Statin Use and the Risk of Incident Dementia: Title and subTitle BreakThe Cardiovascular Health Study FREE

Dementia is associated with excess morbidity and mortality.¹ The most common causes of dementia are Alzheimer disease (AD) and cerebrovascular changes that produce vascular dementia (VaD). Although each type may evolve from distinct pathophysiologies, evidence suggests that the 2 may exist along a spectrum.² For both causes, many demographic, socioeconomic, genetic, and clinical risk factors have been identified, perhaps reflecting different pathological processes.^{2 - 3} In particular, some evidence supports a role of lipid metabolism or inflammation in the development of dementia, specifically AD.⁴

The 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors, or statins, prevent cardiovascular disease in at-risk populations. These agents may reduce cardiovascular risk by inhibiting cholesterol synthesis or through anti-inflammatory effects.⁵ Statin use could therefore affect the development of dementia. In vivo, in vitro, and epidemiological investigations have reported a possible protective association between statin use and the dementia process, specifically AD.^{6 - 12} In contrast, results from trials designed to assess the cardiovascular effects of statin use, in which a surrogate measure was used to assess dementia, have not shown a protective statin association.¹³ Using data from the Cardiovascular Health Cognition Study, an ancillary study of the Cardiovascular Health Study, we assessed the risk of incident all-cause and type-specific dementia associated with statin therapy. We hypothesized that statin therapy would be associated with a lower risk of dementia compared with never use of lipid-lowering therapy.

The Cardiovascular Health Study¹⁴ is a prospective population-based cohort study of risk factors for cardiovascular disease among community-dwelling older adults. The Cardiovascular Health Cognition Study included the subset of Cardiovascular Health Study participants 65 years and older who underwent magnetic resonance imaging from 1991 to 1994 and a concurrent Mini-Mental State Examination. The present analyses include 2798 participants who were free of dementia at the time of the baseline magnetic resonance imaging. The study was approved by appropriate institutional review boards.

At baseline, the Cardiovascular Health Study collected data on demographics, clinical characteristics, medication use, diagnostic testing, laboratory assessments, and clinical outcomes. The clinical assessment included functional health status, cognitive function, smoking status, alcohol intake, blood pressure, diabetes mellitus status, and clinical and subclinical measures of cardiovascular disease. The laboratory assessment included baseline fasting lipid profile, APOE ε4 genotype, and C-reactive protein level. To assess medication use, participants brought all prescription medications used within the prior 2 weeks to each annual study visit.¹⁵ Participants were followed up annually with regard to functional health status, symptoms, physical examination, clinical characteristics, and medication use.

The criteria and process used to identify dementia in the Cardiovascular Health Cognition Study¹⁶ have been detailed previously. The main outcomes of the present analysis were incident all-cause dementia and dementia due solely to AD as defined by the criteria of the National Institute of Neurological and Communicative Diseases and Stroke.¹⁷ The State of California Alzheimer’s Disease Diagnostic and Treatment Centers criteria¹⁸ were used to identify VaD.

The main exposure was the use of lipid-lowering agents (LLAs), specifically statins, as assessed by the annual medication inventory. Lipid-lowering agents were grouped according to statin and nonstatin classifications. Statins were further classified as more lipophilic (lovastatin, simvastatin, and cervistatin) or less lipophilic (atorvastatin calcium, pravastatin sodium, and fluvastatin sodium).¹⁹ Nonstatin LLAs included fibrates, niacin, bile acid sequestrants, and probucol, with fibrate use constituting three quarters of nonstatin use. The baseline cognitive assessment was performed using the Modified Mini-Mental State Examination.²⁰

For the main analyses, the use of LLAs was modeled as a time-dependent covariate and categorized as never use, statin use, or nonstatin LLA use. The use of statins and nonstatin LLAs was modeled as ever use, current use, or former use. Time-dependent Cox proportional hazards regression analysis was used to compute multivariable-adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) for the association of LLA use with dementia, using never users as the referent group.

