- Email: [email protected]
Simple predictors for new onset atrial fibrillation
International Journal of Cardiology 221 (2016) 515–520
Contents lists available at ScienceDirect
International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard
Simple predictors for new onset atrial fibrillation Sandra Cabrera a,b,1, Ermengol Vallès a,b,⁎,1, Begoña Benito a,b,1, Óscar Alcalde a,b,1, Jesús Jiménez a,b,1, Roger Fan c,1, Julio Martí-Almor a,b,1 a b c
Electrophysiology Unit, Department of Cardiology, Hospital del Mar, Universitat Autònoma de Barcelona, Barcelona, Spain Heart Diseases Biomedical Research Group, IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain Stony Brook University School of Medicine Health Science Center, T16-80, Stony Brook, NY 11794, United States
a r t i c l e
i n f o
Article history: Received 12 May 2016 Accepted 4 July 2016 Available online 08 July 2016 Keywords: Atrial fibrillation Holter monitoring Risk calculator
a b s t r a c t Background: Predicting atrial fibrillation is a tremendous challenge. Only few studies have included 24 h-Holter monitoring characteristics to predict new onset AF (NOAF). Objectives: Our aim is to define simple predictors for NOAF. Methods: The study population included 468 patients undergoing Holter for any cause. After excluding 169 patients for history of AF prior to or during the Holter monitoring period, 299 patients were assessed for incidence of NOAF. Results: Age at inclusion was 62.5 ± 18 years (53.5% male). After a median follow up of 39.1 [IQI 36.6–40] months, the incidence of NOAF was 10.4%. With univariate analysis, age, hypertension, diabetes, renal impairment, heart failure/cardiomyopathy, left ventricle ejection fraction ≤50%, left atrium diameter ≥40 mm, CHA2DS2 VASc ≥4, premature atrial complexes (PAC) ≥0.2%, and PR interval were associated with NOAF. With multivariate analysis, age (HR 1075; p = 0.001 per year), presence of heart failure/cardiomyopathy (HR 6,16; p b 0.001), PAC ≥ 0.2% (HR 3,32; p = 0.003) and PR interval (HR 1.011; p = 0.006 per millisecond) were independent predictors for NOAF. Those predictors were used to create a risk calculator for NOAF, which was validated in an independent cohort of 200 consecutive patients with similar baseline characteristics. This new tool resulted in good discrimination capacity calculated by the C index for NOAF prediction: Area under curve (AUC) (95% CI) 0.794 (0.714– 0.875) at 2 years and 0.794 (0.713–0.875) at 3 years. Conclusions: Simple clinical, ECG and Holter monitoring parameters are able to predict NOAF in a broad population and may help guide more rigorous monitoring for atrial fibrillation. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Atrial fibrillation (AF) is the most common sustained arrhythmia in clinical practice. Its prevalence in the developed world is approximately 1.5–2% [1], and increases with age. AF is associated with a higher risk of stroke, heart failure and mortality. However, AF may be asymptomatic, and the diagnosis is often made after an adverse event has already occurred. For these reasons, AF is considered a tremendous medical challenge associated with elevated economic and social costs. Early identification of populations at higher risk for new-onset AF (NOAF) can possibly help to prevent a number of AF related complications. Risk factors already known to be associated with NOAF include age, hypertension, diabetes mellitus, obesity, ischemic heart disease,
⁎ Corresponding author at: Hospital del Mar, Department of Cardiology, Electrophysiology Unit, 25-27 Passeig Marítim, Barcelona 08003, Spain. E-mail address: [email protected] (E. Vallès). 1 This author takes responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation.
http://dx.doi.org/10.1016/j.ijcard.2016.07.077 0167-5273/© 2016 Elsevier Ireland Ltd. All rights reserved.
valvular heart disease, heart failure/cardiomyopathy (HF/CM), atrioventricular conduction impairment, chronic obstructive pulmonary disease and obstructive sleep apnea [2–8]. Several of these risk factors have been used to develop risk scores for AF prediction [9–11]. However, only few studies have used 24-hour Holter monitoring (HM) [12–16], and its role in a broader, more general population remains to be established. The aim of our study was to define and validate clinical, ECG and HM predictors of NOAF in a broad population of patients undergoing a 24hour HM for a number of different indications.
2. Methods 2.1. Original cohort We retrospectively studied a cohort of consecutive patients referred from Primary Care Physicians or the Cardiology Department for HM to investigate symptoms, ECG abnormalities or structural heart disease, between March 2011 and October 2011. The only exclusion criteria were a prior history of AF or the documentation of AF during the index HM. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institution's human research committee.
516
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
Clinical characteristics were collected in all patients. Similar to the description used by the CHA2DS2 VASC score, we recorded “heart failure” and “cardiomyopathy” in the same variable for practical purposes [17]. Heart failure was considered in the presence of 2 major Framingham criteria or 1 major criterion in conjunction with 2 minor criteria. Cardiomyopathy comprised any structural heart disease and/or at least moderate left ventricle systolic disfunction, including patients with ischemic and at least moderate valvular heart disease. Channelopaties were not included in this condition. The majority of patients had a recent transthoracic echocardiogram. If available, we recorded the left atrium diameter (LAD) measured in the parasternal axis, and the left ventricle ejection fraction (LVEF) estimated by Simpson's.
2.2. Holter monitoring and ECG characteristics ECG parameters were obtained from the baseline ECG in all patients. Twenty-fourhour Holter recordings were performed with the use of 3-channel SpaceLabs tape recorders (DMS 300–7 Holter recorder, Beijing, China). All HM were reviewed by an electrophysiologist using the ECG Holter Analyzer (CardioScan Premiere 12 Holter system, Beijing, China). The total number of premature atrial complexes (PAC), the percentage of PAC (obtained by dividing the total number of PAC by the total number of beats during the 24-hour period), the total number of episodes of non-sustained supraventricular tachycardia (NSSVT) and the maximum number of beats (MNB) in tachycardia were assessed in all patients. We defined Atrial Burden as the product of NSSVT by the MNB of the longest NSSVT.
2.3. Follow up All patients were followed in the Cardiology Outpatient Clinics or by their Primary Care Physician, and all events and reports were registered in a common electronic medical record, which included the occurrence of NOAF, adverse events and death. Occurrence of NOAF was defined as the documentation of an AF episode lasting at least 30 s, recorded by ECG, repeated HM, pacemaker, or internal loop recorder.
Table 1 Baseline characteristics of the original cohort. Population characteristics
N = 299
Clinical characteristics Male; n (%) Age; mean, SD (years) Hypertension; n (%) Diabetes mellitus; n (%) Ischemic heart disease; n (%) Valvular heart disease; n (%) HF/CM; n (%) Creatinine clearance b60 ml/min; n (%) Cerebrovascular accident; n (%) CHA2DS2 VASc score; n (%) 0 1 2–3 ≥4 Atrioventricular node blockers; n (%) Antiarrhythmic drugs; n (%) Echocardiographic parameters Left atrium diameter; mean, SD (mm) LVEF; mean, SD (%) ECG and Holter findings Percentage of PAC; median [IQI] NNSVT; median [IQI] MNB; median [IQI] Atrial burden (NSSVT × MNB); median [IQI] PR interval; mean, SD
160 (53.5%) 62.5, 17.9 156 (52.3%) 52 (17.4%) 45 (15.1%) 8 (2.6%) 27 (8%) 37 (12.6%) 21 (7%) 39 (14.2%) 76 (27.8%) 95 (34.7%) 63 (23%) 72 (24.5%) 11 (3.7%) 36.7, 6.4 62.3, 9.9 0.05 [0.05–0.24] 0 [0–2] 0 [0–5] 0 [0–10] 175, 42.5
HF/CM: heart failure/cardiomyopathy; LVEF: left ventricle ejection fraction. MNB: maximum number of beats; NNSVT: number of non sustained supraventricular tachycardia; PAC: premature atrial complexes.
