You are here

Comparative Stroke, Bleeding, and Mortality Risks in Older Medicare Patients Treated with Oral Anticoagulants for Nonvalvular Atrial Fibrillation

Am J Med. 2019;132(5):596-604.e11

Abstract

Background

Nonvitamin K antagonist oral anticoagulants (NOACs) are alternatives to warfarin in patients with nonvalvular atrial fibrillation. Randomized trials compared NOACs with warfarin, but none have compared individual NOACs against each other for safety and effectiveness.

Methods

We performed a retrospective new-user cohort study of patients with nonvalvular atrial fibrillation enrolled in US Medicare who initiated warfarin (n = 183,318), or a standard dose of dabigatran (150 mg twice daily; n = 86,198), rivaroxaban (20 mg once daily; n = 106,389), or apixaban (5 mg twice daily; n = 73,039) between October 2010 and September 2015. Propensity score-adjusted Cox proportional hazards regression was used to estimate adjusted hazard ratios (HR) and 95% confidence intervals (CI) for the outcomes of thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality, comparing each NOAC with warfarin, and with each other NOAC.

Results

Compared with warfarin, each NOAC was associated with reduced risks of thromboembolic stroke (20%-29% reduction; P = .002 [dabigatran], P < 0.001 [rivaroxaban, apixaban]), intracranial hemorrhage (35%-62% reduction; P < 0.001 [each NOAC]), and mortality (19%-34% reduction; P < .001 [each NOAC]). The NOACs were similar for thromboembolic stroke but rivaroxaban was associated with increased risks of intracranial hemorrhage (vs dabigatran: HR = 1.71; 95% CI, 1.35-2.17), major extracranial bleeding (vs dabigatran: HR = 1.32; 95% CI, 1.21-1.45; vs apixaban: HR = 2.70; 95% CI, 2.38-3.05), and death (vs dabigatran: HR = 1.12; 95% CI, 1.01-1.24; vs apixaban: HR = 1.23; 95% CI, 1.09-1.38). Dabigatran was associated with reduced risk of intracranial hemorrhage (HR = 0.70; 95% CI ,0.53-0.94) and increased risk of major extracranial bleeding (HR = 2.04; 95% CI, 1.78-2.32) compared with apixaban.

Conclusions

Among patients treated with standard-dose NOAC for nonvalvular atrial fibrillation and warfarin users with similar baseline characteristics, dabigatran, rivaroxaban, and apixaban were associated with a more favorable benefit–harm profile than warfarin. Among NOAC users, dabigatran and apixaban were associated with a more favorable benefit–harm profile than rivaroxaban.

Keywords: Anticoagulants, Apixaban, Atrial fibrillation, Comparative effectiveness, Comparative safety, Dabigatran, Rivaroxaban, Warfarin.

 

 

  • Nonvitamin K oral anticoagulants (NOACs) are used for stroke prevention in patients with atrial fibrillation, but no clinical trials have compared them against each other for effectiveness or safety.

  • In an observational study of Medicare patients with nonvalvular atrial fibrillation, risks of thromboembolic stroke, intracranial bleeding, and all-cause mortality were reduced in patients treated with apixaban, dabigatran, and rivaroxaban compared with warfarin patients having similar baseline characteristics.

  • Among NOACs, apixaban and dabigatran had a more favorable benefit–harm balance than rivaroxaban.

 

Clinical Significance

 

Introduction

In the pivotal randomized clinical trials of the nonvitamin K oral anticoagulants (NOACs) in patients with nonvalvular atrial fibrillation, dabigatran and apixaban were superior to, and rivaroxaban and edoxaban were noninferior to, warfarin for stroke and systemic embolization prevention. 1 2 3 4 Each of the NOACs also reduced risk of intracranial hemorrhage. 1 2 3 4 There have been no head-to-head randomized trials to determine if any of the NOACs differ from the others with respect to their effects on stroke, bleeding, and mortality, and only a handful of observational studies have compared dabigatran, rivaroxaban, and apixaban, the 3 most widely marketed NOACs, to each other in the same study. 5 6 7 8 9 10 11 These were uniformly small, rarely examined thromboembolic stroke, and none assessed mortality. The question remains unanswered whether the NOACs are therapeutically similar or if clinically important differences exist, which might lead prescribers and patients to prefer one over the others.

To better assess the comparative safety and effectiveness of the commonly marketed oral anticoagulants, we performed a large observational study in older US patients with atrial fibrillation enrolled in fee-for-service Medicare, that compared each NOAC with warfarin and with each other NOAC for the outcomes of hospitalized thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality.

Methods

Each of the NOACs is available in the United States at a standard dose intended for most nonvalvular atrial fibrillation patients and a lower dose for those with impaired renal function (dabigatran, rivaroxaban, edoxaban), or a combination of at least 2 of impaired renal function, lower body weight, or advanced age (apixaban). We employed a new user design 12 to compare patients initiating treatment with warfarin or a standard dose of dabigatran (150 mg twice daily), rivaroxaban (20 mg once daily), or apixaban (5 mg twice daily). Edoxaban use was too low to study.

From October 2010, when dabigatran was approved, through September 2015, patients enrolled in fee-for-service Medicare Part A (hospitalization), Part B (office-based care), and Part D (prescription drug coverage) were eligible for study inclusion if, on the date of a qualifying anticoagulant dispensing (index date), they had at least 6 months of continuous Medicare enrollment, were 65 years or older, and during the preceding 6 months, had an inpatient or outpatient diagnosis of atrial fibrillation or flutter based on International Classification of Diseases, Ninth Revision (ICD-9) coding, and had not filled a prescription for any of the study drugs or edoxaban. Patients were excluded if they were in a skilled nursing facility or nursing home or were receiving hospice care on the index date. Patients with a hospitalization extending beyond the index date were also excluded, as were kidney transplant recipients and dialysis patients. We also excluded patients with a potential alternative indication for anticoagulation in the 6 months preceding study entry (mitral valve disease, heart valve repair or replacement, deep vein thrombosis, pulmonary embolism, or joint replacement).

During the 6 months preceding cohort entry, Medicare claims data on chronic medical conditions, risk factors for cardiovascular and bleeding events (including CHA 2 DS 2 -VASc [Congestive heart failure, Hypertension, Age ≥75 years, Diabetes mellitus, prior Stroke or TIA or thromboembolism, Vascular disease, Age 65-74 years, Sex category] and HAS-BLED [Hypertension, Abnormal renal and liver function, Stroke, Bleeding, Labile international normalized ratio, Elderly, Drugs or alcohol] scores), 13 14 and health care utilization were collected for each patient, as were data on dispensed medications.

To improve the validity of the NOAC vs warfarin comparisons, we performed 1:1 propensity score matching of warfarin to NOAC users with replacement, 15 thereby restricting the study population to all eligible NOAC users and warfarin users with characteristics very similar to the NOAC users ( Supplementary Table 1 , available online). To adjust for potential confounding, we used multinomial logistic regression to estimate stabilized inverse probability of treatment weights for the warfarin and individual NOAC cohorts, using the 3 NOAC cohorts pooled together as the reference for weighting. 16 Covariate balance was assessed using standardized mean differences, with a value ≤0.1 indicating a negligible difference between groups. 17

Follow-up began the day after cohort entry and continued until disenrollment from Medicare, a gap in anticoagulant days of supply exceeding 3 days, dispensing of another anticoagulant, kidney transplantation or initiation of dialysis, admission to a skilled nursing facility or nursing home, transfer to hospice care, end of the study period, or study outcome occurrence, whichever came first. We chose a 3-day gap because of the short half-lives of the individual NOACs (∼12 hours). We censored for admission to a skilled nursing facility or nursing home due to concerns about incomplete capture of outcomes in these settings, and for transfer to hospice care because deaths in these patients were expected and therefore unlikely to be anticoagulant related.

