Skip main navigation

Comparative Effectiveness and Safety of Direct Oral Anticoagulants Versus Warfarin Among Adults With Cancer and Atrial Fibrillation

Originally publishedhttps://doi.org/10.1161/CIRCOUTCOMES.122.008951Circulation: Cardiovascular Quality and Outcomes. 2022;15

Abstract

Background:

While clinical guidelines recommend direct-acting oral anticoagulants (DOAC) over warfarin to treat isolated nonvalvular atrial fibrillation, guidelines are silent regarding nonvalvular atrial fibrillation treatment among individuals with cancer, reflecting the paucity of evidence in this setting. We quantified relative risk of ischemic stroke or systemic embolism and major bleeding (primary outcomes), and all-cause and cardiovascular death (secondary outcomes) among older individuals with cancer and nonvalvular atrial fibrillation comparing DOACs and warfarin.

Methods:

This retrospective cohort study used Surveillance, Epidemiology, and End Results cancer registry and linked US Medicare data from 2010 through 2016, and included individuals diagnosed with cancer and nonvalvular atrial fibrillation who newly initiated DOAC or warfarin. We used inverse probability of treatment weighting to control confounding. We used competing risk regression for primary outcomes and cardiovascular death, and Cox proportional hazard regression for all-cause death.

Results:

Among 7675 individuals included in the cohort, 4244 (55.3%) received DOACs and 3431 (44.7%) warfarin. In the inverse probability of treatment weighting analysis, there was no statistically significant difference among DOAC and warfarin users in the risk of ischemic stroke or systemic embolism (1.24 versus 1.19 events per 100 person-years, adjusted hazard ratio 1.41 [95% CI, 0.92–2.14]), major bleeding (3.08 versus 4.49 events per 100 person-years, adjusted hazard ratio 0.90 [95% CI, 0.70–1.17]), and cardiovascular death (1.88 versus 3.14 per 100 person-years, adjusted hazard ratio 0.82 [95% CI, 0.59–0.1.13]). DOAC users had significantly lower risk of all-cause death (7.09 versus 13.3 per 100 person-years, adjusted hazard ratio 0.81 [95% CI, 0.69–0.94]) compared to warfarin users.

Conclusions:

Older adults with cancer and atrial fibrillation exposed to DOACs had similar risks of stroke and systemic embolism and major bleeding as those exposed to warfarin. Relative to warfarin, DOAC use was associated with a similar risk of cardiovascular death and a lower risk of all-cause death.

What is Known

  • Individuals with cancer and atrial fibrillation are at greater risk of ischemic stroke or systemic embolism and major bleeding than the general population with atrial fibrillation.

  • Exploratory analyses of 3 clinical trials were limited in their ability to determine the safety and effectiveness of direct-acting oral anticoagulants versus warfarin among older adults with cancer and nonvalvular atrial fibrillation.

What the Study Adds

  • Among individuals with cancer and nonvalvular atrial fibrillation, direct-acting oral anticoagulants had similar risks of ischemic stroke or systemic embolism and major bleeding compared to warfarin.

  • Direct-acting oral anticoagulants were associated with lower risk of all-cause death compared to warfarin.

Atrial fibrillation affects between 2.7 and 6.1 million adults in the US and the prevalence is expected to double by 2030.1–3 Cancer is an independent risk factor for the development of atrial fibrillation.4–7 Due to increasing cancer incidence and rising life expectancy of individuals with cancer, the incidence of atrial fibrillation among individuals with cancer will continue to increase.8,9 Among individuals with atrial fibrillation, those with cancer are at 2-fold increased risk of stroke than those without due to hypercoagulability and noninfective endocarditis.10,11 Also, individuals with cancer are at a higher risk of bleeding than the general population due to cancer- and treatment-related thrombocytopenia and disseminated intravascular coagulation.12,13

While clinical guidelines recommend direct-acting oral anticoagulants (DOAC) over warfarin to treat isolated nonvalvular atrial fibrillation (NVAF), guidelines are silent regarding NVAF treatment among individuals with cancer, reflecting the paucity of evidence in this setting.14,15 Although exploratory analyses of clinical trials tried to determine the safety and effectiveness of DOAC versus warfarin in this population, these trial findings were limited by small sample sizes, heterogenous cancer diagnoses, and an absence of key cancer-related characteristics such as stage, grade, size, and time of cancer diagnosis.16–18 Three single-institutional observational studies found that DOAC are safe and effective in patients with cancer and NVAF, but lacked an active control group, preventing any direct comparisons.19–21 Results from 2 real-world studies have provided contradictory evidence. Shah et al22 found similar risks of ischemic stroke and major bleeding comparing DOAC and warfarin users, although apixaban users had a lower risk of bleeding; whereas Kim et al23 found lower rates of ischemic stroke or systemic embolism, major bleeding, and all-cause death among DOAC users compared to warfarin users.

We used the Surveillance Epidemiology and End Results (SEER) cancer registry and linked Medicare claims data to compare the effectiveness and safety of DOAC versus warfarin for the risk of ischemic stroke or systemic embolism, major bleeding, all-cause death, and cardiovascular death among a large, diverse population of older adults with cancer and NVAF. We hypothesized that individuals using DOACs have lower risk of ischemic stroke, systemic embolism and bleeding, and similar risk of all-cause and cardiovascular death, compared to warfarin.

Methods

Data Source

We used SEER linked to Medicare claims data from 2009 through 2016. The SEER-Medicare data is a linkage of 2 large population-based sources of data that provides detailed information about Medicare beneficiaries with cancer.24 The SEER data collects information on cancer incidence, treatment, and survival from 21 cancer registries and covers approximately 35% of the US population. This has been linked by the National Cancer Institute, the SEER registries, and the Centers for Medicare and Medicaid to the Medicare data, which are the healthcare claims for all Medicare beneficiaries.

The Johns Hopkins Bloomberg School of Public Health Institutional Review Board approved this study. The data that support the findings of this study are available from the National Cancer Institute, and additional materials including SAS codes can be provided by the corresponding author for the purposes of reproducible results upon reasonable request.

