Introduction

Dabigatran is a novel, orally available, direct thrombin inhibitor that is currently approved to prevent stroke in patients with atrial fibrillation. The anticoagulant is easier to use than warfarin because laboratory monitoring is not required and there are fewer drug–drug interactions.¹ Accordingly, dabigatran use in the United States increased from 3.1% of all oral anticoagulation visits in 2010 to 18.9% in 2011.² By virtue of the growing availability of dabigatran, the drug is sometimes prescribed for other situations where the drug has not been rigorously studied. For example, in 2011, Kirkly et al² reported that 37% of treatment visits for dabigatran were for coronary heart disease, hypertensive heart disease, and venous thromboembolism. Similar trends will likely be seen with rivaroxaban, another novel anticoagulant that is directed against Factor Xa, which became available in 2011.³

Clinical Perspective on p 979

Dabigatran and rivaroxaban have not been studied in end-stage renal disease (ESRD) patients who were specifically excluded from Randomized Evaluation of Long-Term Anticoagulation Therapy (RE-LY) and Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET AF), 2 large randomized trials that were conducted in patients without renal failure.^4,5 Prescription of these novel oral anticoagulant drugs (NOAC) in dialysis patients is currently contraindicated because the drugs are cleared via the kidneys and drugs levels can bioaccumulate to precipitate bleeding.^6,7 Dabigatran is cleared by dialysis, which leads to a precipitous drop in the drug level, whereas rivaroxaban is not cleared by dialysis because 95% of rivaroxaban is protein bound.

We surveyed a large ESRD population, to describe the prescribing patterns of dabigatran and rivaroxaban in chronic hemodialysis patients with atrial fibrillation. We also compared the rate of bleeding in dialysis patients taking warfarin, dabigatran, or rivaroxaban.

Methods

Study Population and Data Sources

We generated prevalence plots, from October 2010 to October 2014, to describe the prevalence of dabigatran and rivaroxaban within a population of chronic hemodialysis patients with atrial fibrillation. In a secondary analysis, we identified the time point at which patients from the above hemodialysis population were initiated (de novo) on dabigatran, rivaroxaban, warfarin, or aspirin only. These patients were then followed for up to 2 years for bleeding and stroke outcomes.

Data for the study were abstracted from the Fresenius Medical Care North America (FMCNA) ESRD database which prospectively captures 1922 clinically pertinent and standardized data elements for the purpose of conducting large-scale outcomes studies in the dialysis population. More than 1500 clinics in 48 States, the District of Columbia, and the Territory of Puerto Rico were represented in the Database, which corresponds to ≈30% of the U.S. chronic dialysis population. The study protocol was approved and granted a waiver of informed consent by the New England Institutional Review Board.

All subjects registered in the Database are followed longitudinally, where data parameters are actively collected and entered at the point-of-care for each patient at every visit, which typically occurs 3 times per week. Demographic, comorbid, medication, vascular access, and dialysis treatment data are inputted by trained healthcare personnel into standardized electronic forms provided on touchscreen monitors located near each treatment station. The data collection tool includes predefined logic features and user alerts to check data as it is entered. Required fields are structured so that valid data must be entered before the system can authorize the hemodialysis treatment to begin. At the end of each dialysis shift, automatic edit checks are executed to identify inconsistent or out-of-range data which require immediate reconciliation by the facility charge nurse before the shift can be closed and submitted for billing. All laboratory processing was performed by a single accredited provider (Spectra Laboratories, Rockleigh, NJ), and the results were directly downloaded into the Database.

Data auditing occurred on the second week of each month when facility data quality reports were generated for assessment by FMCNA Clinical Quality Managers. The reports were also disseminated back to facilities for auditing by the clinical manager and senior clinical staff.

