Treatment of Atrial Fibrillation and Concordance With the American Heart Association/American College of Cardiology/Heart Rhythm Society Guidelines
Findings From ORBIT-AF (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation)
Background: It is unclear how frequently patients with atrial fibrillation receive guideline-concordant (GC) care and whether guideline concordance is associated with improved outcomes.
Methods and Results: Using data from ORBIT-AF (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation), we determined how frequently patients received care that was concordant with 11 recommendations from the 2014 American Heart Association/American College of Cardiology/Heart Rhythm Society atrial fibrillation guidelines pertaining to antithrombotic therapy, rate control, and antiarrhythmic medications. We also analyzed the association between GC care and clinical outcomes at both the patient level and center level. A total of 9570 patients were included. The median age was 75 years (interquartile range, 67–82), and the median CHA2DS2-VASc score was 4 (interquartile range, 3–5). A total of 5977 patients (62.5%) received care that was concordant with all guideline recommendations for which they were eligible. Rates of GC care were higher in patients treated by providers with greater specialization in arrhythmias (60.0%, 62.4%, and 67.0% for primary care physicians, cardiologists, and electrophysiologists, respectively; P<0.001). During a median of 30 months of follow-up, patients treated with GC care had a higher risk of bleeding hospitalization (hazard ratio=1.21; P=0.021) but a similar risk of death, stroke, major bleeding, and all-cause hospitalization.
Conclusions: Over a third of patients with atrial fibrillation in this large outpatient registry received care that differed in some respect from guideline recommendations. There was no apparent association between GC care and improved risk-adjusted outcomes.
See Editorial by Gula et al
WHAT IS KNOWN?
Clinical practice guidelines are commonly used to guide the management of patients with atrial fibrillation.
There are limited data on how frequently providers prescribe care that is consistent with guideline recommendations and whether guideline-concordant care is associated with improved patient outcomes.
WHAT THE STUDY ADDS?
Approximately one third of patients with atrial fibrillation receive care that is not in agreement with at least one guideline recommendation.
Guideline concordance is higher in patients treated by providers with greater specialization in arrhythmias.
Guideline-concordant care, as measured in this study, was not associated with improved clinical outcomes.
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia and is associated with significant morbidity, mortality, and socioeconomic burden.1 Management of AF encompasses several key aspects of care, including stroke prevention, rate control, and rhythm control. Since 2001, the American College of Cardiology (ACC) and American Heart Association (AHA) have published clinical practice guidelines on the management of AF.2 Recently, the Heart Rhythm Society (HRS) joined as key stakeholder. These guidelines provide evidence-based recommendations on the evaluation and management of patients with AF.
Although the AHA/ACC/HRS AF guidelines are widely used to guide patient care, it is unclear how frequently patients with AF receive care that is consistent with guideline recommendations. Previous studies on antithrombotic therapy and antiarrhythmic medications in patients with AF have found widely varying rates of guideline consistency by providers.3–8 Furthermore, outside of oral anticoagulation, the association between guideline-concordant (GC) care and clinical outcomes in AF is largely unexplored.4,9
The primary objective of this study was to assess how frequently patients with AF are prescribed therapies consistent with guideline recommendations using a nationwide cohort. We also sought to determine whether GC care, both within specific areas of AF care and overall, is associated with improved patient outcomes.
The ORBIT-AF (Outcomes Registry for Better Informed Treatment of AF) has been described previously.10 Briefly, ORBIT-AF is a registry of outpatients with AF enrolled at 176 different sites within the United States. Sites included a diverse array of provider specialties, including primary care, general cardiology, and electrophysiology. An adaptive design was used to recruit a nationally representative sample of sites, with heterogeneity of geography and provider type. Patients were enrolled from June 29, 2010 to August 09, 2011 and were followed up every 6 months for at least 2 years. Inclusion criteria were age ≥18 years and electrocardiographically documented AF. Patients were excluded if AF was because of a reversible cause (eg, in the setting of cardiac surgery or hyperthyroidism) or if life expectancy was <6 months. A web-based case report form was used to gather data. The primary sources were the patients’ medical records and the treating physician. Data entry was performed by trained clinical research coordinators. All case report forms were approved by the primary investigator before incorporation into the main database. Site management and study coordination were performed by the Duke Clinical Research Institute.
