Chen LY, Sotoodehnia N, Bužková P, et al. Atrial fibrillation and the risk of sudden cardiac death: the Atherosclerosis Risk in Communities Study and Cardiovascular Health Study. JAMA Intern Med. Published online November 26, 2012. doi:10.1001/2013.jamainternmed.744.
eMethods. Ascertainment of Atrial Fibrillation and Covariates
eTable 1. Risk of Definite Sudden Cardiac Death by Atrial Fibrillation Status, Atherosclerosis Risk in Communities Study (ARIC) and Cardiovascular Health Study (CHS)
eTable 2. Risk of Sudden Cardiac Death and Nonsudden Cardiac Death by Atrial Fibrillation Status with Adjustment for Left Ventricular Fractional Shortening, Atherosclerosis Risk in Communities Study, Jackson Field Center, Mississippi
eTable 3. Risk of Sudden Cardiac Death and Nonsudden Cardiac Death by Atrial Fibrillation Status with Adjustment for Left Ventricular Ejection Fraction, Cardiovascular Health Study
Chen LY, Sotoodehnia N, Bůžková P, Lopez FL, Yee LM, Heckbert SR, Prineas R, Soliman EZ, Adabag S, Konety S, Folsom AR, Siscovick D, Alonso A. Atrial Fibrillation and the Risk of Sudden Cardiac DeathThe Atherosclerosis Risk in Communities Study and Cardiovascular Health Study. JAMA Intern Med. 2013;173(1):29-35. doi:10.1001/2013.jamainternmed.744
Author Affiliations: Cardiovascular Division, Department of Medicine, University of Minnesota Medical School (Drs Chen and Konety), and Division of Epidemiology and Community Health, School of Public Health (Ms Lopez and Drs Folsom and Alonso), University of Minnesota, Minneapolis; Cardiovascular Health Research Unit (Drs Sotoodehnia, Heckbert, and Siscovick) and Division of Cardiology (Dr Sotoodehnia), Department of Medicine (Dr Siscovick), and Departments of Biostatistics (Dr Bůžková and Ms Yee) and Epidemiology (Drs Heckbert and Siscovick), University of Washington, Seattle; Division of Public Health Sciences (Dr Prineas) and Epidemiological Cardiology Research Center (Dr Soliman), Wake Forest School of Medicine, Winston-Salem, North Carolina; and Division of Cardiology, Veterans Affairs Medical Center, Minneapolis (Dr Adabag).
Background It is unknown whether atrial fibrillation (AF) is associated with an increased risk of sudden cardiac death (SCD) in the general population. This association was examined in 2 population-based cohorts.
Methods In the Atherosclerosis Risk in Communities (ARIC) Study, we analyzed data from 15 439 participants (baseline age, 45-64 years; 55.2% women; and 26.6% black) from baseline (1987-1989) through December 31, 2001. In the Cardiovascular Health Study (CHS), we analyzed data from 5479 participants (baseline age, ≥65 years; 58.2% women; and 15.4% black) from baseline (first cohort, 1989-1990; second cohort, 1992-1993) through December 31, 2006. The main outcome was physician-adjudicated SCD, defined as death from a sudden, pulseless condition presumed to be due to a ventricular tachyarrhythmia. The secondary outcome was non-SCD (NSCD), defined as coronary heart disease death not meeting SCD criteria. We used Cox proportional hazards models to assess the association between AF and SCD/NSCD, adjusting for baseline demographic and cardiovascular risk factors.
Results In the ARIC Study, 894 AF, 269 SCD, and 233 NSCD events occurred during follow-up (median, 13.1 years). The crude incidence rates of SCD were 2.89 per 1000 person-years (with AF) and 1.30 per 1000 person-years (without AF). The multivariable hazard ratios (HRs) (95% CIs) of AF for SCD and NSCD were 3.26 (2.17-4.91) and 2.43 (1.60-3.71), respectively. In the CHS, 1458 AF, 292 SCD, and 581 NSCD events occurred during follow-up (median, 13.1 years). The crude incidence rates of SCD were 12.00 per 1000 person-years (with AF) and 3.82 per 1000 person-years (without AF). The multivariable HRs (95% CIs) of AF for SCD and NSCD were 2.14 (1.60-2.87) and 3.10 (2.58-3.72), respectively. The meta-analyzed HRs (95% CIs) of AF for SCD and NSCD were 2.47 (1.95-3.13) and 2.98 (2.52-3.53), respectively.
