Effectiveness and safety of apixaban versus rivaroxaban for prevention of recurrent venous thromboembolism and adverse bleeding events in patients with venous thromboembolism: a retrospective population-based cohort analysis

Articles

Effectiveness and safety of apixaban versus rivaroxaban for prevention of recurrent venous thromboembolism and adverse bleeding events in patients with venous thromboembolism: a retrospective population-based cohort analysis Ghadeer K Dawwas, Joshua Brown, Eric Dietrich, Haesuk Park

Summary

Background Apixaban and rivaroxaban, both direct-acting oral anticoagulants, are being increasingly used in routine clinical practice because of their fixed dosing and favourable pharmacological profiles. Differences in the risk of recurrent venous thromboembolism and major bleeding events between the two drugs are currently unknown. We aimed to compare the effectiveness and safety of apixaban and rivaroxaban in prevention of recurrent venous thromboembolism and major bleeding events in patients with venous thromboembolism. Methods We did a retrospective cohort analysis of data from the Truven Health MarketScan commercial and Medicare Supplement claims databases in the USA. We analysed data for adult patients with newly diagnosed venous thromboembolism (deep vein thrombosis or pulmonary embolism) who were new users of apixaban or rivaroxaban between Jan 1, 2014, and Dec 31, 2016. Patients who did not initiate the study drugs within 30 days of their diagnosis, those without 12 months of continuous enrolment in medical and pharmacy benefits, and those who used other anticoagulants during the baseline period were excluded. The primary effectiveness outcome was the incidence of recurrent venous thromboembolism and the primary safety outcome was the incidence of major bleeding events. Cox-proportional hazard models after propensity score matching were used to calculate the hazard ratio (HR) and 95% CI. Findings After propensity score matching, 15 254 patients were included in the cohort (3091 apixaban users and 12 163 rivaroxaban users). The crude incidence of recurrent venous thromboembolism was three per 100 person-years in the apixaban group and seven per 100 person-years in the rivaroxaban group. The incidence of major bleeding was three per 100 person-years in the apixaban group and six per 100 person-years in the rivaroxaban group. In multivariable Cox regression models, the use of apixaban compared with rivaroxaban was associated with decreased risk of recurrent venous thromboembolism (HR 0·37 [95% CI 0·24–0·55]; p<0·0001) and major bleeding events (0·54 [0·37–0·82]; p=0·0031).

Lancet Haematol 2018 Published Online December 14, 2018 http://dx.doi.org/10.1016/ S2352-3026(18)30191-1 See Online/Comment http://dx.doi.org/10.1016/ S2352-3026(18)30211-4 Department of Pharmaceutical Outcomes and Policy (G K Dawwas MBA, J Brown PhD, H Park PhD) and Department of Pharmacotherapy and Translational Research (E Dietrich PharmD), College of Pharmacy, University of Florida, Gainesville, FL, USA Correspondence to: Dr Haesuk Park, College of Pharmacy, University of Florida, Gainesville, FL 32610, USA [email protected]

Interpretation Based on our findings, apixaban seems to be more effective than rivaroxaban in preventing the development of recurrent venous thromboembolism and major bleeding events. Our data might give some assurance to clinicians that apixaban can be an effective and safe therapeutic option for treatment of patients with venous thromboembolism. Funding None. Copyright © 2018 Elsevier Ltd. All rights reserved.

Introduction Venous thromboembolism, comprising both pulmonary embolism and deep vein thrombosis, affects approx-​ imately 200 000 individuals each year in the USA.1,2 30% of these individuals will develop a recurrent venous thromboembolism within 10 years of their initial event.2 Continuing anticoagulation treatment can reduce the risk of recurrent venous thromboembolism but is associated with increased bleeding risk.3 Although warfarin has been the drug of choice for decades, because of some constraints associated with its use, direct-acting oral anticoagulants (DOACs) such as rivaroxaban and apixaban are being increasingly used in routine clinical practice because of

their conventional dosing and favourable pharmacological profiles.4 These advantages have led to a change in the utilisation patterns of anticoagulants in patients with venous thromboembolism, as demonstrated in an analysis of data from Danish nationwide registries.5 The study reported that, in September, 2016, 70% of patients with venous thromboembolism initiated rivaroxaban, 16% initiated apixaban, 2% initiated vitamin K antagonists (VKAs), and 2% initiated dabigatran.5 The American College of Chest Physicians (CHEST) guidelines recommend the use of DOACs over VKAs in patients with venous thromboembolism without an associated cancer diagnosis; however, head-to-head

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

1

Articles

Research in context Evidence before this study Venous thromboembolism, a common chronic disorder that comprises both pulmonary embolism and deep vein thrombosis, affects approximately 200 000 individuals annually in the USA. Patients with venous thromboembolism are at increased risk of recurrence in the first 10 years after their initial diagnosis. We searched MEDLINE from Jan 1, 2006, to Aug 11, 2018, using the search terms “venous thromboembolism” and “apixaban”, and restricted our search to studies done in humans. Although we did not identify any previous head-to-head comparisons from randomised controlled trials or observational data, we identified two network-meta analyses of randomised controlled trials. These meta-analyses, which were restricted by the small number of trials and low event rates, suggested no difference in the risk of recurrent venous thromboembolism or major bleeding events between apixaban and rivaroxaban. Added value of this study To the best of our knowledge, this study is the first population-based cohort analysis to compare the effectiveness and safety of apixaban and rivaroxaban for the treatment of venous thromboembolism. This study of 15 254 patients with

