Author + information
- Received December 3, 2012
- Revision received March 1, 2013
- Accepted March 19, 2013
- Published online June 25, 2013.
- Jonas Bille Nielsen, MD∗,†∗ (, )
- Claus Graff, MSc, PhD‡,
- Adrian Pietersen, MD§,
- Bent Lind, MD, DMSci§,
- Johannes Jan Struijk, MSc, PhD†,
- Morten Salling Olesen, MSc, PhD∗,†,
- Stig Haunsø, MD, DMSci∗,†⋮,
- Thomas Alexander Gerds, PhD⋮,
- Jesper Hastrup Svendsen, MD, DMSci∗,†,¶,
- Lars Køber, MD, DMSci¶ and
- Anders Gaarsdal Holst, MD, PhD∗,†
- ∗Danish National Research Foundation Centre for Cardiac Arrhythmia, Copenhagen, Denmark
- †Laboratory for Molecular Cardiology, The Heart Centre, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark
- ‡Department of Health Science and Technology, Aalborg University, Aalborg, Denmark
- §Copenhagen General Practitioners' Laboratory, Copenhagen, Denmark
- ⋮Department of Biostatistics, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark
- ¶Department of Medicine and Surgery, Faculty of Health Sciences, University of Copenhagen, Copenhagen, Denmark
- ↵∗Reprint requests and correspondence:
Dr. Jonas Bille Nielsen, Department of Cardiology, Rigshospitalet, Laboratory for Molecular Cardiology, 9312, Juliane Mariesvej 20, 2100 Copenhagen Ø, Denmark.
Objectives The aim of this study was to investigate whether the heart rate-corrected QT (QTc) interval on the electrocardiogram (ECG) is associated with the onset of atrial fibrillation (AF).
Background Patients with hereditary short-QT or long-QT syndromes, representing the very extremes of the QT interval, both seem to have a high prevalence of AF.
Methods A total of 281,277 subjects were included, corresponding to one-third of the population of the greater region of Copenhagen. These subjects underwent digital ECG recordings in a general practitioner’s core facility from 2001 to 2010. Data on drug use, comorbidities, and outcomes were collected from Danish registries.
Results After a median follow-up period of 5.7 years, 10,766 subjects had developed AF, of whom 1,467 (14%) developed lone AF. Having a QTc interval lower than the first percentile (≤372 ms) was associated with a multivariate-adjusted hazard ratio of 1.45 (95% confidence interval: 1.14 to 1.84; p = 0.002) of AF, compared with the reference group (411 to 419 ms). From the reference group and upward, the risk of AF increased with QTc interval duration in a dose-response manner, reaching a hazard ratio of 1.44 (95% confidence interval: 1.24 to 1.66, p < 0.001) for those with QTc intervals ≥99th percentile (≥464 ms). The association with respect to longer QTc intervals was stronger for the outcome of lone AF, as evidenced by a hazard ratio of 2.32 (95% confidence interval: 1.52 to 3.54, p < 0.001) for having a QTc interval ≥99th percentile (≥458 ms).
Conclusions In this large ECG study, a J-shaped association was found between QTc interval duration and risk of AF. This association was strongest with respect to the development of lone AF.
Atrial fibrillation (AF) is the most frequent cardiac arrhythmia encountered in clinical practice. The arrhythmia occurs in about 1% of the general population and is responsible for considerable morbidity and mortality, primarily because of an increased risk of stroke and heart failure (1). Despite years of extensive research, the electrophysiological mechanisms underlying AF are still far from understood.
The electrocardiographic QT interval mainly reflects cardiac ventricular repolarization, and prolongation of the heart rate–corrected QT (QTc) interval has long been recognized as a marker of sudden cardiac death, cardiovascular death, and all-cause mortality (2). However, subjects with QTc intervals at the short end of the normal range also seem to be at increased risk of death (3). The hereditary short-QT and long-QT syndromes represent the very extremes of the QTc interval. In addition to an increased risk of sudden cardiac death, patients with these rare syndromes also have a high prevalence of AF (4,5). However, whether the duration of the QTc interval is associated with AF is unknown.
This study was designed to investigate the possible association of the QTc interval with the onset of AF in the general population. This could lead to a better understanding of the pathophysiology underlying AF and could also allow the development of clinically relevant prognostic models.
