Abstract
Background and objectives: Chronic kidney disease (CKD) increases systemic inflammation, which is implicated in development and maintenance of atrial fibrillation (AF); therefore, we hypothesized that the prevalence of AF would be increased among nondialysis patients with CKD. This study also reports independent predictors of the presence of AF in this population.
Design, setting, participants, & measurements: A retrospective, cross-sectional analysis of 1010 consecutive nondialysis patients with CKD from two community-based hospitals was conducted. Estimated GFRs (eGFRs) were calculated using the Modification of Diet in Renal Disease (MDRD) equation. Multivariate logistic regression was used to determine independent predictors.
Results: Of 1010 nondialysis patients with CKD, 214 (21.2%) had AF. Patients with AF were older than patients without AF (76 ± 11 versus 63 ± 15 yr). The prevalence of AF among white patients (42.7%) was higher than among black patients (12.7%) or other races (5.7%). In multivariate analyses, age, white race, increasing left atrial diameter, lower systolic BP, and congestive heart failure were identified as independent predictors of the presence of AF. Although serum high-sensitivity C-reactive protein levels were elevated in our population (5.2 ± 7.4 mg/L), levels did not correlate with the presence of AF or with eGFR. Finally, eGFR did not correlate with the presence of AF in our population.
Conclusions: The prevalence of AF was increased in our population, and independent predictors were age, white race, increasing left atrial diameter, lower systolic BP, and congestive heart failure.
Atrial fibrillation (AF) is the most common arrhythmia in clinical practice (1). Cardiac comorbidities that are associated with AF include hypertension, coronary artery disease (CAD), valvular heart disease (VHD), congestive heart failure (CHF), cardiomyopathy, pericarditis, congenital heart disease (CHD), and cardiac surgery (2–9). Noncardiac comorbidities that are associated with AF include acute pulmonary embolism, chronic obstructive pulmonary disease (COPD), obstructive sleep apnea, hyperthyroidism, and obesity (10–14).
Evidence suggests that inflammation is involved in the pathogenesis of AF (15–20). For example, AF after cardiac surgery is associated with proinflammatory cytokine and complement activation (16,19). Moreover, patients with refractory lone AF have inflammatory infiltrates, myocyte necrosis, and fibrosis on biopsy (18). Several studies also reported elevated serum high-sensitivity C-reactive protein (hsCRP) levels in patients with AF (15–17,20).
Evidence suggests that inflammation is associated with renal dysfunction (21–24). Proposed mechanisms include decreased proinflammatory cytokine clearance, endotoxemia, oxidative stress, and reduced antioxidant levels (23,24). Moreover, hsCRP levels are higher among elderly patients with renal insufficiency (24). In hemodialysis (HD) patients with ESRD, hsCRP, IL-6, and fibrinogen levels are elevated (21,22).
HD patients with ESRD have an increased prevalence of AF; however, prevalence among nondialysis patients with CKD has not been investigated (25–30). Because CKD promotes inflammation, which promotes AF, we hypothesized the prevalence of AF would be increased among nondialysis patients with CKD. This study reports the prevalence and independent predictors of the presence of AF in a nondialysis population with CKD.
Materials and Methods
We conducted a retrospective, cross-sectional analysis of consecutive inpatients and outpatients at two community-based teaching hospitals between January and July 2008. Patients had CKD as defined by the Kidney Disease Outcomes Quality Initiative (K/DOQI): (1) Evidence of structural or functional kidney damage for ≥3 mo, with or without decreased GFR, manifest by markers of kidney damage (including blood, urine, and imaging abnormalities) or (2) GFR <60 ml/min per 1.73 m2 for ≥3 mo with or without evidence of kidney damage (31).
CKD was stratified as stage 1 (kidney damage, GFR ≥90), stage 2 (GFR 60 to 89), stage 3 (GFR 30 to 59), stage 4 (GFR 15 to 29), or stage 5 (kidney failure, GFR <15 or dialysis) (31). Patients who were in acute renal failure, in a postoperative period, or receiving dialysis were excluded. Moreover, we hypothesized that decreased GFR was associated with inflammation; therefore, patients with stage 1 CKD were also excluded.