2.4. Validation cohort
3.2. Follow-up and predictors of new-onset AF The clinical predictors for AF were used to create a risk calculator for NOAF. In order to validate our model, we retrospectively studied a second independent cohort of consecutive patients undergoing HM for any cause between November 2011 and March 2012, without previous history of AF or AF during the index Holter recording. This cohort was evaluated for the same clinical, ECG and HM parameters as described in the original cohort.
2.5. Statistical analyses Continuous quantitative variables are described as mean ± standard deviation (SD) if they had a Gaussian distribution, or as median and interquartilic interval (IQI) if the distribution was not normal, while categorical variables are described as frequencies and percentages. Cox-regression models were used to establish predictors associated with the development of NOAF in univariate and multivariate analyses. In the multivariate analysis, we used a stepwise backward elimination (including initial variables with P values less than 0.1 in univariate analysis). The hazard ratio (HR) was expressed with a confidence interval at 95% (95% CI). The predicted probability of NOAF at 2 and 3 years for an individual patient was calculated using the AF predictors obtained in the multivariate analysis, combined in an equation, in accordance to their HR. We validated the calculator on the new validation cohort, using Hosmer–Lemeshow test for survival data. The discrimination capacity was calculated by the C-index and the corresponding generalization of Somers' Dxy rank correlation for a censored response variable. All the analyses were carried out using SPSS 18.0 software package (SPSS Inc., Chicago, Illinois).
3. Results 3.1. Original cohort From an initial population of 468 consecutive patients undergoing 24 h Holter monitoring, 169 were excluded because of a history of prior AF or because of the presence of at least one sustained run (N30 s) of AF during the index HM. The resulting study cohort included 299 patients. The indication for HM was for symptoms in 65% of patients, predominantly for palpitations (50%) or syncope (40%). In 29%, the Holter was used to assess for arrhythmic disorders or conduction disturbances in patients with abnormal ECG at baseline, and in 6%, to evaluate for ventricular arrhythmias in patients with structural heart disease. Original cohort baseline characteristics are shown in Table 1.
All patients had repeated ECGs during the follow-up, with a mean of 6.72 (SD 3.51) ECGs per patient. In addition, 13% of patients had at least one repeated HM, and in 7%, a pacemaker or an internal loop recorder was implanted. Of the 299 patients, 31 (10.4%) developed AF during a median follow-up of 39.12 months [IQI 36.6–40]. Clinical predictors associated with the development of NOAF in univariate analysis are shown in Table 2. Multivariate analysis identified age (HR 1.09 per year; 95% CI 1.05–1.14; p b 0.001), history of HF/CM (HR 5.4; 95% CI 2.3–12.4; p b 0.001), percentage of PAC ≥0.2% (HR 2.7; 95% CI 1.2–5.8; p = 0.01) and increasing PR interval (HR 1.011; 95% CI 1.0–1.02; p = 0.006 per millisecond) as independent predictors for NOAF (Table 3). We used these 4 independent predictors to create a calculator to predict NOAF at mid-term (2 years and 3 years), with each of the 4
Table 2 Univariate analysis: variables associated with NOAF. Variables
HR (95% CI)
P
Male Age Hypertension Diabetes Creatinine clearance b 60 ml/min Ischaemic heart disease Valvular heart disease HF/CM LVEF ≥ 50% Left atrium diameter ≥ 40 mm Stroke CHA2DS2 VASc ≥ 4 PAC ≥ 0.2% Atrial burden ≥ 15 PR interval
1.7 (0.8–3.47) 1.09 (1.0–1.1) 4.71 (1.8–12.32) 2.78 (1.32–5.85) 3.29 (1.5–7.2) 2.16 (0.96–4.87) 2.7 (0.65–11.44) 4.36 (1.95–9.75) 0.33 (0.13–0.84) 2.85 (1.2–6.37) 0.93 (0.2–3.91) 7.11 (3.16–15.99) 3.64 (1.78–7.4) 2.85 (1.39–5.84) 1.08 (1.05–1.11)
0.144 b0.001 0.002 0.007 0.003 0.06 0.17 b0.001 0.02 0.01 0.92 b0.001 b0.001 0.004 b0.001
HF/CM: heart failure/cardiomyopathy; LVEF: left ventricle ejection fraction; PAC: premature atrial complexes.
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520 Table 3 Multivariate analysis: variables predicting NOAF.
Table 4 Comparison of baseline characteristics: original versus validation cohort.
Variables
HR
CI
P
Age Heart failure/cardiomyopathy PAC ≥ 0.2% PR interval
1.08 6.16 3.33 1.01
1.03–1.12 2.61–14.56 1.49–7.43 1.0–1.02
0.001 b0.001 0.003 b0.006
PAC: premature atrial complexes.
parameter's value balanced according to its HR, resulting in the following equation: PAF
at 2 or 3 years
¼ 1‐S0ðtÞ exp:ðPrognostic
517
indexÞ
where S0 (t) is the average survival probability at time and the prognostic index is the sum of the products of the predictors and their coefficients. Two examples of the use of the calculator are shown in Fig. 1. 3.3. Validation cohort In order to assess the reproducibility and external validation of our calculator, we applied it to an independent cohort of 200 consecutive patients undergoing 24 h HM without prior history of AF or AF during index HM. Baseline characteristics of the validation cohort are shown and compared with the original cohort in Table 4. Overall, there were no significant differences between the two cohorts, except for a slightly lower percentage of PAC in the validation cohort and a higher percentage of patients with renal impairment in the original cohort. When applied to the validation cohort, the calculator resulted in good performance in the prediction of NOAF. The Hosmer−Lemeshow
Variables
Original cohort (N = 299)
Validation cohort (N = 200)
Male; n (%) Age; mean, SD Hypertension; n (%) Diabetes mellitus; n (%) Ischemic heart disease; n (%) Valvular heart disease; n(%) HF/CM; n (%) Stroke; n (%) CHA2DS2VASc score; mean, SD Left atrium diameter; mean, SD (mm) LVEF; mean, SD (%) Creatinine clearance b 60 ml/min; n (%) PAC; median [IQI] NSSVT; median [IQI] MNB; median [IQI] Atrial burden ≥ 15; n (%) PR interval; mean, SD
160 (53.5%) 62.5, 17.9 156 (52.3%) 52 (17.4%) 45 (15.1%) 8 (2.7%) 24(8.1%) 21 (7.1%) 2.39, 1.68 36.7, 6.4 62.3, 9.9 37 (12.6%) 0.05 [0.05–0.24] 0 [0–2] 0 [0–5] 0 (0–10) 175.2, 42.6
104 (52%) 61.8, 17.9 97 (49%) 37 (18.5%) 21 (10.5%) 5 (2.5%) 11 (5.5%) 15 (7.5%) 2.43, 2.76 36.6, 5.9 63.2, 8.5 14 (7.29) 0.03 [0.006–0.12] 0 [0–2] 0 [I 0–4] 0 (0–9) 171.6, 50.5
HF/CM: heart failure/cardiomyopathy; LVEF: left ventricle ejection fraction; NSSVT: non sustained supraventricular tachycardia; MNB: maximum number of beats; PAC: premature atrial complexes.