In the greater region of Copenhagen, Denmark, the vast majority of general practitioners refer their patients to 1 core facility (the Copenhagen General Practitioners’ Laboratory [CGPL]) for clinical tests, such as biochemistry and electrocardiogram (ECG) recordings. The study population consists of all subjects who underwent ECG recording at CGPL at the behest of their general practitioners from 2001 to 2010. Subjects were excluded for reasons shown in Figure 1.
Further details on the ECG study population, including a comparison with the general population with respect to the incidence of AF and mortality rate, are provided in the Online Appendix.
Because our study was registry based, with no active participation from study subjects, no approval from an ethics committee was required according to Danish law. The use of registry data was approved by the Danish Data Protection Agency.
All ECGs were digitally recorded and stored in the MUSE Cardiology Information System (GE Healthcare, Wauwatosa, Wisconsin) and later processed using version 21 of the Marquette 12SL algorithm. With the use of the 12SL algorithm, we excluded ECGs with findings not consistent with a valid measurement of the QTc interval (e.g., AF and branch blocks; see the Online Appendix for validation and details on this algorithm). The QT interval was obtained as a representative median beat from all PQRST complexes in the 12 leads of the 10-s ECG tracing (see the Online Appendix for details). QT intervals were corrected for heart rate using the Framingham linear regression formula (QTcFram = QT + 154[1-60/heart rate]) or Bazett’s formula (QTc = QT/RR1/2). Left ventricular hypertrophy was defined according to Sokolow-Lyon electrocardiographic criteria as follows: 1) R-wave in lead V5 or V6 > 2.6 mV; or 2) S-wave in lead V1 + R-wave in lead V5 or V6 ≥ 3.5 mV.
Baseline variables and follow-up
With the use of Danish registries and a unique personal identification number assigned to all persons with permanent residence in Denmark, it is possible to follow subjects with respect to death, emigration, the use of prescription medications, and any hospital, ambulatory, or emergency department discharge diagnoses (6–10). With the use of such registry data and data from biochemical tests conducted at CGPL, we identified subjects with the following baseline variables: hypertension, heart failure, myocardial infarction, valvular heart disease, diabetes, hyperthyroidism, and treatment with QT interval–prolonging drugs on the day of ECG recording (yes or no) (Online Table 1). In brief, hypertension was defined as present if a subject at some point before study inclusion was treated simultaneously with at least 2 kinds of antihypertensive drugs (11). Heart failure was defined as a combination of a discharge diagnosis of heart failure and treatment with loop diuretics (11). Myocardial infarction and valvular heart disease were defined from discharge diagnoses. Diabetes and hyperthyroidism were defined as present if a subject at some point before study inclusion had a discharge diagnosis of diabetes or hyperthyroidism, in case of a blood sample taken at the behest of a general practitioner indicative of disease, or in case of a purchase of prescription medication used for management of 1 of the 2 diseases. A hospital, ambulatory, or emergency department discharge diagnosis of AF or atrial flutter was the primary event of interest. Lone AF was defined as the occurrence of AF before the age of 65 years and in the absence of hypertension, heart failure, myocardial infarction, valvular heart disease, diabetes, and hyperthyroidism. Detailed information on the identification of covariates and clinical outcomes in the Danish registries is provided in the Online Appendix.
Age was used as the time scale in all survival analyses. Events of AF were recorded after the day of the first ECG. Follow-up ended in case of AF, death, or emigration or on October 8, 2011, which was the end of complete follow-up data acquisition. Crude incidence rates of AF per 1,000 person-years were calculated using Poisson regression. Cox regression was used to assess the association of QTc interval, measured on the day of inclusion (first ECG), with the instantaneous risk (hazard) of AF. Cox regression was adjusted for conventional risk factors that were obtained at the age of inclusion. To provide detailed analysis of the functional relationship between QTc interval duration and the risk of AF, we used 2 alternative approaches. First, we divided the population (or subpopulations) into 9 risk categories on the basis of the population QTc interval distribution, with cutoffs at the first, fifth, 20th, 40th, 60th, 80th, 95th, and 99th percentiles. In these analyses, the QTc interval category with the lowest risk of AF was used as the reference group. Second, to describe the functional relationship between QTc interval duration and the risk of AF in a more unbiased way, we also performed restricted cubic regression spline analysis with 4 knots located at the fifth, 35th, 65th, and 95th percentiles (12).
