|Home | About | Journals | Submit | Contact Us | Français|
Chronic kidney disease (CKD) is associated with the incidence of cardiovascular disease. CKD may also increase the risk of atrial fibrillation (AF), but existing studies have reported inconsistent results.
We estimated cystatin C-based glomerular filtration rate (eGFRcys) and measured urinary albumin-creatinine ratio (ACR) in 10,328 men and women free of AF from the Atherosclerosis Risk in Communities (ARIC) Study in 1996–98. Incidence of AF was ascertained through the end of 2007. During a median follow-up of 10.1 years, we identified 788 incident AF cases. Compared to individuals with eGFRcys ≥90 mL/min/1.73 m2, multivariable hazard ratios (HR) and 95% confidence intervals (CI) of AF were 1.3 (1.1–1.6), 1.6 (1.3–2.1), and 3.2 (2.0–5.0) (p for trend <0.0001) in those with eGFRcys of 60–89, 30–59 and 15–29 mL/min/1.73 m2, respectively. Similarly, presence of macroalbuminuria (ACR ≥300 mg/g, HR 3.2, 95% CI 2.3–4.5) and microalbuminuria (ACR 30–299 mg/g, HR 2.0, 95% CI 1.6–2.4) were associated with higher AF risk compared to those with ACR <30 mg/g. Risk of AF was particularly elevated in those with both low eGFRcys and macroalbuminuria (HR 13.1, 95% CI 6.0–28.6, comparing individuals with ACR≥300 mg/g and eGFRcys 15–29 vs. ACR<30 mg/g and eGFRcys ≥90 mL/min/1.73 m2).
In this large population-based study, reduced kidney function and presence of albuminuria were strongly associated with the incidence of AF independently of other risk factors.
Individuals with chronic kidney disease (CKD) are at higher risk for coronary heart disease (CHD), heart failure, peripheral artery disease, and venous thromboembolism, independently of other risk factors.1–4 In addition, CKD leads to hypertension, left ventricular hypertrophy, inflammation, and cardiovascular disease5–factor spotentially associated with an increased risk of atrial fibrillation (AF), a common cardiac arrhythmia.6 CKD can also lead to alterations in the renin-angiotensin-aldosterone (RAA) system which, as recent evidence suggests, may produce atrial fibrosis and increase the risk of AF.7 Finally, CKD causes sympathetic activation, a potential trigger of AF.8, 9
Evidence shows that patients with end-stage renal disease on hemodialysis are more likely to develop AF than the general population.10 Similarly, cross-sectional studies have found a higher prevalence of AF in individuals with non-dialysis-dependent CKD.11–13 To date, three prospective population-based studies have evaluated the association of kidney function with AF, reporting conflicting results.14–16 Additionally, none of these studies explored whether this association might differ by race, gender, or presence of cardiovascular disease and risk factors.
Therefore, we examined the association of kidney function and urinary albumin excretion with the incidence of AF in the Atherosclerosis Risk in Communities (ARIC) cohort, a community-based study of cardiovascular disease in the US. We hypothesized that individuals with worse kidney function and more severe albuminuria would have an increased risk of AF independent of other cardiovascular risk factors.
In 1987–89, the ARIC Study recruited 15,792 men and women, aged 45–64, from 4 communities in the US (Washington County, MD; suburbs of Minneapolis, MN; Jackson, MS; Forsyth County, NC).17 Participants were mostly white in the Minneapolis and Washington County field centers, white and African-American in Forsyth County, while only African-Americans were recruited in the Jackson field center. Study participants had three additional exams, each approximately three years apart (the last in 1996–98). Response rates among survivors for the successive examinations were 93%, 86%, and 80%. In addition, ARIC participants have received annual follow-up calls since the first visit (>90% response rate). Because cystatin C and urinary albumin were measured in samples collected during the 1996–98 exam (ARIC visit 4), we considered that visit as baseline. Institutional review boards at participating institutions approved the study protocol. All study participants provided written informed consent.
