Atrial Fibrillation and Cognitive Decline–The Role of Subclinical Cerebral Infarcts
The Atherosclerosis Risk in Communities Study
Background and Purpose—The mechanism underlying the association of atrial fibrillation (AF) with cognitive decline in stroke-free individuals is unclear. We examined the association of incident AF with cognitive decline in stroke-free individuals, stratified by subclinical cerebral infarcts (SCIs) on brain MRI scans.
Methods—We analyzed data from 935 stroke-free participants (mean age±SD, 61.5±4.3 years; 62% women; and 51% black) from 1993 to 1995 through 2004 to 2006 in the Atherosclerosis Risk in Communities Study, a biracial community-based prospective cohort study. Cognitive testing (including the digit symbol substitution and the word fluency tests) was performed in 1993 to 1995, 1996 to 1998, and 2004 to 2006 and brain MRI scans in 1993 to 1995 and 2004 to 2006.
Results—During follow-up, there were 48 incident AF events. Incident AF was associated with greater annual average rate of decline in digit symbol substitution (−0.77; 95% confidence interval, −1.55 to 0.01; P=0.054) and word fluency (−0.80; 95% confidence interval, −1.60 to −0.01; P=0.048). Among participants without SCIs on brain MRI scans, incident AF was not associated with cognitive decline. In contrast, incident AF was associated with greater annual average rate of decline in word fluency (−2.65; 95% confidence interval, −4.26 to −1.03; P=0.002) among participants with prevalent SCIs in 1993 to 1995. Among participants who developed SCIs during follow-up, incident AF was associated with a greater annual average rate of decline in digit symbol substitution (−1.51; 95% confidence interval, −3.02 to −0.01; P=0.049).
Conclusions—The association of incident AF with cognitive decline in stroke-free individuals can be explained by the presence or development of SCIs, raising the possibility of anticoagulation as a strategy to prevent cognitive decline in AF.
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, and its prevalence is increasing over time.1,2 AF is associated with an increased risk of stroke,3 heart failure,4 and death.5–7 Evidence is emerging that AF is also associated with another growing public health problem, namely, cognitive impairment or dementia. Earlier studies have shown that the evidence for the association between AF and dementia is strongest in individuals with prevalent stroke8–10; in the broader population, the evidence is less clear.11–13 A meta-analysis reported a 2.4-fold increase in the odds of dementia associated with AF when restricted to patients with stroke.14 For studies in the broader population, the association between AF and dementia is of borderline significance with substantial heterogeneity in the risk estimates.14 Later studies, however, provided evidence that AF is associated with cognitive impairment or dementia even in individuals without clinical stroke. Kalantarian et al15 reported in a meta-analysis that AF was associated with an increased risk of cognitive impairment and dementia, with or without a history of clinical stroke. In addition, recent reports from the Cardiovascular Health Study16 and Ongoing Telmisartan Alone and in Combination With Ramipril Global End Point Trial (ONTARGET) and Telmisartan Randomised Assessment Study in ACE Intolerant Subjects With Cardiovascular Disease (TRANSCEND) studies17 indicated that in the absence of clinical stroke, participants with AF experienced faster cognitive decline than those without AF. Although compelling, the mechanism underlying the association between AF and cognitive decline or dementia in the absence of clinical stroke was not elucidated in the aforementioned studies. Understanding the mechanism underlying the association between AF and cognitive decline or dementia in stroke-free individuals is critical in informing prevention strategies for cognitive decline in these individuals.
Based on a previous report that 14.7% of patients with AF had silent cerebral infarcts on cranial computerized tomographic (CT) scans,18 we hypothesized that incident AF is associated with greater cognitive decline in individuals without clinical stroke via subclinical cerebral infarcts (SCIs). To test our hypothesis, we evaluated in a stroke-free population the association between incident AF and rate of cognitive decline, stratified by the presence or absence of SCIs on brain MRI scans. We tested our hypothesis in the Atherosclerosis Risk in Communities (ARIC) Study, a large US community-based cohort study of cardiovascular disease.
