Trend of Reported Bleeding in Warfarin Compared with Direct Oral Anticoagulants in Japan

Introduction

A decade has passed since the direct oral anticoagulants (DOACs)-dabigatran (Prazaxa), edoxaban (Lixiana), rivaroxaban (Xarelto), and apixaban (Eliquis) became available on the market in Japan. DOACs are categorized into two classes: dabigatran is a direct thrombin inhibitor, while edoxaban, rivaroxaban, and apixaban are direct factor Xa inhibitors. Anticoagulation pharmacotherapy with DOACs has been widely used to prevent thrombosis in several cardiovascular contexts. Those drugs demonstrated superiority or noninferiority to prior standards of anticoagulation pharmacotherapy with vitamin K antagonists (warfarin) with similar or reduced bleeding risks in Phase III trials.^1–6 Although the efficacy and safety of DOACs have been demonstrated, those pivotal randomized clinical trials were powered to assess efficacy; thus, the sample sizes may not have been enough to detect safety signals. In addition, patients enrolled in those clinical trials had to meet strict inclusion criteria and may not reflect the general patient population seen in real-world clinical practice. In addition, most patients in those trials were non-Japanese, and reports on Japanese populations treated with DOACs remain limited.

Since the evidence of the safety of newly approved drugs, especially, is often limited, post-marketing surveillance and pharmacovigilance studies are essential to assess the safety profile of drugs used in routine clinical practice. The Japanese Adverse Drug Event Report Database (JADER) is a spontaneous reporting system for drug adverse events managed by the Pharmaceuticals and Medical Devices Agency (PMDA) in Japan.⁷ This database has been widely used for pharmacovigilance studies in Japan. A few studies have reported on the bleeding risks of DOACs using JADER;^8,9 however, to the best of our knowledge, no pharmacovigilance studies have compared the bleeding risks of DOACs and warfarin in Japan. Moreover, the incidence rate of bleeding events associated with those drugs in real-world settings remains uncertain.

In this study, we investigated the number of oral anticoagulant-associated bleeding reports in the JADER. We also calculated the estimated incidence rate of oral anticoagulant-associated bleeding using the nationwide health insurance claims database provided by JMDC Inc. (JMDC database).¹⁰

Methods

Study Design and Data Source

This was a retrospective observational cohort study using the JADER⁷ and the JMDC databases.¹⁰ The JADER database contains serious adverse events which have even been noticed in the package inserts. All adverse drug events recorded in the JADER are reported mainly by marketing authorization holders and healthcare institutions. The database consists of four files: patient background information (DEMO), drug information (DRUG), adverse reactions (REAC), and medical history (HIST).

The nationwide estimated number of patients was obtained from the JMDC database. This database includes patients’ characteristics (age and sex), prescribed or dispensed medications, procedures, and diagnoses for 17 million insured Japanese people. It is the largest anonymized database of claims data in Japan, and it has been widely used for health economics, epidemiology, and outcomes research.

Definitions of Case Reports and Drugs of Interest

From the JADER, adverse event data recorded between fiscal year (FY) 2004 and FY2023 (fiscal year: from April 1 to March 31 of the following year) were obtained. Case reports with bleeding were identified using the preferred terms “bleeding” and “hemorrhage” according to MedDRA version 28.0J. The drugs of interest included warfarin, dabigatran, edoxaban, rivaroxaban, and apixaban (Table S1).