Study Design and Population

In this new-user retrospective cohort study, we included adults aged 66 years or older diagnosed with cancer who used at least 1 oral anticoagulant from January 1, 2010 through December 31, 2016 (Figure S1). We included individuals with breast, bladder, colorectal, esophagus, lung, ovary, kidney, pancreas, prostate, stomach, and uterus cancer because they are common and have a high prevalence of atrial fibrillation.6,25 The first prescription of an oral anticoagulant was considered the individual’s index date. We excluded the following individuals: (1) those who were not continuously enrolled in Medicare parts A, B, and D or enrolled in a health maintenance organization within the 12 months prior to the index date; (2) had no diagnosis code for atrial fibrillation during the 12 months prior to the index date; (3) had a diagnosis code for mitral or aortic disease, heart valve surgery, or mitral/aortic valve surgery; (4) received oral anticoagulants within 12 months prior to the index date; (5) received both a DOAC and warfarin on the index date; and (6) were receiving hospice care on the index date. We identified atrial fibrillation based on the presence of at least 1 inpatient or 2 outpatient claims in 1 year prior to the index date, a definition with a sensitivity of 79% and positive predictive value of 89% (Table S1).26,27

Exposure to Oral Anticoagulants

We used prescription fill information from the Medicare Part D data to define exposure to DOACs (apixaban, dabigatran, rivaroxaban, or edoxaban) and warfarin. We included all possible doses of DOAC and warfarin for defining drug exposure. Based on the first oral anticoagulant prescription, we considered patients exposed to DOAC or to warfarin. In the primary analysis, we defined drug exposure using an as-treated approach where individuals were censored when they were no longer exposed to their initial drug class.

Outcomes

The primary outcomes were ischemic stroke or systemic embolism, and major bleeding (Table S1). Secondary outcomes were all-cause death and deaths due to cardiovascular diseases. We identified patients hospitalized for ischemic stroke or systemic embolism as the primary discharge diagnosis in the inpatient claims. These codes have positive predictive value >80% for ischemic stroke definition.28,29 Major bleeding included hospitalization due to intracranial bleeding, gastrointestinal bleeding, or nongastrointestinal bleeding based on the primary discharge diagnosis in any inpatient claims. The positive predictive values are 75.5% for intracranial bleeding and 91.8% for major gastrointestinal and nongastrointestinal bleeding.30 We used SEER cause of death and identified cardiovascular death if individuals died from heart diseases, hypertension, cerebrovascular disease, atherosclerosis, aortic aneurysm, or other diseases of arteries, arterioles, or capillaries. We included hospitalization due to sepsis as a negative control outcome to assess the influence of unmeasured confounding.31,32

Covariates

We identified covariates for adjustment using information from the 12 months prior to the index date (Table S2). We used SEER data to identify age, sex, race, median household income, Medicaid eligibility, residential census region, and cancer-related characteristics such as cancer site, stage, grade, tumor size, and the use of chemotherapy. Active cancer was defined as receipt of chemotherapy, radiation, or cancer surgery in 12 months prior to the oral anticoagulation initiation. We operationalized 20 comorbidities, 18 measures of medication use, a set of frailty indicators, and prior healthcare use from the Medicare claims as potential confounders.14,22,23 Among these were the diagnoses required to calculate CHA₂DS₂-VASc and HAS-BLED scores, which we operationalized as individual covariates.33,34 Healthcare utilization was quantified by the count of hospital admissions and physician visits.

Statistical Analyses

We calculated a propensity score with logistic regression to predict DOAC versus warfarin use conditional on individual’s sociodemographic, cancer, and clinical characteristics. We then generated stabilized inverse probability of treatment weights (IPTW), which we trimmed at the 1st and 99th percentiles.35 We evaluated the balance of covariates between groups in the original and IPTW weighted samples, where a standardized difference of <0.10 suggests acceptable balance in covariates across groups.36 We separately modeled the outcomes of ischemic stroke or systemic embolism, major bleeding, and cardiovascular death using Fine and Gray competing risk regression.37–39 Individuals were censored when they discontinued their index drug class (gap of >30 days between 2 prescriptions), switched to a different anticoagulant drug class, lost fee-for-service Medicare coverage, were admitted to hospice, or concluded the study period. We considered all-cause death to be a competing risk for ischemic stroke or systemic embolism and major bleeding, and noncardiovascular death to be a competing risk for cardiovascular death. We used a multivariable Cox proportional hazards regression model for all-cause death. The IPTW-weighted competing risk regression and Cox regression included calendar year of drug initiation, time from cancer diagnosis to oral anticoagulant initiation, and any unbalanced covariates. We also used Fine and Gray competing risk regression model for the negative control outcome of sepsis, expecting a null association. We reported adjusted hazard ratios (aHR) and 95% CIs for each outcome of interest.

Subgroup and Sensitivity Analyses

First, we interacted treatment with age (<75 years and ≥75 years), sex, active cancer, and cancer type, each in separate regression models. In sensitivity analyses, we compared each unique DOAC to warfarin. Second, we performed intention-to-treat analysis in which individuals were analyzed in their index drug exposure group and not censored at the time of switching or discontinuation. Third, we used Cox proportional hazards regression model for ischemic stroke or systemic embolism, bleeding, and sepsis, with censoring upon death. Fourth, to evaluate for a dose-response relationship, we stratified by duration of oral anticoagulant use (<90 days, 91–180 days, and >180 days) in the as-treated analysis. We performed quantitative bias assessment by calculating an E-value, which represents the strength of association that an unmeasured confounder needs to have with both the exposure and outcome variables, conditional on the measured covariates, to change the interpretation of our findings.40

We conducted all analyses using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Study Cohort

We identified 105 923 individuals diagnosed with primary cancer who filled at least 1 prescription of oral anticoagulants from 2010 through 2016. After applying exclusion criteria, the final cohort included 7675 individuals diagnosed with cancer and atrial fibrillation who newly initiated DOAC or warfarin (Figure). Overall, 4244 (55.3%) individuals received DOAC and 3431 (44.7%) warfarin. The use of DOAC increased over time, from 2% in 2010 through 76.2% in 2016 (Figure S2). The mean (SD) age of the cohort was 76.6 (6.9) years, 48.1% were female, and 88.0% were non-Hispanic white. Individuals diagnosed with prostate (22.2%), breast (19.6%), lung (19.3%), and colorectal (14.5%) cancers accounted for approximately three-fourths of the cohort.

