Author + information
- Received December 15, 2016
- Revision received March 3, 2017
- Accepted March 10, 2017
- Published online May 15, 2017.
- Lucas N. Marzec, MDa,b,∗ (, )
- Jingyan Wang, MSc,
- Nilay D. Shah, PhDd,
- Paul S. Chan, MD, MScc,
- Henry H. Ting, MDe,
- Kensey L. Gosch, MSc,
- Jonathan C. Hsu, MD, MASf and
- Thomas M. Maddox, MD, MScb
- aUniversity of Colorado School of Medicine, Aurora, Colorado
- bVeterans Affairs Eastern Colorado Health Care System, Denver, Colorado
- cMid America Heart Institute, Kansas City, Missouri
- dMayo Clinic, Rochester, Minnesota
- eNew York Presbyterian Hospital, The University Hospital of Columbia and Cornell University, New York, New York
- fUniversity of California, San Diego, La Jolla, California
- ↵∗Address for correspondence:
Dr. Lucas N. Marzec, University of Colorado School of Medicine, 12401 East 17th Avenue, Campus Stop B-132, Aurora, Colorado 80045.
Background Oral anticoagulation (OAC) with warfarin is underused for atrial fibrillation (AF). The availability of direct oral anticoagulants (DOACs) may improve overall OAC rates in AF patients, but a large-scale evaluation of their effects has not been conducted.
Objectives This study assessed the effect of DOAC availability on overall OAC rates for nonvalvular AF.
Methods Between April 1, 2008 and September 30, 2014, we identified 655,000 patients with nonvalvular AF and a CHA2DS2-VASc score of >1 in the National Cardiovascular Data Registry PINNACLE registry. Temporal trends in overall OAC and individual warfarin and DOAC use were analyzed. Multivariable hierarchical logistic regression identified patient factors associated with OAC and DOAC use. Practice variation of OAC and DOAC use was also assessed.
Results Overall OAC rates increased from 52.4% to 60.7% among eligible AF patients (p for trend <0.01). Warfarin use decreased from 52.4% to 34.8% (p for trend <0.01), and DOAC use increased from 0% to 25.8% (p for trend <0.01). An increasing CHA2DS2-VASc score was associated with higher OAC use (odds ratio [OR]: 1.06; 95% confidence interval [CI]: 1.05 to 1.07), but with lower DOAC use (OR: 0.97; 95% CI: 0.96 to 0.98). Significant practice variation was present in OAC use (median odds ratio [MOR]: 1.52; 95% CI: 1.45 to 1.57) and in DOAC use (MOR: 3.58; 95% CI: 3.05 to 4.13).
Conclusions Introduction of DOACs in routine practice was associated with improved rates of overall OAC use for AF, but significant gaps remain. In addition, there is significant practice-level variation in OAC and DOAC use.
Oral anticoagulation (OAC) is recommended to reduce the risk of stroke associated with nonvalvular atrial fibrillation (AF) in patients at moderate to high risk of thromboembolism (1). Warfarin is the most widely prescribed oral anticoagulant in the United States (2) and is the standard of care to reduce the risk of stroke associated with AF in at-risk patients (3,4). However, previous work has demonstrated low rates of warfarin use in these patients (5–9). Adherence to warfarin therapy is complex and is related to several factors, including the need for frequent blood monitoring, frequent dose adjustments, and myriad drug and food interactions, all of which may contribute to underuse of OAC with warfarin (10).
Direct oral anticoagulants (DOACs), such as dabigatran, rivaroxaban, and apixaban, offer effective alternatives to warfarin without many of the disadvantages of warfarin. Previous studies have suggested that the availability of DOACs may improve rates of use of overall OAC in patients with AF (11,12), but large-scale studies that have investigated the effect of DOAC use on overall OAC use and the influence of patient-level factors is unknown.
Accordingly, we analyzed data from the National Cardiovascular Data Registry (NCDR) PINNACLE (Practice Innovation and Clinical Excellence) Registry to assess how the availability of DOACs has affected overall OAC use. We described the temporal trends in OAC use, including both warfarin and DOACs, and patient factors associated with prescription of warfarin, DOACs, and overall OAC use in patients with nonvalvular AF at moderate to high risk of thromboembolism. In addition, we assessed the extent of practice-level variation in OAC and DOAC use.