Outcome Measures

Among atrial fibrillation (AF) patients who started de novo oral anticoagulation after October 2010, we reported the percentage of patients who were initiated on dabigatran or rivaroxaban. Drug use was also reported by dose for patients on the full drug dose (dabigatran 150 mg BID and rivaroxaban 20 mg qD) and for patients on a reduced dose for moderate renal impairment (dabigatran 75 mg BID and rivaroxaban 15 mg qD).We used prevalence plots to graphically depict the longitudinal point prevalence of dabigatran and rivaroxaban in the full AF population on anticoagulation. Patients with a previous diagnosis of warfarin skin necrosis (n=4), protein S deficiency (n=0), protein C deficiency (n=0), or calciphylaxis (n=2) were excluded from the analysis. Caciphylaxis is a syndrome of vascular calcification, thrombosis, and skin necrosis that can be precipitated by warfarin use.

The end points of the secondary analysis were (1) bleeding (major and minor), and (2) embolic stroke or arterial embolism, within 2 years of medication initiation. Bleeding events were categorized into major or minor, and then further classified by the anatomic location of the bleed. Major bleeding was defined as a hemorrhagic event resulting in hospitalization or death. Cause of hospitalization or death was extracted from hospital discharge summaries or Form 2746-ESRD death notification, which is mandated by The Centers for Medicare & Medicaid Services for all dialysis deaths. The accuracy of the major bleeding end point was previously validated among 75 patients sampled from the Database (К=0.969).^7a Minor bleeding was defined as a hemorrhagic event where hemostasis was achieved without the need for hospitalization and did not result in death. Minor bleeding events were identified by reviewing dialysis nursing notes. For this, Dr Chan reviewed >900 000 pages of nursing notes for all study patients to identify documentation of patient or nurse reported bleeding events where hemostasis was achieved without the need of medical attention. The adjudicator was blinded to patient, treatment group, and outcome. Access bleeding was defined as (1) spontaneous bleeding from the arteriovenous shunt or exit site between dialysis sessions or (2) prolonged bleeding where >30 minutes of compression was required to achieve hemostasis after the needles were withdrawn from the vascular access at the end of the dialysis treatment. Pulmonary bleeding included hemoptysis and hemorrhagic effusion. Hematuria, vaginal, or penile bleeding was defined as urologic. Embolism was defined as the acute occlusion of an arterial vessel, excluding the heart and brain.

Medication Prescription and Patient Covariates

Patient prescriptions for oral and home medications were charted in the Research Database on standardized medication tables. Registered nurses at the dialysis facilities transcribed each drug name with its route, start date, and stop date as labeled on the medication bottle into the Database forms. Medication records were reconciled when patients initiated dialysis, routinely once per month, and in the immediate posthospitalization period.

For this study, medication records for each patient were electronically parsed for occurrences of dabigatran, rivaroxaban, warfarin, aspirin, Aggrenox, and clopidogrel, which were preceded by a diagnosis of atrial fibrillation in the patient record. The diagnostic accuracy of atrial fibrillation in the Database was previously found to be 75% when compared with electrocardiograms obtained from the same patient.⁸ Because the drug names in the medication records were entered as free text, Dr Chan reviewed each drug order for accuracy before being included in the study analysis.

Patients were categorized into study groups based on their first ever prescribed anticoagulant drug (de novo). Patients were then enrolled in the analysis only if this first prescription date of dabigatran, rivaroxaban, warfarin, or aspirin occurred after October 10, 2010, which was the day dabigatran was approved for use in the United States. Patient stroke risk was quantified by the CHADS₂ Score, whereas the risk for major bleeding was estimated using the Outpatient Bleeding Risk Index.^9,10

Statistical Analysis

We calculated and graphed the point prevalence (monthly, October 2010 to October 2014) of dabigatran and rivaroxaban, among the AF hemodialysis population on anticoagulation. The point prevalence was defined as the number of hemodialysis patients on dabigatran or rivaroxaban divided by the number of patients on dabigatran, rivaroxaban, or warfarin on the first day of each month (per 100 patients). A second series of prevalence plots with 95% confidence bands was generated as Figures I–III in the online-only Data Supplement.

