Risks and Benefits of Direct Oral Anticoagulants across the Spectrum of GFR among Incident and Prevalent Patients with Atrial Fibrillation

Abstract

Background and objectives

All randomized trials of direct oral anticoagulants in atrial fibrillation excluded patients with severe kidney disease. The safety and effectiveness of direct oral anticoagulants across the range of eGFR in real-world settings is unknown. Our objective is to quantify the risk of bleeding and benefit of ischemic stroke prevention for direct oral anticoagulants compared with warfarin in patients with atrial fibrillation with and without CKD.

Design, setting, participants, & measurements

We created a propensity score–matched cohort of 3206 patients with atrial fibrillation and direct oral anticoagulant use and 3206 patients with atrial fibrillation using warfarin from October of 2010 to February of 2017 in an electronic health record (Geisinger Health System). The risks of bleeding and ischemic stroke were compared between direct oral anticoagulant and warfarin users using Cox proportional hazards regression, stratified by eGFR (≥60 and <60 ml/min per 1.73 m²).

Results

The mean (SD) age of the 6412 participants was 72 (12) years, 47% were women, and average eGFR was 69 (21) ml/min per 1.73 m². There were 1181 bleeding events and 466 ischemic strokes over 7391 person-years of follow-up. Compared with warfarin use, the hazard ratios (HRs) (95% confidence interval [95% CI]) of bleeding associated with direct oral anticoagulant use were 1.01 (0.88 to 1.17) and 1.23 (1.02 to 1.48) for those with eGFR≥60 and eGFR<60 ml/min per 1.73 m², respectively (P-interaction=0.10). There was no difference between direct oral anticoagulant and warfarin users in the risk of ischemic stroke: HRs (95% CI) of 0.94 (0.74 to 1.18) and 1.02 (0.76 to 1.37) for those with eGFR≥60 and eGFR<60 ml/min per 1.73 m², respectively (P-interaction=0.70). Similar findings were observed with individual drugs.

Conclusions

In a large health care system, patients with eGFR<60 ml/min per 1.73 m² who took direct oral anticoagulants for atrial fibrillation had slightly higher risk of bleeding compared with those on warfarin, but similar benefits from prevention of ischemic stroke.

Introduction

Atrial fibrillation, the most common type of heart arrhythmia, increases a person’s risk of ischemic stroke; thus, oral anticoagulant therapy is recommended to reduce the risk (1,2). Until 2009, warfarin, a vitamin K antagonist, was the most commonly used anticoagulant. Although warfarin is effective in preventing ischemic stroke (3), it also increases the risk of bleeding, ranging from minor to fatal hemorrhage (4). Moreover, warfarin use is limited by a narrow therapeutic index which requires frequent monitoring, resulting in substantial burden to patients (5,6).

Several direct oral anticoagulants, including direct thrombin inhibitors (e.g., dabigatran) and direct factor Xa inhibitors (e.g., rivaroxaban and apixaban), became available over the last decade as alternatives to warfarin. They offer potential advantages over warfarin, including rapid onset and offset of action, fewer drug and food interactions, and predictable anticoagulant effects without the need for frequent monitoring. Clinical trials reported that direct oral anticoagulants have similar or greater efficacy and safety in relation to warfarin in general atrial fibrillation populations (7–9).

All of the direct oral anticoagulants undergo kidney clearance to varying degrees. As a result, elimination of direct oral anticoagulants is slower in patients with CKD, which may predispose patients to drug accumulation and a greater risk of bleeding events (10). All randomized trials of direct oral anticoagulants for treatment of atrial fibrillation excluded patients with severe kidney disease (creatinine clearance <25–30 ml/min) (7–9). Despite the exclusion of such patients, direct oral anticoagulants have been approved by the US Food and Drug Administration (FDA) for use in patients with severe kidney disease (i.e., dabigatran and rivaroxaban for creatinine clearance 15–30 ml/min and apixaban for creatinine clearance <15 ml/min). Given that 25% of patients with atrial fibrillation have CKD (11), the real-world safety and effectiveness of direct oral anticoagulants in patients with atrial fibrillation across the spectrum of GFR is of great public health importance.

Using a large, tertiary health system, we investigated the prevalence of direct oral anticoagulant use in patients with atrial fibrillation by eGFR category and used propensity score–matching techniques to evaluate the risk of bleeding and benefit of ischemic stroke prevention for direct oral anticoagulants compared with warfarin within categories of eGFR.

