Apixaban vs rivaroxaban in patients with atrial fibrillation at high or low bleeding risk: A population-based cohort study

Abstract

BACKGROUND

Despite many atrial fibrillation (AF) patients being at risk of bleeding, very limited data are available on bleeding rates of different direct oral anticoagulants based on the spectrum of bleeding risk.

OBJECTIVE

We aimed to compare the risk of major bleeding and thromboembolic events with apixaban vs rivaroxaban for AF patients stratified by bleeding risk.

METHODS

We conducted a population-based, retrospective cohort study of all adult patients (66 years or older) with AF in Ontario, Canada, who were treated with apixaban or rivaroxaban between April 1, 2011, and March 31, 2020. Bleeding risk was estimated by the HAS-BLED score, with high bleeding risk defined as a score of ≥3. The primary safety outcome was major bleeding, and the primary efficacy outcome was thromboembolic events. Comparisons were adjusted for baseline comorbidities by inverse probability of treatment weighting.

RESULTS

This study included 18,156 AF patients with high bleeding risk and 55,186 AF patients with low bleeding risk. Apixaban use was more common in patients with high bleeding risk; 63% of high-risk patients used apixaban compared with 56% of low-risk patients. Apixaban users had lower rates of major bleeding in high-risk patients (2.9% vs 4.2% per year; hazard ratio [HR], 0.69; 95% CI, 0.58–0.81) and in low-risk patients (1.8% vs 2.9% per year; HR, 0.63; 95% CI, 0.56–0.70) compared with rivaroxaban. There were no differences in rates of thromboembolic events, 3.1% vs 3.0% per year (HR, 1.02; 95% CI, 0.86–1.22) in high-risk patients and 1.9% vs 1.9% per year (HR, 1.00; 95% CI, 0.89–1.14) in low-risk patients.

CONCLUSION

In older AF patients with high or low bleeding risk, treatment with apixaban was associated with lower rates of major bleeding with no difference in risk for thromboembolic events compared with rivaroxaban.

Introduction

Risk factors for stroke and systemic embolism in atrial fibrillation (AF) patients are also risk factors for bleeding, and thus many AF patients with high stroke risk also have high bleeding risk.^1,2 Direct oral anticoagulants (DOACs) have supplanted oral vitamin K antagonists as the preferred oral anticoagulant for stroke prevention in most patients with nonvalvular AF, and apixaban and rivaroxaban are most commonly used.^3–6 DOAC-associated bleeding continues to be one of the most common adverse drug reactions in older patients, accounting for 6.6% of adverse drug reaction–related hospitalizations in those older than 65 years.⁷

However, there are no clinical trials directly comparing their efficacy and safety for stroke and bleeding. Whereas several observational studies have compared the efficacy and safety of apixaban vs rivaroxaban in the general AF patients,^8–13 no studies have reported on outcomes of these 2 drugs in patients with high or low bleeding risk.⁹ There are limited comparative data between apixaban and rivaroxaban to assist in the appropriate choice between these drugs for patients with AF and high risk of gastrointestinal (GI) bleeding.¹⁴

We aimed to compare the rates of major bleeding and thromboembolic events in older AF patients at high or low bleeding risk who were prescribed either apixaban or rivaroxaban, including subgroup analyses of site-specific bleeding (such as subarachnoid or intracranial, GI, and genitourinary bleeding) and standard or reduced dose regimens.

Methods

Data sources

We conducted a population-based retrospective cohort study by linking multiple administrative databases in Ontario, Canada, including the following:

• the Ontario Drug Benefit Program, which records prescription of drugs dispensed to Ontario residents aged ≥65 years;

• hospitalization records from the Canadian Institute for Health Information Discharge Abstract Database, which contain a detailed record of all hospital admissions in Ontario, including diagnostic and procedural information;

• t-he Ontario Health Insurance Plan, which provides information on physician claims for inpatient and outpatient services;

• the Registered Persons Database, which captures demographic and date of death information; and

• the National Ambulatory Care Reporting System, which reports information from emergency department visits.¹³

These data sets were linked by unique encoded identifiers and analyzed at ICES (formerly the Institute for Clinical Evaluative Sciences). The use of the data in this project is authorized under Section 45 of Ontario’s Personal Health Information Protection Act and does not require review by a research ethics board.