For the primary analyses, follow-up started at the date of the baseline magnetic resonance imaging and extended to 1 year before the date of dementia diagnosis or censoring. The main models a priori incorporated this 1-year lag because dementia typically manifests in a gradual manner. Therefore, from a biological perspective, exposures that provide a protective effect would typically be present months or years before clinical dementia. A secondary set of models included the final year of exposure status and modeled LLA use as current use vs nonuse because this model best approximates the case-control approach.^{9 - 12}

Additional analyses were restricted to persons with a diagnosis of coronary heart disease or a baseline total cholesterol level of 200 mg/dL or higher (≥5.18 mmol/L), which are indications for LLA therapy. Differences between subgroups defined by sex, age (≤75 vs >75 years), race, smoking status, hypertension, diabetes mellitus, clinical cerebrovascular disease, APOE ε4 genotype, or quartiles of C-reactive protein levels were assessed with the addition of a cross-product term between the covariate of interest and the categorical LLA variable. Analyses were done using Stata version 7.0 (StataCorp LP, College Station, Tex).

On average, person-years of LLA use were associated with alcohol abstinence, a higher score on the Mini-Mental State Examination, diabetes mellitus, clinical heart disease, and elevated total cholesterol and C-reactive protein levels (Table 1). In addition, person-years of statin use were associated with female sex and the presence of an APOE ε4 allele.

During 15 030 person-years of follow-up, there were 480 incident cases of dementia, including 245 attributable to AD alone, 151 attributable to a combination of AD and VaD, 62 attributable to VaD alone, and 22 attributable to other causes. In the primary analyses with the 1-year lag in exposure classification (Table 2), ever use of statins was not associated with the risk of all-cause dementia (HR, 1.08; 95% CI, 0.77-1.52), AD alone (HR, 1.21; 95% CI, 0.76-1.91), mixed AD and VaD (HR, 0.87; 95% CI, 0.44-1.72), or VaD alone (HR, 1.36; 95% CI, 0.61-3.06) compared with never users. In analyses that recoded ever use of statin therapy into current or former use, current use was not associated with the risk of dementia. However, former use of statins was associated with an elevated risk of all-cause dementia (HR, 1.88; 95% CI, 1.05-3.36) and AD alone (HR, 2.54; 95% CI, 1.24-5.20) compared with never users. Ever use of nonstatin LLAs was not associated with the risk of all-cause dementia (HR, 0.86; 95% CI, 0.50-1.46) or AD alone (HR, 1.05; 95% CI, 0.51-2.13) compared with never use of LLAs. Further adjustment for race, smoking status, hypertension, diabetes mellitus, atrial fibrillation, low-density lipoprotein and total cholesterol levels, APOE ε4 genotype, C-reactive protein level, carotid intima-media thickness, aspirin use, benzodiazepine use, or total number of medications used did not appreciably change the estimates.

No dose response was detected with regard to risk and duration of statin use. Compared with never use of LLAs, the HRs of all-cause dementia were 0.98 (95% CI, 0.55-1.74) for ever use of statins for less than 1 year, 1.41 (95% CI, 0.89-2.25) for 1 to 3 years, and 0.74 (95% CI, 0.35-1.57) for greater than 3 years. For AD alone, the HRs were 1.52 (95% CI, 0.78-2.98) for ever use of statins for less than 1 year, 1.05 (95% CI, 0.49-2.24) for 1 to 3 years, and 1.04 (95% CI, 0.42-2.56) for greater than 3 years. Similarly, no difference was detected between use of statins that were more or less lipophilic. Compared with never use of LLAs, ever use of more lipophilic statins was associated with HRs of 0.94 (95% CI, 0.61-1.44) for all-cause dementia and 1.03 (95% CI, 0.57-1.86) for AD alone, and ever use of less lipophilic statins was associated with HRs of 1.38 (95% CI, 0.83-2.29) for all-cause dementia and 1.58 (95% CI, 0.80-3.11) for AD alone. When analyses were restricted to subjects with clinical coronary heart disease or a total cholesterol level of 200 mg/dL or higher (≥5.18 mmol/L), the HRs associated with ever use of statins were 0.97 (95% CI, 0.67-1.42) for all-cause dementia and 1.01 (95% CI, 0.61-1.69) for AD alone compared with never use. Similarly, no association differences were detected across the specific subgroups detailed in the “Methods” section.

In secondary analyses that included the final year of exposure, current use of statins compared with nonuse of LLAs was associated with HRs of 0.69 (95% CI, 0.46-1.02) for all-cause dementia and 0.59 (95% CI, 0.32-1.10) for AD alone. The HR for the association between current statin use and any AD, combining the AD alone and mixed AD and VaD events (n = 396), was 0.56 (95% CI, 0.35-0.92).

