Relation of Resting Heart Rate to Incident Atrial Fibrillation (From ARIC [Atherosclerosis Risk in Communities] Study)

Accepted Manuscript Title: Relationship of Resting Heart Rate to Incident Atrial Fibrillation (From ARIC [Atherosclerosis Risk in Communities] Study) Author: Weijia Wang, Alvaro Alonso, Elsayed Z. Soliman, Wesley T. O'Neal, Hugh Calkins, Lin Yee Chen, Marie Diener-West, Moyses Szklo PII: DOI: Reference:

S0002-9149(18)30179-6 https://doi.org/10.1016/j.amjcard.2018.01.037 AJC 23115

To appear in:

The American Journal of Cardiology

Received date: Accepted date:

3-11-2017 22-1-2018

Please cite this article as: Weijia Wang, Alvaro Alonso, Elsayed Z. Soliman, Wesley T. O'Neal, Hugh Calkins, Lin Yee Chen, Marie Diener-West, Moyses Szklo, Relationship of Resting Heart Rate to Incident Atrial Fibrillation (From ARIC [Atherosclerosis Risk in Communities] Study), The American Journal of Cardiology (2018), https://doi.org/10.1016/j.amjcard.2018.01.037. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Relationship of Resting Heart Rate to Incident Atrial Fibrillation (From ARIC [Atherosclerosis Risk in Communities] Study) Weijia Wang, MD, MPHa, Alvaro Alonso, MD, MPH, PhDb, Elsayed Z. Soliman, MD, MSc, MSc, Wesley T. O’Neal, MD, MPHd, Hugh Calkins, MDe, Lin Yee Chen, MD, MSf, Marie Diener-West, PhDa, Moyses Szklo MD, DrPH, MPHa Affiliations a

The Johns Hopkins University Bloomberg School of Public Health, Baltimore, MD, bEmory

University Rollins School of Public Health, Atlanta, GA, cWake Forest University School of Medicine, Winston-Salem, NC, dEmory University School of Medicine, Atlanta, GA, eThe Johns Hopkins University School of Medicine, Baltimore, MD, fUniversity of Minnesota School of Medicine, Minneapolis, MN Corresponding author Moyses Szklo, Telephone Number: 410-955-3462; Email address: [email protected]; Mailing address: 615 N. Wolfe Street Room W6009 Baltimore, Maryland 21205 Running title: Resting HR and AF in ARIC List of Support/Grant Information The Atherosclerosis Risk in Communities study has been funded in whole or in part with Federal funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Department of Health and Human Services, under Contract nos. (HHSN268201700001I, HHSN268201700003I, HHSN268201700005I, HHSN268201700004I, HHSN2682017000021). This work was also supported by American Heart Association grant 16EIA26410001 (Alonso) and NHLBI/NIH F32HL134290 (O’Neal WT).

1

Page 1 of 18

Abstract The evidence on the association between resting heart rate (HR) and incident atrial fibrillation (AF) is conflicting. Whether change in resting HR is associated with incident AF is unknown. We evaluated 11,545 participants (mean (±SD) age: 57  5.7 years) free of AF at baseline (1990-1992). Resting HR was obtained from 10-second electrocardiograms at baseline and 3 years later. AF diagnosis was ascertained from visit electrocardiograms, hospital discharge records, and death certificates through 2013. High and low resting HR were defined as  80 and < 55 beats/minute, respectively. Increase and decrease in HR were defined as a 3-year HR difference > 15 and < -15 beats/minute, respectively. Over a median follow-up of 22.5 years, 1746 (15%) participants developed AF. Both baseline high resting HR and increase in HR were independently associated with incident AF (Hazard ratio = 1.2, 95% confidence interval = 1.0-1.5 and hazard ratio = 1.4, 95% confidence interval = 1.1-1.9). Increase in HR was no longer associated with incident AF after additional adjustment for incident heart failure. In stratified analyses, increase in HR was only associated with AF in participants < 60 years, with bachelor’s degree or above, without diabetes, and without hypertension (P values for interaction ≤ 0.05). In conclusion, in a middle-aged population, high resting HR is associated with higher AF risk. Increase in resting HR is also associated with higher AF risk, especially in individuals without traditional AF risk factors. Whether interventions to decrease HR can prevent AF remains to be examined.

Word count : 249 Key words Heart rate, atrial fibrillation, ARIC

2

Page 2 of 18

Introduction Atrial fibrillation (AF) increases the risk of stroke and overall mortality 1. Common risk factors of AF include age, hypertension, diabetes, heart failure, myocardial infarction, sleep apnea, obesity and thyroid disease2,3,4. Its prevalence is projected to double by 20505, underscoring the need for identifying simple markers and novel risk factors for better prediction and prevention. Resting heart rate (HR) is easy to monitor and high resting HR is an established risk factor for cardiovascular mortality and morbidity6,7,8,9. It also offers a potential target for therapy, either via life style modification or medications10. However, the results on the association between resting HR and incident AF have been conflicting. Low resting HR was associated with higher risk of incident AF in athletes11 and elderly population12. On the other hand, high resting HR was also independently associated with incident AF13, especially in hypertensive population14. In addition, the association between temporal change in HR and incident AF is unclear. Thus, we established the following 2 aims: (1) to examine the associations between baseline resting HR and incident AF; (2) to explore the association of change in HR over time with incident AF. The Atherosclerosis Risk in Communities (ARIC) study, a population-based cohort with 25-year follow-up and a large number of incident AF cases, provides an excellent opportunity for this study.