Data elements collected included patient demographics, medical history, components of AF history (including prior treatment and symptoms), medical therapies, vital signs, laboratory and imaging measures, and incident procedures and adverse events. Detailed medication data were collected, including oral anticoagulation use, monitoring, and international normalized ratio levels. Hospitalization and cardiovascular procedures (catheter ablation, cardioversion, and cardiac device implantation) were also recorded in follow-up. The primary outcome in ORBIT was the occurrence of stroke. All primary outcome events were confirmed with source documentation (eg, hospital discharge summary). Secondary outcomes included major adverse cardiovascular events (death, myocardial infarction, new-onset heart failure, revascularization, and systemic embolism), bleeding events, and cause-specific hospitalization. All clinical events (including hospitalizations) were recorded from the patient’s medical record. In the event of healthcare events outside of the primary institution, sites were instructed to obtain documentation from the outside institutions. Other potentially drug-related adverse events, such as organ toxicity, proarrhythmia, or other intolerance, were not recorded, unless they resulted in one of the above.
A total of 10 137 patients with electrocardiographically documented AF were identified who were enrolled in ORBIT-AF from June 29, 2010 to August 9, 2011 with complete baseline data. Patients were excluded if no follow-up data were available (n=388), a rate or rhythm control strategy was not specified by the provider (n=25), or if the patient was taking an oral anticoagulant other than warfarin or dabigatran (n=18). Patients were also excluded if overall guideline concordance could not be assessed based on patient characteristics or medical history, as described below (n=136). All subjects provided written, informed consent. Institutional review boards of all participating enrollment sites approved the study.
Selection and Assessment of Guideline Recommendations
Recommendations in clinical practice guidelines vary in strength and level of supporting evidence. For this analysis, we chose to include only strong recommendations (class I and class III) with a high level of supporting evidence (level A and level B). Class II recommendations, which are subject to uncertainty about the recommended procedure or treatment, were not included. Similarly, we did not include recommendations supported by level C evidence because there is little or no evidence from high-quality clinical trials that following these recommendations impacts clinical outcomes.
The process used to select recommendations for this analysis was as follows. All class I and class III recommendations with level A or level B supporting evidence from the 2014 AHA/ACC/HRS guidelines on AF were abstracted.2 Recommendations were excluded if they could not be assessed based on the clinical data available from ORBIT-AF. For example, we could not assess recommendations pertaining to inpatient management of AF. Additional recommendations on appropriate use of specific antiarrhythmic medications were derived from the guideline text, figures, and tables, as detailed in the right-most column of Table I in the Data Supplement. The recommendation on the absolute creatinine clearance cutoff for sotalol was taken from the package insert.11 The 2014 guidelines use varied language on coronary artery disease and the use of flecainide and propafenone. The guideline text states that flecainide should be avoided in patients with myocardial infarction from coronary artery disease, whereas Table 13 and Figure 7 in the guideline document list coronary artery disease as a contraindication.2 For the purposes of this study, any history of coronary artery disease was deemed to be a contraindication to flecainide and propafenone.
This process resulted in a total of 11 recommendations that could be assessed using data from ORBIT-AF (Table 1). Of note, while patients were enrolled in ORBIT-AF before publication of the 2014 guidelines, nearly all recommendations in this analysis were present in similar form in the 2006 and 2011 guideline editions (Table I in the Data Supplement). A notable difference is the use of the CHADS2 score in the 2006 and 2011 guidelines, which was replaced by the CHA2DS2-VASc score in the 2014 guidelines. Recommendation 11 in Table 1 was not present in the 2006 or 2011 editions.
