Standard and reduced doses of dabigatran, rivaroxaban and apixaban for stroke prevention in atrial fibrillation: a nationwide cohort study

Transcription

1 Original Article doi: /joim Standard and reduced doses of dabigatran, rivaroxaban and apixaban for stroke prevention in atrial fibrillation: a nationwide cohort study L. Staerk 1, T. A. Gerds 2,G.Y.H.Lip 3,4, B. Ozenne 2, A. N. Bonde 1, M. Lamberts 1, E. L. Fosbøl 5, C. Torp-Pedersen 6,7, G. H. Gislason 1,8,9,10 & J. B. Olesen 1 From the 1 Department of Cardiology, Copenhagen University Hospital Herlev and Gentofte, Hellerup, Denmark; 2 Department of Public Health, Section of Biostatistics, University of Copenhagen, Copenhagen K, Denmark; 3 Institute of Cardiovascular Sciences, University of Birmingham, Birmingham, UK; 4 Aalborg Thrombosis Research Unit, Department of Clinical Medicine, Aalborg University, Aalborg, Denmark; 5 Department of Cardiology, Copenhagen University Hospital Rigshospitalet, Copenhagen Ø, Denmark; 6 Department of Health, Science and Technology, Aalborg University, Aalborg, Denmark; 7 Department of Cardiology and Epidemiology/Biostatistics, Aalborg University Hospital, Aalborg, Denmark; 8 The Danish Heart Foundation, Copenhagen, Denmark; 9 Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen N, Denmark; and 10 The National Institute of Public Health, University of Southern Denmark, Copenhagen K, Denmark Abstract. Staerk L, Gerds TA, Lip GYH, Ozenne B, Bonde AN, Lamberts M, Fosbøl EL, Torp-Pedersen C, Gislason GH, Olesen JB (Copenhagen University Hospital Herlev and Gentofte, Hellerup; University of Copenhagen, Copenhagen K, Denmark; University of Birmingham, Birmingham, UK; Aalborg University, Aalborg; Copenhagen University Hospital Rigshospitalet, Copenhagen Ø; Aalborg University Hospital, Aalborg, Denmark; The Danish Heart Foundation, Copenhagen; University of Copenhagen, Copenhagen N; University of Southern Denmark, Copenhagen K, Denmark). Standard and reduced doses of dabigatran, rivaroxaban and apixaban for stroke prevention in atrial fibrillation: a nationwide cohort study. JInternMed2018; 283: Background. Comparative data of non-vitamin K antagonist oral anticoagulants (NOAC) are lacking in patients with atrial fibrillation (AF). Objective. We compared effectiveness and safety of standard and reduced dose NOAC in AF patients. Methods. Using Danish nationwide registries, we included all oral anticoagulant-na ıve AF patients who initiated NOAC treatment ( ). Outcome-specific and mortality-specific multiple Cox regressions were combined to compute average treatment effects as 1-year standardized differences in stroke and bleeding risks (g-formula). Results. Amongst AF patients, the distribution of NOAC/dose was as follows: dabigatran standard dose (22.4%), dabigatran-reduced dose (14.0%), rivaroxaban standard dose (21.8%), rivaroxaban reduced dose (6.7%), apixaban standard dose (22.9%), and apixaban reduced dose (12.2%). The 1-year standardized absolute risks of stroke/thromboembolism were % and % with standard and reduced NOAC dose, respectively, without statistically significant differences between NOACs for given dose level. Comparing standard doses, the 1-year standardized absolute risk (95% CI) for major bleeding was for rivaroxaban 2.78% ( %); corresponding absolute risk differences (95% CI) were for dabigatran 0.93% ( 1.45% to 0.38%) and apixaban, 0.54% ( 0.99% to 0.05%). The results for major bleeding were similar for reduced NOAC dose. The 1-year standardized absolute risk (95% CI) for intracranial bleeding was for standard dose dabigatran 0.19% ( %); corresponding absolute risk differences (95% CI) were for rivaroxaban 0.23% ( %) and apixaban, 0.18% ( %). Conclusions. Standard and reduced dose NOACs, respectively, showed no significant risk difference for associated stroke/thromboembolism. Rivaroxaban was associated with higher bleeding risk compared with dabigatran and apixaban and dabigatran was associated with lower intracranial bleeding risk compared with rivaroxaban and apixaban. Keywords: apixaban, atrial fibrillation, dabigatran, dose, NOAC, rivaroxaban. ª 2017 The Association for the Publication of the Journal of Internal Medicine 45