Figure.

Figure. Cohort derivation.aFirst prescription of oral anticoagulants is considered as an index date. DOAC indicates direct-acting oral anticoagulants; HMO, health maintenance organization; and NVAF, nonvalvular atrial fibrillation.

Balance of Patient Characteristics

DOAC users differed significantly from warfarin users by sociodemographic, cancer diagnoses, and clinical characteristics (Table 1 and Table S3). After IPTW, both groups were well balanced, with all standardized differences below −0.022 (or 2.2%). There was a considerable overlap in the propensity score distribution among warfarin and DOAC users after IPTW (Figure S3).

Table 1. Baseline Patient Characteristics Among Warfarin and DOAC Users, Before, and After IPTW*

Patient characteristicsOriginal sampleIPTW sample
Warfarin (n=3431)DOAC (n=4244)SMDWarfarin (n=3407)DOAC (n=4200)SMD
Sociodemographic
 Age, mean±SD77.1±7.176.5±6.8−0.09176.7±7.076.7±6.9−0.002
Sex
 Male1727 (50.3)2253 (53.1)1774 (52.1)2183 (52.0)
 Female1704 (49.7)1991 (46.9)−0.0551633 (47.9)2017 (48.0)0.002
Race
 White3005 (87.6)3752 (88.4)0.0252991 (87.8)3687 (87.8)0.00004
 Black238 (6.9)34 (5.5)−0.059217 (6.4)267 (6.4)−0.0006
 Hispanic51 (1.5)41 (1.0)−0.04744 (1.3)53 (1.3)−0.002
 Other137 (4.0)217 (5.1)0.054155 (4.6)193 (4.6)0.002
Marital status
 Married1742 (50.8)2286 (53.9)0.0621779 (52.2)2209 (52.6)0.008
 Other1689 (49.2)1958 (46.1)1628 (47.8)1991 (47.4)
Medicaid eligibility842 (24.5)828 (19.5)−0.122747 (21.9)917 (21.8)0.008
Census region
 West1282 (37.4)1639 (38.6)0.0261286 (37.7)1596 (38.0)0.006
 Northeast737 (21.5)997 (23.5)0.048767 (22.5)956 (22.8)0.006
 Midwest576 (16.8)459 (10.8)−0.174463 (13.6)559 (13.3)−0.008
 South814 (23.7)1104 (26.0)0.053863 (25.3)1053 (25.1)−0.006
 Missing22 (0.6)45 (1.1)0.04628 (0.8)36 (0.9)0.003
Median household income per capita
 1st Quartile909 (26.5)955 (22.5)−0.093834 (24.5)1015 (24.2)−0.007
 2nd Quartile930 (27.1)936 (22.1)−0.118839 (24.6)1026 (24.4)−0.005
 3rd Quartile812 (23.7)1051 (24.8)0.026837 (24.6)1026 (24.4)−0.003
 4th Quartile698 (20.3)1166 (27.5)0.168801 (23.5)1015 (24.2)0.015
 Missing82 (2.4)136 (3.2)0.04996 (2.8)119 (2.8)0.0007
Cancer
Cancer site
 Bladder294 (8.6)365 (8.6)0.001296 (8.7)363 (8.7)−0.002
 Breast600 (17.5)906 (21.4)0.098660 (19.4)831 (19.8)0.010
 Colorectal593 (17.3)521 (12.3)−0.141491 (14.4)588 (14.0)−0.012
 Esophagus53 (1.5)58 (1.4)−0.01546 (1.3)56 (1.3)−0.0006
 Kidney170 (5.0)204 (4.8)−0.007165 (4.9)205 (4.9)0.002
 Lung770 (22.4)712 (16.8)−0.143667 (19.6)811 (19.3)−0.007
 Ovary79 (2.3)57 (1.3)−0.07258 (1.7)70 (1.7)−0.003
 Pancreas97 (2.8)70 (1.7)−0.08077 (2.3)95 (2.3)0.0008
 Prostate591 (17.2)1112 (26.2)0.219752 (22.1)939 (22.4)0.007
 Stomach53 (1.5)68 (1.6)0.00553 (1.6)66 (1.6)0.002
 Uterus131 (3.8)171 (4.0)0.011142 (4.2)175 (4.2)−0.00002
Cancer stage
 Stage 0295 (8.6)416 (9.8)0.042314 (9.2)387 (9.2)0.00006
 Stage I1058 (30.8)1504 (35.4)0.0981144 (33.6)1416 (33.7)0.003
 Stage II841 (24.5)1195 (28.2)0.083889 (26.1)1109 (26.4)0.007
 Stage III551 (16.1)549 (12.9)−0.089488 (14.3)587 (14.0)−0.010
 Stage IV502 (14.6)383 (9.0)−0.174402 (11.8)489 (11.6)−0.005
 Unknown184 (5.4)197 (4.6)−0.033171 (5.0)212 (5.0)0.001
Grade
 Grade I386 (11.3)553 (13.0)0.055411 (12.1)510 (12.1)0.002
 Grade II1192 (34.7)1604 (37.8)0.0641250 (36.7)1539 (36.6)−0.001
 Grade III949 (27.7)1184 (27.9)0.005945 (27.7)1163 (27.7)−0.001
 Grade IV172 (5.0)173 (4.1)−0.045153 (4.5)190 (4.5)0.002
 Cell type not determined732 (21.3)730 (17.2)−0.105648 (19.0)798 (19.0)−0.001
Tumor size
 0 cm20 (0.68)26 (0.6)0.00420 (0.6)24.1 (0.6)−0.001
 0.1–1 cm239 (7.0)402 (9.5)0.091272 (8.0)351 (8.4)0.014
 1.1–2 cm503 (14.7)627 (14.8)0.003502 (14.8)617 (14.7)−0.002
 2.1–3 cm427 (12.5)492 (11.6)−0.026417 (12.3)509 (12.1)−0.004
 3.1–4 cm355 (10.4)375 (8.8)−0.051326 (9.6)400 (9.5)−0.002
 4.1–5 cm263 (7.7)279 (6.6)−0.042238 (7.0)289 (6.9)−0.004
 >5 cm592 (17.3)526 (12.4)−0.137502 (14.7)603 (14.4)−0.011
 Not available/site-specific codes1032 (30.1)1517 (35.7)0.1211129 (33.1)1407 (33.5)0.007
Receipt of chemotherapy1000 (29.2)1193 (28.1)−0.023981 (28.8)1203 (28.7)−0.003
Active cancer1272 (37.1)1476 (34.8)−0.0481253 (36.8)1484 (35.3)−0.030
Comorbidities*
 Anemia2030 (59.2)1969 (46.4)−0.2581779 (52.2)2183 (52.0)−0.005
 Dementia274 (8.0)260 (6.1)−0.073235 (6.9)288 (6.9)−0.002
 Liver disease600 (17.5)567 (13.4)−0.114528 (15.5)642 (15.3)−0.006
 Peptic ulcer disease169 (4.9)129 (3.0)−0.097132 (3.9)157 (3.7)−0.008
 Prior major bleeding or predisposition to bleeding2418 (70.5)2568 (60.5)−0.2112230 (65.5)2728 (65.0)−0.010
 Renal disease1284 (37.4)1243 (29.3)−0.1731136 (33.3)1388 (33.0)−0.006
Stroke, TIA, thromboembolism history1742 (50.8)1393 (32.8)−0.3701401 (41.1)1686 (40.2)−0.020
Vascular disease history1471 (42.9)1478 (34.8)−0.1661316 (38.6)1609 (38.3)−0.006
Co-medication use*
 Antiplatelets850 (24.8)1031 (24.3)−0.011853 (25.0)1038 (24.7)−0.008
 Heparin936 (27.3)428 (10.1)−0.452609 (17.9)715 (17.0)−0.022

