Author + information
- Received April 26, 2017
- Revision received July 6, 2017
- Accepted July 17, 2017
- Published online September 4, 2017.
- Jordan B. King, PharmD, MSa,b,
- Peyman N. Azadani, MDb,c,
- Promporn Suksaranjit, MD, MSb,
- Adam P. Bress, PharmD, MSd,
- Daniel M. Witt, PharmDe,
- Frederick T. Han, MDb,
- Mihail G. Chelu, MD, PhDb,
- Michelle A. Silver, MSPHb,
- Joseph Biskupiak, PhD, MBAe,
- Brent D. Wilson, MD, PhDb,
- Alan K. Morris, MSb,
- Eugene G. Kholmovski, PhDb,f and
- Nassir Marrouche, MDb,∗ ()
- aPharmacy Department, Kaiser Permanente Colorado, Aurora, Colorado
- bCARMA Center, Division of Cardiovascular Medicine, School of Medicine, University of Utah, Salt Lake City, Utah
- cCedars-Sinai Medical Center, Los Angeles, California
- dDepartment of Population Health Sciences, School of Medicine, University of Utah, Salt Lake City, Utah
- eDepartment of Pharmacotherapy, College of Pharmacy, University of Utah, Salt Lake City, Utah
- fUCAIR, Department of Radiology and Imaging Sciences, University of Utah, Salt Lake City, Utah
- ↵∗Address for correspondence:
Dr. Nassir F. Marrouche, CARMA Center, Division of Cardiology, University of Utah Health Sciences Center, 30 North 1900 East, Suite 4A100 SOM, Salt Lake City, Utah 84132.
Background Severity of left atrial (LA) fibrosis is a strong predictor of atrial fibrillation (AF) ablation success and has been associated with a history of stroke, hypertension, and heart failure (HF). However, it is unclear whether more severe LA fibrosis independently increases the risk of major adverse cardiovascular and cerebrovascular events (MACCE) among those with AF.
Objectives The goal of this study was to evaluate the occurrence and frequency of MACCE by strata of LA fibrosis severity in patients with AF.
Methods This was a retrospective cohort study of 1,228 patients with AF who underwent late gadolinium enhancement (LGE)–cardiac magnetic resonance imaging to quantify LA fibrosis severity between January 2007 and June 2015. Patients were stratified according to Utah stage of LA LGE criteria, and observed for the occurrence of MACCE, which included a composite of stroke or transient ischemic attack (TIA), myocardial infarction, acute decompensated HF, or cardiovascular death. Disease risk score (DRS) stratification was used to control for between-group differences in baseline characteristics and risk.
Results During follow-up, 62 strokes or TIAs, 42 myocardial infarctions, 156 HF events, and 38 cardiovascular deaths occurred. In DRS stratified analysis, the hazard ratio comparing patients with stage IV versus stage I LA LGE was 1.67 (95% confidence interval: 1.01 to 2.76) for the composite MACCE outcome. The only individual component of the MACCE outcome to remain significantly associated with advanced LGE following DRS stratification was stroke or TIA (hazard ratio: 3.94; 95% confidence interval: 1.72 to 8.98).
Conclusions This retrospective analysis demonstrated that more severe LA LGE is associated with increased MACCE risk, driven primarily by increased risk of stroke or TIA.
Left atrial (LA) fibrosis, a contributing factor to the development and progression of atrial fibrillation (AF), is increasingly recognized as an important risk marker for adverse outcomes in patients with AF (1–3). Specifically, severity of LA fibrosis is associated with previous stroke, congestive heart failure (HF), hypertension, and diabetes mellitus (4). However, a direct, temporal relationship between the severity of LA fibrosis and the incidence of major adverse cardiovascular and cerebrovascular events (MACCE) has yet to be demonstrated.
The primary objective of this study was to investigate the association of LA fibrosis, a marker of the severity of the atrial myopathy underlying AF, with the subsequent risk of MACCE. We hypothesized that among patients with AF, more severe fibrosis in the LA would be associated with a higher risk of experiencing MACCE.