Subgroup analyses were performed, including an analysis with lone AF as an outcome. For this analysis, patients with hypertension, heart failure, myocardial infarction, valvular heart disease, diabetes, hyperthyroidism, and/or age >65 years at baseline were excluded. An event (diagnosis) of hypertension, heart failure, myocardial infarction, valvular heart disease, diabetes, or hyperthyroidism or surpassing 65 years of age during the follow-up period was considered a competing risk of lone AF, and event time was set to the earliest of these events. To explore the association of QTc interval and the risk of AF in patients with cardiovascular disease, another subgroup analyses was performed including only subjects who had hypertension, heart failure, myocardial infarction, and/or valvular heart disease at the age of inclusion. Sensitivity analyses are described in the Online Appendix.
We considered a 2-tailed p value <0.05 as statistically significant. Proportional hazards assumptions were checked and accepted for all covariates. All analyses were conducted using Stata version 12.0 (StataCorp LP, College Station, Texas).
The greater region of Copenhagen has a current population of 1.18 million citizens. Among them, 326,959 subjects (about 28%) had 1 or more ECGs recorded at CGPL during the 10-year period from 2001 to 2010. Of the subjects referred for ECG recording, a total of 281,277 (86%) fulfilled the study inclusion criteria (Fig. 1). Baseline clinical characteristics of the study population are shown in Table 1. Follow-up was 100% with regard to clinical end points, mortality, and emigration.
Risk of AF
Total Study Population
The median follow-up time for the total study population was 5.7 years (interquartile range: 3.2 to 8.4 years), corresponding to 1,614,832 person-years. During follow-up, 10,766 subjects developed AF, and 26,974 died. The multivariate-adjusted analysis exploring the association between categories of QTc interval duration and the risk of AF revealed that the risk of AF increased in a dose-response manner from 420 ms and upward (Fig. 2). Compared with the reference group (40th to <60th percentile, 411 to 419 ms), the risk of AF reached a hazard ratio of 1.44 (95% confidence interval: 1.24 to 1.66; p < 0.001) for those with QTcFram intervals ≥99th percentile (≥464 ms). At the lower range of the interval, only subjects with QTcFram intervals lower than the first percentile (≤372 ms) had a statistically significant increased risk of AF (hazard ratio: 1.45; 95% confidence interval: 1.14 to 1.84, p = 0.002). The multivariate-adjusted restricted cubic spline analysis confirmed the J-shaped association obtained from the analysis based on QTc categories (Fig. 3). In this spline-based analysis, subjects with QTcFram intervals of 404 ms had the lowest risk of AF, whereas the risk increased for both longer and shorter QTcFram intervals.
For the lone AF subgroup analysis, 175,738 subjects were included at baseline, and of these, 1,467 developed lone AF. The median follow-up time for the lone AF model was 4.3 years (interquartile range: 2.1 to 7.0 years), corresponding to 816,322 person-years. This subgroup analysis revealed that the association between QTc interval and the outcome of lone AF was stronger compared with AF, at least for QTc intervals in the upper range (Figs. 2 and 3).
AF in Patients With Cardiovascular Disease
For the subgroup analysis of risk of AF in patients with cardiovascular disease, 47,777 subjects were included at baseline, 3,879 developed AF, and the median follow-up time was 5.5 years (interquartile range: 2.9 to 8.3 years), corresponding to 266,177 person-years. Exploring the association between QTc interval and the risk of AF in this subgroup revealed a weaker association with respect to longer QTc intervals compared with results from the total study population. For shorter QTc intervals, the association was similar to the results from the total population (Figs. 2 and 3).
Fifty-six percent of the study population had available cholesterol profiles at baseline. Adding the total/high-density lipoprotein cholesterol ratio into the model for this subgroup did not change the pattern of association (Online Fig. 1). In a time-dependent model in which status regarding hypertension, heart failure, myocardial infarction, valvular heart disease, diabetes, and hyperthyroidism was updated over time, the association between longer QTc intervals and the risk of AF was slightly weaker than the model including only baseline data (Online Fig. 1).
Because QTc interval prolongation is a well-known predictor of cardiovascular mortality (2), it is important to raise the question of whether QTc interval prolongation is just a marker of cardiac disease, such as ischemic heart disease and heart failure, and not directly associated with AF. Therefore, beyond adjusting for the noted comorbidities, we also investigated this aspect by looking at lone AF, that is, younger subjects (<65 years of age) free of cardiovascular disease. As noted, we found that the association for QTc intervals in the longer range was strongest for subjects who developed lone AF, less so for the whole population, and weakest, but still present, for the population with cardiovascular comorbidities. This result indicates that the QTc interval’s association with AF is not due simply to confounding by traditional cardiovascular risk factors; instead, it seems to be due to inherent characteristics or remodeling of a patient’s cardiac electrophysiology.
