Carotid Atherosclerosis and Stroke in Atrial Fibrillation: The Atherosclerosis Risk in Communities (ARIC) Study

Abstract

Background and Purpose

Whether consideration of carotid intima-media thickness (cIMT) and carotid plaque would improve risk prediction of ischemic stroke in persons with atrial fibrillation (AF) is unknown. The purpose of this study was to assess the improvement in risk prediction of stroke by adding cIMT and carotid plaque to the CHA₂DS₂-VASc score.

Methods

We included participants from the ARIC study (mean age, 63 years) who developed AF within 5 years after carotid measurement, were not on warfarin, and had no prior stroke at AF diagnosis. AF was ascertained from study ECGs and diagnosis codes, and stroke was physician-adjudicated. Multivariable Cox models were used to assess association between carotid indices and ischemic stroke. Improvement in 10-year risk prediction of stroke was assessed by the C-statistic, net reclassification improvement (NRI), and relative integrated discrimination improvement (IDI).

Results

There were 81 (11.2%) stroke events that occurred among 724 participants with AF during a mean follow-up of 8.5 years. Increased cIMT and presence of carotid plaque were significantly associated with increased stroke risk. The addition of cIMT + plaque to the CHA₂DS₂-VASc score marginally increased the C statistic (95% CI) from 0.685 (0.623–0.747) to 0.698 (0.638–0.759). The NRI and IDI for cIMT + plaque were 0.091 (95% CI, 0.012–0.170) and 0.101 (95% CI, 0.002–0.226), respectively.

Conclusions

Increased cIMT and presence of carotid plaque are associated with increased risk of ischemic stroke in individuals with AF. Further, they may improve risk prediction of stroke, over and above the CHA₂DS₂-VASc score.

Introduction

The association of carotid atherosclerosis (carotid intima-media thickness [cIMT] and presence of carotid plaque) with the risk of stroke in a large cohort of community-dwelling adults with atrial fibrillation (AF) has not been examined. It is unknown whether addition of cIMT and carotid plaque would improve risk prediction of stroke over the CHA₂DS₂-VASc risk score (1) among individuals with AF.

We evaluated the association of carotid atherosclerosis with ischemic stroke among participants with AF in the Atherosclerosis Risk in Communities (ARIC) study.

Methods

Study Population

The ARIC study is a population-based prospective study of cardiovascular disease (CVD) in a bi-racial cohort of 15,792 participants aged 45 to 64 years at enrollment (1987-1989) sampled from four US communities (2).

Measurements

AF was ascertained using standardized methods (3). cIMT and carotid plaque were measured by ultrasound (4). Carotid indices and covariates were taken from the most recent visit before AF ascertainment. Covariates included age at time of AF ascertainment, sex, race, ARIC field center, hypertension, heart failure, diabetes, CHD, peripheral arterial disease (PAD), and warfarin use.

Statistical Analyses

cIMT was evaluated in quintiles and as a continuous variable. To estimate the association of cIMT and carotid plaque with time to incident ischemic stroke, we calculated hazard ratios (HRs) and 95% confidence intervals (CIs) using multivariable Cox models. Person-years at risk were calculated from AF ascertainment until date of development of stroke, death, loss to follow up, or end of follow up (December 2011), whichever occurred first. The first multivariate model was adjusted for age, sex, race, and center. The second model additionally adjusted for CHA₂DS₂-VASc variables (heart failure, hypertension, diabetes, PAD, and CHD). Race was included in the models, because AF risk factors, incidence, and AF-related outcomes differ between blacks and whites (5).

We utilized the CHA₂DS₂–VASc score just prior to AF diagnosis as the benchmark to assess the role of arterial indices in enhancing risk prediction for stroke in AF. We considered 4 models: the benchmark alone, addition of cIMT or carotid plaque alone, and addition of both cIMT and carotid plaque to the benchmark. To assess model discrimination, we computed the C-statistic using methods which accounted for censoring for each of the models. We also calculated net reclassification improvement (NRI), Grønnesby-Borgan statistic, and the integrated discrimination improvement (IDI). Statistical analyses were performed with SAS v 9.3 (SAS Inc, Cary, NC).

Results

Participants who suffered strokes during follow up were more likely to be female, black, had traditional atherosclerotic risk factors, higher cIMT, and presence of carotid plaque (Table 1).