Assessment of recommendation concordance was separated into 2 categories: antithrombotic therapy and rate/rhythm control. Full details on how each category was assessed are provided in Table II in the Data Supplement. With respect to anticoagulation, if a mechanical valve was present, the patient must have been treated with warfarin. If a mechanical valve was not present, the CHA2DS2-VASc score was used to assess stroke risk. If the CHA2DS2-VASc score was ≥2, the patient must have been treated with an oral anticoagulant if no significant contraindication was present (defined as history of intracranial hemorrhage, significant allergy to oral anticoagulants, or pregnancy). If a patient was specified as receiving a rate control strategy by the provider, the patient must have been treated with either a β blocker or calcium channel blocker. Digoxin use for rate control was not considered. If the patient was receiving rhythm control, the patient’s medical history was assessed for contraindications to the specific antiarrhythmic drug that the patient was receiving. For example, a patient on sotalol with a creatinine clearance <40 mL/min was considered in violation of guideline recommendations. In the case of amiodarone, first-line therapy was determined by the record of prior antiarrhythmic therapy status in the case report form. Treatment with amiodarone as first-line therapy without concomitant left ventricular hypertrophy or congestive heart failure was considered discordant with the guidelines.2 Left ventricular hypertrophy was defined as a posterior wall thickness >1.5 cm.
A patient was deemed not eligible for assessment of antithrombotic therapy if a significant contraindication to anticoagulation was present or the patient had a CHA2DS2-VASc of 0 or 1. In that case, the patient’s guideline concordance was determined only by rate/rhythm control. Thus, patients with a CHA2DS2-VASc of 0 or 1 were neither penalized nor rewarded for receiving antithrombotic therapy. An illustration of this methodology is given in Table III in the Data Supplement. Patients specified as receiving rhythm control who were either not receiving an antiarrhythmic medication or had permanent AF and were treated with amiodarone were deemed not eligible for assessment of rhythm control concordance. In that case, the patient’s guideline concordance was determined by antithrombotic therapy. This was done to avoid penalizing patients with permanent AF who may be using amiodarone for rate control or patients with infrequent paroxysmal AF who successfully maintain normal sinus rhythm without an antiarrhythmic medication.
Assignment to Guideline Concordance Groups
Each patient was assigned to 1 of 2 groups based on the degree to which their care was concordant with guideline recommendations. Patients who received care that was concordant with all recommendations for which they were eligible were assigned to the GC group. Patients who received care that was not concordant with at least one recommendation for which they were eligible were assigned to the nonguideline-concordant (NGC) group. For analyses within specific AF treatment categories (antithrombotic therapy, rate control, and rhythm control), this assignment strategy was repeated by assessing only the recommendations within the specific category. For example, patients who were in concordance with all antithrombotic recommendations for which they were eligible were assigned to the GC antithrombotic group, and those who were not, were assigned to the NGC antithrombotic group. All clinical variables were assessed at baseline, and group assignments were not modified based on data collected at follow-up.
Follow-up data were collected at 6-month intervals for a minimum of 2 years. The primary outcomes were death, stroke or systemic embolism, International Society on Thrombosis and Haemostasis major bleeding,12 and hospitalization. Hospital admissions were subclassified into bleeding, cardiovascular, and other. Outcome events were assessed at each individual site through review of the medical record. Stroke and systemic embolism events were verified by single-source document submission (eg, hospital discharge report) and central review at the data coordinating center. Outcomes were assessed as the first event within last follow-up. For patients without any clinical events or who were lost to follow-up, the censoring date was the last follow-up.