Study design and sources of data
We conducted a historical cohort study of patients with AF to assess the temporal relationship between severity of LA fibrosis, defined as the quantity of late gadolinium enhancement (LGE) observed on cardiac magnetic resonance (CMR), and the incidence of MACCE. The primary exposure of interest, severity of LA LGE, was categorized according to Utah staging criteria, as described previously (5). The date of the LGE-CMR defined the study index date. The primary study outcome was the incidence of the composite of MACCE, defined as: 1) ischemic stroke or transient ischemic attack (TIA); 2) myocardial infarction (MI); 3) acute decompensated HF; or 4) cardiovascular death. Each component of the MACCE composite was also examined individually. Finally, we assessed the incidence of all-cause mortality. Patients were observed for up to 5 years from the time of LGE-CMR until occurrence of an outcome of interest, death, last encounter within the health system, or September 30, 2015, whichever occurred first. We chose to censor patients who did not experience an event of interest during follow-up at the time of their last encounter within the University of Utah Health Care system to ensure that each patient was still having clinical interactions with the health care system throughout the entire at-risk period. This minimized the risk of false negative misclassification of outcomes caused by patients leaving the state.
We used health data crosslinked from 3 sources to create the final study cohort for analysis. First, patients with AF and quantified LA LGE severity via CMR were identified from the AFib research database at the University of Utah. The AFib research database is a clinical registry of patients with AF and quantified LA LGE severity (described in detail in the following). We then used unique medical record identifiers to crosslink identified patients with complete electronic health record and administrative data within the system of the University of Utah Health Care. Finally, we used existing data linkages with the Utah Population Database to obtain vital status, including incidence and cause of death.
The AFib research database has been used and described elsewhere (4,6–10). Briefly, the AFib research database is a cohort of patients referred to the Comprehensive Arrhythmia Research and Management (CARMA) center at the University of Utah for clinical management of AF. This database is observational in nature; therefore, interactions with the CARMA center are driven by medical need. Although no fixed periods of follow-up interactions are guaranteed, several general standards are followed as outlined. Initial evaluation of patients referred for AF management includes an electrocardiogram and an ambulatory 7- to 30-day event monitor. Patients return to the outpatient clinic 4 to 6 weeks after the initial visit to review medication and/or ablation options. Patients initiated on antiarrhythmic medications receive a follow-up visit in 1 to 3 months to ensure tolerability and acceptable symptom control. Stable patients with AF who have rare episodes or are controlled with medications without ablation are routinely evaluated annually after they are initially stabilized on medications. Patients who undergo ablation return to the clinic at 4 to 6 weeks, 6 months, and 1 year post-ablation, then annually thereafter. Patients who report any symptoms are seen within a week to have an electrocardiogram and to provide an ambulatory 7- to 30-day event monitoring report.
Patients in the database are identified both as part of routine clinical care and through prospective clinical trial enrollment. Most patients (89%) undergo LGE-CMR. Database maintenance is performed regularly and systematically by 3 trained research assistants. Data collection and maintenance occurs predominantly through retrospective review of clinical records. In addition to CMR results, the database also contains detailed information regarding type, duration, and treatment of AF, echocardiography results, and catheter ablation characteristics and outcomes. Patient data are collected at each cardiovascular clinic, cardiac imaging, and electrophysiology laboratory visit or procedure.
All patients age 18 years or older at the time of LGE-CMR with a previous diagnosis of AF between January 2007 and June 2015 were potentially eligible for inclusion. Baseline clinical history, including comorbid conditions and recent health care use, was obtained from the administrative records and based on International Classification of Diseases-Ninth Revision-Clinical Modification (ICD-9-CM) and Current Procedural Terminology codes identified in the year immediately preceding the index date (Online Table 1). Patients with unquantifiable LGE severity (poor LGE-CMR image quality or LGE-CMR not performed), a previous cardiac catheter ablation, or <7 days of available follow-up were excluded from the study. Because hospitalizations that occurred outside of Utah were unlikely to be identified within the medical records of the University of Utah Health Care System, patients whose primary residences were outside of Utah were also excluded. After identifying the final study group, we stratified the cohort according to Utah stage of LA LGE for comparison (5). This study was declared exempt by the University of Utah Institutional Review Board.
Quantification of LA LGE
All studies were performed either on 1.5-T Avanto or 3-T Verio scanners (Siemens Medical Solutions, Erlangen, Germany). High-resolution LGE-CMR images for assessment of LA LGE severity using 3-dimensional electrocardiography-gated, respiratory-navigated, inversion-recovery prepared, gradient-recalled, echo pulse sequences were acquired 15 min after injecting 0.1 mmol/kg of gadolinium contrast (Mutihance, Bracco Diagnostics Inc., Princeton, New Jersey). The scan parameters for assessment of LA LGE-CMR at 3-T were as follows: axial imaging volume with field of view = 400 × 400 × 110 mm, flip angle = 14°, and echo time/repetition time = 3.1/1.4 ms. Scan parameters for assessment of LA LGE-CMR at 1.5-T were as follows: axial imaging volume with field of view = 360 × 360 × 100 mm, flip angle = 20°, and echo time/repetition time = 5.2/2.4 ms. The voxel size was 1.25 × 1.25 × 2.5 mm on both 1.5- and 3-T scanners.
