Article Text

Download PDFPDF

Original article
N-terminal pro-B-type natriuretic peptide as a predictor of incident atrial fibrillation in the Multi-Ethnic Study of Atherosclerosis: the effects of age, sex and ethnicity
  1. Kristen K Patton1,
  2. Susan R Heckbert2,
  3. Alvaro Alonso3,
  4. Hossein Bahrami4,
  5. Joao A C Lima5,
  6. Gregory Burke6,
  7. Richard A Kronmal7
  1. 1Division of Cardiology, University of Washington, Seattle, Washington, USA
  2. 2Department of Epidemiology, University of Washington, Seattle, Washington, USA
  3. 3Division of Epidemiology and Community Health, School of Public Health, University of Minnesota, Minneapolis, Minnesota, USA
  4. 4Cardiology Division, Stanford University, Stanford, California, USA
  5. 5Department of Cardiology, Johns Hopkins University, Baltimore, Maryland, USA
  6. 6Division of Public Health Sciences, Wake Forest University School of Medicine, Winston-Salem, North Carolina, USA
  7. 7Department of Biostatistics, University of Washington, Seattle, Washington, USA
  1. Correspondence to Dr Richard A Kronmal, Collaborative Health Studies Coordinating Center, University of Washington, Bldg. 29, Suite 310, 6200 NE 74th Street, Seattle, WA 98115, USA; kronmal{at}


Objective Atrial fibrillation (AF) is a common source of medical costs, morbidity and mortality. NT pro-brain natriuretic peptide (BNP) is a remarkably strong predictor of AF in older whites; we aimed to assess if this biomarker was as predictive in other racial groups.

Design We used covariate-adjusted Cox model regressions to estimate the HRs of developing AF as a function of NT proBNP, and tested for interactions of NT proBNP with age, gender and race/ethnicity.

Setting The Multi-Ethnic Study of Atherosclerosis (MESA).

Patients 5518 subjects were followed over a median of 7.6 years. During this time, 267 developed AF.

Results NT proBNP was statistically significantly associated with incident AF; for ln NT proBNP, the adjusted HR was 2.2 (95% CI 1.9 to 2.5). Assessed by quintiles, the relationship between NT proBNP was strong and graded; the unadjusted HR for the highest quintile of NT proBNP was 23.7 (95% CI 11.1 to 50.6) and adjusted was 11.4 (95% CI 5.1 to 25.3). NT proBNP was an excellent predictor of incident AF in the younger and older age groups, in men and women and in the different race/ethnicity groups: the HR for ln NT proBNP as a predictor of incident AF ranged from 2.0 to 3.9 in each subgroup.

Conclusions NT proBNP is a robust predictor of incident AF; its prognostic value is more significant in younger patients and women compared with older patients and men. NT proBNP was also as strongly predictive in black patients, Hispanics and Asian/Chinese as in white patients despite a lower incidence of arrhythmia.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Atrial fibrillation (AF) is the most common cardiac arrhythmia and a major source of both cardiovascular and cerebrovascular morbidity and mortality.1 Over 2 million people in the USA have been diagnosed with AF, and this number is expected to increase 2.5 fold over the next 50 years.2 AF is associated with impaired quality of life, increased risk of heart failure and worsened mortality, and is thought to account for nearly half of all embolic strokes.3 Much of the current understanding of the pathophysiology of AF is based on a handful of well-characterised community cohorts, with limited ethnic, socioeconomic or geographic variability.4–6

These important studies of predominantly white populations have revealed AF to be a heterogeneous disorder, and have established multiple clinical risk factors: age, heart failure, valvular heart disease, hypertension, diabetes, obesity, cigarette smoking, obstructive sleep apnoea, metabolic syndrome, inflammation, structural heart disease and several genetic loci.7–9 Less is known about AF in non-white populations,10 ,11 but what is known is paradoxical. Unlike other well described cardiovascular diseases, AF is more frequent in white populations despite a higher prevalence of clinical risk factors for AF in black patients versus white patients.12–14 Data in other ethnic populations are even more limited, yet also suggest a lower incidence of AF in non-white populations.14 ,15 These findings are particularly puzzling in the case of black individuals, given the markedly increased incidence of stroke compared with white individuals and the strong association of stroke with AF.16 It is unclear whether these findings are due under-ascertainment of the occurrence of AF episodes17 or differing expression of AF symptomatology or pathophysiology in different ethnicities. We explore whether differences in amino terminal (NT) pro-brain natriuretic peptide (BNP) among the race/ethnicity/ethnic groups in the Multi-Ethnic Study of Atherosclerosis (MESA) might explain this phenomenon.