The ARIC cohort is a biracial population-based sample from 4 communities in North Carolina, Mississippi, Minnesota, and Maryland, consisting of 15 792 men and women, and 45−64 years of age at baseline or visit 1 (1987−1989).19 After the visit 1 examination, there were 3 additional exams: visit 2 (1990−1992), visit 3 (1993−1995), and visit 4 (1996−1998).
Cognitive assessments were performed in the entire ARIC cohort at visit 2 (n=14 128) and visit 4 (n=11 093). Of the 6715 participants who returned for ARIC visit 3 (1993–1995) at 2 ARIC Field Centers (ie, Forsyth County, NC, and Jackson, MS), 2891 participants aged ≥55 years were invited for brain MRI scan and cognitive assessment (Figure I in the online-only Data Supplement). Usual safety exclusion criteria for MRI excluded 2% of women and 6% of men; brain MRI studies and cognitive assessments were performed in 1920 and 2053 participants, respectively, at visit 3. All participants with MRI scanning at visit 3 were invited to undergo another brain MRI scan and cognitive assessment in 2004 to 2006; 1129 and 1097 participants completed brain MRI scan and cognitive assessment, respectively, in 2004 to 2006 (Figure I in the online-only Data Supplement). We included in our analysis cohort, participants who had cognitive assessments performed at visit 3, visit 4, and 2004 to 2006 and brain MRI studies at visit 3 and 2004 to 2006 (n=1024). Therefore, visit 3 (1993−1995) was the baseline of our analysis, and the follow-up was through 2004 to 2006. After exclusions, the final analysis cohort comprised 935 participants. Figure II in the online-only Data Supplement shows the exclusion criteria and the final number of participants enrolled in the analysis cohort. The ARIC Study protocol was approved by the institutional review board of each participating center, and informed consent was obtained from each study participant.
Ascertainment of AF
AF diagnoses were obtained from ECGs at 4 study visits and hospital discharge records through December 31, 2009.20 Hospitalizations in ARIC are identified by participant or proxy report in the annual follow-up and by surveillance of local hospital discharge lists. All ECG recordings automatically coded as AF were visually rechecked by a cardiologist to confirm the diagnosis.21 A trained abstractor obtained and recorded all International Classification of Diseases, Ninth Revision hospital discharge diagnoses from each hospitalization. AF was defined as the presence of International Classification of Diseases, Ninth Revision code 427.31 or 427.32 in the discharge codes. By physician review of 125 discharge summaries with International Classification of Diseases codes indicating possible AF, we confirmed the presence of AF in ≈90%.20
Ascertainment of Ischemic Stroke
Prevalent stroke at ARIC visit 1 was defined as a self-reported history of physician-diagnosed stroke. To identify incident stroke, cohort participants were followed over time through annual telephone interviews, triennial field center examinations, surveillance of the ARIC community hospitals for all cohort members’ hospitalizations, and the review of death certificates, physician questionnaires, coroner/medical examiner reports, and informant interviews. Hospital reports were reviewed for evidence of acute stroke if the discharge diagnosis included a cerebrovascular disease code (International Classification of Diseases, Ninth Revision codes 430–438), if a cerebrovascular procedure was mentioned in the summary, or if the CT or magnetic resonance report showed evidence of cerebrovascular disease. Medical records for potential stroke events were forwarded to a single nurse abstractor at a central ARIC office who abstracted each record for number, type, and severity of neurological deficits and supporting angiographic, CT, MRI, spinal tap, or autopsy evidence. ARIC adapted National Survey of Stroke criteria for its stroke definition.22 A computerized algorithm and physician reviewer independently confirmed the diagnosis of stroke with disagreements adjudicated by a second physician reviewer. Stroke cases were further classified as definite versus probable and into further subtypes as embolic versus thrombotic stroke.23 For this study, incident ischemic stroke included definite or probable ischemic strokes (embolic or thrombotic).
Cognitive assessments, baseline performance, and longitudinal characteristics of the cognitive assessment battery have been previously reported.24,25 This battery included the delayed word recall (DWR) test, the digit symbol substitution (DSS) subtest of the Wechsler Memory Scale-Revised, and the first-letter word fluency (WF) test. Details of the tests are presented in the Methods in the online-only Data Supplement.
