A trial fibrillation (AF) is a common condition that causes cardiac rhythm disturbance because of abnormal impulse formation, which can lead to numerous heart-related complications.1
AF is one of the key risk factors for ischemic stroke, increasing the risk up to 5-fold.2
In 2010, the prevalence of AF in the United States was 2.7 million and is expected to grow to 5.6 million by 2050.3
AF accounts for a substantial portion of the U.S. health care burden, with more than 467,000 hospitalizations annually and more than 99,000 deaths per year. AF is also responsible for adding $26 billion to U.S. health care spending annually, which is mostly driven by inpatient and outpatient costs.1
Adherence to anticoagulant therapy and accurate assessment of stroke and bleeding risk are vital for reducing the related health care burden and ensuring treatment success.
A study comparing the costs of rivaroxaban and warfarin using a Humana claims database reported that health care costs were comparable. The mean total costs for rivaroxaban was slightly lower compared with warfarin ($17,590 vs. $18,676), but higher pharmacy cost for rivaroxaban was offset by lower hospitalization cost.10
Similar cost comparison results were found between dabigatran and warfarin, using HealthCore data.11
Overall, based on recently published literature, NOACs tend to demonstrate better or comparable economic outcomes than warfarin, but no data exist to highlight the effect of better adherence to NOACs on cost savings.
With inadequate real-world evidence, the challenge for drug reimbursement policy decision makers is to consider whether the high drug costs and copays for NOACs are offset by lower overall medical costs, better adherence, and quality of life. A comprehensive evaluation of overall cost and its components (drug and medical costs) in patients treated with NOACs across different levels of adherence, comorbidities, and stroke and bleeding risk is not yet available. Furthermore, the association of improved adherence NOACs and consequent cost savings has not yet been examined in real-world studies. This analysis addressed these gaps in the current literature.
This study found that the economic burden of NOAC users based on total annual health care cost was substantial (> $30,000). However, better long-term adherence to NOACs directly translated in overall cost savings. The medical (inpatient and outpatient) costs for adherent users were significantly lower (-$6,941) than nonadherent patients, which offset the higher drug cost (+$3,362) among adherent users. The drug cost was comparable between groups categorized based on stroke risk, bleeding risk, and CCI scores. However, total cost was driven mostly by medical cost, which increased proportionally with a higher risk of bleeding and comorbidity. A similar trend was not observed for cost by stroke risk.
We also found that adherence to NOACs in our sample was suboptimal (< 60% with PDC ≥ 80%) and was better among patients with higher comorbidity and a higher risk of stroke and bleeding. The high-risk population benefited from frequent physician visits, focused interventions, aggressive efforts by physicians for improving care and adherence to treatment guidelines, better awareness among severe patients, and motivation to feel better, which can potentially lead to better patient-provider interaction and improvement in overall adherence among these patients.
Based on a large database study by Yao et al. (2016), adherence to NOACs at the end of follow-up at 1.1 years was low (47.5%), which is similar to our estimate of adherence (54%) at 12 months.6
Based on another study using commercial insured patients taking NOACs, the proportion of patients with PDC ≥ 80% at 9 months was suboptimal, with 55.0% for rivaroxaban, 56.8% for apixaban, and 46.7% for dabigatran users.21
It is noteworthy that advantages of NOAC over warfarin are promising and that NOAC therapy has a potential to improve overall medication adherence, which is evident from recently published studies.22-24
In our sample, it was clearly observed that adherence to NOACs was driven by baseline risk of stroke and bleeding and CCI scores. A study by Brown et al. (2016) reported an increase in adherence to NOAC medication from 58% in patients with CHA2
VASc scores of 2-3 and 62% in patients with a higher risk of stroke (CHA2
VASc score of 4+) based on the 9-month follow-up.21
Another study based on insurance claims data concluded that patients with a higher risk of stroke were more adherent to NOAC medication than patients at lower risk (CHA2
VASc score 0 or 1, 30.5%; CHA2
VASc score 2 or 3, 43.4%; and CHA2
VASc score ≥ 4, 45.3%).6
Our study differs from these studies because it presents new evidence in the form of better adherence from groups based on a risk of bleeding and comorbidity, which are integral to the selection of optimal anticoagulation treatment.