Data Analysis

All bleeding case reports associated with the suspected drugs between FY2004 and FY2023 were identified. The patient characteristics and the number of oral anticoagulant-associated bleeding reported to the JADER were presented in a descriptive manner. The overall number of patients receiving study drugs was estimated from the JMDC database between FY2011 and FY2023, and the incidence rate was calculated from the number of bleeding reports extracted from JADER. To estimate the number of patients prescribed the drugs of interest, we used government statistical data to calculate the estimated numbers, adjusted for age (one-year increments) and gender (male and female), based on prescription rates obtained from the JMDC database using the direct standardization method.¹¹ In the following formula, “N Japan male/female” means the number of males and females in Japan for each analysis year, “N jmdc male/female” means the number of males and females in the JMDC database for each analysis year, and “N’ jmdc male/female” means the number of prescriptions per male and female in the JMDC database for each analysis year, and “i” means age. If there was any prescription record within the fiscal year, it counted as one, and each drug of interest counted separately. The rate of oral anticoagulant-associated bleeding incidences between patients who were prescribed warfarin and those who were prescribed dabigatran, edoxaban, rivaroxaban, or apixaban between FY2011 and FY2023 was compared. Since statistical power was not calculated and bleeding incidence was rare, we decided to present the data descriptively.

Nationwide estimated number of patients =

Ethics Approval

Ethics approval was not applicable to this study based on the Ethical Guidelines for Medical and Health Research Involving Human Subjects issued by the Ministry of Health, Labour and Welfare (MHLW), since only completely encrypted data were used.

Results

Between FY2004 and FY2023 (between April 1, 2004, and March 31, 2024), there were 903,869 case reports, of which we identified 50,276 with bleeding. Among them, we identified 16,125 oral anticoagulant-associated bleeding in 15,970 case reports. The characteristics of patients in those case reports are summarized in Table 1. More than half (54.3%) were male, and 69.5% was over 70 years old. The most common suspected oral anticoagulant was apixaban (33.4%), followed by rivaroxaban (26.0%), warfarin (16.6%), and edoxaban (13.3%).

Table 1 Patient Characteristics of 15,970 Case Reports

The trend in the reported number of oral anticoagulant-associated bleeding between FY2011 and FY2023 is shown in Figure 1. The number of reported oral anticoagulant-associated bleeding for warfarin was within the 100–200 range throughout the period. On the other hand, the number of reported oral anticoagulant-associated bleeding for rivaroxaban and apixaban was high; both reached a peak of over 700 in FY2017. Overall, the number of reported oral anticoagulant-associated bleeding tended to decrease from FY2020.

Figure 1 Number of oral anticoagulant-associated bleeding between FY2011 and FY2023. Blanks (-) indicate no available data because these drugs were not yet available on the market.

The estimated number of patients prescribed oral anticoagulants, calculated from the JMDC database by gender and age group, is summarized in Table 2. In this table, data for FY2023, the latest available, are presented. The proportion of male patients who were prescribed oral anticoagulants tended to be higher than that of females, and the majority of patients were aged 70 years or older. The incidence rates of oral anticoagulant-associated bleeding in patients who were prescribed oral anticoagulants are summarized in Table 3. The incidence rates of anticoagulant-associated bleeding in patients who were prescribed dabigatran, edoxaban, rivaroxaban, and apixaban were higher compared to those in patients who were prescribed warfarin. The incidence rate of oral anticoagulant-associated bleeding in patients prescribed warfarin remained below 1000 per 1,000,000/year throughout the study period. In contrast, the incidence exceeded this threshold for dabigatran, edoxaban, rivaroxaban, and apixaban in several fiscal years.

Table 2 Number of Estimated Patients Prescribed Oral Anticoagulants Calculated Using the JMDC Database in FY2023

Table 3 Incidence Rates of Oral Anticoagulant-Associated Bleeding in Patients Who Were Prescribed Oral Anticoagulants

Discussion

We conducted a retrospective observational study using the JADER and JMDC databases to investigate the number of oral anticoagulant-associated bleeding reports and their annual incidence rate. Our study showed that oral anticoagulants were mainly prescribed to patients aged 70 years or older, and the risk of oral anticoagulant-associated bleeding was higher in those prescribed dabigatran, edoxaban, rivaroxaban, or apixaban than in those prescribed warfarin.