Post-processing of LGE-CMR images to assess LA LGE severity was performed with Corview software (Marrek Inc., Salt Lake City, Utah). The quantification of LA LGE severity was obtained using LA segmentation and quantification protocols, as described previously (5,11). In brief, LA wall volumes were manually segmented using the subtraction of epicardial and endocardial segmentation, and manually edited to exclude the mitral valve and pulmonary veins. The final LA segmentation included the 3-dimensional extent of both the LA wall and the pulmonary vein antra. Severity of LA LGE was quantified using a threshold-based algorithm (5,11). The number of voxels in the LA wall segmentation with values above the threshold divided by the total number of voxels in the LA wall segmentation was calculated to derive the percentage of LA LGE. LA LGE severity was categorized based on Utah stages (I to IV), as described previously (5) (Figure 1).
Study outcomes and ascertainment
The primary outcome was the occurrence of MACCE >7 days after the index date. To identify the occurrence of MACCE, we cross referenced patients in the final cohort with hospitalization discharge records at the University of Utah, using a unique medical record identifier. Both fatal and nonfatal events were included in the analysis. Stroke or TIA was defined by ICD-9-CM discharge diagnosis codes 433.xx, 434.xx, or 436.xx (ischemic stroke) and 435.xx (TIA). Hemorrhagic stroke was not included in our definition. MI was defined by ICD-9-CM code 410.xx, excluding 410.x2, which refers to a subsequent episode of care. Acute decompensated HF was defined by ICD-9-CM code 428.xx, excluding 428.x2, which refers to chronic HF. Cases of both new-onset HF or exacerbation of existing HF were included. Occurrence and cause of death were determined based on death certificates obtained from the Utah Population Database. An ICD-10 code in any position on the death certificate, beginning with “I” was used to define cause of cardiovascular death.
Adjustment for confounding variables
Adjusting for confounding variables using multivariable regression strategies required at least 8 to 10 MACCE per each variable entered into the model to avoid biased effect estimates (12). To allow for more confounding variables than would be feasible with multivariable modeling, we used disease risk score (DRS) and propensity score (PS) methods (13). The DRS and PS methods are complementary model adjustment approaches that aim to balance between group differences on either the risk of the outcome (DRS) or the probability of exposure (PS) by creating summary variables that are conditional on measured patient characteristics.
We chose a nonparsimonious approach to model variable selection, as described by Glynn et al. (14). Comorbidities included in the model were components of the CHA2DS2-VASc and HAS-BLED scores, as well as other characteristics deemed clinically relevant. All variables were identified in the pre-index period, and the final models included age (as a continuous variable), sex, smoking history, previous ischemic stroke or TIA, HF, diabetes mellitus, hypertension, previous MI, coronary artery disease, AF type (paroxysmal vs. persistent), valvular AF, chronic kidney disease, hepatic dysfunction, previous venous thromboembolism, previous major bleed, year the LGE-CMR was performed, and use of oral anticoagulant, angiotensin-converting enzyme inhibitor or angiotensin II receptor blocker, beta-blocker, class I or III antiarrhythmic drug, nonsteroidal anti-inflammatory drug, aspirin, or statin therapy.
A summary DRS was calculated for each outcome of interest. Each score was estimated in the entire study group from a logistic regression analysis of the effects of the previously listed variables on the outcome of interest. This predicted the probability of experiencing the outcome of interest during the study follow-up. Regression coefficients from this model were then multiplied by individual patient characteristics, except for Utah stage, which was set to stage I (lowest severity of LA LGE) for all participants. The DRS is the sum of these products. DRS scores were classified into quintiles, with 1 representing the least risk in the cohort. Each analysis was adjusted by stratifying on DRS quintiles.
PS with matching weights
We also calculated a series of 3 PSs (Utah stage II vs. Utah stage I; Utah stage III vs. Utah stage I; Utah stage IV vs. Utah stage I) derived from 3 separate logistic regression models with Utah Stage as the dependent variable and including the previously described covariates as predictors. The PSs were implemented in our analysis by the method of matching weights (15). This method, which provides a weighting analogue of PS matching, ensures that all measured confounding variables included in the weighting model are balanced between treatment groups after weighting and minimizes selection bias.