DOAC indicates direct-acting oral anticoagulants; IPTW, inverse probability of treatment weights; SMD, standardized mean difference; and TIA, transient ischemic attack.

*Comorbidities, frailty indicator, co-medications and health service use were used as covariates in logistic regression model to derive IPTW weights. The balance of these covariates before and after IPTW weighting are reported in Table S3.

Risk of Ischemic Stroke or Systemic Embolism, Major Bleeding, and Mortality

The cumulative incidence of ischemic stroke or systemic embolism was not significantly different among DOAC and warfarin users (Figure S4A). Over a median follow-up of 7.7 months (interquartile range, 3.1–16.7 months), the rate of ischemic stroke or systemic embolism per 100 person years among DOAC and warfarin users were 1.24 and 1.19, respectively. In the IPTW weighted analysis, DOAC use was not associated with a significantly greater risk of ischemic stroke or systemic embolism (aHR 1.41 [95% CI, 0.92–2.14]) compared to warfarin use (Table 2). Over a median follow-up of 7.4 months (interquartile range, 3.0–16.2 months), the rate or major bleeding per 100 person years among DOAC users and warfarin users was 3.08 and 4.49, respectively (Figure S4B and Table S4). In the IPTW weighted analysis, there was not a significant difference in rates of bleeding (aHR 0.90 [95% CI, 0.70–1.17]; Table 2). The E-values were 2.17 for ischemic stroke or systemic embolism and 1.46 for major bleeding.

Table 2. Adjusted Association of DOAC With Outcomes Relative to Warfarin

WarfarinDOAC
Number of patients34314244
Ischemic stroke or systemic embolism (primary outcome)
 Events, n4252
 Total person time follow up3525.394178.02
 Events per 100 patient years, n (95% CI)1.19 (0.86–1.61)1.24 (0.93–1.63)
 Unadjusted HR (95% CI)Reference1.11 (0.74–1.66)
 IPTW adjusted HR (95% CI)Reference1.41 (0.92–2.14)
 E-value for IPTW adjusted HR2.17
Major bleeding (primary outcome)
 Events, n153126
 Total person time follow up3406.554088.97
 Events per 100 patient years, n (95% CI)4.49 (3.81–5.26)3.08 (2.57–3.67)
 Unadjusted HR (95% CI)Reference0.69 (0.55–0.88)
 IPTW adjusted HR (95% CI)Reference0.90 (0.70–1.17)
 E-value for IPTW adjusted HR1.46
All-cause death (secondary outcome)
 Events, n471298
 Total person time follow up3539.034201.64
 Events per 100 patient years, n (95% CI)13.3 (12.1–14.6)7.09 (6.31–7.95)
 Unadjusted HR (95% CI)Reference0.51 (0.44–0.59)
 IPTW adjusted HR (95% CI)Reference0.81 (0.69–0.94)
 E-value for IPTW adjusted HR1.77
Cardiovascular death (secondary outcome)
 Events, n11179
 Total person time follow up3539.034201.64
 Events per 100 patient years, n (95% CI)3.14 (2.18–3.78)1.88 (1.49–2.34)
 Unadjusted HR (95% CI)Reference0.61 (0.46–0.81)
 IPTW adjusted HR (95% CI)Reference0.82 (0.59–1.13)
 E-value for IPTW adjusted HR1.74
Sepsis (negative control outcome)
 Events, n70105
 Total person time follow up3517.814171.78
 Events per 100 patient years, n (95% CI)1.99 (1.55–2.51)2.52 (2.06–3.05)
 Unadjusted HR (95% CI)Reference1.34 (0.99–1.81)
 IPTW adjusted HR (95% CI)Reference1.07 (0.79–1.46)
 E-value for IPTW adjusted HR1.34

DOAC indicates direct-acting oral anticoagulants; HR, hazard ratio; and IPTW, inverse probability of treatment weighting.

In the IPTW weighted analysis, DOAC use was associated with fewer all-cause deaths relative to warfarin use (aHR 0.81 [95% CI, 0.69–0.94]), with E-value of 1.77 (Figure S5 and Table 2). Overall, 190 individuals died due to cardiovascular diseases (Table S5). In the IPTW weighted analysis, DOAC use was not associated with reduced risk of cardiovascular deaths relative to warfarin (aHR 0.82 [95% CI, 0.59–1.13]), with E-value of 1.74 (Figure S6 and Table 2). We did not observe any statistically significant difference between DOAC use and warfarin use and the risk of sepsis (aHR 1.07 [95% CI, 0.79–1.46]).

