Safety and Effectiveness of Apixaban versus Warfarin by Kidney Function in Atrial Fibrillation

Visual Abstract

Abstract

Key Points

• This real-world study involved a large cohort of 38,598 adults with atrial fibrillation from five jurisdictions across Australia and Canada.

• This study supports the use of apixaban as a safe and effective alternative to warfarin for atrial fibrillation across differing levels of kidney function.

• This study also adds important safety data on the use of apixaban in patients with reduced kidney function.

Background

Evidence to guide the use of apixaban in people with atrial fibrillation (AF) and CKD in routine clinical practice has been limited. We assessed comparative safety (major bleeding) and effectiveness (ischemic stroke and death) of apixaban versus warfarin in patients with AF across the spectrum of non–dialysis-dependent CKD using large, routinely collected data.

Methods

We combined findings from five retrospective cohorts (2013–2018) across Australia and Canada. Adults with AF, new dispensation of apixaban or warfarin, and a recorded eGFR grouped as ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m² were included. Patients on dialysis or kidney transplant recipients were excluded. We assessed outcomes within 1 year of initiating either therapy: (1) composite of all-cause death, ischemic stroke, or transient ischemic attack and (2) first hospitalization for major bleeding (intracranial, gastrointestinal, or other). Cox models estimated hazard ratios (HRs; 95% confidence intervals) for outcomes across eGFR categories, after 1:1 matching using propensity scores. We summarized center-level data using random effects meta-analysis.

Results

Among 38,598 matched apixaban and warfarin users, there were 4130 (10.7%) ischemic and 697 (1.8%) bleeding events within 1 year. Apixaban was associated with lower or similar risk for the ischemic outcome compared with warfarin in all eGFR categories (pooled HRs [95% confidence interval]: 0.78 [0.64 to 0.94], 0.77 [0.62 to 0.97], 0.82 [0.68 to 0.98], and 0.99 [0.68 to 1.45] for eGFR ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m², respectively). Apixaban was associated with lower or similar risk of bleeding across the range of kidney function (pooled HRs: 0.55 [0.43 to 0.69], 0.73 [0.52 to 1.02], 0.55 [0.31 to 0.97], and 0.68 [0.47 to 0.99], respectively). There was no significant heterogeneity across jurisdictions or eGFR categories.

Conclusions

In adults with AF and non–dialysis-dependent CKD, apixaban compared with warfarin was associated with lower or similar risk of ischemic and bleeding outcomes. Our results suggest that apixaban offers a favorable risk-benefit ratio in patients with AF independent of kidney function.

Introduction

Atrial fibrillation (AF) and CKD are increasingly common and coincident conditions.^1,2 Both conditions are major risk factors for cardiovascular disease including stroke. Patients with both CKD and AF are at increased risk of thromboembolic events and mortality compared with patients with either CKD or AF alone.³

Direct oral anticoagulants (DOACs) are recommended as first-line treatment in people with AF for stroke prevention. However, they remain suboptimally used in people with concurrent CKD, particularly those with CKD stages 4/5, attributable to uncertainties in the benefit-risk profile in people with CKD.⁴ This is because people with CKD are predisposed to a higher bleeding risk because of impaired platelet recruitment and activation,^5,6 which is further complicated by concerns of DOAC accumulation in renal impairment despite dose adjustment.^7,8

Data on the risks and benefits of apixaban across varying levels of kidney dysfunction are limited. DOAC trials of the broader AF population have been restricted to those with creatinine clearance >25 ml/min. A post hoc analysis of the Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation (ARISTOTLE) trial showed that compared with warfarin, apixaban reduced the risk of stroke, death, and major bleeding across individuals with an eGFR of >80, 50–80, and ≤50 ml/min.⁹ However, only 269 patients had creatinine clearance 25–30 ml/min, representing only 1.5% of the overall ARISTOTLE trial population.¹⁰ Although observational studies on the safety and effectiveness of apixaban have reported similar and consistent findings in those with advanced nondialysis CKD, these studies have been limited by their geographical coverage.^11–14

We therefore performed a multicenter, propensity score-matched cohort study of Australian and Canadian adults with AF to assess the effectiveness (all-cause death, ischemic stroke, or transient ischemic attack [TIA]) and safety (major bleeding) of apixaban compared with warfarin according to levels of kidney function.