Estimated GFR (eGFR) was calculated using the Modification of Diet in Renal Disease (MDRD) equation: eGFR (ml/min per 1.73 m2) = 1.86 × serum creatinine−1.154 × age−0.203 × 0.742 (if female) × 1.210 (if black) (31,32). Patients with AF were identified on the basis of medical record documentation and/or electrocardiographic evidence and classified as paroxysmal, persistent, or permanent. Patients with hypertension were identified on the basis of medical record documentation, and mean systolic (SBP) and diastolic BP (DBP) were systematically calculated from five measurements. Other comorbid conditions were identified on the basis of medical record documentation. Left ventricular (LV) hypertrophy (LVH) was defined as intraventricular septum and/or posterior wall thickness >11 mm. Left atrial (LA) diameter and LV ejection fraction (LVEF) were derived from echocardiographic data. LV systolic dysfunction was defined as LVEF <50%. Similar to a previous study, VHD likely to be associated with AF was defined as any degree of mitral stenosis or moderate to severe mitral regurgitation, aortic stenosis, or aortic regurgitation (26).
Data from patients with and without AF were compared using χ2 and Wilcoxon rank-sum analyses. Univariate linear regression examined the relationship between eGFR and hsCRP levels. Univariate logistic regression identified variables that were associated with the presence of AF, and those with P < 0.1 were included in multivariate analysis using a backward stepwise logistic regression model with a stay criterion of 0.10. A multiplicative model including age-race interaction terms adjusted for significant variables estimated the effect of age (stratified as <50, 50 to 59, 60 to 69, 70 to 79, and ≥80 yr of age) and race (white versus nonwhite) on the prevalence of AF. Odds ratios and 95% confidence intervals were calculated using nonwhite patients who were younger than 50 yr (lowest prevalence of AF) as the denominator. P < 0.05 was considered statistically significant. Statistics were calculated using Stata statistics software (Stata Corp., College Station, TX).
Results
Prevalence of AF
Of 1010 nondialysis patients with CKD, 214 (21.2%) had AF, classified as permanent (38.8%), persistent (18.2%), or paroxysmal (43.0%). The prevalence of AF stratified by age, gender, and race is summarized in Table 1. When stratified by age, the prevalence was 8.1% among those who were younger than 65, 31.6% among those who were aged ≥65, and 45.8% among those who were aged ≥80 yr. When stratified by gender, the prevalence was similar among men (20.6%) and women (21.8%). When stratified by race, the prevalence was 42.7% among white patients, 12.7% among black patients, and 5.7% among other races. When stratified by age and race, the prevalence of AF increased with age, irrespective of race, and was higher among white patients of a given age group (Figure 1). Moreover, the adjusted odds ratio was highest among white patients who were aged ≥80 yr and least among nonwhite patients who were younger than 50 yr (Table 2). When patients with known risk factors for development of AF (e.g., CAD, VHD, CHF, COPD, hyperthyroidism) were excluded from analysis, the prevalence was 6.3% but still increased with age with 3.4% among those who were younger than 65, 10.4% among those who were aged ≥65, and 18.5% among those who were aged ≥80 yr. When stratified by CKD stage, the prevalences were 17.9, 25.2, 20.8, and 8.0% for stages 2 through 5, respectively (Table 1).
Prevalence of atrial fibrillation in nondialysis patients with CKD
Prevalence of atrial fibrillation among nondialysis patients with CKD stratified by age and race.
Effect of age and race on the presence of AF among non-dialysis CKD patients
Demographic and Clinical Characteristics
Demographic and clinical characteristics of nondialysis patients who had CKD with and without AF are summarized in Table 3. Patients with AF were on average older than patients without AF (76 ± 11 versus 63 ± 15 yr; P < 0.001). The proportions of male and female patients with AF (50.0% male versus 50.0% female) and patients without AF (51.9% male versus 48.1% female) were similar. Moreover, the proportion of white race was higher among patients with than without AF (66.8 versus 24.1%; P < 0.001), the proportion of black race was lower among patients with than without AF (27.6 versus 51.1%; P < 0.001), and the proportion of other races was also lower among patients with than without AF (5.6 versus 24.8%; P < 0.001).
