CLINICAL INVESTIGATION Dabigatran, Rivaroxaban, and Warfarin in the Oldest Adults with Atrial Fibrillation in Taiwan Chao-Lun Lai, MD, PhD,*†‡§ Ho-Min Chen, MS,† Min-Tsun Liao, MD,* and Ting-Tse Lin, MD* mortality and cardiovascular mortality than those who OBJECTIVES: To compare the effectiveness and safety of used warfarin. Reduced-dose dabigatran was also associ- reduced-dose dabigatran, reduced-dose rivaroxaban, and warfa- ated with lower risk of intracranial hemorrhage than war- rin in individuals aged 85 and older with atrial fibrillation (AF). farin. J Am AmGeriatr GeriatrSoc Soc66:1567–1574, 2018. 2018. DESIGN: Retrospective cohort study. SETTING: Taiwan National Health Insurance claims database, 20112015. Key words: dabigatran; rivaroxaban; warfarin; effec- PARTICIPANTS: Individuals with AF aged 85 and older tiveness; safety; octogenarian (mean 88.6) with incident use of oral anticoagulants between June 1, 2012 and May 31, 2015 (N54,722; dabi- gatran 110 mg, n51,489; rivaroxaban 15 mg/10 mg, n51,736; warfarin, n51,497). MEASUREMENTS: Clinical outcomes included all-cause death, cardiovascular death, ischemic stroke, acute myocardial he risk of ischemic stroke is 5 times as high in individ- infarction, arterial embolism or thrombosis, intracranial hem- T uals with atrial fibrillation (AF) than in those with- orrhage, and gastrointestinal hemorrhage needing transfusion. out.1 Warfarin, the classic vitamin K antagonist, can reduce 2 Propensity score–matched analysis was performed, and the the risk of ischemic stroke by approximately 60%, but the marginal proportional hazards model was used to estimate the narrow therapeutic window and risk of bleeding complica- relative risk of various clinical outcomes in a matched tions associated with warfarin therapy have led to its being 1 dabigatran-warfarin cohort (n51,180 in each group) and a underused. Direct oral anticoagulants (DOACs) such as rivaroxaban-warfarin cohort (n51,207 in each group) dabigatran and rivaroxaban are approved for prevention of RESULTS: Mean follow-up was 6.6 months for the overall ischemic stroke and systemic embolism in individuals with population. Dabigatran group participants had lower risks of nonvalvular AF, and their clinical efficacy and safety have 3,4 all-cause death (hazard ratio (HR)50.59, 95% confidence been established in large-scale clinical trials. Several 5–9 10 interval (CI)50.45–0.77) and cardiovascular death investigations, including one report from Taiwan, using (HR50.45, 95% CI50.30–0.68) than warfarin group partic- real-world databases have also supported that DOACs are ipants. Rivaroxaban users also had lower risks of all-cause more effectiveness and safer than warfarin, but most partic- death (HR50.61, 95% CI50.47–0.79) and cardiovascular ipants in those studies were aged 65 to 80, with a mean death (HR50.52, 95% CI50.35–0.75) than warfarin users. age of 70 to 75, with the landmark study enrolling individ- Dabigatran users also had a lower risk of intracranial hemor- uals with AF aged 75 and older or 65 to 74 plus major car- diovascular comorbidities.3 Thus, individuals aged 80 and rhage than warfarin users (HR50.31, 95% CI50.10–0.97). older have been underrepresented in clinical studies. CONCLUSION: Individuals with AF aged 85 and older The aim of this study was to provide real-world data to who used reduced-dose dabigatran or reduced-dose rivar- compare the effectiveness and safety of dabigatran, rivarox- oxaban had statistically significantly lower all-cause aban, and warfarin in individuals aged 85 and older using a retrospective cohort study design based on claims data from the National Health Insurance (NHI) program in Taiwan. From the *Department of Internal Medicine and †Center for Critical Care Medicine, National Taiwan University Hospital Hsin-Chu Branch, Hsin-Chu, Taiwan; ‡Department of Internal Medicine, College of Medicine and §Institute of Epidemiology and Preventive Medicine, College METHODS of Public Health, National Taiwan University, Taipei, Taiwan. Address correspondence to Chao-Lun Lai, MD, PhD, No. 25, Lane 442, Data sources Sec. 1, Jingguo Rd., Hsinchu City 30059, Taiwan. E-mail: chaolunlai@ ntu.edu.tw Taiwan has provided compulsory, universal insurance cov- DOI: 10.1111/jgs.15430 erage for all citizens since 1995. Identification number, JAGS 66:1567–1574, 2018 2018 V©C 2018, Copyrightthe Author Authors Journal compilation ©VC 2018, The American AmericanGeriatrics GeriatricsSociety Society 0002-8614/18/$15.00 15682 LAILAI ETET AL.AL. AUGUST 2018–VOL. 66, NO.20188 JAGS sex, birthdate, dates of outpatient clinic visits, dates of hospital admissions and discharges, diagnoses, procedures administered, dates of pharmacy dispensing, and drugs dis- pensed are available in the NHI claims