For the secondary analysis, subjects were classified and followed from the time they initiated dabigatran, rivaroxaban, warfarin, or aspirin; by which ever drug was started first. Patient time was counted up until death, withdrawal, transplantation, transfer to another dialysis unit, or discontinuation of dabigatran, rivaroxaban, warfarin, or aspirin for >14 days. Baseline characteristics of patients initiated on dabigatran, rivaroxaban, warfarin, and aspirin were tabulated and compared using ANOVA (continuous variable) or χ² (categorical variable) testing.

Study end points were initially reported by event count and unadjusted Poisson event rate (per 100 patient-years) for each of the dabigatran, rivaroxaban, warfarin, or aspirin groups. Formal comparisons between the drug groups were done by rate ratio (RR) with 95% confidence intervals (CIs) using covariate adjusted Poisson regression. The final adjusted model included all parameters listed in Table 1 after backward variable selection with an exit criteria of P<0.05.

A second analysis was performed to further evaluate the robustness of the main conclusion with covariate matching to decrease the potential bias from unmeasured confounding. For this, each dabigatran (1:2 matching ratio) and rivaroxaban (1:2) subject was matched to a warfarin subjects on the 20 baseline characteristics listed in Table I in the online-only Data Supplement. Data matching was accomplished through a greedy algorithm supplied by Kosanke et al.¹¹ In short, cases were matched to controls by the shortest matching distance between case and control variables. Matches were made based on the current available choices without consideration of future case–control selection, hence the term greedy algorithm. Groups were examined for covariate balance, before rate ratios for major and minor bleeding were calculated between the drug groups.

A third sensitivity analysis was performed to determine the effect of warfarin management on the outcome of study; moreover, the increased bleeding in subjects on dabigatran and rivaroxaban may have been an artifact of suboptimal warfarin management where patients were being under dosed (ie, international normalized ratio [INR]<2). As such, we recalculated the rate ratios for major and minor bleeding including only warfarin patients with ≥60% of their INR readings between 2 to 3. Zoppo reported 60% of patients have an INR within the recommended range at any given time in usual clinical practice.¹²

We also explored the effect of drug dose on the risk of major bleeding associated with dabigatran and rivaroxaban. For this, we calculated the adjusted rate ratio for major bleeding in patients on full dose dabigatran (150 mg BID), reduced dose dabigatran (75 mg BID), full dose rivaroxaban (20 mg qD), and reduced dose rivaroxaban (15 mg qD) referent to patients on warfarin. Our final analysis explored the impact of dialysis efficiency on the risk of major bleeding associated with dabigatran and rivaroxaban. For this, we recalculated these rate ratios in dabigatran and rivaroxaban patients that were divided by urea clearance and dialysis time (top 50th percentile versus bottom 50th percentile).

Statistical analyses were executed with SAS v9.1 (Cary, NC).

Results

Prevalence of Anticoagulant Drugs

Among 316 859 chronic hemodialysis patients from October 2010 to October 2014, we identified 29 977 (9.5%) patients with AF. Within this AF cohort, the first record of dabigatran prescription occurred 45 days after the drug became available in the United States. The first prescription of rivaroxaban occurred 161 days after it became Food and Drug Administration–approved to prevent stroke from AF (Figure 1A). Of patients on dabigatran, 15.3% were prescribed the full dose (150 mg twice daily), whereas 84.7% of subjects were prescribed a reduced dose (75 mg twice daily) intended for patients with moderate chronic kidney impairment. Similarly for rivaroxaban, 32.1% of patients received the full dose (20 mg per day) whereas 67.8% received an adjusted dose (15 mg per day) for patients with a creatinine clearance between 15 to 50 cc/min (Figure 1B). Patients on dialysis typically have a creatinine clearance of <10 cc/min.

• Download figure
• Download PowerPoint

Among atrial fibrillation patients on dialysis who received anticoagulation over the 4-year study period, 3.1% were prescribed dabigatran, and 2.8% rivaroxaban. Overall, the use of dabigatran and rivaroxaban continues on an upwards trajectory where the prevalence of these drugs was 4.6 per 100 AF patients on anticoagulation in October 2014 (Figure 1A).