Summary statistics and demographics
Descriptive statistics of the baseline patient characteristics were calculated overall and after stratifying by Utah stage. Trends across Utah stages were calculated by modeling the Utah stages of LA LGE as an ordinal variable. Linear regression was used for continuous variables, and logistic or multinomial logistic was used for categorical variables. Patient characteristics were also presented stratified by occurrence of MACCE in the observation period. Independent Student t tests or chi-square tests of association were used to compare means or proportions, respectively. All statistical tests were 2-sided, and p < 0.05 was considered significant. We conducted all analyses using Stata/SE software, Version 14.1 (StataCorp, College Station, Texas).
Incidence rates for all outcomes were determined as events per 1,000 years with associated 95% confidence intervals (CIs). The Kaplan-Meier method was used to calculate the cumulative incidence of each outcome by Utah stage. Cox proportional hazards models were used to calculate DRS-stratified or PS-weighted hazard ratios (HRs) for each outcome associated with the Utah stage, with stage I serving as the referent group. Plots of Schoenfeld residuals and log(−log[survival]) were examined to verify that the proportional hazards assumption was not violated. In addition, restricted cubic splines with 4 knots were used to evaluate the shape of the relationship of the hazard for the outcomes of interest with LA LGE severity as a continuous variable.
We performed 2 sensitivity analyses to assess the stability of study results to potential sources of bias. First, because LA LGE severity is known to strongly predict success of catheter ablation procedures that modify the risk of MACCE (5), we repeated all analyses that censored patients who underwent catheter ablation from the analysis at the time of procedure. Second, we repeated all analyses after excluding patients with a history of the outcome of interest to assess the impact of LA LGE on outcomes in an incident cohort.
Overall, 1,228 patients were included in the primary analysis. Characteristics of study patients by stage of LA LGE severity are described in Table 1. Most of the patients presented with lower severity of LA LGE (stage I: 34.5%; stage II: 41.4%; stage III: 19.1%; stage IV: 4.9%). Older age and female sex were associated with being in Utah stage IV, whereas patients with paroxysmal AF were more likely to be in Utah stage I. In addition, increased mean and median CHADS2 and CHA2DS2-VASc scores were associated with higher Utah stages. Patient baseline characteristics, stratified by the occurrence of MACCE during follow-up, are presented in Online Table 2. Older patients, those with persistent or permanent AF, and those with CHA2DS2-VASc scores of ≥3 were more likely to experience the primary endpoint. All measured patient characteristics were well-balanced following adjustment (Online Figure 1).
During the 5-year follow-up, 202 (16.4%) patients experienced any MACCE (62 experienced stroke or TIA, 42 experienced MI, 154 experienced an acute decompensated HF event, and 38 died from cardiovascular causes). Median follow-up was 2.8 years (maximum 5 years). Unadjusted incidence rates of MACCE and their individual components were progressively higher as LA LGE increased (Table 2). These relationships were statistically significant for the MACCE, stroke or TIA, and HF outcomes. There was a trend toward significance for both MI (p = 0.056) and cardiovascular mortality (p = 0.079) outcomes. Compared with patients with Utah stage I LA LGE severity, patients with stage IV severity had a 67% increase in risk of MACCE (HR: 1.67; 95% CI: 1.01 to 2.76) and a nearly 4-fold increased risk of stroke or TIA (HR: 3.94; 95% CI: 1.72 to 8.98) in the DRS stratified analysis. However, only the linear trend across Utah stages for stroke or TIA remained statistically significant after adjustment by DRS stratification. The HRs for the risk of MI, HF, and cardiovascular death in the DRS-adjusted analysis did not achieve statistical significance when assessing linear trends across Utah stages or when comparing stage IV with stage I severities. Results were qualitatively similar, regardless of whether a PS or DRS confounding variable adjustment modeling approach was used for all outcomes, except for stroke or TIA. The HR associated with risk of stroke or TIA among stage IV versus stage I severity was 3 times larger when using a PS strategy (HR: 9.54; 95% CI: 3.21 to 28.37) as opposed to a DRS strategy. In both analytical approaches, the risk was significantly elevated among stage IV patients. Cumulative incidence plots for MACCE and their individual components (Figure 2) also demonstrated increased rates of MACCE and stroke or TIA in Utah stage IV LA LGE severity compared with stage I.