Materials and Methods

Study Population

We identified 20,727 patients with a diagnosis of atrial fibrillation, on the basis of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code 427.31, 427.32, or 427.3, who filled at least one prescription for anticoagulants and had a serum creatinine measured between October 19, 2010 (United States dabigatran approval date) and February 2, 2017 in the Geisinger Health System. Geisinger is a fully integrated rural health care system that serves a stable patient population (1% annual outmigration rate) with more than 3 million residents throughout 45 counties in central and northeastern Pennsylvania.

Exposure and Baseline Covariates

The primary exposure was direct oral anticoagulant use, ascertained from prescriptions. Warfarin use was considered an active comparator. A gap in prescriptions <60 days was not considered a medication discontinuation to allow for the possibility of stockpiling of medications.

A priori–selected covariates included in the propensity score model comprised factors potentially associated with treatment selection and/or outcomes, including demographics, eGFR, prescription year, history/duration of previous warfarin use, number of prior hospitalizations, comorbidities, and medication use (Table 1). The CHA₂DS₂–VASc (congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, stroke/transient ischemic attack/thromboembolism, vascular disease, age 65 to 74 years, sex category) score (12) and the HAS-BLED (hypertension, abnormal kidney/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly [age ≥65 years], drugs/alcohol concomitantly) score (13), which predict the risk of stroke and bleeding in patients with atrial fibrillation, were also included in the propensity score model.

Table 1.

Baseline characteristics in propensity score–matched patients with atrial fibrillation using direct oral anticoagulants or warfarin from October of 2010 to February of 2017 in the Geisinger Health System

Variable	Direct Oral Anticoagulants (n=3206)	Warfarin (n=3206)	Standardized Mean Difference
Mean age (SD), yr	73 (11)	72 (12)	0.02
Female, %	47	46	0.01
White, %	98	98	0.01
Mean eGFR (SD), ml/min per 1.73 m2a	69 (21)	68 (22)	0.01
eGFR category, %
≥60 ml/min per 1.73 m2	65	65	0.01
30–59 ml/min per 1.73 m2	31	31	0.000
<30 ml/min per 1.73 m2	4	4	0.01
Year, %
2010–2012	22	22	0.001
2013–2014	35	34	0.02
2015–2017	43	44	0.02
Type of direct oral anticoagulant, %
Dabigatran	27	—	—
Rivaroxaban	41	—	—
Apixaban	32	—	—
History of prior anticoagulation, %	37	37	0.01
Median duration of prior anticoagulation (IQR), yrb	3 (1–6)	3 (1–6)	0.03
Mean CHA2DS2–VASc score (SD)	4 (2)	4 (2)	0.01
Mean HAS-BLED score (SD)	2 (1)	2 (1)	0.02
Number of prior hospitalizations, %
None	33	34	0.03
1–3	36	37	0.02
4–7	19	17	0.06
≥8	12	12	0.003
Comorbidities, %
Hypertension	83	83	0.02
Diabetes	34	34	0.01
Congestive heart failure	31	31	0.01
Valvular heart disease	24	26	0.03
Valvular atrial fibrillationc	4	5	0.04
Myocardial infarction	15	15	0.01
Coronary artery disease	41	41	0.01
Peripheral artery disease	16	16	0.01
Deep vein thrombosis	10	10	0.004
Pulmonary embolism	5	5	0.01
Bleeding	35	35	0.004
Anemia	37	37	0.002
Transient ischemic attack	13	14	0.01
Ischemic stroke	9	9	0.01
ESKD	0.4	0.5	0.02
Medication use, %
NSAIDs	24	24	0.01
Antiplatelet agents	22	23	0.03
Statins	32	31	0.001
RAS blockers	30	30	0.02

All P values are >0.05. —, not applicable; IQR, interquartile range; CHA₂DS₂–VASc, congestive heart failure, hypertension, age ≥75 years, diabetes mellitus, stroke/transient ischemic attack/thromboembolism, vascular disease, age 65 to 74 years, sex category; HAS-BLED, hypertension, abnormal kidney/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly [age ≥65 years], drugs/alcohol concomitantly; NSAIDs, nonsteroidal anti-inflammatory drugs; RAS blockers, renin-angiotensin system blockers.