Conclusions Incident AF is associated with an increased risk of SCD and NSCD in the general population. Additional research to identify predictors of SCD in patients with AF is warranted.
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia, and its prevalence is increasing.1,2 Atrial fibrillation is associated with an increased risk of stroke,3 heart failure,4 and death.5- 7 The Framingham Heart Study5 reported that AF increases the risk of death by 1.5-fold in men and 1.9-fold in women. Similarly, a study6 from Olmsted County, Minnesota, showed that new-onset AF doubles the risk of mortality. More recently, the Women's Health Study7 showed that the risk of all-cause death was doubled and cardiovascular death quadrupled by new-onset AF in initially healthy women.
The common causes of death in individuals with AF in these studies were coronary heart disease (CHD) and stroke.5,6 Sudden cardiac death (SCD) was not specifically reported.5- 7 However, there is evidence from patients who had experienced a myocardial infarction or those with heart failure that AF is associated with an increased risk of SCD.8- 10 It is unknown, however, whether AF increases the risk of SCD in the general population. Moreover, since AF and SCD share many risk factors, such as heart failure11,12 and CHD,13,14 it is unknown whether AF is independently associated with an increased risk of SCD in the general population.
We hypothesized that incident AF is associated with an increased risk of SCD in the general population. We tested our hypothesis in the Atherosclerosis Risk in Communities (ARIC) Study and the Cardiovascular Health Study (CHS), 2 large community-based cohort investigations of cardiovascular disease in the United States.
The ARIC Study cohort is a biracial sample consisting of 15 792 men and women, aged 45 to 64 years at baseline (1987-1989), from 4 communities in North Carolina, Mississippi, Minnesota, and Maryland.15 After the baseline examination, participants had 3 additional examinations, the last in 1996-1998. In addition to study examinations, ARIC Study participants have received annual follow-up telephone calls since their first visit (>90% response rate) collecting information on general health and hospitalizations. The present study used data obtained from baseline (1987-1989) through December 31, 2001. We excluded participants with missing or uninterpretable electrocardiograms at baseline (n = 243), missing covariates (n = 73), or prevalent AF (n = 37). The final analysis cohort consisted of 15 439 ARIC Study participants.
The CHS is a cohort study of risk factors for CHD and stroke in older people.16 Between 1989 and 1990, 4 field centers (North Carolina, California, Maryland, and Pennsylvania) recruited 5201 participants aged 65 years or older from Medicare eligibility lists. To enhance minority representation, during 1992-1993, 687 black participants were recruited. The present study used data obtained from baseline (1989-1990 for the first cohort and 1992-1993 for the second cohort) through December 31, 2006. We excluded participants with missing covariates (n = 260) or prevalent AF (n = 149). The final analysis CHS cohort consisted of 5479 participants. The CHS and ARIC study protocols were approved by the institutional review board of each participating center, and informed consent was obtained from each study participant.
In the ARIC Study and the CHS, AF was ascertained from study electrocardiograms and hospital discharge records.14,17 Details are reported in the eMethods.
In the ARIC Study and the CHS, comprehensive data were gathered on cardiovascular events and deaths from hospital records; interviews with physicians, next of kin, and/or witnesses; death certificates; and autopsy reports, where available. Causes of death were adjudicated by respective ARIC Study and the CHS events committees. An independent review of CHD deaths18 was conducted by each cohort study to identify SCD events. The primary outcome, SCD, was similarly defined in the ARIC Study and the CHS: a sudden pulseless condition presumed to be due to a ventricular tachyarrhythmia in a previously stable individual without evidence of a noncardiac cause of cardiac arrest. All SCD cases occurred outside the hospital or in the emergency department, and the individuals could not have a life-threatening noncardiac comorbidity or be under hospice care. For unwitnessed deaths, the participant must have been seen within 24 hours of the arrest in a stable condition and without evidence of a noncardiac cause of cardiac arrest.