The ARIC Study MRI screening protocol and image analyses have been previously described.26 Briefly, 1.5-T magnetic resonance scanners (GE and Picker) were used. Axial images were angled to be parallel to the anterior commissure–posterior commissure line. The digital scan images, including SD/T2 weighted (3000, 30–100, repetition time, and echo time) and T1 weighted images, were evaluated at the MRI Reading Center on a Vortech Personal Display System (PDS-4) workstation. The image evaluation was conducted by trained and certified MRI readers. SCIs were defined as focal, nonmass lesions ≥3 mm that were bright on T2 and proton density and dark on T1 images. All scans were subjected to double-reads for infarct scoring.
Definitions of covariates are presented in the Methods in the online-only Data Supplement.
We report means with SDs or medians and interquartile ranges for continuous variables and counts with percentages for categorical variables.
To test the association between AF and cognitive decline rate, we used mixed effects linear models (PROC MIXED). The models consisted of AF status, time of follow-up (in years), and a term for the interaction of AF×time. Participants who developed incident stroke after visit 3 were excluded from the analysis. Model 1 was adjusted for age, sex, race, and interaction terms: age×time, sex×time, and race×time. Model 2 was additionally adjusted for field center, educational level (<high school, completed high school, some college), time-varying smoking status (current, not current), body mass index, hypertension, diabetes mellitus, coronary heart disease, and heart failure. The coefficient for time estimates the average annual rate of change in the cognitive test score, and the coefficient for the interaction term (AF×time) estimates the difference in average annual rate of change associated with the presence of AF.
To determine whether the association between AF and cognitive decline is mediated by SCIs, we performed 3 additional analyses. First, we repeated the analysis in participants who did not have SCIs at both visit 3 and in 2004 to 2006. Second, we evaluated the association between incident AF and cognitive decline rate in participants with prevalent SCIs at visit 3. Third, we evaluated the association between incident AF and cognitive decline rate in participants without SCIs at visit 3 and who developed SCIs by 2004 to 2006.
Finally, to evaluate further whether SCIs mediate the cognitive decline associated with AF, we conducted a secondary analysis by assessing the association of prevalent or incident SCIs with change in cognitive scores, stratified by presence or absence of incident AF.
Statistical analysis was performed using SAS version 9.2 (SAS Institute Inc, Cary, NC). All P values reported were 2 sided, and statistical significance threshold was chosen as 5%.
The analysis cohort consisted of 935 participants (mean age, 61.5 [SD, 4.3] years at baseline; 62.0% women; 50.7% black). During a median (interquartile range) follow-up of 10.6 (9.8−11.2) years, 48 participants developed AF. Compared with participants who did not develop AF, there were more men and whites among participants who developed AF. Participants who developed AF were also older and had higher prevalence of current smoking, hypertension, and coronary heart disease. Table 1 shows the baseline characteristics of study participants.
Table I in the online-only Data Supplement stratifies participants with AF by CHA2DS2−VASc (congestive heart failure, hypertension, age, diabetes, prior stroke/transient ischemic attack, vascular disease, age, sex category) score and describes the proportion of participants with AF who were anticoagulated. Of 48 participants with AF, 28 (58.3%) had CHA2DS2−VASc score ≥2, meeting definite criteria for anticoagulation. The rate of anticoagulation, however, was low even among participants with CHA2DS2−VASc score ≥2. Of 48 participants with incident AF, 20 (41.7%) had a CHADS2 score of 0, 23 (47.9%) had a score of 1, and 5 (10.4%) had a score of ≥2. Therefore, based on CHADS2 score, 89.6% of participants had CHADS2 score <2 and would not have met definite criteria for anticoagulation.
AF and Cognitive Decline
Table II in the online-only Data Supplement shows the summary cognitive test scores of study participants at visit 3, visit 4, and 2004 to 2006. Participants who developed AF had greater annual average rate of decline in DSS (−0.77; 95% confidence interval [CI], −1.55 to 0.01; P=0.054) and WF (−0.80; 95% CI, −1.60 to −0.01; P=0.048) than those who did not develop AF (Table 2). Incident AF was not associated with a greater rate of decline in DWR.