Methods

Study Design and Data Sources

We conducted an international multicenter propensity score-matched cohort study using health care data in five jurisdictions across Australia and Canada. In Australia, we used data from the EXamining ouTcomEs in chroNic Disease in the 45 and Up (EXTEND45) Study,^15,16 a population-based cohort study built on the Sax Institute's 45 and Up Study, a cohort of 267,357 residents aged 45 years or older in the state of New South Wales, Australia, recruited between 2005 and 2009. The Medicare claims data and Pharmaceutical Benefits Scheme (PBS; recorded claims for subsidized pharmaceutical products in Australia) data were provided to the Sax Institute by Services Australia. In Canada, provincial data sources from British Columbia (BC) Medical Services Plan and Patient Records and Outcome Management Information System, Alberta (AB) Health, Manitoba (MB) Health, and Ontario (ON) ICES were used (see Supplemental Table 1 for details on data sources). In ON, these datasets were linked using unique encoded identifiers and analyzed at ICES.

Ethics Approval

In New South Wales, Australia, ethical approval for the EXTEND45 Study was obtained from the New South Wales Population and Health Services Research Ethics Committee (HREC/13/CIPHS/69). The 45 and Up Study received ethics approval from the University of NSW Human Research Ethics Committee. In the Canadian provinces of AB, BC, and MB, the study was approved by the Institutional Review Boards of the University of Calgary (REB18-0471_REN2), University of BC Providence Health Care Research Institute (H18-02319), and University of MB (HS22072), respectively. In ON, Canada, the use of administrative health data was authorized under section 45 of ON's Personal Health Information Protection Act, which does not require review by a Research Ethics Board. This study complies with the Declaration of Helsinki and was performed according to ethics committee approval.

Identification of Study Cohort

We included adults (aged 18 years or older) who received a new prescription of either apixaban or warfarin from January 1, 2013, and the end date of the available data in each jurisdiction (latest end date was December 31, 2018). Eligible patients included those with at least 1 outpatient serum creatinine measurement within 1 year before the cohort entry date and a diagnosis of AF or atrial flutter (International Classification of Diseases (ICD), Ninth Revision (ICD-9) code 427.31/2¹⁷; Tenth Revision (ICD-10) code I48; within 5 years before the date of the cohort entry). The cohort entry date was defined as the date of the new prescription for apixaban or warfarin. We excluded patients with a diagnosis of venous thromboembolism, had a prescription for DOAC or warfarin within 365 days before cohort entry, had a history of mitral or aortic valve disease (valvular AF) or valve surgery in the period 5 years before cohort entry, or had ESKD defined as receiving chronic dialysis or kidney transplantation. A new apixaban user identified within the study period was matched using propensity scores (the propensity to receive apixaban) to a new warfarin user receiving warfarin with one-to-one matching within each eGFR category. Exposure to apixaban or warfarin was treated as a time fixed variable throughout the study follow-up (intention-to-treat).

Assessment of Kidney Function

Eligible participants were those with ≥1 outpatient serum creatinine measurement within 1 year before the cohort entry date. eGFR was calculated using the 2009 CKD Epidemiology Collaboration equation. Participants were grouped into eGFR categories ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m².

Covariates

Demographic information including age, sex, and comorbidities were determined from health care data sources of the participating sites (the 45 and Up Study baseline questionnaire for EXTEND45 and the population registry for the Canadian sites). Validated ICD-9 and ICD-10 algorithms were used where available to define comorbidities^18,19 from hospitalization and medical services databases (and also the 45 and Up Study baseline questionnaire²⁰ for EXTEND45—see Supplemental Tables 1 and 2 for details on demographic data used).

These included history of cardiovascular disease, major bleeding requiring admission to hospital, peripheral vascular disease, cerebrovascular disease, chronic obstructive pulmonary disease, liver disease, and cancer.²¹ We established baseline use of prescription antiplatelet agent, nonsteroidal anti-inflammatory drug (NSAID), and proton pump inhibitor defined as ≥1 dispensed prescription within 90 days of cohort entry.

We assessed the risk of stroke and bleeding at the time of the cohort entry date. The CHA₂DS₂-VASc score was used to assess the risk of stroke, calculated on the basis of the history of congestive heart failure, hypertension, age (64–74 years and 75 years or older), vascular disease, diabetes, and history of ischemic stroke.²² The modified HAS-BLED score was calculated to determine the risk of bleeding.^23,24 The components of the modified HAS-BLED score included hypertension, abnormal renal or liver function, history of stroke, history of major bleeding, age 65 years or older, and concomitant use of antiplatelet or NSAIDs.

