Effectiveness, safety, and healthcare costs associated with rivaroxaban versus warfarin among venous thromboembolism patients with obesity: a real-world study in the United States

Prior observational studies suggest rivaroxaban is safe and effective among patients with morbid obesity who suffered a venous thromboembolism (VTE) event, but existing data are more limited in the broader population of VTE patients with obesity. This study assessed VTE recurrence, major bleeding, healthcare resource utilization, and healthcare costs among VTE patients with obesity who received rivaroxaban versus warfarin. VTE patients with obesity who initiated rivaroxaban or warfarin after a first VTE (index date) were identified from the IQVIA PharMetrics® Plus database (01/02/2011–09/30/2019). The follow-up period spanned from the index date until health plan disenrollment, end of data availability, cancer diagnosis/treatment, end of the 12 month post-index period, or (for the analysis of major bleeding) anticoagulant discontinuation or switch. Patient characteristics were balanced using inverse probability of treatment weighting. The weighted rivaroxaban (N = 8666) and warfarin cohorts (N = 5946) were well balanced (mean age = 51 years, females = 52%). Over a 9.6 months mean observation period, rivaroxaban users had a significantly lower risk of VTE recurrence [7.0% vs. 8.2%, HR(95% CI) = 0.85(0.75;0.97)] and a similar risk of major bleeding [4.1% vs. 3.6%, HR(95% CI) = 1.11(0.89;1.37)] relative to warfarin users at 12 months. Relative to warfarin users, rivaroxaban users had significantly fewer all-cause outpatient visits [RR(95% CI) = 0.71(0.70;0.74)]. The higher pharmacy costs incurred by rivaroxaban recipients (cost difference = $1252) were offset by lower medical costs (cost difference = − $2515, all p < 0.05) compared with warfarin recipients. Our findings suggest that rivaroxaban is safe and effective versus warfarin, and associated with lower medical costs among VTE patients with obesity. Supplementary Information The online version contains supplementary material available at 10.1007/s11239-022-02661-1.


Introduction
Obesity is a serious public health issue that affects a growing number of individuals in the United States (US) [1]. In 2008, US medical costs associated with obesity were estimated at $147 billion [2]. The condition is associated with a chronic hypercoagulable state that increases the risk of venous thromboembolism (VTE) by at least two fold [3,4].
Direct acting oral anticoagulants (DOACs) are the standard treatment to prevent VTE recurrence in patients who previously experienced a VTE event [5]. Until recently, data on their use in obese population were more limited, and current labeling information does not recommend any dose adjustments in VTE patients with obesity. Mounting evidence from observational studies suggests DOACs may be a safe and effective alternative to vitamin K antagonists (VKAs; such as warfarin) in patients with morbid obesity (i.e., BMI ≥ 40 kg/m 2 ) [6][7][8][9][10][11], and some studies further support that they may have a similar profile in the broader population of patients with overall obesity (i.e., BMI ≥ 30 kg/ m 2 ) [12][13][14][15][16]. Rivaroxaban may be particularly effective in VTE patients with morbid obesity [11]. However, data on the comparative safety and effectiveness of rivaroxaban versus warfarin as a VTE treatment are limited in the broader population of patients with overall obesity. Furthermore, healthcare costs associated with the use of rivaroxaban and warfarin remain uncertain in this population. To fill this knowledge gap, the current study sought to assess VTE recurrence, major bleeding, healthcare resource utilization (HRU), and healthcare costs among patients with obesity who had an acute VTE event and received treatment with rivaroxaban or warfarin.

Data source
Patients were identified from IQVIA PharMetrics® Plus database (study period: 11/02/2011-09/30/2019; IQVIA database). This database comprises data on enrollees and is representative across US regions, with medical and pharmacy benefits available in any given recent year. Information on ~ 40 million patients is available and is generally representative of the less-than-65 years of age, commerciallyinsured population with respect to both age and sex. The IQVIA database contains information on demographics; plan enrollment; and inpatient, outpatient, and pharmacy claims and associated costs. Available data are fully deidentified and therefore compliant with the Health Insurance Portability and Accountability Act.