Subgroup and Sensitivity Analyses

In subgroup analyses by age, sex, active cancer, and cancer sites, the results for ischemic stroke or systemic embolism and major bleeding were similar to the main analysis indicating little heterogeneity in the association between the exposure and the outcome across subsets of patients. We noted all-cause survival benefit in some subgroups. However, when we evaluated cardiovascular death, benefit favoring DOAC was only seen among individuals with no active cancer (Table S6), which was not seen in the comparison group.

In analyses by individual DOAC, apixaban, dabigatran, and rivaroxaban demonstrated no significantly increased risk for ischemic stroke or systemic embolism compared to warfarin. The risk of bleeding was similar for dabigatran and rivaroxaban, but lower for apixaban, compared to warfarin. All 3 products were associated with a lower risk of all-cause death compared to warfarin, but only rivaroxaban was associated with a lower risk of cardiovascular death (Table 3). Our findings were consistent in 3 sensitivity analyses where we evaluated intent-to-treat drug exposure, employed Cox regression models, and stratified drug exposure by duration of oral anticoagulant use (Table S7).

Table 3. Adjusted Association of Individual DOAC Drug With Outcomes Relative to Warfarin*

WarfarinApixabanDabigatranRivaroxaban
Number of patients343118305261874
Ischemic stroke or systemic embolism (primary outcome)
 Total person time follow up3525.391357.01730.252081.99
 Events per 100 patient years, n (95% CI)1.19 (0.86–1.61)1.25 (0.73–2.01)1.23 (0.56–2.34)1.25 (0.82–1.83)
 Unadjusted HR (95% CI)Reference1.09 (0.62–1.91)1.10 (0.54–2.25)1.14 (0.70–1.86)
 IPTW adjusted HR (95% CI)Reference1.22 (0.65–2.28)1.19 (0.56–2.51)1.14 (0.66–1.98)
 E-value for IPTW adjusted HR1.741.671.54
Major bleeding (primary outcome)
 Total person time follow up3406.551346.42697.192036.6
 Events per 100 patient years, n (95% CI)4.49 (3.81–5.26)2.45 (1.69–3.44)3.87 (2.55–5.64)3.19 (2.46–4.07)
 Unadjusted HR (95% CI)Reference0.48 (0.33–0.70)1.03 (0.68–1.55)0.76 (0.57–1.02)
 IPTW adjusted HR (95% CI)Reference0.57 (0.38–0.86)0.84 (0.52–1.36)0.82 (0.60–1.12)
 E-value for IPTW adjusted HR2.901.671.74
All-cause death (secondary outcome)
 Total person time follow up3539.031360.79738.832093.24
 Events per 100 patient years, n (95% CI)13.3 (12.1–14.6)8.23 (6.78–9.90)5.41 (3.87–7.37)6.93 (5.85–8.15)
 Unadjusted HR (95% CI)Reference0.52 (0.42–0.64)0.45 (0.33–0.62)0.52 (0.43–0.63)
 IPTW adjusted HR (95% CI)Reference0.70 (0.56–0.87)0.68 (0.51–0.90)0.67 (0.55–0.81)
 E-value for IPTW adjusted HR2.212.302.35
Cardiovascular death (secondary outcome)
 Total person time follow up3539.031360.79738.832093.24
 Events per 100 patient years, n (95% CI)3.14 (2.58–3.78)2.35 (1.61–3.32)2.17 (1.24–3.52)1.48 (1.01–2.10)
 Unadjusted HR (95% CI)Reference0.69 (0.46–1.03)0.80 (0.47–1.34)0.49 (0.33–0.73)
 IPTW adjusted HR (95% CI)Reference0.86 (0.53–1.40)1.10 (0.68–1.77)0.62 (0.40–0.95)
 E-value for IPTW adjusted HR1.601.432.61

DOAC indicates direct-acting oral anticoagulants; HR, hazard ratio; and IPTW, inverse probability of treatment weighting.

*Very few people received edoxaban. Therefore, we did not compare edoxaban vs warfarin.

Discussion

In this retrospective cohort study of a large, diverse group of US Medicare beneficiaries with cancer and NVAF, although we hypothesized lower risk of stroke or systemic embolism, we did not find statistically significant difference among DOAC versus warfarin users in the risk of ischemic stroke or systemic embolism. We also did not find significant differences for major bleeding among DOAC versus warfarin users, except for apixaban, which showed reduced risk. DOACs were associated with lower risk of all-cause mortality compared to warfarin. These findings persisted in subgroup and several sensitivity analyses.

Our findings add to a growing evidence base regarding the safety and effectiveness of DOACs among individuals with cancer and NVAF. Analyses of ROCKET-AF (Rivaroxaban Once Daily Oral Direct Factor Xa Inhibition Compared with Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation)16 the ARISTOTLE (Apixaban for Reduction in Stroke and Other Thromboembolic Events in Atrial Fibrillation),17 and ENGAGE AF-TIMI (Effective Anticoagulation With Factor Xa Next Generation in Atrial Fibrillation -Thrombolysis in Myocardial Infarction Study 48),18 and an observational study using Marketscan claims data found that DOACs were as effective as warfarin to prevent stroke and major bleeding among cancer patients with NVAF.22 Data from our real-world study complements trial findings and a previous real-word study. Although a study from Korea reported protective effect of DOACs compared with warfarin on clinical outcomes, the study did not adequately control for confounding.23

Among individual DOACs, we found that apixaban compared to warfarin reduced the risk of major bleeding by 43%. The ARISTOTLE trial did not find any association of apixaban with major bleeding, whereas Shah et al found that apixaban reduced the risk of major bleeding by 63% compared to warfarin. As noted by the authors, the ARISTOTLE trial may be underpowered to detect any significant differences. Our estimates may be more precise given we had large cohort and we better controlled confounding by including cancer characteristics such as stage, grade, and tumor size, which were available in the SEER cancer registry.

