Author + information
- Received March 31, 2015
- Accepted April 23, 2015
- Published online August 1, 2015.
- Thomas A. Dewland, MD∗,†,
- Eric Vittinghoff, PhD, MPH‡,
- Tamara B. Harris, MD, PhD§,
- Jared W. Magnani, MD, MSc‖,
- Yongmei Liu, MD, PhD¶,
- Fang-Chi Hsu, PhD#,
- Suzanne Satterfield, MD, DrPH∗∗,
- Christina Wassel, PhD††,
- Gregory M. Marcus, MD, MAS∗∗ (, )
- Health ABC Study Investigators
- ∗Electrophysiology Section, Division of Cardiology, Department of Medicine, University of California, San Francisco, California
- †Knight Cardiovascular Institute, Oregon Health and Science University, Portland, Oregon
- ‡Department of Epidemiology and Biostatistics, University of California, San Francisco, San Francisco, California
- §Laboratory of Epidemiology and Population Sciences, National Institute on Aging, Bethesda, Maryland
- ‖Department of Medicine, Section of Cardiovascular Medicine, Boston University School of Medicine, Boston, Massachusetts
- ¶Division of Public Health Sciences, Wake Forest School of Medicine, Winston-Salem, North Carolina
- #Department of Biostatistical Sciences, Wake Forest School of Medicine, Winston-Salem, North Carolina
- ∗∗Department of Preventive Medicine, University of Tennessee Health Science Center, Memphis, Tennessee
- ††Department of Epidemiology, University of Pittsburgh, Pittsburgh, Pennsylvania
- ↵∗Reprint requests and correspondence:
Dr. Gregory M. Marcus, Electrophysiology Section, Division of Cardiology, Department of Medicine, University of California, San Francisco, 505 Parnassus Avenue, M-1180B, Box 0124, San Francisco, California 94143-0124.
Objectives This study sought to determine the degree to which racial differences in atrial fibrillation (AF) risk are explained by differences in inflammation and adiposity.
Background Despite having a lower prevalence of established AF risk factors, whites exhibit substantially higher rates of this arrhythmia than blacks. The mechanism underlying this observation is not known. Both inflammation and obesity are risk factors for AF, and adipose tissue is a known contributor to systemic inflammation.
Methods Baseline serum inflammatory biomarker concentrations and abdominal adiposity (assessed by computed tomography) were quantified in a subset of black and white participants without prevalent AF in the Health ABC (Health, Aging, and Body Composition) study. Participants were prospectively followed for the diagnosis of AF, using study electrocardiography and Medicare claims data. Cox proportional hazards models were used to determine the adjusted relative hazard of incident AF between races before and after biomarker adjustment.
Results Among 2,768 participants (43% black), 721 participants developed incident AF over a median follow-up of 10.9 years. White race was associated with a heightened adjusted risk of incident AF (HR: 1.55; 95% confidence interval [CI]: 1.30 to 1.84, p < 0.001). Abdominal adiposity was not associated with AF when added to the adjusted model. Among the biomarkers studied, adiponectin, tumor necrosis factor (TNF)-α, TNF-α soluble receptor (SR) I, and TNF-α SR II concentrations were each higher among whites and independently associated with a greater risk of incident AF. Together, these inflammatory cytokines mediated 42% (95% CI: 15% to 119%, p = 0.004) of the adjusted association between race and AF.
Conclusions Systemic inflammatory pathways significantly mediate the heightened risk of AF among whites. The higher level of systemic inflammation and concomitant increased AF risk in whites is not explained by racial differences in abdominal adiposity or the presence of other proinflammatory cardiovascular comorbidities.
Despite having a lower prevalence of traditional atrial fibrillation (AF) risk factors, whites exhibit substantially higher rates of this arrhythmia than blacks (1,2). A better understanding of the intermediary factors responsible for this association could yield important insight into the pathogenesis of the most commonly encountered cardiac arrhythmia.