All ECGs were analyzed digitally using clinically validated software, thus avoiding any intraobserver or interobserver variability. The 12SL algorithm is widely used, has been approved by the U.S. Food and Drug Administration for use in pharmaceutical studies, and has been validated extensively with regard to QT interval measurement (13–15). Additionally, the 12SL algorithm has recently been shown to have a much better intrasubject reproducibility than manual QTc measurement (15). In addition, our algorithm for excluding ECGs not suited for measurement of QTc interval was found to be valid (Online Appendix).
We applied the Framingham formula for heart rate correction of the QT interval because this formula is widely used, because it is based on empirical data from a large population sample rather than on hypothetical reasoning (16), and because linear regression functions rather than Bazett’s formula are recommended in recent guidelines (17). However, for comparability with previous studies of QTc interval, we also provide results for the widely used Bazett’s formula. Using the 2 different formulas resulted in similar conclusions (Fig. 2).
Although the present study was not designed to evaluate the biology linking QTc interval duration and AF, our data raise several interesting pathophysiological questions. To our knowledge, there is no definitive evidence of direct proportionality between atrial and ventricular action potential duration (closely correlated with the QTc interval). However, messenger ribonucleic acid expression studies have shown that the expression of important ion channel complexes involved in cardiac repolarization is very similar in atrial and ventricular tissues (18,19), and one must assume that atrial and ventricular repolarization duration are related. The prevailing conceptual model for AF genesis describes shortened atrial action potential duration as a substrate for multiple-circuit reentry excitation in the atria as the mechanism of arrhythmia (20). In support of this model, many drugs used for rhythm control or termination of AF prolong the action potential duration and, with this, the QTc interval (21). This is in line with our data showing that subjects with QTc intervals at the very short end of the normal range have increased risk of AF and also that patients with the rare short-QT syndrome have a high prevalence of AF (5). In contrast, our data, which mainly show an increased risk of AF with longer QTc intervals, are in conflict with the prevailing theory of AF genesis. We speculate that a mechanism of “atrial torsades de pointes” could mediate the susceptibility to AF, similar to the mechanism thought to underlie ventricular arrhythmias in long-QT syndrome. In support of this hypothesis, a recent mouse model of long-QT syndrome type 3 has indicated that a substrate of prolonged atrial action potential duration and the occurrence of triggers in form of early after-depolarizations can promote AF (22). In addition, it has been documented that patients with long-QT syndrome, not only have prolonged atrial action potential duration (23), but also have a much higher prevalence of early-onset lone AF compared with the background population (4). In addition, we recently showed that patients with early-onset lone AF carry a high prevalence of genetic variants previously associated with long-QT syndrome type 3 (24). Finally, both patients with AF after heart failure and those with lone AF remote from arrhythmic episodes seem to have prolonged atrial refractory periods obtained by intracardiac recordings compared with controls (25,26). Accordingly, our data indicate that both shortened and prolonged action potential duration could be a substrate for AF not only in rare cases of short-QT and long-QT syndromes but also in the general population.
Interestingly, it has lately been documented that QTc interval prolongation is associated with an increased risk of stroke (27,28). However, the mechanism for this association has not yet been identified. Our data indicate that AF could be the link between QTc interval prolongation and stroke.
For data on morbidity, mortality, and medication use, we relied on Danish registries, and for some diagnoses, we do not know their validity in these registries. However, with regard to the most important diagnoses used in this study, these have been validated. Accordingly, the Danish registry-based diagnosis of AF has recently been found to have a positive predictive value of 93% for electrocardiographically documented AF by manual review of patient records (29), and the diagnosis has additionally been found to be valid in a number of other studies (30–32). This is also the case with regard to our registry-based definitions of hypertension (11), heart failure (33), and myocardial infarction (34).