Study outcomes were hospitalized thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality. To facilitate comparison with other studies, results for major gastrointestinal bleeding, a large component of major extracranial bleeding, are also presented in the Supplement (available online). Outcomes were defined using ICD-9 codes from the first hospital discharge diagnosis position ( Supplementary Table 2 , available online). Thromboembolic stroke was defined using a validated algorithm with a positive predictive value (PPV) of 88%-95%. 18 19 20 Intracranial hemorrhage was defined using a validated algorithm for nontraumatic intracranial hemorrhage (PPV 89%-97%), 18 19 20 to which we added codes for intracranial hemorrhage accompanied by nonpenetrating head trauma to capture events that might have been preceded by a fall. Major extracranial bleeding was based on a validated algorithm for hospitalized bleeding by Cunningham et al (PPV 87%), 21 with a requirement that the bleeding event was treated with red blood cell or whole blood transfusion, or involved a critical site (ie, intraarticular, pericardial, retroperitoneal), 21 or resulted in death. Mortality was ascertained by linkage to Social Security files, which capture over 95% of deaths for persons 65 years or older in the United States. 22 Our death outcome included deaths occurring as the first study outcome or within 30 days after hospitalization for another outcome event.

Weighted Kaplan-Meier cumulative incidence plots were generated to characterize risk over time. 23 For each outcome, a single weighted Cox proportional hazards model with robust estimation was used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for all NOAC–warfarin and NOAC–NOAC pairwise comparisons. Adjusted incidence rates and incidence rate differences were also estimated. For thromboembolic stroke, intracranial hemorrhage, and major extracranial bleeding, the adjusted 30-day case fatality rate was determined.

Prespecified subgroup analyses were performed for all outcomes in categories defined by age, sex, use of antiplatelet agents , and CHA 2 DS 2 -VASc and HAS-BLED scores. Prespecified sensitivity analyses included increasing the gap between anticoagulant prescriptions from 3 to 14 days; restricting analysis to patients with at least 2 dispensings of their study drug; restriction to patients initiating an anticoagulant on or after apixaban's approval date (December 28, 2012) to examine for potential time-period bias; and repeating the analysis using Cox regression without and with multivariable adjustment. We also performed several post hoc sensitivity analyses ( Appendix , available online).

This study was classified as public health surveillance by the Food and Drug Administration (FDA) and was exempt from review by its Research in Human Subjects Committee. Analyses were performed using R 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria) and SAS v. 9.4 (SAS Institute Inc., Cary, NC).

Results

A total of 448,944 anticoagulant initiators contributed 159,927 person-years of on-treatment follow-up (mean duration 130 days). The mean age across cohorts was 75.4 years, of whom 47.4% were female. Prior to adjustment, there were minor differences for age, kidney failure, obesity, smoking, cardioversion, emergency department visits, prescriber specialty, and use of injectable anticoagulants, digoxin, and loop diuretics ( Table , Supplementary Table 3 , available online) After inverse probability of treatment weights adjustment, cohorts were closely balanced for all covariates ( Supplementary Table 4 , available online).

Characteristic, % Warfarin(n = 183,318) Dabigatran(n = 86,198) Rivaroxaban(n = 106,389) Apixaban(n = 73,039) Maximum Pairwise SMD *
Age, years (mean) 75.8 75.5 74.9 75.2 0.15
Female 48.0 47.6 46.1 47.8 0.04
Race/ethnicity          
White 91.8 91.8 91.9 92.4 0.02
Black 3.8 3.5 3.5 3.7 0.02
Other 4.4 4.7 4.6 3.9 0.04
Medical comorbidities          
Diabetes 34.2 33.4 32.0 33.9 0.05
Hypercholesterolemia 37.7 39.5 38.1 38.1 0.04
Hypertension 86.3 86.3 85.9 87.6 0.05
Kidney failure          
Acute 4.9 3.3 3.6 5.5 0.11
Chronic 12.1 9.6 8.5 13.5 0.16
Obesity 15.1 13.5 16.6 18.7 0.14
Peptic ulcer disease 0.4 0.4 0.4 0.4 0.01
Prior hospitalized bleeding 0.6 0.5 0.4 0.5 0.03
Smoking 20.5 17.9 22.3 24.4 0.16
Cardiovascular disease          
Acute myocardial infarction          
Past 1-30 days 1.5 1.0 1.3 1.4 0.04
Past 31-183 days 0.8 0.7 0.7 1.0 0.03
Coronary revascularization 15.6 14.4 14.5 16.4 0.05
Heart failure          
Hospitalized 3.6 3.1 2.9 3.5 0.04
Outpatient 12.6 12.8 10.7 11.8 0.06
Other ischemic heart disease 44.6 43.9 42.1 45.1 0.05
Stroke          
Past 1-30 days 2.2 1.9 1.9 2.1 0.02
Past 31-183 days 1.2 1.3 1.1 1.5 0.03
Other          
Transient ischemic attack 6.3 6.4 6.0 6.6 0.03
Cardioablation 1.8 2.2 2.1 2.4 0.04
Cardioversion 6.9 9.2 9.1 10.5 0.13
Other medical conditions          
Falls 5.0 4.5 4.6 4.7 0.02
Fractures 1.4 1.3 1.3 1.3 0.02
Syncope 8.7 8.5 8.8 9.7 0.04
Walker use 2.7 2.4 2.2 2.5 0.04
CHA 2 DS 2 -VASc score          
0-1 2.9 3.3 3.4 2.6 0.04
2 14.6 16.0 17.5 15.6 0.08
3 27.9 28.4 29.9 28.9 0.05
4 28.8 28.4 27.2 27.9 0.04
5 16.0 14.9 14.0 15.4 0.05
≥6 9.9 9.1 8.0 9.6 0.07
HAS-BLED score          
1 9.3 9.2 9.6 8.3 0.05
2 44.9 46.1 46.7 43.9 0.06
3 31.0 31.4 31.3 32.2 0.02
≥4 14.8 13.3 12.4 15.6 0.09
Medication use          
General          
Estrogen replacement 1.9 2.3 2.1 1.9 0.03
H2-antagonists 5.3 5.0 4.9 5.2 0.02
NSAIDs 13.0 14.0 14.3 13.9 0.04
Proton pump inhibitors 26.8 25.6 26.6 29.0 0.08
SSRI antidepressants 12.9 12.7 12.5 13.3 0.02
Cardiovascular          
ACEI/ARB 58.8 58.6 58.1 60.2 0.04
Antiarrhythmics 15.2 16.8 16.7 17.7 0.07
Anticoagulants (injectable) 9.1 6.9 9.0 9.9 0.11
Antiplatelets 14.2 14.3 13.9 15.6 0.05
Beta-blockers 67.6 66.3 67.2 69.9 0.08
Calcium channel blockers 41.8 41.8 41.6 42.1 0.01
Digoxin 13.1 14.8 10.6 8.9 0.18
Diuretics          
Loop 26.7 24.8 21.8 25.1 0.11
Potassium sparing 8.5 8.3 7.5 8.2 0.04
Thiazide 28.1 28.8 28.4 28.8 0.02
Nitrates 10.0 9.3 8.5 9.7 0.05
Statins 58.3 57.3 57.6 60.8 0.07
Fibrates 4.6 4.6 4.1 4.6 0.03
Diabetes-related          
Insulin 7.1 6.3 6.0 7.4 0.06
Metformin 15.0 14.8 15.1 15.7 0.02
Sulfonylureas 9.1 9.1 8.0 8.4 0.04
Other 5.9 6.2 5.6 6.2 0.03
Metabolic inhibitors §          
Amiodarone 9.6 9.0 9.2 9.6 0.02
Dronedarone 2.7 4.5 3.2 3.2 0.09
Prescriber specialty          
Cardiology 49.5 52.8 54.7 59.1 0.19
Family medicine 9.9 9.9 8.2 6.2 0.14
Internal medicine 22.0 19.7 20.0 17.2 0.12
Other 18.7 17.5 17.0 17.5 0.04

SMD = Standardized mean difference. The maximum SMD among all pairwise comparisons is shown for each covariate prior to inverse probability of treatment weighting.