2 Introduction Patients with non-valvular atrial fibrillation (AF) at increased risk of stroke are recommended oral anticoagulant (OAC) treatment as stroke prophylaxis [1]. For decades, the only OAC option was vitamin K antagonist (VKA), for example warfarin. As introduction of non-vka oral anticoagulants (NOACs), the initiation rates and the utilization patterns have changed, [2 5] and in 2015, AF patients were more frequently initiated on NOACs than VKA treatment for stroke prophylaxis in Denmark [2]. Current AF guidelines also advocate NOACs over VKA [1] due to lower intracranial bleeding rates, no need for continuous anticoagulation monitoring and dose adjustment, and fewer drug-drug interactions, when compared with VKA [6 10]. Most prior studies have primarily focused on the comparison of NOACs versus VKA [6 13]. The diversely designed Phase III randomized controlled trials (RCTs) make it difficult to directly compare efficacy and safety of the NOACs, and there are no head-to-head trials comparing individual NOAC against each other [6 9]. Moreover, in the Phase III RCTs for rivaroxaban and apixaban, there was no randomized arm of high or low dose NOAC, [7, 8] in contrast to the dabigatran and edoxaban Phase III RCTs [6, 9]. Nonetheless, AF guidelines recommend dose reductions of NOAC in selected patients, which is not based on solid evidence from Phase III RCTs [1]. Notably, there is a gap in knowledge about the efficacy and safety of reduced dose NOAC despite the disproportionately high usage of reduced dose NOAC, which is a cause of concern [14]. Therefore, studies examining the effectiveness and safety of standard dose or reduced dose NOAC in direct comparisons are warranted. Our study aimed to compare first standard dose dabigatran, rivaroxaban and apixaban, and secondly, reduced dose dabigatran, rivaroxaban and apixaban, and the associated risks of thrombotic and bleeding outcomes. Materials and methods Data source All citizens in Denmark are allocated a unique personal identification number, which we used to cross-link data on individual level from the following administrative registries: (i) The Danish National Patient Register holds information on all hospital contacts. Each hospital contact is coded with diagnosis codes according to the International Classification of Disease System 10th revision (ICD-10); [15] (ii) The Danish Register of Medicinal Product Statistics keeps track of all filled prescriptions including expedition date, package size and dispensed dosage; [16] and (iii) The Danish Civil Registration System administrates data on vital status [17]. Study population In Denmark, dabigatran was approved 22 August 2011, rivaroxaban on 6 February 2012 and apixaban on 10 December Patients were included if they had an AF diagnosis and subsequently filled a first-time prescription of OAC between 1 March 2012 and 31 December Inclusion day was the day a first-time prescription of OAC was filled. The AF diagnosis code has been validated in The Danish National Patient Register with a high positive predicted value [18]. Exclusion criteria are displayed in Fig. 1. Comorbidities and concomitant pharmacotherapy We identified comorbidities from ICD-10 codes recorded within 10 years prior to inclusion day and concomitant pharmacotherapy from prescriptions filled within 180 days prior to inclusion day (ICD-10 and ATC codes are listed in Table S1), as described previously [2, 10]. At time of inclusion in the study, stroke and bleeding risk stratification scores were calculated according to the CHA 2 DS 2 -VASc and HAS-BLED scores, respectively (Table S2). Outcomes Outcomes of interest were stroke/thromboembolism (TE), ischaemic stroke, major bleeding, intracranial bleeding and gastrointestinal bleeding (ICD-10 codes are listed in Table S1). Validation studies have shown high positive predictive values for stroke and bleeding [19, 20]. The 2-year followup period began at day of inclusion and patients were followed until the occurrence of outcome, death, emigration, shift or discontinuation of initiated treatment, or 31 December 2016, whichever came first. Shift was defined as if a patient shifted treatment from the initiated NOAC agent to another 46 ª 2017 The Association for the Publication of the Journal of Internal Medicine

3 Fig. 1 Selection of the study population from 1 March 2012 to 31 December NOAC or VKA. Discontinuation was defined as if a patient did not fill a consecutive prescription within 30 days after the expiration of a dispensed prescription [10, 21]. Statistical analysis Using the Aalen Johansen estimator, the dynamics of the outcome-specific absolute risks (cumulative incidence) were calculated. Outcome-specific Cox models were combined with Cox models for mortality, discontinuation and shift of treatment to predict the absolute 1-year risks of outcome [22]. All Cox models were adjusted for calendar year and stratified for age (<65, 65 74, and 85 years) and sex. For the stroke/te and ischaemic stroke outcomes, the models were adjusted for heart failure, hypertension, diabetes mellitus, prior stroke/te and vascular disease. For the bleeding outcomes, the models were additionally adjusted for abnormal renal or liver function, prior bleeding, antiplatelet drugs, non-steroidal anti-inflammatory drugs and alcohol abuse. All analyses were performed separately for the subset of patients who initiated standard and reduced dose