^aMedian time (IQR) between eGFR measurements and initial dosing of direct oral anticoagulant was 2 d (0–45).

^bOnly among those with a history of prior anticoagulation.

^cValvular atrial fibrillation was defined as atrial fibrillation with rheumatic mitral stenosis or prosthetic heart valve.

eGFR was estimated from the serum creatinine measured (outpatient or inpatient) within 1 year before the baseline prescription date or up to 90 days after baseline, using the CKD Epidemiology Collaboration Equation (14). eGFR was categorized into three groups (≥60, 30–59, and <30 ml/min per 1.73 m²) according to the Kidney Disease Improving Global Outcomes guideline and CKD was defined as eGFR<60 ml/min per 1.73 m² (15). Baseline year was categorized into three groups (2010–2012, 2013–2014, and 2015–2017). Prior hospitalizations was categorized as none, 1–3, 4–7, and ≥8. Comorbid conditions were identified by the presence of relevant ICD-9-CM codes at any time before baseline (Supplemental Table 1). History of anemia was defined as blood hemoglobin <11 g/dl at any time before baseline. ESKD was identified by linkage to the United States Renal Data System. Baseline use of nonsteroidal anti-inflammatory drugs, antiplatelet drugs, statins, and renin-angiotensin system (RAS) blockers was captured from prescription data within 90 days before inclusion in the study. Creatinine clearance by the Cockcroft–Gault Equation (16) on the basis of serum creatinine and body weight measured within 1 year before or up to 90 days after enrollment was used for FDA dose recommendation for dabigatran and rivaroxaban in secondary analyses. During follow-up among warfarin users, time in therapeutic range was determined by the Rosendaal method (17). Labile international normalized ratios (INRs) were defined as time in therapeutic range <60%.

Outcomes Definitions

The outcomes were bleeding and ischemic stroke, ascertained by ICD-9-CM codes adapted from previous validation studies (18,19). A definite bleeding code without a corresponding trauma was used to define a bleeding event, and both major (i.e., bleeding at critical site [intracranial, retroperitoneal, intraspinal, intraocular, pericardial, or intraarticular] or bleeding requiring transfusion) and minor bleeding were included (18). The codes for bleeding and ischemic stroke have a positive predictive value of 89% and 90%, respectively (18,19).

Propensity Matching Procedure

A total of 3655 patients initiated direct oral anticoagulants (dabigatran, rivaroxaban, or apixaban) since October 19, 2010. Eight were ineligible due to a missing or invalid death date (n=8). We then used 1:1 nearest-neighbor propensity score–matching techniques (a caliper of 0.25 propensity score SD) to match direct oral anticoagulant users to warfarin users (20). This procedure was repeated separately for each drug. A propensity score match was obtained for 3206 direct oral anticoagulant users. The baseline date was considered the first direct oral anticoagulant prescription date for direct oral anticoagulant users and the matched warfarin prescription date for warfarin users.

Statistical Analyses

The balance of baseline characteristics between the matched cohorts was assessed using the standardized mean difference, a measure not influenced by sample size; values<0.1 indicate a negligible difference between groups (21). Incidence rates for each outcome were compared between the matched cohorts by category of eGFR. Cox proportional hazards regression models, stratified by CKD status, were used to estimate the hazard ratios (HRs) and 95% confidence intervals (95% CIs) of each outcome associated with direct oral anticoagulant use. The proportional hazards assumption was tested using Schoenfeld residuals (Supplemental Table 2).

In primary analysis, we censored participants at the first occurrence of a study outcome, death (ascertained via linkage to the National Death Index), prescription for a different anticoagulant, medication discontinuation, or end of the study period (i.e., as-treated analyses). In sensitivity analyses, we: (1) used an intention-to-treat approach, (2) used time-varying drug exposure, (3) excluded those on dialysis and censored at dialysis initiation, (4) excluded those with an AKI diagnostic code within 3 months of the baseline serum creatinine, (5) used an average eGFR instead of a single eGFR, (6) used a 30-day gap for defining medication discontinuation, (7) excluded those with rheumatic mitral stenosis or prosthetic heart valve, and (8) stratified by prior anticoagulation history.