The neurohormone BNP is a regulator of cardiovascular function, and is elevated in acute myocardial infarction and heart failure.18 In an analysis of the Cardiovascular Health Study, a single baseline measurement of NT proBNP was a remarkably strong predictor of incident AF, independent of traditional clinical risk factors.11 Currently, it is unclear if elevated BNP represents a subclinical marker of atrial dysfunction in AF development or if it plays another role in pathogenesis. Likewise, it is unknown whether the strong association is present in younger populations or in other race/ethnicity groups, since prior studies largely comprised white subjects. The MESA population provides a unique opportunity to study the relationship of NT proBNP and AF in a younger, multi-ethnic population.


Study population

A detailed description of the MESA study, a prospective observational cohort investigating the characteristics of subclinical cardiovascular disease, has previously been published.19 In brief, 6814 participants between the ages of 45 and 84 years, free of clinically apparent cardiovascular disease, were recruited from six communities (Forsyth County, North Carolina; Northern Manhattan and the Bronx, New York; Baltimore and Baltimore County, Maryland; St Paul, Minnesota; Chicago, Illinois; and Los Angeles County, California, USA) between 2000–2002. Each site recruited an approximately equal number of men and women according to prespecified age and race/ethnicity/ethnicity proportions (approximately 38% white, 28% Black, 22% Hispanic and 12% Asian, predominantly Chinese). All participants gave informed consent, and the protocol was approved by the institutional review board at each site. At enrolment, a detailed questionnaire collected information on demographics, family history, medical history, medication use, lifestyle habits and psychosocial factors. The physical examination included height, weight and blood pressure measurements, and blood was drawn for laboratory measurements. Participants were contacted approximately every 9 months to inquire about all hospitalisations since the last contact. Information on hospital discharge diagnoses were obtained from individual hospitals for all reported hospitalisations. This information was supplemented with Medicare inpatient discharge diagnosis codes for those MESA participants who were Medicare fee-for-service beneficiaries at any time during follow-up. Of the 6814 participants enrolled in MESA, 36 had no follow-up data and 58 were determined from the Centers for Medicare and Medicaid Services (CMS) events ascertainment to have AF prior to entry in MESA. Of the remaining participants, 1202 did not have a NT proBNP measurement due to scarcity of baseline serum. Thus, the sample available for the analyses that appears in this paper is 5518.

Natriuretic hormone assay

NT-proBNP levels were determined from frozen serum samples drawn at enrolment and stored at −70°C. All measurements were performed at the University of California, San Diego; in a CLIA certified laboratory using an FDA-approved commercially available immunoassay from Roche Diagnostics Corporation (Roche Diagnostic Elecsys proBNP Assay, Indianapolis, Indiana, USA) on the Elecsys 2010 instrument. A 250 μL serum sample previously unthawed or only thawed once was used for analysis. The intraassay and interassay coefficients of variation were as follows: at 175 pg/mL, 2.7% and 3.2%; at 355 pg/mL, 2.4% and 2.9%; at 1068 pg/mL, 1.9% and 2.6%; and at 4962 pg/mL, 1.8% and 2.3%.20 The measuring range is 5–35 000 pg/mL. Values below the detection limit are reported as <5 pg/mL, and values above are reported as >35, 000 pg/mL. The core laboratory reported certified data to the central data repository and was blinded to patient outcome.

Determination of incident AF

Participants with a self-report of a diagnosis of current AF were excluded from enrolment in MESA. Incident AF was identified from hospital discharge diagnosis (International Classification of Diseases, Ninth Revision [ICD-9]) codes for AF or atrial flutter, (classified as ICD-9 diagnosis codes 427.31 and 427.32) ascertained by the MESA events detection protocol or from CMS hospital billing information. A validation substudy reviewing a random sample of 45 of 185 MESA participants with hospital discharge diagnosis codes for AF showed that AF was confirmed in 93% of hospitalisations, implying a high specificity for the diagnosis.