comparisons between the different DOACs have not been done in patients with venous thromboembolism.6 Most of the available evidence comes from randomised clinical trials, which compared these agents to either standard therapy or placebo. For example, evidence from the AMPLIFY trial found apixaban to be noninferior to standard therapy (subcutaneous enoxaparin followed by warfarin) for prevention of recurrent venous thromboembolism and to be associated with a lower bleeding risk (relative risk [RR] 0·31 [95% CI 0·17–0·55]).7 Similarly, results from the EINSTEIN trial, which evaluated rivaroxaban for the acute treatment and secondary prevention of venous thromboembolism, found rivaroxaban to be noninferior to standard therapy (subcutaneous enoxaparin followed by a VKA, either warfarin or acenocoumarol) and to have a similar safety profile to standard therapy.8 Results from two randomised controlled trials comparing apixaban and rivaroxaban head to head, the COBRA (NCT03266783) and CANVAS (NCT02744092) trials, which are still recruiting participants, are awaited; until these results become available, evidence generated from real-world data can help to guide selection between apixaban and rivaroxaban in routine clinical practice. We aimed to assess the effectiveness of apixaban versus rivaroxaban for the prevention of recurrent venous thromboembolism and to examine whether there are any differences in major bleeding risk between apixaban and rivaroxaban among patients with venous thrombo-​ embolism. 2

venous thromboembolism provides new evidence about the effectiveness and safety of apixaban in this population. By doing a propensity score-matched analysis, we show that apixaban use was associated with a lower risk of recurrent venous thromboembolism, major bleeding events, and minor bleeding events than was rivaroxaban use. These findings remained consistent in several sensitivity analyses. Implications of all the available evidence Until the results of randomised controlled trials comparing the efficacy and safety of apixaban versus rivaroxaban in patients with venous thromboembolism become available, clinicians have to rely on robust observational data. Our findings suggest that apixaban might be preferred over rivaroxaban in the venous thromboembolism population for reducing the risk of recurrence, major bleeding, and minor bleeding events. Additionally, our subgroup analyses by baseline chronic kidney disease, presence of active cancer, and among patients with provoked and unprovoked venous thromboembolism suggest that the beneficial effect of apixaban extends across several subgroups that might be at increased risk of recurrent venous thromboembolism.

Methods

Study design and data sources We did a retrospective population-based cohort analysis using data from the Truven Health MarketScan commercial and Medicare Supplement claims databases. Data from Jan 1, 2014, to Dec 31, 2016, were used. These databases contain information about outpatient and inpatient claims (including inpatient deaths), health expenditures, enrolment, and prescription drug claims for more than 57 million individuals in the USA. The commercial data include privately insured employees and their dependents who are covered by employersponsored health insurance programmes. The Medicare data represent retirees who are covered by Medicare Supplement insurance. An institutional review board was approved for the study by the University of Florida Health Science Center (Gainesville, FL, USA).

Participants Patients were included if they were aged 18 years or older and had a diagnosis of venous thromboembolism based on the presence of a primary or secondary diagnosis of venous thromboembolism presenting on inpatient or outpatient claims with codes that have been validated previously (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM], 415.1, 451.1, 453.2, 453.4, 453.5, 453.8, or 453.9).9 Patients were selected between 2014 and 2015, but were followed up retrospectively through 2016. Patients were required to be treatment naive and newly initiated on apixaban or rivaroxaban (new users) within

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

Articles

30 days of their first venous thromboembolism diagnosis and to have 12 months of continuous enrolment in medical and pharmacy benefits before treatment initiation. A period of 30 days was selected to provide patients with sufficient time to be discharged from hospital and pick up their prescription from an outpatient pharmacy. We were not able to compare apixaban to other DOACs such as edoxaban or dabigatran since edoxaban was approved in 2015, and the number of dabigatran users in the venous thromboembolism population was low (n=716). The treatment initiation date was assigned as the index date.  Patients were excluded if they had used any anticoagulant therapy during the 12-month pre-index period. Because patients with unprovoked venous thromboembolism are at higher risk of developing recurrent venous thromboembolism than are those with provoked venous thromboembolism, we classified patients as having provoked or unprovoked venous thromboembolism at baseline.10 Patients were considered to have provoked venous thromboembolism if they had malignancy-associated venous thromboembolism (a cancer diagnosis within the preceding 6 months of venous thromboembolism) or any of the following within the 90 days preceding a diagnosis of venous thromboembolism: pregnancy-related venous thromboembolism, traumarelated venous thromboembolism, surgery-related venous thrombo​ embolism, and a hospital admission for 3 or more consecutive days.10 Another subgroup of interest in this population is patients with active cancer as they have been found to have a two to three times increased risk of developing recurrent venous thromboembolism and major bleeding events.11 Therefore, we stratified the analysis by baseline active cancer, which was defined according to the presence of a cancer diagnosis within the 6 months preceding venous thromboembolism or ongoing treatment with radiotherapy or chemotherapy.

Outcomes The primary effectiveness outcome was the incidence of recurrent venous thromboembolism (deep vein thrombosis or pulmonary embolism), which was defined according to the presence of primary discharge diagnoses codes that have been validated previously and found to have a positive predictive value of 73–83%.12 The primary safety outcome was the incidence of major bleeding events presenting on hospital admission with primary discharge diagnoses including intracranial haemorrhage (ICD-9-CM: 430, 431, 432.0, 432.1, and 432.9), gastrointestinal bleeding (ICD-9-CM: 455.2, 455.5, 455.8, 456.0, 456.20, 530.7, 530.82, 531.0–531.6, 532.0–532.6, 533.0–533.6, 534.0–534.6, 535.01–535.61, 537.83, 562.02, 562.03, 562.12, 562.13, 568.81, 569.3, 569.85, 578.0, 578.1, and 578.9), and other major bleeding events (ICD-9-CM: 423.0x, 459.0x, 599.7x, 719.11, 784.7x, 784.8x, and 786.3x). The full list of ICD-10 codes used for defining bleeding events is included in the appendix.

In each effectiveness and safety analysis, patients were followed up from the index date to the occurrence of an outcome, treatment discontinuation (a 7-day gap was allowed between the end of the supply of the last prescription and the date of collection of a new prescription), switch to the study comparator, end of enrolment, or end of the study period. Effectiveness and the safety outcomes were analysed separately. In a secondary analysis, we examined minor bleeding events defined according to the presence of bleeding events in outpatient claims (ie, not requiring admission to hospital).