We also assessed the proportion of direct oral anticoagulant dose by eGFR and proportion taking the FDA recommended level of direct oral anticoagulant dose. Cumulative incidence of treatment change during the study period was estimated accounting for competing events of treatment change (i.e., discontinuation, switch to warfarin, or switch to other direct oral anticoagulants), because these three were mutually exclusive. All analyses were conducted using Stata/MP 14.1 (StataCorp, College Station, TX). This study was reviewed and deemed exempt by the Geisinger Institutional Review Board as well as the Johns Hopkins University Institutional Review Board.

Results

Prevalence of Direct Oral Anticoagulant Use over Time by eGFR Category

Among the 20,727 patients with atrial fibrillation, prescription of direct oral anticoagulants increased from 0.05% to 21% from 2010 to 2016, including within each category of eGFR (all P<0.001) (Figure 1). Among those with eGFR≥60 ml/min per 1.73 m², the most frequently prescribed drug in 2016 was rivaroxaban (11%), followed by apixaban (10%) and dabigatran (3%). Among those with eGFR<60 ml/min per 1.73 m², apixaban was the most frequently prescribed (8% versus 6% and 2% for rivaroxaban and dabigatran, respectively). Thirty-five percent of direct oral anticoagulant users had eGFR<60 ml/min per 1.73 m².

Figure 1.

Direct oral anticoagulant prescriptions increased over time among patients with atrial fibrillation in all stages of CKD in the Geisinger Health System (n=20,727). The total number of patients in each year from 2010 to 2016 was 6622, 6826, 7282, 7964, 8445, 8781, and 9001, respectively. CKD stage G1–2: eGFR≥60, G3: eGFR 30–59, and G4–5: eGFR<30 ml/min per 1.73 m².

Characteristics of Direct Oral Anticoagulant Users

The mean (SD) age of new direct oral anticoagulant users was 70 (12) years, 45% were female, 97% were white, and mean (SD) eGFR was 70 (22) ml/min per 1.73 m² (Supplemental Table 3). At baseline, 2484 (68%), 1038 (28%), and 130 (4%) had eGFR≥60, eGFR 30–59, and eGFR<30 ml/min per 1.73 m², respectively. By type of drugs, 880 (24%), 1560 (43%), and 1207 (33%) were initially prescribed dabigatran, rivaroxaban, and apixaban. On average, dabigatran and rivaroxaban users were similar; however, apixaban users were older, more often women, and had slightly lower eGFR, higher CHA₂DS₂–VASc score, and higher HAS-BLED score than the other direct oral anticoagulant users. Approximately 34% of direct oral anticoagulant users had a history of warfarin use (median duration [interquartile range], 3 years [1–6]). Direct oral anticoagulant users with a history of warfarin use, on average, had a higher risk profile than warfarin-naïve direct oral anticoagulant users (Supplemental Table 4).

Analyses Using the Matched Cohorts

Appropriate warfarin-using matches were found for 3206 (88%) of the direct oral anticoagulant users, resulting in matched cohorts closely balanced for all baseline covariates (Supplemental Tables 5–7, Table 1, all standardized mean differences <0.1). Among 3206 warfarin users, 3016 (94%) had the laboratory data of INR control. Overall, 57% of warfarin users had labile INRs and those with lower eGFR had worse anticoagulation control (67%, 54%, and 57% had labile INRs with eGFR<30, 30–59, and ≥60 ml/min per 1.73 m², respectively). There were 1181 bleeding events (incidence rate [direct oral anticoagulants versus warfarin]=20.0 versus 17.9 per 100 person-years) and 466 ischemic strokes (incidence rate=6.9 versus 6.6) among 6412 patients with 7391 person-years of follow-up (average [SD] follow-up time=1.2 [1.3] years). The incidence rates (95% CI) of bleeding were 17.3 (15.7 to 19.3) versus 16.8 (15.3 to 18.6) among the 4169 people with eGFR≥60, 25.2 (21.9 to 29.0) versus 19.0 (16.4 to 21.9) among the 1990 people with eGFR 30–59, and 38.1 (26.3 to 55.2) versus 30.4 (21.5 to 43.0) for the 253 people with eGFR<30 ml/min per 1.73 m², respectively (Figure 2A). When stratified by CKD status, the HRs (95% CI) of bleeding for direct oral anticoagulant compared with warfarin users were 1.01 (0.88 to 1.17) and 1.23 (1.02 to 1.48) in eGFR≥60 and eGFR<60 ml/min per 1.73 m², respectively (P-for-interaction=0.10, Table 2). The incidence rates of ischemic stroke (95% CI) were 6.0 (5.1 to 7.1) versus 6.0 (5.1 to 7.1), 8.8 (7.0 to 10.9) versus 7.5 (6.0 to 9.3), and 9.2 (4.6 to 18.4) versus 10.6 (6.0 to 18.7) for those with eGFR≥60, eGFR 30–59, and eGFR<30 ml/min per 1.73 m², respectively (Figure 2B). The HRs (95% CI) of ischemic stroke were 0.94 (0.74 to 1.18) and 1.02 (0.76 to 1.37) for those with eGFR≥60 and eGFR<60 ml/min per 1.73 m², respectively (P-for-interaction=0.70). Similar findings were observed when each drug was analyzed individually. The results were consistent when intention-to-treat or time-varying drug exposure analyses were done (Table 2). Additional sensitivity analyses showed similar findings: analyses excluding those on dialysis at baseline and censoring at dialysis initiation (Supplemental Table 8), exploring potential effects of misclassification of baseline eGFR (Supplemental Table 9), using an alternative gap (30 days) for defining medication discontinuation (Supplemental Table 10), and excluding those with valvular atrial fibrillation (Supplemental Table 11). There was limited sample size in stratified analyses by prior anticoagulation history with no significant results (Supplemental Table 12).