Study design and study population
A retrospective, observational cohort study was conducted. Index date was defined as the date of initiation of rivaroxaban or warfarin ≤ 30 days after a first VTE event [i.e., ≥ 1 medical claim with VTE diagnosis in any position (Table S1 and S2 for codes)]. The baseline period was defined as the 12-month period before the index date; patient characteristics were evaluated during this period. Patients with an index date on or after 01/01/2014 were included to account for potential differences associated with early adopters of rivaroxaban (approved 11/2012 by the Food and Drug Administration) and delays before the wider use of the drug. Patients were additionally required to have: (1) ≥ 12 months of continuous health plan enrollment pre-index date, (2) ≥ 1 medical claim with a diagnosis of obesity/BMI ≥ 30 kg/ m 2 (Table S3 for BMI-related International Classification of Diseases [ICD] codes) during the baseline period or on index date, and (3) ≥ 18 years old at index date.
The following exclusion criteria were applied: (1) presence of claims for multiple oral anticoagulants on index date, (2) ≥ 1 pharmacy claim for an oral anticoagulant during the baseline period, (3) ≥ 1 medical claim for VTE before the first VTE event during the baseline period, (4) recurrent VTE after the first observed VTE event but prior to the index date, (5) knee or hip replacement surgery during the baseline period, (6) ≥ 1 medical claim with a diagnosis of atrial fibrillation during the baseline period, (7) cancer diagnosis and treatment during the baseline period or on the index date.
Effectiveness, HRU, and costs were assessed using an intention-to-treat (ITT) approach, whereas safety was assessed using an on-treatment approach. The ITT follow-up spanned from the index date until health plan disenrollment, end of data availability, presence of both cancer diagnosis and treatment (at the later of the two dates), or 12 months, whichever came first. The on-treatment follow-up was censored similarly to the ITT approach and additionally upon anticoagulant discontinuation or anticoagulant switch, so patients were continuously treated with the index anticoagulant. Treatment discontinuation was defined as a gap of ≥ 60 days of supply between the end of an anticoagulant dispensing and the next medication refill or end of data availability. Effectiveness was also assessed using an ontreatment approach in a sensitivity analysis.

Study outcomes
The effectiveness outcome was VTE recurrence, defined as a hospitalization with a primary diagnosis of VTE. Major bleeding was the safety outcome and was identified using hospitalizations with indicators (diagnoses and procedures) of a bleeding episode based on the Cunningham algorithm [17].
All-cause and VTE-related HRU and healthcare costs were assessed during follow-up. HRU outcomes included hospitalizations and days of hospital stay, emergency room (ER) visits, and outpatient visits. Outpatient visits were further broken down into office, outpatient hospital, and other outpatient (including patient home and other unlisted facilities). All-cause healthcare costs included medical and pharmacy costs, with medical costs further broken down into the same categories as HRU. VTE-related HRU and costs were 1 3 defined as visits/costs with primary or secondary diagnosis (i.e., identified in any other diagnosis fields) of VTE [18].

Statistical analysis
Inverse probability of treatment weighting (IPTW) was used to balance the characteristics of cohorts. IPTW uses weights derived from the propensity score (PS) to create pseudopopulations, so that covariates are distributed independently of treatment assignment. PS was defined as the conditional probability of receiving rivaroxaban based on observable covariates. The following covariates were included in the PS estimation: age, sex, year of index date, region, type of insurance plan, morbid obesity (i.e., BMI ≥ 40 kg/m 2 ), time between VTE event and index date, type of VTE event, baseline major bleeding, cardiovascular-related medications, cardiovascular procedures, use of non-oral anticoagulants, number of unique prescription drugs used during baseline, baseline healthcare resource utilization and costs, and baseline risk factors for VTE and bleeding events (with ≥ 1% prevalence in either cohort). Weights were truncated at the 99% of the distribution to limit the effect of extreme weights. The balancing of patient baseline characteristics was assessed using standardized differences, with a threshold < 10% considered not clinically meaningful [19].
Weighted Kaplan-Meier (KM) survival analysis was used to assess time to VTE recurrence and time to major bleeding events. Cumulative KM rates were reported at 12 months post-index date. Rates of VTE recurrence and major bleeding were compared between cohorts using weighted Cox proportional hazards regression models; corresponding hazard ratios (HR), 95% confidence intervals (CI), and p-values were reported.
HRU and healthcare costs were evaluated per patientyears (PPY) to account for the variable duration of followup among individual patients. Rates of HRU were compared between cohorts using rate ratios (RR) obtained from Poisson regression models. Costs were compared between cohorts using mean cost differences. Costs were inflated to 2019 US dollars using the medical care component of the Consumer Price Index. Non-parametric bootstrap procedures were used to estimate 95%CI and p-values for comparisons, since HRU and cost data have positive values that follow a non-normal distribution and commonly include zero values.

Recurrence of venous thromboembolism and major bleeding
Based on an ITT approach, patients in the rivaroxaban cohort had a significantly lower risk of VTE recurrence than those in the warfarin cohort at  Patients who received rivaroxaban incurred significantly lower total all-cause medical costs than those who received    Fig. 3B and Table S4). Similar trends were observed for VTE-related healthcare costs (Table S4).