Baseline Patient Characteristics

From the AF cohort, we identified 8064 subjects who were started de novo on warfarin after October 10, 2010 for comparison with 6018 subjects start on aspirin, 281 subjects started on dabigatran, and 244 subjects started on rivaroxaban. There were statistically significant differences in patient characteristics between the drug groups (Table 1). Dabigatran patients were more likely to have higher hemoglobin levels and heparin doses. Rivaroxaban patients were less likely to be Caucasian, less likely to dialyze with a catheter, and had higher systolic blood pressure readings relative to the warfarin group. All patients on NOACs were also younger, on higher erythropoietin doses, and more likely to have a history of bleeding relative to the warfarin group. Aspirin patients had spent the least time on chronic dialysis. There were no clinically meaningful differences in CHADS₂, bleeding index risk, and Charlson scores between the groups. The average time on drug was 168 days for dabigatran, 211 days for aspirin, and 175 days for warfarin subjects. The average time on drug for rivaroxaban was 106 days and expectedly lower because the drug did not become available until 2011. Among warfarin patients, the mean INR was 2.2 (standard deviation=0.9), but only 13.7% of patients had ≥60% of their INR readings within the target of 2 to 3.

Unadjusted Event Rates

The event rate of major bleeding was highest in the dabigatran (83.1 events per 100 patient-years) and rivaroxaban group (68.4 events per 100 patient-years), although lower in the warfarin group at 35.9 events per 100 patient-years (Table 2). The mortality rate from bleeding was higher in patients on NOACs: dabigatran (19.2 deaths per 100 patient-years), rivaroxaban (16.2), warfarin (10.2), and aspirin (7.7). The highest event rate for minor bleeding also occurred in dabigatran (120.6 events per 100 patient-years) and rivaroxaban patients (149.4 events per 100 patient-years) as shown in Table 3. The lowest major and minor bleeding rates were seen in patients on aspirin. The composite rate of stroke or arterial embolism was 6.2 events per 100 patient-years among warfarin users, yet the rate was significantly lower for patients aspirin (5.0 events per 100 patient-years). There were too few stroke and arterial embolism events in the study to detect any meaningful associations between warfarin, dabigatran, and rivaroxaban (Table 4).

Covariate Adjusted Rate Ratios

In covariate adjusted Poisson regression, dabigatran (RR, 1.48; 95% CI, 1.21–1.81; P=0.0001) and rivaroxaban (RR, 1.38; 95% CI, 1.03–1.83; P=0.04) associated with a higher risk of major bleeding when compared with warfarin. The risk of hemorrhagic death was even larger with dabigatran (RR, 1.78; 95% CI, 1.18–2.68; P=0.006) and rivaroxaban (RR, 1.71; 95% CI, 0.93–3.12; P=0.07) referent to warfarin. Similar associations were seen with minor bleeding: dabigatran (RR, 1.17; 95% CI, 1.00–1.38; P=0.05) and rivaroxaban (RR, 1.35; 95% CI, 1.11–1.65; P=0.001) when compared with warfarin. Patients treated with aspirin had the lowest risk for major bleeding (RR, 0.78; 95% CI, 0.72–0.84; P<0.0001), minor bleeding (RR, 0.39; 95% CI, 0.34–0.46; P<0.0001), and hemorrhagic death (RR, 0.42; 95% CI, 0.28–0.64; P<0.0001) referent to warfarin.

Sensitivity Analyses

Two additional analyses were conducted to evaluate the robustness of the primary analysis. To diminish the measured and unmeasured differences in patient characteristics between the study groups, we matched each dabigatran and rivaroxaban subject to 2 warfarin subjects on 20 data parameters. The resulting cohort achieved balance across treatment groups (Table I in the online-only Data Supplement) where the risk of major bleeding was still associatively higher with dabigatran (RR, 1.64; 95% CI, 1.27–2.12; P=0.03) and rivaroxaban (RR, 1.39; 95% CI, 1.00–1.94; P=0.05) referent to warfarin. In our second analysis, we excluded warfarin patients who were potentially under or over anticoagulated by only including patients with ≥60% of their INR readings between 2 and 3. Here, dabigatran (RR, 1.46; 95% CI, 1.10–1.93; P=0.0001) and rivaroxaban (RR, 1.32; 95% CI, 0.93–1.87; P=0.03) still associated with an increased risk for major bleeding when compared with warfarin.