Baseline characteristics are represented as median (interquartile range) for continuous variables and occurrence rate for dichotomous variables. Characteristics were compared using the Wilcoxon rank-sum test and χ2 tests, respectively. A P<0.05 was considered significant. Risk estimates, including hazard ratio (HR) and 95% confidence interval (CI) of patient-level concordance, were computed using random effect Cox regression after stratification for 23 guideline opportunities. The final regression model for each end point was developed previously based on selected risk factors from candidate baseline characteristics using backward selection, with an α for exclusion of 0.05 in the first imputed data set of 5 imputed data sets. Final covariates for each outcome are listed in the Appendix in the Data Supplement. All continuous variables were tested for linearity, and nonlinear relationships were accounted for using linear splines.13 Missing data on covariates were handled using multiple imputations. Random-intercept Cox regression was used to adjust for site variation. Combined results from the 5 imputed data sets were used in the calculation of the final risk estimates and standard errors. Analyses were conducted comparing the GC and NGC groups overall and within antithrombotic, rate control, and rhythm control. We also conducted a sensitivity analysis excluding the rhythm control recommendations that were derived from the guideline text, tables, and figures (ie, recommendations that are not given a level of evidence or class of recommendation designation in the guideline document, recommendations 4–10 in Table 1) and excluding recommendation 11, which was not present in the 2011 guideline.
A sensitivity analysis was conducted to determine the association between GC care and outcomes at the center level. The continuous center-level concordance proportion was calculated by the number of total therapies divided by the number of total eligibilities for the patients in each center. We adjusted for the same covariates used in the patient-level analysis in the final regression model, including continuous center-level concordance proportion rather than binary patient-level concordance (GC versus NGC). HR and 95% CI for each 0.2 increase in concordance proportion were presented. Patients were grouped into enrollment sites (N=174), and HRs were calculated for each 0.2 increase in concordance proportion.
Study Population and Recommendation Concordance
A total of 9570 patients with AF were eligible for the study and could be assessed for overall guideline concordance. Among these patients, 5977 of 9570 (62.5%) were in the GC group and 3593 of 9570 (37.5%) were in the NGC group. Patient characteristics at baseline are shown in Table 2. The median age was 75 years (interquartile range, 67, 82), 42.9% were women, and the median CHA2DS2-VASc score was 4 (interquartile range, 3, 5). GC patients tended to be younger, more likely to be men, had higher rates of hypertension, prior stroke, valvular disease, prior antiarrhythmic drug therapy, and warfarin usage, and had lower rates of gastrointestinal bleeding, sinus rhythm on ECG, normal left ventricular ejection fraction, normal left atrial size, paroxysmal AF, and antiplatelet usage (Table 2).
To further elucidate potential associations between clinical characteristics and guideline concordance, the proportion of GC care in different patient subgroups was determined. For example, the proportion of GC care in patients with a left ventricular ejection fraction of <30% was 72.9% (n=299/410). These data are shown in the Figure. Several patient characteristics were associated with higher rates of guideline concordance. Patients with a reduced left ventricular ejection fraction were more likely to be in the GC group (concordance 72.9% for ejection fraction <30% versus 61.2% for ejection fraction ≥50%; P<0.001). Electrophysiologists had higher rates of guideline concordance compared with general cardiologists and primary care physicians (67%, 62.4%, and 59%; P<0.001). Patients who were not in sinus rhythm on ECG were more likely to be in the GC group than patients in sinus rhythm (concordance 66.9% versus 53.6%; P<0.001). Associations were also observed in recent versus no history of gastrointestinal bleeding, type of AF, and antiplatelet usage.
Concordance rates within the categories of antithrombotic therapy, rate control, and rhythm control were 6929 of 8781 (78.9%), 4983 of 6623 (75.2%), and 1505 of 2091 (72.0%). Concordance with rate control recommendations was slightly higher than concordance with rhythm control recommendations (75.2% versus 72.0%; P=0.029). Concordance rates for individual recommendations ranged from 44.4% to 93.4% (Table 1). Within antithrombotic therapy, 78.9% of patients with a CHA2DS2-VASc score ≥2 received an oral anticoagulant. Conversely, 501 of 864 (60.0%) of patients with a CHA2DS2-VASc score of 0 or 1 were prescribed an oral anticoagulant. Within rhythm control, the lowest rate of concordance was observed for the recommendation that patients with coronary artery disease not be treated with flecainide or propafenone (concordance 44.4%). Amiodarone was used as first-line therapy in many patients without heart failure, left ventricular dysfunction, or left ventricular hypertrophy (65.4% concordance). Concordance to all other antiarrhythmic recommendations was ≥70%.