The association between LA LGE severity expressed as a percentage (as opposed to Utah stage) and MACCE, or stroke or TIA (modeled as a continuous variable and adjusted by stratifying on the DRS) was J-shaped, with the lowest risk at approximately 10% LA LGE severity (Figure 3). The association with MI, HF, and cardiovascular death and LA LGE severity was not pronounced.
Results of the sensitivity analyses are presented in the Online Appendix. Estimates did not alter qualitatively in the subgroup of patients without a history of an event corresponding to the individual components of MACCE (Online Table 3). In all cases, the risk associated with Utah staging was more elevated when restricted to this population. Point estimates also remained qualitatively similar when censoring at time of first ablation, but confidence intervals were much wider, and no outcomes achieved statistical significance (Online Table 4). However, nearly one-half the population underwent ablation, with a mean time of 2 months from LGE-CMR to ablation. As such, there was a much smaller overall follow-up period (median follow-up of 0.5 years) and number of observed events during this analysis, which created less certainty in estimates. In the present analysis, the proportion of patients who underwent catheter ablation increased with stages of LA LGE (stage I: 42%; stage II: 45%; stage III: 54%; stage IV: 65%).
In the present observational analysis of patients with AF, there was a strong and graded association between LA LGE severity and MACCE (Central Illustration). The observed differences in MACCE were primarily driven by the stroke or TIA component of the composite, which was also the only individual component of MACCE to remain significantly associated with LA LGE after DRS and PS adjustment. MI, HF, and cardiovascular death all demonstrated numerically higher risk as LA LGE severity increased, but these increases were not statistically significant. These results add to the growing body of evidence that suggests that atrial cardiomyopathy, as quantified by LA LGE severity, rather than the AF rhythm, might be the physiological trigger associated with adverse AF sequelae (16).
It is clear that AF is strongly associated with an increased risk of ischemic stroke and thromboembolic events; conventional wisdom and some data suggest this is a causal relationship (17–20). However, some conflicting observations regarding the temporal relationship of AF and thromboembolic activity contradict this dogma (21,22). For example, while comparing AF rate control versus rhythm control strategies, Van Gelder et al. (23) noted that the patients in the rhythm control strategy group remained at risk for cardiovascular events even when sinus rhythm was maintained. Furthermore, this risk was similar regardless of whether sinus rhythm was maintained or AF recurred (23). These data are more readily understood when examined under the paradigm of LA LGE severity as an independent risk factor for stroke, rather than AF rhythm alone (16,24).
Two previous studies within the same clinical registry used here attempted to measure the association between LA LGE severity and stroke risk. First, Daccarett et al. (4) demonstrated that patients with more severe LA LGE were significantly more likely to have a history of stroke. Similarly, Akoum et al. (6) demonstrated that patients with a thrombus or spontaneous echocardiographic contrast identified by transesophageal echocardiography had significantly more severe LA LGE than those with normal findings. Taken together, these studies provided quantitative evidence that the pathophysiology and risk of stroke in AF might be associated with the severity of the LA LGE. However, both studies were limited by small sample sizes and cross-sectional study designs, which weakened the reliability of conclusions derived from the findings. The study by Daccarett et al. (4), which most closely associated LA LGE severity and stroke risk, had a main limitation that assessment of LA LGE was obtained after the stroke, with a mean time from stroke to CMR of nearly 2 years. Our study was able to overcome many of the key limitations in the previous work, namely: 1) it included a relatively large sample size; 2) it assessed LA LGE severity before the outcome of interest; and 3) it used a relatively long follow-up period.
To the best of our knowledge, this is the first cohort study that compared the incidence of MACCE in patients with AF, stratified by degrees of LA LGE severity. As such, we were able to establish a temporal and direct relationship whereby the LA myopathy occurred first, and might have predisposed patients to an elevated risk of MACCE. An additional strength of our study was the several sensitivity analyses that were performed to assess the stability of the observed associations to analytical assumptions. Because LA LGE is significantly associated with many risk factors for MACCE that might confound or mediate the relationship between LA LGE and MACCE, we altered our approach to confounding variable adjustment using both DRS and PS techniques. We also modified the follow-up approach, censoring patients at the time of first catheter ablation. Finally, we altered our exclusion criteria by removing patients with a history of the outcome of interest to identify incident cases. The association between LA LGE severity and MACCE remained significant after performing these analyses.
