Comparison of the Safety and Effectiveness of Four Direct Oral Anticoagulants in Japanese Patients with Nonvalvular Atrial Fibrillation Using Real-World Data

Abstract

Direct oral anticoagulants (DOACs) are widely used for the prevention of ischemic stroke and systemic embolism in patients with nonvalvular atrial fibrillation (NVAF). However, the differences in safety and effectiveness among four DOACs, dabigatran, rivaroxaban, apixaban, and edoxaban, in Japanese patients have not been clarified. Therefore, we conducted a retrospective cohort study to directly compare the safety and effectiveness among the four DOACs using the Japan Medical Data Center (JMDC) claims database. We identified 3823 patients with NVAF who started receiving a DOAC between March 2011 and June 2017. The safety outcome was major bleeding (a composite outcome of intracranial, gastrointestinal, respiratory, or renal/urinary tract bleeding) and the effectiveness outcome was the composite of ischemic stroke including transient ischemic attack (TIA) or systemic embolism. We constructed a Cox proportional hazard model to calculate the hazard ratio (HR) for all four DOAC combinations. The risk of major bleeding was significantly lower in the dabigatran group than in the apixaban group (HR, 0.55; 95% confidence interval (CI), 0.31–0.93; p = 0.03). In contrast, there was no significant difference in the risk of major bleeding among the other DOACs. In the composite risk of ischemic stroke including TIA or systemic embolism, there was no significant difference among the four DOACs. This study suggested that in the current use of DOACs in Japanese patients with NVAF, dabigatran had a significantly lower risk of major bleeding than apixaban, but there was no significant difference in effectiveness among the four DOACs.

INTRODUCTION

Atrial fibrillation (AF) increases the risk of ischemic stroke by five-fold.¹⁾ Furthermore, stroke associated with AF has higher risk of recurrence and mortality than other stroke types, and is more likely to lead to a poor functional outcome.^2–4) Therefore, it is important for patients with AF to take oral anticoagulants to prevent stroke and systemic embolism.

In randomized clinical trials and their meta-analysis, direct oral anticoagulants (DOACs), including dabigatran, rivaroxaban, apixaban, and edoxaban, were demonstrated to have comparable or better efficacy than warfarin, with a significant reduction of the risk of intracranial hemorrhage (ICH) in patients with nonvalvular atrial fibrillation (NVAF).^5–9) Thus, DOACs are now recommended over the conventionally used warfarin.¹⁰⁾

In a meta-analysis, the degree of reduction of the risk of the composite of stroke or systemic embolism with standard-dose DOACs was reported to be greater in Asian patients than in non-Asian patients.¹¹⁾ In addition, standard-dose DOACs reduced the incidence of major bleeding more in Asian patients than in non-Asian patients. Moreover, they increased the risk of gastrointestinal bleeding compared with warfarin only in non-Asian patients.¹¹⁾ Thus, Asian patients are likely to benefit more from DOACs.

Previous studies suggested differences in safety among individual DOACs. However, the results differ among reports. Overall, rivaroxaban has a higher risk of major bleeding than apixaban and dabigatran,^12–16) whereas other reports found no significant difference between rivaroxaban and dabigatran.¹⁷⁾ Regarding the safety of apixaban and dabigatran, several reports demonstrated that apixaban is safer,^{13,16,18–20)} whereas others noted no difference.^12,14,15,17) Although there are few studies that included edoxaban, they revealed that apixaban has a significantly lower risk of major bleeding than rivaroxaban and dabigatran, and that edoxaban has no significant difference in major bleeding risk from apixaban.^21,22) In addition, edoxaban has a significantly lower risk of major bleeding than rivaroxaban and is not significantly different from dabigatran.²¹⁾ On the other hand, many studies that directly compared the effectiveness of DOACs reported no difference among the drugs.^{12,13,15,22,23)}

These are all reports from overseas, and the differences in the safety and effectiveness of the four DOACs in Japanese patients have yet to be clarified. In addition, some DOACs have different approved doses and criteria for dosage reduction or contraindications between in Japan and overseas. For example, overseas, rivaroxaban is used at 20 and 15 mg in patients with NVAF depending on renal function,²⁴⁾ whereas in Japan, the doses of 15 and 10 mg are approved. Dabigatran is contraindicated with creatinine clearance (CrCl) of <30 mL/min in Japan, but in the U.S.A., it is contraindicated with CrCl of <15 mL/min.²⁵⁾ Thus, it is difficult to directly adapt the results of overseas studies to Japanese patients.

We conducted a retrospective cohort study to compare the safety and effectiveness of the four DOACs in Japanese patients with NVAF using real-world data from health insurance claims.

MATERIALS AND METHODS

Data Source

We conducted a retrospective cohort study using a large claims database constructed by the Japan Medical Data Center Co., Ltd,.²⁶⁾ (JMDC; Tokyo, Japan). The JMDC claims database used in this study consists of claims data from medical institutions and pharmacies submitted to health insurance associations and health examination data between January 2005 and June 2017, which includes approximately 4.2 million insured persons. In addition, the JMDC claims database is highly encrypted using irreversible anonymization technology and cannot be used to identify individuals. The JMDC claims database provides information on the beneficiaries, including encrypted personal identifiers, age, sex, procedures, such as surgery or endoscopy, diagnosis, prescription drug name, dose, prescription days, prescription date, and dispensing date. In this database, diagnoses were recorded by the International Classification of Diseases 10th Revision (ICD-10) and all drugs were coded according to the Anatomical Therapeutic Chemical (ATC) classification. This database has the advantage of being able to track patients even if they change hospitals or go to multiple facilities.

Study Population

Patient selection procedures are shown in Fig. 1. We identified all new users of dabigatran, rivaroxaban, apixaban, and edoxaban for NVAF between March 2011 and June 2017. The date of the first prescription of DOACs during the study period was defined as the index date. This first DOAC was defined as the index medication and patients were assigned to each medication group. Patients were required to have at least 6 months of claims data prior to the index date to obtain a medical and prescription history as patient baseline characteristics. In addition, patients were required to be diagnosed with NVAF prior to the index date. No minimum duration of medication was set for eligible patients.