Adjustment for confounders To adjust for differences in baseline characteristics and disease risk factors, propensity score matching (1:4) was used. Patients were matched to the nearest neighbour using a caliper of 0·01. A logistic regression model estimated the predicted probability of initiating apixaban compared with rivaroxaban given baseline covariates. These covariates included patient demographics (sex and age), presence of comorbidities (cancer, surgery, trauma, antiphospholipid syndrome, hyperlipidaemia, abnormal coagulation, tobacco use, respiratory diseases, liver diseases, chronic kidney disease, anaemia, alcohol use disorder, drug use disorder, history of bleeding, ischaemic heart disease, myocardial infarction, stroke, heart failure, varicose veins, and thrombocytopenia), previous use of medications (antiplatelet therapy, corticosteroids, non-steroidal anti-inflammatory drugs [NSAIDs], angiotensin-converting enzyme [ACE] inhibitors, aspirin, β blockers, calcium channel blockers, selective serotonin reuptake inhibitors [SSRIs], protonpump inhibitors [PPIs], loop diuretics, potassiumsparing diuretics, thiazide diuretics, vasodilators, oestrogens, and cyclo-oxygenase-2 [COX-2] inhibitors), and measure of health-care utilisation (mean total number of outpatient visits). These variables were set a priori and extracted from previous studies that assessed risk factors for development of venous thromboembolism or bleeding events.13–16 Furthermore, we calculated the HAS-BLED score, a commonly used bleeding score in atrial fibrillation and validated in venous thrombo-​ embolism, for both groups.17 HAS-BLED was calculated according to the presence of the following components: hypertension, abnormal renal function, abnormal liver function, stroke, bleeding, age older than 65 years, alcohol use disorder, drug use disorder, and use of medications predisposing patients to bleeding (eg, aspirin, clopidogrel, and NSAIDs).

Statistical analysis Demographics and clinical characteristics were summarised with means for continuous variables and with proportions for categorical variables. The differences in baseline covariates between users of apixaban and rivaroxaban before matching were assessed with χ² tests

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

See Online for appendix

3

Articles

Apixaban (n=3387)

Rivaroxaban (n=35 243)

p value

56·5 (15·6)

Men

1673 (49·4%)

17 807 (50·5%)

0·22

Antiplatelet agents

Women

1714 (50·6%)

17 436 (49·5%)

··

Corticosteroids

Comorbidities Cancer

598 (17·7%)

6664 (18·9%)

0·08

Active cancer*

474 (14·0%)

6131 (17·4%)

<0·0001

Surgery

593 (17·5%)

6220 (17·6%)

0·85

Trauma

1086 (32·1%)

11 370 (32·3%)

0·83

Hyperlipidaemia

1 (0·0%)

16 (0·0%)

1·00

1523 (45·0%)

13 649 (38·7%)

<0·0001

Abnormal coagulation

128 (3·8%)

1035 (2·9%)

0·01

Tobacco use

327 (9·7%)

3241 (9·2%)

0·28

Respiratory diseases

667 (19·7%)

6194 (17·6%)

0·0021

1495 (4·2%)

<0·0001

12 200 (34·6%)

<0·0001

NSAIDs

808 (23·9%)

7677 (21·8%)

0·01

ACE inhibitors

740 (21·8%)

5806 (16·5%)

<0·0001

Aspirin β blockers Calcium-channel blockers

0·0018 <0·0001

Loop diuretics

672 (19·8%)

4024 (11·4%)

<0·0001

Potassium-sparing diuretics

138 (4·1%)

910 (2·6%)

<0·0001

Thiazide

69 (2·0%)

471 (1·3%)

0·0009

Vasodilators

71 (2·1%)

360 (1·0%)

<0·0001

55 (1·6%)

493 (2·1%)

0·29

130 (3·8%)

981 (2·8%)

0·0002

Oestrogens

Anaemia

558 (16·5%)

4973 (14·1%)

<0·0001 0·02

Mean number of outpatient visits

0·05

History of bleeding

464 (13·7%)

3947 (11·2%)

<0·0001

Ischaemic heart disease

739 (21·8%)

5079 (14·4%)

<0·0001

Myocardial infarction

148 (4·4%)

970 (2·8%)

<0·0001

Stroke

330 (9·7%)

2059 (5·8%)

<0·0001

Heart failure

345 (10·2%)

1477 (4·2%)

<0·0001

Varicose veins

82 (2·4%)

744 (2·1%)

0·24

Thrombocytopenia

17 (0·5%)

201 (0·6%)

0·72

Mean HAS-BLED score

0·2 (0·4)

0·1 (0·4)

0·0071

456 (13·5%)

4565 (13·0%)

0·21

Pulmonary embolism

1026 (30·3%)

9738 (27·6%)

··

Deep vein thrombosis

1905 (56·2%)

20 940 (59·4%)

··

Provoked venous thromboembolism

1731 (51·1%)

17 110 (48·5%)

0·71

Unprovoked venous thromboembolism

1656 (48·9%)

18 133 (51·5%)

··

Both pulmonary embolism and deep vein thrombosis

(Table 1 continues in next column)

for categorical variables and with independent t tests for continuous variables. After matching, the balance in baseline covariates was assessed with standardised differences where differences less than 0·1 were considered as well balanced.18,19 After propensity score matching, the incidence of recurrent venous thromboembolism and major bleeding events was reported as the total number of events per 100 person-years. The Cox proportional hazards model was used to compare effectiveness and safety outcomes

<0·0001

4908 (13·9%)

0·96

175 (0·5%)

5607 (15·9%)

8480 (24·1%)

1097 (3·1%)

1170 (3·3%)

814 (24·0%) 538 (15·9%)

3003 (8·5%)

28 (0·8%)

0·01 <0·0001

983 (29·0%)

106 (3·1%)

134 (4·0%)

400 (1·1%) 7586 (21·5%)

PPIs

580 (17·1%)

Drug use disorder

56 (1·7%) 1141 (33·7%)