Methods The ARIC study is a mostly biracial, prospective cohort study of 15,792 men and women 45-64 years of age at baseline, recruited from 4 US communities: Forsyth County, North Carolina; Jackson, Mississippi; suburbs of Minneapolis, Minnesota; and Washington County, Maryland. Four visits, one every 3 years, were conducted between 1987 and 1998; a fifth visit was conducted from 2011 to 2013. The ARIC study design has been previously described15. Visit 2 (1990-1992) was selected as the baseline for the present study because thyroid function and NTproBNP were measured at Visit 2 and considered as important covariates. Survival bias from the choice of baseline visit should be negligible since < 5% (111) of the incident AF cases occurred before Visit 2. Of the 14,348 participants who attended Visit 2, we excluded those with prevalent AF(n=110); prevalent sick sinus syndrome, use of pacemaker, bundle branch block, Wolf-Parkinson White syndrome, and idioventricular rhythm (n=430); taking medication affecting HR (beta-blockers, non-dihydropyridine calcium channel blockers, digoxin and other antiarrhythmics) (n= 1934); being non-white or non-black from all study sites, and non-white

3

Page 3 of 18

from the Minneapolis and Washington County sites (n=81); missing values on resting HR or incident AF (n=248). A total of 11,545 individuals were eligible for the analysis of resting HR and incident AF. For the separate analysis of change in HR and incident AF, Visit 3 (1993-1995) was chosen as the time origin. After further excluding participants with missing HR at Visit 3 (n=1351), incident AF cases between Visit 2 and 3 (n=64), 10,128 individuals were included. The Institutional Review Boards at all ARIC study sites and the Coordinating Center approved study protocols. All participants provided written informed consent at each study visit. At each follow-up visit, a standard supine 12-lead resting electrocardiogram (ECG) of 10 seconds was recorded after a 12-hour fast followed by a light snack and at least 1 hour after smoking or ingestion of caffeine. HRs measured from ECGs at Visit 2 (1990-1992) and Visit 3 (1993-1995) were retrieved. The difference in resting HR was defined as resting HR at Visit 3 minus resting HR at Visit 2. AF diagnosis was identified between the date of Visit 2 through December 31, 2013 from 3 sources16: ECGs at visits 2 to 5, hospital discharge records, and death certificates. AF was defined as the presence of ICD-9 code 427.31 or 427.32 in the discharge codes. We excluded AF hospitalization diagnoses occurring simultaneously with heart revascularization surgery or other cardiac surgery involving heart valves and septa without evidence of AF in subsequent hospitalization or study exams. If the underlying cause of death was AF (ICD-9 code 427.3) and no AF was documented prior to death, participants were considered as AF cases. (Three cases of incident AF were identified by death certificates only in ARIC16.) Comparing to a physician reviewing participants’ medical records, the diagnosis of incident AF based on ICD-9 code had a sensitivity of 84% (African American 80%, white 85%) and specificity of 98% (African American 99%, white 97%)16. Race/center, gender, cigarette smoking status, pack-years of smoking, health insurance status, and alcohol drinking status were self-reported at Visit 2. Sports, leisure, and work physical activity (via a modified Baecke questionnaire) and education level were assessed at Visit 1 and were carried over to Visit 2. Body mass index was calculated as weight in kilograms divided by the square of height in meters. Blood pressure was measured 3 times, and the average of the last 2 measurements was considered in these analyses. Hypertension was defined as blood pressure of >140 mm Hg systolic or >90 mm Hg diastolic from visit measurements 17, or use of blood pressure medication in the past 2 weeks. Diabetes was defined as fasting glucose of ≥ 126 mg/dl (or non-fasting glucose ≥ 200 mg/dl)18, self-report of a physician diagnosis, or current use of diabetes medications. Impaired fasting glucose