In the ARIC Study, all CHD deaths that occurred by December 31, 2001, were reviewed by a panel of 5 physicians (including N.S. and R.P.) to identify SCD events. Each event was independently adjudicated by 2 physicians. If there was disagreement, a third investigator reviewed the event to provide final classification. After review of available data, CHD deaths were classified as definite sudden arrhythmic death, possible sudden arrhythmic death, not sudden arrhythmic death, or unclassifiable.19- 21 For the present analysis, SCD was defined as definite or possible sudden arrhythmic deaths in the ARIC Study.
In the CHS, all CHD deaths through December 31, 2006, were reviewed by a cardiologist (N.S.) to classify SCD cases. A blinded second physician review of a random sample of 70 of these death records showed an 88% interreviewer agreement and κ = 0.74 for SCD. Both of these physicians also participated on the ARIC Study SCD review panel. After review of available data in the CHS, CHD deaths were classified as definite, possible, or not SCD. For the present analysis, the CHS SCD definition included definite and possible SCD. In both cohorts, the secondary outcome, non-SCD (NSCD), was defined as CHD death not meeting SCD criteria.
For the main analysis, we used covariates measured at baseline. Definitions of the covariates are detailed in the eMethods.
We report means (SDs) for continuous variables and counts with percentages for categorical variables. Person-years at risk were calculated from the date of baseline until the date of SCD/NSCD, other death, loss to follow-up, or end of follow-up, whichever occurred first.
To estimate the association of AF with risks of SCD and NSCD, we calculated hazard ratios (HRs) and 95% CIs, using a Cox proportional hazards model with AF as a time-dependent exposure variable. We ran 2 models: in model 1, we adjusted for age, sex, race, and field center. In model 2, we additionally adjusted for the baseline covariates: heart rate, smoking status, body mass index, hypertension, diabetes mellitus, CHD, heart failure, electrocardiogram-determined left-ventricular hypertrophy, use of β-blockers, use of digoxin, and use of antiarrhythmic drugs. We adjusted for antiarrhythmic drugs because their use may be a risk factor for SCD. We conducted 2 additional analyses, using time-dependent covariates in the ARIC Study. First, to account for confounding by covariates changing over time, we updated covariates to the time just before ascertainment of AF, censoring, or SCD incidence, whichever occurred earliest. Second, to assess whether the association between AF and SCD is mediated by shared cardiovascular risk factors, we updated the covariates to the end of follow-up.
In addition, we conducted 2 sensitivity analyses. First, we restricted the definition of SCD to include only cases that were classified as definite sudden arrhythmic death (252 participants in the ARIC Study and 194 in the CHS). Second, to control for possible confounding by left-ventricular systolic dysfunction, we adjusted for left-ventricular fractional shortening on 2-dimensional echocardiogram in ARIC Study participants at the Mississippi field center (2028 participants; all were black). In the CHS, we adjusted for left-ventricular ejection fraction (<45%, 45%-54%, and ≥55%) on 2-dimensional echocardiogram (n = 4816). The proportional hazards assumption was assessed with scaled Schoenfeld residuals for both graphical and numerical tests, time interaction terms, and inspection of log-negative log survival curves. Modeling assumptions were not violated in any model.
The ARIC and CHS results were meta-analyzed using fixed-effect analysis. The meta-analysis results were considered the primary results. To determine whether the HR of AF for SCD differed from the HR of AF for NSCD, we performed a proportional hazards competing-risk analysis.22
Commercial software was used for statistical analysis of ARIC Study data (SAS, version 9.2; SAS Institute Inc) and CHS data (R; R Foundation for Statistical Computing; and Stata, version 11.2; StataCorp). All P values reported were 2-sided, and the statistical significance threshold was chosen as 5%.