Table 1.

Participant Characteristics at the time of Atrial Fibrillation Diagnosis, ARIC Study, 1987-2011.

		Incident stroke
Characteristics	Total sample (n=724)	No (n=643)	Yes (n=81)
Age, mean(SD), y	63.3 (5.8)	63.3 (5.9)	63.3 (5.6)
Female sex	40.1%	39.5%	44.4%
Black race	16.4%	15.4%	24.7%
Diabetes	23.1%	21.8%	33.3%
Hypertension	51.8%	49.8%	67.9%
Previous MI	4.83%	4.20%	9.88%
Heart failure	3.18%	2.80%	6.17%
Peripheral artery disease	6.77%	6.53%	8.64%
cIMT, mean(SD), mm	0.922 (0.31)	0.916 (0.31)	0.969 (0.28)
Plaque	38.1%	37.1%	46.9%

cIMT, carotid intimal medial thickness.

Association of cIMT and carotid plaque with ischemic stroke risk in AF (Tables 1, 2)

Table 2.

Association of Carotid Intima-Media Thickness and Carotid Plaque with Ischemic Stroke in Participants with Atrial Fibrillation, ARIC Study, 1987-2011.

	cIMT Quintiles (mm) (n=713)						No Plaque	Plaque
	<0.705	0.705-0.789	0.790-0.899	0.900-1.079	1.080+	P for trend‡
Stroke cases | AF (N)	13 | 147	10 | 139	18 | 136	18 | 140	22 | 151		43 | 448	38 | 276
Person-years	1472.8	1313.1	1110.4	1157.3	977.1		3959	2164
Incidence rate (95%CI)*	8.8 (4.7-15.1)	7.6 (3.7-14.0)	16.2 (9.6-25.6)	15.6 (9.2-24.6)	22.5 (14.1-34.1)		10.9 (8.0-14.5)	17.6 (12.6-23.8)
HR (95%CI), Unadjusted	1(ref.)	0.88 (0.38-2.00)	1.78 (0.87-3.65)	1.74 (0.85-3.57)	2.42 (1.21-4.82)	0.003	1(ref.)	1.59 (1.02-2.45)
HR (95%CI), M1	1(ref.)	0.82 (0.36-1.88)	1.63 (0.79-3.39)	1.68 (0.80-3.52)	2.30 (1.12-4.74)	0.005	1(ref.)	1.56 (1.00-2.41)
HR (95%CI), M2	1(ref.)	0.73 (0.32-168)	1.42 (0.68-2.97)	1.25 (0.58-2.67)	1.56 (0.74-3.32)	0.10	1(ref.)	1.56 (1.00-2.45)
	cIMT Continuous per 1 SD†
HR (95%CI), Unadjusted	1.26 (1.08-1.46); p=0.003
HR (95%CI), M1	1.26 (1.08-1.48); p=0.003
HR (95%CI), M2	1.23 (1.04-1.46); p=0.02

^*Incidence rates are crude stroke incidence rates, per 1000person-years

^†1-SD=0.281

^‡P-values for trend were calculated across quintile categories using the quintile term

Model 1: adjusted for age, sex, race, and center

Model 2: adjusted for M1+CHA₂DS₂-VASc variables (heart failure, hypertension, diabetes, PAD, MI)

There were 81 (11.2%) stroke events occurring among 724 participants with AF during a mean follow-up of 8.5 years (supplemental Figure I). Compared with participants in the lowest quintile, the HR (95% CI) of stroke for those in the highest quintile of cIMT was 2.30 (1.12-4.74), p-value for trend=0.005, after adjustment for age, sex, race and field-center (supplemental figure II). This association was attenuated after additional adjustment for CHA₂DS₂-VASc variables. Each 1 SD increase in cIMT was associated with a 23% increased risk of stroke. The relationship between cIMT and stroke did not differ by sex (interaction p=0.25) or race (p=0.53).

Compared to participants without plaque, presence of plaque was associated with a 56% increased risk of stroke after adjustment for age, sex, race, and field-center. These results remained unchanged after additional adjustment for CHA₂DS₂-VASc variables, HR (95% CI), 1.56 (1.00-2.45). There was no interaction of the relationship between carotid plaque and stroke by sex (p=0.31) or race (p=0.43).