test did not reveal any difference between the predicted risk of NOAF according to the calculator and observed NOAF in the validation cohort after a follow-up period of 2 and 3 years (p = 0.281 at 2 years; p = 0.327 at 3 years). Results are shown in Fig. 2. The C index was used to measure the appropriateness of the calculator for NOAF prediction during follow-up. Area under curve (AUC) (95% CI) was 0.794 (0.714– 0.875) at 2 years and 0.794 (0.713–0.875) at 3 years. To assess the behavior of each one of the variables included in the calculator, we constructed an estimate of the risk of NOAF at 3 years according to age (as a continuous variable) and the four possible categories based on PR b or ≥ 180 ms and PAC b or ≥ 0.2, both for the population with and without structural heart disease (Fig. 3). The risk of NOAF increased exponentially with aging, but the increment was higher when adding PAC ≥0.2 and PR ≥180 ms, in both groups; for all categories NOAF appeared at a younger age in the group of patients with heart failure/cardiomyopathy. 4. Discussion This study allowed us to identify simple clinical, ECG and Holter monitoring factors associated with NOAF, which combined in a risk calculator, can provide accurate prediction of 2 and 3 year risk for NOAF. The prevalence of atrial fibrillation has increased considerably in the last decade, becoming epidemic in developed countries. New technology for early detection of patients with AF is often too aggressive or expensive to be used on a widespread basis. This new prediction model can identify patients at high risk for developing NOAF in the following years, which can help to selectively guide the use of more advanced monitoring tools such as implantable loop recorders for early arrhythmia detection. 4.1. Clinical predictors for NOAF
Fig. 1. Examples using the calculator. Estimated risk of NOAF using the calculator for two different patients. HF/CM: heart failure/cardiomyopathy, NOAF: new onset atrial fibrillation, PAC: premature atrial complexes.
In our study, age N65 years and the presence of heart failure/cardiomyopathy were the only significant clinical predictors of NOAF identified with multivariate analysis, and were included in our calculator. Age N65 years was the strongest predictor for NOAF. These results are consistent with major previous studies [1–4], which have shown an increased prevalence of AF in older patients. Schnabel et al. [9] developed a risk score in a community-based cohort (from Framingham Heart Study) to predict the individual absolute risk of developing AF within the following 10 years. Similarly to our results, an increasing age correlated linearly with a higher risk of AF, with maximum risk at age
518
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
marker of atrial disease and reflect a delay in atrial electrical conduction, which can predispose to AF. Other studies have evaluated the prognostic value of the P wave characteristics for the development of AF. Interatrial block (Bachman bundle block) is manifested in an ECG with P wave duration N120 ms, and is considered advanced when a biphasic morphology appears in inferior leads. Advanced interatrial block is associated with a higher incidence of supraventricular arrhythmias, of which 50% have been reported as AF [21,22]. Several P wave features, such as P maximum, P minimum and P wave dispersion [23,24], or P wave index [25] (standard deviation of P wave duration across all the leads) have been identified as being useful predictors for paroxysmal or recurrent AF. De Bacquer et al. [26,27] highlighted the importance of the combination of a long P wave duration and morphologic changes as a potential independent risk marker for the long-term development of NOAF. We did not evaluate P wave features, because our aim was to identify simple parameters available in a standard HM. The PR interval, which includes P-wave duration, was easily identifiable and, therefore, more useful for the purpose of the study. 4.3. Holter monitoring predictors for NOAF
Fig. 2. A. Comparison of expected versus observed probability of NOAF at 2 years in the validation cohort. Differences between expected and observed probability of NOAF at 2 years using the calculator. The x axis represents the risk of AF expressed in sixtiles, from low [1] to high [6]. The y axis represents the number of patients expected to develop NOAF according to the calculator (blue) in comparison to the number of patients that actually developed AF (red) in the follow-up. No significant differences were observed between the two groups (Hosmer−Lemeshow test, p = 0.281). B. Comparison of expected versus observed probability of NOAF at 3 years in the validation cohort. Differences between expected and observed probability of NOAF at 3 years using the calculator. The X axis represents the risk of AF expressed in sixtiles, from low [1] to high [6]. The Y axis represents the number of patients expected to develop NOAF according to the calculator (blue) in comparison to the number of patients that actually developed AF (red) in the follow-up. No significant differences were observed between the two groups (Hosmer−Lemeshow test, p = 0.327).
N85 years. The anatomical and electrophysiological changes that take place in the atria related to aging can explain well the association between age and AF [1]. The presence of HF/CM has also been correlated with a higher prevalence of AF in a number of studies [5–7]. The frequent association between AF and heart failure is not only explained by the high prevalence of both conditions, but by shared pathophysiological processes and the fact that one can contribute to the development or worsening of the other.
4.2. ECG predictors for NOAF A long PR interval was an independent risk factor for NOAF. This has been validated in other studies as well [18–19]. This association can be explained by genetic and autonomic nervous system factors. Similarly PR prolongation is associated with some channelopathies, such as Brugada syndrome, which has a higher prevalence of AF as well [20]. Particularly in the elderly, PR interval prolongation may be an indirect
In our study we found a correlation between an increased percentage of PACs in the Holter monitoring and the incidence of NOAF in a broad population. Previous studies [12–13] have shown this association in patients with prior ischemic stroke. For this reason, their authors argued the efficiency of performing serial long-term ECG recordings in order to identify AF in this subgroup of high-risk patients. On the same basis, Kochhäuser et al. [14] went a step further and implanted an interval loop recorder in 70 patients with previous acute stroke and baseline HM. A strong relationship was found between PACs or non sustained supraventricular tachycardias in the Holter monitoring and NOAF diagnosed by loop recorder during long-term follow-up. Binici et al. [15] studied a healthy middle-age population with no history of cardiovascular disease or AF, who underwent a 48 h Holter monitoring and found a strong relationship between PACs and AF development, death and stroke at long-term follow up. A multicenter study [16] with more than 1200 participants found that the PAC count provided significantly superior AF risk discrimination and improved risk reclassification than the previously validated Framingham AF risk model. In our study we used the percentage of PACs in a 24 h HM, while previous studies used the number of PACs per hour or number of PACs during a 24 h period. We believe this method homogenizes the sample, adjusting for recording time. A definite cutoff value could not be found in the literature, varying from 5 to 30 per hour or N 70/24 h [12–16]. We chose our cutoff based on the best discriminatory capacity calculated from the ROC curve, which correspond roughly to 8 PACs per hour, comparable to prior studies. 4.4. Limitations We studied a broad population, including patients undergoing HM for cardiac symptoms, ECG abnormalities or prior history of structural heart disease, which induces a potential selection bias. This could explain the high incidence of NOAF observed in our sample (10.4%), which is higher than the expected for the general population at this age range [1]. NOAF was diagnosed by repeated ECGs or Holter in the majority of patients. The criteria for indication of these tests relied on the physician's discretion. More frequent or prolonged monitoring, which could have been achieved with the use of implantable loop recorders, might have increased the likelihood of detecting AF. The intermittent and often subclinical nature of AF could have caused an underestimation of the incidence of NOAF during the follow up. Finally, we did not include all known predictors for AF in our univariate analysis such as previous history of atrial flutter [28,29] or obstructive sleep
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
519
Fig. 3. A and B. Calculator behavior regarding heart failure/cardiomyopathy status. Risk of NOAF at 3 years in patients without (Fig. 3A) and with heart failure/cardiomyopathy (Fig. 3B). The X axis represents age as a continuous variable. The Y axis represents the calculated risk of NOAF at 3 years according to the calculator. There were four categories for each group considering PRb or ≥180 ms and percentage of PACb or ≥0.2. The risk of NOAF increased exponentially with aging, but the increment was higher when adding PAC ≥0.2 and PR ≥180 ms, in both groups; for all categories NOAF appeared at a younger age if heart failure/cardiomyopathy was present.
apnea, as there was a very low prevalence of these variables in our study or information was missing in other cases. 5. Conclusions We have identified simple clinical, ECG and HM risk factors that predict NOAF at mid-term follow up in a broad population of patients referred for any cardiac cause. Using these independent predictors, a precise risk calculator has been created. This model can be used in the future to identify patients at high risk for developing NOAF, who may
benefit from closer follow up and extended monitoring for early arrhythmia detection Further studies need to be performed with internal loop recorder implantation in high risk patients in order to assess whether NOAF can be detected earlier in a cost-effective and efficient manner. Conflict of interest This research was performed without financial support. There were no relationships with industry. There is no conflict of interest.