SSRIs

Chronic kidney disease

Initial presentation of venous thromboembolism

4

247 (7·3%) 1345 (39·7%)

Liver disease

Alcohol use disorder

p value

Baseline medications

Sex

Antiphospholipid syndrome

Rivaroxaban (n=35 243)

(Continued from previous column)

61·5 (16·6)

Mean age, years

<0·0001

Apixaban (n=3387)

COX-2 inhibitors

20·6 (22·3)

19·1 (19·2)

0·0005 <0·0001

Data are mean (SD) or n (%). p values were generated with the χ2 test for categorical variables and the t test for continuous variables. NSAIDs=non-steroidal anti-inflammatory drugs. ACE=angiotensin-converting enzyme. SSRIs=selective serotonin reuptake inhibitors. PPIs=proton-pump inhibitors. COX-2=cyclooxygenase-2. *Active cancer was defined on the basis of the presence of a cancer diagnosis within the 6-month period preceding venous thromboembolism or in treatment with radiotherapy or chemotherapy.

Table 1: Demographic and clinical characteristics of new users of apixaban and rivaroxaban before propensity score matching

between patients with apixaban and rivaroxaban and the proportionality assumption was tested by use of Schoenfeld residuals. In subgroup analyses, we examined the potential heterogeneity of treatment effects in selected subgroups of patients with venous thromboembolism, including those with active cancer versus those without, those with chronic kidney disease versus those without, those aged 65 years or younger versus those aged older than 65 years, those with provoked versus those with unprovoked venous thromboembolism, those with pulmonary embolism versus those with deep vein thrombosis (for the recurrent venous thromboembolism outcome), those with gastrointestinal bleeding versus those with intracranial bleeding (for the major bleeding outcome), those with early events (<90 days) versus those with late events (≥90 days), and those who were on treatment for 3 or more months versus those on treatment for less than 3 months. For the subgroup analyses, a pinteraction value less than 0·05 was used to denote a significant difference between the two groups. Matching was done again within each of the selected subgroup analyses. All analyses were done with SAS, version 9.4.

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

Articles

Apixaban (n=3091)

Rivaroxaban (n=12 163)

Standardised difference

Apixaban (n=3091)

Rivaroxaban (n=12 163)

Standardised difference

(Continued from previous column)

61·6 (16·4)

59·9 (16·2)

0·10

Men

1526 (49·4%)

6038 (49·6%)

0·00

Women

1565 (50·6%)

6125 (50·4%)

··

Corticosteroids NSAIDs

736 (23·8%)

2781 (22·9%)

0·02

546 (17·7%)

2300 (18·9%)

–0·03

ACE inhibitors

675 (21·8%)

2429 (20·0%)

0·04

Active cancer*

415 (13·4%)

1250 (10·3%)

0·05

Surgery

542 (17·5%)

2172 (17·9%)

–0·01

β blockers Calcium-channel blockers

Mean age, years

Baseline medications

Sex

Comorbidities Cancer

Trauma Hyperlipidaemia Abnormal coagulation

991 (32·1%)

3963 (32·6%)

–0·01

1391 (45·0%)

5282 (43·4%)

0·03

117 (3·8%)

435 (3·6%)

0·01

Antiplatelet agents

Aspirin

223 (7·2%)

785 (6·5%)

0·03

1230 (39·8%)

4631 (38·1%)

0·03

50 (1·6%)

187 (1·5%)

0·01

1039 (33·6%)

3735 (30·7%)

0·06

743 (24·0%)

2647 (21·8%)

0·05

SSRIs

491 (15·9%)

1896 (15·6%)

0·01

PPIs

896 (29·0%)

3353 (27·6%)

0·03

Tobacco use

299 (9·7%)

1141 (9·4%)

0·01

Loop diuretics

611 (19·8%)

2164 (17·8%)

0·05

Respiratory diseases

609 (19·7%)

2386 (19·6%)

0·00

Potassium-sparing diuretics

125 (4·0%)

440 (3·6%)

0·02

97 (3·1%)

378 (3·1%)

0·00

Chronic kidney disease

Liver disease

527 (17·1%)

1856 (15·3%)

0·05

Anaemia

509 (16·5%)

1986 (16·3%)

0·00

Oestrogens

26 (0·8%)

101 (0·8%)

0·00

COX-2 inhibitors

Alcohol use disorder Drug use disorder

122 (4·0%)

465 (3·8%)

0·01

History of bleeding

423 (13·7%)

1597 (13·1%)

0·02

Ischaemic heart disease

673 (21·8%)

2408 (19·8%)

0·05

Myocardial infarction

135 (4·4%)

504 (4·1%)

0·01

Stroke

301 (9·8%)

1071 (8·8%)

0·03

Heart failure

312 (10·1%)

1077 (8·9%)

0·04

Varicose veins

75 (2·4%)

285 (2·3%)

0·01

Thrombocytopenia

16 (0·5%)

72 (0·6%)

–0·01

(Table 2 continues in next column)

Thiazide

63 (2·0%)

223 (1·8%)

0·01

Vasodilators

64 (2·1%)

222 (1·8%)

0·02

Mean number of outpatient visits

50 (1·6%)

212 (1·7%)

0·02

118 (3·8%)

440 (3·6%)

0·01

20·7 (22·3)

20·7 (20·4)

0·00

Data are mean (SD) or n (%). Standardised difference calculated as described previously by Yang and Dalton;20 standardised difference of 0·1 indicates similarity. NSAIDs=non-steroidal anti-inflammatory drugs. ACE=angiotensin-converting enzyme. SSRIs=selective serotonin reuptake inhibitors. PPIs=proton-pump inhibitors. COX-2=cyclooxygenase-2. *Active cancer defined according to presence of cancer diagnosis within the 6-month period preceding venous thromboembolism or in treatment with radiotherapy or chemotherapy.