Demographic and clinical characteristics of nondialysis patients who had CKD with and without AF
Overall, diabetic nephropathy was the most common known cause of CKD in our nondialysis population with CKD (46.0%), followed by hypertensive nephrosclerosis (25.4%). Of known causes of CKD, only ischemic nephropathy occurred more frequently among patients with AF. Comorbidities that occurred more frequently among patients with AF included diabetes, CHF, CAD, VHD, peripheral vascular disease (PVD), cerebrovascular accident (CVA), COPD, and hyperthyroidism (Table 3). Patients with AF also had lower SBP and DBP measurements than patients without AF (127/67 ± 17/10 versus 138/73 ± 19/11 mmHg; P < 0.001 each). AF patients had higher eGFRs than patients without AF (36.5 ± 13.9 versus 33.4 ± 16.5 ml/min per 1.73 m2; P = 0.004). Finally, patients with AF were treated more frequently with angiotensin-converting enzyme inhibitors (ACEIs) and/or angiotensin receptor blockers (ARBs). Patients with AF also tended to be treated more frequently with β blockers, whereas statin treatment was similar in both groups.
Echocardiographic and Laboratory Data
Echocardiographic data were obtained from 621 of 1010 nondialysis patients with CKD (Table 4). Patients with AF had lower LVEFs (50.7 ± 15.6 versus 56.8 ± 13.6%; P < 0.001), increased frequency of LV systolic dysfunction (37.2 versus 20.0%; P < 0.001), increased LA diameter (46.4 ± 25.4 versus 40.8 ± 6.5 mm; P < 0.001), and increased frequency of VHD (26.6 versus 6.0%; P < 0.001) than patients without AF; however, there was no difference in frequency of LVH or pulmonary artery systolic pressure between groups.
Echocardiographic data of nondialyhsis patients who have CKD with and without AF
Laboratory data were obtained from all nondialysis patients with CKD (Table 5). Patients with AF had lower serum potassium, calcium, phosphorus, creatinine, albumin, cholesterol, and triglyceride levels and higher serum bicarbonate levels. Levels of hsCRP were obtained from 76 of 1010 nondialysis patients with CKD. Although data were limited, average hsCRP levels were elevated above the reference value (<3.0 mg/L) of our nondialysis population with CKD. Moreover, levels tended to be lower in patients with than without AF (4.3 ± 5.7 versus 5.7 ± 8.2 mg/dl; P = 0.420), although not statistically significant. Finally, to examine the potential relationship between impaired renal function and inflammation, we compared eGFRs with hsCRP levels; however, there was no correlation in our population (Figure 2).
Laboratory data of nondialysis patients who had CKD with and without AF
eGFR does not correlate with hsCRP levels in nondialysis patients with CKD. The correlation coefficient is R = −0.36, P = 0.757.
Independent Predictors of AF
Clinical, echocardiographic, and laboratory variables that were associated with the presence of AF in our nondialysis population with CKD identified by univariate logistic regression analyses are summarized in Table 6. Significant variables that were positively associated with AF included age, white race, dyslipidemia, CHF, CAD, PVD, CVA, COPD, hyperthyroidism, increasing LA diameter, VHD, and eGFR. Significant variables that were negatively associated with AF included diabetes, mean SBP and DBP, LVEF, serum potassium, calcium, phosphorus, and albumin levels. Multivariate analysis of significant variables that were identified by univariate logistic regression analyses identified age, white race, increasing LA diameter, lower SBP, and CHF as independent predictors of the presence of AF in our population (Table 7).