Serum creatinine was measured in samples collected during visit 4 using a modified kinetic Jaffé reaction. Reliability coefficient for 439 blinded quality control replicates was 0.95. Creatinine values were calibrated to the Cleveland Clinic Laboratory.18, 19 Serum cystatin C was measured in 2008 from stored frozen samples collected in visit 4 by a particle-enhanced immunophelometric assay (N Latex Cystatin C, Siemens Healthcare Diagnostics, Deerfield, IL) with a BNII nephelometer (Siemens Healthcare Diagnostics, Deerfield, IL). The reliability coefficient for 421 blinded quality control replicates of cystatin C was 0.65, but was 0.94 after removing 10 pair outliers (>3 standard deviations). Cystatin C was calibrated to the Cleveland Clinic after a relatively constant difference of 16% was found between ARIC and Cleveland Clinic values (Cleveland Clinic=1.16 × ARIC). Estimated glomerular filtration rate (eGFR) based on creatinine (eGFRcreat) was calculated using the CKD Epidemiology Collaboration (CKD-EPI) equation for creatinine,20 while eGFR by cystatin C (eGFRcys) was estimated with the CKD-EPI equation for cystatin: eGFRcys (mL/min/1.73 m2) =127.7 × cystatin C (mg/dL)−1.17 × age−0.13 × 0.91 (if female) × 1.06 (if African-American).21
Urinary albumin was measured by a nephelometric method either on the Dade Behring BN100 (assay sensitivity, 2.0 mg/l) or on the Beckman Image Nephelometer, and urinary creatinine using the Jaffé method in order to determine the albumin-to-creatinine ratio (ACR; mg/g). Blinded samples (n= 516) analyzed for quality assurance showed the correlation coefficient of the loge-transformed ACR to be 0.95.
For the main analysis, we used visit 4 (1996–1998) as the baseline visit, and considered only incident AF events occurring after that time. Individuals with AF identified at visit 4 or before were considered to have prevalent AF and, therefore, excluded from this analysis.
Identification of AF events after visit 4 was done through hospital discharge codes and death certificates.22, 23 Hospitalizations in ARIC participants were identified through annual follow-up calls and review of local hospital discharges through the end of 2007.24 AF was identified when ICD-9-CM codes 427.31 or 427.32 were listed as a discharge diagnostic code. AF events identified during hospitalizations for cardiac surgery were excluded. We confirmed approximately 90% of the cases in a physician review of discharge summaries from 125 possible AF cases.22 Finally, an AF diagnosis was assigned if AF was listed as a cause of death (ICD-9 427.3 or ICD-10 I48). Most incident AF cases (>99%) in the current analysis were identified from hospital discharge codes.
AF events before ARIC visit 4 were identified from electrocardiograms (ECG) done at study visits, in addition to hospitalizations and death certificates. Specifically, at each study visit, a 10-second 12-lead ECG was done using a MAC PC cardiograph (Marquette Electronics Inc, Milwaukee, WI) and transmitted to the ARIC ECG Reading Center for coding and interpretation. ECGs automatically coded as AF were visually checked by a trained cardiologist to confirm the diagnosis.25 AF events before visit 4 were not included in the main analysis, but were considered in sensitivity analyses with visit 1 as baseline (see below).
At each study visit, participants reported information on smoking and alcohol intake, underwent a physical exam, and provided blood samples. For the present analysis, we used covariates measured at visit 4. Body mass index was calculated as the weight in kilograms divided by the height in meters squared. Blood pressure was measured using a random-zero sphygmomanometer after 5 minutes of rest in the sitting position and was defined as the average of 2 measurements taken. Diabetes was defined as fasting glucose ≥126 mg/dL, non-fasting glucose ≥200 mg/dL, treatment for diabetes, or a self-reported diagnosis of diabetes. High sensitivity C-reactive protein was measured using an immunonephelometric assay (Siemens Healthcare Diagnostics, Deerfield, IL). Heart failure at baseline was defined as the reported use of heart failure medications in the previous 2 weeks or presence of heart failure according to Gothenburg criteria, while incident heart failure was defined as the presence of ICD-9-CM code 428 in any hospitalization or death certificate during follow-up.26 Similarly, prevalent CHD was defined as physician-diagnosed CHD or presence of a previous myocardial infarction by ECG, while incident CHD was adjudicated by the ARIC Morbidity and Mortality Classification Committee using information obtained from follow-up calls, hospitalization records and death certificates, as previously published.24 Prevalent heart failure and CHD for this study were defined as prevalent disease at baseline plus incident events prior to visit 4.
Of 11,656 individuals who attended ARIC visit 4 we excluded individuals who were of a racial/ethnic group other than white or African-American and nonwhites in the Minneapolis and Washington County field centers (n=69), those with prevalent AF at visit 4 (n=304), unreadable ECG (n=202), or prevalent stroke (n=226), those with missing eGFRcys, eGFRcreat or ACR (n=420), or other missing covariates (n=92), and those with eGFRcys or eGFRcreat <15 ml/min/1.73 m2 (n=15). The final sample included 10,328 participants (8143 whites and 2185 African-Americans).