The CHA 2 DS 2 -VASc score assigns points for the presence of Congestive heart failure, Hypertension, Age 65-74 years and age ≥75 years, Diabetes mellitus, Stroke or transient ischemic attack, Vascular disease, and female sex. 13

The HAS-BLED score assigns points for the presence of Hypertension, Abnormal renal or liver function, Stroke, Bleeding history, Labile international normalized ratio (INR), age ≥ 65 years, and antiplatelet Drug or alcohol use. 14 Labile INR was not included in our estimation of the HAS-BLED score because INR laboratory results were not available in Medicare claims data.

§ Day's supply of use overlapped with the date of first prescription for dabigatran or rivaroxaban.

ACEI/ARB = angiotensin-converting enzyme inhibitor/angiotensin-receptor blocker; NSAID = nonsteroidal antiinflammatory drug; SSRI = selective serotonin reuptake inhibitor.

TableDemographic and Clinical Characteristics of Medicare Beneficiaries Initiating Warfarin, Dabigatran, Rivaroxaban, or Apixaban for Nonvalvular Atrial Fibrillation from 2010 to 2015, Prior to Inverse Probability of Treatment Weighting

  Warfarin Dabigatran(300 mg/d) Rivaroxaban(20 mg/d) Apixaban(10 mg/d)
Received oral anticoagulant fill between Oct 19, 2010 and Sept 30, 2015 4,805,044 254,242 339,721 179,205
Age ≥65 y and enrolled in Medicare Parts A, B, & D for ≥6 months 2,582,587 152,425 194,390 106,579
No oral anticoagulant use in 6 months prior to cohort-defining (index) fill, only in one cohort on the index date, and no missing dosage 1,439,770 108,210 156,494 96,057
Not in nursing home, skilled nursing, or hospice on index, or hospitalized beyond index 1,213,948 103,368 144,989 90,936
Atrial fibrillation/flutter diagnosis during 6 months prior to index 714,421 96,684 122,981 83,792
No diagnoses of DVT/PE, valvular heart disease, joint replacement, dialysis, kidney transplant in 6 months prior to index 523,264 86,198 106,389 73,039
Propensity score matching of warfarin to pooled NOACs users with replacement 183,318 86,198 106,389 73,039
DVT = deep vein thrombosis; NOAC = nonvitamin K antagonist oral anticoagulant; PE = pulmonary embolism.

Supplementary Table 1Number of Patients Meeting Study Eligibility Criteria at Various Steps in the Cohort Development Process

Outcome ICD-9 Codes Position Setting
Ischemic stroke 433.x1, 434.x (except subcode: x0), 436 1st IP only
Intracranial hemorrhage 430, 431, 432, + (852.0, 852.2, 852.4, 853.0) 1st IP only
Extracranial bleeding events A bleeding event is defined as a definite bleeding code, or a possible bleeding code (primary) supported by a definite bleeding code (secondary); without a corresponding trauma code (as defined in Cunningham et al 21 )
Definite bleeding: 531.0x, 531.2x, 531.4x, 531.6x, 532.0x, 532.2x, 532.4x, 532.6x, 533.0x, 533.2x, 533.4x, 533.6x, 534.0x, 534.2x, 534.4x, 534.6x, 535.01, 535.11, 535.21, 535.31, 535.41, 535.51, 535.61, 537.83, 456.0, 456.20, 530.7,530.82, 578.0, 455.2, 455.5, 455.8, 562.02, 562.03, 562.12, 562.13, 568.81, 569.3, 569.85, 578.1, 578.9, 593.81, 599.7, 623.8, 626.2, 626.6, 423.0, 459.0, 719.1x, 784.7, 784.8, 786.3
Possible bleeding: 531.1, 531.3, 531.5, 531.7, 531.9, 532.1, 532.3, 532.5, 532.7, 532.9, 533.1, 533.3, 533.5, 533.7, 533.9, 534.1, 534.3, 534.5, 534.7, 534.9, 535.00, 535.10, 535.20, 535.30, 535.40, 535.50, 535.60, 455.x, 562.00, 562.01, 562.10, 562.11, 530.1, 280.0, 285.1, 285.9, 790.92
1st IP only
Major bleeding events Major bleeding is defined as a hospitalized bleeding event with (i) a critical site code, (ii) a transfusion, or (iii) death.
Critical site: Intracranial: 430, 431, 432, 852.0, 852.2, 852.4, 853.0; Extracranial: 423.0, 568.81, 719.1x
Transfusions: a) ICD-9 PRC: 9903, 9904, 9905, 9906, 9907, 9909, b) HCPC: P9010, P9011, P9016, P9017, P9019, P9020, P9021, P9022, P9023, P9031-P9040, P9044, P9051 - P9060, c) Revenue Center Codes: 0380-0392, 0399, d) Additional Value Codes: 37, 38, 39
N/A IP only
Major GI bleeding events Major GI bleeding is defined as a major bleeding event at a GI site N/A IP only

Intracranial hemorrhage was defined using codes for atraumatic hemorrhage (430-432), with a PPV of 89%-97%, 18 19 20 and codes for hemorrhage with closed head trauma (852-853), which have not been validated. We included these latter codes to capture situations where a bleeding event preceded by a fall may have been coded as trauma related.

GI = gastrointestinal; IP = inpatient.

Supplementary Table 2International Classification of Disease, Ninth Edition, Clinical Modification (ICD-9-CM) Codes Used to Define Study Outcomes

Characteristic, % Warfarin(n = 183,318) Dabigatran(n = 86,198) Rivaroxaban(n = 106,389) Apixaban(n = 73,039) Maximum Pairwise SMD Prior to Weighting *
Low income 18.1 19.0 16.1 15.6 0.09
Geographic region          
Northeast 19.6 20.2 18.6 18.3 0.05
South 41.2 41.9 42.9 45.6 0.09
Midwest 22.7 21.1 21.5 20.8 0.04
West 16.6 16.8 16.9 15.2 0.05
Rural metropolitan statistical area 24.0 23.6 22.3 23.6 0.04
Medical comorbidities          
Alcohol abuse 1.8 1.8 2.1 1.9 0.02
Anemia 21.5 20.1 18.9 20.5 0.06
Chronic liver disease 1.9 1.8 2.0 2.1 0.02
COPD 20.1 18.6 18.9 19.2 0.04
Dementia 3.9 3.5 3.6 3.4 0.03
Gout 5.9 5.5 5.3 6.1 0.04
Malignancy 23.1 22.7 23.4 23.9 0.03
Peripheral vascular disease 17.7 16.8 16.0 17.2 0.05
Other medical conditions          
Home health care 10.6 9.5 8.8 9.7 0.06
Home oxygen 8.2 7.4 7.2 7.7 0.04
Wheelchair use 2.3 2.3 1.8 1.7 0.04
Hospitalizations (nonbleeds) 38.0 36.9 37.6 37.8 0.02
Emergency department visits in past 0-30 days (n)          
0 87.1 88.6 84.8 85.9 0.11
1 11.5 10.3 13.6 12.7 0.10
≥2 1.4 1.1 1.6 1.4 0.04
Emergency department visits in past 31-183 days (n)          
0 83.7 84.7 84.4 82.2 0.07
1 12.6 12.0 12.3 13.8 0.05
≥2 3.7 3.4 3.4 4.1 0.04
Medication use          
Corticosteroids 15.8 14.7 16.3 16.9 0.06
Thyroid replacement 19.7 19.2 18.7 20.2 0.04
Prescriber characteristics          
Age group, y          
<40 18.3 17.5 19.4 20.4 0.07
40-59 60.7 62.2 60.8 60.3 0.05
≥60 19.8 18.6 19.0 18.4 0.04
Unknown 1.2 1.7 0.9 0.9 0.07
Female 19.6 17.0 19.1 19.8 0.07

SMD = Standardized mean difference. The maximum SMD among all pairwise comparisons is shown for each covariate prior to inverse probability of treatment weighting.