Fig. 1. Patient Selection Procedure

We identified patients with non-valvular atrial fibrillation who were newly prescribed any DOAC between March 2011 and June 2017. The date of the first DOAC prescription was defined as the index date. Patients with a history of warfarin prescription within 6 months before the index date were excluded to avoid the effects of warfarin-induced bleeding. Patients with a history of deep venous thrombosis, pulmonary embolism, major bleeding, ischemic stroke, or systemic embolism within 6 months prior to the index date, hip fracture surgery, or knee or hip replacement surgery within 6 weeks prior to the index date were excluded. Pregnant patients and patients with end-stage kidney disease or valvular heart disease were also excluded. ^† Recent is defined as within 6 months prior to the index date. ^‡ Recent is defined as within 6 weeks prior to the index date. ^§ Major bleeding is a composite outcome of intracranial, gastrointestinal, respiratory, or renal/urinary tract bleeding. AF, atrial fibrillation; DOAC, direct oral anticoagulant; JMDC, Japan Medical Data Center; NVAF, nonvalvular atrial fibrillation.

We excluded patients with a history of warfarin prescription within 6 months prior to the index date. Incidentally, the starting month of this study, March 2011, was when dabigatran, the first DOAC, was launched, thus there were no patients using DOACs before the index date. Thus, the population consisted of new DOAC users. Furthermore, patients with a diagnosis of deep venous thrombosis, pulmonary embolism, major bleeding, ischemic stroke or systemic embolism within 6 months prior to the index date, and patients who underwent hip fracture surgery, or knee or hip replacement surgery within 6 weeks prior to the index date were excluded. We also excluded patients with valvular heart disease, end-stage kidney disease, including dialysis, or pregnancy.

These diagnostic and surgical histories were examined by extracting patients with relevant codes from the database. The relevant codes were defined as shown in Table 1.

Table 1. Codes Used to Define Patient Selection and Study Outcomes

Inclusion criteria	ICD-10/ATC/Medical procedure/Surgery codes	Details
Atrial fibrillation	ICD-10: I48
Exclusion criteria
Recent deep venous thrombosis and pulmonary embolism	ICD-10: I26, I800, I801†, I802, I803†, I808†, I809†, I82	Of the patients marked with “†,” only those with thrombophlebitis disease were excluded.
Recent orthopedic surgery	Surgery codes: K082, K082-3, K073, K073-2, K046, K046-2	Patients who recently underwent hip fracture surgery, or knee or hip replacement surgery were excluded.
Recent major bleeding	ICD-10: I60, I61, I62, K228‡, K250‡, K252, K254‡, K256, K260‡, K262, K264, K266, K284, K290, K571‡, K573‡, K625,K920, K921, K922, J942, R04, N02, R31	Of the patients marked with “‡,” only those considered to be associated with bleeding were excluded.
Recent ischemic stroke	ICD-10: I63, G45
Recent systemic embolism	ICD-10: I74
Valvular heart disease	ICD-10: I050, I052, I080§, I081§, I083§, I088§, I342, I489||, Z952, Z953, Z954	Of the patients marked with “§,” only those with the disease name of mitral valve stenosis were excluded.
		|| Patients with I489 coding but with valvular atrial fibrillation in their disease name were excluded.
End-stage kidney disease	ICD-10: E102¶, E112¶, N180¶, N083¶, Z49, Z992¶	Of the patients marked with “¶,” only those with end-stage renal failure, including those on dialysis, were excluded.
	Medical procedure codes: J038, J042
Pregnancy	ICD-10: O00-O99, Z33, Z34, Z35
Comorbidity
Congestive heart failure	ICD-10: I50, I42, I110, I130, I132, J81
Hypertension	ICD-10: I10, I11, I12, I13, I15	Hypertensive patients were defined as those who received antihypertensive drugs within 6 months before the index date and also had a diagnosis of hypertension.
	ATC: C03AA, C03BA, C08CA, C08DB01, C09AA, C09CA, C09DA, C09DB, C10BX03
Diabetes	ATC: A10	A history of diabetes was determined by whether the patient had a history of receiving diabetic medications within 6 months before the index date.
Liver dysfunction	ICD-10: B15, B16, B17, B18, B19, C22, D684**, I982, K70, K71, K72, K73, K74, K75, K76, K77, Z944	Of the patients marked with “**,” only those with coagulation factor deficiency due to liver disease were excluded.
Concomitant medication
NSAIDs	ATC: M01A
Antiplatelet drugs	ATC: B01AC04, B01AC05, B01AC06, B01AC22, B01AC23, B01AC24, B01AC30, B01AC56
Outcome
Major bleeding
Intracranial bleeding	ICD-10: I60, I61, I62
Gastrointestinal bleeding	ICD-10: K228‡, K250‡, K252, K254‡, K256, K260‡, K262, K264, K266, K284, K290, K571‡, K573‡, K625, K920, K921, K922	Of the patients marked with “‡,” only those considered to be associated with bleeding were excluded. Gastrointestinal bleeding was defined as endoscopy performed in the month before and after the diagnosis.
	Medical procedure codes: D306, D308, D310, D312, D313, J017, K646, K654, K722
Respiratory bleeding	ICD-10: J942, R04
Renal/Urinary tract bleeding	ICD-10: N02, R31
Ischemic stroke	ICD-10: I63, G45
Systemic embolism	ICD-10: I74

ICD, International Classification of Diseases; NSAIDs, nonsteroidal anti-inflammatory drugs.