NOAC, respectively. Results were reported as standardized differences in 1-year absolute risks of outcome between NOACs (average treatment effects) based on g-formula together with 95% bootstrap confidence limits (1000 bootstrap data sets) [22, 23]. Within the limitations of the available data, the average treatment effects can be interpreted as what one would have observed had the patients been randomized to dabigatran, rivaroxaban or apixaban [24]. The level of significance was set at 5%. Statistical analyses were performed in SAS (version 9.4, SAS Institute Inc. Gary, NC) and R [25]. Retrospective registry-based studies do not require approval from the Research Ethics Committee System. The Danish Data Protection Agency had approved use of data for this study (ref.no: /GEH I-Suite no: 02720). Results Study population The final study population (Fig. 1) included AF patients initiating dabigatran standard dose (150 mg, n = 7078), dabigatran reduced dose (110 mg, n = 4414), rivaroxaban standard dose ª 2017 The Association for the Publication of the Journal of Internal Medicine 47

4 (20 mg, n = 6868), rivaroxaban reduced dose (15 mg, n = 2098), apixaban standard dose (5 mg, n = 7203) and apixaban reduced dose (2.5 mg, n = 3861). Baseline characteristics are shown in Table 1. The median age was higher in patients treated with standard dose rivaroxaban (71 years) and apixaban (71 years) than standard dose dabigatran (67 years). Moreover, the prevalence of chronic kidney disease and usage of antiplatelets was more frequent in patients treated with standard dose rivaroxaban and apixaban compared with standard dose dabigatran. A similar difference in patient characteristics was observed between patients treated with reduced dose rivaroxaban and apixaban versus reduced dose dabigatran. Absolute risks The unadjusted absolute risks including numbers at risks are illustrated in Fig. S1. The standardized absolute risks of stroke/te, major bleeding, intracranial bleeding and gastrointestinal bleeding for patients treated with standard and reduced dose are illustrated in Figs 2, 3, S2 and S3, respectively. Patients receiving standard dose of NOAC were associated with a lower standardized absolute risk of all outcomes compared with reduced dose of NOAC. Overall, the standardized absolute risk of major bleeding was highest for rivaroxaban. Comparisons of standard dose NOAC For dabigatran, rivaroxaban and apixaban, the 1- year standardized absolute risks of stroke/te were 1.73%, 1.73% and 1.98%, respectively, and of ischaemic stroke 1.02%, 0.91% and 0.99%, respectively (Table 2). The absolute risk differences for stroke/te and ischaemic stroke did not differ significantly when comparing NOACs pairwise. For the major bleeding outcome, the 1-year standardized absolute risk was for rivaroxaban 2.78% (95% CI, 2.42% to 3.17%); the corresponding absolute risk differences were significantly lower for dabigatran 0.23% (95% CI, 0.41% to 0.06%) and for apixaban, 0.18% (95% CI, 0.34% to 0.01%). For intracranial bleeding, the 1-year standardized absolute risk was for dabigatran 0.19% (95% CI, 0.10 to 0.32%); the corresponding absolute risk differences were significantly increased for rivaroxaban with 0.23% (95% CI, 0.06% to 0.41%), and for apixaban with 0.18% (95% CI, 0.01% to 0.34%), but there was no significant difference in the absolute risks between rivaroxaban and apixaban. Comparisons of reduced dose NOAC The 1-year standardized absolute risks for reduced dose dabigatran, rivaroxaban and apixaban were for stroke/te 2.51%, 2.70% and 2.78%, respectively, and for ischaemic stroke 1.28%, 1.52% and 1.29%, respectively (Table 3). Again, the absolute risk differences were not significant in pairwise comparisons between NOACs. The 1-year standardized absolute risks of major bleeding varied between NOACs, and rivaroxaban was associated with a significantly higher standardized absolute risk of major bleeding compared with dabigatran and apixaban. The risk of gastrointestinal bleeding was significantly lower for apixaban, and the 1-year standardized absolute risk was 1.50% (95% CI, 1.12% to 1.94%); the corresponding absolute risk differences were for dabigatran 0.68% (95% CI, 0.02% to 1.35%), and for rivaroxaban 0.87% (95% CI, 0.15% to 1.58%). Discussion In our nationwide cohort study of OACna ıve AF patients initiating NOAC treatment, the associated stroke risks were similar in comparison with standard dose NOACs or in comparison with reduced dose NOACs. The bleeding risks varied between dabigatran, rivaroxaban and apixaban treatment. Rivaroxaban treatment was associated with an increased risk of major bleeding compared with the other NOACs. In a comparison of standard dose NOAC, dabigatran was associated with significantly lower risk of intracranial bleeding, and in a comparison of reduced dose NOAC, apixaban was associated with significantly lower risk of gastrointestinal bleeding. Effectiveness between NOACs We estimated the average treatment effects to imitate an RCT comparing standard dose NOACs or reduced dose NOACs. Other comparative observational studies, including a recently published study from our group, have compared NOACs versus VKA, and the studies found no difference in the association between stroke risk and 48 ª 2017 The Association for the Publication of the Journal of Internal Medicine

