Author + information
- Received August 8, 2016
- Revision received September 19, 2016
- Accepted September 20, 2016
- Published online December 12, 2016.
- Benjamin A. Steinberg, MD, MHSa,b,c,∗ (, )
- Peter Shrader, MAc,
- Laine Thomas, PhDc,
- Jack Ansell, MDd,
- Gregg C. Fonarow, MDe,
- Bernard J. Gersh, MB, ChB, DPhilf,
- Peter R. Kowey, MDg,
- Kenneth W. Mahaffey, MDh,
- Gerald Naccarelli, MDi,
- James Reiffel, MDj,
- Daniel E. Singer, MDk,
- Eric D. Peterson, MD, MPHb,c,
- Jonathan P. Piccini, MD, MHSb,c,
- ORBIT-AF Investigators and Patients
- aDivision of Cardiovascular Medicine, University of Utah Health Sciences Center, Salt Lake City, Utah
- bDepartment of Medicine, Duke University Medical Center, Durham, North Carolina
- cDuke Clinical Research Institute, Durham, North Carolina
- dDepartment of Medicine, Hofstra Northwell School of Medicine, Long Island, New York
- eAhmanson-UCLA Cardiomyopathy Center, University of California Los Angeles Division of Cardiology, Los Angeles, California
- fDepartment of Medicine, Mayo Clinic, Rochester, Minnesota
- gLankenau Institute for Medical Research, Wynnewood, Pennsylvania
- hDepartment of Medicine, Stanford University School of Medicine, Palo Alto, California
- iDepartment of Medicine, Penn State University School of Medicine, Hershey, Pennsylvania
- jDepartment of Medicine, Columbia University College of Physicians and Surgeons, New York, New York
- kDepartment of Medicine, Harvard Medical School and Massachusetts General Hospital, Boston, Massachusetts
- ↵∗Reprint requests and correspondence:
Dr. Benjamin A. Steinberg, Division of Cardiovascular Medicine, University of Utah Health Sciences Center, 30 North 1900 East, Room 4A100, Salt Lake City, Utah 84132.
Background Although non-vitamin K antagonist oral anticoagulants (NOACs) do not require frequent laboratory monitoring, each compound requires dose adjustments on the basis of certain clinical criteria.
Objectives This study assessed the frequency of off-label NOAC doses among AF patients and the associations between off-label dose therapy and clinical outcomes in community practice.
Methods We evaluated 5,738 patients treated with a NOAC at 242 ORBIT-AF II (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation phase II) sites. NOAC doses were classified as either underdosed or overdosed, consistent with Food and Drug Administration labeling. Longitudinal outcomes (median follow-up: 0.99 years) included stroke or systemic embolism, myocardial infarction, major bleeding (International Society of Thrombosis and Haemostasis criteria), cause-specific hospitalization, and all-cause mortality.
Results Overall, 541 NOAC-treated patients (9.4%) were underdosed, 197 were overdosed (3.4%), and 5,000 were dosed according to U.S. labeling (87%). Compared with patients receiving the recommended dose, those who were receiving off-label doses were older (median: 79 and 80 years of age vs. 70 years of age, respectively; p < 0.0001), more likely female (48% and 67% vs. 40%, respectively; p < 0.0001), less likely to be treated by an electrophysiologist (18% and 19% vs. 27%, respectively; p < 0.0001), and had higher CHA2DS2-VASc scores (96% and 97% ≥2 vs. 86%, respectively; p < 0.0001) and higher ORBIT bleeding scores (25% and 31% >4 vs. 11%, respectively; p < 0.0001). After dose adjustment, NOAC overdosing was associated with increased all-cause mortality compared with recommended doses (adjusted hazard ratio: 1.91; 95% confidence interval [CI]: 1.02 to 3.60; p = 0.04). Underdosing was associated with increased cardiovascular hospitalization (adjusted hazard ratio: 1.26; 95% CI: 1.07 to 1.50; p = 0.007).