To evaluate whether or not incident AF is associated with greater cognitive decline in the absence of SCIs, we repeated the analysis restricting to participants who did not have SCIs at both visit 3 and 2004 to 2006. In this subset comprising 699 participants of whom 29 developed AF, incident AF was not associated with decline in all 3 cognitive test scores (Table 3).
Next, we evaluated the association of incident AF and cognitive decline in participants who had prevalent SCIs at visit 3. There were 67 participants in this subset of whom 3 developed AF. In this subset, compared with participants who did not develop AF, participants who developed AF had greater annual average rate of decline in WF (−2.65; 95% CI, −4.26 to −1.03; P=0.002) but not in DWR and DSS (Table 4).
Finally, we evaluated the association of incident AF and cognitive decline in participants who did not have SCIs at visit 3 but developed them by 2004 to 2006. In this subset comprising 169 individuals, participants who developed AF (n=16) had greater annual average rate of decline in DSS (−1.51; 95% CI, −3.02 to −0.01; P=0.049) than those who did not develop AF. Incident AF was not associated with decline in DWR or WF (Table 5).
Risk Factors and Impact of Subclinical Cerebral Infarcts
Table III in the online-only Data Supplement shows selected baseline characteristics (at visit 3) of participants without prevalent or incident SCIs versus those with prevalent or incident SCIs. The distribution of stroke risk factors was not significantly different between the 2 groups with the exception of hypertension and AF; the prevalence of hypertension and AF incidence were significantly higher among participants with prevalent or incident SCIs. Further, in participants with incident AF, the proportion who developed SCIs was 16 of 48 (33.3%); this proportion was almost double that in participants without incident AF, 153 of 887 (17.3%). Table IV in the online-only Data Supplement shows the number (%) of participants with SCIs on their brain MRI scans at ARIC visit 3 and 2004 to 2006. Collectively, the data suggest that AF is an important risk factor for SCIs.
To evaluate further whether SCIs mediate the cognitive decline associated with AF, we conducted a secondary analysis. In this secondary analysis, we assessed the association of prevalent or incident SCIs with change in cognitive scores, stratified by presence or absence of incident AF. We found that compared with absence of prevalent or incident SCIs, presence of prevalent or incident SCIs was significantly associated with greater decline in DSS, even in participants without AF (Tables 6 and 7). This observation suggests that it is not AF per se but rather SCIs that are related to cognitive decline.
In a large biracial population-based cohort study that comprised middle-aged individuals, we observed that participants without clinical stroke who developed AF had greater cognitive decline compared with participants who did not develop AF. However, this association was true only in participants who had prevalent SCIs or who developed SCIs during follow-up. In individuals without prevalent SCIs or who did not develop SCIs during follow-up, incident AF was not associated with cognitive decline. Further, prevalent or incident SCI was significantly associated with greater cognitive decline, even in the absence of AF. Collectively, our observations suggest that the association between incident AF and cognitive decline is mediated by the presence or development of SCIs.
A recent meta-analysis of 21 studies has provided compelling evidence for an association between AF and cognitive impairment or dementia.15 AF was significantly associated with an increased risk of cognitive impairment in patients with first-ever or recurrent stroke (relative risk, 2.70; 95% CI, 1.82−4.00) and in a broader population including patients with or without a history of stroke (relative risk, 1.40; 95% CI, 1.19−1.64). Of note, among studies that excluded patients with prevalent stroke or adjusted the risk estimates for prevalent stroke, AF was also significantly associated with an increased risk of cognitive impairment (relative risk, 1.34; 95% CI, 1.13−1.58). However, the possibility of silent stroke or SCIs as the mechanism underlying the association in patients without clinical stroke cannot be excluded.