Table 2: Demographics and clinical characteristics in propensity-score matched cohorts of new users of apixaban and rivaroxaban

Role of the funding source

99 days (146) in the rivaroxaban group. We identified 25 cases of recurrent venous thromboembolism among apixaban users and 254 cases among rivaroxaban users. The crude incidence of recurrent venous thromboembolism was three per 100 person-years among apixaban users and seven per 100 person-years among rivaroxaban users. The Kaplan-Meier curve comparing the risk of recurrent venous Results We identified 3387 new users of apixaban and 35 243 new thromboembolism between apixaban and rivaroxaban is users of rivaroxaban (appendix). Table 1 summarises the shown in the appendix. In the Cox proportional hazards differences in demographics and clinical characteristics model, the use of apixaban was associated with lower risk among users of apixaban and rivaroxaban before of recurrent venous thromboembolism than was use of propensity score matching. After matching, 15 254 patients rivaroxaban (hazard ratio [HR] 0·37 [95% CI 0·24–0·55]; were included in the cohort (3091 apixaban users and p<0·0001). In subgroup analyses (table 4; appendix), we 12 163 rivaroxaban users; table 2). In the propensity score- found this association to be consistent in patients with matched cohorts, patients’ characteristics, including active cancer and in those without; in patients with chronic age, sex, and the presence of comorbid conditions kidney disease and those without; in patients aged 65 years (eg, hyperlipidaemia, respiratory diseases, and liver or younger and those aged older than 65 years; in those diseases), and previous medication use (eg, NSAIDs and with provoked venous thromboembolism and those with COX-2 inhibitors) were similar between apixaban and unprovoked venous thromboembolism; in those with pulmonary embolism and those with deep vein thrombosis; rivaroxaban users (all standardised differences <0·1). Table 3 shows the risk of recurrent venous thrombo-​ in comparisons of those with early events and those with embolism in apixaban users versus rivaroxaban users in late events, and in comparisons of patients on treatment for the propensity score-matched analysis. The mean follow- 3 or more months and those on treatment for less than up time was 99 days (SD 173) in the apixaban group and 3 months. There was no funding source for this study. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

5

Articles

Patients Person-years

Events

Crude incidence per 100 person-years

Adjusted hazard ratio (95% CI)

p value

Recurrent venous thromboembolism Apixaban Rivaroxaban

3091

861

25

3

0·37 (0·24–0·55)

<0·0001

12 163

3394

254

7

Ref

··

Major bleeding* Apixaban Rivaroxaban

3091

862

28

3

0·54 (0·37–0·82)

0·0031

12 163

3400

188

6

Ref

··

Minor bleeding* Apixaban Rivaroxaban

3091

839

166

20

0·57 (0·48–0·67)

<0·0001

12 163

3186

1082

34

Ref

··

Data are n, unless otherwise stated. *Major bleeding events were defined as bleeding events requiring admission to hospital, whereas minor bleeding events were defined according to the presence of bleeding in outpatient settings.

Table 3: Risk of recurrent venous thromboembolism, major bleeding events, and minor bleeding events with apixaban versus rivaroxaban in propensity score-matched analyses

Patients

Person-years Events

Crude incidence Adjusted hazard per 100 person- ratio (95% CI) years

pinteraction

Active cancer Present Apixaban Rivaroxaban

420

108

4

4

0·44 (0·16–1·27)

1660

425

33

8

Ref

0·87 ··

Absent Apixaban Rivaroxaban

2667

751

21

3

0·37 (0·23–0·57)

··

10 462

2968

216

7

Ref

··

Chronic kidney disease Present Apixaban Rivaroxaban

522

162

3

2

0·25 (0·08–0·80)

1856

585

40

7

Ref

0·22 ··

Absent Apixaban Rivaroxaban

2553

699

22

3

0·38 (0·25–0·59)

··

10 167

2694

218

8

Ref

··

Age ≤65 years Apixaban

1935

442

22

5

Rivaroxaban

7677

1776

183

10

Apixaban

1144

417

3

Rivaroxaban

4281

1548

91

0·47 (0·30–0·73)

0·01

Ref

··

1

0·11 (0·04–0·36)

··

6

Ref

··

>65 years

Baseline venous thromboembolism Provoked venous thromboembolism Apixaban

1511

397

10

3

0·26 (0·14–0·50)

Rivaroxaban

5868

1495

137

9

Ref

0·35 ··

Unprovoked venous thromboembolism Apixaban

1576

462

15

3

0·48 (0·28–0·82)

··

Rivaroxaban

6205

1881

120

6

Ref

··

Type of event Pulmonary embolism Apixaban Rivaroxaban

3091

863

27

3

0·57 (0·38–0·85)

12 163

3412

181

5

Ref

(Table 4 continues on next page)

6

0·66 ··

Study results remained consistent in the sensitivity analyses, including after restriction of the time between diagnosis and first prescription to 2 days and to 7 days (table 5). Table 3 shows the risk of major bleeding in apixaban users versus rivaroxaban users in the propensity score-matched analysis. The primary safety outcome, the incidence of major bleeding, was three per 100 person-years in the apixaban group and six per 100 person-years in the rivaroxaban group. The Kaplan-Meier curve comparing the risk of major bleeding between apixaban and rivaroxaban is shown in the appendix. In the Cox proportional hazards model, the use of apixaban was associated with lower risk of major bleeding than was use of rivaroxaban (HR 0·54 [95% CI 0·37–0·82]; p=0·0031). In subgroup analyses (table 6; appendix), this association was consistent in patients with active cancer and those without; in patients with baseline chronic kidney disease and those without; in those aged 65 years or younger and those aged older than 65 years; in those with provoked venous thromboembolism and in those with unprovoked venous thromboembolism; in those with gastrointestinal bleeding and those with intracranial bleeding; in comparisons of those with early events and those with late events; and in comparisons of patients on treatment for 3 or more months and those on treatment for less than 3 months. Study results remained consistent in sensitivity analyses, including after restricting the time between diagnosis and treatment initiation to 2 days and to 7 days (table 5). Deaths in the inpatient setting were higher among rivaroxaban users than among apixaban users (1·5% vs 1·0%). For the secondary safety outcome, the incidence of minor bleeding was 20 per 100 person-years in the apixaban group and 34 per 100 person-years in the rivaroxaban group (table 3). In the Cox proportional hazards model, the use of apixaban was associated with lower risk of minor bleeding events than use of rivaroxaban (HR 0·57 [95% CI 0·48–0·67]; p<0·0001; table 3).