Cohort

The initial cohort included all AF patients, aged ≥66 years, with at least 1 DOAC prescription ≥28 days between April 1, 2011, and March 31, 2020. AF patients were identified by a previously validated algorithm.^13,15 New DOAC users were identified by the first DOAC prescription in the Ontario Drug Benefit Program database. We excluded patients with a warfarin prescription within 1 year before DOAC use, dementia, severe liver disease, metastatic cancer, deep venous thrombosis or pulmonary embolism, hip or knee replacement surgery, renal dialysis, and high hospital frailty risk.¹⁶ We further excluded AF diagnoses that occurred in the hospital because they might have indicated secondary AF from other diagnoses. Finally, patients receiving dabigatran or edoxaban were excluded (Figure 1).

Figure 1.

Cohort creation. AF = atrial fibrillation; DOAC = direct oral anticoagulant.

Definition of DOAC use

Apixaban and rivaroxaban users were identified as those who had at least a single apixaban or rivaroxaban prescription with supply of ≥28 days with 1-week grace period or consecutive supply of ≥28 days for either a standard dose (5 mg twice daily for apixaban and 20 mg once daily for rivaroxaban) or reduced dose (2.5 mg twice daily for apixaban and 15 mg once daily for rivaroxaban).¹³ Other doses that are not for standard use by AF patients were excluded.

Definition of bleeding risk

The HAS-BLED score has been shown to be the optimal choice to assess major bleeding risk in clinical practice for patients with AF taking a DOAC.¹⁷ Patients with high bleeding risk were identified with a score of ≥3 on the HAS-BLED (hypertension, abnormal renal or liver function, previous stroke, bleeding history or predisposition, labile international normalized ratio [applies only to a vitamin K antagonist user, otherwise not applicable for a non–vitamin K antagonist user], elderly [age ≥65 years], and concomitant drugs or alcohol excess), and those with low bleeding risk were identified with a score of <3 on the HAS-BLED.¹⁷

Outcomes

The primary safety outcome was major bleeding, and the primary efficacy outcome was thromboembolic events (stroke, transient ischemic attack, or systemic embolism). Our secondary outcomes were any bleeding; thromboembolic events excluding transient ischemic attack; death; myocardial infarction; a composite of major bleeding, thromboembolic events, or myocardial infarction; and finally, a composite of any bleeding, thromboembolic events, or myocardial infarction.^12,13,18

Major bleeding was defined as a hospitalization with an International Classification of Diseases, Tenth Revision, Canada code that included a primary diagnosis of subarachnoid, intracranial, or GI bleeding. This definition of major bleeding has been validated with a sensitivity of 94% and a positive predictive value of 87%.⁷ Cases of traumatic hemorrhage were not included. Any bleeding included major bleeding and nonmajor bleeding, including urogenital or other bleeding locations (Supplemental Table 1). Outcomes were captured at 90 days and 365 days from the first prescription dispensing date of apixaban or rivaroxaban.

Statistical analysis

The index date was defined as the date of the first DOAC prescription dispensed. We controlled for confounding and selection bias by propensity scores based on logistic regression models and inverse probability of treatment weighting (IPTW) to create a weighted cohort, with average treatment effect weights derived from the propensity score.¹⁹ The propensity score was created with patient demographic characteristics: age, sex, and geographic factors (rural vs urban residence); and cardiovascular risk factors and comorbidities, including hypertension, diabetes, stroke, transient ischemic attack, heart failure, myocardial infarction, percutaneous coronary intervention, coronary artery bypass surgery, chronic obstructive pulmonary disease, chronic kidney disease, liver disease (mild or moderate), cognitive impairment/dementia, peripheral vascular disease, rheumatic disease, any cancer as a primary diagnosis, anemia, and peptic ulcer disease. Baseline medications were also included, such as angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, beta blockers, calcium channel blockers, digoxin, clopidogrel, insulin, oral hypoglycemic agents, lipid-lowering agents, nonsteroidal anti-inflammatory drugs, and proton pump inhibitors. The subanalyses for site-specific bleeding and for standard and reduced dose regimens were performed by creating separate weighted cohorts, using IPTW method, and adjusted hazard ratios (HRs) were calculated.