We evaluated both all-cause and cardiovascular deaths. Contrary to our hypothesis of similar risk of death among DOAC and warfarin users, we noted lower risk of all-cause death among DOAC users. This was consistent with a recent Cochrane review of 10 randomized controlled trials (n=65 624), which showed a significant all-cause mortality reduction effect of factor Xa inhibitors compared with warfarin (OR 0.89 [95% CI, 0.83–0.95]) among individuals with atrial fibrillation.41 Similar results of lower all-cause mortality were seen among Medicare beneficiaries with atrial fibrillation who were using dabigatran,42 and commercially insured patients with atrial fibrillation using rivaroxaban.43 Our results of no benefits of DOACs, except for rivaroxaban, for cardiovascular deaths were not in agreement with an analysis of 4 trials (n=30 907) that showed a significant reduction in cardiovascular mortality among individual with atrial fibrillation who received DOACs (risk ratio 0.87 [95% CI, 0.80–0.95]).44 Although we controlled for baseline measured confounders, it is possible that unmeasured confounding or differences in cancer characteristics and treatment received after the oral anticoagulation initiation may impact survival. Further work might explore the basis for this association and disentangle causes of deaths among cancer patients with atrial fibrillation.

Our study had some strengths. This was a new user active comparator study, where we included 11 cancers and controlled confounding due to cancer-related and clinical factors. We used propensity score method to control selection bias and confounding. To assess the role of unmeasured confouding, we used negative control outcomes and E-values. Finally, we performed several subgroup and sensitivity analyses to assess the robustness of study findings.

Our study had some limitations. First, our study findings may not be generalizable to patients with other cancers and those who are not enrolled in Medicare fee-for-service. Second, we cannot rule out the possibility of unmeasured confounding given relatively low E-value for some outcomes. Third, some subgroup analysis may be underpowered because of low number of events. Fourth, due to small number of patients in some subgroups such as those with upper gastrointestinal cancers, we did not evaluate outcomes in such population. Fifth, nearly one-third of individuals discontinued DOACs or warfarin and this may introduce bias if the reason for discontinuation differed between these 2 groups. However, our results were consistent in an intention-to-treat analysis where we did not censor individuals upon switching or discontinuation. Sixth, we relied on administrative Medicare claims data to identify drug exposure and clinical outcomes and we cannot rule out the possibility of coding error as these data are not collected for research purposes. Seventh, because we looked at the class effect of DOACs in the primary analysis, we did not evaluate the effect of different dosing on clinical outcomes. Finally, we could not control for some factors that may play role in the prescription of DOACs such as frailty.

In conclusion, among individuals diagnosed with cancer and NVAF, DOACs demonstrated similar effectiveness in preventing stroke and systemic embolism without increasing the risk of major bleeding. Moreover, DOACs were associated with similar risk of cardiovascular death and lower risk of all-cause death. DOACs may be safe and effective alternatives to warfarin among individuals with cancer and NVAF.

Article Information

Acknowledgments

We thank the Applied Research Program of the National Cancer Institute; the Office of Research, Development and Information of the Centers for Medicaid Services; Information Management Services, Inc; and the Surveillance, Epidemiology, and End Results (SEER) program tumor registries for their work in creating the SEER database.

Supplemental Material

Figures S1–S6

Tables S1–S7

Nonstandard Abbreviations and Acronyms

aHR

adjusted hazard ratio

DOAC

direct-acting oral anticoagulants

IPTW

inverse probability of treatment weights

NVAF

nonvalvular atrial fibrillation

SEER

surveillance epidemiology and end results

Disclosures Dr Alexander is a current member and past Chair of FDA’s Peripheral and Central Nervous System Advisory Committee, is a co-founding Principal and equity holder in Monument Analytics, a health care consultancy whose clients include the life sciences industry as well as plaintiffs in opioid litigation, and is a past member of OptumRx’s National P&T Committee. These arrangements have been reviewed and approved by Johns Hopkins University in accordance with its conflict of interest policies. The other authors report no conflicts.

Footnotes

Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/CIRCOUTCOMES.122.008951.

For Sources of Funding and Disclosure, see page 1003 & 1004.

Correspondence to: Hemalkumar B. Mehta, MS, PhD, Johns Hopkins Bloomberg School of Public Health, Department of Epidemiology, 615 N. Wolfe St, E7640 Baltimore, MD 21205. Email

