Comparative Clinical and Economic Outcomes Associated with Warfarin Versus Apixaban in the Treatment of Patients with Venous Thromboembolism in a Large U.S. Commercial Claims Database

BACKGROUND: Venous thromboembolism (VTE), constituting deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common cause of vascular-related morbidity and mortality, resulting in a significant clinical and economic burden in the United States each year. Clinical guidelines recommend that patients with DVT and PE without cancer should be initiated on anticoagulation therapy with a direct oral anticoagulant over a vitamin K antagonist. Yet there is limited real-world evidence comparing the economic burden of warfarin and apixaban in treating VTE patients in a large commercially insured population. OBJECTIVE: To compare safety and effectiveness of warfarin and apixaban and evaluate associated economic burden in treating VTE patients in a large U.S. commercial health care claims database. METHODS: The PharMetrics Plus database was used to identify oral anticoagulant (OAC)-naive patients aged ≥ 18 years who initiated apixaban or warfarin within 30 days of a qualifying VTE encounter and had continuous health plan enrollment with medical and pharmacy benefits for 6 months before treatment initiation. Apixaban initiators and warfarin initiators were matched using the propensity score matching (PSM) technique. Cox proportional hazard models were used to assess and compare the risk of major bleeding (MB), clinically relevant nonmajor (CRNM) bleeding, and recurrent VTE. Generalized linear models were used to assess and compare the all-cause health care costs. A 2-part model with bootstrapping was used to evaluate MB- and recurrent VTE-related medical costs. RESULTS: Among 25,193 prematched patients, 13,421 (53.3%) were prescribed warfarin and 11,772 (46.7%) were prescribed apixaban. After 1:1 PSM, 8,858 matched warfarin-apixaban pairs were selected with a mean follow-up of 109 days and 103 days, respectively. Warfarin was associated with a significantly higher risk of MB (HR = 1.52, 95% CI = 1.14-2.04), CRNM bleeding (HR = 1.27, 95% CI = 1017.15-1.40), and recurrent VTE (HR = 1.50, 95% CI = 1.24-1.82) compared with apixaban. Warfarin patients had significantly higher all-cause medical costs per patient per month (PPPM; $2,333 vs. $1,992; P = 0.001), MB-related costs PPPM ($112 vs. $65; P = 0.020), and recurrent VTE-related costs PPPM ($287 vs. $206; P = 0.014) compared with apixaban patients. Warfarin patients had similar all-cause total health care costs PPPM ($2,630 vs. $2,420; P = 0.051) compared with apixaban patients. CONCLUSIONS: Warfarin use was associated with a higher risk of MB, CRNM bleeding, and recurrent VTE compared with apixaban. Warfarin use was also associated with higher all-cause medical costs, MB-related medical costs, and recurrent VTE-related costs PPPM compared with apixaban.

in an average incremental direct medical cost of $18,000 to $23,000 per incident case among survivors. 10 Overall, the economic burden of VTE ranges between $13.5 and $27.2 billion per year (in 2011 U.S. dollars) in the United States. 11 According to the American College of Chest Physician guidelines, DVT and PE patients without evidence of cancer should be initiated on anticoagulation therapy with a direct oral anticoagulant (DOAC: dabigatran, rivaroxaban, apixaban, or edoxaban) over a vitamin K antagonist (VKA) such as warfarin. 12 This recommendation is supported with evidence from the AMPLIFY randomized clinical trial program that demonstrated that a 5-mg twice-daily fixed-dose oral regimen of apixaban alone (after administering apixaban 10 mg twice daily for 7 days) was noninferior to conventional therapy (subcutaneous enoxaparin, overlapped and followed by warfarin) and was associated with significantly less bleeding in VTE patients. 13 Similar to other DOACs, treatment with apixaban reduces the complexity of VTE treatment by eliminating the need for an initial parenteral anticoagulant (PAC) as required in conventional therapy. 14 A recent large-scale, real-world evaluation of VTE patients receiving outpatient treatment with apixaban or warfarin in U.S. clinical practice reported that patients treated with apixaban had a significantly lower risk of major bleeding (MB), clinically relevant nonmajor (CRNM) bleeding, and recurrent VTE compared with those treated with warfarin over a mean of 152-day follow-up but without economic assessment. 15 Furthermore, studies regarding the economic burden of VTE have been conducted but solely within the Medicare fee-for-service population. 16 The economic burden of warfarin versus apixaban in treatment of VTE has not been well established using commercially insured data in the literature. Therefore, the aim of this study was to compare safety and effectiveness of warfarin versus apixaban and evaluate the associated economic burden in treating VTE patients in a large, commercially insured U.S. population.