The cohort at risk for SCD in the ARIC Study consisted of 8524 women and 6915 men aged 45 to 64 years at baseline and, in the CHS, 3189 women and 2290 men aged 65 years or older at baseline (Table 1). Of the 5479 participants in the CHS, 4857 (88.6%) were from the first cohort and 622 (11.4%) were from the second cohort. During follow-up in the ARIC Study (median, 13.1; interquartile range [IQR], 12.4-13.9 years), 894 AF, 269 SCD, and 233 NSCD events occurred through 2001. During follow-up in the CHS (median, 13.1 years; IQR, 8.0-16.3 years), 1458 AF, 292 SCD, and 581 NSCD events occurred through 2006.
Compared with participants without AF, those with incident AF had higher incidence rates of SCD and NSCD (Table 2). After adjustment for age, sex, race, and ARIC Study field center, AF was significantly associated with an increased risk of SCD and NSCD (Table 2, model 1). Although additional adjustment for cardiovascular risk factors attenuated these risk estimates (Table 2, model 2), the associations remained statistically significant. Overall, the presence of incident AF was associated with a tripling of the risk of SCD and doubling of the risk of NSCD.
To account for confounding by change of covariates over time, we adjusted the main analysis for time-dependent covariates by updating the covariates to the time just before ascertainment of AF, censoring, or SCD incidence, whichever occurred earliest. We found that AF remained significantly associated with SCD (Table 3). Atrial fibrillation also remained significantly associated with NSCD (Table 3). To investigate whether shared risk factors could mediate the association between AF and SCD, we conducted another time-dependent regression analysis by updating the covariates to the time before the end of follow-up. We found that even after adjusting for risk factors that are potentially on the causal pathway, AF was significantly associated with an increased risk of SCD (Table 3). By contrast, AF was no longer significantly associated with NSCD. The nonsignificant association between AF and NSCD, however, could be the result of inadequate power. Moreover, the 2 HRs (2.03 vs 1.48) did not differ significantly (P = .33).
From sex-stratified analysis (Table 4), we found the risk of SCD associated with AF in women (HR, 4.12; 95% CI, 1.91-8.90; P < .001) to be comparable to the risk in men (3.12; 1.93-5.04; P < .001) (for interaction by sex, P = .60). Race-stratified analysis showed that the risk of SCD associated with AF was higher in black participants (HR, 5.77; 95% CI, 2.96-11.24; P < .001) than in nonblack participants (2.49; 1.49-4.17; P < .001) (for interaction by race, P = .02) (Table 4).
We performed 2 sensitivity analyses. First, we restricted the analysis to definite cases of SCD. The presence of AF was associated with a doubling of the risk of definite SCD (HR, 2.00; 95% CI, 1.22-3.28; P = .006) (eTable 1). Second, to control for possible confounding by left-ventricular systolic dysfunction, we restricted the analysis to participants at the ARIC Study Mississippi field center with fractional shortening measured by 2-dimensional echocardiogram. Of 2028 participants in this sample, there were 53 AF and 30 SCD events through 2001. After adjustment for left-ventricular fractional shortening, the HRs (95% CI) of AF for SCD and NSCD were 13.59 (4.20-43.93; P < .001) and 10.74 (2.87-40.24; P < .001), respectively (eTable 2).
Incident AF was associated with a doubling of the risk of SCD (HR, 2.14; 95% CI, 1.60-2.87; P < .001) and a tripling of the risk of NSCD (3.10; 2.58-3.72; P < .001) in the CHS (Table 2). Similar to the ARIC Study, the risk of SCD associated with AF in women (HR, 2.49; 95% CI, 1.57-3.95; P < .001) was comparable to that in men (1.99; 1.37-2.91; P < .001) (for interaction by sex, P = .33) (Table 4). In contrast to the ARIC Study, we did not find an interaction of race with AF risk (for interaction by race, P = .46) (Table 4).
From sensitivity analysis, we found that AF was associated with a doubling of the risk of definite SCD in the CHS (HR, 2.25; 95% CI, 1.56-3.23; P < .001) (eTable 1). After adjustment for left-ventricular ejection fraction (<45%, 45%-54%, and ≥55%) on 2-dimensional echocardiogram, AF was still associated with a significantly increased risk of SCD (HR, 2.07; 95% CI, 1.52-2.82; P < .001) and NSCD (2.92; 2.42-3.54; P < .001) (eTable 3).