COPD = chronic obstructive pulmonary disease.

Supplementary Table 3Distribution of Additional Baseline Sociodemographic and Medical Factors Included in Cohorts of Medicare Beneficiaries Initiating Warfarin, Dabigatran, Rivaroxaban, or Apixaban for Nonvalvular Atrial Fibrillation from 2010-2015

Characteristic, % Warfarin(n = 183,003) * Dabigatran(n = 86,293) * Rivaroxaban(n = 106,369) * Apixaban(n = 72,921) * Maximum Pairwise SMD after Weighting
Age, years (mean) 75.2 75.1 75.1 75.1 0.01
Age group, y          
65-74 50.5 50.7 50.2 50.1 0.01
75-84 39.1 39.8 40.6 40.5 0.03
85+ 10.3 9.5 9.2 9.3 0.04
Female 47.2 46.9 47.1 47.2 0.01
Race/ethnicity          
White 92.0 92.1 92.0 92.0 0.00
Black 3.6 3.5 3.5 3.6 0.01
Other 4.4 4.4 4.5 4.4 0.00
Medical comorbidities          
Diabetes 33.4 33.0 33.1 33.0 0.01
Hypercholesterolemia 38.6 38.7 38.6 38.6 0.00
Hypertension 86.7 86.5 86.6 86.5 0.01
Kidney failure          
Acute 4.2 4.0 4.0 4.2 0.01
Chronic 10.5 10.1 10.2 10.4 0.01
Obesity 16.3 16.3 16.2 16.3 0.00
Peptic ulcer disease 0.4 0.4 0.4 0.4 0.00
Prior hospitalized bleeding 0.5 0.4 0.5 0.5 0.00
Smoking 21.6 21.5 21.5 21.7 0.00
Cardiovascular disease          
Acute myocardial infarction          
Past 1-30 days 1.3 1.3 1.3 1.3 0.01
Past 31-183 days 0.8 0.8 0.8 0.8 0.01
Coronary revascularization 15.4 15.0 15.1 15.2 0.01
Heart failure          
Hospitalized 3.3 3.1 3.1 3.2 0.01
Outpatient 11.8 11.6 11.8 11.7 0.01
Other ischemic heart disease 44.1 43.6 43.7 43.7 0.01
Stroke          
Past 1-30 days 2.0 1.9 2.0 2.0 0.00
Past 31-183 days 1.3 1.3 1.3 1.3 0.00
Other          
Transient ischemic attack 6.4 6.3 6.3 6.3 0.00
Cardioablation 2.2 2.2 2.2 2.2 0.00
Cardioversion 9.3 9.6 9.5 9.7 0.01
Other medical conditions          
Falls 4.7 4.6 4.6 4.6 0.01
Fractures 1.3 1.3 1.3 1.3 0.00
Syncope 9.1 8.9 9.0 9.0 0.01
Walker use 2.4 2.3 2.3 2.3 0.01
CHA 2 DS 2 -VASc score          
0-1 3.1 3.2 3.1 3.1 0.01
2 16.1 16.6 16.4 16.3 0.01
3 28.9 29.2 29.1 29.0 0.01
≥4 51.9 51.0 51.4 51.5 0.01
HAS-BLED score          
1 8.9 9.1 9.0 9.0 0.01
2 45.3 45.9 45.7 45.5 0.01
3 31.8 31.5 31.6 31.7 0.01
≥4 13.9 13.5 13.7 13.8 0.01
Medication use          
General          
Estrogen replacement 2.1 2.1 2.1 2.1 0.00
H2-antagonists 5.1 5.0 5.1 5.0 0.00
NSAIDs 14.2 14.1 14.2 14.1 0.00
Proton pump inhibitors 27.2 26.8 27.0 27.0 0.01
SSRI antidepressants 12.9 12.8 12.8 12.8 0.00
Cardiovascular          
ACEI/ARB 59.1 58.9 59.0 58.9 0.00
Antiarrhythmics 17.0 17.1 17.0 17.1 0.00
Anticoagulants (injectable) 8.7 8.6 8.6 8.6 0.00
Anti-platelets 14.9 14.4 14.6 14.5 0.01
Beta-blockers 67.7 67.7 67.6 67.8 0.00
Calcium channel blockers 41.9 41.8 41.8 41.8 0.00
Digoxin 11.7 11.5 11.5 11.4 0.01
Diuretics          
Loop 24.1 23.6 23.7 23.9 0.01
Potassium sparing 8.0 8.0 8.0 8.0 0.00
Thiazide 28.6 28.6 28.6 28.6 0.00
Nitrates 9.4 9.1 9.1 9.1 0.01
Statins 58.6 58.5 58.5 58.4 0.00
Fibrates 4.4 4.4 4.4 4.3 0.00
Diabetes-related          
Insulin 6.6 6.5 6.5 6.5 0.01
Metformin 15.3 15.2 15.3 15.1 0.00
Sulfonylureas 8.6 8.5 8.6 8.5 0.00
Other 6.0 5.9 6.0 6.0 0.00
Metabolic inhibitors §          
Amiodarone 9.4 9.2 9.2 9.3 0.01
Dronedarone 3.6 3.6 3.6 3.6 0.00
Prescriber specialty          
Cardiology 55.1 55.6 55.3 55.4 0.01
Family medicine 8.4 8.2 8.3 8.2 0.01
Internal medicine 19.2 19.1 19.1 19.2 0.00
Other 17.4 17.1 17.3 17.2 0.01
Low income 17.3 16.8 17.0 16.9 0.01
Geographic region          
Northeast 18.9 19.0 19.0 19.1 0.00
South 43.3 43.4 43.4 43.2 0.00
Midwest 21.3 21.2 21.2 21.3 0.00
West 16.4 16.4 16.5 16.5 0.00
Rural metropolitan statistical area 23.4 23.0 23.2 23.0 0.01
Medical comorbidities          
Alcohol abuse 1.9 1.9 1.9 2.0 0.00
Anemia 20.0 19.6 19.7 19.9 0.01
Chronic liver disease 2.0 2.0 1.9 2.0 0.00
COPD 19.3 18.8 19.0 19.0 0.01
Dementia 3.6 3.5 3.5 3.5 0.01
Gout 5.6 5.6 5.6 5.6 0.00
Malignancy 23.3 23.3 23.3 23.4 0.00
Peripheral vascular disease 16.9 16.5 16.6 16.7 0.01
Other medical conditions          
Home health care 9.6 9.2 9.3 9.4 0.01
Home oxygen 7.6 7.4 7.5 7.5 0.01
Wheelchair use 2.0 1.9 2.0 1.9 0.01
Hospitalizations (non-bleeds) 38.1 37.4 37.4 38.0 0.01
Emergency department visits in past 0-30 days (n)          
0 86.2 86.3 86.4 86.3 0.01
1 12.4 12.3 12.2 12.3 0.00
≥2 1.4 1.4 1.4 1.4 0.00
Emergency department visits in past 31-183 days (n)          
0 83.6 83.9 83.8 83.9 0.01
1 12.7 12.6 12.6 12.5 0.01
≥2 3.7 3.5 3.6 3.5 0.01
Medication use          
Corticosteroids 16.1 16.0 16.0 16.0 0.00
Thyroid replacement 19.4 19.2 19.3 19.3 0.00
Prescriber characteristics          
Age group, y          
<40 19.0 19.0 19.0 19.1 0.00
40-59 61.1 61.2 61.2 61.1 0.01
≥60 18.8 18.7 18.7 18.7 0.00
Unknown 1.1 1.1 1.1 1.0 0.01
Female 18.6 18.6 18.6 18.7 0.00

Weighted cohort sample size is calculated by summing the stabilized inverse probability of treatment weights from each patient in the cohort. The size of this adjusted pseudo-population can differ slightly from the unadjusted count.