■■ Methods Data Source
This retrospective cohort study used the IQVIA PharMetrics Real-World Data Adjudicated Claims (PharMetrics Plus, PMTX+) database from March 1, 2014, to September 30, 2018. PMTX+ data were selected primarily due to national representativeness of the data over other available commercial claims databases. 17 It provides fully adjudicated health plan claims information for 150 million nonidentified lives in the United States, sourcing information from several commercial health plans in 50 states. It includes patients with 97% commercial, 2% Medicare Advantage, and 1% Medicaid coverage.

Patient Selection Criteria
Claims indicating a VTE diagnosis in primary or secondary position in an inpatient or outpatient setting during the identification period (September 1, 2014-September 30, 2018); the first VTE encounter was designated as the index VTE encounter N = 504,434 Claims for apixaban or warfarin during the 30-day period on or after the index VTE encounter during the identification period; the first claim of apixaban or warfarin was designated as the index date n = 82,601 Aged ≥ 18 years and had continuous health plan enrollment with medical and pharmacy benefits for 6 months before or on the index date n = 58,376 No claims indicating atrial fibrillation/atrial flutter or mechanical heart valve during the 6 months before or on the index date n = 51,374 No pharmacy claim for any OAC during the 6 months before or on the index date or during the period between the index VTE encounter and the index date n = 50,588 No claims indicating active malignancy 6 months before, or 30 days after, the index VTE encounter; a VTE event during the baseline period; or an inferior vena cava filter or pregnancy during the study period (March 1, 2014-September 30, 2018) n = 29,135 Among warfarin patients with their index VTE encounter in the outpatient setting, patients without evidence of PAC use 14 days before or after the index date were excluded n = 25,193 intracranial hemorrhage (ICH); and bleeding at other key sites (e.g., hepatic, splenic, and ocular hemorrhage; Appendix A, available in online article). 19 CRNM bleeding was identified using secondary diagnosis codes in the inpatient setting for noncritical sites for bleeding (not including patients with MB in the inpatient setting) or a diagnosis code for GI bleeding or other bleeding in the outpatient setting. 16 Recurrent VTE was defined by primary diagnosis in an inpatient setting, excluding admissions that occurred within 7 days of the index VTE encounter. 15 All-cause hospitalization was defined as an acute inpatient stay related to any cause at any time during the follow-up period.
Economic Outcomes. Economic outcomes, including all-cause inpatient costs, all-cause emergency room (ER) costs, all-cause office visits costs, all-cause outpatient costs, all-cause medical costs, all-cause pharmacy costs, all-cause health care costs, MB-related costs, and VTE-related costs were evaluated in this study. All-cause inpatient, ER, and office costs were defined as costs related to any inpatient, ER, or office stay respectively during the follow-up. All-cause outpatient costs were defined as the sum of costs in the outpatient, ER, or office setting during the follow-up. All-cause medical costs were defined as the sum of costs in inpatient and outpatient setting during the follow-up. All-cause pharmacy costs included costs for all prescriptions during the follow-up, not just for OAC medications. All-cause health care costs were defined as the sum of medical and pharmacy costs during the follow-up. MB-related medical costs were defined as hospitalization costs associated with the first MB event plus all subsequent bleeding costs occurring in the inpatient or outpatient setting (primary and secondary diagnoses). Recurrent VTE-related medical costs were defined as the hospitalization costs associated with the first recurrent VTE event plus all subsequent VTE costs in the inpatient or outpatient (primary and secondary diagnoses) setting. Because a fixed follow-up was not used, the total costs were calculated per patient per month (PPPM) and adjusted to 2018 U.S. dollars. The numerator for PPPM calculation was the aggregate of all costs in each category, and the denominator was the aggregated patient-months.