This was an observational and retrospective study. As such, observed effects were limited to associations, because causality could not be determined due to potential unmeasured confounding. In addition, misclassification might be present, both if the outcome was reported inadvertently or not reported. Specifically, outcomes in this analysis were based on hospitalization ICD-9 discharge diagnosis codes. Under-reporting of outcomes due to out-of-system hospitalizations was an inherent limitation in research of emergent events using a single health care system. We minimized this limitation by excluding patients who were unlikely to present to the University of Utah Health Care system during an emergent event (i.e., patients whose primary residence was outside of Utah). Also, there was no reason to assume that rates of out-of-system hospitalizations would occur differentially between the LA LGE groups. Severity of AF and baseline risk of MACCE were associated with LA LGE severity, and despite robust statistical methods, there was a risk that observed associations were caused by unmeasured confounding. We did not adjust the p values of our secondary endpoints (i.e., the individual components of the MACCE composite and mortality) for multiple comparisons. As such, the results from these comparisons were exploratory in nature and should be confirmed in an independent patient cohort. Finally, the single-center nature of the present study, and because not all centers that treat AF perform LGE-CMR or quantify LA myopathy, limited the extrapolation of our findings to other settings or systems.
In this observational cohort of patients with AF, more severe LA LGE was associated with increased risk of MACCE, particularly stroke or TIA. This study adds to the growing body of evidence on the prognostic importance of the LA as a risk factor for adverse outcomes in AF. Future longitudinal studies are needed to determine the nature of LA LGE progression, identify therapies that mitigate or reverse progression, and determine if LA LGE reversal lowers risk. Understanding these relationships could have significant implications for clinical practice.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In patients with AF, the risk of MACCE, particularly ischemic stroke, is associated with the severity of structural remodeling of the LA, as assessed by LGE on CMR.
TRANSLATIONAL OUTLOOK: Additional research is needed to assess whether LGE can improve risk stratification and guide antithrombotic therapy for patients with AF.
The authors thank the University of Utah Center for Clinical and Translational Science (CCTS) (funded by NIH Clinical and Translational Science Awards), the Pedigree and Population Resource, University of Utah Information Technology Services, and Biomedical Informatics Core for establishing the Master Subject Index between the Utah Population Database, the University of Utah Health Sciences Center, and Intermountain Health Care. Preliminary data from this project was presented in poster format at the 2015 American Heart Association Scientific Sessions.
For supplemental tables and a figure, please see the online version of this article.
Dr. Bress was supported by 1K01HL133468-01 from the National Heart, Lung, and Blood Institute; and has received research support from Novartis not related to the current project. Dr. Han has received research funding from St. Jude Medical and Boston Scientific. Dr. Chelu has received research funding from Wavelet Health, Biotronik, Medtronic, and Boston Scientific. Mr. Morris has an ownership interest in Marrek Inc. Dr. Kholmovski has been a consultant for and has ownership interest in Marrek, Inc. Dr. Marrouche has an ownership interest in Marrek, Inc. and Cardiac Designs; has performed contracted research for Biosense Webster, Medtronic, St. Jude, Biotronik, and Boston Scientific; and has received consulting fees from Biotronik and Preventice. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- confidence interval
- cardiac magnetic resonance
- disease risk score
- heart failure
- hazard ratio
- International Classification of Diseases
- left atrial
- late gadolinium enhancement
- major adverse cardiovascular and cerebrovascular event(s)
- myocardial infarction
- propensity score
- transient ischemic attack
- Received April 26, 2017.
- Revision received July 6, 2017.
- Accepted July 17, 2017.
- 2017 American College of Cardiology Foundation
- Goldberger J.J.,
- Arora R.,
- Green D.,
- et al.
- Daccarett M.,
- Badger T.J.,
- Akoum N.,
- et al.
- Segerson N.M.,
- Daccarett M.,
- Badger T.J.,
- et al.
- Badger T.J.,
- Daccarett M.,
- Akoum N.W.,
- et al.
- Oakes R.S.,
- Badger T.J.,
- Kholmovski E.G.,
- et al.
- Li L.,
- Greene T.
- Glotzer T.V.,
- Daoud E.G.,
- Wyse D.G.,
- et al.
- Vanassche T.,
- Lauw M.N.,
- Eikelboom J.W.,
- et al.
- Brambatti M.,
- Connolly S.J.,
- Gold M.R.,
- et al.,
- ASSERT Investigators
- Van Gelder I.C.,
- Hagens V.E.,
- Bosker H.A.,
- et al.,
- Rate Control versus Electrical Cardioversion for Persistent Atrial Fibrillation Study Group