4

Page 4 of 18

(pre-diabetes) was defined as fasting glucose  100 and <126 mg/dl and no diagnosis of diabetes18. Prevalent coronary heart disease was defined as presence of a self-report of myocardial infarction, coronary bypass, angioplasty, or myocardial infarction suggested on the baseline ECG and adjudicated incident coronary heart disease between Visits 1 and 2. Prevalent heart failure was identified using the Gothenburg criteria or self-report of heart failure medication use in the past 2 weeks19 and incident heart failure between Visits 1 and 2. Incident coronary heart disease and heart failure were ascertained at study visits, hospital records, and death certificates. Participants were asked to bring all the medications they were currently taking to the field center visit. Left ventricular hypertrophy on ECG was determined by Cornell Voltage Criteria20. Abnormal P-terminal force in V1, a measurement of left atrial enlargement, was defined as the product of the duration and amplitude of the terminal portion of P-wave at V1 that is  −4 mVms21.Total cholesterol, high-density lipoprotein cholesterol, and triglycerides were measured using standardized enzymatic assays. Low-density lipoprotein cholesterols were then calculated based on Friedewald formula22. Estimated glomerular filtration rate was calculated from serum creatinine using the CKD-Epi equation23. NT-proBNP and thyroid-stimulating hormone were measured by a sandwich immunoassay method in 2012-2013 from thawed Visit 2 serum samples that had been stored at -70 degrees Celsius. Baseline characteristics of participants were compared between groups using the Chi-squared test, KruskalWallis test, t-test, and Wilcoxon Rank-Sum test as appropriate. To study the association between baseline resting HR and incident AF, follow-up time was calculated from the date of Visit 2 to the date of incident AF or censoring (either death, loss to follow-up, or administrative censoring on December 31, 2013). Incidence rates of AF were calculated according to resting HR categories (low: < 55 beats/minute [bpm], intermediate: 55-79 bpm, and high:  80 bpm). The cutoffs were chosen based on clinical relevance and the distribution of resting HR (Figure 1). Cox proportional hazard models were used to estimate the hazard ratios for the association between resting HR and incident AF. Because the log hazard of incident AF did not appear to be linearly associated with the exposure, resting HR was treated as a categorical variable as above. Model 1 was adjusted for age, race/center and gender; Model 2 was additionally adjusted for smoking, alcohol, physical activity, income, education, health insurance status; Model 3 was adjusted for variables in model 2 plus body mass index, systolic blood pressure, prevalent coronary heart disease, prevalent heart failure, prevalent diabetes or impaired fasting glucose, left ventricular hypertrophy on ECG, left atrial enlargement on ECG, low-density lipoprotein cholesterol, thyroid function, renal function and NT pro BNP. We also examined the graphical association between baseline resting HR and incident

5

Page 5 of 18

AF using a restricted cubic spline model and incorporated knots at the 5 th, 50th and 95th percentiles. In the analysis of the association between change in resting HR and incident AF, follow-up time was calculated from the date of Visit 3. Increase in HR was defined as the difference in HR between Visit 3 and Visit 2 > 15 bpm. Decrease was defined similarly as < -15 bpm. The intermediate values were defined as stable. These cutoffs were chosen based on clinical relevance and the distribution of change in HR from Visit 2 to Visit 3 (Figure 2). Cox proportional hazard models were used to estimate the hazard ratios for the association between change in resting HR and incident AF. In addition to the covariates in models 1, 2 and 3, resting HR at Visit 2 and medications affecting HR at Visit 3 were also adjustment variables. Because incident heart failure was considered as a very important mediator/ confounder which changes over time in follow up, it was treated as a time-varying variable. Because AF incidence varies significantly with age, gender, race/center, education, diabetes and hypertension statuses1, interactions between baseline resting HR (or change in resting HR) and these variables were tested by including interaction terms in the Cox models. Statistical analyses were performed using Stata/IC 14.2 and SAS 9.4. A 2-sided p value of <0.05 was considered statistically significant. Results Baseline characteristics are shown in Table 1. Mean ( SD) age was 57  5.7 years (range: 46-70 years). Fifty seven percent were women and 24% were black. Mean ( SD) resting HR was 65  10 bpm. Participants with resting HR  80 bpm were more likely to be female and black, and to have fewer years of education and less physical activity scores, but more pack years of smoking, and less likely to be current drinkers and have health insurance. They also had a higher body-mass index and a more adverse lipid profile. A higher prevalence of hypertension, diabetes, coronary heart disease, heart failure, left ventricular hypertrophy and left atrial enlargement on ECG, and hyperthyroidism was observed in the high resting HR group as well.

Over a median follow-up of 22.5 years, a total of 1746 (15%) participants developed AF (Table 2). Compared to intermediate resting HR, only high resting HR was independently associated with incident AF. In a restricted cubic spline model, an increased hazard of AF was observed with increasing HRs but not decreasing HRs (Figure 3). There was no difference in the relationship by age, race/center, diabetes or hypertension (Supplemental Materials: Table S1).

6

Page 6 of 18

Mean change in HR from Visit 2 to 3 was 0  8 bpm. Increase but not decrease in HR was associated with an increased hazard of incident AF. (Table 3). The association was attenuated after adjusting for incident heart failure. When compared with their counterparts, higher hazards with increase in HR versus stable HR were found in patients who were < 60 years (p for interaction, pint= 0.05), females (pint= 0.06), blacks (pint= 0.15) and those with bachelor’s degree or above (pint= 0.02), without diabetes (pint= 0.02), and without hypertension (pint= 0.01) (Table 4). Sensitivity analysis excluding participants with prevalent and incident myocardial infarction yielded similar results. The association between change of resting HR and incident AF remained unchanged after further excluding participants taking medications affecting HR at Visit 3. (Supplemental Materials: Table S2-S8). Discussion In this large middle-aged, population based cohort, high but not low resting HR was associated with higher incidence of AF. An increase in HR over 3 years was not associated with AF in the entire cohort after adjustment for incident heart failure. However, heterogeneities were found for several variables: increases in HR were found to be associated with incident AF in participants who were < 60 years, those with bachelor’s degree or above, nonhypertensive, or non-diabetic. Although not reaching the conventional p= 0.05 cut-off point, nominal interactions were also found for race/center and sex. The observed association between high resting HR and incident AF is in line with previous studies13,14. High resting HR was thought to represent poor physical fitness24 and was associated with increased morbidity and mortality6,7,8. Potential mechanisms include increased sympathetic activity resulting in high resting HR25 and/or subclinical reductions in left ventricular function manifested as high resting HR26. Both were reported to promote the development of AF14,27. This is the first study examining the association between change in resting HR and incident AF in a populationbased cohort. Increase in HR was associated with increased cardiovascular mortality 7,28. In the trial of Losartan Intervention for End Point Reduction in Hypertension, every 10 bpm increase in the in-treatment heart rate was associated with 19% increase risk of AF14. In our study, the association disappeared upon adjustment for incident heart failure, which suggests that ventricular dysfunction may mediate the relationship between HR and AF. However, HR could be just a marker of ventricular dysfunction and not in its causal pathway leading to AF. Of note, 90% AF patients’ heart rate were stable in our cohort, suggesting other major mechanisms in AF.