Model Discrimination, Calibration, and Reclassification (Table 3)

Table 3.

C-Statistic, NRI, and Relative IDI of Predictive Models for Ischemic Stroke in Participants with Atrial Fibrillation, ARIC Study, 1987-2011.

	C-statistic (95%CI)	Calibration Χ2 (P Value)	Category-based NRI (95%CI)	Relative IDI (95%CI)
Benchmark	0.685(0.623-0.747)	5.05(0.83)	-	-
Benchmark+cIMT	0.697(0.636-0.758)	6.42(0.70)	0.073(−0.029-0.182)	0.085(−0.007-0.198)
Benchmark+plaque	0.690(0.629-0.751)	5.85(0.76)	0.108(−0.007 to 0.229)	0.059(−0.028-0.158)
*Benchmark+cIMT+plaque	0.698(0.638-0.759)	11.12(0.27)	0.131(0.010-0.226)	0.101(0.002-0.226)

Based on a 10 year risk prediction

^*Results in BOLD denote the best model for prediction of stroke in AF

cIMT is added as per 1 SD of measure

NRI is category-based; categories of <5%, 5-10%, >10%

Benchmark: race and CHA₂DS₂-VASc variables(age, heart failure, hypertension, diabetes, PAD, MI)

C-statistic increased from 0.685 (95% CI 0.623-0.747) for the model including race and CHA₂DS₂-VASc variables to 0.697 (95% CI 0.636-0.758) after addition of cIMT alone; 0.690 (95% CI 0.629-0.751) after addition of plaque information alone; and 0.698 (95% CI 0.638-0.759) after addition of both cIMT and plaque information. Addition of both plaque and cIMT to the model including race and CHA₂DS₂-VASc variables resulted in a significant NRI, 0.091 (95% CI 0.012-0.170). Overall, the risk levels for participants reclassified because of cIMT and carotid plaque data were more accurate (Supplemental Table I).

The IDI showed improved risk classification after addition of cIMT and plaque information to model 2. The Grønnesby-Borgan statistic also showed good model fit for model 2, after addition of cIMT and plaque information. Overall, the best model for prediction of stroke in AF was one comprising race and CHA₂DS₂-VASc variables, with addition of cIMT (as per 1-SD of measure) and plaque information.

Discussion

In this population-based prospective study of participants with incident AF, cIMT and carotid plaque were found to be independent predictors of 10-year ischemic stroke risk in middle-aged adults. Addition of cIMT and carotid plaque information provided incremental predictive value for risk of stroke in adults with AF, over the CHA₂DS₂-VASc risk score. Direct evaluation of carotid plaque and cIMT using non-invasive and readily available imaging modalities like carotid ultrasonography may contribute to improvement in risk prediction of stroke in AF over the CHA₂DS₂-VASc risk score which uses clinical variables alone.

For patients with intermediate risk of stroke (CHA₂DS₂-VASc score of 1), the 2012 ESC guidelines (6) recommend oral anticoagulation; while the 2014 AHA/ACC/HRS guidelines recommend no antithrombotic therapy, treatment with oral anticoagulant or aspirin (7). Hence there is equipoise in management of these patients. Our findings provide preliminary evidence that carotid ultrasonography may be useful in reclassifying these patients. These results also reinforce the concept that atherosclerosis plays an important role in the etiology of stroke in AF.

Strengths of this study include the extensive measurement of covariates, large number of AF cases, over 40% female, and inclusion of non-white participants. There are however, a few limitations. First, it is possible that non-hospitalized stroke events that were not validated in the study could influence the results. However, the magnitude of any potential underestimation of the rate of stroke is likely to be small (<5%) (8). Second, there may be misclassification of AF. However, prior analysis within the ARIC cohort to determine the validity of hospital discharge diagnoses for AF reported 84% sensitivity and 98% specificity in AF ascertainment (3). Third, our study population has a mean age of 63 followed by an 8.5 year mean follow up. This may limit generalizability to the elderly.

Conclusions

Further studies are needed to validate these findings in other populations and develop a scoring system that would enhance stroke risk prediction by incorporating cIMT and carotid plaque into the CHA₂DS₂-VASc score.