Outcomes

The effectiveness outcome was the composite of all-cause mortality or first hospitalization for ischemic stroke, TIA, within 1 year from the cohort entry date (see Supplemental Table 2 for the list of ICD-9 and ICD-10 codes). The safety outcome of major bleeding was defined as the composite of first hospitalization for intracranial, upper or lower gastrointestinal, or other bleeding within 1 year from the cohort entry date. The follow-up period for each patient commenced on cohort entry date and continued until the earliest occurrence of the outcomes of interest, death, end of the 1-year follow-up period, or study end (December 31, 2018).

Statistical Analysis

We constructed a multivariable logistic regression model within each eGFR category which included demographic information (age and sex), comorbidities (listed in Covariates), year of cohort entry and use of prescription antiplatelet agents, NSAIDs, and proton pump inhibitors to determine the conditional odds of being prescribed apixaban. We used one-to-one matching without replacement and a caliper width of 0.2 of the SD of the logit of the propensity score.²⁵ Matching was conducted within each eGFR category. The balance of covariates between apixaban and warfarin users was determined using standardized differences with a meaningful imbalance defined as >10%.²⁶

The study outcome event rates (expressed as per 100 person-years) were calculated using Poisson regression by oral anticoagulant use (apixaban or warfarin) and eGFR category. The rates of the composite effectiveness and safety outcomes were used to estimate the absolute risk difference of study outcomes between apixaban and warfarin users by eGFR category. Cox proportional hazards models (accounting for clustering within matched pairs using a robust variance estimator) were constructed to estimate the association between apixaban/warfarin use and outcomes within 1 year from the cohort entry date by eGFR category. Warfarin users were considered the reference category in the calculation of hazard ratios (HRs) and their corresponding 95% confidence intervals (CIs) for each study outcome within each eGFR category. Analyses (including propensity score matching) were performed independently within each jurisdiction according to a common analytical protocol using SAS, version 9.4 (SAS Institute Inc.). We considered a two-sided P value < 0.05 as statistically significant.

Meta-Analysis

Summary estimates of HRs and 95% CIs for each eGFR category were obtained from each jurisdiction and pooled using random effects meta-analysis (DerSimonian and Laird). The percentage of variability across sites attributable to heterogeneity beyond chance was estimated using the I² statistic where values of ≤25%, >25%–75%, and >75% correspond to low, moderate, and high levels of heterogeneity, respectively.²⁷ Meta-regression was performed to test for modification of the association between apixaban use, compared with warfarin use, and study outcomes by eGFR category. Meta-analysis was performed with Stata software, version 16.1 (Stata Corp., College Station, TX).

Sensitivity Analysis

We conducted a series of prespecified sensitivity analyses. We repeated all analyses after adjusting for baseline eGFR and variables observed to be meaningfully imbalanced when covariate balance was assessed within each eGFR category, using an as-treated approach (patients were censored at the time of oral anticoagulant treatment switching or discontinuation) and using both hospital and emergency department data in jurisdictions where emergency data were available (Canadian provinces of AB, BC, and ON). Finally, we assessed the relationship between apixaban/warfarin use and the outcome of myocardial infarction (both as part of the composite outcome and individually) by eGFR category.

Results

Patient Characteristics

We identified 106,102 adults with a diagnosis of AF with a new prescription for apixaban (n=69,049) or warfarin (n=37,053) who had at least one outpatient baseline serum creatinine measurement (Figure 1). A total of 38,598 patients (19,299 apixaban users and 19,299 warfarin users) were included in the 1:1 propensity score matched analysis (EXTEND45=332, AB=5,066, BC=130, MB=1,874, and ON=31,196; Supplemental Figure 1). Propensity score matching achieved good overall balance in covariates between the apixaban and warfarin groups (Table 1 and Supplemental Table 3). The mean age of the cohort was 75 years, and 45% were female. For baseline kidney function, 57.3% (n=22,096) had eGFR ≥60 ml/min per 1.73 m², 19.2% (n=13,212) had eGFR 30–59 ml/min per 1.73 m², and 8.5% (n=3290) had eGFR <30 ml/min per 1.73 m². Overall, 36% (n=13,944) of participants had diabetes mellitus, and 76% had hypertension (n=29,348). CHA₂DS₂-VASc scores were evenly distributed among groups with 93.4% (n=18,032) and 94.5% (n=18,211) with a score ≥2 in the apixaban and warfarin groups, respectively. Modified HAS-BLED scores were also similar in both groups with 62% of participants with scores 0–2 and 37% with a score ≥3.