7

Page 7 of 18

HR increase was not significantly associated with incident AF among participants exposed to known risk factors for AF, namely the elder, white, male, less educated, and patients with diabetes and hypertension. One possible explanation is that the incidence of those exposed to these risk factors and with stable HR is much higher than that in the unexposed groups (Supplemental Materials: Table S9), which would dilute the hazard ratios. Another explanation is that cohort participants may develop AF before the increase in HR, resulting in survival bias. Low resting HR was associated with higher incidence of AF in trained athletes26 but the ARIC participants were not nearly as athletic. Also, low resting HR was associated with incident AF in the elderly in Cardiovascular Health Study12. Alternations of sinus node function, which may predispose to AF 29,30 , is more prevalent with aging31. The younger age profile in ARIC (mean age at baseline 57 years versus 72 years in Cardiovascular Health Study) could account for this discrepancy. Also, we excluded participants taking medications which affect the HR. It is a population with lower HR from medication effect but at higher AF risk due to underling comorbidities such as hypertension and heart failure. The clinical implications are 2-fold. First, resting HR > 80 beats per minute should alert physicians of excessive AF risk and address risk factors (such as obesity, smoking, and sleep apnea) thoroughly. Second, since resting HR is potentially modifiable10, whether intervention to decrease HR can prevent AF effectively and safely is of interest. Some limitations should be mentioned. Paroxysmal AF was likely to be underestimated with the 3 sources of ascertainment. Also, echocardiography was not available so we did not have structural information on atria and ventricles. We used P terminal force in V1 as a surrogate for left atrial enlargement 21 and Cornell Voltage Criteria20 for left ventricular hypertrophy. Given the circadian variability of the HR, a single-point 10-second ECG may not represent the mean/median resting HR. Lastly, residual confounding was always possible. In conclusion, high but not low resting HR is associated with higher hazard of incident AF in middle-aged population. An increase in resting HR is also associated with higher hazard of incident AF, especially in participants without traditional risk factors for AF. Whether interventions to decrease HR, via either lifestyle or medications, can prevent AF remains to be examined. Acknowledgement The authors thank the staff and participants of the ARIC study for their important contributions.

8

Page 8 of 18

Reference 1. Benjamin EJ, Wolf PA, D’Agostino RB, Silbershatz H, Kannel WB, Levy D. Impact of atrial fibrillation on the risk of death: the Framingham Heart Study. Circulation 1998;98:946–952. 2. Benjamin EJ, Levy D, Vaziri SM, D’Agostino RB, Belanger AJ, Wolf PA. Independent risk factors for atrial fibrillation in a population-based cohort. The Framingham Heart Study. JAMA 1994;271:840–844. 3. Gami AS, Pressman G, Caples SM, Kanagala R, Gard JJ, Davison DE, Malouf JF, Ammash NM, Friedman PA, Somers VK. Association of atrial fibrillation and obstructive sleep apnea. Circulation 2004;110:364–367. 4. Frost L, Vestergaard P, Mosekilde L. Hyperthyroidism and risk of atrial fibrillation or flutter: a population-based study. Arch Intern Med 2004;164:1675–1678. 5. Lloyd-Jones DM, Wang TJ, Leip EP, Larson MG, Levy D, Vasan RS, D’Agostino RB, Massaro JM, Beiser A, Wolf PA, Benjamin EJ. Lifetime risk for development of atrial fibrillation: the Framingham Heart Study. Circulation 2004;110:1042–1046. 6. Cooney MT, Vartiainen E, Laatikainen T, Laakitainen T, Juolevi A, Dudina A, Graham IM. Elevated resting heart rate is an independent risk factor for cardiovascular disease in healthy men and women. Am Heart J 2010;159:612– 619.e3. 7. Jouven X, Empana JP, Escolano S, Buyck JF, Tafflet M, Desnos M, Ducimetière P. Relation of heart rate at rest and long-term (>20 years) death rate in initially healthy middle-aged men. Am J Cardiol 2009;103:279–283. 8. Tverdal A, Hjellvik V, Selmer R. Heart rate and mortality from cardiovascular causes: a 12 year follow-up study of 379,843 men and women aged 40-45 years. Eur Heart J 2008;29:2772–2781. 9. O’Neal WT, Qureshi WT, Judd SE, Meschia JF, Howard VJ, Howard G, Soliman EZ. Heart rate and ischemic stroke: the REasons for Geographic And Racial Differences in Stroke (REGARDS) study. Int J Stroke Off J Int Stroke Soc 2015;10:1229–1235. 10. Tardif J-C, Ford I, Tendera M, Bourassa MG, Fox K, INITIATIVE Investigators. Efficacy of ivabradine, a new selective I(f) inhibitor, compared with atenolol in patients with chronic stable angina. Eur Heart J 2005;26:2529– 2536. 11. Grimsmo J, Grundvold I, Maehlum S, Arnesen H. High prevalence of atrial fibrillation in long-term endurance cross-country skiers: echocardiographic findings and possible predictors--a 28-30 years follow-up study. Eur J Cardiovasc Prev Rehabil Off J Eur Soc Cardiol Work Groups Epidemiol Prev Card Rehabil Exerc Physiol 2010;17:100–105. 12. O’Neal WT, Almahmoud MF, Soliman EZ. Resting heart rate and incident atrial fibrillation in the elderly. Pacing Clin Electrophysiol PACE 2015;38:591–597. 13. Skov MW, Bachmann TN, Rasmussen PV, Olesen MS, Pietersen A, Graff C, Lind B, Struijk JJ, Køber L, Haunsø S, Svendsen JH, Gerds TA, Holst AG, Nielsen JB. Association Between Heart Rate at Rest and Incident Atrial Fibrillation (from the Copenhagen Electrocardiographic Study). Am J Cardiol 2016;118:708–713. 14. Okin PM, Wachtell K, Kjeldsen SE, Julius S, Lindholm LH, Dahlöf B, Hille DA, Nieminen MS, Edelman JM, Devereux RB. Incidence of atrial fibrillation in relation to changing heart rate over time in hypertensive patients: the LIFE study. Circ Arrhythm Electrophysiol 2008;1:337–343. 15. Anon. The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. The ARIC investigators. Am J Epidemiol 1989;129:687–702.