The CHA 2 DS 2 -VASc score assigns points for the presence of congestive heart failure, hypertension, age 65-74 years and age ≥75 years, diabetes mellitus, stroke or transient ischemic attack, vascular disease, and female sex. 13

The HAS-BLED score assigns points for the presence of hypertension, abnormal renal or liver function, stroke, bleeding history, labile international normalized ratio (INR), age ≥65 years, and antiplatelet drug or alcohol use. 14 Patients were not treated with warfarin, and INR testing was not performed, so labile INR was excluded from our scoring.

§ Days supply of use overlapped with the date of first prescription for dabigatran or rivaroxaban.

ACEI/ARB = angiotensin-converting enzyme inhibitor/angiotensin-receptor blocker; COPD = chronic obstructive pulmonary disease; NSAID = nonsteroidal antiinflammatory drug; SMD = standardized mean difference; SSRI = selective serotonin reuptake inhibitor.

Supplementary Table 4Demographic and Clinical Characteristics of Medicare Beneficiaries Initiating Warfarin, Dabigatran, Rivaroxaban, or Apixaban for Nonvalvular Atrial Fibrillation from 2010 to 2015, after Inverse Probability of Treatment Weighting

 

During follow-up, there were 11,263 outcome events ( Supplementary Table 5 , available online). Adjusted incidence rates for thromboembolic stroke, intracranial hemorrhage, and all-cause mortality were highest with warfarin, while major extracranial bleeding rates were highest with rivaroxaban ( Figure 1 ).

  Person-Years of On-Drug Follow-Up Thromboembolic Stroke Intracranial Hemorrhage Major Extracranial Bleed Major GI Bleed All-Cause Mortality
Warfarin 68,785 814 605 1876 1488 2234
Dabigatran 30,368 279 104 839 737 677
Rivaroxaban 39,583 312 213 1335 1119 904
Apixaban 21,191 190 96 329 264 456
Total 159,927 1595 1018 4379 3608 4271

Supplementary Table 5Thromboembolic Stroke, Intracranial Hemorrhage, Major Extracranial Bleeding, Major Gastrointestinal (GI) Bleeding, and All-Cause Mortality Event Counts (Unweighted) During On-Treatment Follow-Up, By Anticoagulant Cohort

gr1
 

Figure 1
Adjusted incidence rates per 1000 person-years of thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality in Medicare beneficiaries with nonvalvular atrial fibrillation treated with warfarin, dabigatran, rivaroxaban, or apixaban. Weighted cohort sizes are shown.

 

Compared with warfarin, each NOAC was associated with significantly lower HRs for thromboembolic stroke (20%-29% reduction; P = .002 [dabigatran], P < .001 [rivaroxaban, apixaban]), intracranial hemorrhage (35%-62% reduction; P < .001 [each NOAC]), and mortality (19%-34% reduction; P < .001 [each NOAC]) ( Figure 2 , Supplementary Table 6 , available online).

gr2
 

Figure 2
Adjusted hazard ratios with 95% confidence intervals from comparisons of each nonvitamin K antagonist oral anticoagulant (NOAC) vs warfarin (A) and each NOAC vs each other NOAC (B) for the outcomes of thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality.

  Thromboembolic Stroke Intracranial Hemorrhage Major Extracranial Bleed Major Gastrointestinal Bleed All-Cause Mortality
NOAC vs warfarin          
Dabigatran vs warfarin 0.80 (0.70-0.93) 0.38 (0.31-0.47) 1.04 (0.96-1.14) 1.16 (1.06-1.27) 0.73 (0.67-0.80)
Rivaroxban vs warfarin 0.72 (0.63-0.83) 0.65 (0.56-0.77) 1.38 (1.29-1.49) 1.48 (1.36-1.60) 0.81 (0.75-0.88)
Apixaban vs warfarin 0.71 (0.60-0.83) 0.54 (0.43-0.68) 0.51 (0.45-0.58) 0.52 (0.45-0.60) 0.66 (0.60-0.74)
NOAC vs NOAC          
Rivaroxaban vs dabigatran 0.90 (0.76-1.06) 1.71 (1.35-2.17) 1.32 (1.21-1.45) 1.27 (1.16-1.40) 1.12 (1.01-1.24)
Rivaroxaban vs apixaban 1.02 (0.85-1.23) 1.21 (0.94-1.55) 2.70 (2.38-3.05) 2.83 (2.47-3.25) 1.23 (1.09-1.38)
Dabigatran vs apixaban 1.14 (0.94-1.37) 0.70 (0.53-0.94) 2.04 (1.78-2.32) 2.23 (1.93-2.58) 1.10 (0.97-1.24)
NOAC = nonvitamin K antagonist oral anticoagulant.

Supplementary Table 6Adjusted Hazard Ratios (95% Confidence Intervals) for Pairwise Comparisons of Each NOAC vs Warfarin and Each NOAC vs Each Other NOAC for Thromboembolic Stroke, Intracranial Hemorrhage, Major Extracranial (Including Major Gastrointestinal) Bleeding, and All-Cause Mortality

 

Major extracranial bleeding risk was increased with rivaroxaban, decreased with apixaban, and was similar to warfarin for dabigatran. Major gastrointestinal bleeding, which accounted for 82% of major extracranial bleeding, was increased with dabigatran and rivaroxaban and decreased with apixaban, compared with warfarin ( Supplementary Table 6 ). Although the proportional hazards assumption was violated for major extracranial and major gastrointestinal bleeding, a time-dependent analysis met the proportional hazards assumption and generally confirmed results from the primary analysis, with the most important difference being that risks for both bleeding outcomes were significantly increased with dabigatran compared with warfarin only after more than 60 days of treatment ( Supplementary Table 7 , available online).

