Anticoagulant therapy in elderly patients with atrial brillation : A time series study using the nationwide claim database in Japan

Background and aim: The introduction of direct oral anticoagulants (DOACs) has greatly changed the use of anticoagulant therapy in patients with non-valvular atrial brillation (Af). However, few studies have reported on the current state of anticoagulant therapy at the national level. This study aimed to examine changes in the proportions of oral anticoagulant (OAC) prescriptions (4 DOACs and a vitamin K antagonist: VKA) in patients with non-valvular Af aged ≥ 65 years, taking into consideration the risk of cerebral infarction and bleeding. Methods: Anticoagulant prescriptions in outpatients with Af were temporally analyzed using the nationwide claims database in Japan. The proportion of anticoagulants prescribed to Af patients aged ≥ 65 years was determined. Trends in anticoagulant prescriptions were examined according to cerebral infarction and bleeding risk. Results: The proportion of anticoagulant prescriptions for 12,076 Af patients aged ≥ 65 years in Japan increased from 41% in 2011 to 56% in 2015. The proportion of prescriptions for DOACs surpassed that of VKA. An increase in DOAC prescriptions was accompanied by an increase in the proportion of anticoagulant prescriptions in each group according to the CHA2DS2-VASc and HAS-BLED scores. The proportion of anticoagulant prescriptions for patients with a high risk of developing cerebral infarction and bleeding showed a marked increase. Conclusion: Trends in anticoagulant prescriptions in patients with non-valvular AfAf in Japan showed a marked increase in DOAC prescriptions. The widespread use of DOACs greatly changes the proles of cerebral infarction and bleeding risks, which are related to the prescription of anticoagulant therapy in patients with non-valvular Af. agent. With the release of dabigatran in 2011, prescriptions for direct oral anticoagulants (DOACs) have increased rapidly 3–9 . While recent guidelines around the world recommend prescribing DOACs 10–13 , it remains unclear how indications and risk of bleeding, a serious side effect of anticoagulants, are considered when these drugs are prescribed. In fact, there have been no reports on the current state of anticoagulant prescriptions at the national level. Accordingly, the present study aimed to reveal the state of anticoagulant prescriptions in Japanese elderly Af patients aged ≥ 65 years, as well as the risk of cerebral infarction and bleeding. To this end, trends in anticoagulant prescriptions were examined using data obtained from the National Database of Health Insurance Claims and Specic Health Checkups of Japan (NDB), a database that covers more than 95% of all claims data across Japan.


Background
Atrial brillation (Af) is a common arrhythmia in the elderly. Af is associated with the development of cardiogenic cerebral infarction 1 , with serious outcomes including physical dysfunction and death 2 .
In order to prevent the development of cardiogenic cerebral infarction, Af patients are conventionally prescribed warfarin potassium (vitamin K antagonist: VKA), an anticoagulant agent. With the release of dabigatran in 2011, prescriptions for direct oral anticoagulants (DOACs) have increased rapidly [3][4][5][6][7][8][9] . While recent guidelines around the world recommend prescribing DOACs [10][11][12][13] , it remains unclear how indications and risk of bleeding, a serious side effect of anticoagulants, are considered when these drugs are prescribed. In fact, there have been no reports on the current state of anticoagulant prescriptions at the national level. Accordingly, the present study aimed to reveal the state of anticoagulant prescriptions in Japanese elderly Af patients aged ≥ 65 years, as well as the risk of cerebral infarction and bleeding. To this end, trends in anticoagulant prescriptions were examined using data obtained from the National Database of Health Insurance Claims and Speci c Health Checkups of Japan (NDB), a database that covers more than 95% of all claims data across Japan.

1) Study design and database used
This time-series study used a sampling dataset of NDB 14 . Of the entire dataset available at the time of the study (including data from inpatients, outpatients, and pharmacy claims, as well as DPC data; note that pharmacy claims include data for November, which are linked to outpatient data), we applied to and obtained approval from the Ministry of Health, Labour and Welfare for the use of patients claims data from October of each year from 2011 to 2015. In this dataset we used outpatients and pharmacy claims data. This NDB sampling dataset comprises data (10% each for inpatients and DPC data, and 1% each for outpatients and pharmacy claims data) extracted from a highly exhaustive claims database that stores claims information collected from over 95% of the Japanese population, from infants to elderly individuals [14][15][16][17][18] . This database includes data on patient age and sex, ambulatory services and diagnoses, medical treatment, and aid, as well as drug dispensation. All diagnoses in the database were coded according to the 10th revision of the International Classi cation of Diseases (ICD-10). Pharmacy claims include various information on outpatient prescriptions dispensed from Japan pharmacies, such as dispensing/prescription dates and the number of tablets dispensed. Prescription handouts were coded according to the Anatomical Therapeutic Chemical Classi cation System.