We analyzed data from the NCDR PINNACLE Registry, the details of which have been previously described (13). In brief, the PINNACLE Registry is the first national, prospective outpatient registry for quality improvement in cardiovascular care. It contains data collected at the point of care during office-based visits related to cardiovascular disease through a validated electronic health record mapping algorithm. Periodic data quality checks, standardized data collection, and clear data element definitions ensure reliable data quality. The current PINNACLE data dictionary and the data dictionary forms used during our study period are included in the Online Appendix.
Study population and eligibility
We evaluated all patients aged 18 years or older with nonvalvular AF and with a CHA2DS2-VASc score >1 with clinical encounters between April 1, 2008 and September 30, 2014. We excluded patients with previous cardiac valve surgery, a documented contraindication to OAC, and those with a CHA2DS2-VASc score ≤1.
The primary outcome was prescription of any OAC (warfarin, dabigatran, rivaroxaban, or apixaban). Secondary outcomes included prescription of any DOAC (dabigatran, rivaroxaban, or apixaban) and of the individual DOACs. Prescription data were obtained from PINNACLE outpatient encounter documentation. For patients with multiple encounters, the last encounter was used in the analysis.
Patients were initially categorized as having received warfarin, any DOAC (dabigatran, rivaroxaban, or apixaban), or no OAC based on the prescription data from their last encounter. Baseline demographic and clinical characteristics associated with the use of any OAC, warfarin, and any DOAC were then described.
We then examined temporal trends in the rates of use of any OAC, warfarin, any DOAC (including dabigatran, rivaroxaban, and apixaban), and of the individual DOACs. For each quarter during the study period, we determined the proportion of patients who received therapeutic OAC with warfarin, dabigatran, rivaroxaban, or apixaban. If patients had >1 visit per quarter, the last study visit from each patient in each quarter was used for the analysis. A Cochrane-Armitage test for trend analysis was then performed to evaluate for changes in rates of use of any OAC, any DOAC, and the individual DOACs over time.
We then examined the association between patient-level variables and the use of OAC and DOACs. We did this first with univariate regression models to assess the unadjusted association of patient factors on OAC use. We then used separate hierarchical, multivariable logistic regression models to better understand the individual contributions of patient factors to OAC and DOAC use. Patient-level variables for this model included demographics (age and sex), vital signs (height, weight, heart rate, systolic and diastolic blood pressure), comorbid conditions and risk factors (hypertension, diabetes mellitus, dyslipidemia, heart failure, previous stroke or transient ischemic attack [TIA], coronary artery disease [CAD], peripheral arterial disease [PAD]), and concomitant P2Y12 receptor antagonist use (clopidogrel, prasugrel, or ticlopidine). We also calculated the total CHA2DS2-VASc score (2 points each for age 75 years or older and previous TIA or stroke; and 1 point each for heart failure, hypertension, age 65 years or older, diabetes mellitus, vascular disease, and female sex). As a sensitivity analysis, we specified a regression model using the same patient factors and the CHADS2, score rather than the CHA2DS2-VASc score, because the CHADS2 score was the predominant AF risk score in use during portions of our study period. In this model, the total CHADS2 score (2 points for previous TIA or stroke, 1 point each for heart failure, hypertension, age 65 years or older, and diabetes mellitus) was calculated for each patient, and patients with a CHADS2 score >1 were included.
To assess the effect of practice site on OAC use, independent of patient factors, we also calculated median odds ratios (MORs) from our hierarchical model. The MOR can be interpreted as the odds that 2 patients with identical patient-level covariates from 2 randomly chosen practice sites will receive OAC. An MOR of 1.0 indicates that no variation exists between practices. Thus, the MOR is always ≥1. For example, a MOR of 1.5 indicates a 50% likelihood that a similar patient would receive different OAC prescriptions at 2 different practices. It provides an estimate of the effect size of the practice on the outcome, much as the odds ratio (OR) estimates the effect size of patient factors on the outcome. Based on previous literature, a MOR >1.2 indicates a clinically significant variation (14).
Because colinearity concerns prevent inclusion of the CHA2DS2-VASc score and its individual component conditions in the same model, we constructed our primary model to include the individual components of the scores, but not the scores themselves. We then performed sensitivity analyses in which we constructed secondary models, including each of the scores in addition to their individual components, to determine the relationship of the CHA2DS2-VASc score, independent of its individual component conditions, with OAC use. Variables with missing data in >30% of patients were excluded from the analysis, and included patient race, insurance type, chronic liver disease, intracranial hemorrhage, alcohol use, and nonsteroidal anti-inflammatory use. Variables with <30% missing rates were included in our model using the IMPUTE module for multivariate imputation of IVEware software (Institute for Social Research, University of Michigan, Ann Arbor, Michigan) (variables excluded from the analysis due to >30% missing data are included in Online Table 1).