9

Page 9 of 18

16. Alonso A, Agarwal SK, Soliman EZ, Ambrose M, Chamberlain AM, Prineas RJ, Folsom AR. Incidence of atrial fibrillation in whites and African-Americans: the Atherosclerosis Risk in Communities (ARIC) study. Am Heart J 2009;158:111–117. 17. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL, Jones DW, Materson BJ, Oparil S, Wright JT, Roccella EJ, National Heart, Lung, and Blood Institute Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure, National High Blood Pressure Education Program Coordinating Committee. The Seventh Report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure: the JNC 7 report. JAMA 2003;289:2560–2572. 18. Menke A, Casagrande S, Geiss L, Cowie CC. Prevalence of and Trends in Diabetes Among Adults in the United States, 1988-2012. JAMA 2015;314:1021–1029. 19. Rosamond WD, Chang PP, Baggett C, Johnson A, Bertoni AG, Shahar E, Deswal A, Heiss G, Chambless LE. Classification of heart failure in the atherosclerosis risk in communities (ARIC) study: a comparison of diagnostic criteria. Circ Heart Fail 2012;5:152–159. 20. Casale PN, Devereux RB, Alonso DR, Campo E, Kligfield P. Improved sex-specific criteria of left ventricular hypertrophy for clinical and computer interpretation of electrocardiograms: validation with autopsy findings. Circulation 1987;75:565–572. 21. Tereshchenko LG, Shah AJ, Li Y, Soliman EZ. Electrocardiographic deep terminal negativity of the P wave in V1 and risk of mortality: the National Health and Nutrition Examination Survey III. J Cardiovasc Electrophysiol 2014;25:1242–1248. 22. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concentration of low-density lipoprotein cholesterol in plasma, without use of the preparative ultracentrifuge. Clin Chem 1972;18:499–502. 23. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, Coresh J, CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration). A new equation to estimate glomerular filtration rate. Ann Intern Med 2009;150:604–612. 24. Sandvik L, Erikssen J, Thaulow E, Erikssen G, Mundal R, Rodahl K. Physical fitness as a predictor of mortality among healthy, middle-aged Norwegian men. N Engl J Med 1993;328:533–537. 25. Bettoni M, Zimmermann M. Autonomic tone variations before the onset of paroxysmal atrial fibrillation. Circulation 2002;105:2753–2759. 26. Palatini P, Julius S. Heart rate and the cardiovascular risk. J Hypertens 1997;15:3–17. 27. Agarwal SK, Norby FL, Whitsel EA, Soliman EZ, Chen LY, Loehr LR, Fuster V, Heiss G, Coresh J, Alonso A. Cardiac Autonomic Dysfunction and Incidence of Atrial Fibrillation: Results From 20 Years Follow-Up. J Am Coll Cardiol 2017;69:291–299. 28. Vazir A, Claggett B, Jhund P, Castagno D, Skali H, Yusuf S, Swedberg K, Granger CB, McMurray JJV, Pfeffer MA, Solomon SD. Prognostic importance of temporal changes in resting heart rate in heart failure patients: an analysis of the CHARM program. Eur Heart J 2015;36:669–675. 29. Alonso A, Jensen PN, Lopez FL, Chen LY, Psaty BM, Folsom AR, Heckbert SR. Association of sick sinus syndrome with incident cardiovascular disease and mortality: the Atherosclerosis Risk in Communities study and Cardiovascular Health Study. PloS One 2014;9:e109662. 30. Chang H-Y, Lin Y-J, Lo L-W, Chang S-L, Hu Y-F, Li C-H, Chao T-F, Yin W-H, Chen S-A. Sinus node dysfunction in atrial fibrillation patients: the evidence of regional atrial substrate remodelling. Eur Eur Pacing

10

Page 10 of 18

Arrhythm Card Electrophysiol J Work Groups Card Pacing Arrhythm Card Cell Electrophysiol Eur Soc Cardiol 2013;15:205–211. 31. Jensen PN, Gronroos NN, Chen LY, Folsom AR, deFilippi C, Heckbert SR, Alonso A. Incidence of and risk factors for sick sinus syndrome in the general population. J Am Coll Cardiol 2014;64:531–538.