5 Table 1 Patient characteristics at baseline Dabigatran Rivaroxaban Apixaban Dose Standard Reduced Standard Reduced Standard Reduced No. (%) 7078 (22.4) 4414 (14.0) 6868 (21.8) 2098 (6.7) 7203 (22.9) 3861 (12.2) Males, No. (%) 4507 (63.7) 1914 (43.4) 3807 (55.4) 896 (42.7) 4193 (58.2) 1400 (36.3) Age, median [Q1, Q3] 67 [61, 71] 81 [76, 85] 71 [65, 78] 83 [76, 88] 71 [65, 77] 84 [80, 89] Age groups, No. (%) < (38.1) 219 (5.0) 1567 (22.8) 108 (5.1) 1666 (23.1) 106 (2.7) 65 to (49.5) 641 (14.5) 2750 (40.0) 332 (15.8) 3156 (43.8) 376 (9.7) 75 to (12.2) 2269 (51.4) 1898 (27.6) 722 (34.4) 1941 (26.9) 1453 (37.6) (0.2) 1285 (29.1) 653 (9.5) 936 (44.6) 440 (6.1) 1926 (49.9) CHA 2 DS 2 -VASc, 2 [1, 3] 4 [3, 5] 3 [2, 4] 4 [3, 5] 3 [2, 4] 4 [3, 5] median [Q1, Q3] HAS-BLED, median [Q1, 2 [1, 2] 2 [2, 3] 2 [1, 3] 3 [2, 3] 2 [1, 3] 3 [2, 3] Q3] Comorbidities, No. (%) Prior stroke/te 833 (11.8) 848 (19.2) 1025 (14.9) 414 (19.7) 1254 (17.4) 906 (23.5) Myocardial infarction 381 (5.4) 415 (9.4) 360 (5.2) 232 (11.1) 427 (5.9) 403 (10.4) Ischaemic heart disease 1158 (16.4) 1088 (24.6) 1135 (16.5) 598 (28.5) 1306 (18.1) 1003 (26.0) Peripheral artery disease 121 (1.7) 148 (3.4) 156 (2.3) 94 (4.5) 164 (2.3) 179 (4.6) Heart failure 843 (11.9) 784 (17.8) 821 (12.0) 532 (25.4) 978 (13.6) 814 (21.1) Diabetes mellitus 751 (10.6) 512 (11.6) 774 (11.3) 295 (14.1) 966 (13.4) 475 (12.3) Hypertension 2939 (41.5) 2411 (54.6) 2989 (43.5) 1246 (59.4) 3160 (43.9) 2146 (55.6) Chronic kidney disease 94 (1.3) 123 (2.8) 105 (1.5) 184 (8.8) 182 (2.5) 320 (8.3) Abnormal liver function 77 (1.1) 41 (0.9) 74 (1.1) 31 (1.5) 108 (1.5) 65 (1.7) Prior bleeding 583 (8.2) 617 (14.0) 656 (9.6) 327 (15.6) 807 (11.2) 647 (16.8) Alcohol abuse 249 (3.5) 121 (2.7) 224 (3.3) 69 (3.3) 277 (3.8) 107 (2.8) Cancer 707 (10.0) 707 (16.0) 917 (13.4) 347 (16.5) 985 (13.7) 691 (17.9) Concomitant medication, No. (%) ADP receptor antagonists 460 (6.5) 523 (11.8) 657 (9.6) 306 (14.6) 731 (10.1) 604 (15.6) Aspirin 2180 (30.8) 1897 (43.0) 2149 (31.3) 941 (44.9) 2183 (30.3) 1539 (39.9) Dipyridamole 113 (1.6) 145 (3.3) 136 (2.0) 71 (3.4) 142 (2.0) 129 (3.3) Non-steroidal antiinflammatory 1130 (16.0) 568 (12.9) 996 (14.5) 283 (13.5) 1074 (14.9) 441 (11.4) drugs Loop diuretics 677 (9.6) 963 (21.8) 805 (11.7) 649 (30.9) 872 (12.1) 1135 (29.4) Beta-blockers 2518 (35.6) 1775 (40.2) 2385 (34.7) 862 (41.1) 2429 (33.7) 1432 (37.1) Calcium channel 1632 (23.1) 1315 (29.8) 1693 (24.7) 601 (28.6) 1744 (24.2) 1136 (29.4) blockers Renin-angiotensin 2791 (39.4) 1946 (44.1) 2705 (39.4) 984 (46.9) 3020 (41.9) 1678 (43.5) system inhibitors Digoxin 268 (3.8) 410 (9.3) 309 (4.5) 205 (9.8) 269 (3.7) 320 (8.3) dabigatran, rivaroxaban or apixaban compared with VKA [10 12, 26]. In the present study, we extend our prior observations [10] by showing that there was no absolute risk difference in the associated stroke risk between standard dose of NOAC or between reduced dose of NOAC. ª 2017 The Association for the Publication of the Journal of Internal Medicine 49

6 Fig. 2 Standardized absolute risks of stroke/thromboembolism and major bleeding for standard dose dabigatran, rivaroxaban and apixaban. Stroke/thromboembolism Major bleeding Standardized absolute risk (%) Reduced dose Dabigatran Rivaroxaban Apixaban Standardized absolute risk (%) Reduced dose Dabigatran Rivaroxaban Apixaban Time since oral anticoagulation was initiated (months) Time since oral anticoagulation was initiated (months) Fig. 3 Standardized absolute risks of stroke/thromboembolism and major bleeding for reduced dose dabigatran, rivaroxaban and apixaban. Graham et al. [27] and Hernandez et al. [28] both used Medicare data (U.S.) when comparing standard and reduced dose of rivaroxaban (20 mg/15 mg) versus dabigatran (150 mg/75 mg). Similar to our results, the two American studies found no significant difference in associated stroke risk between rivaroxaban and dabigatran. Contrary, a Chinese study estimated a lower associated risk of ischaemic stroke for dabigatran 110 mg compared with rivaroxaban 15 mg and 20 mg [29]. The American and Chinese studies were restricted to only compare rivaroxaban versus dabigatran, 50 ª 2017 The Association for the Publication of the Journal of Internal Medicine