Figure 1.

Identification of study cohort. DOAC, direct oral anticoagulant.

Table 1.

Baseline characteristics of patients with atrial fibrillation, by apixaban and warfarin use, before and after propensity score matching

Covariate	Unmatched Cohort		Standardized Differenceb	Matched Cohort		Standardized Differenceb
	Apixaban	Warfarin		Apixaban	Warfarin
No. of patients	69,049	37,053		19,299	19,299
Age, yr, mean (SD)a	77.6 (16.9)	75.3 (10.5)	0.208	75.5 (10.2)	75.3 (9.6)	0.017
Female	33,574 (48.6%)	16,672 (45.0%)	0.080	8728 (45.2%)	8694 (45.0%)	0.013
Year of cohort entry, n (%)
2013–2014	7641 (11.1%)	16,123 (43.5%)	0.832	5641 (29.2%)	5818 (30.1%)	0.100
2015–2016	22,499 (32.6%)	10,164 (27.4%)	0.238	8674 (44.9%)	8516 (44.1%)	0.106
2017–2019	38,908 (56.3%)	5179 (14.0%)	0.682	4983 (25.8%)	4964 (25.7%)	0.015
eGFR, ml/min per 1.73 m 2 , mean (SD)a	65.1 (18.7)	61.8 (22.7)	0.198	63.3 (21.6)	63.0 (21.2)	0.021
≥60	41,805 (60.5%)	20,483 (55.2%)		11,048 (57.2%)	11,048 (57.2%)
45–59	16,166 (23.4%)	7084 (19.1%)		3702 (19.2%)	3702 (19.2%)
30–44	9005 (13.0%)	5339 (14.4%)		2904 (15.0%)	2904 (15.0%)
<30	2073 (3.0%)	4147 (11.2%)		1645 (8.5%)	1645 (8.5%)
Diabetes	23,595 (34.1%)	11,698 (31.6%)	0.140	6930 (36.0%)	7014 (36.3%)	0.013
Hypertension	53,298 (77.2%)	27,580 (74.4%)	0.060	14,675 (76.0%)	14,673 (76.0%)	0.003
MI	5587 (8.1%)	5737 (15.5%)	0.148	2253 (11.7%)	2273 (11.8%)	0.005
Congestive heart failure	19,246 (27.9%)	13,627 (36.8%)	0.202	6548 (34.0%)	6537 (33.9%)	0.001
Cerebrovascular disease	14,171 (20.5%)	7931 (21.4%)	0.034	4102 (21.3%)	4041 (20.9%)	0.005
Peripheral vascular disease	1897 (2.7%)	2562 (6.9%)	0.080	857 (4.4%)	867 (4.5%)	0.003
Chronic obstructive pulmonary disease	6336 (9.2%)	7241 (19.5%)	0.058	2459 (12.7%)	2476 (12.8%)	0.002
Liver disease	3.053 (4.4%)	1662 (4.5%)	0.039	893 (4.6%)	918 (4.8%)	0.010
Cancer	22,816 (33.0%)	10,400 (28.1%)	0.026	5721 (29.6%)	5826 (30.2%)	0.010
Prior major bleeding requiring hospital admissionb	1167 (1.7%)	657 (1.8%)	0.020	333 (1.7%)	343 (1.8%)	0.025
Prescription drug use
Antiplatelet agents	6084 (8.8%)	3412 (9.2%)	0.018	1762 (9.1%)	1678 (8.7%)	0.077
Nonsteroidal anti-inflammatory drugs	14,339 (20.8%)	5365 (14.5%)	0.025	3380 (17.5%)	3335 (17.3%)	0.034
Proton pump inhibitors	19,280 (27.9%)	11,068 (29.9%)	0.046	5233 (27.1%)	5290 (27.4%)	0.078
CHA 2 DS 2 -VASc score
0–1	3059 (4.4%)	2463 (6.6%)	0.039	1267 (6.6%)	1088 (5.5%)	0.029
≥2	65,990 (95.6%)	34,590 (93.4%)	0.032	18,032 (93.4%)	18,211 (94.5%)	0.015
Modified HAS-BLED score
0–2	42,694 (61.8%)	24,023 (64.8%)	0.071	12,014 (62.3%)	12,141 (62.9%)	0.025
≥3	26,372 (38.2%)	13,030 (35.2%)	0.041	7285 (37.7%)	7158 (37.1%)	0.017