520
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ijcard.2016.07.077. References [1] A.J. Camm, G.Y. Lip, R. De Caterina, et al., ESC Committee for practice guidelines (CPG). 2012 focused update of the ESC guidelines for the management of atrial fibrillation: an update of the 2010 ESC guidelines for the management of atrial fibrillation. Developed with the special contribution of the European Heart Rhythm Association, Eur. Heart J. 33 (21) (2012) 2719–2747. [2] R. Nieuwlaat, A. Capucci, A.J. Camm, et al., Atrial fibrillation management: a prospective survey in ESC member countries: the Euro heart survey on atrial fibrillation, Eur. Heart J. 26 (2005) 2422–2434. [3] M. Nabauer, A. Gerth, T. Limbourg, et al., The registry of the German competence NETwork on atrial fibrillation: patient characteristics and initial management, Europace 11 (2009) 423–434. [4] E.J. Benjamin, D. Levy, S.M. Vaziri, R.B. D'Agostino, A.J. Belanger, P.A. Wolf, Independent risk factors for atrial fibrillation in a population-based cohort. The Framingham Heart Study, JAMA 271 (1994) 840–844. [5] W.H. Maisel, L.W. Stevenson, Atrial fibrillation in heart failure: epidemiology, pathophysiology and rationales for therapy, Am. J. Cardiol. 91 (2003) 2D–8D. [6] D. Kotecha, J.P. Piccini, Atrial fibrillation in heart failure: what should we do? Eur. Heart J. 36 (2015) 3250–3257. [7] K.C. Siontis, J.B. Geske, K. Ong, R.A. Nishimura, S.R. Ommen, B.J. Gersch, Atrial fibrillation in hypertrophic cardiomyopathy: prevalence, clinical correlations and mortality in a large high-risk population, J. Am. Heart Assoc. 3 (2014), e001002. [8] G.C. Digby, A. Baranchuk, Sleep apnea and atrial fibrillation: 2012 update, Curr. Cardiol. Rev. 8 (2012) 265–272. [9] R.B. Schnabel, L.M. Sullivan, D. Levy, et al., Development of a risk score for atrial fibrillation (Framingham Heart Study): a community-based cohort study, Lancet 373 (2009) 739–745. [10] A.M. Chamberlain, S.K. Agarwal, A.R. Folsom, et al., A clinical risk score for atrial fibrillation in a biracial prospective cohort (from the atherosclerosis risk in communities (ARIC) study), Am. J. Cardiol. 107 (1) (2011) 85–91. [11] A. Alonso, B.P. Krijthe, T. Aspelund, et al., Simple risk model predicts incidence of atrial fibrillation in a racially and geographically diverse population: the CHARGEAF consortium, J. Am. Heart Assoc. 2 (2) (2013). [12] D. Wallmann, D. Tüller, N. Kucher, J. Fuhrer, M. Arnold, E. Delacretaz, Frequent atrial premature contractions as a surrogate marker for paroxysmal atrial fibrillation in patients with acute ischemic stroke, Heart 89 (2003) 1247–1248. [13] D. Wallmann, D. Tüller, K. Wustmann, et al., Frequent atrial premature beats predict paroxysmal atrial fibrillation in stroke patients: an opportunity for a new diagnostic strategy, Stroke 38 (2007) 2292–2294.
[14] S. Kochhäuser, D.G. Dechering, R. Dittrich, et al., Supraventricular premature beats and short atrial runs predict atrial fibrillation in continuously monitored patients with cryptogenic stroke, Stroke 45 (3) (2014) 884–886. [15] Z. Binici, T. Intzilakis, O.W. Nielsen, L. Køber, A. Sajadieh, Excessive supraventricular ectopic activity and increased risk of atrial fibrillation and stroke, Circulation 121 (2010) 1904–1911. [16] T.A. Dewland, E. Vittinghoff, M.C. Mandyam, et al., Atrial ectopy as a predictor of incident atrial fibrillation: a cohort study, Ann. Intern. Med. 159 (11) (2013) 721–728. [17] G.Y. Lip, R. Nieuwlaat, R. Pisters, D.A. Lane, H.J. Crijns, Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the Euro heart survey on atrial fibrillation, Chest 137 (2) (2010) 263–272. [18] M. Cheng, X. Lu, J. Huang, S. Zhang, D. Gu, Electrocardiographic PR prolongation and atrial fibrillation risk: a meta-analysis of prospective cohort studies, J. Cardiovasc. Electrophysiol. 26 (1) (2015) 36–41. [19] J.B. Nielsen, A. Pietersen, C. Graff, et al., Risk of atrial fibrillation as a function of the electrocardiographic PR interval: results from the Copenhagen ECG study, Heart Rhythm. 10 (2013) 1249–1256. [20] K.P. Letsas, R. Weber, K. Astheimer, et al., Predictors of atrial tachyarrhythmias in subjects with type 1 ECG pattern of Brugada syndrome, Pacing Clin. Electrophysiol. 32 (4) (2009) 500–505. [21] A. Bayés de Luna, M. Cladellas, R. Oter, et al., Interatrial conduction block and retrograde activation of the left atrium and paroxysmal supraventricular tachyarrhythmia, Eur. Heart J. 9 (1988) 1112–1118. [22] D. Conde, A. Baranchuk, Interatrial block as anatomical-electrical substrate for supraventricular arrhythmias: Bayes' syndrome, Arch. Cardiol. Mex. 84 (2014) 32–40. [23] P.E. Dilaveris, E.J. Gialafos, G.K. Andrikopoulos, et al., Clinical and electrocardiographic predictors of recurrent atrial fibrillation, Pacing Clin. Electrophysiol. 23 (3) (2000) 352–358. [24] P.E. Dilaveris, J.E. Gialafos, P-wave dispersion: a novel predictor of paroxysmal atrial fibrillation, Ann. Noninvasive Electrocardiol. 6 (2) (2001) 159–165. [25] M.V. Perez, F.E. Dewey, R. Marcus, et al., Electrocardiographic predictors of atrial fibrillation, Am. Heart J. 158 (4) (2009) 622–628. [26] D. De Backquer, J. Willekens, G. De Backer, Long-term prognostic value of p-wave characteristics for the development of atrial fibrillation in subjects aged 55 to 74 years at baseline, Am. J. Cardiol. 100 (2007) 850–854. [27] D. De Backquer, J. Willekens, G. De Backer, Long-term prognostic value of p-wave characteristics for the development of atrial fibrillation in subjects aged 55 to 74 years at baseline, Am. J. Cardiol. 100 (2007) 850–854. [28] F. Philippon, V.J. Plumb, A.E. Epstein, G.N. Kay, The risk of atrial fibrillation following radiofrequency catheter ablation of atrial flutter, Circulation 92 (1995) 430–435. [29] J.S. Chinitz, E.P. Gerstenfeld, F.E. Marchlinski, D.J. Callans, Atrial fibrillation is common after ablation of isolated atrial flutter during long-term follow-up, Heart Rhythm. 4 (2007) 1029–1033.