Cardiac MRI

Cardiac MRI was performed at six MESA field centres using scanners with 1.5 Tesla magnets, a four-element, phased-array surface coil placed anteriorly and posteriorly, ECG gating and brachial artery blood pressure monitoring.21 Left ventricular structure and function were determined at baseline in 5004 participants (73.4%) who agreed to undergo the study. The MRI protocol and analysis methods have been previously described.19

Analysis plan and methods

We used covariate-adjusted Cox model regressions to estimate the HRs of developing AF as a function of NT proBNP, and tested for interactions of NT proBNP with age, gender and race/ethnicity. NT proBNP levels were analysed both as a continuous variable, where the natural log (ln) of NT proBNP was used and as categories of NT proBNP in quintiles. Failure time for the Cox models for those who developed AF was the time between the participant's baseline exam and the first hospitalisation with an AF diagnosis. For those who did not develop AF, failure time was the time between the participant's baseline exam and the last known follow-up.

Cox models were adjusted for significant predictors from a backward selection from among the following variables: age, sex, race/ethnicity, ln NT proBNP, body mass index (weight/height2), height, cigarette smoking, hypertension, systolic and diastolic blood pressure, hypertensive medication use and diabetes. Stratified analyses were performed for age, sex and race/ethnicity.

In supplementary analyses, the cardiac MRI variables left ventricular mass, left ventricular end diastolic volume and ejection fraction were also used for adjustment.

All HRs for NT proBNP are given in units of SD, which is equal to 1.2. Analyses were performed using STATA V.11.0 (Statacorp, College Station, Texas, USA).


Over a median of 7.6 years of follow-up (with 99% of the participants having less than 8.25 years of follow-up), of 5518 subjects, 267 developed AF. The demographic and clinical characteristics of the participants according to AF status are compared in table 1.

Table 1

Demographic and clinical characteristics of the study participants by incident atrial fibrillation (AF) status

As expected, older age was associated with AF; the average age in the AF group was 72 years, compared with 62 years in the no AF group. Similarly expected, subjects with AF were more likely to be male and white. A higher systolic blood pressure, a diagnosis of hypertension and medical therapy for hypertension were all more common in subjects with AF. Body mass index and diagnosis of diabetes did not differ significantly between the two groups. Although ejection fraction was similar, the left ventricular mass was higher in those with AF. The average NT proBNP level was also higher in the subjects with AF.

BNP was statistically significantly associated with incident AF; for ln NT proBNP, the unadjusted HR was 2.6 (95% CI 2.4 to 3.0) per SD, and when fully adjusted for age, sex, race/ethnicity and the other significant predictors for AF, the HR was 2.2 (95% CI 1.9 to 2.5), (table 2).

Table 2

Cox proportional hazards regression for NT ProBNP as a predictor of incident atrial fibrillation

All HR for NT proBNP are given in units of SD, which is equal to 1.2. When assessed by quintiles, the relationship between NT proBNP was significant and graded. The unadjusted HR for the highest quintile of NT proBNP was 23.7 (95% CI 11.1 to 50.6, and adjusted for age, sex, race/ethnicity and significant clinical variables was 11.4 (95% CI 5.1 to 25.3), compared with the lowest quintile. Of all the cardiac MRI variables assessed, only left ventricular mass and ejection fraction were significant predictors of AF. The HR of ln NT proBNP in the model including the MRI variables was HR 2.1, compared with HR 2.2 in the clinically adjusted model without the MRI variables.

NT proBNP was significantly associated with incident AF in the younger and older age groups, in men and women, and in the different ethnic groups (the p values in each group were <0.005; see table 3).