7 Table 2 Standard dose At 1 year Standardized absolute Absolute risk Absolute risk No. of event risk (95% CI) difference (95% CI) a difference (95% CI) a Stroke/thromboembolism Dabigatran % (1.41% to 2.14%) Reference Rivaroxaban % (1.43% to 2.03%) 0.00% ( 0.50% to 0.44%) Reference Apixaban % (1.67% to 2.31%) 0.25% ( 0.33% to 0.72%) 0.24% ( 0.20% to 0.68%) Ischaemic stroke Dabigatran % (0.77% to 1.35%) Reference Rivaroxaban % (0.70% to 1.12%) 0.12% ( 0.51% to 0.22%) Reference Apixaban % (0.77% to 1.24%) 0.03% ( 0.47% to 0.35%) 0.09% ( 0.22% to 0.39%) Major bleeding Dabigatran % (1.53% to 2.23%) Reference Rivaroxaban % (2.42% to 3.17%) 0.93% (0.38% to 1.45%) a Reference Apixaban % (1.91% to 2.60%) 0.40% ( 0.09% to 0.85%) 0.54% ( 0.99% to 0.05%) a Intracranial bleeding Dabigatran % (0.10% to 0.32%) Reference Rivaroxaban % (0.29% to 0.59%) 0.23% (0.06% to 0.41%) a Reference Apixaban % (0.25% to 0.51%) 0.18% (0.01% to 0.34%) a 0.05% ( 0.24% to 0.12%) Gastrointestinal bleeding Dabigatran % (0.79% to 1.34%) Reference Rivaroxaban % (0.95% to 1.43%) 0.15% ( 0.24% to 0.51%) Reference Apixaban % (0.78% to 1.23%) 0.05% ( 0.42% to 0.29%) 0.20% ( 0.50% to 0.10%) a The corresponding absolute risk differences test for significant differences between standardized absolute risks. Statistical significance is obtained when the 95% confidence limits for difference in absolute risks does not contain zero. [27 29] whereas a recently published American study by Noseworthy et al. compared dabigatran, rivaroxaban and apixaban in three one-to-one propensity-score-matched cohorts. The study found no difference in associated risk of stroke in the pairwise comparisons [30]. A major limitation in the study by Noseworthy et al. was the inability to compare across dabigatran, rivaroxaban and apixaban in only one analysis, as they used three one-to-one propensity-score-matched cohorts. Moreover, Noseworthy et al. did not separately compare standard or reduced dose NOAC. Major bleeding risk When studies point in a direction of no difference in effectiveness between the different NOACs, the succeeding clinical question is how to choose the safest NOAC treatment. In our study, rivaroxaban was associated with a higher risk of major bleeding compared with dabigatran or apixaban. The previously discussed American studies and a recently published meta-analysis also found an increased risk of bleeding associated with rivaroxaban compared with dabigatran [27, 28, 30, 31]. Two Danish studies by Larsen et al. [11] and Nielsen et al. [12] used VKA as reference and reported a significantly lower major bleeding risk associated with standard dose dabigatran and apixaban [11] and reduced dose dabigatran, respectively [12]. The two Danish studies did not compare across NOACs. Moreover, baseline covariates of patients receiving VKA influenced the inverse probability of treatmentweighted analysis. Other differences were that in our study all patients were first-time initiators of OAC, whereas in the studies by Larsen et al. [11] and Nielsen et al. [12] the inclusion criterion was those free of OAC treatment 1 year before entering ª 2017 The Association for the Publication of the Journal of Internal Medicine 51

8 Table 3 Reduced dose At 1 year No. of event Standardized absolute risk (95% CI) Absolute risk difference (95% CI) a Absolute risk difference (95% CI) a Stroke/thromboembolism Dabigatran % (2.05% to 2.97%) Reference Rivaroxaban % (2.04% to 3.37%) 0.19% ( 0.60% to 0.98%) Reference Apixaban % (2.25% to 3.39%) 0.28% ( 0.43% to 1.10%) 0.08% ( 0.77% to 1.97%) Ischaemic stroke Dabigatran % (0.98% to 1.62%) Reference Rivaroxaban % (1.04% to 2.00%) 0.24% ( 0.28% to 0.81%) Reference Apixaban % (0.94% to 1.67%) 0.01% ( 0.52% to 0.53%) 0.23% ( 0.81% to 0.38%) Major bleeding Dabigatran % (3.31% to 4.54%) Reference Rivaroxaban % (4.13% to 5.84%) 1.08% (0.03% to 2.09%) a Reference Apixaban % (3.14% to 4.35%) 0.19% ( 1.04% to 0.72%) 1.27% ( 2.19% to 0.22%) a Intracranial bleeding Dabigatran % (0.30% to 0.71%) Reference Rivaroxaban % (0.51% to 1.24%) 0.39% ( 0.002% to 0.79%) Reference Apixaban % (0.47% to 1.01%) 0.26% ( 0.06% to 0.59%) 0.13% ( 0.55% to 0.28%) Gastrointestinal bleeding Dabigatran % (1.73% to 2.68%) Reference Rivaroxaban % (1.74% to 3.04%) 0.20% ( 0.55% to 0.96%) Reference Apixaban % (1.12% to 1.94%) 0.68% ( 1.35% to 0.02%) a 0.87% ( 1.58% to 0.15%) a a The corresponding absolute risk differences test for significant differences between standardized absolute risks. Statistical significance is obtained when the 95% confidence limits for difference in absolute risks does not contain zero. the study. Rivaroxaban is dosed once daily, whereas dabigatran and apixaban are dosed twice daily, and this could potentially contribute to the higher bleeding risk associated with rivaroxaban. Intracranial bleeding Meta-analyses based on