Contents lists available at ScienceDirect
International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard
Simple predictors for new onset atrial fibrillation Sandra Cabrera a,b,1, Ermengol Vallès a,b,⁎,1, Begoña Benito a,b,1, Óscar Alcalde a,b,1, Jesús Jiménez a,b,1, Roger Fan c,1, Julio Martí-Almor a,b,1 a b c
Electrophysiology Unit, Department of Cardiology, Hospital del Mar, Universitat Autònoma de Barcelona, Barcelona, Spain Heart Diseases Biomedical Research Group, IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain Stony Brook University School of Medicine Health Science Center, T16-80, Stony Brook, NY 11794, United States
a r t i c l e
i n f o
Article history: Received 12 May 2016 Accepted 4 July 2016 Available online 08 July 2016 Keywords: Atrial fibrillation Holter monitoring Risk calculator
a b s t r a c t Background: Predicting atrial fibrillation is a tremendous challenge. Only few studies have included 24 h-Holter monitoring characteristics to predict new onset AF (NOAF). Objectives: Our aim is to define simple predictors for NOAF. Methods: The study population included 468 patients undergoing Holter for any cause. After excluding 169 patients for history of AF prior to or during the Holter monitoring period, 299 patients were assessed for incidence of NOAF. Results: Age at inclusion was 62.5 ± 18 years (53.5% male). After a median follow up of 39.1 [IQI 36.6–40] months, the incidence of NOAF was 10.4%. With univariate analysis, age, hypertension, diabetes, renal impairment, heart failure/cardiomyopathy, left ventricle ejection fraction ≤50%, left atrium diameter ≥40 mm, CHA2DS2 VASc ≥4, premature atrial complexes (PAC) ≥0.2%, and PR interval were associated with NOAF. With multivariate analysis, age (HR 1075; p = 0.001 per year), presence of heart failure/cardiomyopathy (HR 6,16; p b 0.001), PAC ≥ 0.2% (HR 3,32; p = 0.003) and PR interval (HR 1.011; p = 0.006 per millisecond) were independent predictors for NOAF. Those predictors were used to create a risk calculator for NOAF, which was validated in an independent cohort of 200 consecutive patients with similar baseline characteristics. This new tool resulted in good discrimination capacity calculated by the C index for NOAF prediction: Area under curve (AUC) (95% CI) 0.794 (0.714– 0.875) at 2 years and 0.794 (0.713–0.875) at 3 years. Conclusions: Simple clinical, ECG and Holter monitoring parameters are able to predict NOAF in a broad population and may help guide more rigorous monitoring for atrial fibrillation. © 2016 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Atrial fibrillation (AF) is the most common sustained arrhythmia in clinical practice. Its prevalence in the developed world is approximately 1.5–2% [1], and increases with age. AF is associated with a higher risk of stroke, heart failure and mortality. However, AF may be asymptomatic, and the diagnosis is often made after an adverse event has already occurred. For these reasons, AF is considered a tremendous medical challenge associated with elevated economic and social costs. Early identification of populations at higher risk for new-onset AF (NOAF) can possibly help to prevent a number of AF related complications. Risk factors already known to be associated with NOAF include age, hypertension, diabetes mellitus, obesity, ischemic heart disease,
⁎ Corresponding author at: Hospital del Mar, Department of Cardiology, Electrophysiology Unit, 25-27 Passeig Marítim, Barcelona 08003, Spain. E-mail address: [email protected] (E. Vallès). 1 This author takes responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation.
http://dx.doi.org/10.1016/j.ijcard.2016.07.077 0167-5273/© 2016 Elsevier Ireland Ltd. All rights reserved.
valvular heart disease, heart failure/cardiomyopathy (HF/CM), atrioventricular conduction impairment, chronic obstructive pulmonary disease and obstructive sleep apnea [2–8]. Several of these risk factors have been used to develop risk scores for AF prediction [9–11]. However, only few studies have used 24-hour Holter monitoring (HM) [12–16], and its role in a broader, more general population remains to be established. The aim of our study was to define and validate clinical, ECG and HM predictors of NOAF in a broad population of patients undergoing a 24hour HM for a number of different indications.
2. Methods 2.1. Original cohort We retrospectively studied a cohort of consecutive patients referred from Primary Care Physicians or the Cardiology Department for HM to investigate symptoms, ECG abnormalities or structural heart disease, between March 2011 and October 2011. The only exclusion criteria were a prior history of AF or the documentation of AF during the index HM. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki as reflected in a priori approval by the institution's human research committee.
516
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
Clinical characteristics were collected in all patients. Similar to the description used by the CHA2DS2 VASC score, we recorded “heart failure” and “cardiomyopathy” in the same variable for practical purposes [17]. Heart failure was considered in the presence of 2 major Framingham criteria or 1 major criterion in conjunction with 2 minor criteria. Cardiomyopathy comprised any structural heart disease and/or at least moderate left ventricle systolic disfunction, including patients with ischemic and at least moderate valvular heart disease. Channelopaties were not included in this condition. The majority of patients had a recent transthoracic echocardiogram. If available, we recorded the left atrium diameter (LAD) measured in the parasternal axis, and the left ventricle ejection fraction (LVEF) estimated by Simpson's.
2.2. Holter monitoring and ECG characteristics ECG parameters were obtained from the baseline ECG in all patients. Twenty-fourhour Holter recordings were performed with the use of 3-channel SpaceLabs tape recorders (DMS 300–7 Holter recorder, Beijing, China). All HM were reviewed by an electrophysiologist using the ECG Holter Analyzer (CardioScan Premiere 12 Holter system, Beijing, China). The total number of premature atrial complexes (PAC), the percentage of PAC (obtained by dividing the total number of PAC by the total number of beats during the 24-hour period), the total number of episodes of non-sustained supraventricular tachycardia (NSSVT) and the maximum number of beats (MNB) in tachycardia were assessed in all patients. We defined Atrial Burden as the product of NSSVT by the MNB of the longest NSSVT.
2.3. Follow up All patients were followed in the Cardiology Outpatient Clinics or by their Primary Care Physician, and all events and reports were registered in a common electronic medical record, which included the occurrence of NOAF, adverse events and death. Occurrence of NOAF was defined as the documentation of an AF episode lasting at least 30 s, recorded by ECG, repeated HM, pacemaker, or internal loop recorder.
Table 1 Baseline characteristics of the original cohort. Population characteristics
N = 299
Clinical characteristics Male; n (%) Age; mean, SD (years) Hypertension; n (%) Diabetes mellitus; n (%) Ischemic heart disease; n (%) Valvular heart disease; n (%) HF/CM; n (%) Creatinine clearance b60 ml/min; n (%) Cerebrovascular accident; n (%) CHA2DS2 VASc score; n (%) 0 1 2–3 ≥4 Atrioventricular node blockers; n (%) Antiarrhythmic drugs; n (%) Echocardiographic parameters Left atrium diameter; mean, SD (mm) LVEF; mean, SD (%) ECG and Holter findings Percentage of PAC; median [IQI] NNSVT; median [IQI] MNB; median [IQI] Atrial burden (NSSVT × MNB); median [IQI] PR interval; mean, SD
160 (53.5%) 62.5, 17.9 156 (52.3%) 52 (17.4%) 45 (15.1%) 8 (2.6%) 27 (8%) 37 (12.6%) 21 (7%) 39 (14.2%) 76 (27.8%) 95 (34.7%) 63 (23%) 72 (24.5%) 11 (3.7%) 36.7, 6.4 62.3, 9.9 0.05 [0.05–0.24] 0 [0–2] 0 [0–5] 0 [0–10] 175, 42.5
HF/CM: heart failure/cardiomyopathy; LVEF: left ventricle ejection fraction. MNB: maximum number of beats; NNSVT: number of non sustained supraventricular tachycardia; PAC: premature atrial complexes.