Discussion
In this retrospective study based on health insurance claims data, rivaroxaban was associated with a statistically AF atrial fibrillation, BMI body mass index, CAD coronary artery disease, COPD chronic obstructive pulmonary disease, DVT deep vein thrombosis; NSAID nonsteroidal anti-inflammatory drugs, PAD peripheral artery disease, Quan-CCI Quan-Charlson comorbidity index, RIETE registro informatizado enfermedad tromboembolica, SERM selective estrogen receptor modulator, Std. diff standard difference, VTE venous thromboembolism a Rivaroxaban and warfarin patients were weighted using the inverse probability of treatment weighting approach based on the propensity score significant reduction in VTE recurrence relative to warfarin among VTE patients with obesity. Likewise, the rates of major bleeding were similar between both cohorts. Furthermore, rivaroxaban was associated with statistically significant lower all-cause medical costs and a nonsignificant numerically lower all-cause total healthcare costs vs. warfarin.
The results of this study build on the existing literature. In the EINSTEIN-DVT trial (rivaroxaban vs. enoxaparin/ VKA for DVT), patients were stratified based on weight rather than BMI (the highest weight category was > 90 kg), and the number of VTE recurrences was low in both arms among patients > 90 kg (rivaroxaban:11/491, enoxaparin/ VKA:11/486) [20]. A similar weight stratification was used in the EINSTEIN-PE trial (rivaroxaban vs. enoxaparin/VKA for PE); likewise, the number of recurrences was low in patients > 90 kg (rivaroxaban:13/683, enoxaparin/ VKA:10/672) [21]. Therefore, despite the wealth of data provided by the EINSTEIN trials, the relative safety and efficacy of rivaroxaban versus warfarin remains uncertain in the obese population.
Several observational studies subsequently evaluated the safety and effectiveness of rivaroxaban in patients with morbid obesity [7][8][9]13]. However, to the best of our knowledge, only Costa et al. addressed this research question in the broader population of patients with obesity (i.e., BMI ≥ 30 kg/m 2 ) rather than patients with morbid obesity (i.e., BMI ≥ 40 kg/m 2 ) [13]. Consistent with the current study, the authors found that the rates of VTE recurrence were significantly lower in rivaroxaban-treated patients than warfarin-treated patients and that the rates of major bleeding were similar [13]. Interestingly, the effect size observed for VTE recurrence at 12 months was larger in the study by  [13]. This might be driven by the severity of VTE events in the two studies; in the present study, 49.5-50.8% of patients experienced a PE (with or without DVT) whereas this proportion was only 20.7-24.4% in the Costa et al. study [13].
The risk of VTE recurrence has been shown to increase linearly with BMI, with each 1-unit increase in BMI associated with a 4.4% higher risk of VTE recurrence [22]. Although the present study focused on a more inclusive population of VTE patients with obesity (rather than VTE patients with morbid obesity), the risk of VTE observed (~ 7-9% at 12 months) was within the range observed in previous studies of patients with morbid obesity (~ 1-17%) [6][7][8][9], but inconsistent length of follow-up and differences in patient characteristics limit comparisons across studies. Notwithstanding this limitation, this suggests that the risk of recurrent VTE may remain substantial in the broader population of VTE patients with obesity. Taken together, the results of the current study are consistent with the growing body of literature which suggests that rivaroxaban is a safe and effective option to reduce the risk of VTE recurrence in this population [23].  In the current study, patients initiated on rivaroxaban incurred significantly lower medical costs than those initiated on warfarin, which offset the higher pharmacy costs associated with rivaroxaban. The major driver of the difference in medical costs were outpatient costs, which may be lower among rivaroxaban users due to DOACs not requiring international normalized ratio monitoring [24]. Likewise, this may also explain the large difference in outpatient visits, which were more than two times less frequent among rivaroxaban users compared with warfarin users. These data suggest that rivaroxaban may be a cost-neutral alternative to VKAs among VTE patients with obesity.

Limitations
The present study should be interpreted considering some limitations inherent to the retrospective nature of the analysis. First, height and weight data are not available in health insurance claims; thus, obesity was identified using ICD-9-CM and ICD-10-CM codes for high BMI rather than actual BMI values. Because of this, some VTE patients with obesity may not have been captured. However, research has shown that patients with codes for obesity are likely to be obese (i.e., high positive predictive value and high specificity) [25][26][27][28]. Second, VTE recurrences were defined based on diagnosis recorded during a hospitalization; therefore, recurrences in the outpatient setting were not captured. Third, mortality data were not available. Fourth, coding inaccuracies in administrative claims data may have led to the misidentification of some patients, although this limitation is expected to similarly impact both study cohorts. Similarly, the database may not contain information on all medications, particularly those administered in inpatient settings and over-the-counter medications (e.g., aspirin). Fifth, it was not possible to know whether all tablets supplied were actually taken by the patients. Sixth, while IPTW mitigated the risk of confounding due to observed variables, unmeasured confounders may have impacted results. Seventh, confounding by indication cannot be eliminated, because of the lack of randomization in an observational study. However, careful choice of study design and patient inclusion/exclusion criteria can help mitigate the potential risk of selection bias. Eighth, patients included were working age adults with commercial insurance; thus, results may not be generalized to other populations. Lastly, healthcare costs were assessed from the payer's perspective and do not include indirect costs (e.g., productivity costs).

Conclusion
In this retrospective cohort study, VTE patients with obesity that were initiated on rivaroxaban had a lower risk of VTE recurrence and a similar risk of major bleeding compared with those initiated on warfarin. The higher pharmacy costs associated with rivaroxaban were fully offset by reduced medical costs, resulting in similar total healthcare costs between rivaroxaban and warfarin users. Altogether, these data suggest that rivaroxaban is a safe, effective, and cost-neutral alternative to VKAs among VTE patients with obesity.