Although we found the strongest association with respect to the outcome of lone AF, we cannot exclude some residual confounding, because we lack information on important anthropometric data on our study population, such as body mass index, blood pressure, and smoking status. However, we were able to adjust for several cardiovascular risk factors and diseases that are likely intermediate phenotypes for the possible confounding effect of body mass index, blood pressure, and smoking status on the association between QTc interval duration and the risk of AF. We adjusted for treatment with antihypertensive medication as a proxy for hypertension, and by using this approach, we cannot exclude residual confounding by untreated hypertension. However, we believe that our definition of hypertension might be as good as controlling for a single blood pressure measurement, as done in many cohort studies, because such a blood pressure measurement is not always a good estimate of a subject’s true resting blood pressure (35–37). Additionally, we developed a model looking at the subgroup with cholesterol level measurements and found that adding the total/high-density lipoprotein cholesterol ratio to the model for this subgroup did not change the pattern of association (Online Fig. 1). Moreover, applying a time-dependent model in which statuses regarding covariates were updated over time only attenuated the association slightly with respect to longer QTc intervals (Online Fig. 1).
We observed an increased incidence rate of AF in the electrocardiographic study population compared with the general population (6.7 vs. 5.3 per 1,000 person-years, respectively) (Online Appendix), indicating a slight selection bias with regard to the outcome of AF. However, our incidence rate of AF is in line with those from previous reports (38,39). Additionally, the study population was not much sicker than the general population, as demonstrated by an only slightly increased mortality rate in the study population compared with the general population (16.7 vs. 16.0 per 1,000 person-years, respectively; see the Online Appendix). Although our study population represents a sample with a higher rate of the outcome in question (AF) and a slightly higher mortality, most nonopportunistic cohort studies likely suffer from the opposite issue in the form of an often pronounced healthy responder bias (40,41).
Finally, because our population was recruited from a mainly white population, it is uncertain whether our findings can be extrapolated to other ethnicities.
The present study population is the largest primary care population studied to date, encompassing about a third of the population in a capital region, with high-quality digital ECGs and almost complete follow-up. In this population, we found a robust J-shaped association between QTc interval duration and the occurrence of AF. With regard to the upper range of the QTc interval, this association was most pronounced for lone AF.
This study was supported by the University of Copenhagen, the Danish National Research Foundation, The Danish Council for Independent Research (Grant No. 11-107456), The John and Birthe Meyer Foundation, The Foundation of 17-12-1981, The Willadsen Family Foundation, The Beckett Foundation, and the Copenhagen Medical Society. Dr. Svendsen is a member of the advisory board of Medtronic; and has received research grants from Medtronic, Biotronik, St. Jude Medical, and Boston Scientific Corporation. Dr. Kober was a speaker at symposia organized by Servier and AstraZeneca. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- Copenhagen General Practitioners’ Laboratory
- heart rate–corrected QT interval
- QT interval corrected for heart rate using the Framingham formula (QTcFram = QT + 154[1-60/heart rate])
- Received December 3, 2012.
- Revision received March 1, 2013.
- Accepted March 19, 2013.
- American College of Cardiology Foundation
- Giustetto C.,
- Schimpf R.,
- Mazzanti A.,
- et al.
- Frank L.
- Lynge E.,
- Sandegaard J.L.,
- Rebolj M.
- Helweg-Larsen K.
- Kildemoes H.W.,
- Sørensen H.T.,
- Hallas J.
- Pedersen C.B.
- Olesen J.B.,
- Lip G.Y.H.,
- Hansen M.L.,
- et al.
- Orsini N.,
- Greenland S.
- Rautaharju P.M.,
- Surawicz B.,
- Gettes L.S.,
- et al.
- Zimetbaum P.
- Lemoine M.D.,
- Duverger J.E.,
- Naud P.,
- et al.
- Olesen M.S.,
- Yuan L.,
- Liang B.,
- et al.
- Sanders P.,
- Morton J.B.,
- Davidson N.C.,
- et al.
- Stiles M.K.,
- John B.,
- Wong C.X.,
- et al.
- Cardoso C.R.L.,
- Salles G.F.,
- Deccache W.
- Soliman E.Z.,
- Howard G.,
- Cushman M.,
- et al.
- Mukamal K.J.,
- Tolstrup J.S.,
- Friberg J.,
- Jensen G.,
- Grønbaek M.
- Stergiou G.S.,
- Baibas N.M.,
- Gantzarou A.P.,
- et al.
- Warren R.E.,
- Marshall T.,
- Padfield P.L.,
- Chrubasik S.
- Heeringa J.,
- Van der Kuip D.A.M.,
- Hofman A.,
- et al.
- Drivsholm T.,
- Eplov L.F.,
- Davidsen M.,
- et al.