Discussion To the best of our knowledge, this is the first real-world assessment, based on data from routine clinical practice, of the effectiveness and safety of apixaban compared with rivaroxaban in patients with venous thrombo-​ embolism. In this USA-based, propensity score-matched cohort analysis of patients with venous thrombo-​ embolism, the use of apixaban was associated with a significantly decreased risk of recurrent venous thromboembolism and major bleeding events. These results were consistent in patients with baseline active cancer versus those without, those with baseline chronic kidney disease versus those without, and in those with provoked versus those with unprovoked venous thromboembolism, in comparisons of early and late events, and when examining the event type. We also

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

Articles

found the use of apixaban to be associated with a decreased risk of minor bleeding events. To the best of our knowledge, no head-to-head comparisons have been done of apixaban and rivaroxaban in patients with venous thromboembolism, and the only available evidence comes from two network metaanalyses of randomised controlled trials. Cohen and colleagues21 indirectly assessed the efficacy and safety of DOACs (by comparing apixaban and rivaroxaban against standard therapy). They reported no difference in the risk of recurrent venous thromboembolism between apixaban and rivaroxaban. These results were supported by another meta-analysis (of the same trials), which found no significant difference in the risk of recurrent venous thromboembolism between apixaban and rivaroxaban (RR 0·57 [95 % CI 0·29–1·15]).22 The results from these analyses seem to be contradictory to our findings; however, several limitations of these analyses could explain the conflicting findings. The conclusion that apixaban was not significantly associated with decreased risk of recurrent venous thromboembolism compared with rivaroxaban was based on three trials, which reported a small number of venous thromboembolism events (59 for apixaban vs 86 for rivaroxaban). Furthermore, the reported results from both metaanalyses were not significant between apixaban and rivaroxaban when the risk of major bleeding was examined alone (which was similarly affected by the small number of events; 15 for apixaban and 40 for rivaroxaban). However, when major bleeding events were grouped with clinically relevant non-major bleeding events (ie, minor bleeding) the risk was lower with apixaban than with rivaroxaban (115 for apixaban and 388 for rivaroxaban; RR 0·47 [95 % CI 0·37–0·61]). Since these analyses are likely to be underpowered as reflected by the wide confidence intervals, we believe that our conclusions are in fact complementary to the findings of the two network meta-analyses. In this study, 474 (14·0%) of 3387 patients who initiated apixaban had active cancer, compared with 6131 (17·4%) of 35 243 who initiated rivaroxaban. A comparison of the characteristics of patients with active cancer who initiated apixaban and rivaroxaban is provided in the appendix. The proportion of users with an active cancer was higher than reported previously in the AMPLIFY (2·5%) and the ENSTEIN trials (6·8%) and in an observational analysis (11·4–14·0%).23 Furthermore, results from the subgroup analysis for the primary outcome of recurrent venous thromboembolism suggest some differences between patients aged 65 years or younger and those aged older than 65 years (pinteration=0·01). We were unable to identify any potential pharmacokinetic or pharmacodynamic mechanism that can explain this interaction. Because of the small number of events in the apixaban group (n=3), future studies are needed to confirm whether there is an additional benefit of apixaban relative to rivaroxaban in patients older than 65 years.

Patients

Crude incidence Adjusted hazard per 100 person- ratio (95% CI) years

Person-years Events

pinteraction

(Continued from previous page) Deep vein thrombosis Apixaban Rivaroxaban

3091

861

13

2

0·48 (0·27–0·87)

··

12 163

3443

103

3

Ref

··

Events Early <90 days Apixaban Rivaroxaban

3091

862

20

2

0·71 (0·44–1·14)

12 163

3411

108

3

Ref

0·43 ··

Late ≥90 days Apixaban Rivaroxaban

3091

864

5

1

0·13 (0·05–0·31)

··

12 163

3412

146

4

Ref

··

Treatment duration ≥3 months Apixaban Rivaroxaban

860

536

7

1

0·34 (0·16–0·75)

3322

2051

74

4

Ref

0·71 ··

0·35 (0·22–0·57)

··

Ref

··

<3 months Apixaban

2226

323

18

6

Rivaroxaban

8788

1489

160

11

Data are n, unless otherwise stated.

Table 4: Risk of recurrent venous thromboembolism with apixaban versus rivaroxaban in propensity score-matched subgroup analyses

Patients

Person-years

Events

Crude incidence per 100 person-years

Adjusted hazard ratio (95% CI)

Primary outcome (recurrent venous thromboembolism) 2 days between diagnosis and first refill Apixaban

1376

357

19

5

Rivaroxaban

5433

1366

159

12

0·45 (0·28–0·72) Ref

7 days between diagnosis and first refill Apixaban

1824

490

20

4

Rivaroxaban

7182

1853

211

11

0·35 (0·22–0·55) Ref

Primary outcome (major bleeding) 2 days between diagnosis and first refill Apixaban

1376

386

5

1

0·19 (0·08–0·48)

Rivaroxaban

5433

1378

96

7

Ref

7 days between diagnosis and first refill Apixaban

1824

492

22

4

0·34 (0·18–0·63)

Rivaroxaban

7182

1878

121

6

Ref

Data are n, unless otherwise stated.