Conclusions A significant minority (almost 1 in 8) of U.S. patients in the community received NOAC doses inconsistent with labeling. NOAC over- and underdosing are associated with increased risk for adverse events. (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation II [ORBIT-AF II]; NCT01701817)
One of the principal objectives in the treatment of patients with atrial fibrillation (AF) is the prevention of stroke or systemic embolism. Although dose-adjusted warfarin was the preferred therapy for many years, nonvitamin K antagonist oral anticoagulants (NOACs) have since been developed and approved for thromboembolism prevention in patients with nonvalvular AF. As a class, these agents have been demonstrated to be at least as safe and effective as dose-adjusted warfarin (1), despite not requiring routine or frequent laboratory monitoring. On the basis of pivotal, phase 3 randomized controlled trials in patients with nonvalvular AF (2–5), the U.S. Food and Drug Administration (FDA) approved each NOAC at a specific dose, with recommended adjustments based on selected patient factors (e.g., renal function, age) or concomitant medications.
However, whether these dose recommendations are adhered to in community practice remains a major concern. The extent to which NOAC doses are given according to U.S. FDA label recommendations is not known, and the consequences of off-label doses are also uncertain. Therefore, this study sought to assess the frequency at which patients received off-label doses for a NOAC; to identify factors associated with under- and overdosing of NOACs; and to analyze the associations between under- and/or overdosing of NOACs and clinical outcomes.
Patient cohort and data collection
We used data from phase II of the ORBIT-AF II (Outcomes Registry for Better Informed Treatment of Atrial Fibrillation AF II) trial, a U.S. national, prospective registry of patients with AF. The registry enrolled only patients with recent, new diagnosis of AF (within 6 months) and/or who had recently begun NOAC therapy for stroke prevention in AF (within 3 months). All patients had to be 21 years of age or older, have electrographically documented AF not due to a reversible cause (e.g., pulmonary embolism or acute thyrotoxicosis), and be capable of clinically indicated follow-up every 6 months. All eligible consecutive patients were enrolled from a nationally representative sample of sites. Enrolling physicians and sites included primary care providers, cardiologists, electrophysiologists, and neurologists. Follow-up was 12 to 24 months at 6-month intervals.
We used a Web-based case report form and collected patient demographics, medical and surgical history, medications, vital signs, laboratory data, and imaging and electrocardiographic parameters. Changes in pharmacotherapy, cardiac rhythm, and subsequent cardiovascular events and procedures were documented in follow-up. Follow-up data included oral anticoagulant and antiplatelet agent doses throughout the study. Additional details of the ORBIT-AF II registry have been published previously (6).
For the purpose of this analysis, we included patients who were treated with a NOAC for stroke prevention. Patients were excluded if they had been treated with a specific anticoagulant that was used in fewer than 100 patients. Patients without dose information available, those with significant mitral stenosis or mechanical valve replacement, patients without any follow-up, and those in whom compliance with labeled dosing could not be assessed were excluded.
Patients were then categorized by NOAC dose at baseline: dosing consistent with, underdosed, and overdosed according to U.S. FDA-approved package inserts (PIs). Approved dose criteria were specific to each NOAC, according to the following patient characteristics: renal function, weight, age, and concomitant medications, as indicated (Online Table 1) (7–9). Patients for whom the selected drug was contraindicated were classified as overdosed (e.g., patients with end-stage renal disease undergoing dialysis).
Baseline characteristics were compared among the dose groups. Rates of under- and overdosing, by drug-specific dose criteria (e.g., renal function), were calculated for each drug to assess drug-specific trends.
Outcomes during follow-up were analyzed according to and between dose categories, which included doses according to the U.S. PI, those who were underdosed, and patients who were overdosed. The outcomes of interest included all-cause death, stroke or systemic embolism (adjudicated from primary source documentation), myocardial infarction, cause-specific hospitalization (cardiovascular, bleeding, or noncardiovascular, nonbleeding); and major bleeding classified by International Society on Thrombosis and Haemostasis criteria (10).
Univariate data are shown as median (with interquartile range) or mean ± SD for continuous variables and as numbers with percentages for categorical variables. Comparison tests were performed using the Kruskal-Wallis test for continuous variables and the chi-square test for categorical variables.
