Association of Atrial Fibrillation with White Matter Disease: The Atherosclerosis Risk in Communities (ARIC) Study

Abstract

Background and Purpose:

Evidence suggests that atrial fibrillation (AF) is associated with increased risk of cognitive decline and dementia, even in the absence of stroke. White matter disease (WMD) is a potential mechanism linking AF to cognitive impairment. In this study, we explored the association between prevalent AF and WMD.

Methods:

We performed a cross-sectional analysis of participants attending the ARIC-Neurocognitive Study in 2011–2013 who underwent brain magnetic resonance imaging. AF was ascertained from study visit electrocardiograms or prior hospitalization codes. Extent of WMD was defined by measures of white matter microstructural integrity and white matter hyperintensity (WMH) volume. Multivariable linear regression models were used to assess the association between AF and WMD.

Results:

Among 1899 participants (mean age 76 years, 28% black, 60% female), 133 (7%) had prevalent AF. After multivariable adjustment, differences between participants with and without AF were −0.001 (95% confidence interval (CI) −0.006, 0.004) for global white matter fractional anisotropy, 0.031*10⁻⁴ mm²/s (95%CI −0.075, 0.137) for global white matter mean diffusivity, and 0.08 mm³ (95%CI −0.14, 0.30) for WMH volume.

Conclusions:

The results suggest that there is no association between prevalent AF and WMD.

Introduction

Increasing evidence suggests a direct association between atrial fibrillation (AF) and cognitive decline.¹ Cerebral small vessel disease is one of the pathologic mechanisms by which AF could lead to cognitive impairment. White matter (WM) loss, WM hyperintensities (WMH), and impaired WM microstructural integrity are all MRI-based markers of WM disease (WMD), a manifestation of small vessel disease, and are predictive of cognitive function among older adults.^{2, 3} Adverse effects of AF on cerebral small vessel disease may be one mechanism by which AF contributes to cognitive impairment.

Previous studies suggested an association between AF and WMH volume.³ However, none of them investigated whether AF is associated with microstructural changes to WM, an early marker of WMD. Two novel MRI measures—fractional anisotropy (FA) and mean diffusivity (MD)—are often used jointly to capture white matter microstructural changes with high sensitivity.⁴ Increased WM MD and decreased WM FA are observed in patients with damaged WM tissue and adverse cognitive outcomes.^5,6 Thus, we evaluated the cross-sectional association of AF with measures of white matter microstructural integrity and WMH volume in a subset of participants from the Atherosclerosis Risk in Communities–Neurocognitive Study (ARIC-NCS).

Methods

The data that support the findings of this study are available from the ARIC Coordinating Center at the University of North Carolina, Chapel Hill. As part of an ancillary study, 1951 participants from the ARIC-NCS Visit 5 (2011–2013) underwent brain MRI, including Diffusion Tensor Imaging (DTI). Details on the ARIC study are available in the Online-only Supplement. Participants with incomplete AF data (n= 38), non-white or non-black persons and non-white persons from the Minneapolis and Washington County sites (n= 14) were excluded. The study was approved by institutional review boards at the participating institutions and all participants provided written consent.

White matter microstructural integrity measures were obtained from axial DTI imaging sequence (Online-only Supplement). Six regions of interest (ROIs) were selected for analysis: frontal, temporal, parietal, occipital lobes as well as anterior and posterior corpus callosum. We calculated region-specific WM FA and WM MD as the weighted average of left and right voxel measures for each ROI. Global measures of WM FA and WM MD were calculated as the weighted average of all ROIs. Overall WMH volume, another marker for WMD, was calculated using algorithms developed by Mayo Clinic.⁷

Prevalent AF was considered to be present if AF was diagnosed at any ARIC study visit preceding and including Visit 5, concurrent with brain MRI. AF diagnosis was based on the following criteria: standard electrocardiograph or presence of International Classification of Diseases, Ninth Revision Clinical Modification codes 427.3x in the absence of open heart surgery in any prior hospitalization.⁸ Covariates included in the analysis are: age, sex, race/center, systolic and diastolic blood pressure, high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C), education, smoking, alcohol use, body mass index (BMI), hypertension medication, diabetes, electrocardiogram-based left ventricular hypertrophy, plasma C-reactive protein (CRP), estimated glomerular filtration rate (eGFR), APOE genotype, and anticoagulant use at Visit 5. Covariates measurements and definitions are described in the Online-only Supplement.

Statistical Analysis

Associations of AF with WMH volume and regional or overall WM FA and WM MD were assessed using weighted linear regression with robust estimators of the variance, with weights based on the probabilities of completion of the brain MRI among visit 5 ARIC participants. For all models, we adjusted for covariates listed above. We performed additional analyses restricting to participants without prior history of stroke or silent cerebral infarct (SCI) in their brain MRI.