2.4. Validation cohort
3.2. Follow-up and predictors of new-onset AF The clinical predictors for AF were used to create a risk calculator for NOAF. In order to validate our model, we retrospectively studied a second independent cohort of consecutive patients undergoing HM for any cause between November 2011 and March 2012, without previous history of AF or AF during the index Holter recording. This cohort was evaluated for the same clinical, ECG and HM parameters as described in the original cohort.
2.5. Statistical analyses Continuous quantitative variables are described as mean ± standard deviation (SD) if they had a Gaussian distribution, or as median and interquartilic interval (IQI) if the distribution was not normal, while categorical variables are described as frequencies and percentages. Cox-regression models were used to establish predictors associated with the development of NOAF in univariate and multivariate analyses. In the multivariate analysis, we used a stepwise backward elimination (including initial variables with P values less than 0.1 in univariate analysis). The hazard ratio (HR) was expressed with a confidence interval at 95% (95% CI). The predicted probability of NOAF at 2 and 3 years for an individual patient was calculated using the AF predictors obtained in the multivariate analysis, combined in an equation, in accordance to their HR. We validated the calculator on the new validation cohort, using Hosmer–Lemeshow test for survival data. The discrimination capacity was calculated by the C-index and the corresponding generalization of Somers' Dxy rank correlation for a censored response variable. All the analyses were carried out using SPSS 18.0 software package (SPSS Inc., Chicago, Illinois).
3. Results 3.1. Original cohort From an initial population of 468 consecutive patients undergoing 24 h Holter monitoring, 169 were excluded because of a history of prior AF or because of the presence of at least one sustained run (N30 s) of AF during the index HM. The resulting study cohort included 299 patients. The indication for HM was for symptoms in 65% of patients, predominantly for palpitations (50%) or syncope (40%). In 29%, the Holter was used to assess for arrhythmic disorders or conduction disturbances in patients with abnormal ECG at baseline, and in 6%, to evaluate for ventricular arrhythmias in patients with structural heart disease. Original cohort baseline characteristics are shown in Table 1.
All patients had repeated ECGs during the follow-up, with a mean of 6.72 (SD 3.51) ECGs per patient. In addition, 13% of patients had at least one repeated HM, and in 7%, a pacemaker or an internal loop recorder was implanted. Of the 299 patients, 31 (10.4%) developed AF during a median follow-up of 39.12 months [IQI 36.6–40]. Clinical predictors associated with the development of NOAF in univariate analysis are shown in Table 2. Multivariate analysis identified age (HR 1.09 per year; 95% CI 1.05–1.14; p b 0.001), history of HF/CM (HR 5.4; 95% CI 2.3–12.4; p b 0.001), percentage of PAC ≥0.2% (HR 2.7; 95% CI 1.2–5.8; p = 0.01) and increasing PR interval (HR 1.011; 95% CI 1.0–1.02; p = 0.006 per millisecond) as independent predictors for NOAF (Table 3). We used these 4 independent predictors to create a calculator to predict NOAF at mid-term (2 years and 3 years), with each of the 4
Table 2 Univariate analysis: variables associated with NOAF. Variables
HR (95% CI)
P
Male Age Hypertension Diabetes Creatinine clearance b 60 ml/min Ischaemic heart disease Valvular heart disease HF/CM LVEF ≥ 50% Left atrium diameter ≥ 40 mm Stroke CHA2DS2 VASc ≥ 4 PAC ≥ 0.2% Atrial burden ≥ 15 PR interval
1.7 (0.8–3.47) 1.09 (1.0–1.1) 4.71 (1.8–12.32) 2.78 (1.32–5.85) 3.29 (1.5–7.2) 2.16 (0.96–4.87) 2.7 (0.65–11.44) 4.36 (1.95–9.75) 0.33 (0.13–0.84) 2.85 (1.2–6.37) 0.93 (0.2–3.91) 7.11 (3.16–15.99) 3.64 (1.78–7.4) 2.85 (1.39–5.84) 1.08 (1.05–1.11)
0.144 b0.001 0.002 0.007 0.003 0.06 0.17 b0.001 0.02 0.01 0.92 b0.001 b0.001 0.004 b0.001
HF/CM: heart failure/cardiomyopathy; LVEF: left ventricle ejection fraction; PAC: premature atrial complexes.
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520 Table 3 Multivariate analysis: variables predicting NOAF.
Table 4 Comparison of baseline characteristics: original versus validation cohort.
Variables
HR
CI
P
Age Heart failure/cardiomyopathy PAC ≥ 0.2% PR interval
1.08 6.16 3.33 1.01
1.03–1.12 2.61–14.56 1.49–7.43 1.0–1.02
0.001 b0.001 0.003 b0.006
PAC: premature atrial complexes.
parameter's value balanced according to its HR, resulting in the following equation: PAF
at 2 or 3 years
¼ 1‐S0ðtÞ exp:ðPrognostic
517
indexÞ
where S0 (t) is the average survival probability at time and the prognostic index is the sum of the products of the predictors and their coefficients. Two examples of the use of the calculator are shown in Fig. 1. 3.3. Validation cohort In order to assess the reproducibility and external validation of our calculator, we applied it to an independent cohort of 200 consecutive patients undergoing 24 h HM without prior history of AF or AF during index HM. Baseline characteristics of the validation cohort are shown and compared with the original cohort in Table 4. Overall, there were no significant differences between the two cohorts, except for a slightly lower percentage of PAC in the validation cohort and a higher percentage of patients with renal impairment in the original cohort. When applied to the validation cohort, the calculator resulted in good performance in the prediction of NOAF. The Hosmer−Lemeshow
Variables
Original cohort (N = 299)
Validation cohort (N = 200)
Male; n (%) Age; mean, SD Hypertension; n (%) Diabetes mellitus; n (%) Ischemic heart disease; n (%) Valvular heart disease; n(%) HF/CM; n (%) Stroke; n (%) CHA2DS2VASc score; mean, SD Left atrium diameter; mean, SD (mm) LVEF; mean, SD (%) Creatinine clearance b 60 ml/min; n (%) PAC; median [IQI] NSSVT; median [IQI] MNB; median [IQI] Atrial burden ≥ 15; n (%) PR interval; mean, SD
160 (53.5%) 62.5, 17.9 156 (52.3%) 52 (17.4%) 45 (15.1%) 8 (2.7%) 24(8.1%) 21 (7.1%) 2.39, 1.68 36.7, 6.4 62.3, 9.9 37 (12.6%) 0.05 [0.05–0.24] 0 [0–2] 0 [0–5] 0 (0–10) 175.2, 42.6
104 (52%) 61.8, 17.9 97 (49%) 37 (18.5%) 21 (10.5%) 5 (2.5%) 11 (5.5%) 15 (7.5%) 2.43, 2.76 36.6, 5.9 63.2, 8.5 14 (7.29) 0.03 [0.006–0.12] 0 [0–2] 0 [I 0–4] 0 (0–9) 171.6, 50.5
HF/CM: heart failure/cardiomyopathy; LVEF: left ventricle ejection fraction; NSSVT: non sustained supraventricular tachycardia; MNB: maximum number of beats; PAC: premature atrial complexes.