Initially, we explored the association between measures of kidney function and AF risk modeling eGFR and ACR using restricted cubic splines. Then, we categorized individuals according to their eGFR following the National Kidney Foundation guidelines: ≥90 mL/min/1.73 m2 (normal), 60–89 mL/min/1.73 m2 (mildly decreased kidney function), 30–59 mL/min/1.73 m2 (CKD stage 3), 15–29 mL/min/1.73 m2 (CKD stage 4).27 Separate categorization was done using eGFRcreat and eGFRcys. ARIC participants were also categorized according to their ACR levels: normal (<30 mg/g), microalbuminuria (30–299 mg/g), and macroalbuminuria (≥300 mg/g). P-values for trend were calculated including measures of kidney function or albuminuria as continuous variables in the models.
We estimated the association of eGFR and ACR levels with the incidence of AF using Cox proportional hazards models with time to AF as the main outcome variable. Follow-up time was defined as the time elapsed between visit 4 and date of AF incidence, death, lost to follow-up or December 31, 2007, whichever came first. Initial models were adjusted for age, gender, and race. In a second model we additionally adjusted for study site, education (did not complete high school, high school diploma, at least some college), income (<$16,000, $16,000–<$35,000, $35,000–<$50,000, ≥$50,000), height (continuous), smoking (never, former, current), alcohol drinking (never, former, current), diabetes (dichotomous), systolic blood pressure (continuous), use of antihypertensive medications (none, angiotensin converting enzyme inhibitors (ACEI) or angiotensin receptor blockers (ARB), or other medication), body mass index (continuous), and high sensitivity C-reactive protein (continuous). In additional models, we adjusted for prevalent CHD and heart failure at visit 4, for maximum P wave duration, and ECG-defined left ventricular hypertrophy (which are correlates of echocardiographic left atrial enlargement and left ventricular hypertrophy, respectively),28 and fasting blood glucose, and for incident myocardial infarction, heart failure, or any hospitalization after visit 4 (as time-varying covariates). Adjusted AF-free survival curves were estimated using the method proposed by Zhang et al.29 To avoid associations resulting from reverse causation (AF leading to kidney disease), we conducted a sensitivity analysis excluding cases of AF identified during the first 2 years of follow-up. Finally, an additional sensitivity analysis used eGFRcreat measured at the baseline ARIC visit (1987–89) as the main exposure and AF identified in follow-up ECGs as the main outcome to ensure that our results were independent of the method of AF ascertainment. The proportional hazards assumption was tested using time interaction terms and inspection of log negative log survival curves.
We conducted stratified analyses by age, gender, race, prevalent cardiovascular disease, use of antihypertensive medications, and hypertension. Effect modification was formally tested including an interaction term between the previous variables and eGFRcys levels or ACR, in separate analyses.
Table 1 shows selected characteristics of ARIC participants at visit 4 by categories of eGFRcys and ACR. In this study sample, older individuals were more likely to have worse kidney function and higher levels of ACR. Elevated CRP, a history of cardiovascular disease, diabetes, and hypertension also were associated with markers of CKD.
During a median follow-up of 10.1 years, we identified 788 incident cases of AF. Lower levels of eGFRcys were associated with higher risk of AF, even after adjustment for multiple potential confounders (table 2). Incidence of AF in individuals with eGFRcys 15–29mL/min/1.73 m2 was approximately 3 times higher compared to those with eGFRcys in the optimal range (hazard ratio (HR) 3.2, 95% confidence interval (CI) 2.0–5.0) after adjustment for potential confounders and history of cardiovascular disease. The association of eGFRcreat and AF incidence was J-shaped, with the lowest risk in individuals with eGFRcreat 60–89 mL/min/1.73 m2 (table 2). Multivariable HR (95% CI) of AF associated with 10 mL/min/1.73 m2 lower eGFRcreat or eGFRcys were, respectively, 1.06 (1.01–1.12) and 1.16 (1.11–1.21). Results were only slightly attenuated after adjusting for incident CHD and heart failure. AF-free survival curves in whites and African-Americans by eGFRcys categories are shown in figure 1.