The meta-analyzed HR (95% CI) of AF for SCD in the ARIC Study and the CHS was 2.47 (1.95-3.13; P < .001). The meta-analyzed HR (95% CI) of AF for NSCD was 2.98 (2.52-3.53; P < .001) (Table 2). These 2 HRs did not differ significantly (P = .21).
In 2 large population-based cohort studies in the United States including middle-aged and elderly individuals, we found that participants who developed incident AF had an increased risk of SCD compared with participants who did not develop AF. Incident AF was also associated with an increased risk of NSCD. The strength of the association between AF and SCD was comparable to that between AF and NSCD. From meta-analysis results of the ARIC Study and the CHS, compared with participants without AF, the risks of SCD and NSCD were more than doubled in participants with AF.
Although prospective cohort studies have compellingly demonstrated an association between AF and increased risk of total and cardiovascular mortality,5- 7 none has shown that AF increases the risk of SCD. However, AF may increase the risk of SCD in specific patient subgroups, such as those who have had a myocardial infarction or those with heart failure.8- 10 In the Trandolapril Cardiac Evaluation registry,8 development of AF during hospitalization for myocardial infarction was associated with a 1.3-fold higher risk of SCD during subsequent follow-up. In a study of patients who were discharged after hospitalization for myocardial infarction, AF development during hospitalization increased the risk of SCD by 2.7-fold during follow-up.10 In patients with severe heart failure, AF was associated with a higher risk of SCD compared with patients without AF—the 1-year SCD-free survival in patients with AF was 69% compared with 82% in patients without AF.9
To the best of our knowledge, the present study is the first to demonstrate that incident AF is associated with an increased risk of SCD in 2 independent population-based cohorts. This association was observed in men and women, as well as in blacks and nonblacks.
Several mechanisms might explain our observation. First, AF may facilitate the induction of ventricular tachyarrhythmias. A rapid ventricular rate during an atrial tachyarrhythmia will directly reduce ventricular refractoriness,23 promoting ventricular tachyarrhythmias. In addition, the irregular rhythm of AF leads to short-long-short sequences that may be intrinsically proarrhythmic.24 Evidence of AF facilitating induction of ventricular tachyarrhythmias comes from several sources. Somberg et al25 reported from canine experiments that ventricular tachycardia was induced in 25 of 26 dogs by programmed electrical stimulation only in AF and not in sinus rhythm. Grönefeld et al26 reported that AF is an independent predictor of implantable cardioverter-defibrillator therapy for ventricular tachycardia or ventricular fibrillation. Analysis of device-stored electrograms revealed a higher incidence of short-long-short cycles preceding ventricular arrhythmias in patients with AF compared with patients in sinus rhythm (50% vs 16%;P = .002). Collectively, these observations suggest that atrial tachyarrhythmias may increase susceptibility to ventricular tachyarrhythmias.
Second, because AF is also associated with an increased risk of NSCD, it is possible that the association between AF and SCD is mediated by shared risk factors, such as CHD or heart failure. To assess this possibility, we conducted a secondary analysis in the ARIC Study by updating covariates to the end of follow-up. Although the association between AF and SCD was attenuated in this analysis, AF remained significantly associated with SCD after adjustment for factors potentially in the causal pathway. This observation suggests that the association between AF and SCD is only partially explained by the measured shared risk factors.