the large Phase III RCTs have found lower risk of intracranial bleeding with NOACs and in conclusion the meta-analyses stated that the risk-benefit profile was more favourable with NOACs compared with VKA [13, 32]. Our group has previously published a study showing that dabigatran and apixaban were associated with lower risk of intracranial bleeding compared with VKA [10]. In a perspective of previously published studies [10, 27, 30] and the results from this study, dabigatran may have the lowest associated risk of intracranial bleeding compared with the other NOACs and VKA [6, 10, 26]. In Denmark, dabigatran was selected in less than 10% amongst all first-time users of OAC in December 2015 [2]. Gastrointestinal bleeding Intracranial bleedings more frequently causes death than gastrointestinal bleedings, but gastrointestinal bleedings are more frequent than intracranial bleedings amongst OAC-treated AF patients [6 8, 33]. In the RE-LY trial, [6] dabigatran 150 mg had an increased risk of gastrointestinal bleeding compared with warfarin, but other observational studies have not been able to find this association [34, 35]. In our study, the associated risk of gastrointestinal bleeding did not significantly differentiate between standard dose of NOACs, but reduced dose apixaban was associated with lower risk of gastrointestinal bleeding. Opposing results 52 ª 2017 The Association for the Publication of the Journal of Internal Medicine

9 were found in the American studies, which reported increased gastrointestinal bleeding risk with rivaroxaban versus dabigatran [27, 28]. The results in our study reflect all AF patients initiating NOAC treatment in a nationwide population, whereas the American studies were limited by including mostly patients >65 years. This may contribute to the differences in results as well as the fact that dabigatran 75 mg is recommended reduced dose in U.S., whereas in Denmark it is 110 mg [36]. Reduced dose In our study, the standardized absolute risks of stroke and bleeding were overall higher amongst patients initiating reduced dose NOAC than standard dose NOAC. This was also expected, because the Danish guidelines recommend dose reduction in selected patients, for example dabigatran 110 mg when age 80 years, concomitant treatment with verapamil, or high bleeding risk and low stroke risk; apixaban 2.5 mg when 2 criteria present of age 80 years, body weight 60 kg or serum creatinine 133 lmol L 1 [36]. We compared standard dose of NOAC and reduced dose of NOAC in two separate cohorts to avoid implanting a major confounding by indication bias; however, the differential thresholds for NOAC dose reduction could influence the risks of outcomes. We have previously shown that amongst those receiving reduced dose NOAC, 66.2% were 80 years (dabigatran 61.3%, rivaroxaban 64.3% and apixaban 77.1%) [10]. We do not have available data on body weight or blood tests in the Danish registries, and of that reason we could not examine if reduced dose was given according to guideline criteria. However, under- or overdosing of NOAC may have had an affect on our results, and a recent study showed that amongst NOAC-treated patients, 9.4% were underdosed, and 3.4% were overdosed [14]. Clinical practice Our study together with other comparative NOAC studies may contribute to guide AF patients and medical doctors to select between NOACs, for example choosing between dabigatran, rivaroxaban and apixaban in a case, where an AF patient has a high bleeding risk [27, 28, 30]. Moreover, our study reports standardized absolute risks of stroke and bleeding one year after initiation of NOAC treatment. Quantifying risk into an absolute risk instead of a relative risk might help AF patients and medical doctors in clinical practice to more intuitively understand the risk of stroke and bleeding when treated with NOACs. Limitations The major limitations of this study were due to the observational nature of the data. Residual confounding may have biased our results. We controlled for known confounders that were obtainable through the Danish registries, but we did not have data on potential confounders such as body weight, haemoglobin, international normalized ratio, serum creatinine and CrCl. The lack of data on kidney function may have influenced our results, as dabigatran is 80% renally excreted, thus, contraindicated for use when CrCL<30. Confounding by indication is a concern in pharmacoepidemiologic studies. We attempted to minimize this limitation using g-formula and narrowing the selection of the study population to a more homogenous study population by only including patients with an AF diagnosis code registered prior to OAC initiation, exclude VKA and only compare NOACs, and include only firsttime initiators of OAC treatment. Patients initiating