We also wanted to determine if the availability of DOACs resulted in their use among patients who were started on de novo OAC, patients who were transitioned from warfarin, or both. To select patients who were eligible for either therapy throughout the study period, we included 16,812 AF patients who had at least 1 visit before and after October 31, 2010 (the date when dabigatran—the first available DOAC—was approved by the U.S. Food and Drug Administration) and received a DOAC after that time. Previous OAC use was determined by the last recorded visit before October 31, 2010. Patients who received de novo DOACs were those who did not receive OAC before October 31, 2010 or those who received a DOAC after October 31, 2010 without having previously received warfarin. Patients who transitioned to DOACs were those who received OAC (i.e., warfarin) before October 31, 2010 or who received warfarin before receiving a DOAC after October 31, 2010. Patient characteristics between the 2 groups were collected and compared using a Student t test for continuous variables and a chi-square or the Fisher exact test for categorical variables. Statistical analyses were performed using SAS (version 9.4, SAS Institute, Cary, North Carolina).
Between April 1, 2008 and September 30, 2014, we identified 740,255 patients with nonvalvular AF. After exclusion of 53,367 (7.2%) with a CHA2DS2-VASc score of ≤1, 25,918 (3.5%) patients with previous cardiac valve surgery, 658 (0.09%) with documented contraindication to OAC, and 5,312 (0.72%) with multiple OAC agents recorded in 1 visit, 655,000 (88.5%) patients representing 3,164,236 outpatient encounters were included in the primary analysis (Figure 1). Patients were categorized by OAC use, and their characteristics were compared (Table 1). In general, patients who received DOACs compared with warfarin were younger, had lower rates of comorbidities, and lower CHA2DS2-VASc scores.
The rate of use of overall OAC increased from 52.4% to 60.7% over the course of the study period (p for trend <0.01) (Central Illustration). Warfarin use decreased from 52.4% to 34.8% (p for trend <0.01) during the study period, and any DOAC use increased from 0% to 25.8% (p for trend <0.01). Dabigatran use increased from 0% to 7.2% (p for trend <0.01), rivaroxaban use increased from 0% to 12.3% (p for trend <0.01), and apixaban use increased from 0% to 6.3% (p for trend <0.01). Similar results were seen in the sensitivity analysis using CHADS2 >1 as the inclusion criteria, in which overall OAC use increased from 52.9% to 62.1% (p for trend <0.01), and DOAC use increased from 0% to 25.5% (p for trend <0.01).
Patient characteristics associated with OAC and DOAC use are shown in Figures 2 and 3. In the multivariate models, higher rates of use of any OAC included risk factors for stroke related to AF (increased age, heart failure, hypertension, previous stroke/TIA, PAD, and diabetes mellitus), as well as male sex, dyslipidemia, and increased weight, heart rate, and diastolic blood pressure. Patient factors associated with lower rates of use of any OAC included use of P2Y12 receptor antagonists, CAD, and increased systolic blood pressure.
Patient factors associated with higher rates of use of a DOAC included higher systolic and diastolic blood pressures, dyslipidemia, and hypertension. Patient factors associated with lower rates of use of a DOAC included risk factors for stroke related to AF (including increased age, previous stroke/TIA, CAD, heart failure, PAD, and diabetes mellitus), as well as male sex and increased heart rate. In a sensitivity analysis that included the composite CHA2DS2-VASc score, a higher CHA2DS2-VASc score was associated with increased rates of use of overall OAC use (OR: 1.06 per unit increase in CHA2DS2-VASc score; 95% confidence interval [CI]: 1.05 to 1.07), although this was associated with lower rates of using DOACs in those who received OAC (OR: 0.97; 95% CI: 0.96 to 0.98).
We found significant practice-level variation in the use of any OAC, any DOAC, and the individual DOACs. At the practice level, the use of any OAC in eligible patients ranged from 11% to 78.8% (MOR: 1.52; 95% CI: 1.45 to 1.57), and the use of any DOAC ranged from 0% to 40.4% (MOR: 3.58; 95% CI: 3.05 to 4.13) (Figure 4). There was also significant practice-level variation in the use of the individual DOACs. Dabigatran use ranged from 0% to 22% (MOR: 2.91; 95% CI: 2.54 to 3.28), rivaroxaban use ranged from 0% to 21.1% (MOR: 3.44; 95% CI: 2.95 to 3.94), and apixaban use ranged from 0% to 10.7% (MOR: 3.97; 95% CI: 3.33 to 4.64).