The propensity score distributions in the apixaban and rivaroxaban groups displayed good overlap. Standardized differences describe differences between group means or proportions relative to the pooled standard deviation.²⁰ The magnitude of covariate imbalance was assessed with the standardized difference; and a difference was considered potentially meaningful if it was 0.10 or greater.

The adjusted HRs of the outcomes were calculated using IPTW cause-specific hazards models,²¹ which estimate the effect of covariates on the cause-specific hazard function. Death was considered a competing risk for the nonfatal outcomes. Cox regression was used for end points that included death. A robust variance estimator was used to account for the within-person homogeneity in outcomes induced by weighting.²² Weighted cumulative incidence curves were used to calculate percentages. Statistical significance was defined as P < .05. All analyses were conducted with SAS version 9.4 software (SAS Institute, Cary, NC).

Results

Study cohort

A total of 195,624 AF patients with at least 1 DOAC prescription dispensed between April 1, 2011, and March 31, 2020, were identified. From this initial cohort, we excluded 51,483 patients who had warfarin prescription within 1 year before the index DOAC prescription and 26,280 who had significant comorbidities as detailed in the exclusion criteria. We then excluded patients with dabigatran and edoxaban use and those with in-hospital diagnosis of AF (Figure 1). This study included 18,156 AF patients (25%) with high bleeding risk (HAS-BLED score ≥3) and 55,186 AF patients (75%) with low bleeding risk (HAS-BLED score <3).

High bleeding risk cohort

Baseline characteristics in high-risk patients

Before IPTW, there were 11,461 patients in the apixaban group and 6695 in the rivaroxaban group in the high bleeding risk cohort. After IPTW using propensity scores, patients in the apixaban and rivaroxaban groups were well balanced for baseline values of demographics, comorbidities, and medications with standardized differences consistently <0.1. Both groups had a similar mean age of 78.5 years and a similar proportion of female patients, 48% in the apixaban group vs 49% in the rivaroxaban group (Table 1).

Table 1.

Baseline characteristics comparing apixaban vs rivaroxaban in AF patients with high bleeding risk