Figure legends Figure 1. Distribution of resting heart rate among study participants at baseline (Visit 2)

Figure 2. Distribution of change in resting heart rate among study participants from Visit 2 to Visit 3.

Figure 3. Adjusted hazard ratios of atrial fibrillation by baseline resting heart rate. Each hazard ratio was computed with the median heart rate value of 65 beasts/ minute as the reference. The hazard ratio was adjusted for age (quintiles), race/center (white or black) , gender, smoking (<15 pack years, 15-35 pack years, or >35 pack years), alcohol (currently, formerly, or never), physical-activity (both sport and work) index score, education (less than high school, high school or equivalent, or college or above), annual income (< 25000 dollars, 25000-49999 dollars,  50000 dollars), insurance (yes or no), body mass index (<25kg/m2, 25-30kg/m2, >30kg/m2), systolic blood pressure (mmHg), coronary heart disease (yes or no), heart failure (yes or no), diabetes (yes or no), pre-diabetes (yes or no), left ventricular hypertrophy on electrocardiography (yes or no), left atrial enlargement on ECG (yes or no), low-density lipoprotein cholesterol (<100mg/dl, 100-159 mg/dl, or >159mg/dl), thyroidstimulating hormone (<0.56mIU/L, 0.56-1.69mIU/L, 1.7-5.1mIU/L, >5.1mIU/L), estimated glomerular filtration rate (<30milliliter/minute/1.72meter2, 30-60milliliter/minute/1.72meter2, >60milliliter/minute/1.72meter2) and Nterminal pro b-type natriuretic peptide (quintiles). Dotted lines represent the 95% confidence interval.

11

Page 11 of 18

Table 1 Characteristics of Study Participants at Visit 2 (1990-1992) Baseline Resting Heart Rate (beats per minute) pVariable <55 55-79 ≥ 80 value (n=1212) (n=9240) (n=1093) Age (years) 56.6 (5.7) 56.6 (5.7) 57.0 (5.8) 0.25 <0.00 Women 503 (41.5%) 5335 (57.7%) 697 (63.8%) 1 <0.00 Black 268 (22.1%) 2089 (22.6%) 380 (34.8%) 1 Current smoker 266 (22.0%) 2066 (22.4%) 289 (26.5%) <0.00 1 Former smoker 519 (42.9%) 3430 (37.2%) 356 (32.7%) <0.00 Pack-Years of smoking 160 (0, 525) 63 (0, 520) 100 (0, 600) 1 Current alcohol drinker 731 (60.4%) 5385 (58.4%) 545 (50.0%) <0.00 Former alcohol drinker 253 (20.9%) 1793 (19.4%) 253 (23.2%) 1 Never alcohol drinker 228 (18.8%) 2062 (22.3%) 295 (27.0%) <0.00 Sport index* 2.5 (2, 3.3) 2.3 (1.8, 3) 2.0 (1.8, 2.8) 1 <0.00 Work index* 2.3 (1.6, 3) 2.3 (1, 2.9) 2.0 (1, 2.8) 1 <0.00 Leisure index* 2.5 (2, 2.8) 2.25 (2, 2.8) 2.3 (1.8, 2.5) 1 Education Less than high school 261 (21.6%) 1830 (19.8%) 279 (25.5%) <0.00 High school degree 361 (29.9%) 3082 (33.4%) 374 (34.2%) 1 Some college 101 (8.4%) 826 (8.9%) 82 (7.5%) Bachelor’s degree or above 484 (40.1%) 3492 (37.8%) 357 (32.7%) <0.00 Health insurance 1107 (91.4%) 8462 (91.7%) 960 (88.0%) 1 26.3 26.9 28.3 (24.7, <0.00 2 Body mass index (kilogram/meter ) (23.8,29.3) (24.1,30.4) 32.5) 1 <0.00 Hypertension 245 (20.3%) 2555 (27.7%) 457 (42.1%) 1 <0.00 Diabetes mellitus 77 (6.4%) 1102 (12.0%) 328 (30.2%) 1 <0.00 Impaired fasting glucose 512 (42.5%) 4101 (44.6%) 445 (40.9%) 1 <0.00 Prevalent coronary heart disease 28 (2.3%) 240 (2.6%) 45 (4.2%) 1 <0.00 Prevalent heart failure 24 (2.0%) 277 (3.0%) 88 (8.1%) 1 <0.00 left ventricular hypertrophy on electrocardiograph 17 (1.4%) 152 (1.6%) 43 (3.9%) 1 <0.00 Left atrial enlargement on electrocardiograph 85 (7.0%) 628 (6.8%) 141 (12.9%) 1 low-density lipoprotein-cholesterol (milligram/deciliter) 130 (37) 133 (36) 134(41) 0.011 high-density lipoprotein-cholesterol 50 (17) 51 (17) 50 (18) 0.27 (milligram/deciliter) <0.00 Triglycerides (milligram/deciliter) 112(67) 131(83) 163(132) 1 thyroid-stimulating hormone <0.56 milli-international 37 (3.3%) 424 (4.9%) 80 (7.8%) units/liter <0.00 thyroid-stimulating hormone ≥5.1 1 81 (7.1%) 575 (6.6%) 53 (5.2%) milli-international units/ liter