Table 3

Unadjusted Cox proportional hazards model for ln NT proBNP as a predictor of incident atrial fibrillation stratified by age, sex and race/ethnicity/ethnicity

There was evidence of a differential association between NT proBNP and AF among the age and gender groups, but not across the different racial groups. There was also evidence that NT proBNP is a stronger predictor among females compared with males (HR 3.5 vs HR 2.5, p value for interaction of <0.01) and in younger patients compared with older (HR 3.0 vs HR 2.0, p value for interaction of <0.05). For race/ethnicity there was insufficient evidence to conclude that the HRs were different (table 3). The cumulative incidence curves for AF by quintile of NT proBNP are shown for the entire cohort, and for each of the race/ethnicity/ethnic groups (figures 1A–E). The curves are truncated at 8 years of follow-up for better display of the data. For all of the race/ethnicity groups the 8 year cumulative incidence for the top quintile of NT proBNP is between 10% and 15% compared with less than 1% for the 1st quintile.

Figure 1

Cumulative incidence curves for atrial fibrillation by quintile of NT proBNP. (A) Entire cohort. (B) Whites. (C) Blacks. (D) Hispanics. (E) Chinese.


We found a single baseline measurement of NT proBNP was as significantly predictive of incident AF in younger subjects as in older, for both men and women, and as effective in Black, Hispanic, and Chinese race/ethnicity as it is in Whites. In fact, NT proBNP was an even more robust predictor of incident AF in women than men and in participants under 65 years of age than those aged 65 or older. Notably, enrolment into MESA was predicated on absence of clinical cardiovascular disease, which reveals that NT proBNP is a highly statistically significant predictor in subjects without clinically recognised atherosclerotic disease. Furthermore, adjustment for comorbidities associated with the development of AF did not significantly attenuate this association.

Our results also confirm prior findings of a lower AF incidence in Black, Hispanic, and Asian compared with White. However, despite the relatively small number of events, the cumulative risk of incident AF among all participants at 8 years of follow-up was greater than 12.5% in subjects in the highest quintile of NT proBNP compared with less than 1% in the lowest. The baseline NT proBNP level was a substantially stronger predictor of incident AF than any other clinical covariate.

Our findings are consistent with several prior reports11 ,22 ,23 showing a similar striking association between NT proBNP and AF, and also extending this finding to younger populations and other racial groups. The paradox of a lower prevalence of AF in certain race/ethnicity groups has prompted considerable attention,24 and it is unclear whether this is due to under-ascertainment of AF in different race/ethnicity groups or if protective environmental factors or genetics are playing a role.17 However, NT-pro BNP differences among different race/ethnicities do not appear to explain the difference in AF prevalence according to race/ethnicity, as these differences persist after adjustment for this potent predictor of AF risk.

AF is a notoriously difficult endpoint to assess due to its often paroxysmal and asymptomatic nature, and this is a limitation in our analysis. Thus, we cannot be sure that the relationship of NT proBNP we show will be the same for paroxysmal or asymptomatic AF due to our ascertainment of events only through ICD codes from hospitalisation records. However, some fraction of our AF events will be paroxysmal or even asymptomatic (diagnosed from a hospital ECG) and thus it is likely that NT proBNP is also strongly associated with these forms of AF.

Given the ongoing AF epidemic, and the resultant effects on morbidity, mortality, and healthcare costs, there has been increasing interest in prevention.7 Unfortunately, the complexities of the underlying physiology, coupled with the difficulty in establishing epidemiologic patterns, have contributed to our inability to reduce this burden on public health.


NT proBNP is a robust predictor of incident AF; its prognostic value is even greater in younger patients and women compared with older patients and men. NT proBNP was also as robustly predictive in Blacks, Hispanics and Asian/Chinese as in Whites, again despite a lower incidence of arrhythmia. Given the high medical and financial burden of AF, NT proBNP may be useful to distinguish a high-risk group to target for preventative measures.



  • Contributors Each author contributed significantly to the submitted work. KKP and RAK participated in the conception and design of the study, drafting and revising the manuscript. RAK performed the analysis. SRH, AA, HB, JACL and GB participated in the data interpretation and revised the manuscript critically for important intellectual content. All authors had final approval of the manuscript.

  • Funding This research was supported by contracts N01-HC-95159 through N01-HC-95169 from the National Heart, Lung, and Blood Institute and by grants UL1-RR-024156 and UL1-RR-025005 from NCRR. The authors thank the other investigators, the staff and the participants of the MESA study for their valuable contributions. A full list of participating MESA investigators and institutions can be found at

  • Competing interests None.

  • Ethics approval University of Washington IRB.

  • Provenance and peer review Not commissioned; externally peer reviewed.