	Unweighted			Weighted
Characteristics	Apixaban	Rivaroxaban	SD	Apixaban	Rivaroxaban	SD
No. of participants	11,461	6695
Age at index, y, mean	79.2	77.5	0.2	78.5	78.5	0
Women	49.2	47.3	0.04	47.9	49.3	0.03
Neighborhood income quintile
Quintile 1 (lowest)	21.9	20.6	0.03	21.4	21.3	0.003
Quintile 5 (highest)	17.3	19.1	0.05	17.7	18.3	0.01
Rural residence	11.3	13.6	0.07	12.1	12.6	0.02
Charlson score, mean	1.6	1.3	0.2	1.5	1.5	0.04
Cardiovascular risk factorsa
Diabetes	42.6	40.4	0.04	41.5	42.4	0.02
Hypertension	99.1	99.3	0.02	99.1	99.3	0.02
Heart failure	37.1	31.2	0.1	34.8	35.4	0.01
Stroke	16.7	12.9	0.1	15.0	15.7	0.02
Transient ischemic attack	2.4	2.0	0.03	2.3	2.3	0.001
Myocardial infarction	12.4	12.0	0.01	11.9	12.9	0.03
Chronic obstructive pulmonary disease	32.1	31.2	0.02	31.6	32.1	0.009
Peripheral vascular disease	19.6	17.0	0.07	18.3	19.5	0.03
Percutaneous coronary intervention	6.9	7.4	0.02	6.9	7.3	0.01
Coronary artery bypass graft	2.5	2.6	0.006	2.6	2.8	0.01
Major bleeding	12.0	10.2	0.06	11.6	10.9	0.02
HAS-BLED bleeding risk score, mean	3.2	3.1	0.09	3.2	3.2	0.006
HAS-BLED bleeding risk score
3	84.1	87.5	0.1	85.4	85.4	0.002
4	14.5	11.3	0.09	13.0	13.0	0.01
5 or 6	1.4	1.2	0.02	1.6	1.6	0.03
CHA2DS2-VASc score, mean	4.7	4.4	0.2	4.6	4.6	0.04
Alcohol use Prior medical conditionsb	0.9	0.9	0.009	0.9	0.9	0.007
Chronic kidney disease within 1 year	11.4	8.3	0.1	10.1	10.6	0.02
Chronic kidney disease	33.4	27.3	0.1	32.0	29.7	0.05
Liver disease	1.5	1.6	0.01	1.5	1.5	0.005
Anemia	11.0	8.3	0.09	10.0	10.1	0.003
Cancer, primary	7.1	6.9	0.009	6.9	7.5	0.03
Cancer, metastatic	0.6	0.5	0.002	0.6	0.5	0.02
Rheumatic disease	1.0	1.3	0.03	0.9	1.2	0.03
Peptic ulcer disease	2.0	1.5	0.04	1.8	1.9	0.005
Medication usec
Antiplatelets, anticoagulants,	54.0	61.6	0.2	55.8	58.2	0.05
nonsteroidal anti-inflammatory   within 100 days before index date
ACE inhibitors	30.1	31.6	0.03	30.4	30.9	0.01
Angiotensin II receptor blockers	25.8	25.9	0.001	26.2	25.6	0.01
Beta blockers	49.9	48.8	0.02	49.3	49.8	0.01
Calcium channel blockers	39.7	37.9	0.04	39.1	39.1	0
Clopidogrel	23.8	24.1	0.006	23.7	24.3	0.01
Digoxin	4.9	5.3	0.02	4.9	5.3	0.02
Insulin	7.2	6.5	0.03	6.7	7.3	0.02
Lipid-lowering agents	61.3	60.2	0.02	61.0	61.1	0.002
Oral antiglycemics	22.1	21.2	0.02	21.4	22.5	0.03
Proton pump inhibitors	41.0	40.2	0.02	40.4	41.4	0.02

Values are reported as percentage unless otherwise indicated.

ACE = angiotensin-converting enzyme; AF = atrial fibrillation.

^aLookback period for cardiovascular risk factors was 5 years unless otherwise stated.

^bLookback period for prior medical conditions was 5 years unless otherwise stated.

^cLookback period for medication use was 90 days before index date unless otherwise stated.

Outcomes associated with apixaban vs rivaroxaban in high-risk IPTW cohort

Within 90 days of the first prescription, the apixaban group had lower rates of both major bleeding (1.1% vs 1.6%; HR, 0.66; 95% CI, 0.51–0.86) and any bleeding (4.0% vs 5.2%; HR, 0.78; 95% CI, 0.67–0.90). There was no difference in rates of thromboembolic events (1.2% vs 1.2%; HR, 1.04; 95% CI, 0.78–1.40), myocardial infarction (0.6% vs 0.6%; HR, 0.98; 95% CI, 0.67–1.45), and death (2.6% vs 2.8%; HR, 0.93; 95% CI, 0.76–1.12). The composite outcome of major bleeding, thromboembolic events, or myocardial infarction rate for apixaban was 2.9% vs 3.3% for rivaroxaban (HR, 0.86; 95% CI, 0.72–1.03; Figure 2).

Figure 2.

Outcomes within 90 and 365 days comparing apixaban vs rivaroxaban in atrial fibrillation (AF) patients with high bleeding risk. Thromboembolic event includes stroke, systemic embolism, and transient ischemic attack. ^aEvent rates are generated from the IPTW cohort. IPTW = inverse probability of treatment weighting; MI = myocardial infarction; TIA = transient ischemic attack.