Oral anticoagulant-associated bleeding was more commonly reported in elderly patients (those aged 70 years or older) in our study (78.7%). Our study also showed that the majority of patients prescribed DOACs were elderly, ranging from 66.8% to 71.6% of the total (based on FY2023 data); in contrast, the proportion of elderly patients prescribed warfarin was only 58.6% (based on FY2023 data). These findings confirm several studies showing that anticoagulation pharmacotherapy increases the bleeding risk in elderly patients.^12–17 The Japanese local clinical guidelines also mention that age over 75 years is one of the factors associated with significant bleeding during anticoagulation therapy.¹⁸ However, anticoagulation pharmacotherapy using DOACs in the elderly population is less extensively studied, and the clinical impact of anticoagulation-related bleeding in elderly patients remains uncertain. Thus, the risk of bleeding, particularly in the elderly population treated with DOACs, needs further investigation.

Although package inserts of DOACs precaution against the risk of bleeding,^19–22 a number of serious bleeding events have been reported in the JADER. In our study, bleeding risk was higher in those who were prescribed DOACs compared to those who were prescribed warfarin. Compared to the Western population, the bleeding risk in the East Asian population treated with anticoagulation pharmacotherapy has been reported to be higher.^23–26 Moreover, the pharmacokinetic profiles of some DOACs differ in the Japanese population compared to those in Western populations.^27,28 Therefore, the risk of bleeding, specifically in the Japanese population treated with DOACs, should be more carefully determined.

Previous real-world studies conducted outside Japan have consistently shown that DOACs are associated with a lower or comparable risk of bleeding compared with warfarin. In a large observational study conducted in the United States, apixaban and dabigatran were associated with a lower risk of major bleeding compared with warfarin.²⁹ Furthermore, a large cohort study involving approximately 200,000 patients in the United Kingdom also reported that apixaban was associated with a significantly lower risk of major bleeding compared with warfarin in clinical practice.³⁰ Similar findings have been reported in Asian populations. A meta-analysis of studies involving Asian patients with atrial fibrillation demonstrated that DOAC therapy was associated with a lower risk of major bleeding compared with warfarin.³¹ Furthermore, a cohort study using electronic health records and claims data from 372 acute care hospitals in Japan showed that DOACs were associated with a lower risk of major bleeding compared with warfarin.³² In contrast to these findings, our analysis using the JADER database showed that the frequency of reported serious bleeding events tended to be higher with DOACs than with warfarin. This apparent discrepancy may be explained by the inherent limitations of spontaneous reporting systems, including reporting bias,^33,34 whereby adverse events associated with newly marketed drugs may be reported more frequently, whereas well-known adverse events related to long-established drugs may be underreported. Therefore, the low frequency of reported bleeding events with warfarin in JADER may not reflect the actual bleeding risk and should be interpreted in conjunction with findings from other studies.

The majority of oral anticoagulant-associated bleeding was reported between FY2015 and FY2019, while far fewer cases were reported between FY2020 and FY2023. A decrease in the oral anticoagulant-associated bleeding events in FY2020 and thereafter might be due to the COVID-19 pandemic. MHLW reports that the average daily outpatient visits fell dramatically in 2020 and then did not recover to pre-COVID-19 pandemic levels even in 2023.³⁵

Limitations

Our study had several limitations. Underreporting, overreporting, and data entry errors can occur in any spontaneous adverse event reporting system, including the JADER. To interpret our study results using the JADER, reporting and notoriety bias should be taken into account. Regarding the JMDC database, it primarily includes health insurance data for employees of large companies. Also, the proportion of the elderly aged 65 years and older included in the database is significantly lower than in the Japanese population. Thus, caution is needed when generalizing our findings from the JMDC database. Although our study had such limitations, showing differences in oral anticoagulant-associated bleeding risks among drugs and their incidence rates among Japanese patients receiving oral anticoagulants in real-world settings is informative.

Future Directions

Further studies with more inclusive datasets would be warranted for a better understanding of oral anticoagulant-associated bleeding risks.

Conclusions

Our study demonstrated that oral anticoagulant-associated bleeding was more commonly reported in elderly patients aged 70 years or older. Moreover, the risk of oral anticoagulant-associated bleeding might be higher with DOACs than with warfarin in the real-world setting in Japan. Further assessment would be warranted to understand the risk-benefit of DOACs.