Advertisement

Dabigatran and Warfarin for Secondary Prevention of Stroke in Atrial Fibrillation Patients: A Nationwide Cohort Study

Published:September 01, 2014DOI:https://doi.org/10.1016/j.amjmed.2014.07.023

      Abstract

      Background

      This register-based observational study compares dabigatran to warfarin for secondary stroke prevention in atrial fibrillation patients among both “new starters” on dabigatran and “switchers” to dabigatran from warfarin.

      Methods

      We identified, in nationwide Danish registries, 2398 patients with atrial fibrillation and a history of stroke/transient ischemic attack, making a first-time purchase of dabigatran 110 mg twice a day (bid; D110) and 150 mg bid (D150). Patients were categorized as either vitamin K antagonist (VKA) naive or experienced. Warfarin controls were identified using a complete (for VKA-naive dabigatran patients) or matched sampling approach (for VKA-experienced dabigatran patients). Subjects were followed for an average of 12.6 months for stroke and transient ischemic attacks. Confounder-adjusted Cox regression models were used to compare event rates between treatments.

      Results

      Among patients with a history of stroke/transient ischemic attack and prior VKA experience, switching to dabigatran was associated with an increased stroke/transient ischemic attack rate for both dabigatran doses compared with continuing on warfarin (D110 hazard ratio [HR] 1.99; 95% confidence interval [CI], 1.42-2.78; D150 HR 2.34; 95% CI, 1.60-3.41). Among prior stroke/transient ischemic attack patients who were new starters on dabigatran or warfarin, the rate of stroke/transient ischemic attack for both doses of dabigatran was similar to or lower than warfarin (D110 HR 0.64; 95% CI, 0.50-0.80; D150 HR 0.92l; 95% CI, 0.73-1.15).

      Conclusions

      In this register-based study, VKA-experienced patients with a history of stroke or transient ischemic attack who switched to dabigatran therapy had an increased rate of stroke compared with patients persisting with warfarin therapy.

      Keywords

      Clinical Significance
      • In patients with a history of stroke/transient ischemic attack who are VKA-naive, both dabigatran doses provided similar protection to warfarin against recurrent stroke/TIA.
      • Vitamin K antagonist-experienced patients with a history of stroke or transient ischemic attack who switch to dabigatran therapy may have an increased rate of a recurrent stroke compared to patients persisting with vitamin K antagonist therapy.
      • Caution is recommended when switching prior VKA-experienced atrial fibrillation patients to dabigatran.
      The risk of stroke in atrial fibrillation patients is high, especially among patients with a history of ischemic stroke or transient ischemic attack.
      • Albertsen I.E.
      • Rasmussen L.H.
      • Overvad T.F.
      • Graungaard T.
      • Larsen T.B.
      • Lip G.Y.H.
      Risk of stroke or systemic embolism in atrial fibrillation patients treated with warfarin: a systematic review and meta-analysis.
      Conventionally, stroke prevention in atrial fibrillation has involved anticoagulant therapy with vitamin K antagonists (VKA) such as warfarin,
      • Banerjee A.
      • Marín F.
      • Lip G.Y.H.
      A new landscape for stroke prevention in atrial fibrillation: focus on new anticoagulants, antiarrhythmic drugs, and devices.
      but a narrow therapeutic range makes optimal treatment challenging. Dabigatran etexilate is a direct thrombin inhibitor whose rapid and predictable response can mitigate the complexity of conventional anticoagulant therapy. Trials and subsequent observational studies in mixed patient populations have shown dabigatran to provide comparable or slightly improved protection against stroke and systemic embolism, without compromising bleeding safety.
      • Connolly S.J.
      • Ezekowitz M.D.
      • Yusuf S.
      • et al.
      Dabigatran versus warfarin in patients with atrial fibrillation.
      • Larsen T.B.
      • Rasmussen L.H.
      • Skjøth F.
      • et al.
      Efficacy and safety of dabigatran etexilate and warfarin in “real-world” patients with atrial fibrillation: a prospective nationwide cohort study.
      • Sørensen R.
      • Gislason G.
      • Torp-Pedersen C.
      • et al.
      Dabigatran use in Danish atrial fibrillation patients in 2011: a nationwide study.
      Trial-based findings within the clinically important subgroup of patients with prior stroke have been consistent with the overall efficacy and safety profile results for dabigatran. A predefined sub-study of the Randomized Evaluation of Long Term Anticoagulant Therapy (RE-LY) study found similar or slightly lower rates of stroke and systemic embolism among patients with a prior stroke/transient ischemic attack (2.78% for warfarin, 2.32% for dabigatran 110 mg twice a day [bid]; 2.07% for dabigatran 150 mg bid).
      • Diener H.-C.
      • Connolly S.J.
      • Ezekowitz M.D.
      • et al.
      Dabigatran compared with warfarin in patients with atrial fibrillation and previous transient ischaemic attack or stroke: a subgroup analysis of the RE-LY trial.
      In an indirect comparison of 3 new oral anticoagulants (apixaban, rivaroxaban, and dabigatran) using secondary prevention (previous stroke) trial subgroups,
      • Rasmussen L.H.
      • Larsen T.B.
      • Graungaard T.
      • Skjøth F.
      • Lip G.Y.H.
      Primary and secondary prevention with new oral anticoagulant drugs for stroke prevention in atrial fibrillation: indirect comparison analysis.
      we found that dabigatran had similar efficacy to warfarin for stroke/systemic embolism, ischemic stroke, and all-cause mortality. However, the extent to which these trial-based subgroup findings transfer to a practical clinical setting is currently unclear.
      Analysis of observational data, while insufficient for providing unequivocal treatment recommendations, can provide valuable insight into medication effectiveness in a practical clinical setting. The present nationwide observational study shares design and data similarities with recent studies on myocardial infarction and bleeding risk with dabigatran and warfarin, but in the present study we focus only on patients with previous stroke.
      • Larsen T.B.
      • Rasmussen L.H.
      • Gorst-Rasmussen A.
      • et al.
      Myocardial ischemic events in “real world” patients with atrial fibrillation treated with dabigatran or warfarin.
      • Larsen T.B.
      • Gorst-Rasmussen A.
      • Rasmussen L.H.
      • Skjøth F.
      • Rosenzweig M.
      • Lip G.Y.H.
      Bleeding events among new starters and switchers to dabigatran compared with warfarin in atrial fibrillation.
      Specifically, we used prescription purchase data to assess the effectiveness of dabigatran relative to warfarin for secondary prevention in a real-world atrial fibrillation population with a history of stroke or transient ischemic attack. A substantial proportion of dabigatran initiators are patients who are switched from warfarin – so-called warfarin survivors.
      • Lane D.A.
      • Lip G.Y.H.
      Dabigatran in atrial fibrillation: balancing secondary stroke prevention against bleeding risk.
      In RE-LY, no differences were found in the prognosis between such warfarin survivors and new starters.
      • Ezekowitz M.D.
      • Wallentin L.
      • Connolly S.J.
      • et al.
      Dabigatran and warfarin in vitamin K antagonist-naive and -experienced cohorts with atrial fibrillation.
      However, this finding may not transfer to a nontrial setting with more autonomous treatment management, particularly not in a high-risk patient group. Accordingly, we assessed the risk of stroke among patients with a history of stroke/transient ischemic attack separately within the stratum of “switchers” from warfarin to dabigatran (compared with warfarin persisters); and within the stratum of “new starters” on dabigatran (compared with new starters on warfarin).