At 1 year, the apixaban group had lower rates of both major bleeding (2.9% vs 4.2%; HR, 0.69; 95% CI, 0.58–0.81) and any bleeding (10.4% vs 13.4%; HR, 0.76; 95% CI, 0.70–0.84) with no difference in rates of thromboembolic events (3.1% vs 3.0%; HR, 1.02; 95% CI, 0.86–1.22), myocardial infarction (1.7% vs 1.7%; HR, 1.01; 95% CI, 0.80–1.29), and death (9.1% vs 9.0%; HR, 1.02; 95% CI, 0.92–1.14). The composite outcome of major bleeding, thromboembolic events, or myocardial infarction rate for apixaban was 7.3% vs 8.5% for rivaroxaban (HR, 0.86; 95% CI, 0.77–0.96; Figure 2).

Site-specific bleeding in high-risk cohort

At 1 year, the apixaban group had lower rates of both GI bleeding (2.4% vs 3.8%; HR, 0.63; 95% CI, 0.53–0.76) and genitourinary bleeding (1.1% vs 1.9%; HR, 0.59; 95% CI, 0.46–0.75) with no difference in rates of subarachnoid or intracranial bleeding compared with rivaroxaban (0.5% vs 0.4%; HR, 1.08; 95% CI, 0.68–1.71; Supplemental Figure 1; Supplemental Table 2).

Standard vs reduced dose in high-risk cohort

At 1 year, patients with standard dose apixaban had lower rates for major bleeding (HR, 0.74; 95% CI, 0.60–0.91) with no difference in rates of thromboembolic events (HR, 0.94; 95% CI, 0.76–1.17) in comparison to standard dose rivaroxaban. At 1 year, patients with reduced dose apixaban had lower major bleeding rates in comparison to those with reduced dose rivaroxaban (HR, 0.62; 95% CI, 0.48–0.81) with no difference in rates of thromboembolic events (HR, 1.21; 95% CI, 0.88–1.68).

Low bleeding risk cohort

Baseline characteristics in low-risk patients

Before IPTW, in the low bleeding risk cohort, 31,156 patients were treated with apixaban and 24,030 patients were treated with rivaroxaban. After IPTW using the propensity score, patients in the apixaban and rivaroxaban groups were well balanced for baseline values of demographics, comorbidities, and medications with standardized differences consistently <0.1. Both groups had similar mean age of 77 years, and 51% were female in the apixaban group vs 50% in the rivaroxaban group (Table 2).

Table 2.

Baseline characteristics comparing apixaban vs rivaroxaban in AF patients with low bleeding risk