2) Subjects
The following selection criteria were used: patients aged ≥ 65 years with a disease code corresponding to the ICD-10 code (I48: atrial brillation) in their medical claims 19 . The exclusion criteria were patients with a disease code for valvular Af or valvular disease recorded in their medical claims (i.e., those with an ICD-10 code I05, I06, I34, or I35), those with venous thromboembolism (I802 or I269), and those who underwent arti cial joint replacement (total hip arthroplasty, knee arthroplasty). The extracted non-valvular Af patients were considered the subjects of this study (hereafter, Af patients). Oral anticoagulants (OACs) were de ned as VKA and four DOACs (dabigatran, rivaroxaban, apixaban, and edoxaban), coded in pharmacy claims data as 3332-D01280, 3339-D07082, D07086, D03213, and D09546, according to the ATC classi cation.
Patients who had been prescribed an antiplatelet agent (AP) were extracted, and those who had been prescribed both anticoagulants and AP were considered candidates for anticoagulant therapy. For drug codes, medical claim codes corresponding to the ATC classi cation numbers were used 19 . The list of drug and disease codes used for subject extraction is shown in Supplementary Table S1.

3) Evaluation indices
The primary evaluation indices were proportions of Af patients who were prescribed VKA, DOACs, and AP in October of each year from 2011 to 2015, as well as those who were not prescribed any of the above-mentioned drugs, among all Af patients.

4) Measurement item
Subject characteristics included age, sex, comorbidities (cerebral infarction, myocardial infarction, ischemic heart disease, peripheral artery disease, heart failure, chronic kidney disease, liver disease, hemorrhagic disease, alcoholic disease, hypertension, and diabetes mellitus), and concomitant drugs (non-steroidal anti-in ammatory drugs [NSAIDs] ). Each disease was extracted using disease codes recorded in medical claims 19 corresponding to ICD-10 codes. Patients with hypertension (de ned as those who were prescribed antihypertensive drugs) and patients with diabetes (de ned as those who were prescribed anti-diabetic drugs) were extracted and included as subjects. Patients with renal dysfunction, de ned as those with a treatment code corresponding to dialysis recorded in their medical claims 19 , were also included. The list of drug and disease codes used for subject extraction is shown in Supplementary Table S1.

5) CHA 2 DS 2 -VASc and HAS-BLED scores
The CHA 2 DS 2 -VASc score was used to assess the risk of developing cerebral infarction, and a total score was obtained by adding 1 point each for the presence of heart failure, hypertension, diabetes mellitus, and vascular disease (history of myocardial infarction, peripheral arterial disease, aortic plaques), female sex, and age 65-74 years, and 2 points each for age ≥ 75 years and prior cerebral infarction/TIA 20 . The HAS-BLED score was used to assess the risk of bleeding during anticoagulant therapy, and a total score was obtained by adding 1 point each for hypertension, prior history of cerebral infarction/TIA, kidney dysfunction (dialysis, kidney transplant), liver dysfunction (cirrhosis), prior bleeding or predisposition to bleeding, age ≥ 65 years, use of AP and NSAIDs, alcoholism, and unstable INR 21 . Since INR values were not available, the highest HAS-BLED score was 8 points.

6) Analyses
Data were analyzed using descriptive statistics, followed by the Cochrane Armitage test to examine changes over time in prescriptions for each drug. The entire Af patient population was divided into four groups: two groups according to the CHA 2 DS 2 -VASc score (< 2 or ≥ 2) and two groups according to the HAS-BLED score (< 3, ≥ 3), based on data from to 2011-2015. The numbers of VKA and DOAC prescriptions and proportions of prescriptions for each of these drugs among all Af patients were calculated for each group. The rst of the four groups comprised patients with "low thrombus risk and low bleeding risk," for whom anticoagulant therapy is only considered an option (Group 1); the second group comprised patients with "low risk of developing cerebral infarction and high risk of bleeding," for whom caution is required when prescribing anticoagulant therapy (Group 2); the third group comprised patients with "high risk of developing cerebral infarction and low risk of bleeding," for whom anticoagulant therapy is recommended (Group 3); and the fourth group comprised patients with "both high risk of developing cerebral infarction and high risk of bleeding," for whom clinical judgment is di cult due to opposing risks and bene ts of anticoagulant therapy for non-valvular Af patients (Group 4).
Statistical analyses were performed using R (version 3.3.2) (Copyright (C) 2016 The R Foundation for Statistical Computing), with a twotailed test at a signi cance level of 5%.