Patient-Level Guideline Concordance and Outcomes
Table 3 details the clinical event rates in the GC and NGC groups. During a median follow-up of 30 months, 1236 (12.9%) patients died, 5048 (52.7%) were hospitalized, and 217 (2.3%) experienced a stroke or systemic embolism. After adjusting for baseline characteristics, overall survival was similar between the 2 groups (Table 4; HR=0.96; 95% CI, 0.85–1.08; P=0.497). There was no difference in time to first stroke, major bleeding, all-cause hospitalization, or cardiovascular hospitalization. However, a higher hazard of bleeding hospitalization was observed in the GC group (HR=1.21; 95% CI, 1.03–1.42; P=0.021). Similar results were obtained in a sensitivity analysis that excluded rhythm control recommendations that were derived from the guideline document text, tables, and figures and recommendations not present in the 2011 guideline (Table IV in the Data Supplement).
Among the 8781 patients with an indication for anticoagulation (CHA2DS2-VASc ≥2 or mechanical valve) and no contraindication, there was no significant difference in any outcome between patients receiving GC versus NGC antithrombotic therapy after adjustment for covariates (Table 5) although there was a trend toward a lower risk of stroke (HR=0.75; 95% CI, 0.53–1.06; P=0.10). In the 6623 patients managed with a rate control strategy, there was a higher risk of bleeding hospitalization in patients receiving GC rate control versus patients receiving NGC rate control (Table 5; HR=1.24; 95% CI, 1.00–1.54; P=0.05). There were no other significant differences in outcomes within rate control. In the 2091 patients receiving rhythm control, there were no differences in outcomes in patients receiving GC rhythm control therapy versus NGC rhythm control therapy (Table 5).
The unadjusted and adjusted associations between center-level guideline concordance and outcomes are presented in Table 6. Increasing guideline concordance at the site-level was not associated with improved outcomes, including risk of death, stroke, major bleeding, or cardiovascular hospitalization. Higher guideline concordance was associated with a higher risk of all-cause hospitalization (HR=1.18; 95% CI, 1.06–1.33; P=0.004), bleeding hospitalization (HR=1.26; 95% CI, 1.00–1.58; P=0.05), and other hospitalization (HR=1.24; 95% CI, 1.08–1.41; P=0.002).
In this analysis of >9000 patients with AF, we found that a significant proportion of patients (37%) in routine clinical practice received care that was not in agreement with at least one of the guideline recommendations assessed in this study. Several patient characteristics were associated with higher rates of guideline concordance, including non–sinus rhythm on ECG, more advanced AF, and lack antiplatelet usage. In addition, guideline concordance was higher in patients treated by providers with greater specialization in arrhythmias. Concordance with guideline recommendations for antiarrhythmic drugs was marginally lower than concordance with recommendations on rate control. After adjustment for clinical characteristics, overall guideline concordance was not associated with improved clinical outcomes at either the patient or center-level.