	Unweighted			Weighted
Characteristics	Apixaban	Rivaroxaban	SD	Apixaban	Rivaroxaban	SD
No. of participants	31,156	24,030
Age at index, y, mean	77.8	76.1	0.2	77.0	77.0	0.001
Women	52.1	48.0	0.08	50.5	50.0	0.01
Neighborhood income quintile
Quintile 1 (lowest)	19.4	18.2	0.03	18.9	18.9	0.001
Quintile 5 (highest)	21.1	22.4	0.03	21.8	21.5	0.006
Rural residence	12.2	13.9	0.05	13.1	12.9	0.004
Charlson score, mean	0.6	0.5	0.1	0.6	0.6	0.006
Cardiovascular risk factorsa
Diabetes	27.6	25.9	0.04	26.9	26.9	0
Hypertension	77.6	75.6	0.05	76.6	76.8	0.005
Heart failure	25.1	20.4	0.1	23.0	23.0	0
Stroke	0.4	0.3	0.02	0.4	0.4	0.001
Transient ischemic attack	0.9	0.8	0.02	0.9	0.9	0.002
Myocardial infarction	4.6	3.9	0.04	4.4	4.2	0.009
Chronic obstructive pulmonary disease	26.2	24.9	0.03	25.7	25.7	0
Peripheral vascular disease	6.8	5.1	0.07	6.1	6.0	0.008
Percutaneous coronary intervention	2.1	2.0	0.008	2.0	2.0	0.004
Coronary artery bypass graft	1.2	1.2	0.006	1.2	1.2	0.003
Major bleeding	0.3	0.2	0.008	0.3	0.3	0.003
HAS-BLED bleeding risk score, mean	1.8	1.8	0.05	1.8	1.8	0.008
HAS-BLED bleeding risk score
1	18.9	20.9	0.05	19.9	19.6	0.008
2	81.1	79.2	0.05	80.1	80.4	0.008
CHA2DS2-VASc score, mean Prior medical conditionsb	3.6	3.4	0.2	3.5	3.5	0.004
Chronic kidney disease within 1 year	0.3	0.2	0.02	0.2	0.2	0.003
Chronic kidney disease	0.5	0.4	0.02	0.5	0.5	0.004
Anemia	4.1	3.1	0.05	3.6	3.7	0.005
Cancer, primary	5.7	4.9	0.04	5.4	5.3	0.006
Cancer, metastatic	0.4	0.5	0.02	0.5	0.5	0.006
Rheumatic disease	0.7	0.6	0.02	0.7	0.6	0.01
Peptic ulcer disease	0.6	0.4	0.03	0.5	0.5	0.002
Medication usec
Antiplatelets, anticoagulants,	2.2	2.6	0.02	2.3	2.5	0.009
nonsteroidal anti-inflammatory   within 100 days before index date
ACE inhibitors	22.8	22.9	0.001	22.8	22.8	0.001
Angiotensin II receptor blockers	19.2	19.2	0	19.1	19.4	0.008
Beta blockers	40.4	38.9	0.03	39.7	39.7	0.001
Calcium channel blockers	29.3	27.6	0.04	28.5	28.7	0.004
Clopidogrel	0.5	0.7	0.02	0.5	0.6	0.02
Digoxin	4.2	4.2	0.002	4.2	4.1	0.005
Insulin	2.9	2.3	0.04	2.7	2.6	0.007
Lipid-lowering agents	41.7	40.9	0.02	41.3	41.5	0.006
Oral antiglycemics	13.5	12.4	0.03	13.1	12.9	0.005
Proton pump inhibitors	25.9	24.0	0.05	25.1	25.0	0.001

Values are reported as percentage unless otherwise indicated.

ACE = angiotensin-converting enzyme; AF = atrial fibrillation.

^aLookback period for cardiovascular risk factors was 5 years unless otherwise stated.

^bLookback period for prior medical conditions was 5 years unless otherwise stated.

^cLookback period for medication use was 90 days before index date unless otherwise stated.

Outcomes associated with apixaban vs rivaroxaban in low-risk IPTW cohort

Within 90 days of the first dispensed prescription, the apixaban group had lower rates of both major bleeding (0.6% vs 1.1%; HR, 0.55; 95% CI, 0.46–0.67) and any bleeding (2.6% vs 4.0%; HR, 0.64; 95% CI, 0.59–0.71) with no difference in rates of thromboembolic events (HR, 1.02; 95% CI, 0.83–1.27). The composite outcome of major bleeding, thromboembolic events, or myocardial infarction rate for apixaban was 1.5% vs 2.0% for rivaroxaban (HR, 0.76; 95% CI, 0.67–0.86; Figure 3).

Figure 3.

Outcomes within 90 and 365 days comparing apixaban vs rivaroxaban in atrial fibrillation (AF) patients with low bleeding risk. Thromboembolic event includes stroke, systemic embolism, and transient ischemic attack. ^aEvent rates are generated from the IPTW cohort. IPTW = inverse probability of treatment weighting; MI = myocardial infarction; TIA = transient ischemic attack.