Univariate logistic regression analyses for the presence of AF among nondialysis patients with CKD
Multivariate logistic regression analyses for the presence of AF among nondialysis patients with CKD
Discussion
Although the prevalence of AF was increased in our nondialysis population with CKD, we did not find an association between AF and inflammatory biomarkers or eGFR. The prevalence of AF in our population (21.2%) was greater than estimates in the general population (1.5 to 6.2%) (1,33–36). The prevalence of AF increased with age and was highest among those who were aged ≥80 yr. Patients with AF were older than patients without AF, and the prevalence among patients who were aged ≥65 yr (31.6%) was greater than estimates for the same age group in the general population (5.9%). Studies that estimated the prevalence of AF in HD patients with ESRD (5.4 to 27.0%) vary likely because of different enrollment criteria (25–29). For example, in one HD population, the prevalence of AF was 27.0%, whereas in another that excluded rheumatic VHD and paroxysmal AF the prevalence was 13.6%, whereas in another that included only permanent AF the prevalence was 5.4% (26,27,29). For comparison with our population, we calculated that, without exclusions, the prevalence of AF was 21.2% (versus 27%); when VHD and paroxysmal AF were excluded, it was 10.1% (versus 13.6%); and when only permanent AF was included, it was 8.2% (versus 5.4%). Therefore, the prevalence of AF in our population (21.2%) is at least triple that reported for the general population (1.5 to 6.2%) and within the broad range reported among various HD populations with ESRD (5.4 to 27.0%).
In Framingham Heart Study patients, the prevalence of AF in the general population was higher among men than women (7:1 ratio) (37,38). In our nondialysis population with CKD, the prevalences of AF among male (20.6%) and female (21.8%) patients were similar, as were the proportions of male and female patients with and without AF. With respect to race, studies that estimated the prevalence of AF vary considerably (34,35,38–41). A higher electrocardiographic prevalence of AF was reported in white (7.8%) compared with black (2.5%) hospitalized patients (41). In another study, a higher prevalence of heart failure–associated AF was reported in white (38.3%) compared with black (19.7%) patients (40). Similarly, the prevalence of AF in our population was higher in white (42.7%) compared with black (12.7%) patients or patients of other races (5.7%). These racial differences may be due to genetic polymorphisms that code for intrinsic differences in atrial membrane stability and/or conduction pathways, resulting in different susceptibilities to development of AF (40). When stratified by age and race, the prevalence of AF increased with age, irrespective of race (Figure 1). It is interesting that, when stratified by CKD stage, there was no notable trend.
Clearly, our nondialysis population with CKD is elderly; has a high prevalence of atherosclerotic, diabetic, and hypertensive disease; and is more prone to inflammatory influences and development of AF. Among our population, we found that comorbidities including diabetes, CHF, CAD, VHD, PVD, CVA, COPD, and hyperthyroidism occurred more frequently among patients with AF. Not surprising, diabetic nephropathy (46.0%) and hypertensive nephrosclerosis (25.4%) were common in our nondialysis population with CKD; however, only ischemic nephropathy occurred more frequently among patients with AF. Diabetes was negatively associated with AF by univariate analysis, possibly because of higher frequency of ACEI and/or ARB use among patients with AF. The higher frequency of ACEI and/or ARB use among patients with AF may also reflect their use in treatment of CHF (42,43). Moreover, ACEIs and/or ARBs have been shown to prevent AF, especially among those with systolic LV dysfunction or LVH.
Among our nondialysis population with CKD, echocardiographic data revealed that patients with AF have significantly lower LVEF, increased LA diameter, and increased frequencies of VHD and LV systolic dysfunction; however, there was no difference in frequency of LVH or pulmonary artery SBP between groups. These findings are partially consistent with a study that reported that LVEF and LVH were associated with AF (44). Laboratory data revealed that patients with AF have lower serum potassium, calcium, phosphorus, creatinine, albumin, cholesterol, and triglyceride levels and higher serum bicarbonate levels. We are not aware of any studies with similar data; however, although the mean concentrations of the proarrhythmogenic electrolytes calcium and potassium were different between groups, they were within the range of normality and therefore likely not related to AF in our population.
We hypothesized that the prevalence of AF would be increased among nondialysis patients with CKD because CKD promotes inflammation, which promotes AF. We reasoned that hsCRP levels might be elevated and associated with decreased eGFR in this population. Although hsCRP levels were elevated, there was no association with eGFR (Figure 2). The extent to which renal dysfunction estimated by GFR is related to inflammatory biomarkers is controversial. Some studies reported hsCRP levels are elevated and increased with progression of CKD, whereas others reported no correlation (24,45,46). Several studies also reported an association between elevated hsCRP levels and AF and that higher baseline levels may predict development of AF (15–17,20). Although hsCRP levels were elevated in our nondialysis population with CKD, comparison between patients with and without AF proved difficult because of considerable variation in serum levels. These findings question the utility of hsCRP as an indicator of inflammation in nondialysis patients with CKD and its relevance to AF.