Our data suggest that many patients with AF are not treated in strict concordance with guideline recommendations. In particular, many patients were treated with antiarrhythmic drug (AAD) therapy outside of guideline recommendations (28%). This finding is largely consistent with prior studies examining rates of inappropriate AAD usage in patients with AF.4–6 Concordance with individual AAD recommendations ranged from 44% to 93% although all but 2 recommendations had concordance rates ≥70% (Table 1). The lowest concordance rate was observed in patients receiving a class IC agent despite the presence of coronary artery disease (56% nonconcordance). It is important to note that there are often important reasons why a given patient is treated outside of guideline recommendations. For instance, the provider may be aware of the contraindication to therapy but may consider the comorbid condition to be relatively mild (eg, nonobstructive coronary artery disease, etc) and thus unlikely to cause complications. In the case of discordant first-line use of amiodarone, it is possible that patients were placed on amiodarone as a temporary therapy, and thus the provider may not be concerned about long-term toxicities. Finally, it is possible that some providers are unaware of the guideline recommendation. Indeed, an analysis of market research data demonstrated that although ≥50% of patients with AF treated with an AAD should receive a class IC agent, approximately 74% receive a class III agent and only 19% receive a class IC agent.14
We were unable to demonstrate an association between concordance to AAD guideline recommendations and superior patient outcomes. This is in contrast to a prior retrospective, observational single-center study of 5976 patients treated with an AAD, which found a lower risk of AF recurrence, fewer hospital admissions for AF, and fewer procedures for recurrent AF (electric cardioversion, pacemaker implantation, atrioventricular nodal ablation) in patients receiving guideline-directed AAD therapy.4 The risk of death, stroke, and cardiovascular hospitalization was similar. We did not examine the risk of AF recurrence, AF-related procedures, or AF-specific hospitalization, but we similarly did not find a difference in risk of death, stroke, or cardiovascular hospitalization. We also conducted a sensitivity analysis excluding rhythm control recommendations that were derived from the guideline text, tables, and figures and therefore were not assigned official classes of recommendation or levels of evidence. This sensitivity analysis yielded similar results (Table IV in the Data Supplement).
The concordance rate for antithrombotic therapy (79%) was higher in this cohort than in many prior studies. A systematic review of 54 studies on antithrombotic therapy in patients with AF found concordance rates of warfarin therapy ranging from 19% to 92.3% in patients deemed high risk for thromboembolism.3 Of the 9 studies using the CHADS2 score for thromboembolism risk estimation, 7 of 9 studies reported anticoagulation treatment levels <70% (range, 39%–92%). A large retrospective cohort study (n=42 834) of Canadian patients with newly diagnosed AF found that only 54.1% in patients with a CHA2DS2-VASc score ≥2 were prescribed warfarin within 90 days of diagnosis.15 In addition, patients with a low or intermediate risk of stroke by CHADS2 and CHA2DS2-VASc were equally likely to receive warfarin as those with a high risk. We observed a relatively favorable concordance rate of 79% in patients with a CHA2DS2-VASc score ≥2 and high concordance (93%) in patients with a mechanical valve. The relatively high rate of concordance in this cohort may be in part because of selection bias from voluntary participation in a national registry focused on AF care. Usage of antiplatelet agents may also have influenced the decision of a given provider to prescribe or not prescribe anticoagulant therapy, and we did not compare antiplatelet usage among AAD concordant and AAD nonconcordant patients. Of note, the vast majority of patients in this cohort were receiving warfarin, with only 473 of 9570 (5%) receiving dabigatran. Concordance in cohorts treated predominantly with novel oral anticoagulants may be different.
Surprisingly, concordance to antithrombotic therapy was not associated with improved outcomes after adjustment for covariates. However, there was a strong trend toward a lower risk of stroke in the GC group (HR, 0.75; 95% CI, 0.53–1.06; P=0.10). The high rate of oral anticoagulation and low stroke rate may have limited our power to detect a difference in outcomes between GC and NGC oral anticoagulation.
Concordance with the rate control recommendation evaluated in this study was 75%. Of note, we interpreted this recommendation strictly and required that all patients treated with a rate control strategy be prescribed a β blocker or a calcium channel blocker. Thus, patients with an adequate level of intrinsic rate control without medications were treated as nonconcordant as were patients treated only with digoxin. We found that treatment with a β blocker or calcium channel blocker was not associated with improved patient outcomes. This is in contrast to a large retrospective study of health insurance claims in Taiwan which found a reduced risk of death in patients with AF treated with β blockers.9 However, the death rate in this study was 32.7% during a mean follow-up of 4.9 years, which is significantly higher than the death rate in our cohort (12.9%). Thus, there may be differences in the patient populations that account for this discrepant finding. We also grouped β blockers and calcium channel blockers together, which may have obscured a possible favorable effect of β blockers.