Effect of Dose and Dialysis on Dabigatran and Rivaroxaban

The unadjusted event rate of major bleeding increased with drug dosage: 69.1 events per patient year in dabigatran 150 mg BID, 25.4 in dabigatran 75 mg BID, 42.6 in rivaroxaban 20 mg qD, and 23.9 in rivaroxaban 15 mg qD. Using covariate adjusted Poisson models, we examined the risk of major bleeding by drug dose for dabigatran and rivaroxaban referent to warfarin (Figure 2). Dialysis patients prescribed the full dosage of dabigatran (RR, 2.90; 95% CI, 1.93–4.34; P<0.0001) or rivaroxaban (RR, 1.93; 95% CI, 1.22–3.04; P=0.04) had a higher risk of major bleeding than patients who were prescribed a lower drug dose intended for patients with moderate renal impairment (Figure 2).

• Download figure
• Download PowerPoint

We also examined dabigatran (dialyzable) versus rivaroxaban (nondialyzable) by dialysis dose and time.¹³ Patients on dabigatran (dialyzable) who achieved less urea clearance (<230 cc/min) had more major bleeding than patients whose urea clearance was ≥230 cc/min (Figure 2). The amount of dialysis urea clearance did not decrease the risk of bleeding with rivaroxaban likely because the drug is not affected by dialysis. Shorter dialysis treatment times also associated with an increased risk of bleeding with both dabigatran and rivaroxaban (Figure 2).

Discussion

Dabigatran and rivaroxaban are promising anticoagulants that overcome the inconveniences of frequent serum level monitoring for dosing with warfarin therapy. Both drugs were shown to be at least equivalent to warfarin at stroke prevention without increasing the risk of bleeding; however, these major clinical trials both specifically excluded ESRD patients from their study.^4,14 Our study is the first to evaluate these drugs in the dialysis population, and suggests concern given the increasing use of dabigatran and rivaroxaban in the ESRD population despite formal Food and Drug Administration warnings of caution in renal failure. In fact, our secondary analyses suggest excess morbidity and mortality from bleeding are associatively higher with dabigatran and rivaroxaban when compared to warfarin.

Dabigatran exerts its anticoagulant effect through the direct inhibition of thrombin activity by binding to the molecule with high affinity and specificity.¹⁵ Approximately 80% to 85% of the drug is excreted by the kidneys mostly through glomerular filtration. The half-life of the drug is 9 hours, but increases to 25 to 30 hours in individuals with a creatinine clearance <30 cc/min.¹³ Dabigatran is also dialyzable, and 50% to 60% of the drug is eliminated during a 4-hour treatment.¹⁶ A dose of 150 mg twice daily was approved for patients with a creatinine clearance >30 cc/min based on the results of the RE-LY trial.⁴ This study showed dabigatran reduced the stroke risk by 34% with no increase in major bleeding in comparison with warfarin, but excluded patients with severe kidney disease. Of note, a dose of 75 mg twice daily can be considered based only on pharmacokinetic data for patients with a creatinine clearance of 15 to 30 cc/min; however, the American College of Chest Physicians recommends dabigatran be contraindicated in patients with severe renal impairment (creatinine clearance ≤30 cc/min).⁶

Rivaroxaban is an oral factor Xa inhibitor where 33% of the drug undergoes renal elimination through proximal tubule secretion. The half-life of the drug is 8 hours and increases minimally to 9.5 hours in patients with severe renal disease (mean creatinine clearance was 22 cc/min).¹⁷ Unlike dabigatran, rivaroxaban is not dialyzable because the drug is highly bound to plasma proteins (92% to 95%). In comparison with warfarin, rivaroxaban was noninferior at preventing stroke in atrial fibrillation without significant difference in bleeding risk in The Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation (ROCKET AF). Like RE-LY, patients with a creatinine clearance <30 cc/min were excluded from ROCKET AF. Currently, the use of rivaroxaban at a reduced dose of 15 mg per day may be considered as second line therapy in patients with a creatinine clearance between 15 to 50 cc/min, even though the safety and efficacy of this approach has not been clinically validated; however, rivaroxaban is contraindicated for patients with ESRD or on hemodialysis.⁶