Inflammation and its downstream effects, including atrial fibrosis, are thought to play a central role in AF development and perpetuation (3,4). Systemic inflammation can be quantified through measurement of serum inflammatory cytokine levels, and previous investigations have associated concentrations of C-reactive protein (CRP) (5–7) and interleukin (IL)-6 (8) with clinical AF risk. Many of these inflammatory markers are secreted or regulated by adipose tissue (9), and a growing body of research has linked obesity with heightened AF susceptibility (10–12). Notably, racial differences in both inflammation and adiposity have also been reported (13–16). Furthermore, whites tend to have greater amounts of visceral abdominal adiposity than blacks (17), providing a potential mechanistic explanation for differences in inflammation by race.
In light of these associations, we hypothesized that racial differences in inflammation, secondary to racial variation in abdominal adiposity, explain the substantially divergent risk of AF between whites and blacks. The Health ABC (Health, Aging, and Body Composition) study (18) was therefore utilized to determine the degree to which inflammatory cytokines and adiposity mediate the association between race and AF.
The Health ABC study design
Health ABC was a population-based cohort study sponsored by the National Institute on Aging. Eligibility, enrollment, and follow-up protocols were previously published (19). Briefly, 3,075 white or black individuals 70 to 79 years of age were recruited between 1997 and 1998 from a random sample of Medicare beneficiaries residing in 2 urban areas (Pittsburgh, Pennsylvania, and Memphis, Tennessee). After participants underwent a baseline medical history and physical examination, laboratory assessment, and electrocardiography (ECG), they were followed with yearly clinic visits and interim telephone contact every 6 months. All participants provided written, informed consent upon study enrollment.
Individuals with prevalent AF were excluded from the overall Health ABC cohort. Participants actively receiving chemotherapy or taking oral steroids were also excluded, as we reasoned that the inflammatory conditions underlying or induced by such treatments could confound the association between race and inflammatory markers. Individual inflammatory cytokine and abdominal adiposity analyses were restricted to participants with available baseline biomarker measurements.
Inflammatory marker measurement
At the baseline study visit, participants underwent morning venipuncture after an overnight fast. Serum samples were frozen at −70°C and stored in a core laboratory at the University of Vermont (Burlington, Vermont). A total of 9 inflammatory markers were measured as part of the Health ABC protocol. Because the relative importance of these candidate inflammatory cytokines in AF risk prediction has not been well established, we considered all biomarkers in our initial analysis. Quantification of adiponectin, IL-6, tumor necrosis factor (TNF)-α, CRP, and plasminogen activator inhibitor (PAI)-1 was attempted in all participants, whereas soluble receptor concentrations (IL-2 soluble receptor [SR], IL-6 SR, TNF-α SR I, and TNF-α SR II) were measured in a subsample of 1,367 individuals. IL-6, TNF-α, IL-2 SR, IL-6 SR, TNF-α SR I, and TNF-α SR II were measured in duplicate, using enzyme-linked immunosorbent assay (ELISA) kits (R&D Systems, Minneapolis, Minnesota). CRP levels also were measured in duplicate by ELISA based on purified protein and polyclonal anti-CRP antibodies (Calbiochem, San Diego, California). PAI-1 was measured from citrated plasma, using a 2-site ELISA (Center of Molecular and Vascular Biology, University of Leuven, Belgium), whereas adiponectin was measured in duplicate by radioimmunoassay (Linco Research, St. Charles, Missouri). Assay detectable limits and performance characteristics have been previously described (20).
Abdominal adiposity measurements
Abdominal subcutaneous and visceral adiposity were quantified for each participant by using computed tomography (CT) images obtained in the axial plane at the level of L4–L5 at 20 kV, 200 to 250 mA, and 10-mm slice thickness (21). Participants enrolled at the Memphis site were imaged with a Somatom Plus 4 (Siemens, Erlangen, Germany) or a Picker PQ 2000S scanner (Marconi Medical Systems, Cleveland, Ohio), whereas the Pittsburgh center used a 9800 Advantage scanner (General Electric, Milwaukee, Wisconsin). Images were analyzed in a core laboratory. The fascial plane of the internal abdominal wall was manually identified on each CT image and used to differentiate visceral from subcutaneous fat. Adiposity area was quantified by multiplying the adipose soft tissue pixel count by pixel area (Interactive Data Language software, ITT Visualization Solutions, Boulder, Colorado).