Similarly, presence of micro-or macroalbuminuria was associated with a higher risk of AF (table 3). Compared to those with ACR<30 mg/g, the HR (95% CI) of AF was 2.0 (1.6–2.4) in those with ACR 30–299 mg/g and 3.2 (2.3–4.5) in those with ACR ≥300 mg/g after adjustment for multiple potential confounders and history of cardiovascular disease. Multivariable HR (95% CI) associated with 100 mg/g higher ACR was 1.04 (1.03–1.05). The higher risk of AF with higher ACR and lower eGFRcys levels was cumulative, with the highest risk among study participants with low eGFRcys (15–29 mL/min/1.73 m2) and macroalbuminuria (ACR ≥300 mg/g) (figure 2). HR (95% CI) in these individuals compared with those with ACR <30 mg/g and optimal eGFRcys was 13.1 (6.0–28.6). Figure 2 also shows higher AF risk with higher ACR across all categories of eGFRcys, and higher risk with lower eGFRcys across all categories of albuminuria. No evidence of multiplicative interaction between ACR and eGFRcys was present (p for interaction=0.35).
Even though lower eGFRcys and higher ACR levels were associated with a higher risk of hospitalization (p for trend<0.0001), adjustment for incident hospitalizations before AF incidence or censoring did not materially change the associations (tables 2 and and3,3, model 4b). Similarly, results did not appreciably change after additional adjustment for maximum P wave duration, ECG-defined left ventricular hypertrophy and fasting blood glucose (tables 2 and and3,3, model 4c). In a sensitivity analysis excluding AF cases identified during the first 2 years of follow-up, we obtained similar results: HR (95% CI) of AF in those with eGFRcys 15–29 mL/min/1.73 m2 vs. ≥90 mL/min/1.73 was 2.8 (1.7–4.7), and in those with ACR ≥300 mg/g vs <30 mg/g was 2.7 (1.8–4.0)(tables 2 and and3,3, model 4d). Finally, no evidence of interaction was present in analyses stratified by age, gender, race, history of cardiovascular disease, use of antihypertensive medication, or history of hypertension (supplemental figures 1 and 2).
We conducted an additional sensitivity analysis using eGFRcreat measured at the first ARIC visit (1987–1989) as the main exposure and incident AF identified from ECGs done in the three ARIC follow-up visits (1990–1992, 1993–1995, 1996–1998) as the outcome. Among 14,839 eligible ARIC participants, 119 cases of AF were observed in follow-up ECGs. Multivariable HR (95% CI) of ECG-defined AF was 2.3 (0.9–6.0) and 1.5 (1.0–2.2) in those with eGFRcreat 15–60 mL/min/1.73 m2 and 60-<90 mL/min/1.73 m2 respectively, compared to those with eGFRcreat>90 mL/min/1.73 m2 (p for trend=0.01)(supplemental table 1). Corresponding HRs including the 1,544 incident AF cases identified through study ECGs or hospitalization and death certificate surveillance from visit 1 through the end of 2007 were 1.7 (1.3–2.3) and 1.1 (1.0–1.2)(p for trend=0.003) (supplemental table 2).
In this population-based prospective study, we found that kidney damage, manifested as micro-or macroalbuminuria, and decreased kidney function were associated with a higher AF risk. An elevated risk of AF was observed even among individuals with mildly decreased kidney function measured by eGFRcys. These associations were independent of lifestyles, clinical factors, and cardiovascular disease, and were similar in men, women, whites, and African-Americans, and in individuals with or without a history of cardiovascular disease or hypertension, and among those taking antihypertensive medications. The somewhat different results using eGFRcys and eGFRcreat are consistent with previous data suggesting that the former presents a more linear association with mortality, probably due to creatinine being lowered by muscle loss.30
Previous studies addressing the relationship of kidney function with AF risk have provided inconsistent results. Lower eGFRcreat was associated with a higher risk of AF in two studies in Japan.14, 31 However, reduced kidney function as measured by both higher cystatin C levels or reduced eGFRcreat was not associated with AF risk in the Cardiovascular Health Study, a population-based study of elderly individuals in the US.16 Potential explanations for the discrepancy with our results is the older age in the Cardiovascular Health Study (average 75 vs. 63 in the ARIC study), differences in the classification of kidney dysfunction (cystatin C quartiles in the Cardiovascular Health Study vs. clinical categories of eGFRcys in our analysis), and in AF ascertainment.16 In a subset of participants in the prospective Framingham Heart Study, ACR was not associated with AF incidence; however their analysis had limited statistical power as it only included 135 AF events.15 In contrast, cross-sectional analyses have consistently shown a higher prevalence of AF among individuals with chronic kidney disease.11–13 Our study compares favorably with previous reports in its large sample size, extended follow-up, and well characterized information on potential confounders.