The findings of our study are consistent with the recent meta-analysis above15 and recent reports from the Cardiovascular Health Study16 and ONTARGET and TRANSCEND studies17 and provide further evidence that incident AF is associated with dementia or cognitive decline in individuals without stroke. The Veterans Affairs Stroke Prevention in Nonrheumatic AF Investigators reported that 14.7% of patients with AF had silent cerebral infarcts on cranial CT scans.18 Further, from a recent case–control study by Gaita et al,27 silent cerebral ischemia was reported to be more prevalent in patients with AF than in controls, and cognitive performance was worse in patients with AF than in controls. However, the study did not demonstrate that the worse cognitive performance in patients with AF was mediated by silent cerebral ischemia. Based on the findings of the Veterans Affairs study and the study by Gaita et al,27 we hypothesized that incident AF may be associated with greater cognitive decline in individuals without clinically manifested stroke via SCIs. To test our hypothesis, we performed additional analyses stratified by the presence or absence of SCIs as determined by brain MRI scans. In individuals without SCIs, incident AF was not associated with cognitive decline. By way of contrast, in participants with prevalent SCIs or who developed SCIs during follow-up, incident AF was related to greater rate of cognitive decline. Thus, the association between AF and cognitive decline can be explained by the presence or development of SCIs.
The majority of SCIs in our study were likely deeply located in the deep nuclear region and deep cerebral white matter.28 Nevertheless, these lesions could still be related to AF. The Veterans Affairs Stroke Prevention in Nonrheumatic AF Investigators reported in a study of 516 patients with AF who underwent cranial CT scans that 46 of 76 (61%) patients with SCIs had lesions that were small and located in the deep hemispheric areas that are typically classified as lacunar infarcts.18 Therefore, SCIs that are related to AF are not exclusively located superficially; they can also be deeply located.
Cognitive decline associated with AF was reflected in a greater annual average change in DSS and WF but not DWR. The pattern of cognitive decline is worth noting. In general, DWR is a test of recent memory, DSS is a test of attention, and WF is a test of executive function. In participants with prevalent SCIs, incident AF was associated with a greater decline in executive function (WF), whereas in participants who developed SCIs during follow-up, incident AF was associated with a greater decline in attention (DSS). The ARIC study had previously reported that during a 14-year follow-up (visit 2 through 2004−2006), diabetes mellitus was associated with a decline in DSS and WF, hypertension was associated with a decline in WF, and systolic blood pressure was related to a decline in DSS.29 The APOE ε4 genotype−most strongly associated with the risk of Alzheimer disease−was associated with declining performance in DWR.29 These patterns are consistent with a model of pathoanatomic relationships where lacunar infarcts30 and vascular risk factors31,32 are associated with attentional and executive function deficits, whereas Alzheimer disease–related risk is associated with decline in memory functioning.33,34 These data suggest that although SCI may be the principal mechanism that mediates cognitive decline associated with AF, the possible contribution of other shared risk factors (eg, hypertension) cannot be entirely excluded.
The strengths of this study include the long follow-up, inclusion of black participants, and extensive measurement of covariates. However, several limitations should be noted. First, incident AF was identified mostly from hospitalization discharges, and we could have missed asymptomatic AF or AF managed exclusively in an outpatient setting. However, we and others have previously shown that the validity of AF ascertainment using hospitalizations is acceptable,20,35 that incidence rates of AF in ARIC are consistent with other population-based studies,20,35 and that the associations between genetic variants in the chromosome 4q25 locus and AF−extremely specific for AF risk−in ARIC are similar to other studies with a more rigorous ascertainment of AF.36 Second, we defined SCIs as focal, nonmass lesions ≥3 mm that were bright on T2 and proton density and dark on T1 images. Therefore, participants with SCIs <3 mm might have been misclassified. Third, ARIC participants who failed to complete the 11 years of observation (ARIC visit 3 through 2004–2006) because of death, an inability to undergo MRI, or refusal had greater burdens of disease (eg, diabetes mellitus, hypertension), lower baseline cognitive test scores, and worse baseline MRI findings than those who completed follow-up. Therefore, participants who completed the longitudinal assessments described here comprised the healthier subset of the original cohort. Because of the loss to follow-up of participants with more disease burden and cognitive impairment at baseline, we may have underestimated the degree of cognitive decline during the 11 years of observation. However, we do not expect this loss to follow-up to introduce a substantial bias on the assessment of the association between AF and cognitive decline because in all our analyses, we have adjusted for determinants of participation in the follow-up exams (eg, diabetes mellitus, hypertension, coronary heart disease, and others). Fourth, because we are unable to classify AF type (paroxysmal, persistent, or permanent AF) accurately in the ARIC study, we did not assess the relationship between AF type and cognitive decline or SCIs. Finally, information on warfarin use from the study visits and annual follow-up phone calls may not be complete.
