interest because a relation between the two has been assumed, even though evidence linking blood pressure to AD pathology (e.g., amyloid-β and phosphorylated tau) is much weaker than evidence linking blood pressure to cerebrovascular pathology. Our analyses show that data from large-scale prospective epidemiologic studies of blood pressure and AD (in 20 distinct study populations) do not provide convincing evidence to support the hypothesis that higher blood pressure, at any point in the lifespan, increases AD risk. However, as we write in our accompanying paper in Epidemiology, preventing or treating hypertension can still benefit the public health, even where dementia is concerned. For example, systematic autopsy studies have shown that the presence of cerebrovascular pathology makes clinical dementia more likely for a given level of Alzheimer's pathology (Schneider et al., 2004). Thus, hypertension may substantially increase the burden of total dementia.

    Intertwined AD and vascular pathologies in the CNS clearly frustrate the ability of epidemiologic studies to tie blood pressure to AD risk. In AlzRisk, when a paper includes results on both AD and total dementia, we record both sets of results on the site. Of the 20 populations noted above, 13 examined total dementia in addition to AD. Studies of two cohorts specifically defined their “total dementia” as AD and vascular dementia, and even without this definition, total dementia includes primarily these two diagnoses and mixed cases. Surprisingly, associations of measures of blood pressure with total dementia were not substantially stronger than corresponding associations with AD. Notably, the studies of total dementia are vulnerable to many of the same limitations found in studies of AD, including reverse causation, selection bias, misclassification of blood pressure, and misspecification of the shape of the association between blood pressure and dementia.

    We agree that cognitive decline has several advantages as an outcome. For example, studies of cognitive decline in persons who are unlikely to have clinical dementia may be less prone to bias from reverse causation and differential selection. And, as Dr. Richard points out, cognitive decline may be especially practical in the context of clinical trials of blood pressure treatment. However, it seems unlikely that such studies could address the etiologic question about blood pressure and AD pathology. AlzRisk does not include studies of cognitive decline, but, in the detailed Discussion for each risk factor, we do consider findings from these studies when we interpret the epidemiologic results on AD.

    Finally, we appreciate the insights of Dr. Knopman on the likely importance of timing. Given the blood pressure-lowering effects of frailty and cognitive decline, reverse causation could explain, in part, the “null” or protective findings in studies of the oldest participants. Studies of elevated blood pressure in midlife are less likely to suffer from bias due to reverse causation, but as of yet, these studies are frustratingly few.


    . Cerebral infarctions and the likelihood of dementia from Alzheimer disease pathology. Neurology. 2004 Apr 13;62(7):1148-55. PubMed.

    View all comments by Melinda Power

Make a Comment

To make a comment you must login or register.


Paper Citations

  1. . The age-dependent relation of blood pressure to cognitive function and dementia. Lancet Neurol. 2005 Aug;4(8):487-99. PubMed.
  2. . Blood pressure and the risk for dementia: a double edged sword. Ageing Res Rev. 2009 Apr;8(2):61-70. PubMed.
  3. . The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA. 2003 May 21;289(19):2560-72. PubMed.

External Citations

  1. AlzRisk
  2. hypertension table
  3. AlzGene
  4. Current Understanding and Discussion

Further Reading

No Available Further Reading

Primary Papers

  1. . The association between blood pressure and incident Alzheimer disease: a systematic review and meta-analysis. Epidemiology. 2011 Sep;22(5):646-59. PubMed.