References

  • 1. January CT, Wann LS, Alpert JS, Calkins H, Cigarroa JE, Cleveland JC, Conti JB, Ellinor PT, Ezekowitz MD, Field ME, et al; ACC/AHA Task Force Members. 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation: executive summary: a report of the American College of Cardiology/American Heart Association Task Force on practice guidelines and the Heart Rhythm Society [published correction appears in Circulation. 2014 Dec 2;130(23):e270-1].Circulation. 2014; 130:2071–2104. doi: 10.1161/CIR.0000000000000040LinkGoogle Scholar
  • 2. Colilla S, Crow A, Petkun W, Singer DE, Simon T, Liu X. Estimates of current and future incidence and prevalence of atrial fibrillation in the U.S. adult population.Am J Cardiol. 2013; 112:1142–1147. doi: 10.1016/j.amjcard.2013.05.063CrossrefMedlineGoogle Scholar
  • 3. Miyasaka Y, Barnes ME, Gersh BJ, Cha SS, Bailey KR, Abhayaratna WP, Seward JB, Tsang TSM. Secular trends in incidence of atrial fibrillation in Olmsted County, Minnesota, 1980 to 2000, and implications on the projections for future prevalence [published correction appears in Circulation. 2006 Sep 12;114(11):e498].Circulation. 2006; 114:119–125. doi: 10.1161/CIRCULATIONAHA.105.595140LinkGoogle Scholar
  • 4. Yun JP, Choi EK, Han KD, Park SH, Jung J-H, Park SH, Ahn H-J, Lim J-H, Lee S-R, Oh S. Risk of atrial fibrillation according to cancer type: a nationwide population-based study.JACC CardioOncol. 2021; 3(2):221–232. Published 2021 Jun 15. doi:10.1016/j.jaccao.2021.03.006CrossrefMedlineGoogle Scholar
  • 5. Yuan M, Zhang Z, Tse G, Feng X, Korantzopoulos P, Letsas KP, Yan BP, Wu WK, Zhang H, Li G, et al. Association of cancer and the risk of developing atrial fibrillation: a systematic review and meta-analysis.Cardiol Res Pract. 2019; 2019:8985273. Published 2019 Apr 14. doi:10.1155/2019/8985273CrossrefMedlineGoogle Scholar
  • 6. O’Neal WT, Lakoski SG, Qureshi W, Judd SE, Howard G, Howard VJ, Cushman M, Soliman EZ. Relation between cancer and atrial fibrillation (from the REasons for Geographic And Racial Differences in Stroke Study).Am J Cardiol. 2015; 115:1090–1094. doi: 10.1016/j.amjcard.2015.01.540CrossrefMedlineGoogle Scholar
  • 7. Farmakis D, Parissis J, Filippatos G. Insights into onco-cardiology: atrial fibrillation in cancer.J Am Coll Cardiol. 2014; 63:945–953. doi: 10.1016/j.jacc.2013.11.026CrossrefMedlineGoogle Scholar
  • 8. Siegel RL, Miller KD, Jemal A. Cancer statistics, 2020.CA Cancer J Clin. 2020; 70:7–30. doi: 10.3322/caac.21590CrossrefMedlineGoogle Scholar
  • 9. Kornej J, Börschel CS, Benjamin EJ, Schnabel RB. Epidemiology of atrial fibrillation in the 21st century: novel methods and new insights.Circ Res. 2020; 127:4–20. doi: 10.1161/CIRCRESAHA.120.316340LinkGoogle Scholar
  • 10. Koene RJ, Prizment AE, Blaes A, Konety SH. Shared risk factors in cardiovascular disease and cancer.Circulation. 2016; 133:1104–1114. doi: 10.1161/CIRCULATIONAHA.115.020406LinkGoogle Scholar
  • 11. Navi BB, Reiner AS, Kamel H, Iadecola C, Okin PM, Elkind MSV, Panageas KS, DeAngelis LM. Risk of arterial thromboembolism in patients with cancer.J Am Coll Cardiol. 2017; 70:926–938. doi: 10.1016/j.jacc.2017.06.047CrossrefMedlineGoogle Scholar
  • 12. Melloni C, Shrader P, Carver J, Piccini JP, Thomas L, Fonarow GC, Ansell J, Gersh B, Go AS, Hylek E, et al; ORBIT-AF Steering Committee. Management and outcomes of patients with atrial fibrillation and a history of cancer: the ORBIT-AF registry.Eur Heart J Qual Care Clin Outcomes. 2017; 3:192–197. doi: 10.1093/ehjqcco/qcx004CrossrefMedlineGoogle Scholar
  • 13. Angelini DE, Radivoyevitch T, McCrae KR, Khorana AA. Bleeding incidence and risk factors among cancer patients treated with anticoagulation.Am J Hematol. 2019; 94:780–785. doi: 10.1002/ajh.25494CrossrefMedlineGoogle Scholar
  • 14. January CT, Wann LS, Calkins H, Chen LY, Cigarroa JE, Cleveland JC, Ellinor PT, Ezekowitz MD, Field ME, Furie KL, et al. 2019 AHA/ACC/HRS Focused Update of the 2014 AHA/ACC/HRS Guideline for the Management of Patients With Atrial Fibrillation: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society in Collaboration With the Society of Thoracic Surgeons [published correction appears in Circulation. 2019 Aug 6;140(6):e285].Circulation. 2019; 140:e125–e151. doi: 10.1161/CIR.0000000000000665LinkGoogle Scholar
  • 15. Sorigue M, Miljkovic MD. Atrial fibrillation and stroke risk in patients with cancer: a primer for oncologists.J Oncol Pract. 2019; 15(12):641–650doi:10.1200/JOP.18.00592CrossrefMedlineGoogle Scholar
  • 16. Chen ST, Hellkamp AS, Becker RC, Berkowitz SD, Breithardt G, Fox KAA, Hacke W, Halperin JL, Hankey GJ, Mahaffey KW, et al. Efficacy and safety of rivaroxaban vs. warfarin in patients with non-valvular atrial fibrillation and a history of cancer: observations from ROCKET AF.Eur Heart J Qual Care Clin Outcomes. 2019; 5:145–152. doi: 10.1093/ehjqcco/qcy040CrossrefMedlineGoogle Scholar
  • 17. Melloni C, Dunning A, Granger CB, Thomas L, Khouri MG, Garcia DA, Hylek EM, Hanna M, Wallentin L, Gersh BJ, et al. Efficacy and safety of apixaban versus warfarin in patients with atrial fibrillation and a history of cancer: insights from the ARISTOTLE trial.Am J Med. 2017; 130:1440–1448.e1. doi: 10.1016/j.amjmed.2017.06.026CrossrefMedlineGoogle Scholar
  • 18. Fanola CL, Ruff CT, Murphy SA, et al. Efficacy and safety of edoxaban in patients with active malignancy and atrial fibrillation: analysis of the ENGAGE AF - TIMI 48 trial.J Am Heart Assoc. 2018; 7:e008987. doi: 10.1161/JAHA.118.008987LinkGoogle Scholar
  • 19. Laube ES, Yu A, Gupta D, Miao Y, Samedy P, Wills J, Harnicar S, Soff GA, Mantha S. Rivaroxaban for stroke prevention in patients with nonvalvular atrial fibrillation and active cancer.Am J Cardiol. 2017; 120:213–217. doi: 10.1016/j.amjcard.2017.04.009CrossrefMedlineGoogle Scholar
  • 20. Russo V, Rago A, Papa AA, Meo FD, Attena E, Golino P, D'Onofrio A, Nigro G. Use of non-vitamin k antagonist oral anticoagulants in atrial fibrillation patients with malignancy: clinical practice experience in a single institution and literature review.Semin Thromb Hemost. 2018; 44:370–376. doi: 10.1055/s-0037-1607436CrossrefMedlineGoogle Scholar