12

Page 12 of 18

Estimated glomerular filtration rate (milliliter/minute/1.73m2)

96.1 (13.8)

97.1 (14.8)

97.9 (18.5)

0.019

<0.00 1 Categorical variables are presented as n (%). Continuous variables are presented as mean (standard deviation) or median (interquartile interval) as appropriate. Baseline characteristics were compared between groups using the Chisquared test, Kruskal-Wallis test, t-test, and Wilcoxon Rank-Sum test as appropriate. *Sports, leisure, and work index were assessed via a modified Baecke questionnaire at Visit 1 and were carried over to Visit 2. Natural log of N-terminal pro b-type natriuretic peptide

3.9 (0.9)

13

3.8 (1.0)

3.7 (1.2)

Page 13 of 18

Table 2. Incidence Rates Per 1000 Person Years and Adjusted Hazard Ratios for Atrial Fibrillation According to Baseline Resting Heart Rate Model1 Model2 Model3 Events/Number IR (95% CI) Hear rate (beats per minute) at Risk per 1000 PYs HR (95%CI) HR (95%CI) HR (95%CI) <55 206/1212 8.8 (7.7,10.1) 1.0 (0.9,1.2) 1.0 (0.9,1.2) 1.0 (0.9, 1.2) Reference 55-79 1368/9240 7.9 (7.5,8.3) Reference Reference 1.3 (1.1,1.6) 172/1093 9.5 (8.2,11.0) 1.4 (1.2, 1.6) 1.2 (1.0, 1.5)  80 Model 1: adjusted for age (quintiles), race/center (white or black) and gender; Model 2: adjusted for model1 covariates plus smoking (<15 pack years, 15-35 pack years, or >35 pack years), alcohol (currently, formerly, or never), physical-activity (both sport and work) index score, education (less than high school, high school or equivalent, or college or above), annual income (< 25000 dollars, 25000-49999 dollars, or  50000 dollars), insurance (yes or no); Model 3: adjusted for model 2 covariates plus body mass index (<25kg/m2, 2530kg/m2, >30kg/m2), systolic blood pressure (mmHg), coronary heart disease (yes or no), heart failure (yes or no), diabetes (yes or no), pre-diabetes (yes or no), left ventricular hypertrophy on electrocardiography (yes or no), left atrial enlargement on ECG (yes or no), low-density lipoprotein cholesterol (<100mg/dl, 100-159 mg/dl, or >159mg/dl), thyroid-stimulating hormone (<0.56mIU/L, 0.56-1.69 mIU/L, 1.7-5.1mIU/L, >5.1mIU/L), estimated glomerular filtration rate (<30milliliter/minute/1.72meter 2, 3060milliliter/minute/1.72meter2, >60milliliter/minute/1.72meter2) and N-terminal pro b-type natriuretic peptide (quintiles).Abbreviations: HR = hazard ratio; CI = confidence interval; IR = incidence rate; PYs = person-years

Table 3. Incidence Rates and Adjusted Hazard Ratios for Atrial Fibrillation by Change in Resting Heart Rate (from Visit 2 to 3) Events/Number IR (95%CI) Model1 Model 2 Model 3 Model 4 Variable at Risk per 1000 PYs HR (95%CI) HR (95%CI) HR (95%CI) HR (95%CI) Decrease* 49/309 10.1 (7.7,13.4) 1.1 (0.8,1.5) 1.1 (0.8,1.5) 1.0 (0.8, 1.5) 1.1 (0.8,1.5) Stable 1396/9522 8.9 (8.5,9.4) Reference Reference Reference Reference 1.1 (0.8,1.5) Increase** 55/299 12.3 (9.4,16.0) 1.6 (1.2, 2.1) 1.6 (1.2, 2.1) 1.4 (1.1, 1.9) Model 1: adjusted for age (quintiles), race/center (white or black), gender, and heart rate at Visit 2; Model 2: adjusted for model1 covariates plus smoking (<15 pack years, 15-35 pack years, or >35 pack years), alcohol (currently, formerly, or never), physical-activity (both sport and work) index score, education (less than high school, high school or equivalent, or college or above), annual income (< 25000 dollars, 25000-49999 dollars, or  50000 dollars), insurance (yes or no); Model 3: adjusted for model 2 covariates plus body mass index (<25kg/m2, 2530kg/m2, >30kg/m2), systolic blood pressure (mmHg), coronary heart disease (yes or no), heart failure (yes or no), diabetes (yes or no), pre-diabetes (yes or no), left ventricular hypertrophy on electrocardiography (yes or no), left atrial enlargement on ECG (yes or no), low-density lipoprotein cholesterol (<100mg/dl, 100-159 mg/dl, or >159mg/dl), thyroid-stimulating hormone (<0.56mIU/L, 0.56-1.69mIU/L, 1.7-5.1mIU/L, >5.1mIU/L), estimated glomerular filtration rate (<30milliliter/minute/1.72meter 2, 302 2 60milliliter/minute/1.72meter , >60milliliter/minute/1.72meter ) ,N-terminal pro b-type natriuretic peptide (quintiles), and medications affecting heart rate at Visit 3 (yes or no). Model 4: adjusted for model 3 covariates plus incident heart failure as a time-varying covariate (yes or no). *Decrease was defined as difference in resting heart rates between Visit 2 and 3 <-15 beats/minute; **Increase was defined as difference in resting heart rates between Visit 2 and 3 >15 beats/minute. Abbreviations: HR = hazard ratio; CI = confidence interval; PY = person-years; IR = incidence rate; AF = atrial fibrillation.