Statistical Methods
Bivariate comparisons of baseline characteristics were conducted using t-tests and chi-square tests. Propensity score matching (PSM) without replacement was conducted between the warfarin and apixaban cohorts to balance the patient characteristics between cohorts. Patients were matched 1:1 based on the propensity scores generated by logistic regression. Some of the baseline variables included in the PSM were demographics, VTE-related variables, Deyo-Charlson Comorbidity Index (CCI) score, baseline comorbidities, selected surgeries, and baseline medication use (Appendix B, available in online article). 20 The nearest neighbor method with a caliper of 0.01 was used to match patients. Covariate balance was checked through standardized differences, with a threshold of 10%; covariates with a standardized difference greater than 10% were considered unbalanced. 21 The incidence rates for the clinical outcomes were calculated as the number of events per 100 person-years. The first event per patient for each clinical outcome was included in the calculation. As all variables were balanced after PSM, OAC was the only independent variable included in the models. Cox proportional models were conducted to compare the risk of MB, CRNM bleeding, recurrent VTE, and all-cause hospitalization among the postmatch warfarin and apixaban cohorts.
The all-cause health care costs among the postmatched population were evaluated using generalized linear models with log-link and gamma distribution, wherein cost was the dependent variable.
Two-part models with bootstrapping were implemented for MB and recurrent VTE-related costs to account for a large number of zeros ($0) because very few patients had MB or recurrent VTE event costs. In the 2-part models, the first part was a logistic regression for the occurrence of the event (MB or recurrent VTE) and the second part was a generalized linear model regression of cost, conditional on the event. 22 To mitigate the impact of extreme cost outliers in the data, MB and recurrent VTE-related costs were adjusted using the winsorization method. Winsorization transforms the cost of the outliers to the preestablished percentile of the data. 23,24 Therefore, the event-related cost values below the 5th percentile were winsorized to costs in the 5th percentile, and costs above the 95th percentile were transformed to costs in the 95th percentile.

Sensitivity Analysis
There is variation in the literature surrounding the definition of recurrent VTE. 9,16,25-28 Therefore, to test the robustness of our findings, 3 sensitivity analyses were conducted surrounding the definition of recurrent VTE for both clinical and economic outcomes. First, recurrent VTE was defined to include events occurring within 7 days of the qualifying index VTE encounter. 26 Second, recurrent VTE was defined to include encounters in the inpatient and emergency room (ER) settings, excluding events within 7 days of the index VTE encounter. 28 Third, recurrent VTE was defined to include both encounters within 7 days of the index VTE encounter and those occurring in the ER setting. 16   [CI] = 1.14-2.04; P = 0.005), CRNM bleeding (HR = 1.27, 95% CI = 1.15-1.40; P < 0.001), and recurrent VTE (HR = 1.50, 95% CI = 1.24-1.82; P < 0.001) compared with apixaban. The risk of all-cause hospitalization was similar for warfarin patients compared with apixaban patients (HR = 1.08, 95% CI = 0.98-1.18; P = 0.113; Figure 2).