Data are presented as the number of patients and the corresponding percentage of the cohort unless otherwise indicated. Cells of tables with patient counts <5 were suppressed by participating centers due to privacy restrictions. CHA₂DS₂-VASc and modified HAS-BLED scores estimate the risk of stroke and bleeding, respectively, in patients with atrial fibrillation. DOAC, direct oral anticoagulant; MI, myocardial infarction.

^aSample size weighted mean or standardized difference.

^bMajor bleeding defined as first hospitalization for intracranial, gastrointestinal, or other bleeding.

Incidence of Outcomes

Over the 1-year follow-up period, the composite effectiveness outcome occurred in 4130 patients (10.7%), while the safety outcome occurred in 717 patients (1.8%) over a total of 73,079 person years of follow-up. When assessed as individual outcomes, there were 3767 (9.8%) deaths, 433 (1.1%) ischemic stroke or TIA events, 452 (1.2%) gastrointestinal bleeding events, 105 (0.3%) intracranial bleeds, and 160 (0.4%) other bleeding events recorded (Supplemental Table 4).

HRs for the Ischemic Outcome

Apixaban initiation was associated with lower or similar risk of death/ischemic events compared with warfarin initiation in all eGFR categories (pooled HRs [95% CI]: 0.78 [0.64 to 0.94]), 0.77 [0.62 to 0.97], and 0.82 [0.68 to 0.98] and 0.99 [0.68 to 1.45] for eGFR ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m², respectively; Figure 2). Heterogeneity across jurisdictions within each eGFR category was low to moderate (I² values between 10.5% and 32.4%). Meta-regression showed that the comparative risk of death/ischemic events when comparing apixaban initiation with warfarin initiation was consistent across the eGFR categories examined (P value = 0.32). Similar results were observed when individual components of the effectiveness outcome were assessed separately, although statistical significance was not reached for the outcome of ischemic stroke or TIA (Supplemental Table 5).

Figure 2.

Hazard ratios (95% CIs) for the effectiveness outcome by eGFR category. CI, confidence interval; EXTEND45, EXamining ouTcomEs in chroNic Disease in the 45 and Up; PROMIS, Patient Records and Outcome Management Information System; S, number of outcome events <5 and cells were suppressed to meet privacy restrictions.

HRs for the Bleeding Outcome

Apixaban initiation was associated with lower or similar hazard for bleeding across all eGFR groups (pooled HRs [95% CI]: 0.55 [0.43 to 0.69], 0.73 [0.52 to 1.02], 0.55 [0.31 to 0.97], and 0.68 [0.47 to 0.99] for eGFR ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m², respectively; Figure 3). We observed low levels of heterogeneity across jurisdictions within each eGFR category. Meta-regression showed that the comparative safety (bleeding) of apixaban initiation compared with warfarin initiation was consistent across the eGFR categories examined (P value = 0.48). Overall, similar results were observed when individual components of the bleeding outcome were assessed separately, although statistical significance was not reached for some bleeding events (Supplemental Table 5).

Figure 3.

Hazard ratios (95% CIs) for the major bleeding outcome by eGFR category.

Absolute Risk Difference

Apixaban users experienced similar or fewer effectiveness outcome events compared with warfarin users across nearly all eGFR categories ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m² (pooled absolute risk difference: −2.0% [95% CI, −3.7% to −0.4%], −2.8% [−9.2% to 3.7%], −1.7% [−7.9% to 4.6%], and 4.7% [−8.01% to 13.4%], respectively). There were similar or fewer major bleeding events in apixaban users compared with warfarin users (pooled absolute risk difference: −0.9% [95% CI, −1.6% to −0.4%], −0.6% [−1.3% to 0.1%], −1.9% [−3.5% to −0.3%], and −1.4% [−2.7% to −0.0%] across eGFR categories ≥60, 45–59, 30–44, and <30 ml/min per 1.73 m², respectively).