  Days on Treatment Hazard Ratio (95% Confidence Interval)
Major extracranial bleeding    
NOAC vs warfarin    
Dabigatran vs warfarin 1-30 0.91 (0.79-1.04)
  31-60 1.10 (0.89-1.35)
  61+ 1.16 (1.02-1.32)
Rivaroxaban vs warfarin 1-30 1.18 (1.05-1.33)
  31-60 1.79 (1.51-2.11)
  61+ 1.43 (1.28-1.59)
Apixaban vs warfarin 1-30 0.48 (0.40-0.58)
  31-60 0.62 (0.48-0.81)
  61+ 0.49 (0.40-0.61)
NOAC vs NOAC    
Rivaroxaban vs dabigatran 1-30 1.30 (1.12-1.50)
  31-60 1.63 (1.32-2.01)
  61 1.23 (1.08-1.40)
Rivaroxaban vs apixaban 1-30 2.45 (2.03-2.96)
  31-60 2.87 (2.19-3.76)
  61+ 2.89 (2.34-3.57)
Dabigatran vs apixaban 1-30 1.89 (1.54-2.30)
  31-60 1.76 (1.31-2.37)
  61+ 2.36 (1.89-2.94)
Major gastrointestinal bleeding    
NOAC vs warfarin    
Dabigatran vs warfarin 1-30 0.98 (0.85-1.14)
  31-60 1.19 (0.94-1.49)
  61+ 1.35 (1.17-1.55)
Rivaroxaban vs warfarin 1-30 1.24 (1.09-1.40)
  31-60 1.87 (1.56-2.25)
  61+ 1.57 (1.39-1.77)
Apixaban vs warfarin 1-30 0.50 (0.41-0.61)
  31-60 0.57 (0.42-0.77)
  61+ 0.52 (0.41-0.66)
NOAC vs NOAC    
Rivaroxaban vs dabigatran 1-30 1.26 (1.08-1.47)
  31-60 1.58 (1.26-1.99)
  61 1.16 (1.01-1.34)
Rivaroxaban vs apixaban 1-30 2.50 (2.03-3.07)
  31-60 3.30 (2.42-4.51)
  61+ 3.01 (2.38-3.81)
Dabigatran vs apixaban 1-30 1.99 (1.60-2.47)
  31-60 2.09 (1.49-2.93)
  61+ 2.59 (2.03-3.30)
NOAC = nonvitamin K antagonist oral anticoagulant.

Supplementary Table 7Adjusted Hazard Ratios (95% Confidence Intervals) for Pairwise Comparisons of Each NOAC vs Warfarin and Each NOAC vs Each Other NOAC, for the Outcome of Major Extracranial Bleeding and Major Gastrointestinal Bleeding, By Days on Anticoagulant Therapy

 

The NOACs were similar to each other for thromboembolic stroke, but rivaroxaban was associated with significantly increased risks of intracranial hemorrhage (vs dabigatran: HR 1.71; 95% CI, 1.35-2.17), major extracranial bleeding (vs dabigatran: HR 1.32; 95% CI, 1.21-1.45; vs apixaban: HR 2.70; 95% CI, 2.38-3.05), and death (vs dabigatran: HR 1.12; 95% CI, 1.01-1.24; vs apixaban: HR 1.23; 95% CI, 1.09-1.38) ( Figure 2 , Supplementary Table 6 ). Dabigatran was associated with significantly reduced risk of intracranial hemorrhage (HR 0.70; 95% CI, 0.53-0.94) and with significantly increased risk of major extracranial bleeding (HR 2.04; 95% CI, 1.78-2.32) compared with apixaban. Adjusted Kaplan-Meier plots ( Figure 3 ) and adjusted incidence rate differences ( Supplementary Table 8 , available online) were consistent with the HR estimates.

gr3
 

Figure 3
Adjusted Kaplan-Meier plots for thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality, in cohorts of Medicare beneficiaries treated with warfarin, dabigatran, rivaroxaban, or apixaban for nonvalvular atrial fibrillation. Weighted cohort sizes are shown.

  Thromboembolic Stroke Intracranial Hemorrhage Major Extracranial Bleed Death
NOAC vs warfarin        
Dabigatran vs warfarin −2.2 (−3.5, −0.8) −5.3 (−6.2, −4.3) 1.0 (−1.2, 3.2) −8.2 (−10.3, −6.1)
Rivaroxaban vs warfarin −3.1 (−4.3, −1.9) −3.0 (−4.0, −1.9) 9.6 (7.4, 11.8) −5.9 (−7.9, −3.9)
Apixaban vs warfarin −2.4 (−3.9, −0.9) −3.8 (−4.9, −2.6) −11.3 (−13.3, −9.3) −8.7 (−11.0, −6.4)
NOAC vs NOAC        
Rivaroxaban vs dabigatran −1.0 (−2.3, 0.4) 2.3 (1.3, 3.3) 8.6 (5.9, 11.2) 2.3 (0.1, 4.5)
Rivaroxaban vs apixaban −0.7 (−2.3, 0.8) 0.8 (−0.4, 2.0) 20.9 (18.4, 23.3) 2.8 (0.4, 5.3)
Dabigatran vs apixaban 0.2 (−1.4, 1.9) −1.5 (−2.7, −0.4) 12.3 (9.8, 14.7) 0.5 (−2.0, 3.1)
NOAC = nonvitamin K antagonist oral anticoagulant.

Supplementary Table 8Adjusted Incidence Rate Differences per 1000 Person-Years of Treatment for Each NOAC vs Warfarin and Each NOAC vs Each Other NOAC

 

Rivaroxaban was associated with 8.6 and 20.9 excess cases of major extracranial bleeding, and 2.3 and 2.8 excess deaths per 1000 person-years of use compared with dabigatran or apixaban, respectively ( Supplementary Table 8 ). The adjusted 30-day case fatality rates across all study drugs combined were 11.1% (thromboembolic stroke), 32.6% (intracranial hemorrhage), and 4.2% (major extracranial bleeding).

There were no clinically meaningful differences in HRs across subgroups except that the risk of major extracranial bleeding in dabigatran users was reduced in 65- to 74-year-olds but increased in patients 75 years or older, compared with warfarin ( P interaction < .001) ( Supplementary Figure 1 , available online). Sensitivity analyses were generally consistent with the primary analysis ( Supplementary Figure 2 , available online). The point estimate for thromboembolic stroke for dabigatran compared with apixaban increased slightly and became statistically significant in the 14-day gap analysis (HR 1.23; 95% CI, 1.05-1.45), but the CIs largely overlapped the main analysis. In the analysis restricted to the period since apixaban's approval, intracranial hemorrhage risk with dabigatran was no longer significantly reduced compared with apixaban (HR 0.92; 95% CI, 0.63-1.32). Post hoc analyses were generally consistent with the primary analysis ( Supplementary Figure 3 , available online), except that some comparisons for intracranial hemorrhage changed as to whether differences were statistically significant ( Supplementary Table 9 , available online).

gr4
 

Supplementary Figure 1
Age-subgroup-specific adjusted hazard ratios and 95% confidence intervals for thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality in elderly patients treated with dabigatran, rivaroxaban, or apixaban compared with warfarin for nonvalvular atrial fibrillation.

gr5a
 

Supplementary Figure 2
Hazard ratios (95% confidence intervals) for thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality from sensitivity analyses in elderly patients treated with warfarin, dabigatran, rivaroxaban, or apixaban for nonvalvular atrial fibrillation. The conditions covered by the sensitivity analyses included using a 14-day gap allowance, restriction to patients who initiated anticoagulant therapy after the approval date for apixaban, restriction to patients who received 2 or more study drug prescriptions, multivariable Cox regression, and a crude (unadjusted) analysis. (A) Each nonvitamin K antagonist oral anticoagulant (NOAC) vs warfarin. (B) NOAC vs NOAC.

gr6
 

Supplementary Figure 3
Hazard ratios (95% confidence intervals) for thromboembolic stroke, intracranial hemorrhage, major extracranial bleeding, and all-cause mortality from a post hoc analysis including all eligible warfarin, dabigatran, rivaroxaban, and apixaban patients in an unweighted multivariable Cox proportional hazards regression model. (A) Nonvitamin K antagonist oral anticoagulant (NOAC) vs warfarin. (B) NOAC vs NOAC.