Patient Characteristics

We collected baseline patient characteristics, including age, sex, concomitant medications (nonsteroidal anti-inflammatory drugs (NSAIDs) and antiplatelet drugs), comorbidities (congestive heart failure, hypertension, diabetes, and liver dysfunction), the calendar period of the first DOAC prescription, and the Charlson–Deyo Comorbidity Index, which we selected as confounding variables. In addition, we examined the proportion of patients who started with a reduced dosage to clarify the current status of DOAC prescription in Japan. Concomitant antiplatelet drugs were defined as aspirin, clopidogrel, prasugrel, ticagrelor, or cilostazol prescribed within 6 months prior to the index date. Hypertension was defined as a history of diagnosis and prescription of antihypertensive drugs within 6 months prior to the index date. Diabetes was defined as the prescription of hypoglycemic drugs or insulin in the 6 months prior to the index date. A history of congestive heart failure and liver dysfunction was defined by the presence of a corresponding ICD10 codes in the claims data within 6 months prior to the index date. Furthermore, we calculated the Charlson–Deyo Comorbidity Index using the ICD-10 codes based on a previous study²⁷⁾ and the updated version of each disease score.²⁸⁾ The codes used to investigate concomitant medications and comorbidities are shown in Table 1.

The reason for selecting these factors as confounding variables is that they are risk factors included in the CHADS₂ score, CHA₂DS₂-VASc score, or HAS-BLED score, which are recommended for risk assessment of stroke or bleeding.²⁹⁾ Furthermore, as DOACs are newly launched drugs, the calendar period of the first prescription was added as a confounding variable. As in previous studies,^19,23) the Charlson–Deyo Comorbidity Index was used as a confounding variable. However, we were unable to examine the patient background of labile Prothrombin Time-International Normalized Ratio (PT-INR), renal failure, and alcohol abuse as confounding variables, which are components of the HAS-BLED score, because we were unable to obtain the data for PT-INR values, information on serum creatinine values and body weight, which are necessary for the calculation of creatinine clearance, an index of renal function, or information on the daily amount or frequency of alcohol consumption from the JMDC claims database. We also eliminated vascular disease, which is a component of the CHA₂DS₂-VASc score, to avoid overfitting because it was not a significant factor in Japanese studies.^30–32)

Outcomes

We examined two categories of treatment outcomes: safety and effectiveness. The primary safety outcome was major bleeding (a composite of intracranial, gastrointestinal, respiratory, or renal/urinary tract bleeding). The primary effectiveness outcome was a composite of ischemic stroke including transient ischemic attack (TIA) or systemic embolism.

The diagnosis of major bleeding (intracranial, respiratory, or renal/urinary tract bleeding), ischemic stroke, including TIA, and systemic embolism were defined by ICD-10 codes (Table 1). Gastrointestinal bleeding was defined by ICD-10 codes (Table 1), and endoscopy performed in the month before and after the diagnosis.

Follow-Up Period

Patients were followed until their first event, discontinuation of medication, medication switching, censoring for loss to follow-up, or end of the study period (June 30, 2017), whichever occurred earliest. The different outcomes of safety and effectiveness were analyzed separately. Specifically, patients who had two events (ischemic stroke and major bleeding) were included in the effectiveness analysis up to the time of the first ischemic stroke and included in the safety analysis up to the time of the first major bleeding event. However, patients who had two ischemic strokes were censored at the time of the first ischemic stroke. We defined discontinuation of medication as a gap of 30 d or more between medication periods. We used the fill dates and prescription days to determine the period of medication. The treatment was considered to be continuing as long as the patient received the following prescriptions within 30 d after the scheduled end date of the remaining medication.

Statistical Analysis

Differences in patient background for each DOAC were statistically calculated using the Pearson chi-squared test, only the follow-up period was analyzed using the Kruskal–Wallis test, and for multiple testing, the Steel–Dwass test was used. Comparisons between each DOAC in the proportion of patients who started with reduced doses were analyzed using the adjusted p-value according to the Holm method for multiple testing. The crude incidence rate was calculated by dividing the number of events by 100 person-years. To calculate the adjusted hazard ratio (HR) and 95% confidence interval (95% CI) of major bleeding and ischemic stroke or systemic embolism across the four treatment groups, we constructed Cox proportional hazard models. Cox models controlled for age, sex, calendar period of first prescription, Charlson–Deyo Comorbidity Index, comorbidities of congestive heart failure, hypertension, diabetes or liver dysfunction, and history of using NSAIDs or antiplatelet drugs as confounding variables.

JMP 11.2.1 (SAS Institute Inc., NC, U.S.A.) was used for all statistical analyses. Values of p < 0.05 were considered significant.

Ethical Considerations

This study was approved by the ethics committee of the Faculty of Medicine, Tokyo Medical and Dental University.

RESULTS

Baseline Patient Characteristics

In total, 3823 patients were identified after applying the selection criteria. Of the total, 984 (25.7%) were in the dabigatran treatment group, 1390 (36.4%) were in the rivaroxaban treatment group, 1067 (27.9%) were in the apixaban treatment group, and 382 (10.0%) were in the edoxaban treatment group. Dabigatran users were the youngest and apixaban users were the oldest, but there was little difference in median age among all groups. The proportion of patients taking concomitant antiplatelet drugs was lowest in the apixaban group. On the other hand, there were no significant differences in patient background among the four groups in terms of hypertension, diabetes, liver dysfunction, sex, or use of NSAIDs. The prevalence of heart failure was highest in the edoxaban group and lowest in the rivaroxaban group. In addition, rivaroxaban users had the lowest Charlson–Deyo Comorbidity Index scores. Regarding the proportion of patients who started with a reduced dose, there was a significant difference among all DOACs. In particular, dabigatran had the largest proportion of patients with dose reduction, whereas apixaban had the smallest proportion of patients with dose reduction. As for the follow-up period, it was significantly longer for rivaroxaban than for apixaban. That for edoxaban was significantly shorter than that for the other three DOACs, which may have been due to the timing of approval (Table 2).