Our study describes the off-label prescribing patterns that may be expected soon after a new drug is approved and released on the market. When dabigatran and rivaroxaban were approved for use in the United States, there was almost no evidence that these drugs were effective or safe in patients on dialysis because such patients were excluded from randomized, controlled trials. Despite the lack of evidence in ESRD, it took only 45 days after the release of the drug before dabigatran use began to show up in our dialysis population. The use of these NOACs were sustained and increased over time. After only 4 years, almost 5% of anticoagulated patients were taking dabigatran or rivaroxaban.

Our study also highlights the potential for harm when dabigatran is used in dialysis patients, where kidney failure impairs the clearance of the drug. This may lead to bioaccumulation of the drug that places patients at increased risk for serious hemorrhage. The risk may be mitigated by increasing the dialysis dose (Figure 2), but it does not entirely eliminate the concern. The periodic nature of hemodialysis will likely result in fluctuating and unpredictable blood levels of dabigatran, which could result in recurring periods of under- and overanticoagulation. It is also well known that dialysis treatments are sometimes shortened, missed, or delayed. This may result in prolonged periods of increased risk for bleeding, especially if the patient continues taking dabigatran in the absence of dialysis. For rivaroxaban, which is not dialyzable, we found little risk reduction when these patients were aggressively dialyzed leaving no means of dealing with excessive exposure to the drug.

We also bring attention to the rate of hemorrhagic stroke (Table 2) in warfarin patients, which was 4 times higher than the rate in dabigatran patients whereas there were no hemorrhagic strokes among rivaroxaban patients. These findings are consistent with trial results in the general population; however, vascular access was the largest cause of morbidity and mortality from bleeding in the dialysis population.^4,5

Finally, comparison of bleeding morbidity and mortality risk between the medications also provided further important information. For dabigatran, the risk of major bleeding increased by 48% and the risk of fatal bleeding increased even more, by 88% referent to warfarin. Similarly, for rivaroxaban, the risk of major bleeding increased 38%, yet increased even higher by 58% for fatal bleeding in comparison with warfarin. Overall, NOACs may magnify the risk of hemorrhagic death more than the risk of nonfatal bleeding. Perhaps the lack of reversal agents for dabigatran and rivaroxaban complicate the management of bleeding events where more patients will die from exsanguination. Taken together, reversal agents are less likely to decrease the risk of bleeding but may have a role in improving the treatment and prognosis of bleeding events in dabigatran and rivaroxaban patients.

The limitations of our study suggest room for further study. We had few stroke and arterial embolic events in the patients on dabigatran and rivaroxaban, such that our cerebrovascular analysis was underpowered. The increased risk of major bleeding with dabigatran and rivaroxaban may be acceptable for dialysis patients if these drugs also substantially decrease the risk of stroke more effectively than warfarin; however, the NOACs did not trend toward lower stroke rates than warfarin in our study. A study with more subjects followed for a longer period of time is needed to completely evaluate the risk–benefit profile of these drugs, and the inclusion of patients with chronic renal insufficiency who are not dialyzed is further worthy of study. Other limitations include the lack of blood transfusion data and the potential for confounding by indication. We mitigated this potential bias from unmeasured factors by performing a matched analysis which supported the main findings of the study.

In conclusion, more dialysis patients are being started on dabigatran and rivaroxaban, even when their use is contraindicated in ESRD. Our data support the current labeling of dabigatran and rivaroxaban, which advise against use in patients on dialysis. Further research is needed into the safety and efficacy of these agents in the ESRD population before they can be recommended.

Acknowledgments

Dr Chan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Footnotes