Comparative Effectiveness of Anticoagulants in Patients With Cancer-Associated Thrombosis

Key Points Question What are the patterns of anticoagulant utilization and the anticoagulants associated with the lowest risk for venous thromboembolism (VTE) recurrence in patients with cancer in a clinical setting? Findings In this comparative effectiveness study with 5100 adult patients, twice as many patients were prescribed direct oral anticoagulants (DOACs) as other classes, and cancer type was associated with the choice of anticoagulant prescription. Use of DOACs was associated with a 50% risk reduction in VTE recurrence compared with low-molecular-weight heparin (LMWH) and warfarin and a 60% risk reduction in all-cause mortality compared with LMWH; DOACs were also associated with reduced risk of major bleeding and gastrointestinal tract bleeding compared with LMWH. Meaning In this study, DOACs were associated with a higher persistence rate, lower risk of VTE recurrence, lower risk of major bleeding, and improved mortality.


Introduction
Management of cancer-associated thrombosis (CAT) is complicated owing to several variables, including cancer-specific thrombotic and bleeding risk, cancer treatment-associated complications, frequent invasive procedures, and constitutional adverse effects such as nausea, vomiting, and anorexia, which may impact medication absorption and adherence. These add to the risk of venous thromboembolism (VTE) recurrence and major bleeding, which both carry high case fatality rates.
Treating patients with cancer-associated VTE is therefore challenging due to the delicate balance between these extremes. 1 For nearly 20 years, guideline recommendations for patients with cancer-associated VTE was low-molecular-weight heparin (LMWH) 2-6 based on the results of the CLOT trial. 7 Trials of other LMWH preparations, however, were not able to duplicate these results. [8][9][10][11] More recently, 4 randomized clinical trials (RCTs) have found that direct oral anticoagulants (DOACs) offer a reasonable alternative to parenteral dalteparin for the acute treatment of cancer-associated VTE with acceptable efficacy and safety outcomes. [12][13][14][15] As such, DOACs have received guideline endorsement for treating acute VTE in this setting. 16,17 Despite these findings and guideline statements, warfarin has remained a common treatment strategy for community-based oncology practices for various reasons, including cost and patient preference for oral over parenteral medications. 18,19 The comparative utilization of these 3 classes of anticoagulant in clinical oncology practices has not been thoroughly explored.
DOACs decrease VTE recurrence and major bleeding as compared with LMWH in clinical trial settings. 13 Whether DOACs are more effective and safer than LMWH and how they compare with warfarin in a clinical oncology context is not established. Thus, we used claims-based data from OptumLabs to assess the utilization patterns and the comparative efficacy and safety of these available anticoagulant classes.

Methods
This comparative effectiveness study was reported in accordance with the Professional Society for Health Economics and Outcomes Research (ISPOR) reporting guideline for comparative effectiveness research. 20 Deidentified administrative claims data from OptumLabs Data Warehouse (OLDW) 21 were queried to identify patients with active cancer and acute VTE from January 1, 2012, to September 30, 2019. The Mayo Clinic institutional review board exempted this study from review and the requirement for informed consent due to the analysis of preexisting, deidentified data. 30 days after the VTE date. Patients were then categorized into 1 of 3 groups ([1] DOAC, [2] LMWH, or [3] warfarin) based on the initial prescription filled. The first fill date of a specific anticoagulant was defined as the index therapy and treatment date. Study drug discontinuation was defined as not refilling a medication after 30 days of the end of last treatment episode, which is calculated based on fill date and supply.
Patients who crossed over to a different anticoagulant within the first year were excluded from the analysis. Patients were also excluded from the analysis for any of the following reasons: (1) crossed over to a different anticoagulant within the first year; (2) prior history of VTE; (3) filled prescription for an oral anticoagulant (warfarin and DOAC) less than 1 year prior to the VTE index date; or (4) less than 1 year of continuous insurance coverage prior to the VTE index date. Eligible individuals with missing data were removed and were not included in statistical analyses. Detailed inclusion and exclusion criteria are outlined in the eMethods in Supplement 1.