Table 2. Baseline Characteristics and Follow-Up Period of Patients by DOAC Treatment Group

Variable	All		Dabigatran		Rivaroxaban		Apixaban		Edoxaban		p-Value*
	n = 3823		n = 984		n = 1390		n = 1067		n = 382
Age (years), median (IQR)	58	(52–64)	57	(51–63)	58	(52–64)	59	(53–64)	58	(52–65)
< 60 (%)	2105	(55.1%)	574	(58.3%)	777	(55.9%)	539	(50.5%)	215	(56.3%)	<0.01
≥ 60 (%)	1718	(44.9%)	410	(41.7%)	613	(44.1%)	528	(49.5%)	167	(43.7%)
Sex (%)
Male	3196	(83.6%)	831	(84.5%)	1174	(84.5%)	868	(81.3%)	323	(84.6%)	0.14
Female	627	(16.4%)	153	(15.5%)	216	(15.5%)	199	(18.7%)	59	(15.4%)
Calendar period of first prescription (%)
2011/03–2012/12	313	(8.2%)	292	(29.7%)	21	(1.5%)	0	-	0	-	<0.01
2013/01–2014/06	872	(22.8%)	344	(35.0%)	408	(29.4%)	119	(11.2%)	1	(0.3%)
2014/07–2015/12	1302	(34.1%)	235	(23.9%)	494	(35.5%)	509	(47.7%)	64	(16.8%)
2016/01–2017/06	1336	(34.9%)	113	(11.5%)	467	(33.6%)	439	(41.1%)	317	(83.0%)
Charlson–Deyo Index median (IQR)	2	(0–2)	2	(0–2)	1	(0–2)	2	(0–2)	2	(0–2)
0–2 (%)	3032	(79.3%)	793	(80.6%)	1131	(81.4%)	818	(76.7%)	290	(75.9%)	<0.01
≥3 (%)	791	(20.7%)	191	(19.4%)	259	(18.6%)	249	(23.3%)	92	(24.1%)
Comorbidity (%)
Congestive heart failure	1643	(43.0%)	418	(42.5%)	544	(39.1%)	496	(46.5%)	185	(48.4%)	<0.01
Hypertension	1811	(47.4%)	453	(46.0%)	662	(47.6%)	525	(49.2%)	171	(44.8%)	0.36
Diabetes	436	(11.4%)	101	(10.3%)	156	(11.2%)	136	(12.7%)	43	(11.3%)	0.36
Liver dysfunction	618	(16.2%)	139	(14.1%)	235	(16.9%)	182	(17.1%)	62	(16.2%)	0.24
Concomitant medication (%)
NSAIDs	799	(20.9%)	182	(18.5%)	292	(21.0%)	245	(23.0%)	80	(20.9%)	0.10
Antiplatelet drugs	471	(12.3%)	132	(13.4%)	187	(13.5%)	107	(10.0%)	45	(11.8%)	0.045
Starting with reduced dosage (%)†	984	(25.7%)	481	(48.9%)	223	(16.0%)	122	(11.4%)	158	(41.4%)	<0.01
Follow-up period (day)‡, median (IQR)	184	(67–432)	182a	(59–487)	205b,c	(79–507)	181b,d	(69–403)	122a,c,d	(45–228)	<0.01
Follow-up period (day)§, median (IQR)	185	(67–437)	177e	(59–489)	208f,g	(77–521)	184f,h	(69–409)	119e,g,h	(45–226)	<0.01

* p-Values were obtained by the Pearson chi-squared test except for the follow-up period. The Kruskal–Wallis test was used to compare the follow-up periods. p-Values <0.05 indicate significant differences among DOAC treatment groups. The Steel–Dwass test was used to compare the follow-up period between each DOAC. The same letter (a, b, c, d, e, f, g, h) indicates a significant difference between the drugs. ^† The Holm method for multiple testing was used to determine which drugs significantly differed in the proportion of patients who started with reduced doses. There was a significant difference in the proportion of patients who started with reduced doses among all drugs, which was below the adjusted p-values using the Holm method for multiple testing. ^‡ Follow-up period used to assess safety. ^§ Follow-up period used to assess effectiveness. As outcomes of safety and efficacy were analyzed separately, there is a follow-up period for each. DOAC = direct oral anticoagulant; IQR = interquartile range; NSAIDs = nonsteroidal anti-inflammatory drugs.

Safety and Effectiveness Outcomes

The crude incidence rate for safety and effectiveness outcomes is shown in Table 3. Before adjustment, the composite risk of major bleeding was highest with apixaban and lowest with edoxaban. In addition, the composite risk of ischemic stroke including TIA or systemic embolism was highest with apixaban and lowest with rivaroxaban.

Table 3. Number of Events and Crude Incidence Rates of Safety and Effectiveness Outcomes by DOAC Treatment Group

Outcome	Dabigatran	Rivaroxaban	Apixaban	Edoxaban
Safety outcome: Major bleeding†
Number of events	33	56	47	6
Event rate per 100 person-years (95% CI)	3.47 (3.46–3.49)	4.13 (4.12–4.14)	5.68 (5.66–5.70)	3.34 (3.31–3.37)
Effectiveness outcome: Ischemic stroke or systemic embolism
Number of events	34	32	32	6
Event rate per 100 person-years (95% CI)	3.58 (3.57–3.59)	2.36 (2.35–2.37)	3.87 (3.85–3.88)	3.34 (3.31–3.37)

^† Major bleeding is a composite outcome of intracranial, gastrointestinal, respiratory, or renal/urinary tract bleeding. DOAC, direct oral anticoagulant; CI, confidence interval.

After adjusting for confounding factors, the risk of major bleeding was significantly lower in the dabigatran group than in the apixaban group (HR, 0.55; 95% CI, 0.31–0.93; p = 0.03). On the other hand, there was no significant difference in the risk of major bleeding among the other DOACs. Regarding the composite risk of ischemic stroke including TIA or systemic embolism, there was no significant difference among DOACs (Table 4).