Our findings have important clinical and public health implications. Because the majority of our participants with AF do not meet criteria for anticoagulation (based on CHADS2) to prevent stroke, our findings raise 2 questions: first, should patients with AF be anticoagulated to prevent cognitive decline, even if they don’t meet criteria for anticoagulation to prevent stroke? Second, should persons with SCIs, irrespective of AF, be prescribed antiplatelet agents or anticoagulants to prevent cognitive decline? In addition, because more participants with AF met definite criteria for anticoagulation based on CHA2DS2−VASc (58.3% of participants with AF had CHA2DS2−VASc score ≥2) than CHADS2 (10.4% of participants with AF had a CHADS2 score of ≥2), the CHA2DS2−VASc score may be a more appropriate tool for ascertaining need for anticoagulation in AF. These questions will need to be addressed in other prospective studies and clinical trials.
In conclusion, in a large biracial population based-cohort study of middle-aged individuals without stroke, incident AF is independently associated with greater cognitive decline. This association can be explained by the presence or development of SCIs. Presently, anticoagulation is indicated in most patients with AF to prevent thromboembolic stroke.37 In patients with AF without risk factors for stroke (and hence without current indication for anticoagulation), further research is needed to evaluate whether anticoagulation would be beneficial in preventing cognitive decline.
Sources of Funding
This research was conducted while Dr Chen was a recipient of a Scientist Development Grant (10SDG3420031) from the American Heart Association and an R21AG042660-01A1 from the National Institute on Aging. Dr Alonso received support from the National Heart, Lung, and Blood Institute (RC1 HL099452) and the American Heart Association (09SDG2280087). The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN268201100010C, HHSN268201100011C, and HHSN268201100012C).
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.114.005243/-/DC1.
- Received February 20, 2014.
- Revision received June 18, 2014.
- Accepted July 1, 2014.
- © 2014 American Heart Association, Inc.
- Go AS,
- Hylek EM,
- Phillips KA,
- Chang Y,
- Henault LE,
- Selby JV,
- et al
- Miyasaka Y,
- Barnes ME,
- Gersh BJ,
- Cha SS,
- Bailey KR,
- Abhayaratna WP,
- et al
- Wolf PA,
- Abbott RD,
- Kannel WB
- Wang TJ,
- Larson MG,
- Levy D,
- Vasan RS,
- Leip EP,
- Wolf PA,
- et al
- Benjamin EJ,
- Wolf PA,
- D’Agostino RB,
- Silbershatz H,
- Kannel WB,
- Levy D
- Barba R,
- Martínez-Espinosa S,
- Rodríguez-García E,
- Pondal M,
- Vivancos J,
- Del Ser T
- Inzitari D,
- Di Carlo A,
- Pracucci G,
- Lamassa M,
- Vanni P,
- Romanelli M,
- et al
- Marzona I,
- O’Donnell M,
- Teo K,
- Gao P,
- Anderson C,
- Bosch J,
- et al
- Ezekowitz MD,
- James KE,
- Nazarian SM,
- Davenport J,
- Broderick JP,
- Gupta SR,
- et al
- 19.↵The Atherosclerosis Risk in Communities (ARIC) study: design and objectives. The ARIC Investigators. Am J Epidemiol. 1989;129:687–702.
- Soliman EZ,
- Prineas RJ,
- Case LD,
- Zhang ZM,
- Goff DC Jr..
- Rosamond WD,
- Folsom AR,
- Chambless LE,
- Wang CH,
- McGovern PG,
- Howard G,
- et al
- 26.↵Magnetic resonance imaging version 1.0. Aric Manual of Operations. Chapel Hill, NC: University of North Carolina; 1993.
- Bryan RN,
- Cai J,
- Burke G,
- Hutchinson RG,
- Liao D,
- Toole JF,
- et al
- Psaty BM,
- Manolio TA,
- Kuller LH,
- Kronmal RA,
- Cushman M,
- Fried LP,
- et al