Given the prevalence of dementia and its associated morbidity and mortality, an effective preventive pharmacological treatment would have important public health implications. In this prospective cohort of older adults, antecedent statin use was not associated with a lower risk of all-cause dementia or AD compared with never use of LLAs after considering other known or suspected risk factors. In contrast, the results suggested a protective statin association in models that simulated the typical case-control design of a cross-sectional statin exposure and dementia outcome.

Several factors may explain why statin use was not associated with a lower risk of dementia. Participants were on average 75 years of age, and statin use was assessed for a median of 5 years. Statin exposure may need to occur earlier in adulthood or for longer periods to prevent dementia, although analyses that stratified the duration of statin use did not suggest a duration-dependent association. The potential benefit may depend on the particular type of statin used or the characteristics of the patient, although analyses that assessed the association by statin lipophilicity or by clinical subgroup did not detect differences.

Alternatively, statin use may not affect the development of dementia. Case-control studies^{9 - 12} have reported a protective association typically based on cross-sectional statin exposures and dementia outcomes and have not considered former use or incorporated a lag period into the analysis. A protective statin association was observed in the present investigation when analyses modeled the statin exposure as current use vs nonuse and did not incorporate a lag period, thereby simulating a case-control approach. In contrast, no association was evident in the primary analyses. Indeed, former use of statins was associated with an elevated risk of dementia. Although starting and stopping statin therapy may somehow trigger the dementia process, the most likely explanation is that former statin use (discontinuation of statin therapy) is a surrogate marker for declining health. Patients and physicians may be less inclined to maintain preventive treatments such as statin use as overall health deteriorates.²¹ Conversely, current statin use may be a surrogate marker for good health, which may be associated with a lower risk of dementia.

This study has limitations. Despite efforts to minimize confounding, we cannot exclude the possibility of uncontrolled confounding in this prospective cohort study. For example, the primary indication for statin therapy was elevated cholesterol in the setting of coronary heart disease or its risk factors. Some authors have reported that heart disease is a risk factor for dementia.³ However, analyses that adjusted for heart disease or restricted the assessment to persons with a clinical indication for statin therapy did not reveal an association. The primary analyses had greater than 80% power to detect a 50% decrease in risk of all-cause dementia or AD (a risk difference comparable to previous reports from epidemiological studies^{9 - 12} ); however, the study had modest power to detect associations among subgroups of interest.

In this investigation, statin therapy was not associated with a lower risk of dementia. Although statin use is an important treatment for cardiovascular disease, additional investigation is needed to determine whether and for whom statin use may affect dementia risk.

Correspondence: Thomas D. Rea, MD, MPH, Department of Medicine, University of Washington, Metropolitan Park, East Tower, 1730 Minor Ave, Suite 1360, Seattle, WA 98101 (rea123@u.washington.edu).

Accepted for Publication: March 1, 2005.

Author Contributions:Study concept and design: Rea, Breitner, Psaty, Hazzard, Burke, Lyketsos. Acquisition of data: Psaty, Lopez, Burke, Lyketsos, Kuller. Analysis and interpretation of data: Rea, Breitner, Psaty, Fitzpatrick, Newman, Zandi, Burke, Lyketsos, Bernick, Kuller. Drafting of the manuscript: Rea, Breitner, Kuller. Critical revision of the manuscript for important intellectual content: Breitner, Psaty, Fitzpatrick, Lopez, Newman, Hazzard, Zandi, Burke, Lyketsos, Bernick, Kuller. Statistical analysis: Rea, Fitzpatrick, Zandi. Obtained funding: Psaty, Lopez, Burke, Kuller. Administrative, technical, and material support: Lopez, Kuller. Study supervision: Breitner, Psaty. Geriatric expertise: Hazzard.

Funding/Support: This study was supported by grant R01 625566 and by contracts N01 HC85079, N01 HC85080, N01 HC85081, N01 HC85082, N01 HC85083, N01 HC85084, N01 HC85085, N01 HC85086, N01 HL35129, and N01 HL15103 from the National Heart, Lung, and Blood Institute, Bethesda, Md, and by grants R01 AG09556 and R01 AG15928 from the National Institute on Aging, Bethesda.