Table 4. Adjusted Hazard Ratio for Comparisons of Safety and Effectiveness Outcomes between Each DOAC Treatment Group

Outcome		Safety outcome (Major bleeding†)		Effectiveness outcome (Ischemic stroke or systemic embolism)
		HR‡ (95% CI)	p-Value	HR‡ (95% CI)	p-Value
Dabigatran	vs. Rivaroxaban	0.75 (0.45–1.23)	0.26	1.24 (0.70–2.17)	0.46
Dabigatran	vs. Apixaban	0.55 (0.31–0.93)	0.03*	0.82 (0.44–1.49)	0.51
Dabigatran	vs. Edoxaban	0.87 (0.35–2.50)	0.78	1.11 (0.43–3.23)	0.84
Rivaroxaban	vs. Apixaban	0.73 (0.48–1.10)	0.13	0.66 (0.39–1.10)	0.11
Rivaroxaban	vs. Edoxaban	1.16 (0.51–3.11)	0.74	0.89 (0.38–2.46)	0.81
Apixaban	vs. Edoxaban	1.59 (0.71–4.24)	0.27	1.35 (0.59–3.67)	0.50

* p-Values <0.05 indicate significant differences. ^† Major bleeding is a composite outcome of intracranial, gastrointestinal, respiratory, or renal/urinary tract bleeding. ^‡ The hazard ratio of the left DOAC to the right DOAC is shown. DOAC, direct oral anticoagulant; CI, confidence interval; HR, hazard ratio.

In terms of the effects of each confounding variable, patients aged 60 years or older had a significantly higher risk of major bleeding and ischemic stroke, including TIA, or systemic embolism than patients younger than 60 years old. In addition, male had a significantly lower risk of ischemic stroke, including TIA, or systemic embolism than female. On the other hand, there was no significant difference in major bleeding between the sexes. There were significant differences in the risk of major bleeding and ischemic stroke, including TIA, or systemic embolism between some of the calendar periods of the first prescription. In this study, the Charlson–Deyo Index, the use of NSAIDs or antiplatelet drugs, and presence of comorbidities did not significantly increase the risk of major bleeding, ischemic stroke, including TIA, or systemic embolism (Table 5).

Table 5. Adjusted Hazard Ratio for Comparisons of Safety and Effectiveness Outcomes for Each Confounding Variable

Outcome		Safety outcome (Major bleeding†)		Effectiveness outcome (Ischemic stroke or systemic embolism)
Confounding variables		HR‡ (95% CI)	p-Value	HR‡ (95% CI)	p-Value
Age (years)
≥60	<60	1.61 (1.14–2.29)	<0.01*	1.56 (1.04–2.36)	0.03*
Sex
Male	Female	0.81 (0.54–1.25)	0.33	0.60 (0.38–0.96)	0.03*
Calendar period of the first prescription
2011/03–2012/12	2013/01–2014/06	1.07 (0.52–2.11)	0.85	1.61 (0.79–3.20)	0.19
2011/03–2012/12	2014/07–2015/12	1.65 (0.76–3.41)	0.20	2.17 (1.02–4.57)	0.04*
2011/03–2012/12	2016/01–2017/06	1.66 (0.72–3.72)	0.23	1.75 (0.77–3.97)	0.18
2013/01–2014/06	2014/07–2015/12	1.54 (1.01–2.34)	0.04*	1.35 (0.79–2.30)	0.27
2013/01–2014/06	2016/01–2017/06	1.55 (0.92–2.66)	0.10	1.09 (0.59–2.02)	0.78
2014/07–2015/12	2016/01–2017/06	1.01 (0.62–1.67)	0.98	0.81 (0.47–1.41)	0.45
Charlson–Deyo Index
0–2	≥3	0.68 (0.42–1.10)	0.11	0.99 (0.57–1.76)	0.97
Comorbidity
Congestive heart failure		1.05 (0.71–1.54)	0.81	1.32 (0.84–2.05)	0.22
Hypertension		1.00 (0.71-1.42)	0.99	1.14 (0.76-1.72)	0.54
Diabetes		1.21 (0.76–1.86)	0.41	1.22 (0.69–2.04)	0.49
Liver dysfunction		1.46 (0.94–2.23)	0.09	1.43 (0.83–2.40)	0.19
Concomitant medication
NSAIDs		1.18 (0.79–1.72)	0.40	0.98 (0.59–1.55)	0.94
Antiplatelet drugs		1.19 (0.76–1.79)	0.44	1.49 (0.90–2.38)	0.12

* p-Values <0.05 indicate significant differences. ^† Major bleeding is a composite outcome of intracranial, gastrointestinal, respiratory, or renal/urinary tract bleeding. ^‡ The hazard ratio of the left confounding variable to the right confounding variable is shown. For comorbidity and concomitant medications, it is the ratio of patients with this background to those without. CI, confidence interval; HR, hazard ratio; NSAIDs, nonsteroidal anti-inflammatory drugs.

DISCUSSION

This was the first study to directly compare the safety and effectiveness of four DOACs using a large Japanese claims database.

Our study suggested the following: (1) Dabigatran has a lower risk of major bleeding than apixaban in the current use of DOACs in Japan. (2) Rivaroxaban is not suggested to increase the risk of major bleeding more than other DOACs in Japanese patients. (3) There is no significant difference in effectiveness among the four DOACs.

The results of this study differed from those of previous studies in terms of safety. Most reports stated that apixaban has a lower risk of bleeding than dabigatran^{13,16,18–20)} or that there is no difference between the two drugs.^12,14,15,17) In contrast, our study demonstrated that the major bleeding risk of dabigatran is lower than that of apixaban. We interpreted these results as follows: In the dabigatran group, patients may have been prescribed the reduced dose despite not meeting the reduced dose criteria because the safety flash report (blue letter) was issued on serious bleeding early after its release in Japan. In contrast, in the apixaban group, many patients may have been prescribed the standard dose because they did not meet the reduced dose criteria. We were only able to examine the initial dose, but 48.9% of patients taking dabigatran started at a reduced dose, whereas this proportion for apixaban was 11.4% (Table 2). Thus, the choice of DOAC type and dosage in Japan may be associated with differences in safety. However, we were only able to examine the initial dosage and it may have been adjusted during the course of the study; therefore, the effects of dosage on the risk of bleeding remain unclear.