1)Subject extraction
The same subject extraction process was followed over a 5-year period. The representative results from 2015 are described below.
A total of 14,363 Af patients aged ≥65 years were extracted from the 2015 outpatient claims data. Of these, 12,076 Af patients were subjected to analysis after excluding 2,049 patients with valvular Af or valvular disease, 173 patients with venous thromboembolism, and 65 patients who were prescribed two or more anticoagulants. The characteristics of the patients are summarized in Table 1. Mean age (SD) was 78.8 (7.7) years, mean CHADS 2 score (SD) was 2.5 (1.4), mean CHA 2 DS 2 -VASc score (SD) was 4.1 (1.5), and mean HAS-BLED score (SD) was 2.6 (1.1). Subjects were divided into those who were prescribed anticoagulants (n=6,761) and those who were not (n=5,315); the former group was further classi ed into those who were prescribed VKA (n=2,991) and those who were prescribed DOACs (n=3,770) ( Figure  1). The results from the datasets (2011 to 2014) were obtained using the same procedure.

2) Time trends in the numbers of oral anticoagulant prescriptions
Time trends in the number of patients who were prescribed VKA, each of the four DOACs, and that of all DOAC-prescribed patients combined, among all Af patients aged ≥ 65 years, are shown in Fig. 2. As the total number of Af patients increased from 2011 to 2015, the number of patients who were prescribed DOACs signi cantly increased (339 to 3770), whereas the number of patients who were prescribed VKA signi cantly decreased (3051 to 2991).  Table 2. The proportion of anticoagulant prescriptions overall changed from 18-26% in Group 1, 37-24% in Group 2, 33-49% in Group 3, and 54-66% in Group 4, with the proportion of DOAC prescriptions reaching 20%, 16%, 29%, and 35% in Groups 1, 2, 3, and 4, respectively, in 2015. In 2015, there were 11,642 Af patients in Groups 3 and 4 combined, and these patients accounted for 96% of all Af patients that year. In these 11,642 Af patients, the proportion of those who were prescribed anticoagulants was 58% and that of those who were prescribed DOACs was 32%. The proportion of OAC prescriptions was higher in Group 4 than in Group 3.

Discussion
With an increase in DOAC prescriptions, the proportion of anticoagulant prescriptions for Af patients aged ≥ 65 years increased from 41-56% from 2011 to 2015. The proportion of anticoagulant prescriptions increased from 33-49% in the group of patients with a CHA 2 DS 2 -VASc score ≥ 2 and HAS-BLED scores < 3 (i.e., for whom anticoagulant therapy is recommended), and from 54-66% in the group of patients with a CHA 2 DS 2 -VASc score ≥ 2 and HAS-BLED scores ≥ 3, from 2011 to 2015. These ndings from highly exhaustive NDB data shed light on how anticoagulants are prescribed in Japan.

1) Changes over time in the numbers of anticoagulant prescriptions
The number of anticoagulant prescriptions overall increased from 2011 to 2015; the number of DOAC prescriptions surpassed that of VKA prescriptions, which showed a decreasing trend. These results were consistent with previous reports from Denmark, Germany, France, and the United States 3 − 7 , as well as in Japanese registry studies 8 − 9 . Clinical studies that examined four DOACs demonstrated no differences between VKA and DOACs 22 − 27 and clinical practice guidelines in Japan, and those of other countries now recommend the use of DOACs. This may explain our results as well as those reported previously. Meanwhile, Behdarvand et al. pointed out that intense sales promotion efforts by pharmaceutical companies that market DOACs toward Australian clinicians might have been a driving force behind the increase in DOAC prescriptions 28 . Although there are no such reports in Japan, since DOACs are more expensive than VKA, we cannot deny the possibility that aggressive information provision by companies may have in uenced clinicians' prescribing behaviors.