CKD may increase the risk of AF through several mechanisms. Individuals with kidney dysfunction are more likely to develop hypertension and have poorer control of their blood pressure.32 The resulting expansion of the extracellular fluid might lead to left ventricular hypertrophy, poor ventricular compliance and, eventually, to atrial stretch and fibrosis, established predictors of AF.5, 33 In addition, CKD leads to pathological activation of the intrarenal RAA system.34 Compelling evidence suggests that an upregulated RAA system causes atrial fibrosis and electrical remodeling, creating the required substrate for the development of AF.35 This effect might be partly mediated through increased secretion of TGF-β1, which is pro-fibrotic.36 Moreover, involvement of the RAA system in AF pathogenesis is also suggested by the potential reduced risk of AF after use of ACEIs or ARBs,35, 37 though this has not been observed in all studies.38 Whether CKD affects AF risk through this mechanism merits further inquiry. Finally, CKD might cause AF through an increased risk of cardiovascular disease, including heart failure and CHD,1, 2, 5 or through sympathetic overactivity.8 Our observation that both low eGFRcys and presence of albuminuria independently increased the risk of AF hints that a variety of mechanisms underlie these associations.
Some study limitations should be noted. In our primary analysis, incident AF was mostly identified from hospitalization discharges; therefore, we could not include asymptomatic AF and AF managed exclusively in an outpatient setting. However, an analysis including only AF events identified from systematic study ECGs found an association between CKD and AF risk of similar magnitude. Moreover, we and others have previously shown that the validity of AF ascertainment using hospitalizations is acceptable,22, 39 that incidence rates of AF in the ARIC Study are consistent with other population-based studies,22 and that the associations between variants in the chromosome 4q25—extremely specific for AF risk—and AF incidence in ARIC are similar to those in other studies with a more intensive ascertainment of AF.40 Still, individuals with reduced renal function may have been more likely to have their AF detected by hospitalization than those with normal function, potentially creating ascertainment bias. Other important limitations include the absence of echocardiographic data, the potential for residual confounding by some CKD risk factors (e.g. hypertension, diabetes), the availability of only 1 measurement of ACR and cystatin C, which could lead to measurement bias, and the lack of standardization of cystatin C assays and equation. Despite some limitations, our study has important strengths: a large sample size, the elevated number of AF events, the long follow-up, the diversity of the study population, and the quality and extent of measured covariates.
In conclusion, we have found that kidney damage and impaired kidney function are associated with an increased risk of AF independently of other risk factors. Given the growing burden of CKD in the general population and the potential for its prevention,41 future studies should focus on understanding the specific mechanisms underlying this association. Furthermore, strategies for the prevention of AF will have to consider CKD as a preventable risk factor for AF in addition to other well-established risk factors.6
Previous research has shown that individuals with end-stage renal disease have a higher risk of developing atrial fibrillation, and some cross-sectional studies have found higher prevalence of atrial fibrillation among those with decreased kidney function. However, evidence from prospective studies in the general population is limited. In an analysis of 10,328 men and women participating in the Atherosclerosis Risk in Communities Study, we observed that impaired kidney function, measured by lower cystatin-based or creatinine-based estimated glomerular filtration rate, was strongly associated with a higher risk of atrial fibrillation. Similarly, individuals with increased levels of urinary albumin-creatinine ratio, a marker of kidney damage, had higher risk of developing atrial fibrillation. Our study highlights the potential role of chronic kidney disease as a risk factor for atrial fibrillation. Interventions aimed to preventing and treating chronic kidney disease could also contribute to reduce the burden of atrial fibrillation in the population.
The authors thank the staff and participants of the ARIC study for their important contributions.
The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts N01-HC55015, N01-HC55016, N01-HC55018, N01-HC55019, N01-HC55020, N01-HC55021, and N01-HC55022. This study was additionally supported by grants RC1-HL099452 from the National Heart, Lung, and Blood Institute and 09SDG2280087 from the American Heart Association. C-reactive protein assays were supported by the National Institute of Diabetes and Digestive and Kidney Diseases (R01-DK076770). Siemens Healthcare Diagnostics provided the reagents and loan of a BNII instrument to conduct these assays.
Alvaro Alonso, Brad Astor, and Josef Coresh have received significant funding from the National Institutes of Health. Alvaro Alonso and Lin Chen have received significant funding from the American Heart Association.