Follow-Up
Follow-up originated at the VTE index date and continued until the end of treatment. This was defined as (1)

Outcomes of Interest
The main efficacy end points included any VTE recurrence and all-cause mortality. The main safety end points included any episode of major bleeding and sites of bleeding (GI, GU, intracranial bleeding) (eTable 2 in Supplement 1). 22

Statistical Analysis
Analyses were performed between April 2020 and September 2021. Baseline characteristics (including but not limited to cancer type, presence of metastatic disease, baseline intervention [chemotherapy and surgery], baseline comorbidities, and Charlson Comorbidity Index) of the treatment cohorts were reported. Multinomial logistic regression was used to assess factors associated with use of DOAC relative to other anticoagulants (LMWH and warfarin) and presented as odds ratios (ORs) and 95% CIs. Kaplan-Meier curves were plotted to assess the differences in time to medication discontinuation among the 3 groups.
Propensity score (PS) with inverse probability of treatment weighting was used to balance differences in baseline characteristics among the 3 treatment groups. 23 All baseline characteristics listed were included in the PS models to derive the PS and the average treatment effect weights (Table 1; eTable 3 in Supplement 1). The standardized mean difference was used to assess the balance of covariates, and a standardized mean difference less than 0.1 was considered acceptable.
Weighted Cox proportional hazards regression with a robust variance estimator was used to assess outcomes. The event rates per 100 person-years and hazard ratios (HRs) were calculated, and the cumulative incidence curves were plotted. P < .05 was considered statistically significant for all 2-sided tests. Sensitivity analyses were also conducted on the cohort with index dates between

Population Characteristics
A total of 5100 patients were included (mean [

Medication Utilization Patterns
By multinomial regression analysis, younger patients were more likely to be prescribed LMWH (OR per 1-year, 0.97; 95% CI, 0.97-0.98; P < .001) ( Table 2). Both LMWH and warfarin were more likely to be prescribed for patients with lung, urological, gynecological, and colorectal cancer. LMWH was more likely to be prescribed for patients with musculoskeletal and brain cancer, while warfarin was more likely to be prescribed in patients with breast cancer compared with DOACs.
Warfarin was more likely to be prescribed for patients with DVT alone (OR, 1.31; 95% CI,  Table 2 and eTable 4 in Supplement 1. The results of sensitivity analyses were consistent and are provided in eTable 6 in Supplement 1.

Medication Comparative Effectiveness in Weighted Cohorts
The results for comparative effectiveness in the weighted cohort are highlighted in  Figure 2B.  was small and precluded any meaningful statistics. Additionally, post hoc sensitivity analysis for GI

DOAC LMWH Warfarin
Study drug discontinuation was defined as not refilling a medication after 30 days of the end of last treatment episode, which was calculated based on fill date and supply. The 1-month drop observed here is likely due to the allowed 30-day gap. DOAC indicates direct oral anticoagulant; LMWH, low-molecular-weight heparin.
bleeding in upper GI malignant neoplasms showed no significant differences among anticoagulants (eTable 7 in Supplement 1).

Reconciliation With Prior Clinical Data
A literature search was conducted to identify similar studies and to reconcile the findings of our study with the results of those studies. Findings are shown in eFigure 3 in Supplement 1.