Table 5: Risk of major bleeding events with apixaban versus rivaroxaban in propensity score-matched sensitivity analyses

The observed effectiveness and safety of apixaban compared with rivaroxaban might be partially explained by differences in the pharmacokinetic profile of the two medications. Persistence of anticoagulation effects beyond the half-life of rivaroxaban allowed for selection of a once-daily dosing regimen, which, according to the manufacturer, would lead to improved rates of adherence compared with a twice-daily dosing regimen.24 However,

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

7

Articles

Medications

Patients Person-years Events Crude incidence per 100 person-years

Adjusted hazard ratio (95% CI)

pinteraction

0·70

Active cancer Present Apixaban Rivaroxaban

420

107

5

5

0·51 (0·20–1·30)

1660

428

36

8

Ref

··

Absent Apixaban Rivaroxaban

2667

752

24

3

0·59 (0·38–0·91)

··

10 462

2981

152

5

Ref

··

Chronic kidney disease Present Apixaban Rivaroxaban

522

162

11

7

1856

195

48

25

0·77 (0·40–1·48)

0·40

Ref

··

Absent Apixaban Rivaroxaban

2553

670

18

3

0·51 (0·31–0·84)

··

10 167

2703

131

5

Ref

··

Age ≤65 years Apixaban

1935

445

11

2

0·56 (0·30–1·06)

Rivaroxaban

7677

1798

74

4

Ref

0·58 ··

>65 years Apixaban

1144

415

17

4

0·51 (0·30–0·84)

··

Rivaroxaban

4281

1551

118

8

Ref

··

Baseline venous thromboembolism Provoked venous thromboembolism Apixaban

1511

396

16

4

0·49 (0·29–0·83)

Rivaroxaban

5868

1501

117

8

Ref

0·47 ··

Unprovoked venous thromboembolism Apixaban

1576

464

12

3

0·62 (0·34–1·14)

··

Rivaroxaban

6205

1901

74

4

Ref

··

Type of event Gastrointestinal bleeding Apixaban Rivaroxaban

3091

866

34

4

0·69 (0·47–1·01)

12 163

3550

190

5

Ref

0·42 ··

Intracranial bleeding Apixaban Rivaroxaban

3091

868

7

1

0·76 (0·33–1·72)

··

12 163

3359

34

1

Ref

··

Events Early <90 days Apixaban Rivaroxaban

3091

864

14

2

0·63 (0·36–1·11)

12 163

3411

84

2

Ref

0·18 ··

Late ≥90 days Apixaban Rivaroxaban

3076

861

14

2

0·48 (0·28–0·83)

··

12 108

3443

104

3

Ref

··

Treatment duration ≥3 months Apixaban Rivaroxaban

860

538

4

1

0·37 (0·13–1·04)

3322

2083

41

2

Ref

0·09 ··

<3 months Apixaban

2226

322

24

7

Rivaroxaban

8788

1450

139

10

0·67 (0·43–1·01)

··

Ref

··

Data are n, unless otherwise stated.

Table 6: Risk of major bleeding events with apixaban versus rivaroxaban in propensity score-matched subgroup analyses

8

to ensure rivaroxaban concentrations remained higher than the minimum concentration necessary to prevent thrombosis with the short half-life, a high Cmax was needed to facilitate once-daily dosing. Accordingly, the peak-to-trough ratio of rivaroxaban was approximately 10 (at a dose of 10–20 mg once daily) whereas for apixaban it was around 3 (at a dose of 5 mg twice daily).25,26 A separate analysis comparing DOACs dosed twice daily (dabigatran and apixaban) to those dosed once daily (rivaroxaban and edoxaban) found a more favourable safety profile with DOACs dosed twice daily, with the benefit proposed to be a result of the decreased peak-to-trough ratios afforded by twice-daily DOACs.25 Our results seem to confirm the results of this analysis, as evidenced by the effectiveness of apixaban compared with rivaroxaban with regard to safety outcomes. Our study had several strengths. We used the Truven Health Marketscan database, which is nationally representative of patients enrolled in commercial or Medicare Supplement health insurance and allowed for longitudinal assessment of drug exposure and clinical outcomes. Additionally, the large sample size allowed for assessment of the heterogeneity of treatment effects in selected subpopulations with venous thromboembolism. Additionally, although patients initiating apixaban were older and more likely to have a higher prevalence of comorbidities (eg, ischaemic heart diseases, hyperlipidaemia, and hypertension) than were those initiating rivaroxaban, propensity score matching allowed for adjustment of differences in baseline demographics and clinical characteristics. Several limitations of this analysis should be noted. First, fatal outcome events that occurred in outpatient settings were not captured because of the absence of linkage to death records, although deaths in the inpatient setting were higher among rivaroxaban users than among apixaban users (1·5% vs 1·0%). Second, because recent studies suggested that most patients with venous thrombo-​ embolism are more likely to be prescribed rivaroxaban or apixaban than other DOACs, assessment of other comparators (eg, warfarin and dabigatran) was not possible because of the smaller numbers of patients taking these drugs. Third, residual confounding because of missing information on laboratory values and other variables (eg, D-dimer) is possible. Fourth, the outcome definition based on ICD codes could have introduced some outcome misclassification. Although non-differential, we used an outcome definition based on the primary diagnosis only to minimise the effect of this bias on the observed estimates. Fifth, several patients were excluded because of unconfirmed continuous eligibility, which is required for ascertainment of baseline covariates and previous medication use. Moreover, selection bias could have been introduced by the inclusion of individuals with continuous eligibility of at least 1 year. Results from this study are therefore only generalisable to patients with venous thromboembolism who are covered by commercial or