At 1 year, the apixaban group had lower rates of both major bleeding (1.8% vs 2.9%; HR, 0.63; 95% CI, 0.56–0.70) and any bleeding (7.3% vs 10.2%; HR, 0.71; 95% CI, 0.67–0.75) with no difference in rates of thromboembolic events (1.9% vs 1.9%; HR, 1.00; 95% CI, 0.89–1.14), myocardial infarction (0.8% vs 0.9%; HR, 0.93; 95% CI, 0.77–1.12), and death (6.1% vs 5.7%; HR, 1.07; 95% CI, 0.99–1.15). The composite outcome of major bleeding, thromboembolic events, or myocardial infarction rate for apixaban was 4.4% vs 5.4% for rivaroxaban (HR, 0.80; 95% CI, 0.74–0.86; Figure 3).

Site-specific bleeding in low-risk cohort

At 1 year, the apixaban group had lower rates of subarachnoid or intracranial bleeding (0.3% vs 0.4%; HR, 0.73; 95% CI, 0.54–0.98), GI bleeding (1.6% vs 2.5%; HR, 0.61; 95% CI, 0.54–0.69), and genitourinary bleeding (1.2% vs 1.9%; HR, 0.62; 95% CI, 0.54–0.72) in comparison to the rivaroxaban group (Supplemental Figure 2; Supplemental Table 2).

Standard vs reduced dose in low-risk cohort

At 1 year, patients with standard dose apixaban had lower rates of major bleeding (HR, 0.64; 95% CI, 0.56–0.73) with no difference in rates of thromboembolic events (HR, 0.99; 95% CI, 0.85–1.15) in comparison to standard dose rivaroxaban. At 1 year, patients with reduced dose apixaban had lower rates of major bleeding (HR, 0.60; 95% CI, 0.48–0.74) with no difference in rates of thromboembolic events (HR, 1.04; 95% CI, 0.83–1.29) in comparison to those with reduced dose rivaroxaban.

Discussion

In AF patients who were aged 66 years and older, treatment with apixaban was associated with lower incidence of major bleeding and a similar incidence of thromboembolic events compared with rivaroxaban. This result was consistent across patients with high and low estimated bleeding risk, for patients receiving standard or reduced dose, and for the different bleeding sites we examined.

Apixaban is taken twice daily with absolute bioavailability of 50% and is not affected by food intake. Rivaroxaban is taken once daily, and bioavailability is dose dependent; the absolute bioavailability of 20 mg in fasting conditions is 66%, with higher bioavailability (≥80%) of 15 mg and 20 mg of rivaroxaban when taken with food. Also, 25% and 33% of unchanged apixaban and rivaroxaban are cleared by the kidney. Those differences in pharmacodynamic and pharmacokinetic profiles between the drugs, coupled with different doses, could affect their safety and efficacy.^23,24 The lower incidence of bleeding with apixaban was of a clinically relevant size. For AF patients with high bleeding risk, apixaban was associated with an absolute major bleeding incidence difference of 21.3% (2.9% with apixaban to 4.2% with rivaroxaban), resulting in a number needed to treat to prevent 1 major bleeding event at 1 year of 77 compared with rivaroxaban. For AF patients with high bleeding risk, we observed an absolute difference in the incidence of hospitalization for any bleeding of 23.0% (10.4% with apixaban to 13.4% with rivaroxaban). Similar observations were noted for AF patients with low bleeding risk with absolute major bleeding incidence difference of 21.1% and absolute difference in any bleeding hospitalization of 22.9% with treatment with apixaban. This lower incidence is substantial at the population level, given the large number of AF patients receiving DOACs in the community. Our findings thus support a preferential use of apixaban, regardless of estimated bleeding risk, with a favorable absolute risk reduction.

Whereas other studies have compared apixaban with rivaroxaban, they were based on smaller cohorts. The study by Ciou and colleagues²⁵ included 964 AF patients comparing the different DOAC regimens in AF patients with high bleeding risk, whereas our study included 18,156 AF patients with high bleeding risk and 55,186 AF patients with low bleeding risk, which represents a large sample of older patients with AF who were entirely 66 years or older.