Interestingly, we found an association between overall GC care and an increased risk for bleeding hospitalization in the patient-level analysis. In addition, we observed an association between rate control concordance and bleeding hospitalization. In the center-level analysis, we found an increased risk of all-cause hospitalization, bleeding hospitalization, and noncardiovascular hospitalization with higher guideline concordance. Although the causes of these associations are unclear, these findings may reflect increased healthcare use in patients who are treated with strict guideline concordance. Regarding the association between rate control and bleeding hospitalization, patients treated with a rate control strategy may have more advanced AF, such as persistent or permanent AF. These patients, in turn, may be more likely to receive oral anticoagulation because of a perceived higher thromboembolic risk. Indeed, patients in the rate control group were more likely to have persistent or permanent AF compared with the rhythm control group (54.4% versus 24.0%; P<0.001). It is also possible that rate control patients had more extensive comorbidities that either led to a higher risk of bleeding or a higher rate of strict antithrombotic therapy. We attempted to minimize this by adjusting for covariates associated with each outcome, but residual confounding is a possibility.
This study has several limitations. This was a retrospective analysis and therefore is subject to confounding. We attempted to minimize confounding by adjusting for all covariates that were associated with each outcome. However, residual measured and unmeasured confounding variables may have influenced our findings. The associations observed between some patient characteristics, such as left ventricular ejection fraction, and guideline concordance may have been confounded by referral bias to more specialized providers. Although patients treated by electrophysiologists and general cardiologists had higher rates of overall guideline concordance, we did not assess whether this was because of improved concordance to antithrombotic therapy, rate/rhythm control, or both. This analysis used recommendations from the 2014 ACC/AHA/HRS guideline, but the patient data were obtained from 2010 to 2011 before publication of this guideline edition. However, as discussed in the Methods section, all but 1 recommendation used in this study was present in the 2006 and 2011 guideline editions (Table I in the Data Supplement). As noted in the Methods section, we assessed thromboembolic risk using the CHA2DS2-VASc, rather than the CHADS2 score. It is possible that some patients at high risk for thromboembolism by CHA2DS2-VASc were not prescribed anticoagulation because they were classified as low/intermediate risk by CHADS2 at the time of enrollment in ORBIT-AF. This fact may account for some degree of the guideline nonconcordance observed within antithrombotic therapy. We did not consider renal dysfunction to be a contraindication to prescription of an oral anticoagulant. As mentioned in the Methods section, we considered any history of coronary artery disease to be a contraindication to the use of flecainide or propafenone. As the guidelines are somewhat ambiguous about this topic, it is possible that many providers considered only a history of myocardial infarction as a contraindication to these medications. This may account for some of the nonconcordance observed with this recommendation.
We only assessed a limited set of class I and class III recommendations and therefore cannot comment on associations between other guideline recommendations and patient outcomes. This patient population and participating practices, in which patients consented to participate in a national registry and each site volunteered to participate, may not be representative of the general patient with AF in other practice populations. Finally, our analysis cannot provide information as to why physicians chose not to follow guideline recommendations in specific patients. It is possible that certain patient-specific factors, such as medication intolerance, perceived bleeding risk, or patient preference, led providers to deviate from guideline recommendations. Such factors could significantly influence outcome events independent of guideline concordance or nonconcordance.