Sensitivity Analysis

The results were consistent across all sensitivity analyses (Figure 4). This included (1) as-treated analysis, (2) models adjusted for covariates identified as imbalanced when covariate balance was evaluated according to eGFR category (Supplemental Table 6), (3) analyses using both hospitalization and emergency department data to assess the occurrence of study outcome events in jurisdictions where these data were available, and (4) an assessment of the effectiveness of apixaban use compared with warfarin use on the outcome of myocardial infarction by eGFR category.

Figure 4.

Sensitivity analyses. (A) Sensitivity analyses assessing the comparative effectiveness of apixaban and warfarin according to eGFR category. (B) Sensitivity analyses assessing the comparative safety (major bleeding) of apixaban and warfarin according to eGFR category.

Discussion

In this multicenter observational study of over 38,000 people with AF, apixaban use, compared with warfarin use, was associated with greater or similar effectiveness and lower or similar bleeding risk across all levels of kidney function. We observed consistent findings across jurisdictions and a series of sensitivity analyses which demonstrate robustness of our study findings. Overall, our results indicate a favorable benefit-risk profile of apixaban in people with CKD, including those with advanced CKD, and support its ongoing utility in this population.

The use of apixaban (as an alternative to vitamin K antagonists) for stroke prevention in patients with AF and normal kidney function has been well supported by robust randomized controlled trial data.^28,29 However, clinical evidence to guide the use of apixaban in people with AF and CKD, with heightened thromboembolic risk, has been limited. Landmark trials assessing the efficacy and safety of apixaban in the broader AF population have shown similarly favorable treatment effects across mild-to-moderate stages of CKD. For example, subgroup analyses of the A Phase III Study of Apixaban in Patients With Atrial Fibrillation study³⁰ (1697 patients with eGFR 30–59 ml/min per 1.73 m²) showed that apixaban when compared with aspirin showed a reduced risk of stroke and systemic emboli (HR, 0.32 [95% CI, 0.18 to 0.55]) and no significant difference in major hemorrhage. This was mirrored in a subgroup analysis of the ARISTOTLE study⁹ (3017 patients with eGFR 25–50 ml/min per 1.73 m²) that showed apixaban had a reduced risk of stroke or systemic emboli (HR, 0.61 [95% CI, 0.39 to 0.94]) and reduction in major bleeding (HR, 0.48 [95% CI, 0.37 to 0.64]) when compared with warfarin. Our findings on those with mild-to-moderate CKD (n=7404 with eGFR 45–59 ml/min per 1.73 m² and n=5808 with eGFR 30–44 ml/min per 1.73 m²) are consistent with both subgroup analyses of large randomized controlled trials and support the ongoing use of apixaban as an alternative to warfarin in this patient group.

Pharmacokinetic data show that apixaban is less dependent on renal excretion compared with rivaroxaban and dabigatran.⁷ Accordingly, apixaban is the only DOAC approved for use in patients with creatinine clearance of <15 ml/min,³¹ and much of the emerging evidence base on the safety and effectiveness of apixaban in people with advanced CKD has focused on patients requiring long-term dialysis.^32,33 However, such data have limited generalizability in the broader group of patients with advanced CKD not receiving chronic dialysis—data in this patient population have been particularly lacking. To this end, our study adds important new clinical information about the balance between the benefits and harms associated with apixaban use across people with varying levels of kidney function, including those with advanced CKD (<30 ml/min per 1.73 m²; n=3290) not receiving chronic dialysis. Our findings showed that apixaban was associated with a lower risk of bleeding compared with warfarin without an increased risk of stroke, TIA, or death in those with eGFR <30 ml/min per 1.73 m². These results are consistent with those reported by Stanifer et al.¹⁰ in a post hoc analysis of the ARISTOTLE study including 269 patients with creatinine clearance 25–30 ml/min, which showed that apixaban use was associated with less major bleeding (HR, 0.34 [95% CI, 0.14 to 0.80]) and no clear difference in stroke or systemic embolism (HR, 0.55 [95% CI, 0.2 to 1.5]). Furthermore, our study extends these findings to patients with eGFR <25 ml/min per 1.73 m² not receiving dialysis, showing a favourable benefit-risk profile of apixaban in this patient group.