For outcomes, unadjusted Cox models were performed to test the association of off-label NOAC dose and outcomes through the last follow-up. A robust covariance estimate was included in order to account for variation within sites. All models (including unadjusted models) were stratified by the type of NOAC and drug-specific regulator criteria for dose (i.e., eligibility to be inappropriately dosed). Thus, hazard ratios represent comparisons of dose status among patients with similar eligibility criteria. A categorical variable for dose category with 3 levels (appropriately dosed, underdosed, overdosed) was included in the model, with patients receiving recommended doses treated as the reference group. Hazard ratios with 95% confidence intervals (CI) are presented, along with the corresponding p value. Multivariate models were adjusted for clinically relevant patient characteristics that might explain the association between dosing and outcomes (Online Table 2). Data missing in the multivariate models were accounted for by multiple imputations. Combined results from 5 imputations are presented. All continuous covariates were tested for linearity any nonlinear associations were accounted for using linear splines.
Subjects in ORBIT-AF II trial provided written, informed consent, and sites received regulatory board approval for this study, pursuant to local regulations. The study was coordinated by the Duke Clinical Research Institute and approved by the Duke University institutional review board. The deidentified, aggregate data were analyzed by the Duke Clinical Research Institute using commercially available software (version 9.3, SAS Institute, Cary, North Carolina).
The baseline ORBIT-AF II population included 11,603 patients enrolled at 242 U.S. ORBIT-AF II sites from February 2013 to January 2016. For this analysis, we excluded 3,085 patients who were not taking a NOAC at baseline; 721 with missing renal function; 42 patients with valvular AF (mechanical valves or moderate/severe mitral stenosis); and 2,017 patients who had not yet been seen in follow-up. This yielded a study cohort of 5,738 patients from 242 sites treated with a NOAC at baseline, including 425 treated with dabigatran (7.4%), 3,078 with rivaroxaban (53.6%), and 2,235 with apixaban (39%). Overall, 5,000 patients were receiving doses consistent with the label (87%), 541 were receiving a dose lower than the PI would dictate (9.4%), and 197 were receiving a NOAC dose higher than the PI would indicate (3.4%). Baseline characteristics of these groups are shown in Table 1. Relative to those whose doses were consistent with the PI, patients who received under- or overdoses of NOACs were significantly older (median 79 and 80 years of age vs. 70 years of age, respectively; p < 0.0001), more likely female (48% and 67% vs. 40%, respectively; p < 0.0001), more likely to have a CHA2DS2-VASc score ≥2 (96% and 97% vs. 86%, respectively; p < 0.0001), and more likely to have high (≥4) ORBIT bleeding scores (25% and 31% vs. 11%, respectively; p < 0.0001). Rates of concomitant aspirin were not different among the dose groups (25% and 24% vs. 26%, respectively; p = 0.7).
Dose groups were subsequently stratified by drug (Figure 1) at baseline. Among patients taking dabigatran, those with an estimated creatinine clearance (CrCl) of 30 to 50 ml/min had the highest rates of off-label dosages: 23% were underdosed (recommended dosage of 150 mg twice a day). Among patients receiving rivaroxaban, those with a CrCl of 15 to 50 ml/min had the highest rates of off-label dosages: 34% were overdosed (recommended dosage of 15 mg daily). Among those treated with apixaban, patients undergoing dialysis had the lowest rates of label-concordant dosing: overall, 7 of 22 patients (32%) undergoing hemodialysis were receiving off-label doses of apixaban (Online Table 3).
Clinical outcomes during a median follow-up of 0.99 years, stratified by NOAC dose, are shown in the Central Illustration. Rates of adverse events were higher in underdosed and overdosed patients than in those receiving the indicated NOAC dose. Unadjusted and adjusted Cox proportional hazards models are shown in Table 2. Compared with NOAC dose according to the PI, overdosing was significantly associated with increased risk of all-cause mortality (adjusted hazard ratio: 1.91: 95% CI: 1.02 to 3.60), and underdosing of NOACs was significantly associated with increased risk of cardiovascular hospitalization (adjusted hazard ratio: 1.26; 95% CI: 1.07 to 1.50; p = 0.007).
To our knowledge, this is the first analysis of NOAC dose across all agents and of the association between NOAC dose and clinical outcomes in patients with AF in community practice. There are 3 major findings from this analysis. First, most patients with AF treated with a NOAC in the community are prescribed doses according to FDA-approved labeling (11). Second, off-label doses occur in a significant minority across NOACs and dose levels (approximately 1 in 8 patients). Finally, off-label doses of NOACs for stroke prevention in AF are associated with worse clinical outcomes in U.S. practice.