Discussion
Consistent with the general theme of recent RCTs, results from this claims-based cohort of more than 5000 patients include a general preference for DOAC therapy use, with nearly twice as many patients receiving this class of medication compared with other classes. Furthermore, these data reinforce the general efficacy and safety of DOACs in this patient population; they were associated with a nearly 50% reduction in VTE recurrence rates and a more than 2-fold reduction in hospitalization for major bleeding compared with LMWH therapy. The odds of GI and intracranial bleeding were likewise reduced among patients receiving DOACs. Not seen in clinical trials, these data showed an association between DOAC therapy and a significant 60% reduction in all-cause mortality rates relative to LMWH. As such, it is anticipated that these data will help facilitate shared decision-making and inform clinical guidelines for the treatment of such patients.
DOACs have emerged as the most prescribed anticoagulant choice for management of cancerassociated VTE. 24 Half of patients in this cohort were treated with a DOAC and less than one-third received LMWH (29.1%) despite contemporary guideline recommendations. 2,5,6 The timeframe of this study (2012-2019) antedated guideline changes endorsing DOAC use in this context. 2,3 However, DOAC popularity might be attributed to pharmacologic benefits including rapid onset of action, convenient oral administration and dosing, short half-life, lack of monitoring, few drug or food interactions, and good bioavailability. 25,26 Compared with LMWH, patient preference for oral apixaban, for example, resulted in fewer discontinuations in a recent RCT of cancer-associated acute VTE treatment. 13 By comparison, warfarin was the least frequently prescribed anticoagulant in this cohort (28.6%). While some studies have shown a decreasing use of warfarin, others have not. In some community oncology practices, nearly half of patients with cancer and acute VTE are still treated with warfarin. 18,19,24 Guideline recommendations preferring LMWH, specifically dalteparin, over warfarin are based on a single trial. 7 Trials of other LMWH preparations have not shown an advantage over warfarin. [8][9][10][11] Warfarin was associated with improved overall survival compared with LMWH in a recent study assessing 9706 propensity score-matched patients with cancer and VTE. 27 Despite a clear benefit for DOACs over LMWH in the current study, neither major bleeding nor survival outcomes favored DOACs over warfarin; VTE recurrence rates were lower with DOACs.
This study revealed low persistence rates of patients with cancer receiving anticoagulant therapy, with a minority continuing treatment beyond 3 months, similar to findings in another analysis of Medicare data. 27 Conversely, Guo et al 28  Contrarily, several meta-analyses have suggested increased risk of major bleeding with DOACs compared with LMWH; however, these results are based on limited data and have wide confidence intervals. 40,41,44,45 Recurrent VTE also presents with an increased risk of mortality, especially in patients with prior history of PE. No anticoagulants to date have been able to achieve overall survival benefit. [7][8][9][10][11][12][13][14][15]44,45 However, contrary to RCTs, which included near-equal proportion of patients with metastatic disease as our study, we observed lower all-cause mortality with DOACs and an increased risk with LMWH compared with DOACs. These findings indicate that most of these RCTs might have been underpowered to detect such differences. Consistently, another study showed a significant reduction in mortality with rivaroxaban when compared with enoxaparin. 46 Taken together, these findings reassure and reinforce prior evidence in terms of VTE risk reduction and may be suggestive of lower risk of mortality with DOACs.
The current study has several strengths. We relied on claims data from OLDW, which contain longitudinal health information on enrollees and patients, representing a diverse mixture of ages, races and ethnicities, and geographical regions across the US; conducted multinominal regression analysis to assess factors associated with treatment patterns of utilization; applied propensity score matching for adjustment of differences across baseline sociodemographic and clinical characteristics; assessed comparative effectiveness of DOACs, LMWH, and warfarin using weighted Cox proportional hazard models; and reconciled our design and findings with previous clinical studies to compare differences and consistency between results (eFigure 3 in Supplement 1).

Limitations
This study has limitations, including information bias (billing inaccuracies and data omissions), the use of ICD codes to identify patients with VTE, and the lack of radiological evidence of VTE in the database, which can potentially lead to classification bias for assessment of VTE. The analyses were conducted using US claim-based data so the results could not be extrapolated to other populations.
We lacked information on uninsured patients or patients receiving insurance from other federal-or state-regulated insurances; hence, the results may not be representative of such populations. Only observable uncontrolled covariates were accounted for in adjusted multivariate analyses; hence, there is a risk of residual confounding bias. Likewise, this analysis predates the pivotal RCTs; therefore, selection bias in the use of different drugs is likely. The proportion of patients still receiving treatment was used to reflect patient adherence to different medications and assumed that medications supplied were being used, which may not be reflective of true patient adherence.
Clinically relevant nonmajor bleeding, which has competing risks among different anticoagulant treatments, was not assessed and may alter the choice of anticoagulant therapy. We assessed all-cause mortality and not VTE or bleeding-specific mortality, which might be more informative to guide the choice of anticoagulant therapy in patients with CAT. We used propensity score matching to address differences in baseline characteristics among the 3 treatment groups. However, it is important to note that there may be additional confounding variables that were not accounted for in outcome assessment. Therefore, careful consideration is warranted when interpreting the results of this study.

Conclusions
In this study, patients with cancer-associated VTE received anticoagulation therapy for a short duration in clinical practice. The findings suggest that DOACs and warfarin may offer better treatment persistence than LMWH in clinical practice. Warfarin may still be considered for patients with contraindications to DOACs and for those who have poor persistence on LMWH.