Randomized trials of apixaban in the broader AF patient population have specifically excluded people with advanced CKD (creatinine clearance <25 ml/min). However, emerging observational data on the safety and effectiveness of apixaban in patients with advanced CKD not receiving dialysis indicate a similarly favorable profile associated with apixaban. A large cohort study of two national US claims databases found a higher risk of major bleeding in warfarin compared with apixaban (HR, 1.85 [95% CI, 1.59 to 2.15]) with no differences for ischemic stroke in patients with AF and CKD stages 4/5.¹¹ However, there were relatively few ischemic stroke events, and inclusion of patients with CKD stages 4/5 was reliant on diagnosis codes subject to misclassification. In another US-based study, apixaban versus warfarin was associated with a lower risk of major bleeding (HR, 0.53 [95% CI, 0.39 to 0.70]), and similar risks for stroke or systemic embolism (HR, 0.80 [95% CI, 0.59 to 1.09]) and death (HR, 1.03 [95% CI, 0.82 to 1.29]) in over 2700 matched apixaban-warfarin pairs with advanced CKD.¹³ Taken together, the broad geographical coverage of accumulating data including those from this study supports the utility of apixaban as an alternative to warfarin in people with advanced CKD and provides reassurance on the balance between benefits and harms associated with apixaban in this high-risk patient population.

This study represents one of the largest assessments on the comparative effectiveness and safety of apixaban and warfarin in patients with AF across the spectrum of CKD. The major strength of this study is that it used a multicenter study design (a large binational cohort of over 38,000 patients from across five jurisdictions) with broad inclusion criteria to identify patients from routine clinical settings which substantially increases the generalizability of our findings, including to those who have been excluded in randomized trials of apixaban in the broader AF population. Our findings were also consistent across all levels of kidney function, jurisdictions, and sensitivity analyses for the composite outcomes and their individual components, supporting the robustness of the overall findings. The incident user design of this study also mitigates the risk of biases that can occur³⁴ when comparisons include prevalent users including an underestimation of safety outcomes associated with DOAC use compared with warfarin use. However, our study has several limitations. First, information on international normalized ratio was not available in all jurisdictions, and therefore, we were unable to determine the time in therapeutic range among warfarin users, which may have affected the estimation of study outcome event rates. Second, the propensity score matching process created a cohort with comparable characteristics between the apixaban and warfarin groups, minimizing treatment selection biases. However, the selection of matched pairs on the basis of incident users of either apixaban or warfarin may have led to the creation of a cohort deemed to be at lower risk of bleeding, and for whom the perceived benefits of anticoagulation outweighed the perceived risks. Although 3290 patients with eGFR <30 ml/min per 1.73 m² were included, the proportion of patients with poor kidney function (eGFR <15 ml/min per 1.73 m²) was relatively low which limited our ability to further assess the safety and effectiveness of apixaban compared with warfarin in this patient group. Third, the appropriateness of apixaban dosing could not be ascertained which may have influenced both ischemic stroke and bleeding events. In addition, our study lacked data on hemorrhagic stroke events. Previous systematic reviews identified a significantly lower risk of hemorrhagic stroke with DOACs compared with warfarin.³⁵ This may have underestimated safety of anticoagulation in the apixaban group. Study outcomes were recorded for 12 months after initiation of warfarin or apixaban, and thus, the long-term risk-benefit ratio for apixaban therapy in patients with AF and CKD is unclear. Finally, although we sought to account for a range of important patient factors that might influence the receipt of apixaban compared with warfarin for the management of AF (e.g., comorbid conditions and year of cohort entry), we were unable to further adjust for other potentially important patient-specific and clinician-specific factors that influence oral anticoagulant choice such as the availability of a reversal agent for apixaban, and variations in prescribing patterns across regions (e.g., provider preference and medication subsidy), and thus, the potential for residual confounding remains.

In this large multicenter, propensity score-matched cohort study of 38,958 adults with AF, the initiation of apixaban therapy was associated with greater or similar effectiveness and lower or similar risk of bleeding when compared with warfarin. Our results support the ongoing utility of apixaban therapy as an alternative to warfarin in patients with AF and CKD, including those with eGFR <30 ml/min per 1.73 m² not receiving long-term dialysis.

Supplementary Material

kidney360-6-01510-s001.pdf ^{(1.6MB, pdf)}

kidney360-6-01510-s002.pdf ^{(369.2KB, pdf)}

Disclosures

Disclosure forms, as provided by each author, are available with the online version of the article at http://links.lww.com/KN9/B26.