Doses consistent with the PI in most of our patients represents success for the prevention of stroke and systemic embolism and compares favorably with time in therapeutic range for dose-adjusted warfarin. Multiple international randomized trials have demonstrated overall improved outcomes in patients receiving NOACs compared with those for warfarin at the pre-specified and tested NOAC doses (1–5,12). Without implementation of these therapies in appropriate patients at the appropriate doses, the expected improvement in population health outcomes and cost effectiveness would be difficult to achieve. The fact that nearly 9 of 10 subjects received doses consistent with the label suggests that use of NOACs is largely consistent with clinical trials and prescribing recommendations from the FDA.
This may raise the question of whether these results are definitively generalizable to broader U.S. or international AF cohorts. Although ORBIT-AF sites were selected to broadly represent a variety of practice settings (private, academic, governmental) and specialties (primary care, cardiology, electrophysiology, neurology), these data may represent an optimistic assessment of doses due to the potential biases of participation in an AF registry. Our findings do require corroboration with other populations, as the implications of off-label dose in clinical practice may be even larger than we observed.
A significant proportion of patients in our study did not receive NOACs according to the label dose for stroke prevention in nonvalvular AF. Although this dose level was present across agents and renal function strata, it was particularly prominent in groups with intermediate renal dysfunction (13). There may be several reasons for this, including variable calculation of CrCl; many electronic medical records estimate renal function via the Modification of Diet in Renal Disease (MDRD) formula (14), whereas drug dose approval and dose adjustments are consistently determined using the Cockcroft-Gault formula (15). Although this difference may not be clinically relevant at very high or very low CrCl strata, calculations for patients with intermediate renal function may yield clinically relevant differences and subsequent dose errors (16). Alternatively, lack of familiarity with dose guidelines and adjustments for concomitant medications, particularly among physicians who use NOACs less frequently, may contribute to unintentional off-label use; we observed significant variability in doses according to provider specialty. Finally, physicians may be looking at a given patient’s trend in renal function or other characteristics and intentionally alter the dose from that indicated by the PI.
It is difficult to ascertain physician intent from these registry data; however, the underlying risks of the cohort suggest potential intentional deviation from approved doses. For example, patients who were overdosed had significantly higher CHA2DS2-VASc scores than those given doses on-label, and patients who were underdosed had significantly higher ORBIT bleeding scores than those given doses on-label. Physicians may be tailoring the doses of these medications to the specific patient’s underlying risk, despite the fact that large clinical trials and subgroup analyses demonstrated favorable risk-benefit profiles across risk strata (2–5,17). This type of physician behavior was cited as the reason the FDA did not approve the lower dose of dabigatran from the RE-LY (Randomized Evaluation of Long-Term Anticoagulation Therapy) trial; regulatory authorities were concerned that physicians would favor the lower dosing in order to minimize bleeding, at the expense of too many thromboembolic events (18). These data suggest such concerns may have been valid. A significant proportion of patients received off-label dosing, and such dosing with these agents was associated with worse events.
Although adverse events were uncommon in our cohort, the outcomes we observed are biologically plausible and consistent with other data sources; overdosing of anticoagulant agents, particularly in the setting of impaired renal function, is associated with worse bleeding outcomes (19). Emerging pharmacokinetic data may provide some reassurance for consistent drug effect in patients on dialysis (20); however, the current clinical data do not support use of these drugs in patients with severe kidney disease. Underdosing also appears to be associated with worse outcomes, but these data should be interpreted with caution, as the mechanism of the effect is unclear.
We studied the relationship between off-label doses and outcomes retrospectively; dose groups were not randomized. Residual and/or unmeasured confounding may exist and introduce bias in the results; furthermore, clinical events may be associated with off-label use but not causally related to dosing (e.g., risk factors for bleeding and stroke). Additionally, not all concomitant medicines that might indicate NOAC adjustment were collected; however, these are likely to be uncommon and/or short-term (e.g., antimicrobial agents). Similarly, trends in off-label doses over time were not assessed due to variability in the availability of dose criteria (e.g., renal function). The definition of on-label dose was derived from the U.S. FDA-approved PIs, and thus these results may not be applicable in other jurisdictions. Last, our multivariate analysis assumed a consistent magnitude of effect for all off-label doses, which may not reflect the relative associations among different drugs and doses.