Race, prevalent health conditions, medication usage, tobacco history, and alcohol consumption were self-reported by participants at baseline, using an interview-administered standardized questionnaire. Alcohol use was dichotomized as <1 or ≥1 drink/week. Diabetes was present if the participant reported a history of diabetes or was taking an antihyperglycemic medication. Coronary artery disease was defined as a history of angina, myocardial infarction, percutaneous coronary intervention, or coronary artery bypass surgery. Height was assessed barefoot with a wall-mounted stadiometer to the nearest 0.1 cm. Weight was quantified using a balance beam scale to the nearest 0.1 kg. Body mass index (BMI) was calculated by dividing weight in kilograms by height in meters squared. Blood pressure was obtained in a seated position and determined by averaging 2 successive measurements. Hypertension was defined as a systolic blood pressure of ≥140 mm Hg, a diastolic blood pressure of ≥90 mm Hg, or a participant’s reported history of hypertension with concomitant antihypertensive use.
Participants were linked to Centers for Medicare Services databases, and AF was identified using International Classification of Diseases-9th edition (ICD-9) code 427.31, recorded during an inpatient hospitalization or ambulatory health care encounter. Prevalent AF was present if pre-enrollment Medicare claims data revealed AF ICD-9 coding (between 1992 and study enrollment) or if the baseline study ECG demonstrated AF. Incident AF was determined using Medicare claim ICD-9 data or the study year 4 ECG. ECGs were analyzed using the Minnesota Coding system and visually inspected for accuracy in a core laboratory at St. Louis University Medical Center (St. Louis, Missouri).
Continuous variables with a normal distribution are presented as mean ± SD and were compared using Student t tests. Non-normally distributed continuous variables are presented as medians with interquartile ranges (IQRs) and were compared using Kruskal-Wallis tests. The association between categorical variables was determined using chi-square tests. For interpretability and to ensure log-linearity, we used log base 2 transformations of sera cytokine concentrations. Successful normalization of the log-transformed variables was subjectively assessed by comparing histogram and Q-Q plots before and after transformation. Cox proportional hazards models were used to determine the association between race and AF both before and after controlling for confounders identified a priori. Participant enrollment site (Memphis versus Pittsburgh) was included in all adjusted models. Hazard ratios for the association between cytokines and AF can be interpreted as the increased hazard for each doubling in cytokine concentration.
The extent to which inflammation mediates the race–AF association was first assessed using an average causal mediation effect methodology (22). Briefly, adjusted pooled logistic regression with annual time periods was used to calculate the percentage of total effect mediated by each inflammatory cytokine. To estimate the overall inflammatory mediation effect, all sera markers that significantly mediated the race–AF association in the above models (p < 0.05) were added to a single adjusted model, and the percent treatment effect methodology (23) was used to facilitate the inclusion of multiple mediators. Confidence intervals (CI) for mediation statistics were obtained using bootstrap resampling with 1,000 repetitions.
Data were analyzed using Stata version 12 software (StataCorp, College Station, Texas). A 2-tailed p value of <0.05 was considered statistically significant. All participants provided written informed consent upon enrollment. Certification to use deidentified Health ABC data was obtained from the University of California, San Francisco Committee on Human Research.
Among the 3,075 Health ABC participants, individuals with prevalent AF (n = 211), those receiving chemotherapy (n = 35), and those treated with oral steroids (n = 61) were excluded. The remaining cohort consisted of 1,179 black adults (43%) and 1,589 white adults (57%). Black participants were more likely to be female, less frequently consumed alcohol, had a greater mean BMI, and had a higher prevalence of medical comorbidities including hypertension and diabetes (Table 1).