Figure 2.

Patients with low eGFR on direct oral anticoagulants for treatment of atrial fibrillation experienced bleeding events more frequently than those on warfarin, but had similar rates of ischemic stroke. Incidence rates of bleeding (A) and ischemic stroke (B) by eGFR category among propensity-score matched patients with atrial fibrillation using direct oral anticoagulants or warfarin. IR, incidence rate; PYs, person-years.

Table 2.

Risk of outcomes among patients receiving direct oral anticoagulants versus warfarin, overall and stratified by eGFR

Outcome	Number of Events		IR, per 100 PYs		HR (95% CI)	P Value	P-inter
	Direct Oral Anticoagulants	Warfarin	Direct Oral Anticoagulants	Warfarin
All bleeding (all/major/minor)
Analyses 1
Overall	574/76/498	607/126/481	20.0/2.7/17.3	17.9/3.7/14.2	1.08 (0.97 to 1.22)	0.16
eGFR≥60	350/39/311	390/76/314	17.3/1.9/15.4	16.8/3.3/13.5	1.01 (0.88 to 1.17)	0.88	0.10
eGFR<60	224/37/187	217/50/167	26.3/4.4/21.9	20.1/4.6/15.5	1.23 (1.02 to 1.48)	0.03
Analyses 2
Overall	863/111/752	805/155/650	14.9/1.9/13.0	13.8/2.7/11.1	1.09 (0.99 to 1.20)	0.08
eGFR≥60	537/63/474	520/93/427	12.9/1.5/11.4	12.7/2.3/10.4	1.03 (0.91 to 1.16)	0.65	0.11
eGFR<60	326/48/278	285/62/223	20.0/2.9/17.1	16.3/3.6/12.7	1.21 (1.03 to 1.42)	0.02
Analyses 3
Overall	574/76/498	740/142/740	16.8/2.2/14.6	15.1/2.9/12.2	1.18 (1.05 to 1.31)	0.004
eGFR≥60	350/39/311	475/87/388	14.7/1.6/13.0	14.1/2.6/11.5	1.11 (0.96 to 1.27)	0.15	0.10
eGFR<60	224/37/187	265/55/210	21.8/3.6/18.2	17.2/3.6/13.6	1.33 (1.11 to 1.58)	0.002
Ischemic stroke
Analyses 1
Overall	222	244	6.9	6.6	0.97 (0.81 to 1.16)	0.72
eGFR≥60	136	152	6.0	6.0	0.94 (0.74 to 1.18)	0.59	0.70
eGFR<60	86	92	8.8	7.8	1.02 (0.76 to 1.37)	0.91
Analyses 2
Overall	305	301	4.4	4.5	1.00 (0.83 to 1.17)	0.99
eGFR≥60	191	193	4.0	4.2	0.97 (0.79 to 1.18)	0.73	0.57
eGFR<60	114	108	5.6	5.3	1.06 (0.82 to 1.38)	0.64
Analyses 3
Overall	222	282	6.9	6.0	0.94 (0.79 to 1.12)	0.48
eGFR≥60	136	180	6.1	5.6	0.89 (0.71 to 1.11)	0.30	0.49
eGFR<60	86	102	8.8	6.8	1.03 (0.77 to 1.37)	0.85

Analyses 1: as-treated analyses; analyses 2: intention-to-treat analyses; analyses 3: time-varying analyses. IR, incidence rate; PY, person-years; HR, hazard ratio; 95% CI, 95% confidence interval; P-inter, P-for-interaction.