  Intracranial Hemorrhage Extracranial Bleeding Gastrointestinal Bleeding
  Sensitivity Analysis Primary Analysis Sensitivity Analysis Primary Analysis Sensitivity Analysis Primary Analysis
NOAC vs warfarin            
Dabigatran vs warfarin 0.39 (0.30-0.51) 0.38 (0.31-0.47) 1.05 (0.97-1.12) 1.04 (0.96-1.14) 1.23 (1.13-1.32) 1.16 (1.06-1.27)
Rivaroxban vs warfarin 0.73 (0.60-0.88) 0.65 (0.56-0.77) 1.34 (1.26-1.43) 1.38 (1.29-1.49) 1.48 (1.38-1.58) 1.48 (1.36-1.60)
Apixaban vs warfarin 0.42 (0.31-0.59) 0.54 (0.43-0.68) 0.60 (0.54-0.66) 0.51 (0.45-0.58) 0.64 (0.57-0.71) 0.52 (0.45-0.60)
NOAC vs NOAC            
Rivaroxaban vs dabigatran 1.86 (1.38-2.50) 1.71 (1.35-2.17) 1.28 (1.19-1.38) 1.32 (1.21-1.45) 1.21 (1.11-1.31) 1.27 (1.16-1.40)
Rivaroxaban vs apixaban 1.71 (1.21-2.41) 1.21 (0.94-1.55) 2.25 (2.03-2.49) 2.70 (2.38-3.05) 2.32 (2.08-2.60) 2.83 (2.47-3.25)
Dabigatran vs apixaban 0.92 (0.62-1.36) 0.70 (0.53-0.94) 1.75 (1.57-1.95) 2.04 (1.78-2.32) 1.92 (1.71-2.16) 2.23 (1.93-2.58)
NOAC = nonvitamin K antagonist oral anticoagulant.

Supplementary Table 9Adjusted Hazard Ratios (95% Confidence Intervals) for Pairwise Comparisons of Each NOAC vs Warfarin and Each NOAC vs Each Other NOAC from Post Hoc Sensitivity Analyses for Intracranial Hemorrhage (International Classification of Diseases, Ninth Revision Codes 430, 431, 432 Only), All Hospitalized Extracranial Bleeding, and All Hospitalized Gastrointestinal Bleeding, Compared with Results from the Primary Analyses for These Same Outcomes

 

Discussion

NOAC vs Warfarin

Among NOAC users with nonvalvular atrial fibrillation and warfarin users with similar baseline characteristics, each NOAC was associated with statistically significant reductions in thromboembolic stroke, intracranial hemorrhage, and all-cause mortality compared with warfarin. Based on these findings, each NOAC was associated with a more favorable benefit–harm balance than warfarin.

Our HR estimates for intracranial hemorrhage and major extracranial or gastrointestinal bleeding with each NOAC vs warfarin were similar to those reported in the pivotal preapproval clinical trials. 1 2 3 For thromboembolic stroke, we found a reduced HR (protective effect) for dabigatran similar to that reported in its pivotal trial, 1 but for rivaroxaban and apixaban, we found substantial reductions in risk that were not found in their respective trials. 2 3 For all-cause mortality, our HR estimates indicated greater risk reduction with each NOAC than was reported in the pivotal trials, where risk reductions of 11%-15% were observed, but was significant only for apixaban. 1 2 3

Despite close balance between cohorts for the large number of covariates we adjusted for, residual confounding due to channeling of sicker patients to warfarin or unmeasured covariates is possible, and could have contributed to the differences between our study and the pivotal NOAC trials with respect to thromboembolic stroke and all-cause mortality risks. Another possibility is that warfarin control was better in the NOAC clinical trials than in the “real-world” community setting represented by our study. The time in therapeutic range among warfarin-treated patients in the NOAC trials was 62% for apixaban, 64% for dabigatran, and 55% for rivaroxaban, 1 2 3 while a national study of time in therapeutic range in the community setting reported an estimate of 48% among patients treated for <6 months, 24 which could apply to much of our study population. A systematic review of published studies also found that warfarin control was significantly better in randomized trials than in retrospective studies. 25 Suboptimal warfarin control is associated with higher stroke, bleeding, and mortality risks. 26 Poorer quality warfarin control in our study population than in the pivotal trials might explain our findings of NOAC superiority for thromboembolic stroke and all-cause mortality.

NOAC vs NOAC

Among NOACs, the risk of thromboembolic stroke was similar, but rivaroxaban was associated with significantly increased risk of intracranial hemorrhage compared with dabigatran and with significantly increased risks of major extracranial bleeding and death, compared with dabigatran or apixaban. Although the NOACs each have a half-life of about 12 hours, dabigatran and apixaban are dosed twice daily, while rivaroxaban is dosed once daily, which may account for our findings. Concern about once-daily dosing of rivaroxaban was raised during an FDA advisory committee meeting convened prior to rivaroxaban's approval. 27 Based on these results, dabigatran and apixaban were associated with a more favorable benefit–harm profile than rivaroxaban.

Dabigatran and apixaban had similar thromboembolic stroke and all-cause mortality risk. Dabigatran was associated with significantly increased major extracranial bleeding risk but significantly reduced intracranial hemorrhage risk, possibly supporting a more favorable benefit–harm profile for it compared with apixaban. Sensitivity analyses for dabigatran vs apixaban were largely consistent with the primary analysis, but indicated a small change in thromboembolic stroke and intracranial hemorrhage risk that might shift the benefit–harm profile in apixaban's favor. Additional studies are needed to clarify the comparative benefit–harm of these drugs.

Seven observational studies compared our study NOACs against each other for thromboembolic stroke or bleeding risks. 5 6 7 8 9 10 11 These were generally small and underpowered. In 2, intracranial hemorrhage and extracranial bleeding were combined as a single outcome and separate estimates for each were not provided. 5 6 For thromboembolic stroke, one study found no difference in risk, consistent with our findings, 7 and one reported increased risk for rivaroxaban compared with apixaban. 8 Of the 4 studies examining intracranial hemorrhage, risk was increased for rivaroxaban compared with dabigatran in 2, 7 9 and was reduced for dabigatran compared with apixaban in one, 9 in agreement with our study. None of these studies examined extracranial bleeding, but 4 did assess gastrointestinal bleeding risks across NOACs. 8 9 10 11 Rivaroxaban risk was increased in 2 studies compared with dabigatran, 10 11 and was increased in 3 studies compared with apixaban. 8 10 11 Dabigatran risk was increased compared with apixaban in 2 studies. 9 11 These were consistent with our findings for gastrointestinal bleeding ( Supplementary Table 6 ).

Our study had several unique strengths. It was national in scope, and among studies comparing the 3 most widely marketed NOACs, was the largest by a factor of 2.7- to 60-fold, depending on the NOAC and comparator involved. Our study assessed a range of relevant clinical outcomes, including mortality, facilitating assessment of the overall balance of benefits and harms, and it applied a single propensity score model to all 4 study cohorts simultaneously, resulting in a single adjusted population being used for all pairwise comparisons, thus providing HR estimates that can be directly compared with each other.

Our study had several limitations. It was observational and may be subject to confounding by factors not adjusted for in the analysis (eg, quality of blood pressure control) or by residual channeling bias. Also, the mean duration of continuous anticoagulant use was <5 months. Despite this, we had a larger number of patients still on therapy at 8 months than the number starting NOAC therapy in most other studies comparing these NOACs to each other. 5 6 7 8 9 10 11 Our study was restricted to patients aged 65 years or older, the age-group accounting for over 80% of atrial fibrillation patients. 28 The comparative effects of NOACs could be different in younger-aged populations. We studied only initiators of standard-dose NOACs. Comparative effects may differ in patients treated with the lower dose. Because we excluded warfarin users less likely to be treated with an NOAC, our results may not be generalizable to them. However, a post hoc analysis including all warfarin users yielded results largely consistent with the primary analysis. Finally, our study examined first-time users of an anticoagulant for stroke prevention in nonvalvular atrial fibrillation. Results could differ in patients switching from warfarin to an NOAC.