A significant proportion of patients with AF in clinical practice receive care that is not consistent with the AHA/ACC/HRS guideline recommendations assessed in this study. Although concordance to oral anticoagulation was higher than has previously been reported, concordance with guideline antiarrhythmic recommendations was relatively limited. Physicians with greater training in treating arrhythmias achieved higher rates of guideline concordance. Overall guideline concordance was associated with some evidence for increased risk of hospitalization and healthcare use but was not associated with statistically significant reductions in death, stroke, or major bleeding. These findings emphasize the need for continuous provider education for both generalists and specialists, as well as continuous incorporation and propagation of emerging and existing knowledge into guideline documents. In addition, there is a strong need for continued research to evaluate contemporary guideline concordance and clinical outcomes in AF.
Dr Kowey is ad hoc consultant to Johnson and Johnson. Dr Reiffel received research funding from Giliead, Janssen, and Medtronic; consulting or speaking services to BI, BMS/Pfizer, Janssen, Daiichi Sankyo, and Portola. Dr Allen received research funding from the National Heart, Lung, and Blood Institute, Patient-Centered Outcomes Research Institute, and American Heart Association; consulting for Novartis, Janssen, St. Jude, and ZS Pharma. Dr Fonarow is consulting to Janssen, Medtronic, and St Jude. Dr Piccini received research funding from Agency for Healthcare Research and Quality, ARCA biopharma, Boston Scientific, Gilead, Janssen Pharmaceuticals, ResMed, Spectranetics, and St Jude Medical and provides consulting to BMS/Pfizer, GSK, Johnson & Johnson, Medtronic, and Spectranetics. The ORBIT-AF (Outcomes Registry for Better Informed Treatment of atrial fibrillation) registry is supported by Janssen Pharmaceuticals, Inc. The other authors report no conflicts.
The Data Supplement is available at http://circep.ahajournals.org/lookup/suppl/doi:10.1161/CIRCEP.117.005051/-/DC1.
Circ Arrhythm Electrophysiol is available at http://circep.ahajournals.org.
- Received January 14, 2017.
- Accepted September 5, 2017.
- © 2017 American Heart Association, Inc.
- Chugh SS,
- Blackshear JL,
- Shen WK,
- Hammill SC,
- Gersh BJ
- January CT,
- Wann LS,
- Alpert JS,
- Calkins H,
- Cigarroa JE,
- Cleveland JC Jr.,
- Conti JB,
- Ellinor PT,
- Ezekowitz MD,
- Field ME,
- Murray KT,
- Sacco RL,
- Stevenson WG,
- Tchou PJ,
- Tracy CM,
- Yancy CW
- Qin D,
- Leef G,
- Alam MB,
- et al
- Chiang CE,
- Goethals M,
- O’Neill JO,
- Naditch-Brûlé L,
- Brette S,
- Gamra H,
- Zharinov O,
- Steg PG
- Reiffel JA,
- Kowey PR,
- Myerburg R,
- Naccarelli GV,
- Packer DL,
- Pratt CM,
- Reiter MJ,
- Waldo AL
- Kowey PR,
- Reiffel JA,
- Myerburg R,
- Naccarelli GV,
- Packer DL,
- Pratt CM,
- Reiter MJ,
- Waldo AL
- Chao TF,
- Liu CJ,
- Tuan TC,
- Chen SJ,
- Wang KL,
- Lin YJ,
- Chang SL,
- Lo LW,
- Hu YF,
- Chen TJ,
- Chiang CE,
- Chen SA
- Piccini JP,
- Fraulo ES,
- Ansell JE,
- Fonarow GC,
- Gersh BJ,
- Go AS,
- Hylek EM,
- Kowey PR,
- Mahaffey KW,
- Thomas LE,
- Kong MH,
- Lopes RD,
- Mills RM,
- Peterson ED
- 11.↵Betapace AF(R) [package insert]. Wayne, NJ: Bayer HealthCare Pharmaceuticals Inc; 2011.
- Schulman S,
- Kearon C
- Harrell FE
- Sandhu RK,
- Bakal JA,
- Ezekowitz JA,
- McAlister FA