test did not reveal any difference between the predicted risk of NOAF according to the calculator and observed NOAF in the validation cohort after a follow-up period of 2 and 3 years (p = 0.281 at 2 years; p = 0.327 at 3 years). Results are shown in Fig. 2. The C index was used to measure the appropriateness of the calculator for NOAF prediction during follow-up. Area under curve (AUC) (95% CI) was 0.794 (0.714– 0.875) at 2 years and 0.794 (0.713–0.875) at 3 years. To assess the behavior of each one of the variables included in the calculator, we constructed an estimate of the risk of NOAF at 3 years according to age (as a continuous variable) and the four possible categories based on PR b or ≥ 180 ms and PAC b or ≥ 0.2, both for the population with and without structural heart disease (Fig. 3). The risk of NOAF increased exponentially with aging, but the increment was higher when adding PAC ≥0.2 and PR ≥180 ms, in both groups; for all categories NOAF appeared at a younger age in the group of patients with heart failure/cardiomyopathy. 4. Discussion This study allowed us to identify simple clinical, ECG and Holter monitoring factors associated with NOAF, which combined in a risk calculator, can provide accurate prediction of 2 and 3 year risk for NOAF. The prevalence of atrial fibrillation has increased considerably in the last decade, becoming epidemic in developed countries. New technology for early detection of patients with AF is often too aggressive or expensive to be used on a widespread basis. This new prediction model can identify patients at high risk for developing NOAF in the following years, which can help to selectively guide the use of more advanced monitoring tools such as implantable loop recorders for early arrhythmia detection. 4.1. Clinical predictors for NOAF
Fig. 1. Examples using the calculator. Estimated risk of NOAF using the calculator for two different patients. HF/CM: heart failure/cardiomyopathy, NOAF: new onset atrial fibrillation, PAC: premature atrial complexes.
In our study, age N65 years and the presence of heart failure/cardiomyopathy were the only significant clinical predictors of NOAF identified with multivariate analysis, and were included in our calculator. Age N65 years was the strongest predictor for NOAF. These results are consistent with major previous studies [1–4], which have shown an increased prevalence of AF in older patients. Schnabel et al. [9] developed a risk score in a community-based cohort (from Framingham Heart Study) to predict the individual absolute risk of developing AF within the following 10 years. Similarly to our results, an increasing age correlated linearly with a higher risk of AF, with maximum risk at age
518
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
marker of atrial disease and reflect a delay in atrial electrical conduction, which can predispose to AF. Other studies have evaluated the prognostic value of the P wave characteristics for the development of AF. Interatrial block (Bachman bundle block) is manifested in an ECG with P wave duration N120 ms, and is considered advanced when a biphasic morphology appears in inferior leads. Advanced interatrial block is associated with a higher incidence of supraventricular arrhythmias, of which 50% have been reported as AF [21,22]. Several P wave features, such as P maximum, P minimum and P wave dispersion [23,24], or P wave index [25] (standard deviation of P wave duration across all the leads) have been identified as being useful predictors for paroxysmal or recurrent AF. De Bacquer et al. [26,27] highlighted the importance of the combination of a long P wave duration and morphologic changes as a potential independent risk marker for the long-term development of NOAF. We did not evaluate P wave features, because our aim was to identify simple parameters available in a standard HM. The PR interval, which includes P-wave duration, was easily identifiable and, therefore, more useful for the purpose of the study. 4.3. Holter monitoring predictors for NOAF
Fig. 2. A. Comparison of expected versus observed probability of NOAF at 2 years in the validation cohort. Differences between expected and observed probability of NOAF at 2 years using the calculator. The x axis represents the risk of AF expressed in sixtiles, from low [1] to high [6]. The y axis represents the number of patients expected to develop NOAF according to the calculator (blue) in comparison to the number of patients that actually developed AF (red) in the follow-up. No significant differences were observed between the two groups (Hosmer−Lemeshow test, p = 0.281). B. Comparison of expected versus observed probability of NOAF at 3 years in the validation cohort. Differences between expected and observed probability of NOAF at 3 years using the calculator. The X axis represents the risk of AF expressed in sixtiles, from low [1] to high [6]. The Y axis represents the number of patients expected to develop NOAF according to the calculator (blue) in comparison to the number of patients that actually developed AF (red) in the follow-up. No significant differences were observed between the two groups (Hosmer−Lemeshow test, p = 0.327).
N85 years. The anatomical and electrophysiological changes that take place in the atria related to aging can explain well the association between age and AF [1]. The presence of HF/CM has also been correlated with a higher prevalence of AF in a number of studies [5–7]. The frequent association between AF and heart failure is not only explained by the high prevalence of both conditions, but by shared pathophysiological processes and the fact that one can contribute to the development or worsening of the other.
4.2. ECG predictors for NOAF A long PR interval was an independent risk factor for NOAF. This has been validated in other studies as well [18–19]. This association can be explained by genetic and autonomic nervous system factors. Similarly PR prolongation is associated with some channelopathies, such as Brugada syndrome, which has a higher prevalence of AF as well [20]. Particularly in the elderly, PR interval prolongation may be an indirect
In our study we found a correlation between an increased percentage of PACs in the Holter monitoring and the incidence of NOAF in a broad population. Previous studies [12–13] have shown this association in patients with prior ischemic stroke. For this reason, their authors argued the efficiency of performing serial long-term ECG recordings in order to identify AF in this subgroup of high-risk patients. On the same basis, Kochhäuser et al. [14] went a step further and implanted an interval loop recorder in 70 patients with previous acute stroke and baseline HM. A strong relationship was found between PACs or non sustained supraventricular tachycardias in the Holter monitoring and NOAF diagnosed by loop recorder during long-term follow-up. Binici et al. [15] studied a healthy middle-age population with no history of cardiovascular disease or AF, who underwent a 48 h Holter monitoring and found a strong relationship between PACs and AF development, death and stroke at long-term follow up. A multicenter study [16] with more than 1200 participants found that the PAC count provided significantly superior AF risk discrimination and improved risk reclassification than the previously validated Framingham AF risk model. In our study we used the percentage of PACs in a 24 h HM, while previous studies used the number of PACs per hour or number of PACs during a 24 h period. We believe this method homogenizes the sample, adjusting for recording time. A definite cutoff value could not be found in the literature, varying from 5 to 30 per hour or N 70/24 h [12–16]. We chose our cutoff based on the best discriminatory capacity calculated from the ROC curve, which correspond roughly to 8 PACs per hour, comparable to prior studies. 4.4. Limitations We studied a broad population, including patients undergoing HM for cardiac symptoms, ECG abnormalities or prior history of structural heart disease, which induces a potential selection bias. This could explain the high incidence of NOAF observed in our sample (10.4%), which is higher than the expected for the general population at this age range [1]. NOAF was diagnosed by repeated ECGs or Holter in the majority of patients. The criteria for indication of these tests relied on the physician's discretion. More frequent or prolonged monitoring, which could have been achieved with the use of implantable loop recorders, might have increased the likelihood of detecting AF. The intermittent and often subclinical nature of AF could have caused an underestimation of the incidence of NOAF during the follow up. Finally, we did not include all known predictors for AF in our univariate analysis such as previous history of atrial flutter [28,29] or obstructive sleep
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
519
Fig. 3. A and B. Calculator behavior regarding heart failure/cardiomyopathy status. Risk of NOAF at 3 years in patients without (Fig. 3A) and with heart failure/cardiomyopathy (Fig. 3B). The X axis represents age as a continuous variable. The Y axis represents the calculated risk of NOAF at 3 years according to the calculator. There were four categories for each group considering PRb or ≥180 ms and percentage of PACb or ≥0.2. The risk of NOAF increased exponentially with aging, but the increment was higher when adding PAC ≥0.2 and PR ≥180 ms, in both groups; for all categories NOAF appeared at a younger age if heart failure/cardiomyopathy was present.
apnea, as there was a very low prevalence of these variables in our study or information was missing in other cases. 5. Conclusions We have identified simple clinical, ECG and HM risk factors that predict NOAF at mid-term follow up in a broad population of patients referred for any cardiac cause. Using these independent predictors, a precise risk calculator has been created. This model can be used in the future to identify patients at high risk for developing NOAF, who may
benefit from closer follow up and extended monitoring for early arrhythmia detection Further studies need to be performed with internal loop recorder implantation in high risk patients in order to assess whether NOAF can be detected earlier in a cost-effective and efficient manner. Conflict of interest This research was performed without financial support. There were no relationships with industry. There is no conflict of interest.