Funding

M. Jun: National Health and Medical Research Council (1148060).

Author Contributions

Conceptualization: Martin Gallagher, Amit X. Garg, Brenda Hemmelgarn, Matthew T. James, Min Jun, Adeera Levin, Anish Scaria, Manish M. Sood, Navdeep Tangri.

Data curation: Peter Birks, Sarah E. Bota, Ognjenka Djurdjev, Amit X. Garg, Brenda Hemmelgarn, Carinna Hockham, Matthew T. James, Meg J. Jardine, Min Jun, Adeera Levin, Manish M. Sood, Navdeep Tangri, Reid Whitlock.

Formal analysis: Anna Campain, Min Jun, Eric McArthur, Anish Scaria, Selena Shao, Zhi Tan, Reid Whitlock.

Funding acquisition: Martin Gallagher, Amit X. Garg, Brenda Hemmelgarn, Matthew T. James, Meg J. Jardine, Min Jun, Adeera Levin, Navdeep Tangri.

Investigation: Jason Andrade, Sunil V. Badve, Peter Birks, Sarah E. Bota, Anna Campain, Ognjenka Djurdjev, Martin Gallagher, Amit X. Garg, Jeffrey T. Ha, Ziv Harel, Brenda Hemmelgarn, Carinna Hockham, Matthew T. James, Meg J. Jardine, Min Jun, Dickson Lam, Adeera Levin, Eric McArthur, Pietro Ravani, Anish Scaria, Selena Shao, Manish M. Sood, Zhi Tan, Navdeep Tangri, Reid Whitlock.

Methodology: Jason Andrade, Sunil V. Badve, Peter Birks, Sarah E. Bota, Anna Campain, Ognjenka Djurdjev, Martin Gallagher, Amit X. Garg, Jeffrey T. Ha, Ziv Harel, Brenda Hemmelgarn, Carinna Hockham, Matthew T. James, Meg J. Jardine, Min Jun, Dickson Lam, Adeera Levin, Eric McArthur, Pietro Ravani, Anish Scaria, Selena Shao, Manish M. Sood, Zhi Tan, Navdeep Tangri, Reid Whitlock.

Project administration: Jeffrey T. Ha, Min Jun.

Supervision: Min Jun.

Writing – original draft: Dickson Lam.

Writing – review & editing: Jason Andrade, Sunil V. Badve, Peter Birks, Sarah E. Bota, Anna Campain, Ognjenka Djurdjev, Martin Gallagher, Amit X. Garg, Jeffrey T. Ha, Ziv Harel, Brenda Hemmelgarn, Carinna Hockham, Matthew T. James, Meg J. Jardine, Min Jun, Dickson Lam, Adeera Levin, Eric McArthur, Pietro Ravani, Anish Scaria, Selena Shao, Manish M. Sood, Zhi Tan, Navdeep Tangri, Reid Whitlock.

Data Sharing Statement

Partial restrictions to the data and/or materials apply. The study was completed using administrative and health care data from five jurisdictions across Australia (in the state of New South Wales) and Canada (AB, BC, MB, and ON). The study jurisdictions do not permit the release of individual participant data or aggregate summary level data. The data underlying this article are available in the article and in its online supplementary material.

Supplemental Material

This article contains the following supplemental material online at http://links.lww.com/KN9/B27.

Supplemental Table 1. Data sources used in the study across the participating centers.

Supplemental Table 2. ICD-9-CM and ICD-10-CA diagnosis codes for the identification of AF, covariates, and study outcomes.

Supplemental Table 3. Baseline characteristics of patients with AF by eGFR category, before and after propensity score matching.

Supplemental Table 4. Incidence rates (95% CIs) of the composite effectiveness and major bleeding outcomes by DOAC and warfarin use and eGFR category.

Supplemental Table 5. Pooled HRs (95% CIs) for the individual components of the composite and major bleeding outcomes by eGFR category.

Supplemental Table 6. Sensitivity analyses—pooled HRs (95% CIs) for the individual components of the composite effectiveness and major bleeding outcomes by eGFR category adjusted for imbalanced covariates; within each eGFR category, warfarin initiation was considered as the reference category in estimating the HRs and their 95% CIs.

Supplemental Figure 1. Identification of propensity score matched new users of apixaban or warfarin in the five jurisdictions.