14

Page 14 of 18

Table 4. Adjusted Hazard Ratios for Atrial Fibrillation According to Increase in Heart Rate* by at least 15 Beats/Minute Stratified by Age, Gender, Race/center, Education, Diabetes and Hypertension Model 1 Model 2 Model 3 Model 4 Interaction ** Variable HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) P-value Age < 60years 1.8(1.3, 2.6) 1.8(1.2, 2.6) 1.9 (1.3,2.7) 1.5 (1,2.2) 0.05 1.4 (0.9, 2) 1.4 (0.9,2.1) 1.2 (0.8,1.8) 0.9 (0.6,1.4) Age  60 years Female 1.8 (1.2, 2.5) 1.8 (1.2, 2.6) 1.6 (1.1, 2.3) 1.5 (1.0, 2.2) 0.06 Male 1.4 (0.94, 2.1) 1.4 (0.92, 2.1) 1.3 (0.87, 2) 0.8 (0.5, 1.3) Black 1.9 (1,3.4) 1.9 (1,3.6) 1.8 (0.9,3.5) 1.8 (0.9,3.5) 0.15 White 1.5 (1.1, 2.1) 1.5 (1.1, 2) 1.4 (1,1.9) 1 (0.7,1.4) ≥ Bachelor’s degree 2.2 (1.5, 3.3) 2.3 (1.6, 3.6) 2.2 (1.5, 3.4) 1.6 (1.1, 2.5) 0.02 < Bachelor’s degree 1.3 (0.9, 1.8) 1.2 (0.8, 1.8) 1.1 (0.8, 1.6) 0.9 (0.6, 1.3) No diabetes 1.7 (1.3, 2.3) 1.8 (1.3, 2.4) 1.7 (1.3, 2.3) 1.4 (1, 1.9) 0.02 Diabetes 0.9 (0.5,1.7) 0.9 (0.4,1.8) 0.8 (0.4,1.6) 0.5 (0.2,1.1) No hypertension 1.7 (1.2, 2.4) 1.8 (1.3, 2.5) 1.9(1.3,2.7) 1.5 (1.1,2.1) 0.01 Hypertension 1.3 (0.8, 2) 1.2 (0.7, 1.9) 1 (0.6,1.7) 0.8 (0.5,1.3) Model 1: adjusted for age (quintiles), race/center (white or black), gender, and heart rate at Visit 2; Model 2: adjusted for model1 covariates plus smoking (<15 pack years, 15-35 pack years, or >35 pack years), alcohol (currently, formerly, or never), physical-activity (both sport and work) index score, education (less than high school, high school or equivalent, or college or above), annual income (less than 25000 dollars, 25000-49999 dollars, or 50000 dollars and more), insurance (yes or no); Model 3: adjusted for model 2 covariates plus body mass index (<25kg/m2, 25-30kg/m2, >30kg/m2), systolic blood pressure (mmHg), coronary heart disease (yes or no), heart failure (yes or no), diabetes (yes or no), pre-diabetes (yes or no), left ventricular hypertrophy on electrocardiography (yes or no), left atrial enlargement on ECG (yes or no), low-density lipoprotein cholesterol (<100mg/dl, 100-159 mg/dl, or >159mg/dl), thyroid-stimulating hormone (<0.56mIU/L, 0.56-1.69mIU/L, 1.7-5.1mIU/L, >5.1mIU/L), estimated glomerular filtration rate (<30 milliliter/minute/1.72meter 2, 30-60 milliliter/minute/1.72meter2, >60 milliliter/minute/1.72meter2), N-terminal pro b-type natriuretic peptide (quintiles), and medications affecting heart rate at Visit 3 (yes or no). Model 4: adjusted for model 3 covariates plus incident heart failure as a time-varying covariate (yes or no). * Stable heart rate (difference in heart rate -15 to 15 bpm) was the reference in the calculation of hazard ratios. **Interaction terms by each stratifier were created and tested in model 4. Abbreviations: HR = hazard ratio; CI = confidence interval; bpm = beats per minute; AF = atrial fibrillation

15

Page 15 of 18

AJC Figure1.tif

16

Page 16 of 18

AJC Figure2.tif

17

Page 17 of 18

AJC Figure3.tif

18

Page 18 of 18