Our results also highlighted stroke and bleeding risk as a major predictor of medication adherence. Other predictors such as age, region, CCI score, and statin and ACE-ARB use were consistent with the previous studies.9,25-27
In addition, younger age, male gender, low stroke risk, poverty, higher education, and poor cognitive function have been associated with lower adherence.28
A recent study based on the Danish National Patient Registry reported an overall 1-year PDC equal to 83.9% and found that females (OR = 1.06), patients using cardiovascular drugs, and a CHA2
VASc score of ≥ 2 (OR = 1.12) were major predictors of adherence among dabigatran users.22
Our study is the first to compare the real-world cost between adherent and nonadherent NOAC patients. We found a positive association of adherence with cost saving (-$3,867) over a 12-month period that can be attributed to lower health care utilization (e.g., fewer hospital visits and shorter hospital stays). Our estimates of health care costs (drug, medical, and total) were consistent with the existing literature. Based on data from the U.S. Department of Defense Military Health System, drug costs were higher ($4,369, P
< 0.001) for dabigatran compared with warfarin, which is similar to our estimated drug costs.29
Also, Bancroft et al. (2016) studied dabigatran costs using a managed care data (2009-2012) and reported annual drug costs as $6,122, medical costs as $19,195, and total costs as $25,370.30
According to another database study by Fonseca et al. (2015), the total cost for patients taking dabigatran and warfarin after propensity score matching was $14,794 vs. $16,826.31
Drug cost was similar across subgroups based on CCI scores and stroke and bleeding risks. Medical and total costs were higher for patients with higher comorbidities and increased risk of bleeding. Conversely, we found that medical and total costs were higher for patients with a low risk of stroke compared with high-risk patients. Better adherence in the high-risk group might lead to lower incidence of stroke and its related lower cost burden. In addition, complications related to low-risk patients might be aggressively treated, leading to higher medical (inpatient and outpatient) costs. Cost and adherence information by stroke and bleeding risks and CCI scores can serve as crucial parameters informing the risk-benefit of treatment.
One of the strengths of this study was the use of a nationwide database with a large sample size. In addition, HAS-BLED was preferred because of its widespread use and better discrimination and prediction of bleeding risk compared with HEMORR2
HAGES and ATRIA.32,33
An inclusion criteria of at least 2 prescription refills ensured the exclusion of possible intermittent users (i.e., multiple periods of use but none exceeding 3 months) or patients who might use NOACs as concomitant therapy or for other indications. Furthermore, 2 prescription refills enabled calculation of adherence using PDC.
A few limitations need to be considered when interpreting our results. Although PDC has been widely used and accepted, it is an indirect method of assessing adherence because it is based on insurance claims (prescription fills). It should also be acknowledged that Optum Clinformatics Data Mart is a nationwide database, but the majority of the study patients were from the southern and midwestern regions. This database also underrepresents the older Medicare population.
As with all studies using claims databases, our data lacked information on race and ethnicity, long-term disease history, reasons for therapy discontinuation, and clinical variables such as INR values and body mass index. However, clinical markers of severity, such as CHA2DS2VASc, HAS-BLED, and CCI, reflected risk due to hypertension, previous cardiovascular disease, diabetes, and other comorbidities. Aspirin use was not comprehensively captured in the claims database because of its availability as an over-the-counter drug; however, since all patients were prescribed NOACs, the differential use of aspirin might be unlikely among patients with a CHA2DS2VASc score > 1, who are recommended anticoagulant therapy based on guidelines from the American College of Cardiology and the American Heart Association.
It is also important to acknowledge the differences in mechanism of action and costs between dabigatran and rivaroxaban. A higher proportion of our sample was prescribed dabigatran, since it was approved earlier. The newer drugs apixaban and edoxaban, which came into the market after 2012, were not included in our analysis. Since the measurement of PDC and cost was in the same period, inferences suggest a cross-sectional association instead of a sequential (longitudinal) causal relationship. This is a question that can be explored in future studies.
Despite these limitations, this study provides new evidence that emphasizes the importance of considering overall cost and adherence information in conjunction with the assessment of stroke and bleeding risk to inform the optimal choice of anticoagulant therapy and ascertain the overall risk-benefit of NOAC therapy.