In our nondialysis population with CKD, multivariate analysis found that age and white race are independent predictors of the presence of AF. Moreover, the prevalence of AF was higher among white patients of a given age group, increased with each decade, and was highest among patients who were aged ≥80 yr. That age and white race are independent predictors of the presence of AF is not surprising given the increasing prevalence with age among white patients. CHF and increasing LA diameter were also independent predictors of the presence of AF in our population. Similarly, others have reported that CHF and increasing LA diameter are risk factors for developing AF (7,37,39,44). Overall, our study suggests that the high prevalence of AF in our nondialysis population with CKD may be due to the presence of numerous cardiovascular comorbidities rather than reduced GFR.
Whereas hypertension is associated with AF in the general population, this is not necessarily the case among HD populations with ESRD (7,26,27,37,39). In our nondialysis population with CKD, hypertension was negatively associated with and not a predictor of AF. Rather, lower SBP was an independent predictor of AF in our population. The prevalence of hypertension was >90% in our population, leaving relatively few normotensive patients for comparison, perhaps contributing to these findings. Moreover, although the association between lower SBP and AF is difficult to interpret, it is likely not due to cardiac inefficiency, because mean LVEF, although statistically different between groups, was within the normal range in both groups. As expected, patients with AF had increased prevalence of cardiovascular comorbidities, including CHF and CAD, and accordingly received more evidence-based medications including ACEIs and ARBs; however, β blocker use was similar between groups and likely noncontributory (43,47,48). Overall, the potential relationship between BP and AF among nondialysis patients with CKD should be further examined in longitudinal studies.
CAD and VHD are associated with development of AF in population-based studies (7,37–39); however, in our population, they failed to reach significance by multivariate analysis even when LA diameter was integrated into our statistical model. Perhaps CAD and VHD influence development of AF by mechanisms other than increased LA diameter. Similarly, diabetes and hyperthyroidism are associated with development of AF; however, they failed to reach significance by multivariate analysis (7,37). Moreover, COPD was not an independent predictor of AF in our nondialysis population with CKD, similar to one study but contrary to another (37,39). Finally, decreased LVEF and LVH have been reported as predictors of AF but failed to reach significance in our nondialysis population with CKD (44).
With respect to limitations, this study was designed to determine independent predictors of the presence of AF in a nondialysis population with CKD and does not make comparisons with a control population with normal renal function. Also, CKD and AF are chronic illnesses, often with unidentifiable times of onset. The extent to which the prevalence of AF in our population can be attributed to CKD is also not clear because other comorbidities likely contribute. Moreover, the retrospective design does not allow determination of cause-and-effect relationships. We can only describe the prevalence, demographic and clinical characteristics, and identify independent predictors of the presence of AF in this population. Moreover, hsCRP samples were collected irrespective of coexisting medical conditions and may not entirely reflect inflammatory status with respect to renal dysfunction. This limits conclusions that can be drawn concerning hsCRP and inflammation and its association with AF in this population. We should also note that increasing LA diameter does not necessarily reflect LVH. Finally, our nondialysis population with CKD includes a substantial number of inpatients who typically have a higher prevalence of AF and a greater number of comorbidities than the general population; therefore, caution must be exercised when making comparisons with the general population. Larger multicenter, prospective studies would be ideal to clarify the relationship among renal dysfunction, inflammation, and AF.
Conclusions
We observed a high prevalence of AF in our nondialysis population with CKD, and age, white race, increasing LA diameter, lower SBP, and CHF were identified as independent predictors of the presence of AF. Notably, hsCRP levels were elevated in our population; however, levels did not correlate with the presence of AF or the degree of renal dysfunction estimated by GFR. Finally, eGFR did not correlate with the presence of AF in our population.
Disclosures
None.
Acknowledgments
We thank Dr. Joseph Oyama for reviewing the manuscript.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
Access to UpToDate on-line is available for additional clinical information at http://www.cjasn.org/
- Received May 12, 2009.
- Accepted October 28, 2009.
- Copyright © 2010 by the American Society of Nephrology