Race and AF
Over a median 10.9 years of follow-up, incident AF was diagnosed in 721 participants. In bivariate analyses, whites demonstrated a significantly increased risk of incident AF compared with blacks (Table 2). After controlling for the known AF risk factors in Table 2, white race remained associated with a 55% increase in AF risk (hazard ratio [HR]: 1.55, 95% CI: 1.30 to 1.84, p < 0.001). Nearly all a priori risk factors were associated with AF in both bivariate and multivariate models (Table 2). Notably, the association between diabetes and AF was of borderline statistical significance in both models, BMI was not significantly associated with increased AF risk, and statin therapy did not appear to have a definitive protective effect. Study site was not associated with incident AF in either bivariate or multivariate model.
Race, inflammatory cytokines, and AF
White participants demonstrated significantly elevated serum concentrations of adiponectin, IL-6 SR, IL-2 SR, TNF-α, TNF-α SR I, and TNF-α SR II compared with blacks (Figure 1). Higher inflammatory cytokine concentrations were associated with increased AF risk after controlling for established AF risk factors, although this association was of borderline significance for IL-2 SR and did not meet statistical significance for IL-6 SR and PAI-1 (Table 3).
To be considered a potential mediator of the race–AF association, a candidate cytokine was required to have a significantly higher concentration among whites and a significant association with AF after adjustment for race and other risk factors. Adiponectin, TNF-α, TNF-α SR I, and TNF-α SR II each met these criteria. When these biomarkers were individually added to a multivariate model containing the known AF risk factors in Table 2, significant mediation of the race–AF association was observed for each biomarker (Figure 2). When these 4 cytokines were simultaneously included in the same multivariate race–AF model, the proportion of mediated association between race and AF (i.e., the proportion of the race–AF relationship explained by racial differences in cytokine concentration) was 42.2% (95% CI: 15.2% to 118.9%, p = 0.004).
Race, adiposity, and AF
Among the 2,768 participants included in incident AF analyses, 2,664 participants (96%) and 2,581 participants (93%) had adequate visceral and subcutaneous CT adiposity measurements, respectively. Whites demonstrated significantly higher mean abdominal visceral adiposity area than blacks (153 ± 70 cm2 vs. 130 ± 62 cm2, respectively, p < 0.001), whereas blacks had a significantly higher mean subcutaneous fat area (314 ± 139 cm2 vs. 267 ± 103 cm2, respectively, p < 0.001). In bivariate analyses, each 10 cm2 increase in visceral abdominal fat area was associated with a 2% increased risk of incident AF (HR: 1.02, 95% CI: 1.01 to 1.03, p < 0.001). Subcutaneous fat area, on the other hand, was not significantly associated with incident AF (HR: 0.99 for each 10-cm2 increase in area, 95% CI: 0.99 to 1.00, p = 0.078). In multivariate models adjusting for the known AF risk factors listed in Table 2, neither visceral nor subcutaneous abdominal fat area was significantly associated with AF. Because adiposity measurements were not independently associated with AF risk, further mediation analyses incorporating these variables were not performed.
In a large, well-characterized, population-based sample of older black and white adults, race was associated with significant differences in AF risk, abdominal adiposity, and serum inflammatory marker concentration. Although abdominal adiposity was not associated with incident AF after controlling for other established risk factors, concentrations of adiponectin, TNF-α, TNF-α SR I, and TNF-α SR II were each higher among whites and associated with a greater adjusted risk of incident AF. Attenuation of the race–AF association after controlling for these cytokines indicates that approximately 40% of the elevated risk of AF among whites may be attributed to racial differences in inflammation.
The relationship between sera inflammatory cytokines and AF risk has been previously described in multiple settings. Both CRP (5–7) and IL-6 (8) are independently associated with AF, even after adjustment for established AF risk factors. Although a significant relationship between adiponectin and AF was not observed in the Framingham Offspring study (24), a more recent investigation from the Busselton Health study did identify a significant association (25). The association between TNF-α and AF was previously described using only a case-control study design (26), and our current investigation represents the first description of the association between this biomarker and incident AF by using a community-based cohort. Because these inflammatory biomarkers were measured at baseline in a cohort of individuals without known AF, these findings strongly support the hypothesis that systemic inflammation contributes to the clinical arrhythmogenesis.