Overall, 21% of direct oral anticoagulant users were prescribed a reduced dose. Only 11% were prescribed a reduced dose among direct oral anticoagulant users with eGFR≥60, whereas 36% and 72% were prescribed a reduced dose among those with eGFR 30–59 and eGFR<30 ml/min per 1.73 m² (Figure 3A). Among direct oral anticoagulant users with eGFR<30 ml/min per 1.73 m², 22% of dabigatran users, 21% of rivaroxaban users, and 31% of apixaban users received full dose. Approximately 11% of direct oral anticoagulant users needed dose reduction according to the FDA recommendations. Among those who needed dose reduction, 17% of dabigatran users, 35% of rivaroxaban users, and 15% of apixaban users received inappropriately a full dose (Figure 3B). At 1 year after initiation of direct oral anticoagulants, 3% had switched to other direct oral anticoagulants, 19% had switched to warfarin, and 28% had discontinued them (Figure 4). In comparison, at 1 year of follow up, 38% of warfarin users had discontinued use.

Figure 3.

Pattern of prescribed direct oral anticoagulant dose by eGFR and the FDA recommended dose level. Adding the sample size of cohort of each drug in (A) does not give 3206, the total number of matched direct oral anticoagulant users, because individuals changing from one type to the others were included in two or three different cohorts of each drug. The sample sizes in (B) are slightly smaller than those in (A) due to missing body weight. The current FDA dosing guidelines were applied to all past dosing of each direct oral anticoagulant: Dabigatran 150 mg twice a day and 75 mg twice a day for those with creatinine clearance >30 and creatinine clearance 15–30 ml/min, respectively. Rivaroxaban 20 mg every day and 15 mg every day for those with creatinine clearance >50 and creatinine clearance 15–50 ml/min, respectively. Apixaban 5 mg twice a day unless dose reduction is recommended. Apixaban 2.5 mg twice a day for those with at least two of the following conditions: age≥80 years, body wt ≤60 kg, or serum creatinine ≥1.5 mg/dl. FDA, Food and Drug Administration.

Figure 4.

Cumulative incidence of treatment change during follow-up among matched direct oral anticoagulant users and warfarin users with atrial fibrillation. *Baseline time for warfarin users (A) is the warfarin prescribed date matched with duration of previous warfarin use, in comparison with direct oral anticoagulant users (B). Thirty-nine percent of discontinuation of anticoagulant occurred after bleeding events.

Discussion

Despite sparse evidence of safety and effectiveness in patients with atrial fibrillation and CKD, we demonstrate that prescription of direct oral anticoagulants has increased substantially within all categories of eGFR in a health care setting. In our real-world data, approximately 20% of all patients with atrial fibrillation treated with anticoagulation received a prescription for a direct oral anticoagulant in 2016, including 17% and 10% of those with eGFR 30–59 and <30 ml/min per 1.73 m², respectively. Compared with propensity-matched warfarin users, direct oral anticoagulant use resulted in similar benefit in terms of ischemic stroke prevention but was associated with a 23% higher risk of bleeding among those with eGFR<60 ml/min per 1.73 m². These findings might suggest the need for heightened caution in prescribing and monitoring of direct oral anticoagulants in patients with CKD.

Our results add to a recent study which showed increased prescription of direct oral anticoagulants in patients with atrial fibrillation and advanced CKD, despite the most recent American Heart Association (AHA), American College of Cardiology (ACC), and European Society of Cardiology (ESC) guidelines recommending against the use of direct oral anticoagulants when creatinine clearance <30 ml/min (1,2). The previous analysis reported the most commonly used direct oral anticoagulant in 2015 was apixaban (10%), followed by rivaroxaban (9%) and dabigatran (3%), but they did not measure bleeding rates or ischemic stroke prevention (22). Our findings of a trend toward higher bleeding risk and similar benefit in ischemic stroke prevention in CKD support the current guidelines, which recommend refraining from direct oral anticoagulant use in patients with atrial fibrillation and advanced CKD.