In our analysis to determine de novo versus transition DOAC usage patterns, we identified 16,812 patients who had at least 1 visit both before and after October 31, 2010, and received a DOAC at some point during the study period. Of these, 7,238 (43.1%) received a DOAC without having previously received warfarin (i.e., as de novo therapy) and 9,574 (56.9%) received a DOAC after having first received warfarin (i.e., as transition therapy). Patient characteristics for transition and de novo DOAC use are listed in Table 2. Patients who received a de novo DOAC were more likely to be prescribed a P2Y12 receptor antagonist. Patients who received a transition DOAC were more likely to be older, men, have a higher CHA2DS2-VASc score and weight, hypertension, heart failure, diabetes mellitus, and a previous myocardial infarction.
We found the rate of use of overall OAC increased from 52.4% to 60.7% following the introduction of DOACs (Central Illustration). The increase in DOAC prescriptions more than offset the decrease in warfarin prescriptions, which suggested that the introduction of DOACs might have contributed to improved overall OAC rates among eligible AF patients. However, analyses of patient factors indicated that DOAC use was largely directed toward younger, healthier AF patients. This observation, combined with the fact that significant numbers of eligible AF patients still remain without OAC prescriptions, suggested that DOAC availability alone is insufficient to achieve optimal OAC use in patients with AF.
Underuse of OAC with warfarin in patients with AF is common (5,7), and is associated with increased risk of stroke and thromboembolism (15,16). However, DOACs provide benefits compared with warfarin, although they are more expensive, and rivaroxaban and apixaban lack a reversal agent. Similar to other recent work (11), we found overall rates of OAC for AF increased following the introduction of DOACs, and the early adoption of DOACs preferentially occurred in patients with lower risk of stroke compared with those who received warfarin (17–20). However, this previous work was limited by either the lack of granular patient data or did not include a study period in which multiple DOACs were available. Our work from a large outpatient registry of cardiology practices (where most of dabigatran prescriptions initially occurred in the United States ) added to these previous studies by providing an analysis of practice patterns during a time in which several DOACs were available in the U.S. market, a description of the temporal trends in OAC and DOAC use among outpatient cardiology practices with the context of granular patient-level data, and an assessment of practice-level variations in OAC and DOAC use. Our results suggested the early trends in adoption of DOACs persist, although they appear to be used more as a transition from warfarin, rather than as de novo therapy for those who were not previously anticoagulated. We also found that cardiology practitioners continue to prescribe DOACs for patients with a lower risk of stroke. The reasons for this association were unknown, although perceived bleeding risk might play a role, because many of the risk factors included in the CHA2DS2-VASc score are also associated with increased bleeding risk (22).
Despite the increase in overall OAC use observed in our study, several challenges remain regarding the effective use of OAC for nonvalvular AF. First, although overall OAC use increased during our study period, DOACs were preferentially prescribed for healthier patients and for patients already on warfarin. Second, significant practice-level variation existed for overall OAC use and DOAC use. Third, a significant number of patients with nonvalvular AF without a documented contraindication to OAC were not receiving OAC. A previous analysis of data from the ORBIT-AF (Outcomes Registry for Better Quality of Care in the Treatment of AF) database demonstrated higher rates of OAC use among patients with AF seen by electrophysiologists compared with cardiologists and internal medicine/primary care providers (23). This might represent a heightened attention to the need for OAC by electrophysiologists and/or a higher likelihood of patients without contraindication to OAC being referred to electrophysiology practices. The extent to which provider specialty is associated with DOAC use over time is not known. Further work is needed to understand provider and practice characteristics associated with OAC and DOAC use to reduce variation and improve use of OAC for patients at moderate to high risk of stroke associated with AF. This is of particular importance, because recent work demonstrated that a significant portion of patients at high risk of stroke related to AF receive no OAC therapy (24). Our data raised concerns that the observed overall increase in OAC use for AF, which is primarily due to an increase in use of DOACs in patients at lower risk of stroke and in those previously anticoagulated with warfarin, would not address this treatment gap.