We found lower rates of site-specific bleeding with apixaban compared with rivaroxaban. In the ROCKET AF trial,²⁶ the rate of GI bleeding was higher in the rivaroxaban group in comparison to warfarin. In the ARISTOTLE trial,²⁷ there was an 11% numerically lower rate of major bleeding from a GI site in the apixaban group in comparison to warfarin, which was not statistically significant. In addition, apixaban is safer than warfarin in reducing major bleeding across the spectrum of risk of bleeding according to the HAS-BLED score.²⁸ This indirect evidence from randomized controlled trial data suggests that apixaban may have a better GI-specific bleeding rate and supports our results of lower GI bleeding incidence in the apixaban group for patients with high and low bleeding risk. Our findings are also consistent with observational data from Medicare, Medicaid, and insurance providers¹⁴ and extend these to report on AF patients with high and low estimated bleeding risk based on HAS-BLED scores, not only for patients with high risk of GI bleeding. This makes our results more generalizable to the AF population.

Current AF guidelines do not provide specific recommendations regarding the choice of individual DOACs for patients with high or low bleeding risk^29,30; with the growing body of evidence around the efficacy and safety of the individual DOACs, we expect this to change. The higher bleeding rates observed with rivaroxaban in our study should be taken into consideration when choosing between apixaban and rivaroxaban for stroke prevention in older AF patients, especially in those with high bleeding risk.

Study limitations

Potential limitations of our study deserve discussion. First, as frequently encountered in nonrandomized studies, unmeasured confounding and selection bias remains a risk. To mitigate this risk, we performed extensive adjustment for important clinical factors as we accounted for multiple possible confounders when estimating the propensity score. After IPTW by the propensity score, patients in the apixaban and rivaroxaban groups were well balanced for measured demographics, comorbidities, and medications, with standardized differences consistently <0.1. Second, given that patients in the apixaban group had higher baseline comorbidities, one should expect worse outcomes. However, bleeding rate was lower in the apixaban group, despite this source of bias in favor of rivaroxaban. Also, our observational study used a new user design, compared 2 treatment groups, and collected outcomes universally, and hence one should expect much less bias in comparison to observational studies comparing treatment vs no-treatment groups. Third, we were not able to verify whether patients took their medications and could not identify short interruptions, switching between medications, actual discontinuation of medications, or if reduced doses were appropriately adjusted; however, this would result in a bias toward the null. Our study was designed as intention-to-treat as in a clinical trial. Also, we focused on outcomes that occurred during the first 90 days and 365 days to avoid medication crossover that may occur with DOACs during longer periods; hence, we believe the impact of switching of DOAC is likely to be low.^31,32 Fourth, we included only patients with AF aged 66 years or older, and hence our results might differ in a younger population. Fifth, given the inclusion criteria for cohort entry, including AF diagnosis in the outpatient setting only, our findings might not be generalizable to the inpatient AF population. Sixth, there is a limitation of the accuracy of measuring HAS-BLED score with administrative data using hospitalization records, but patients receiving both drugs had the same measurement. We were limited to the assessment of bleeding requiring hospitalization or in patients hospitalized with bleeding because there are no validated methods for identifying minor bleeding in the outpatient setting not requiring hospitalization. Finally, although it was used commonly when it was first approved, dabigatran has largely been replaced by apixaban and rivaroxaban³³; thus, there was a small number of patients receiving dabigatran, and we could not adequately examine the outcomes. Also, there were not many patients receiving the newly introduced edoxaban, which would limit the analysis of our outcomes of interest.

Conclusion

For AF patients at high or low bleeding risk, 66 years or older, treatment with apixaban was associated with lower incidence of both major and any bleeding events but with no difference in the rate of thromboembolic events. Our study provides important evidence in favor of the use of apixaban rather than rivaroxaban for treating ambulatory patients with AF, regardless of baseline high or low bleeding risk, who are prescribed a DOAC as their initial anticoagulant.

Abbreviations

AF:

atrial fibrillation

DOAC:

direct oral anticoagulant

GI:

gastrointestinal

IPTW:

inverse probability of treatment weighting