Results

The study included 1899 eligible participants. Participants diagnosed with AF were older (78 yrs vs 76 yrs) and more likely to be male and white (Table 1). Prevalence of stroke, SCI, and diabetes was higher in participants with AF. Prevalence of other vascular risk factors was similar in the two groups. Unadjusted WM FA and WM MD measures by AF status are included in the online-only supplement.

Table 1.

Characteristics of Eligible Study Participants by Atrial Fibrillation (AF) Status

	No AF (N=1766)	AF (N=133)
Age, yrs	76 (5)	78 (5)
Female	61%	45%
Black	29%	15%
Education
≤ 11 yrs	14	16
12 – 16 yrs	41	40
> 16 yrs	45	44
Current drinker, %	45	50
Current smoker, %	5	5
Prevalent stroke or Silent cerebral infarct %	24	30
Diabetes mellitus, %	32	38
Anticoagulant use, %	2	50
Hypertension, %	67	65
LVH presence, %	6	6
BMI, kg/m2	28 (6)	28 (5)
High-sensitivity C-Reactive Protein, mg/L	3.9 (6)	5.1 (9)
Systolic blood pressure, mm Hg	131 (18)	130 (20)
Diastolic blood pressure, mm Hg	66 (11)	65 (12)
eGFR, mL/min per 1.73 m2	66 (18)	62 (17)
HDL cholesterol, mmol/dL	1.4 (0.4)	1.3(0.4)
LDL cholesterol, mmol/dL	2.7(0.9)	2.5(0.9)
APOE e4 allele, %	29	23
White matter hyperintensity volume, mm3	17305 (16864)	21037 (20320)

AF was not associated with WM FA or WM MD across all regions of interest after adjusting for vascular risk factors (Table 2). AF was not associated with WMH volume after adjusting for covariates (Table 2). Results from analysis excluding participants with stroke/SCI were comparable (Online-only Supplement). Additional exploratory analyses showed that age, C-reactive protein and APOE genotype strongly attenuated the association between AF and WM MD whereas the association between AF and WMH was strongly attenuated by age. Of note, AF and WM FA were not associated in the crude model (Online-only Supplement).

Table 2.

Association between AF and White Matter Microstructural Integrity Measures and White Matter Hyperintensity (WMH) volume

	Fractional Anisotropy (FA) *	Mean Diffusivity (MD) *, 10−4 mm2/s
Global	−0.001 (−0.006, 0.004)	0.031 (−0.075, 0.137)
Frontal	−0.001(−0.006, 0.005)	0.026 (−0.091, 0.143)
Temporal	−0.001 (−0.006, 0.005)	0.027 (−0.086, 0.140)
Parietal	−0.002 (−0.006, 0.003)	0.026 (−0.080, 0.132)
Occipital	−0.003 (−0.005, 0.004)	0.061 (−0.051, 0.187)
Anterior corpus callosum	−0.005 (−0.021, 0.011)	0.094 (−0.121, 0.309)
Posterior corpus callosum	0.003 (−0.011, 0.017)	0.121 (−0.070, 0.322)
	WMH Volume, mm3
Model 1 **	0.08 (−0.14, 0.30)

^*Numbers correspond to difference (95% confidence interval) in FA or MD measurements between those with AF and those without after adjustment

^**Numbers correspond to differences (95% confidence interval) in log-transformed WMH volume between those with AF and those without after adjustment

Covariates include: age, sex, race/center, systolic/diastolic blood pressure, HDL-C, LDL-C, education, smoking, alcohol use, BMI, hypertension medication, presence of stroke or infarct, WMH, diabetes, ECG-based left ventricular hypertrophy, plasma CRP, eGFR, APOE genotype, and anticoagulant use at Visit 5.

Discussion

In our study, we did not find an independent association between AF and WMD, using white matter microstructural integrity and WMH as markers. Such finding remained after excluding participants with prior stroke/SCI. Our results are consistent with some prior studies evaluating the association of AF with burden of WMH,^3,9 but not others.¹⁰

Our study has strengths. It is one of the first to explore the association between AF and WMD using highly sensitive imaging markers of white matter microstructural integrity in a large and diverse community-based sample. Nonetheless, the cross-sectional design could lead to selection bias since participants at visit 5 are less likely to suffer from severe WMD than those that were lost to follow-up, or drop out due to competing risks, which may contribute to our null results. Lastly, non-differential misclassification of AF could bias our estimate towards the null. To conclude, our study suggests that there is no association between AF and WMD. Future studies would benefit from larger sample size and a longitudinal study design.