Although the US FDA provides dosing for direct oral anticoagulants when creatinine clearance is as low as 15 ml/min for dabigatran and rivaroxaban and <15 ml/min for apixaban, the FDA labeling was mostly on the basis of small, single-dose pharmacokinetics studies without clinical trial outcomes data (23–26). Moreover, labels by the European Medicines Agency do not support this practice. None of the three pivotal direct oral anticoagulant trials showed higher risk of bleeding in subgroups of patients with moderate CKD (creatinine clearance <50 ml/min), and all showed a trend toward lower risk of ischemic stroke (27–29). However, in one of these trials, when eGFR was assessed continuously, bleeding risk was higher among dabigatran users compared with warfarin users with eGFR<50 ml/min per 1.73 m² (27). In our study, direct oral anticoagulant use was significantly associated with higher risk of bleeding among those with eGFR<60 ml/min per 1.73 m², and a trend toward higher risk of bleeding was consistently observed across three types of drugs with lower eGFR. On the other hand, we found no difference in ischemic stroke prevention between direct oral anticoagulant and warfarin users in any eGFR category. We note that our study population was quite different from that of the clinical trials, all of which excluded patients with advanced CKD and certain types of valvular heart disease, although it is clear that many of the patients who might be eligible for direct oral anticoagulant prescription have these conditions in real-world practice (11,30). We found 28% of patients recommended to receive dose reduction by the FDA recommendation were prescribed full-dose therapy. This might be because the FDA recommends drug dose on the basis of creatinine clearance, whereas GFR is widely being used in clinical practice.

Our study is the first study to evaluate risks and benefits of direct oral anticoagulants across the spectrum of GFR in patients with atrial fibrillation, using a heterogeneous but stable study population from a large health care system. Rigorous propensity score–matching techniques were used to create a cohort with comparable characteristics between treatment groups, minimizing bias. Moreover, our study’s follow-up began right from FDA approval, affording us some of the longest follow-up times possible. However, there are also certain limitations. First, our study is observational. Despite propensity score matching, direct oral anticoagulant users may differ from warfarin users in unidentified ways that relate to the risk of bleeding or ischemic stroke (e.g., treatment adherence). Second, analyzing only one health care system which included few (n=123, 2%) nonwhite patients limits the generalizability of our findings. Third, we had few patients with eGFR<30 ml/min per 1.73 m², precluding many analyses in this subgroup. Forth, our study population included direct oral anticoagulant users both with and without history of previous anticoagulation. Although pharmacoepidemiology studies often utilize a new user design to mitigate chronology bias, we included all users because we aimed to assess safety and effectiveness in real-world settings, where many direct oral anticoagulant users switch from warfarin. Further, to address this potential source of bias, we matched on duration of previous anticoagulation, which enabled comparisons at a comparable time in the natural history of their treatment. Fifth, outcomes were identified through the use of ICD-9-CM codes. However, we adapted ICD-9-CM codes for outcome ascertainment from previous validation studies (18,19). If misclassification of outcomes exists, it would be expected to be nondifferential, thus our observed associations are conservative. Finally, we did not assess risk of site-specific bleeding (e.g., gastrointestinal bleeding or intracranial hemorrhage).

In conclusion, we report a higher bleeding risk and similar ischemic stroke prevention with direct oral anticoagulants compared with warfarin in patients with atrial fibrillation with eGFR<60 ml/min per 1.73 m² in a clinical practice setting. These findings lend support to the current clinical guidelines by AHA/ACC and ESC, which suggest refraining from direct oral anticoagulant use for treatment of atrial fibrillation in advanced CKD, and raise safety concerns regarding the current US FDA labels. Further large-scale studies are warranted to confirm our findings. For now, we suggest caution in prescribing direct oral anticoagulants for patients with advanced CKD.

Disclosures

G.C.A. is Chair of the US Food and Drug Administration’s Peripheral and Central Nervous System Advisory Committee; has served as a paid consultant to IQVIA and serves on the Advisory Board of MesaRx Innovations; holds equity in Monument Analytics, a health care consultancy whose clients include the life sciences industry as well as plaintiffs in opioid litigation; and serves as a member of OptumRx’s Pharmacy and Therapeutics Committee. This arrangement has been reviewed and approved by the Johns Hopkins Bloomberg School of Public Health.