Although we hypothesized that racial differences in abdominal adiposity could account for racial differences in AF risk via inflammatory cytokine production, we did not observe a significant adjusted association between CT-derived adiposity measurements and AF. This suggests that mechanisms distinct from abdominal fat cytokine production, such as genetic differences or environmental exposures, explain the difference in serum biomarker concentrations by race. It also remains possible that CT measurements of abdominal visceral and subcutaneous adiposity do not sufficiently quantify the fat stores responsible for generating the inflammatory cytokines important for AF pathogenesis; recent data suggests pericardial fat is associated with AF (27) and may contribute to an inflammatory response (28). It is also notable that we did not observe an association between BMI and AF. Although BMI has been independently associated with AF in the Framingham (10) and ARIC (Atherosclerosis Risk In Communities) (12) cohorts, BMI did not predict AF in the CHS (Cardiovascular Health Study) (29). The absence of an association in our Health ABC cohort could be explained by cohort characteristics (CHS and Health ABC, for instance, enrolled older participants compared to those in Framingham and ARIC), by differences in comorbidity characterization, or because BMI is an imperfect surrogate for more important pathologic mediators of AF risk, such as accumulation of pericardial fat.
Recent evidence derived from multiple racial and ethnic groups suggests the association between race and AF is due to heightened AF risk among whites rather than a protective effect unique to blacks (1). Consistent with this observation, increasing European ancestry among African Americans is an independent risk factor for AF (29). The underlying mechanism through which white race and European ancestry increase AF risk is unknown. Serum concentrations of TNF-α (13), IL-6 (13), CRP (14), and adiponectin (15) are known to differ by race, and greater European ancestry among African Americans predicts higher concentrations of adiponectin, CRP, IL-2 SR, IL-6 SR, and TNF-α SR II (16). The substantial proportion of the race–AF association mediated by inflammation adds to the understanding of AF pathogenesis and could have important treatment implications. The association between serum inflammatory marker concentration and incident AF persisted after adjustment for other exposures and cardiovascular conditions known to be linked with increased systemic inflammation, including coronary artery disease and heart failure. Although speculative, this finding could suggest that the inflammatory pathways important for AF pathogenesis are distinct from those associated with these other comorbid conditions. Furthermore, our mediation findings indicate that inflammation is a more prominent and, as a result, more important driver of AF risk among whites.
From a clinical standpoint, these results suggest that interventions targeting inflammation specific to AF pathogenesis may be especially important for AF risk reduction and that the efficacy of such therapies may differ by race. Indeed, inflammation-related AF may be a specific disease subtype amenable to tailored therapies. It has recently been shown that randomization to a weight loss intervention results in a reduction both in serum CRP concentration and in AF symptom burden and severity (30). Whether more targeted anti-inflammatory treatments could further improve AF outcomes remains untested.
Potential limitations of this analysis should be recognized. First, race was self-reported by study participants. Importantly, nearly all investigations examining the association between race and AF have used this methodology. In addition, inaccurate self-reporting of race would likely bias our results toward the null hypothesis. Second, because the relative strength of various inflammatory biomarkers with regard to AF prediction is not known, our analysis included multiple inflammatory cytokines and soluble receptors. Although we believe this approach is supported by biological plausibility and is preferable in light of limited prior data, we recognize that broad inclusion of candidate markers increases the likelihood of observing chance associations. It is important to acknowledge that this study does not establish a causal link between inflammation and AF, and the ability to prevent AF through inflammation suppression remains speculative. AF was not a pre-specified, adjudicated Health ABC study outcome, and we therefore relied on screening ECGs and ICD-9 coding to identify prevalent and incident disease. It is notable that administrative ICD-9 coding at a large health maintenance organization exhibited 95% sensitivity and 99% specificity for the diagnosis of AF compared with record review by trained abstractors (31). Nonetheless, we accept that we had a reduced ability to identify participants with AF who were asymptomatic, had paroxysmal disease, or did not seek care. Our findings are unlikely to be secondary to differential access to care, as the association between white race and AF has been replicated in multiple clinical contexts (1,2,32,33) and has been shown to be proportional to European ancestry within a cohort of African Americans (29). Finally, although we attempted to identify and exclude participants with prevalent AF, we cannot be certain that all individuals with AF at study enrollment were identified. In light of prior investigations that suggest AF itself may contribute to a proinflammatory state (34–37), it remains possible that AF was the cause (instead of the consequence) of heightened systemic inflammation. We believe such an “effect-cause” relationship was unlikely given the long pre-enrollment period over which prevalent AF could be ascertained (up to 6 years) and steady rate of AF diagnoses over our long duration of follow-up.