First, our analysis was performed using a registry that included data from outpatient cardiology practices. It was possible that practices that provided data to the PINNACLE Registry might achieve higher rates of OAC use than in cardiology practices that did not contribute data to PINNACLE or in noncardiology practices. Second, we could not account for all variables that might have influenced the use of OAC, such as patient preference or elevated bleeding risk. However, it was unlikely patient preference alone accounted for nontreatment with OAC therapy in such a large proportion of patients in our cohort. Third, most of the data in the PINNACLE Registry are abstracted from electronic health records, which might have incomplete documentation of OAC use, contraindications to OAC treatment, or associated patient factors. However, our OAC use rates and prevalence of patient characteristics were similar to other OAC studies (25–27), which suggested that missing data were not a significant factor in our analyses. Finally, although our analysis provided some insight regarding the underlying mechanisms for the treatment gaps we observed, we could not account for all the factors associated with the use of OAC or DOACs. Further work, including construction of the PINNACLE Registry to better capture data regarding the reasons underlying OAC underuse, is needed to address these mechanisms and development strategies to improve rates of OAC for AF.
We found the overall rate of use of OAC for patients with AF increased following the introduction of DOACs. DOACs were preferentially used in patients with fewer co-morbidities, lower risk of stroke, and among those previously anticoagulated with warfarin. In addition, there was significant practice-level variation in OAC and DOAC use. Further work is needed to better define the factors associated with variation and underuse of OAC for patients with AF at high risk for stroke to inform strategies to lower the risk of stroke for patients with AF.
COMPETENCY IN SYSTEMS-BASED PRACTICE: Since the introduction of DOACs, which include dabigatran, rivaroxaban, apixaban, and edoxaban, overall rates of oral anticoagulation for patients with AF in the United States have increased, but there is considerable practice-level variation. During the early years following regulatory approval, DOACs were prescribed preferentially for patients with fewer comorbidities, lower risk of stroke, and previous treatment with warfarin.
TRANSLATIONAL OUTLOOK: Additional research is needed to understand the factors associated with underuse of oral anticoagulation for patients with AF at high risk for stroke.
For the current PINNACLE data dictionary and the data dictionary forms used during our study period as well as a supplemental table, please see the online version of this paper.
This research was supported by the American College of Cardiology’s National Cardiovascular Data Registry (NCDR). The views expressed in this paper represent those of the authors, and do not necessarily represent the official views of the NCDR or its associated professional societies identified at CVQuality.ACC.org/NCDR.
Dr. Chan has received research support from the National Heart, Lung, and Blood Institute. Dr. Hsu has received honoraria from St. Jude Medical, Medtronic, Biotronik, Janssen Pharmaceuticals, and Bristol-Myers Squibb; and has received research grants from Biosense Webster and Biotronik. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- coronary artery disease
- confidence interval
- direct oral anticoagulant
- mean odds ratio
- National Cardiovascular Data Registry
- oral anticoagulation
- odds ratio
- peripheral arterial disease
- transient ischemic attack
- Received December 15, 2016.
- Revision received March 3, 2017.
- Accepted March 10, 2017.
- 2017 American College of Cardiology Foundation
- January C.T.,
- Wann L.S.,
- Alpert J.S.,
- et al.
- Brown J.D.,
- Shewale A.R.,
- Dherange P.,
- Talbert J.C.
- Chan P.S.,
- Oetgen W.J.,
- Buchanan D.,
- et al.
- Maddox T.M.,
- Chan P.S.,
- Spertus J.A.,
- et al.
- Nieuwlaat R.,
- Olsson S.B.,
- Lip G.Y.,
- et al.
- Gorin L.,
- Fauchier L.,
- Nonin E.,
- Charbonnier B.,
- Babuty D.,
- Lip G.Y.
- Steinberg B.A.,
- Holmes D.N.,
- Piccini J.P.,
- et al.
- Patel P.A.,
- Zhao X.,
- Fonarow G.C.,
- et al.
- Kirley K.,
- Qato D.M.,
- Kornfield R.,
- Stafford R.S.,
- Alexander G.C.
- Fosbol E.L.,
- Holmes D.N.,
- Piccini J.P.,
- et al.
- Hsu J.C.,
- Maddox T.M.,
- Kennedy K.F.,
- et al.
- Camm A.J.,
- Accetta G.,
- Ambrosio G.,
- et al.
- Steinberg B.A.,
- Kim S.,
- Piccini J.P.,
- et al.
- Hsu J.C.,
- Akao M.,
- Abe M.,
- et al.