2) Proportion of anticoagulant prescriptions among all Af patients
As for the cause of the increase in the proportion of anticoagulant prescriptions among all Af patients ( 41% in 2011 to 56% in 2015), it is possible that the introduction of DOACs may have helped reduce the number of patients for whom the attending physician avoided prescribing an anticoagulant. It is conceivable that, when VKA was prescribed as a single therapy, a considerable number of patients were not prescribed anticoagulants. In addition, from 2011 to 2015, there was an increase in the number of Af patients. This suggests that more patients were likely to have been diagnosed with Af, and thus, the number of new patients with Af who were prescribed DOACs also increased. However, the proportion of anticoagulant prescriptions was lower than that reported previously in studies from other countries [3][4][5][6][7] . Given that the risk of developing cerebral infarction is lower among Japanese non-valvular Af patients than among European and American patients 29 , the lower proportion of anticoagulant prescriptions in Japan might re ect a cautious attitude on the part of physicians who prescribe anticoagulant therapy. Moreover, physicians may hold up prescriptions for elderly patients due to a high number of comorbidities and concomitant drugs. Although the e cacy of AP has not been demonstrated 30 , it is still prescribed to some patients with coronary artery disease who have undergone PCI or CABG 31 . However, as concomitant anticoagulants have been reported to be associated with poor prognosis 32 , the use of AP is expected to decrease further in the future.

3) Changes over time in the proportion of prescriptions strati ed by cerebral infarction risk and bleeding risk
The proportion of anticoagulant prescriptions, especially DOACs, increased in all four groups. However, this may re ect possible overuse of anticoagulants in Group 1 and 2 patients, who have a low need to receive anticoagulant therapy. Moreover, the proportion of anticoagulant prescriptions was higher in Group 4 patients who had a high risk of bleeding compared to Group 3 patients, and for whom anticoagulant therapy was recommended. This likely resulted from a prescribing behavior that only considers the risk of cerebral infarction when prescribing OAC; that is, patients with a high HAS-BLED score were administered OAC. Compared with prescription trends in other countries 4 , our results suggest the possibility that anticoagulants may be prescribed based only on cerebral infarction risk. It is also possible that the attitudes of clinicians may change in the direction of "if in doubt, prescribe." With the introduction of anticoagulant therapy, there has emerged a need to consider safety and reduce risks. Efforts should be made to control blood pressure to prevent bleeding, to instruct patients to reduce alcohol consumption, and to avoid the use of NSAIDs and other drugs 33 . Whether the increase in DOAC administration reduced the risk of developing cerebral infarction or increased the risk of bleeding in Af patients should be evaluated in future studies.

Study Limitations
This study has several limitations. First, the NDB sampling dataset comprised single-month data and not a continuous set of data. Accordingly, whether patients were newly prescribed DOACs or switched from VKA to DOACs could not be veri ed. Similarly, at the time of OAC prescription, whether it was prescribed in consideration of CHA 2 DS 2 -VASc and HAS-BLED scores could not be veri ed. Second, since insurance claims data contain no descriptions of laboratory test values (INR, kidney function, blood pressure, body weight, etc.), CHA 2 DS 2 -VASc and HAS-BLED scores were calculated based on disease names and drugs used. This likely resulted in higher scores relative to actual risks, given that patients with hypertension were identi ed based on the presence of administered anti-hypertensive drugs, that is, some patients who were included in this patient group might not have had hypertension. For example, there might have been cases in which patients were assigned a disease code for hypertension so that antihypertensive drugs could be administered for the purpose of treating IgA nephropathy. Even though such cases might not be common, this could have led to high CHA 2 DS 2 -VASc and HAS-BLED scores. Finally, comorbidities and concomitant drugs were also likely confounders, but were not fully assessed in this analysis.

Conclusions
From 2011 to 2015, anticoagulant prescriptions for patients with non-valvular Af rapidly shifted from VKA to DOACs, and the proportion of total anticoagulant prescriptions increased in Japan. While the proportion of anticoagulant prescriptions for patients with a high risk of cerebral infarction increased, that for patients at a high risk of bleeding also increased. Due to the widespread use of DOACs, the pro les of these two risks, which are associated with anticoagulant prescriptions in non-valvular Af patients, are drastically changing.

Declarations Ethical considerations
This study complied with the guidelines of the Ministry of Health, Labour and Welfare regarding the provision of health insurance claims information and special health check-up information (revised April 2015) 16 and was approved by the Ethics Committee of Kyoto University Faculty of Medicine (approval number: R1013).

Consent for publication
Not applicable" in this section.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Con ict of interests
The authors have no con icts of interest regarding the contents of this study.

Funding
The budget of Department of Health Informatics, Graduate School of Medicine and Public Health, Kyoto University, Japan.