edoxaban (available in Denmark 19 June 2016) were not included (n = 33). The number of events was small in some of the treatment groups, and reduced statistical power could have influenced our results in these analyses. The Danish population is predominantly white, and the generalization of the study findings was limited to other populations similar to the Danish population. Conclusions This study examined AF patients, who initiated OAC for the first time. In comparison with standard dose NOAC, and, secondly, in comparison with reduced dose NOAC, the associated ischaemic stroke risk did not differ significantly, respectively. Standard dose and reduced dose of rivaroxaban were associated with increased risk of major bleeding compared with standard dose and reduced dose of the other NOACs, respectively. Standard dose dabigatran had the lowest associated risk of intracranial bleeding compared with standard dose of the other NOACs. There is a ª 2017 The Association for the Publication of the Journal of Internal Medicine 53

10 growing need for an RCT comparing the effectiveness and safety of NOACs amongst AF patients. Funding This work was supported by the Velux Foundation. Dr. Gislason is supported by an unrestricted clinical research fellowship from the Novo Nordisk Foundation. Conflict of interest LS: Has received unrelated research grants from Boehringer Ingelheim; TAG: None; GYHL: Consultant for Bayer/Janssen, BMS/Pfizer, Biotronik, Medtronic, Boehringer Ingelheim, Microlife and Daiichi-Sankyo. Speaker for Bayer, BMS/Pfizer, Medtronic, Boehringer Ingelheim, Microlife, Roche and Daiichi-Sankyo; BO: None; ANB: None; ML: Has received speaker fees from Bayer and Bristol- Myers Squibb; ELF: Has received unrelated research support from Janssen and Janssen and the Lundbeck Foundation; CTP: Has received research support and speaker honoraria for Bayer and Biotronic; GG: Has received research grants from Bayer, Bristol-Myers Squibb, Boehringer Ingelheim and AstraZeneca; JBO: Has received speaker fees from Bristol-Myers Squibb, Boehringer Ingelheim, Bayer and AstraZeneca, previous funding for research from the Lundbeck Foundation, and current funding for research from the Bristol-Myers Squibb and The Capital Region of Denmark, Foundation for Health Research. Author contributions All authors have contributed significantly to the submitted manuscript. LS, TAG, GYHL, GG and JBO have designed the study, performed the analyses and interpreted the data and the results. LS has drafted the manuscript. All authors have contributed with interpretation of data, and revising the manuscript critically for essential intellectual content. All authors have read and approved the submitted manuscript. References 1 Kirchhof P, Benussi S, Kotecha D et al ESC Guidelines for the management of atrial fibrillation developed in collaboration with EACTS. Eur Heart J 2016; 37: Staerk L, Fosbøl EL, Gadsbøll K et al. Non-vitamin K antagonist oral anticoagulation usage according to age among patients with atrial fibrillation: Temporal trends in Denmark. Sci Rep 2016; 6: Gadsbøll K, Staerk L, Fosbøl EL et al. Increased use of oral anticoagulants in patients with atrial fibrillation: temporal trends from 2005 to 2015 in Denmark. Eur Heart J 2017; 38: Barnes GD, Lucas E, Alexander GC, Goldberger ZD. National trends in ambulatory oral anticoagulant use. Am J Med 2015; 128: e2. 5 Weitz JI, Semchuk W, Turpie AGG et al. Trends in prescribing oral anticoagulants in Canada, Clin Ther 2015; 37: e4. 6 Connolly SJ, Ezekowitz MD, Yusuf S et al. Dabigatran versus warfarin in patients with atrial fibrillation. N Engl J Med 2009; 361: Patel MR, Mahaffey KW, Garg J et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med 2011; 365: Granger CB, Alexander JH, McMurray JJV et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med 2011; 365: Giugliano RP, Ruff CT, Braunwald E et al. Edoxaban versus warfarin in patients with atrial fibrillation. N Engl J Med 2013; 369: Staerk L, Fosbøl EL, Lip GYH et al. Ischaemic and haemorrhagic stroke associated with non-vitamin K antagonist oral anticoagulants and warfarin use in patients with atrial fibrillation: a nationwide cohort study. Eur Heart J 2017; 38: Larsen TB, Skjøth F, Nielsen PB, Kjældgaard JN, Lip GYH. Comparative effectiveness and safety of non-vitamin K antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity weighted nationwide cohort study. BMJ 2016; 353: i Nielsen PB, Skjøth F, Søgaard M, Kjældgaard JN, Lip GYH, Larsen TB. Effectiveness and safety of reduced dose nonvitamin K antagonist oral anticoagulants and warfarin in patients with atrial fibrillation: propensity weighted nationwide cohort study. BMJ 2017; 356: j Ruff CT, Giugliano RP, Braunwald E et al. Comparison of the efficacy and safety of new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-analysis of randomised trials. Lancet 2014; 383: Steinberg BA, Shrader P, Thomas L et al. Off-label dosing of non-vitamin K antagonist oral anticoagulants and adverse outcomes: the ORBIT-AF II registry. J Am Coll Cardiol 2016; 68: Schmidt M, Schmidt SAJ, Sandegaard JL, Ehrenstein V, Pedersen L, Sørensen HT. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7: Kildemoes HW, Sørensen HT, Hallas J. The Danish national prescription registry. Scand J Public Health 2011; 39: Pedersen CB. The Danish civil registration system. Scand J Public Health 2011; 39: Frost L, Andersen LV, Vestergaard P, Husted S, Mortensen LS. Trend in mortality after stroke with atrial fibrillation. Am J Med 2007; 120: Krarup L-H, Boysen G, Janjua H, Prescott E, Truelsen T. Validity of stroke diagnoses in a national register of patients. Neuroepidemiology 2007; 28: ª 2017 The Association for the Publication of the Journal of Internal Medicine

11 20 Cunningham A, Stein CM, Chung CP, Daugherty JR, Smalley WE, Ray WA. An automated database case definition for serious bleeding related to oral anticoagulant use. Pharmacoepidemiol Drug Saf 2011; 20: Hellfritzsch M, Husted SE, Grove EL et al. Treatment changes among users of non-vitamin k antagonist oral anticoagulants in atrial fibrillation. Basic Clin Pharmacol Toxicol 2017; 120: Benichou J, Gail MH. Estimates of absolute cause-specific risk in cohort studies. Biometrics 1990; 46: Sato T, Matsuyama Y. Marginal structural models as a tool for standardization. Epidemiology 2003; 14: Robins JM, Hernan MA. Causal Inference. Boca Raton: Chapman & Hall/CRC, forthcoming Anon. R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria:R Foundation for Statistical Computing Lauffenburger JC, Farley JF, Gehi AK, Rhoney DH, Brookhart MA, Fang G. Effectiveness and safety of dabigatran and warfarin in real-world US patients with non-valvular atrial fibrillation: a retrospective cohort study. J Am Heart Assoc 2015; 4: e Graham DJ, Reichman ME, Wernecke M et al. Stroke, bleeding, and mortality risks in elderly medicare beneficiaries treated with dabigatran or rivaroxaban for nonvalvular atrial fibrillation. JAMA Intern Med 2016; 176: Hernandez I, Zhang Y. Comparing stroke and bleeding with rivaroxaban and dabigatran in atrial fibrillation: analysis of the US Medicare Part D data. Am J Cardiovasc Drugs Drugs Devices Interv 2017; 17: Li W-H, Huang D, Chiang C-E et al. Efficacy and safety of dabigatran, rivaroxaban, and warfarin for stroke prevention in Chinese patients with atrial fibrillation: the Hong Kong Atrial Fibrillation Project. Clin Cardiol 2017; 40: 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 2016; 150: Bai Y, Deng H, Shantsila A, Lip GYH. Rivaroxaban versus dabigatran or warfarin in real-world studies of stroke prevention in atrial fibrillation: systematic review and metaanalysis. Stroke 2017; 48: Chatterjee S, Sardar P, Biondi-Zoccai G, Kumbhani DJ. New oral anticoagulants and the risk of intracranial hemorrhage: traditional and Bayesian meta-analysis and mixed treatment comparison of randomized trials of new oral anticoagulants in atrial fibrillation. JAMA Neurol 2013; 70: Gomez-Outes A, Lagunar-Ruız J, Terleira-Fernandez A-I, Calvo-Rojas G, Suarez-Gea ML, Vargas-Castrillon E. Causes of death in anticoagulated patients with atrial fibrillation. J Am Coll Cardiol 2016; 68: Abraham NS, Singh S, Alexander GC et al. Comparative risk of gastrointestinal bleeding with dabigatran, rivaroxaban, and warfarin: population based cohort study. BMJ 2015; 350: h Staerk L, Gislason GH, Lip GYH et al. Risk of gastrointestinal adverse effects of dabigatran compared with warfarin among patients with atrial fibrillation: a nationwide cohort study. Europace 2015; 17: Brandes A, Jensen HK, Husted S. NBV 2016 Dansk Cardiologisk Selskab 15. Atrieflimren og atrieflagren Correspondence: Laila Staerk, MD, Department of Cardiology, Copenhagen University Hospital Herlev and Gentofte, Post 635, Kildegaardsvej 28, 2900 Hellerup, Denmark. (fax: ; Lailastaerk@gmail.com). Supporting Information Additional Supporting Information may be found in the online version of this article: Figure S1. Unadjusted absolute risks (cumulative incidence) of stroke/thromboembolism, major bleeding, intracranial bleeding, and gastrointestinal bleeding for standard and reduced dose NOAC. Figure S2. Standardized absolute risks of intracranial bleeding and gastrointestinal bleeding for standard dose dabigatran, rivaroxaban, and apixaban. Figure S3. Standardized absolute risks of intracranial bleeding and gastrointestinal bleeding for reduced dose dabigatran, rivaroxaban, and apixaban. Table S1. ICD-10 and ATC codes. Table S2. Calculation of CHA 2 DS 2 -VASc and HAS- BLED scores. ª 2017 The Association for the Publication of the Journal of Internal Medicine 55