Apixaban metabolism is predominantly driven by CYP3A4/5.³³⁾ There are several reports stating that the ABCG2 421 A/A and CYP3A5*3 genotypes and renal function are intrinsic factors affecting apixaban pharmacokinetics. In patients with the ABCG2 421C/C or C/A genotype, the population mean of CL/F is 1.49-times higher than that in patients with the ABCG2 421 A/A genotype, whereas in patients with the CYP3A5*1/*1 genotype, the population mean of CL/F is 1.52-times higher than that in patients with the CYP3A5*1/*3 or *3/*3 genotype.³⁴⁾ In addition, there are reports that the plasma trough concentration/dose (C/D) ratio of apixaban is significantly higher in patients with the ABCG2 421 A/A genotype than in patients with the ABCG2 421C/C genotype, and that the plasma trough C/D ratio of apixaban in patients with the CYP3A5*1/*3 or *3/*3 genotype is also significantly higher than that in patients with the CYP3A5*1/*1 genotype.³⁵⁾ In the Japanese population, the genotype frequency of CYP3A5*3/*3, with enzyme protein deficiency, is approximately 60% and that of CYP3A5*1/*3 is approximately 35%.^36,37) Moreover, the frequency of the ABCG2 421C > A allele, which is associated with the reduced expression of drug excretion proteins, is reported to be higher in East Asians (approximately 30–60%) than in Caucasian and African-American populations (approximately 5–10%).³⁸⁾ Therefore, racial differences in genetic polymorphisms that affect the pharmacokinetics of apixaban may be responsible for the differences between our results and those of previous overseas studies.

Rivaroxaban was reported to have a higher risk of major bleeding than apixaban and dabigatran.^12–16) On the other hand, our study did not suggest that rivaroxaban increases the risk of major bleeding more than other DOACs in Japanese patients. We considered two possible reasons for this. The first is the difference in approved doses for rivaroxaban in Japan. Overseas, rivaroxaban is used at 20 and 15 mg in patients with NVAF depending on renal function,²⁴⁾ whereas in Japan, the doses of 15 and 10 mg are approved. The other reason is the difference in patient background from previous studies. In our study, elderly patients were underrepresented and patients with a history of major bleeding within 6 months prior to the index date were excluded. In contrast, in overseas clinical studies, the average age of patients was approximately 70 years^12–14,16) or the proportion of patients over 75 years was 40–50%.^14–16) In addition, several studies included patients with a relatively high mean HAS-BLED score of 3, 15–40% of patients had renal failure,^12,13,16) and 10–30% of patients had a history of bleeding.^13–15) Therefore, caution should be exercised when administering DOACs to elderly patients, patients with a history of bleeding, and patients with CrCl of less than approximately 30 mL/min who were excluded from the randomized clinical trials of DOACs.^5–8)

The follow-up period in this study was comparable to that in previous studies.^14,16,19,23) However, the follow-up period was shorter than that in the randomized clinical trials of DOACs.^5–8) Therefore, we need to be aware of their changing circumstances (age, renal function, weight, concomitant medications, etc.) for long-term DOAC users.

This study has some important limitations. First, the study population had a lower proportion of elderly patients. As the JMDC claims database is constructed using information of beneficiaries covered by the employees’ health insurance system, most data are from working adults or their family members. Accordingly, we need to consider individual risks more carefully when prescribing DOACs to elderly patients. Second, as we used the claims database, we were able to collect the prescription history of DOACs and concomitant medications, but not information on medication compliance, length of time on interacting drugs, or the frequency of concomitant medications. NSAIDs and antiplatelet drugs are reported to increase the risk of bleeding.³⁹⁾ Furthermore, a previous study suggested that the concurrent use of amiodarone, fluconazole, rifampin, and phenytoin compared with the use of DOACs alone is associated with an increased risk of major bleeding.⁴⁰⁾ Therefore, we need to assess patient compliance and gather accurate information about concomitant medications. Third, we were unable to investigate whether the dose of DOACs prescribed to patients was appropriate because the information on body weight and renal function required for dosage determination is not available from the JMDC claims database. Fourth, in general, the event per variable (EPV) should be 10 or more in multivariate analysis, but the EPV was below 10 in the effectiveness analysis. Therefore, there is a possibility of overfitting, which may lead to unstable results. In the future, comparative validation in larger DOAC clinical trials is necessary. Fifth, we defined non-valvular atrial fibrillation and other diseases selected for the outcomes and confounding factors using ICD-10, ATC, or medical procedure codes. In order to improve the accuracy as much as possible, we excluded cases of suspected diseases in the JMDC claims database for all diseases. However, due to a lack of validation studies in Japan for these diseases, the definition of each disease used in this study may not ensure a high positive predictive value. Lastly, we were unable to analyze by dosage because of the small number of events and the possibility of overfitting. We considered that if the reduced risk of bleeding in this study was associated with the use of low-dose medication and this resulted in a reduction in effectiveness, the results would not be clinically meaningful; therefore, we also compared the effectiveness of the four DOACs. The safety and effectiveness of each DOAC by dosage need to be assessed in a larger clinical study. In addition, we were unable to sufficiently examine the results by bleeding site due to the small number of patients and events. DOACs are reported to have a lower risk of ICH than warfarin, but have a higher or comparable risk of gastrointestinal bleeding, except for low-dose edoxaban.^5–9) It thus remains unclear whether there is a difference in the organ-specific bleeding risk among DOACs, although there are numerous reports. In the future, it will be necessary to accumulate data on the Japanese population for more detailed assessment.

In conclusion, in the current use of DOACs in Japanese patients with NVAF, dabigatran had a significantly lower risk of major bleeding than apixaban, but there was no significant difference in effectiveness. This study has limitations due to the small number of elderly patients, the use of claims data, and the lack of dose-specific analysis. We believe that comparative assessment of DOACs in a larger number of patients is necessary for appropriate drug selection.

Conflict of Interest

The authors declare no conflict of interest.