Sensitivity Analyses
Both the clinical and economic outcomes from all 3 sensitivity analyses were consistent with the main analysis (Table 3). Recurrent VTE incidence and recurrent VTE-related costs were significantly higher in warfarin patients compared with apixaban patients when recurrent VTE was defined to include events ( Figure 1). Patients prescribed warfarin had higher mean CCI scores, more selected surgeries, and more history of falls during the baseline period compared with apixaban in the prematched population (Appendix B, available in online article). After 1:1 PSM, 8,858 warfarin-apixaban pairs were selected with a mean follow-up of 109 days and 103 days, respectively (Table 1).
After PSM, the standardized differences of the means of the matched variables were less than 10%, indicating that the covariates were balanced between treatment groups at baseline. Apixaban-and warfarin-matched patients had a mean age of 52.5 years and a mean CCI score of 0.8. Approximately 46% of patients had an index VTE encounter in the inpatient setting, 55% of patients had a diagnosis of DVT only, and 31% of patients had a provoked VTE diagnosis (Table 1).

Clinical Outcomes
In the PSM-matched population, the incidence rates of MB (4. Warfarin was associated with a significantly higher risk of MB (hazard ratio [HR] = 1.52, 95% confidence interval  apixaban-treated patients. The inconsistency in the recurrent VTE risk could be attributed to study setting (controlled trial vs. real world), differences in the methods employed to identify eligible VTE patients (diagnoses via imaging techniques vs. ICD-9-CM and ICD-10-CM codes), and variable follow-up periods. and definition of recurrent VTE (clinical assessment including fatal and nonfatal VTE vs. a primary diagnosis in an inpatient setting). Additionally, the AMPLIFY trial was designed to assess noninferiority in a clinical trial setting, whereas the present real-world study has a much larger sample size for detecting association.

Propensity Score-Matched Incidence Rates and Hazard Ratios of Major Bleeding, CRNM Bleeding, Recurrent VTE, and All-Cause Hospitalization for Warfarin Versus Apixaban
Other retrospective studies comparing DOACs to warfarin have shown similar results. 15,19 Using a commercial pooled database, VTE patients who initiated apixaban were associated with a significantly lower risk of MB (HR = 0.75, 95% CI = 0.64-0.87), CRNM bleeding (HR = 0.77, 95% CI = 0.71-0.83), and recurrent VTE (HR = 0.80, 95% CI = 0.70-0.91) compared with warfarin patients, which is consistent with our study. 15 In another study using a Medicare population, warfarin was associated with a significantly higher risk of MB (HR = 1.31, 95% CI = 1.10-1.57) and CRNM bleeding (HR = 1.31, 95% CI = 1.19-1.43; P < 0.001), but a similar risk of recurrent VTE (HR = 0.96, 95% CI = 0.70-1.33; P = 0.825), compared with apixaban patients. 19 Although the MB and CRNM bleeding risks are consistent with the Medicare study, the risk of recurrent VTE is not. This could be attributed to the different health insurances (i.e., Medicare vs. commercial insurance) or the differences in within 7 days of the index VTE encounter, events in the ER setting, and events both within 7 days of index VTE encounter and the ER setting in aggregate.

■■ Discussion
This retrospective cohort analysis evaluated the clinical and economic burden associated with VTE among patients treated with warfarin and apixaban in a younger, commercially insured population compared with the published literature. 15,16 Warfarin was associated with a higher risk of MB, CRNM bleeding, and recurrent VTE among VTE patients compared with patients prescribed apixaban. In addition to observed clinical benefit, warfarin use was also associated with higher all-cause medical costs PPPM, and MB-and recurrent VTE-related costs PPPM compared with apixaban use. All-cause health care costs were similar among warfarin and apixaban patients.
The results from this study add evidence to supplement results from the AMPLIFY trial where a benefit of reduced risk of MB (0.6% vs. 1.8%; relative risk = 0.31 [0.17-0.55]) and CRNM bleeding (3.8% vs. 8.0%; relative risk = 0.48 [0.38-0.60]) was observed when patients were treated with apixaban compared with conventional therapy (subcutaneous enoxaparin followed by warfarin). 13 The trial also reported a reduced risk of recurrent VTE in apixaban patients; however, the association was not statistically significant (HR = 0.84, 95% CI = 0.60-1.18). In this study, in addition to MB and CRNM bleeding, recurrent VTE was also significantly higher for warfarin-compared with