The principal strength of this study is the reproducible finding of a strong association between AF and SCD in 2 independent, large population-based cohorts. Other strengths include the long follow-up, inclusion of nonwhite participants, extensive measurement of covariates, large number of AF cases, and physician-adjudication of all SCD cases. However, several limitations should be noted. First, incident AF was identified mostly from hospitalization discharges in the ARIC Study and the CHS, and we could not include asymptomatic AF or AF managed exclusively in an outpatient setting. However, previous studies have shown that the validity of AF ascertainment using hospitalizations in the ARIC Study and the CHS is acceptable,14,17 that incidence rates of AF in the ARIC Study and the CHS are consistent with those of other population-based studies,2,11,14,17 and that the associations between genetic variants in the chromosome 4q25 locus and AF—extremely specific for AF risk—in the ARIC Study and the CHS are similar to those in studies with a more rigorous ascertainment of AF.27 Second, we could adjust for left-ventricular systolic function in only a subgroup of the ARIC Study cohort. The small sample size for this subgroup analysis was reflected in the wide confidence intervals. However, even after this adjustment, the association between AF and SCD remained statistically significant. Moreover, after adjustment for left-ventricular ejection fraction in the CHS, AF remained a significant risk factor for SCD. Third, although we adjusted for multiple potential confounders in our analyses, we cannot exclude residual confounding by imperfectly measured and unmeasured factors. Finally, we have limited power to detect a significant difference comparing the strengths of association between AF and SCD vs AF and NSCD.
In conclusion, in 2 large, population-based cohorts of middle-aged and elderly individuals, incident AF independently increased the risk of SCD. This finding should be confirmed in additional studies, and if confirmed, it adds to our evolving understanding that AF is not a benign condition. Not only does AF predispose to stroke, heart failure, and death, but the arrhythmia per se may increase the risk of death from ventricular tachyarrhythmias. The latter is potentially preventable; to this end, additional research to identify predictors of SCD in patients with AF is much needed.
Correspondence: Lin Y. Chen, MD, MS, Cardiovascular Division, Department of Medicine, University of Minnesota Medical School, 420 Delaware St SE, MMC 508, Minneapolis, MN 55455 (email@example.com).
Accepted for Publication: August 8, 2012.
Published Online: November 26, 2012. doi:10.1001/2013.jamainternmed.744
Author Contributions: Drs Alonso and Sotoodehnia had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Chen, Bůžková, Lopez, Heckbert, Soliman, Adabag, and Alonso. Acquisition of data: Sotoodehnia, Heckbert, Prineas, Soliman, Folsom, Siscovick, and Alonso. Analysis and interpretation of data: Chen, Sotoodehnia, Bůžková, Lopez, Yee, Heckbert, Adabag, Konety, Siscovick, and Alonso. Drafting of the manuscript: Chen, Bůžková, and Lopez. Critical revision of the manuscript for important intellectual content: Chen, Sotoodehnia, Bůžková, Lopez, Yee, Heckbert, Prineas, Soliman, Adabag, Konety, Folsom, Siscovick, and Alonso. Statistical analysis: Bůžková, Lopez, and Yee. Obtained funding: Heckbert, Prineas, Siscovick, and Alonso. Administrative, technical, or material support: Alonso. Supervision: Chen, Sotoodehnia, Heckbert, and Alonso.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by grant funding from the National Institutes of Health. The ARIC 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. This CHS research was supported by contracts HHSN268201200036C, N01-HC-85239, N01-HC-85079 through N01-HC-85086, N01-HC-35129, N01 HC-15103, N01 HC-55222, N01-HC-75150, and N01-HC-45133, and grants HL080295 and HL102214 from the National Heart, Lung, and Blood Institute, with additional contribution from the National Institute of Neurological Disorders and Stroke. Additional support was provided through AG-023629, AG-15928, AG-20098, and AG-027058 from the National Institute on Aging. This research was conducted while Dr Chen was a recipient of Scientist Development Grant 10SDG3420031 from the American Heart Association. This study was also supported in part by grants RC1 HL099452 (Dr Alonso), R01HL088456 (Dr Sotoodehnia), and R01 HL102214 (Dr Heckbert) from the National Heart, Lung, and Blood Institute and by grant 09SDG2280087 (Dr Alonzo) from the American Heart Association.
Role of the Sponsors: The funding agencies had no role in the design or conduct of the study; in the collection, management, analysis, or interpretation of the data; or in the preparation, review, or approval of the manuscript.
Previous Presentations: This study was presented at the American Heart Association Scientific Sessions; November 14, 2011; Orlando, Florida; and the Great Wall International Congress of Cardiology; October 13, 2012; Beijing, China.
Additional Information: A full list of principal CHS investigators and institutions can be found at http://www.chs-nhlbi.org/pi.htm.