Supplementary Material

Supplementary material accompanying this article can be found in the online version at https://doi.org/10.1016/j.amjmed.2018.12.023

Appendix

Methods description of post hoc analyses that were performed:

  • To assess whether exclusion of warfarin users during the propensity score-matching step affected the generalizability of hazard ratio estimates for the warfarin vs NOAC comparisons, we performed unweighted multivariable Cox proportional hazards regression for each study outcome based on all eligible warfarin and NOAC users. We also conducted post hoc sensitivity analyses of our outcome definitions by restricting the definition of intracranial hemorrhage to validated International Classification of Diseases, Ninth Revision codes only, 18 19 20 and by expanding the definition of extracranial bleeding to include all hospitalized extracranial bleeding events.

 

Funding: The US Food and Drug Administration , through an interagency agreement between it and the Centers for Medicare & Medicaid Services.

Conflict of Interest: None of the authors have any conflicts of interest to declare.

Authorship: All authors had access to the data and a role in writing the manuscript. The views expressed are the authors’ and not necessarily those of the Centers for Medicare & Medicaid Services, the Food and Drug Administration, or the Department of Health and Human Services.

References

  • 1 SJ Connolly, MD Ezekowitz, S Yusuf, et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med. 2009;161:1139-1151
  • 2 MR Patel, KW Mahaffey, J Garg, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med. 2011;365:883-891
  • 3 CB Granger, JH Alexander, JJV McMurray, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med. 2011;365:981-992
  • 4 RP Giugliano, CT Ruff, E Braunwald, et al. Edoxaban versus warfarinin patients with atrial fibrillation. N Engl J Med. 2013;369:2093-2104
  • 5 GY Lip, X Pan, S Kamble, et al. Major bleeding risk among non-valvular atrial fibrillation patients initiated on apixaban, dabigatran, rivaroxaban or warfarin: a "real-world" observational study in the United States. Int J Clin Pract. 2016;70(9):752-763
  • 6 GYH Lip, A Keshishian, S Kamble, et al. Real-world comparison of major bleeding risk among non-valvular atrial fibrillation patients initiated on apixaban, dabigatran, rivaroxaban, or warfarin. A propensity score matched analysis. Thromb Haemost. 2016;116(5):975-986
  • 7 PA Noseworthy, X Yao, NS Abraham, LR Sangaralingham, RD McBane, ND Shah. Direct comparison of dabigatran, rivaroxaban, and apixaban for effectiveness and safety in nonvalvular atrial fibrillation. Chest. 2016;150(6):1302-1310
  • 8 S Deitelzweig, X Luo, K Gupta, et al. Comparison of effectiveness and safety of treatment with apixaban vs other oral anticoagulants among elderly nonvalvular atrial fibrillation patients. Curr Med Res Opin. 2017;33:1745-1754
  • 9 G Adeboyeje, G Sylwestrzak, JJ Barron, et al. Major bleeding risk during anticoagulation with warfarin, dabigatran, apixaban, or rivaroxaban in patient s with nonvalvular atrial fibrillation. J Manag Care Spec Pharm. 2017;23:968-978
  • 10 I Hernandez, Y Zhang, S Saba. Comparison of the effectiveness and safety of apixaban, dabigatran, rivaroxaban, and warfarin in newly diagnosed atrial fibrillation. Am J Cardiol. 2017;120:1813-1819
  • 11 NS Abraham, PA Noseworthy, X Yao, LR Sangaralingham, ND Shah. Gastrointestinal safety of direct oral anticoagulants: a large population-based study. Gastroenterology. 2017;152:1014-1022
  • 12 WA Ray. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol. 2003;158:915-920
  • 13 GYH Lip, R Nieuwlaat, R Pisters, DA Lane, JGM Crijins. Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the Euro Heart Survey on Atrial Fibrillation. Chest. 2010;137:263-272
  • 14 R Pisters, DA Lane, R Nieuwlaat, CB de Vos, HJ Crijns, GY Lip. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest. 2010;138(5):1093-1100
  • 15 EA Stuart. Matching methods for causal inference: a review and a look forward. Stat Sci. 2010;25:1-21
  • 16 GW Imbens. The role of the propensity score in estimating dose-response functions. Biometrika. 2000;87(3):706-710
  • 17 PC Austin. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28:3083-3107
  • 18 DL Tirschwell, WT Longstreth. Validating administrative data in stroke research. Stroke. 2002;33(10):2465-2470
  • 19 CL Roumie, E Mitchel, PS Gideon, C Varas-Lorenzo, J Castellsague, MR Griffin. Validation of ICD-9 codes with a high positive predictive value for incident stroke resulting in hospitalization using Medicaid health data. Pharmacoepidemiol Drug Saf. 2008;17(1):20-26
  • 20 RA Kokotaila, MD Hill. Coding of stroke and stroke risk factors using, , revision 9 and 10. Stroke. 2005;36(8):1776-1781
  • 21 AW Cunningham, CM Stein, CP Chung, JR Daugherty, WE Smalley, WA Ray. An automated database definition for serious bleeding due to oral anticoagulant use. Pharmacoepidemiol Drug Saf. 2011;20:560-566
  • 22 Hill ME, Rosenwaike I. The Social Security Administration's Death Master File: the completeness of death reporting at older ages. Soc Secur Bull . 2001-2002;64(1):45-51. Available at: <http://www.socialsecurity.gov/policy/docs/ssb/v64n1/v64n1p45.pdf>. Accessed June 14, 2018.
  • 23 J Xie, C Liu. Adjusted Kaplan-Meier estimator and log-rank test with inverse probability of treatment weighting for survival data. Stat Med. 2005;24:3089-3110
  • 24 JS Dlott, RA George, X Huang, et al. National assessment of warfarin anticoagulation therapy for stroke prevention in atrial fibrillation. Circulation. 2014;129:1407-1414
  • 25 Y Wan, C Heneghan, R Perera, et al. Anticoagulation control and prediction of adverse events in patients with atrial fibrillation: a systematic review. Circ Cardiovasc Qual Outcomes. 2008;1:84-91
  • 26 F Björck, J Renlund, GYH Lip, P Wester, PJ Svensson, A Själander. Outcomes in a warfarin-treated population with atrial fibrillation. JAMA Cardiol. 2016;1:172-180
  • 27 US Food and Drug Administration. 2011 Meeting materials, Cardiovascular and Renal Drugs Advisory Committee. Available at: <https://wayback.archive-it.org/7993/20170403223812/https://www.fda.gov/AdvisoryCommittees/CommitteesMeetingMaterials/Drugs/CardiovascularandRenalDrugsAdvisoryCommittee/ucm250287.htm>. Accessed June 14, 2018.
  • 28 AS Go, EM Hylek, KA Phillips, et al. Prevalence of diagnosed atrial fibrillation in adults: national implications for rhythm management and stroke prevention: the AnTicoagulation and Risk factors In Atrial fibrillation (ATRIA) study. JAMA. 2001;285:2370-2375

© 2019 Elsevier Inc. All rights reserved.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds or experiments described herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made. To the fullest extent of the law, no responsibility is assumed by Elsevier for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

This e-print is distributed with the support of BMS.  

Reproduced by:
Elsevier España, S.L.U.
(A member of Elsevier)
Av. Josep Tarradellas, 20-30
08029 Barcelona
Tel.: 932 000 711
Fax: 932 091 136