520
S. Cabrera et al. / International Journal of Cardiology 221 (2016) 515–520
Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ijcard.2016.07.077. References [1] A.J. Camm, G.Y. Lip, R. De Caterina, et al., ESC Committee for practice guidelines (CPG). 2012 focused update of the ESC guidelines for the management of atrial fibrillation: an update of the 2010 ESC guidelines for the management of atrial fibrillation. Developed with the special contribution of the European Heart Rhythm Association, Eur. Heart J. 33 (21) (2012) 2719–2747. [2] R. Nieuwlaat, A. Capucci, A.J. Camm, et al., Atrial fibrillation management: a prospective survey in ESC member countries: the Euro heart survey on atrial fibrillation, Eur. Heart J. 26 (2005) 2422–2434. [3] M. Nabauer, A. Gerth, T. Limbourg, et al., The registry of the German competence NETwork on atrial fibrillation: patient characteristics and initial management, Europace 11 (2009) 423–434. [4] E.J. Benjamin, D. Levy, S.M. Vaziri, R.B. D'Agostino, A.J. Belanger, P.A. Wolf, Independent risk factors for atrial fibrillation in a population-based cohort. The Framingham Heart Study, JAMA 271 (1994) 840–844. [5] W.H. Maisel, L.W. Stevenson, Atrial fibrillation in heart failure: epidemiology, pathophysiology and rationales for therapy, Am. J. Cardiol. 91 (2003) 2D–8D. [6] D. Kotecha, J.P. Piccini, Atrial fibrillation in heart failure: what should we do? Eur. Heart J. 36 (2015) 3250–3257. [7] K.C. Siontis, J.B. Geske, K. Ong, R.A. Nishimura, S.R. Ommen, B.J. Gersch, Atrial fibrillation in hypertrophic cardiomyopathy: prevalence, clinical correlations and mortality in a large high-risk population, J. Am. Heart Assoc. 3 (2014), e001002. [8] G.C. Digby, A. Baranchuk, Sleep apnea and atrial fibrillation: 2012 update, Curr. Cardiol. Rev. 8 (2012) 265–272. [9] R.B. Schnabel, L.M. Sullivan, D. Levy, et al., Development of a risk score for atrial fibrillation (Framingham Heart Study): a community-based cohort study, Lancet 373 (2009) 739–745. [10] A.M. Chamberlain, S.K. Agarwal, A.R. Folsom, et al., A clinical risk score for atrial fibrillation in a biracial prospective cohort (from the atherosclerosis risk in communities (ARIC) study), Am. J. Cardiol. 107 (1) (2011) 85–91. [11] A. Alonso, B.P. Krijthe, T. Aspelund, et al., Simple risk model predicts incidence of atrial fibrillation in a racially and geographically diverse population: the CHARGEAF consortium, J. Am. Heart Assoc. 2 (2) (2013). [12] D. Wallmann, D. Tüller, N. Kucher, J. Fuhrer, M. Arnold, E. Delacretaz, Frequent atrial premature contractions as a surrogate marker for paroxysmal atrial fibrillation in patients with acute ischemic stroke, Heart 89 (2003) 1247–1248. [13] D. Wallmann, D. Tüller, K. Wustmann, et al., Frequent atrial premature beats predict paroxysmal atrial fibrillation in stroke patients: an opportunity for a new diagnostic strategy, Stroke 38 (2007) 2292–2294.
[14] S. Kochhäuser, D.G. Dechering, R. Dittrich, et al., Supraventricular premature beats and short atrial runs predict atrial fibrillation in continuously monitored patients with cryptogenic stroke, Stroke 45 (3) (2014) 884–886. [15] Z. Binici, T. Intzilakis, O.W. Nielsen, L. Køber, A. Sajadieh, Excessive supraventricular ectopic activity and increased risk of atrial fibrillation and stroke, Circulation 121 (2010) 1904–1911. [16] T.A. Dewland, E. Vittinghoff, M.C. Mandyam, et al., Atrial ectopy as a predictor of incident atrial fibrillation: a cohort study, Ann. Intern. Med. 159 (11) (2013) 721–728. [17] G.Y. Lip, R. Nieuwlaat, R. Pisters, D.A. Lane, H.J. Crijns, Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the Euro heart survey on atrial fibrillation, Chest 137 (2) (2010) 263–272. [18] M. Cheng, X. Lu, J. Huang, S. Zhang, D. Gu, Electrocardiographic PR prolongation and atrial fibrillation risk: a meta-analysis of prospective cohort studies, J. Cardiovasc. Electrophysiol. 26 (1) (2015) 36–41. [19] J.B. Nielsen, A. Pietersen, C. Graff, et al., Risk of atrial fibrillation as a function of the electrocardiographic PR interval: results from the Copenhagen ECG study, Heart Rhythm. 10 (2013) 1249–1256. [20] K.P. Letsas, R. Weber, K. Astheimer, et al., Predictors of atrial tachyarrhythmias in subjects with type 1 ECG pattern of Brugada syndrome, Pacing Clin. Electrophysiol. 32 (4) (2009) 500–505. [21] A. Bayés de Luna, M. Cladellas, R. Oter, et al., Interatrial conduction block and retrograde activation of the left atrium and paroxysmal supraventricular tachyarrhythmia, Eur. Heart J. 9 (1988) 1112–1118. [22] D. Conde, A. Baranchuk, Interatrial block as anatomical-electrical substrate for supraventricular arrhythmias: Bayes' syndrome, Arch. Cardiol. Mex. 84 (2014) 32–40. [23] P.E. Dilaveris, E.J. Gialafos, G.K. Andrikopoulos, et al., Clinical and electrocardiographic predictors of recurrent atrial fibrillation, Pacing Clin. Electrophysiol. 23 (3) (2000) 352–358. [24] P.E. Dilaveris, J.E. Gialafos, P-wave dispersion: a novel predictor of paroxysmal atrial fibrillation, Ann. Noninvasive Electrocardiol. 6 (2) (2001) 159–165. [25] M.V. Perez, F.E. Dewey, R. Marcus, et al., Electrocardiographic predictors of atrial fibrillation, Am. Heart J. 158 (4) (2009) 622–628. [26] D. De Backquer, J. Willekens, G. De Backer, Long-term prognostic value of p-wave characteristics for the development of atrial fibrillation in subjects aged 55 to 74 years at baseline, Am. J. Cardiol. 100 (2007) 850–854. [27] D. De Backquer, J. Willekens, G. De Backer, Long-term prognostic value of p-wave characteristics for the development of atrial fibrillation in subjects aged 55 to 74 years at baseline, Am. J. Cardiol. 100 (2007) 850–854. [28] F. Philippon, V.J. Plumb, A.E. Epstein, G.N. Kay, The risk of atrial fibrillation following radiofrequency catheter ablation of atrial flutter, Circulation 92 (1995) 430–435. [29] J.S. Chinitz, E.P. Gerstenfeld, F.E. Marchlinski, D.J. Callans, Atrial fibrillation is common after ablation of isolated atrial flutter during long-term follow-up, Heart Rhythm. 4 (2007) 1029–1033.