REFERENCES

• 1) Wolf PA, Abbott RD, Kannel WB. Atrial fibrillation as an independent risk factor for stroke: the Framingham study. Stroke, 22, 983–988 (1991).
• 2) Savelieva I, Bajpai A, Camm AJ. Stroke in atrial fibrillation: update on pathophysiology, new antithrombotic therapies, and evolution of procedures and devices. Ann. Med., 39, 371–391 (2007).
• 3) Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statistics—2012 update: a report from the American Heart Association. Circulation (New York, N.Y.), 125, e2–e220 (2012).
• 4) Lin HJ, Wolf PA, Kelly-Hayes M, Beiser AS, Kase CS, Benjamin EJ, D’Agostino RB. Stroke severity in atrial fibrillation. the Framingham study. Stroke, 27, 1760–1764 (1996).
• 5) Granger CB, Alexander JH, McMurray JJV, et al. Apixaban versus warfarin in patients with atrial fibrillation. N. Engl. J. Med., 365, 981–992 (2011).
• 6) Giugliano RP, Ruff CT, Braunwald E, et al. Edoxaban versus warfarin in patients with atrial fibrillation. N. Engl. J. Med., 369, 2093–2104 (2013).
• 7) Patel MR, Mahaffey KW, Garg J, Pan G, Singer DE, Hacke W, Breithardt G, Halperin JL, Hankey GJ, Piccini JP, Becker RC, Nessel CC, Paolini JF, Berkowitz SD, Fox KAA, Califf RM. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N. Engl. J. Med., 365, 883–891 (2011).
• 8) Connolly SJ, Ezekowitz MD, Yusuf S, Eikelboom J, Oldgren J, Parekh A, Pogue J, Reilly PA, Themeles E, Varrone J, Wang S, Alings M, Xavier D, Zhu J, Diaz R, Lewis BS, Darius H, Diener H, Joyner CD, Wallentin L. Dabigatran versus warfarin in patients with atrial fibrillation. N. Engl. J. Med., 361, 1139–1151 (2009).
• 9) Ruff CT, Giugliano RP, Braunwald E, Hoffman EB, Deenadayalu N, Ezekowitz MD, Camm AJ, Weitz JI, Lewis BS, Parkhomenko A, Yamashita T, Antman EM. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: A meta-analysis of randomised trials. Lancet, 383, 955–962 (2014).
• 10) January CT, Wann LS, Calkins H, Chen LY, Cigarroa JE, Cleveland JC Jr, Ellinor PT, Ezekowitz MD, Field ME, Furie KL, Heidenreich PA, Murray KT, Shea JB, Tracy CM, Yancy CW. 2019 AHA/ACC/HRS focused update of the 2014 AHA/ACC/HRS guideline for the management of patients with atrial fibrillation. Heart Rhythm, 16, e66–e93 (2019).
• 11) Wang KL, Lip G, Lin S, Chiang C. Non–Vitamin K antagonist oral anticoagulants for stroke prevention in Asian patients with nonvalvular atrial fibrillation: meta-analysis. Stroke, 46, 2555–2561 (2015).
• 12) Hernandez I, Zhang Y, Saba S. Comparison of the effectiveness and safety of apixaban, dabigatran, rivaroxaban, and warfarin in newly diagnosed atrial fibrillation. Am. J. Cardiol., 120, 1813–1819 (2017).
• 13) Noseworthy PA, Yao X, Abraham NS, Sangaralingham LR, McBane RD, Shah ND. Direct comparison of dabigatran, rivaroxaban, and apixaban for effectiveness and safety in nonvalvular atrial fibrillation. Chest, 150, 1302–1312 (2016).
• 14) Adeboyeje G, Sylwestrzak G, Barron JJ, White J, Rosenberg A, Abarca J, Crawford G, Redberg R. Major bleeding risk during anticoagulation with warfarin, dabigatran, apixaban, or rivaroxaban in patients with nonvalvular atrial fibrillation. J. Manag. Care Spec. Pharm., 23, 968–978 (2017).
• 15) Staerk L, Gerds TA, Lip GYH, Ozenne B, Bonde AN, Lamberts M, Fosbøl EL, Torp-Pedersen C, Gislason GH, Olesen JB. Standard and reduced doses of dabigatran, rivaroxaban and apixaban for stroke prevention in atrial fibrillation: a nationwide cohort study. J. Intern. Med., 283, 45–55 (2018).
• 16) Lip GYH, Keshishian A, Li X, Hamilton M, Masseria C, Gupta K, Luo X, Mardekian J, Friend K, Nadkarni A, Pan X, Baser O, Deitelzweig S. Effectiveness and safety of oral anticoagulants among nonvalvular atrial fibrillation patients: the ARISTOPHANES study. Stroke, 49, 2933–2944 (2018).
• 17) Lip GYH, Keshishian A, Kamble S, Pan X, Mardekian J, Horblyuk R, Hamilton M. Real-world comparison of major bleeding risk among non-valvular atrial fibrillation patients initiated on apixaban, dabigatran, rivaroxaban, or warfarin. Thromb. Haemost., 116, 975–986 (2016).
• 18) Lamberts M, Staerk L, Olesen JB, Fosbøl EL, Hansen ML, Harboe L, Lefevre C, Evans D, Gislason GH. Major bleeding complications and persistence with oral anticoagulation in non-valvular atrial fibrillation: Contemporary findings in real-life Danish patients. J. Am. Heart Assoc., 6, e004517 (2017).
• 19) Amin A, Keshishian A, Trocio J, Dina O, Le H, Rosenblatt L, Liu X, Mardekian J, Zhang Q, Baser O, Nadkarni A, Vo L. A real-world observational study of hospitalization and health care costs among nonvalvular atrial fibrillation patients prescribed oral anticoagulants in the U.S. medicare population. J. Manag. Care Speci.Pharm., 24, 911–920 (2018).
• 20) Gupta K, Trocio J, Keshishian A, Zhang Q, Dina O, Mardekian J, Rosenblatt L, Liu X, Hede S, Nadkarni A, Shank T. Real-world comparative effectiveness, safety, and health care costs of oral anticoagulants in nonvalvular atrial fibrillation patients in the U.S. department of defense population. J. Manag. Care Speci. Pharm., 24, 1116–1127 (2018).