  • 21. Ianotto JC, Couturier MA, Galinat H, Mottier D, Berthou C, Guillerm G, Lippert E, Delluc A. Administration of direct oral anticoagulants in patients with myeloproliferative neoplasms.Int J Hematol. 2017; 106:517–521. doi: 10.1007/s12185-017-2282-5CrossrefMedlineGoogle Scholar
  • 22. Shah S, Norby FL, Datta YH, Lutsey PL, MacLehose RF, Chen LY, Alonso A. Comparative effectiveness of direct oral anticoagulants and warfarin in patients with cancer and atrial fibrillation.Blood Adv. 2018; 2:200–209. doi: 10.1182/bloodadvances.2017010694CrossrefMedlineGoogle Scholar
  • 23. Kim K, Lee YJ, Kim TH, Uhm JS, Pak HN, Lee MH, Joung B. Effect of non-vitamin k antagonist oral anticoagulants in atrial fibrillation patients with newly diagnosed cancer.Korean Circ J. 2018; 48:406–417. doi: 10.4070/kcj.2017.0328CrossrefMedlineGoogle Scholar
  • 24. SEER-Medicare:Brief Description of the SEER-Medicare Database.Available at https://healthcaredelivery.cancer.gov/seermedicare/overview/ Accessed February 20, 2021Google Scholar
  • 25. Gross CP, Galusha DH, Krumholz HM. The impact of venous thromboembolism on risk of death or hemorrhage in older cancer patients.J Gen Intern Med. 2007; 22:321–326. doi: 10.1007/s11606-006-0019-xCrossrefMedlineGoogle Scholar
  • 26. Jensen PN, Johnson K, Floyd J, Heckbert SR, Carnahan R, Dublin S. A systematic review of validated methods for identifying atrial fibrillation using administrative data.Pharmacoepidemiol Drug Saf. 2012; 21Suppl 1(0 1):141–147. doi:10.1002/pds.2317CrossrefMedlineGoogle Scholar
  • 27. Yao RJR, Andrade JG, Deyell MW, Jackson H, McAlister FA, Hawkins NM. Sensitivity, specificity, positive and negative predictive values of identifying atrial fibrillation using administrative data: a systematic review and meta-analysis.Clin Epidemiol. 2019; 11:753–767. Published 2019 Aug 23. wdoi:10.2147/CLEP.S206267CrossrefMedlineGoogle Scholar
  • 28. Andrade SE, Harrold LR, Tjia J, Cutrona SL, Saczynski JS, Dodd KS, Goldberg RJ, Gurwitz JH. A systematic review of validated methods for identifying cerebrovascular accident or transient ischemic attack using administrative data.Pharmacoepidemiol Drug Saf. 2012; 21Suppl 1(Suppl 1):100–128. doi:10.1002/pds.2312CrossrefMedlineGoogle Scholar
  • 29. Coding Trend Analyses:Ischemic Stroke.Available at https://www.sentinelinitiative.org/sites/default/files/surveillance-tools/validations-literature/Ischemic_Stroke_Final_Trend_Report.pdf Accessed February 20, 2021Google Scholar
  • 30. Mini-Sentinel Medical Product Assessment. A Protocol For Assessment Of Dabigatran.Available at https://www.sentinelinitiative.org/sites/default/files/Drugs/Assessments/Mini-Sentinel_Protocol-for-Assessment-of-Dabigatran_0.pdf Accessed February 20, 2021Google Scholar
  • 31. Lipsitch M, Tchetgen E, Cohen T. Negative controls: a tool for detecting confounding and bias in observational studies.Epidemiology. 2010; 21:383–388. doi: 10.1097/EDE.0b013e3181d61eebCrossrefMedlineGoogle Scholar
  • 32. Xian Y, Xu H, O’Brien EC, et al. Clinical effectiveness of direct oral anticoagulants vs warfarin in older patients with atrial fibrillation and ischemic stroke: findings from the patient-centered research into outcomes stroke patients prefer and effectiveness research (PROSPER) study.JAMA Neurol. 2019; 76:1192–1202. doi: 10.1001/jamaneurol.2019.2099CrossrefMedlineGoogle Scholar
  • 33. Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. 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. doi: 10.1378/chest.09-1584CrossrefMedlineGoogle Scholar
  • 34. Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. 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; 138:1093–1100. doi: 10.1378/chest.10-0134CrossrefMedlineGoogle Scholar
  • 35. Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies.Stat Med. 2015; 34:3661–3679. doi: 10.1002/sim.6607CrossrefMedlineGoogle Scholar
  • 36. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples.Stat Med. 2009; 28:3083–3107. doi: 10.1002/sim.3697CrossrefMedlineGoogle Scholar
  • 37. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk.J Am Statist Assoc. 1999; 94::496–509.doi: 10.1080/01621459.1999.10474144CrossrefGoogle Scholar
  • 38. Rossello X, González-Del-Hoyo M. Survival analyses in cardiovascular research, part I: the essentials.Rev Esp Cardiol (Engl Ed). 2022; 75:67–76. doi: 10.1016/j.rec.2021.06.003CrossrefMedlineGoogle Scholar
  • 39. Rossello X, González-Del-Hoyo M. Survival analyses in cardiovascular research, part II: statistical methods in challenging situations.Rev Esp Cardiol (Engl Ed). 2022; 75(1):77–85. doi:10.1016/j.rec.2021.07.001CrossrefMedlineGoogle Scholar
  • 40. VanderWeele TJ, Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value.Ann Intern Med. 2017; 167:268–274. doi: 10.7326/M16-2607CrossrefMedlineGoogle Scholar
  • 41. Bruins Slot KM, Berge E. Factor Xa inhibitors versus vitamin K antagonists for preventing cerebral or systemic embolism in patients with atrial fibrillation.Cochrane Database Syst Rev. 2018; 3(3):CD008980. Published 2018 Mar 6. doi:10.1002/14651858.CD008980.pub3CrossrefMedlineGoogle Scholar
  • 42. Graham DJ, Reichman ME, Wernecke M, Zhang R, Southworth MR, Levenson M, Sheu TC, Mott K, Goulding MR, Houstoun M, et al. Cardiovascular, bleeding, and mortality risks in elderly Medicare patients treated with dabigatran or warfarin for nonvalvular atrial fibrillation.Circulation. 2015; 131:157–164. doi: 10.1161/CIRCULATIONAHA.114.012061LinkGoogle Scholar
  • 43. Alberts M, Chen YW, Lin JH, Kogan E, Twyman K, Milentijevic D. Risks of stroke and mortality in atrial fibrillation patients treated with rivaroxaban and warfarin.Stroke. 2020; 51:549–555. doi: 10.1161/STROKEAHA.119.025554LinkGoogle Scholar
  • 44. Chai-Adisaksopha C, Hillis C, Isayama T, Lim W, Iorio A, Crowther M. Mortality outcomes in patients receiving direct oral anticoagulants: a systematic review and meta-analysis of randomized controlled trials.J Thromb Haemost. 2015; 13:2012–2020. doi: 10.1111/jth.13139CrossrefMedlineGoogle Scholar