In a population-based sample of older adults, we have demonstrated that systemic inflammatory pathways significantly mediate the heightened risk of AF among whites compared with that among blacks. The higher level of systemic inflammation and concomitant increased AF risk among whites is not explained by racial differences in abdominal adiposity or the presence of other proinflammatory cardiovascular comorbidities such as diabetes, coronary artery disease, and heart failure. Future research aimed at defining and treating the inflammatory pathways unique to AF risk could identify primary prevention therapies relevant to individuals prone to an inflammatory AF subtype.
COMPETENCY IN MEDICAL KNOWLEDGE: Despite having a lower prevalence of established risk factors, whites exhibit a substantially increased risk of AF compared with blacks. Racial differences in inflammation appear to explain a sizeable proportion of the association between white race and atrial fibrillation risk.
TRANSLATIONAL OUTLOOK: Additional research aimed at defining and treating the inflammatory pathways unique to AF risk could identify primary prevention therapies relevant to individuals prone to an inflammatory AF subtype.
This study was supported by American Heart Association grants 12POST11810036 (Dr. Dewland) and 12GRNT11780061 (Dr. Marcus) and the Joseph Drown Foundation (Dr. Marcus). Also supported by U.S. National Institutes of Health Intramural Research Program, National Institute on Aging (NIA) contracts N01-AG-6-2101, N01-AG-6-2103, and N01-AG-6-2106 and NIA grants R01-AG028050 and R03-AG045075; National Institute of Nursing Research grant R01-NR012459; and NIH National Center for Advancing Translational Sciences award UL1TR000454. Dr. Dewland has received education-related travel reimbursement from Medtronic. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- body mass index
- confidence interval
- C-reactive protein
- computed tomography
- enzyme-linked immunosorbent assay
- hazard ratio
- plasminogen activator inhibitor
- soluble receptor
- tumor necrosis factor
- Received March 31, 2015.
- Accepted April 23, 2015.
- American College of Cardiology Foundation
- Dewland T.A.,
- Olgin J.E.,
- Vittinghoff E.,
- Marcus G.M.
- Frustaci A.,
- Chimenti C.,
- Bellocci F.,
- Morgante E.,
- Russo M.A.,
- Maseri A.
- Issac T.T.,
- Dokainish H.,
- Lakkis N.M.
- Chung M.K.,
- Martin D.O.,
- Sprecher D.,
- et al.
- Aviles R.J.,
- Martin D.O.,
- Apperson-Hansen C.,
- et al.
- Schnabel R.B.,
- Larson M.G.,
- Yamamoto J.F.,
- et al.
- Fain J.N.
- Huxley R.R.,
- Misialek J.R.,
- Agarwal S.K.,
- et al.
- Visser M.,
- Pahor M.,
- Taaffe D.R.,
- et al.
- Khera A.,
- McGuire D.K.,
- Murphy S.A.,
- et al.
- Katzmarzyk P.T.,
- Bray G.A.,
- Greenway F.L.,
- et al.
- ↵National Institute on Aging, Nations Institutes of Health. Health ABC. Available at: http://www.grc.nia.nih.gov/branches/leps/healthabc/index.htm. Accessed July 2015.
- Wong C.X.,
- Abed H.S.,
- Molaee P.,
- et al.
- Mazurek T.,
- Zhang L.,
- Zalewski A.,
- et al.
- Marcus G.M.,
- Alonso A.,
- Peralta C.A.,
- et al.
- Ruo B.,
- Capra A.M.,
- Jensvold N.G.,
- Go A.S.
- Rotter M.,
- Jaïs P.,
- Vergnes M.-C.,
- et al.
- Kallergis E.M.,
- Manios E.G.,
- Kanoupakis E.M.,
- et al.