• 21) López-López JA, Sterne JAC, Thom HHZ, Higgins JPT, Hingorani AD, Okoli GN, Davies PA, Bodalia PN, Bryden PA, Welton NJ, Hollingworth W, Caldwell DM, Savović J, Dias S, Salisbury C, Eaton D, Stephens-Boal A, Sofat R. Oral anticoagulants for prevention of stroke in atrial fibrillation: Systematic review, network meta-analysis, and cost effectiveness analysis. BMJ, 359, j5058 (2017).
• 22) Chan YH, Lee HF, See LH, Tu H, Chao T, Yeh Y, Wu L, Kuo C, Chang S, Lip GYH. Effectiveness and safety of four direct oral anticoagulants in Asian patients with nonvalvular atrial fibrillation. Chest, 156, 529–543 (2019).
• 23) Mueller T, Alvarez-madrazo S, Robertson C, Wu O, Bennie M. Comparative safety and effectiveness of direct oral anticoagulants in patients with atrial fibrillation in clinical practice in Scotland. Br. J. Clin. Pharmacol., 85, 422–431 (2019).
• 24) XARELTO [package insert] 2011. Janssen Pharmaceuticals, Inc.; Titusville, NJ.
• 25) PRADAXA [package insert] 2010. Boehringer Ingelheim Pharmaceuticals, Inc.; Ridgefield, CT.
• 26) “Features of JMDC Claims Database”: ‹https://www.jmdc.co.jp/pharma/database.html›, accessed 14 March, 2021.
• 27) Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi J, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med. Care, 43, 1130–1139 (2005).
• 28) Quan H, Li B, Couris CM, Fushimi K, Graham P, Hider P, Januel J, Sundararajan V. Updating and validating the Charlson comorbidity index and score for risk adjustment in hospital discharge abstracts using data from 6 countries. Am. J. Epidemiol., 173, 676–682 (2011).
• 29) “JCS/JHRS Guideline on pharmacotherapy of cardiac arrhythmias. (2020 edition)”: ‹https://www.j-circ.or.jp/old/guideline/pdf/JCS2020_Ono.pdf›, accessed 12 June, 2021.
• 30) Hamatani Y, Ogawa H, Takabayashi K, Yamashita Y, Takagi D, Esato M, Chun Y, Tsuji H, Wada H, Hasegawa K, Abe M, Lip GYH, Akao M. Left atrial enlargement is an independent predictor of stroke and systemic embolism in patients with non-valvular atrial fibrillation. Sci. Rep., 6, 31042 (2016).
• 31) Suzuki S, Yamashita T, Okumura K, Atarashi H, Akao M, Ogawa H, Inoue H. Incidence of ischemic stroke in Japanese patients with atrial fibrillation not receiving anticoagulation therapy—pooled analysis of the Shinken Database, J-RHYTHM Registry, and Fushimi AF Registry. Circ. J., 79, 432–438 (2015).
• 32) Kodani E, Atarashi H, Inoue H, et al. Impact of blood pressure control on thromboembolism and major hemorrhage in patients with nonvalvular atrial fibrillation: a subanalysis of the J-RHYTHM registry. J. Am. Heart Assoc., 5, e004075 (2016).
• 33) Wang L, Zhang D, Raghavan N, Yao M, Ma L, Frost CA, Maxwell BD, Chen S, He K, Goosen TC, Humphreys WG, Grossman SJ. In vitro assessment of metabolic drug–drug interaction potential of apixaban through cytochrome P450 phenotyping, inhibition, and induction studies. Drug Metab. Dispos., 38, 448–458 (2010).
• 34) Ueshima S, Hira D, Kimura Y, Fujii R, Tomitsuka C, Yamane T, Tabuchi Y, Ozawa T, Itoh H, Ohno S, Horie M, Terada T, Katsura T. Population pharmacokinetics and pharmacogenomics of apixaban in Japanese adult patients with atrial fibrillation. Br. J. Clin. Pharmacol., 84, 1301–1312 (2018).
• 35) Ueshima S, Hira D, Fujii R, Kimura Y, Tomitsuka C, Yamane T, Tabuchi Y, Ozawa T, Itoh H, Horie M, Terada T, Katsura T. Impact of ABCB1, ABCG2, and CYP3A5 polymorphisms on plasma trough concentrations of apixaban in Japanese patients with atrial fibrillation. Pharmacogenet. Genomics, 27, 329–336 (2017).
• 36) Hiratsuka M, Takekuma Y, Endo N, Narahara K, Hamdy SI, Kishikawa Y, Matsuura M, Agatsuma Y, Inoue T, Mizugaki M. Allele and genotype frequencies of CYP2B6 and CYP3A5 in the Japanese population. Eur. J. Clin. Pharmacol., 58, 417–421 (2002).
• 37) Saeki M, Saito Y, Nakamura T, Murayama N, Kim S, Ozawa S, Komamura K, Ueno K, Kamakura S, Nakajima T, Saito H, Kitamura Y, Kamatani N, Sawada J. Single nucleotide polymorphisms and haplotype frequencies of CYP3A5 in a Japanese population. Hum. Mutat., 21, 653 (2003).
• 38) Hira D, Terada T. BCRP/ABCG2 and high-alert medications: Biochemical, pharmacokinetic, pharmacogenetic, and clinical implications. Biochem. Pharmacol., 147, 201–210 (2018).
• 39) Lip GYH, Frison L, Halperin JL, Lane DA. Comparative validation of a novel risk score for predicting bleeding risk in anticoagulated patients with atrial fibrillation: the HAS-BLED (Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR, Elderly, Drugs/Alcohol Concomitantly) score. J. Am. Coll. Cardiol., 57, 173–180 (2011).
• 40) Chang SH, Chou IJ, Yeh YH, Chiou MJ, Wen MS, Kuo CT, See LC, Kuo CF. Association between use of non-vitamin K oral anticoagulants with and without concurrent medications and risk of major bleeding in nonvalvular atrial fibrillation. JAMA, 318, 1250–1259 (2017).