All-Cause Health Care Costs in the Propensity Score-Matched Cohorts
Comparative Clinical and Economic Outcomes Associated with Warfarin Versus Apixaban in the Treatment of Patients with Venous Thromboembolism in a Large U.S. Commercial Claims Database the data structure. The Medicare study also included higherrisk older patients with a mean age of 78 years, compared with a younger population with a mean age of 52 years in this study. Additionally, the current study used more recent data (until September 2018) compared with the Medicare study (until December 2016), which could have affected the outcomes.
Fewer studies have evaluated the economic burden as well as the impact of complications including recurrence of VTE and MB among patients who initiated oral anticoagulants in a younger, commercially insured VTE population. Lin et al. (2014) found that among Medicare and commercially insured VTE patients, recurrent VTE rates were relatively high (11.4% and 15.4%, respectively). 9 A systematic review had estimated that managing an initial episode of acute VTE costs between $12,000 and $15,000, with subsequent complications increasing costs to between $18,000 and 23,000 per incident case. 10 Additionally, the authors found that the primary driver for health care costs was the use of inpatient care, which is similar to the findings of the present study. 10 They concluded that among the VTE patient population, the total health care costs for patients with a recurrent VTE event were approximately triple those for patients who did not experience a recurrent VTE event. 9 The above-mentioned study using the Medicare population found that all-cause ($3,267 vs. $3,033; P < 0.001) and MB-related costs ($147 vs. $75; P = 0.003) were significantly higher for warfarin patients whereas recurrent VTE related costs were similar ($30 vs. $36; P = 0.516) compared with apixaban patients. 19 In this study, warfarin patients had significantly higher MB-and recurrent VTE-related costs compared with apixaban patients using winsorization. A similar trend was seen without applying winsorization; warfarin trended toward higher MB-and recurrent VTE-related costs; however, the MB-related costs were not significant. One likely reason for this may be due to the extreme cost outliers and the smaller sample size associated with MB and recurrent VTE events.

Limitations
To the best of our knowledge, this retrospective analysis is the first study that evaluated economic burden associated with VTE and recurrence of VTE in a commercially insured population. It provides a comprehensive picture of how medications are used by clinicians in routine practice. One of the main strengths of the study is that it has a wide geographic U.S. representation and can be generalized to the U.S. commercially insured population younger than aged 65 years.
There are some inherent limitations to this study and the nature of observational research. Causal relationships cannot be reported; only associations could be assessed. Additionally, administrative claims data are collected for purposes other than research, and the analysis is constrained by diagnosis and procedure codes. The presence of a diagnosis code on a medical claim is not a positive indication of the presence of disease, as the diagnosis code may be incorrectly coded or included as rule-out criteria rather than actual disease. Over-the-counter use of medications such as aspirin could not be captured which could affect the treatment patterns associated with the study anticoagulants.
This study used ICD-9-CM and ICD-10-CM codes to identify DVT and PE, which is different from the imaging assessments used in clinical trials; this may lead to differences in patient populations. Also, although cohorts were matched using PSM, there remains the potential for residual confounding, such as warfarin dose adjustment and over-the-counter use of aspirin, which was not available in the database. Finally, these results may not be generalizable to those who are uninsured, the Medicare population, or other segments of the U.S. population.

■■ Conclusions
In this commercially insured population, warfarin use among VTE patients was associated with higher risk of MB, CRNM bleeding, and recurrent VTE compared with apixaban use.  These results are largely consistent with available clinical trial and recently published real-world data. Warfarin was also associated with higher medical health care costs, higher MB-related costs, and higher recurrent VTE-related costs compared with apixaban. These results may help clinicians, stakeholders, and decision makers in identifying safe and effective treatment for VTE as well as potentially reducing economic burden. 0