Most patients treated with NOACs for stroke prevention in AF received doses according to FDA-approved PI. However, a significant minority did not receive such doses, and off-label doses were associated with increased risk for adverse events. Careful attention to recommended doses and additional studies of these agents in patients under-represented in clinical trials may improve clinical outcomes.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: In a practice-based registry, almost 1 of 8 patients with AF treated with target-specific NOACs received doses other than those recommended in the FDA-approved labels. Patients given doses that were higher than recommended exhibited greater all-cause mortality, and those receiving lower than recommended doses had more cardiovascular hospitalizations than those prescribed the recommended doses of these agents.
TRANSLATIONAL OUTLOOK: Future studies of variability in clinical outcomes would be informed by correlations among doses, serum drug concentrations, and clinical outcomes.
For supplemental tables, please see the online version of this article.
The ORBIT-AF registry was sponsored by Janssen Scientific Affairs, LLC. Dr. Steinberg has received research support from Janssen Scientific; and is a consultant for BMS-Pfizer. Dr. Ansell is a consultant for and advisory board member for Bristol-Myers Squibb, Pfizer, Janssen, Daiichi, Boehringer Ingelheim, and Alere. Dr. Fonarow is a consultant for and receives research support from Janssen. Dr. Gersh has received research support from and is an advisory board member of Medtronic, Baxter Healthcare Corporation, InspireMD, Cardiovascular Research Foundation, PPD Development, LP, Boston Scientific, and St. Jude. Dr. Kowey is a consultant for and advisory board member with Boehringer Ingelheim, Bristol-Myers Squibb, Johnson & Johnson, Portola, Merck, Sanofi, and Daiichi-Sankyo. Dr. Mahaffey has received research grants from Amgen, Daiichi, Johnson & Johnson, Medtronic, Merck, St. Jude Medical, Tenax; has received consulting fees from the American College of Cardiology, AstraZeneca, BAROnova, Bayer, Bio2Medical, Boehringer Ingelheim, Bristol-Myers Squibb, Cubist, Eli Lilly, Elsevier, Epson, Forest, Johnson & Johnson, and Medtronic; and has equity in BioPrintFitness. Dr. Peterson has received research support from Eli Lilly & Co., Janssen Pharmaceuticals, and American Heart Association; and is a consultant and advisory board member for Boehringer Ingelheim, Bristol-Myers Squibb, Janssen Pharmaceuticals, Pfizer, and Genentech Inc. Dr. Piccini has received research support from Johnson & Johnson, Janssen Pharmaceuticals, Bayer HealthCare Pharmaceuticals, Boston Scientific, Johnson & Johnson; and is a consultant and advisory board member for Forest Laboratories, Inc., Medtronic, Johnson & Johnson, and Janssen Pharmaceuticals. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- atrial fibrillation
- creatinine clearance
- nonvitamin K antagonist oral anticoagulant
- Received August 8, 2016.
- Revision received September 19, 2016.
- Accepted September 20, 2016.
- The Authors
- ↵(2011) PRADAXA (dabigatran etexilate mesylate) [package insert] (Boehringer-Ingelheim, Ridgefield, CT).
- (2011) Xarelto (rivaroxaban) [package insert] (Janssen Pharmaceuticals, Titusville, NJ).
- (2013) Eliquis (apixaban) [package insert] (Bristol-Myers Squibb, Princeton, NJ).
- Schulman S.,
- Kearon C.,
- for the Subcommittee on Control of Anticoagulation of the Scientific and Standardization Committee of the International Society on Thrombosis and Haemostasis
- Bell A.D.,
- Gross P.,
- Heffernan M.,
- et al.
- Barra M.E.,
- Fanikos J.,
- Connors J.M.,
- et al.
- Chan K.E.,
- Edelman E.R.,
- Wenger J.B.,
- et al.
- Dias C.,
- Moore K.T.,
- Murphy J.,
- et al.