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

Articles

Medicare Supplement insurance and have had continuous coverage for at least 12 months; generalisability to other populations is restricted. In conclusion, this retrospective, propensity scorematched, cohort analysis is, to our knowledge, one of the first US-based population studies to report a reduced risk of developing recurrent venous thromboembolism and major bleeding events with apixaban compared with rivaroxaban in patients being treated for secondary prevention of venous thromboembolism. Future studies with a larger sample size are needed to confirm these findings. Contributors GKD and HP conceptualised and designed the study. GKD analysed the data. GKD and HP interpreted the data, with assistance from JB and ED. The manuscript was written primarily by GKD; HP, JB, and ED provided assistance and contributed to revisions. All authors substantially contributed to this project, read and approved the manuscript, and assume responsibility for the contents of the manuscript. Declaration of interests We declare no competing interests. Acknowledgments Data used in this study were obtained from the Truven MarketScan Health Analytics under a licence to the University of Florida, College of Pharmacy and are not publicly available. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sector. GKD is a recipient of the American Association of Graduate Women (AAUW) fellowship for the year 2017–18. JB receives funding from the PhRMA Foundation. References 1 Goldhaber SZ, Bounameaux H. Pulmonary embolism and deep vein thrombosis. Lancet 2012; 379: 1835–46. 2 Lloyd-Jones D, Adams RJ, Brown TM, et al, on behalf of the American Heart Association Statistics Committee and Stroke Statistics Subcommittee. Heart disease and stroke statistics—2010 update: a report from the American Heart Association. Circulation 2010; 121: e46–215. 3 Weitz JI, Lensing AWA, Prins MH, et al. Rivaroxaban or aspirin for extended treatment of venous thromboembolism. N Engl J Med 2017; 376: 1211–22. 4 Lee LH. DOACs—advances and limitations in real world. Thromb J 2016; 14 (suppl 1): 17. 5 Sindet-Pedersen C, Pallisgaard JL, Staerk L, et al. Temporal trends in initiation of VKA, rivaroxaban, apixaban and dabigatran for the treatment of venous thromboembolism—a Danish nationwide cohort study. Sci Rep 2017; 7: 3347. 6 Kearon C, Akl EA, Ornelas J, et al. Antithrombotic therapy for VTE disease: CHEST Guideline and Expert Panel Report. Chest 2016; 149: 315–52. 7 Agnelli G, Buller HR, Cohen A, et al. Oral apixaban for the treatment of acute venous thromboembolism. N Engl J Med 2013; 369: 799–808. 8 The EINSTEIN Investigators. Oral rivaroxaban for symptomatic venous thromboembolism. N Engl J Med 2010; 363: 2499–510. 9 White RH, Garcia M, Sadeghi B, et al. Evaluation of the predictive value of ICD-9-CM coded administrative data for venous thromboembolism in the United States. Thromb Res 2010; 126: 61–67.

10 White RH, Chew HK, Zhou H, et al. Incidence of venous thromboembolism in the year before the diagnosis of cancer in 528,693 adults. Arch Intern Med 2005; 165: 1782–87. 11 Streiff MB. Thrombosis in the setting of cancer. Hematology Am Soc Hematol Educ Program 2016; 2016: 196–205. 12 Alotaibi GS, Wu C, Senthilselvan A, McMurtry MS. The validity of ICD codes coupled with imaging procedure codes for identifying acute venous thromboembolism using administrative data. Vasc Med 2015; 20: 364–68. 13 Zhu T, Martinez I, Emmerich J. Venous thromboembolism: risk factors for recurrence. Arterioscler Thromb Vasc Biol 2009; 29: 298–310. 14 Douketis JD, Foster GA, Crowther MA, Prins MH, Ginsberg JS. Clinical risk factors and timing of recurrent venous thromboembolism during the initial 3 months of anticoagulant therapy. Arch Intern Med 2000; 160: 3431–36. 15 Liabeuf S, Scaltieux LM, Masmoudi K, et al. Risk factors for bleeding in hospitalized at risk patients with an INR of 5 or more treated with vitamin K antagonists. Medicine 2015; 94: e2366. 16 Hughes M, Lip GY, Guideline Development Group for the NICE national clinical guideline for management of atrial fibrillation in primary and secondary care. Risk factors for anticoagulation-related bleeding complications in patients with atrial fibrillation: a systematic review. Q JM 2007; 100: 599–607. 17 Brown JD, Goodin AJ, Lip GYH, Adams VR. Risk Stratification for bleeding complications in patients with venous thromboembolism: application of the HAS-BLED bleeding score during the first 6 months of anticoagulant treatment. J Am Heart Assoc 2018; 7: e007901. 18 Yang D, Dalton J. A unified approach to measuring the effect size between two groups using SAS®. SAS Global Forum 2012; 35: 1–6. 19 D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998; 17: 2265–81. 20 Yang D, Dalton JE. A unified approach to measuring the effect size between two groups using SAS®. Paper 335-2012. 2012. httwrt.sas. com/resources/papers/proceedings12/335-2012.pdf (accessed Dec 3, 2018). 21 Cohen AT, Hamilton M, Mitchell SA, et al. Comparison of the novel oral anticoagulants apixaban, dabigatran, edoxaban, and rivaroxaban in the initial and long-term treatment and prevention of venous thromboembolism: systematic review and network meta-analysis. PLoS One 2015; 10: e0144856. 22 Mantha S, Ansell J. Indirect comparison of dabigatran, rivaroxaban, apixaban and edoxaban for the treatment of acute venous thromboembolism. J Thromb Thrombolysis 2015; 39: 155–65. 23 Sindet-Pedersen C, Staerk L, Pallisgaard JL, et al. Safety and effectiveness of rivaroxaban and apixaban in patients with venous thromboembolism—a nationwide study. Eur Heart J Cardiovasc Pharmacother 2018; 4: 220–27. 24 Kubitza D, Berkowitz SD, Misselwitz F. Evidence-based development and rationale for once-daily rivaroxaban dosing regimens across multiple indications. Clin Appl Thromb Hemost 2016; 22: 412–22. 25 Clemens A, Noack H, Brueckmann M, Lip GY. Twice- or once-daily dosing of novel oral anticoagulants for stroke prevention: a fixed-effects meta-analysis with predefined heterogeneity quality criteria. PLoS One 2014; 9: e99276. 26 Frost C, Nepal S, Wang J, et al. Safety, pharmacokinetics and pharmacodynamics of multiple oral doses of apixaban, a factor Xa inhibitor, in healthy subjects. Br J Clin Pharmacol 2013; 76: 776–86.

www.thelancet.com/haematology Published online December 4, 2018 http://dx.doi.org/10.1016/S2352-3026(18)30191-1

9