database. Diagno- ses are coded according to the International Statistical Classification of Diseases and Related Health Problems, Ninth Revision, Clinical Modification (ICD-9-CM) system. Validation studies of diagnosis codes for diabetes melli- tus,11 ischemic stroke,12,13 and acute myocardial infarc- tion14 in Taiwan NHI claims database have been reported. An individual’s record can be linked to the Taiwan National Death Registry to obtain the exact date of death and the officially speculated main cause of death coded according to the ICD-10 system. The NHI has reimbursed for dabigatran for stroke prevention in individuals with AF with an estimated glomerular filtration rate (eGFR) of 30 mL/min per 1.73 m2 or higher since June 1 2012, and for rivaroxaban since February 1, 2013. Ethical approval To comply with Taiwanese privacy regulations, all perso- nal identifiers were encrypted, and all data were analyzed anonymously. The study protocol conforms to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the institutional review board of the National Taiwan University Hospital Hsin-Chu Branch (104–009-E), Figure 1. Participant flow diagram. *Antiplatelet agents which waived requirement for informed consent. included aspirin, clopidogrel, ticlopidine, ticagrelor, dipyrida- mole, and cilostazol. DOAC5direct oral anticoagulant; DVT5deep vein thrombosis; MS5mitral stenosis; PE5pulmo- Study design and cohort definition nary embolism. We used the NHI claims database from 2011 to 2015. 15 The study design was a retrospective cohort study. All gastrointestinal hemorrhage needing transfusion. We adult beneficiaries aged 85 and older with a diagnosis of also included osteoporotic fracture as the falsification 9 AF and flutter at initiation of study medications during analysis to explore the possible residual confounding the June 1, 2012, to May 31, 2015, enrollment period effect (Supplementary Table S1). were identified. The date of the first prescription of dabi- gatran, rivaroxaban, or warfarin was operationally defined Exposures and follow-up as the index date. Subjects with unknown sex; a diagnosis of deep vein thrombosis, pulmonary embolism, mitral ste- According to our analysis, individuals receiving rivaroxa- nosis, or renal failure; or having any procedure including ban 15 mg or 10 mg had comparable risks of various clin- valvular replacement, mitral commissurotomy, heart trans- ical outcomes (Supplementary Table S2, Supplementary plantation, or extracorporeal circulatory support within 6 Figure S1), so rivaroxaban 15 and 10 mg were pooled into months before the index date were excluded (Supplemen- one group in this study. Thus, we defined study groups as tary Table S1). We also excluded individuals receiving two reduced-dose dabigatran (110 mg), reduced-dose rivaroxa- different kinds of study medications on the index date, ban (15 or 10 mg), and warfarin according to initial pre- those receiving concomitant antiplatelet therapy, those scription of study medications. All clinical outcomes were whose DOAC dosage could not be clarified, and those evaluated from inpatient records in the NHI claims data- who had been exposed to warfarin between January 1, base. All individuals were followed from their index date 2011, and May 31, 2012. Because less than 4% (171/ until death, a switch to another oral anticoagulant, discon- 4893) of our study population received a standard dose of tinuation of study medication (30-day treatment gap), or dabigatran (150 mg) or rivaroxaban (20 mg), only those end of follow-up at December 31, 2015, whichever came receiving reduced-dose dabigatran (110 mg), reduced-dose first. rivaroxaban (15 or 10 mg), or warfarin were retained for analysis (Figure 1). Baseline characteristics Age was ascertained at index date. The Taiwan NHI pre- Clinical outcomes mium for each subject was used as a proxy of socioeco- Clinical outcomes included all-cause death, cardiovascular nomic status, and quartiles of the insurance premium in death, ischemic stroke, acute myocardial infarction, arte- the overall study population were used as cut-offs for cate- rial embolism or thrombosis, intracranial hemorrhage, and gorization. We assessed comorbidities as appearance of JAGS AUGUST 2018 2018–VOL. 66, NO. 8 DABIGATRAN,DABIGATRAN,RIVAROXABAN, RIVAROXABAN,AND ANDWARFARIN WARFARIN1569 3 specific diagnosis codes twice in outpatient records or warfarin group to explore the effectiveness and safety of once in inpatient records during the 6-month period before label-adherent dosing in rivaroxaban users. the index date and coded as binary variables. Comorbid- All analysis was performed using SAS version 9.4 ities were evaluated using the Elixhauser Comorbidity (SAS Institute, Inc., Cary, NC). Index,16 except for