      Methods

      We used the civil registration number assigned to all Danish residents to link 3 nationwide databases: 1) the Danish National Prescription Registry,
      • Kildemoes H.W.
      • Sørensen H.T.
      • Hallas J.
      The Danish National Prescription Registry.
      which holds information on purchase date, Anatomical Therapeutic Chemical (ATC) classification code, and package size for every prescription purchase in Denmark since 1994; 2) the Danish National Patient Register
      • Lynge E.
      • Sandegaard J.L.
      • Rebolj M.
      The Danish National Patient Register.
      established in 1977, which includes admission/discharge date and discharge International Classification of Diseases (ICD) diagnoses for >99% of somatic hospital admissions in Denmark; and 3) the Danish Civil Registration System,
      • Pedersen C.B.
      The Danish Civil Registration System.
      which holds information on sex, date of birth, vital and emigration status.

      Study Population

      We identified first-time purchases of dabigatran 110 mg bid (D110) or dabigatran 150 mg bid (D150) during the period August 1, 2011 (dabigatran market entry in Denmark) to May 30, 2013, alongside all purchases of warfarin during the period August 1, 2009 to May 30, 2013 (incorporating predabigatran warfarin purchases in order to determine VKA-experience status). We excluded purchases not preceded by a hospital diagnosis of atrial fibrillation, or preceded by a hospital diagnosis of mitral stenosis, venous thromboembolism, or valvular surgery, or preceded by phenprocoumon use. In accordance with the focus on secondary prevention, we excluded purchases not preceded by a hospital diagnosis of stroke/transient ischemic attack. A person was considered VKA-naive or VKA-experienced if the time since the last warfarin purchase was ≥2 years or <2 years, respectively. A relatively long period was used in order to also reflect naivety in relation to treatment routine.
      From the purchase data, we defined a VKA-naive stratum of all VKA-naive subjects making a first-time purchase of dabigatran. As controls in this stratum, we sampled the full population of VKA-naive subjects making a first-time warfarin purchase. The baseline date in the VKA-naive stratum was set to the date of first purchase.
      We next defined a VKA-experienced stratum of all VKA-experienced subjects purchasing dabigatran for the first time (switchers). Comparable controls were selected for each switcher by matched sampling among VKA-experienced warfarin controls. Specifically, 2 VKA-experienced warfarin users were matched to each switcher according to calendar month of purchase and duration of VKA experience (up to 1 year; 1-5 years; more than 5 years). The baseline date in the VKA-experienced stratum was set to the date of (first) purchase in the calendar month of inclusion.

      Endpoints and Variable Definitions

      Participants were followed until July 31, 2013 in the Danish National Patient Register (using the 10th Revision of ICD codes, ICD-10) for the occurrence of the following endpoints: ischemic stroke (I63, I64.9) or transient ischemic attack (G45); and fatal stroke, not including hemorrhagic stroke (ischemic stroke or transient ischemic attack followed by death within 30 days).
      Demographic data were obtained from the Danish Civil Registration System. Comorbidities and co-medications (listed in Table 1) were ascertained from the Danish National Patient Registry and the Danish National Prescription Registry (for ICD-10 and ATC code definitions, see Supplementary Table 1, available online). We combined covariate information into CHADS2 (Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, and previous stroke or TIA)/CHA2DS2VASc (Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category [female sex]) scores
      • Gooley T.A.
      • Leisenring W.
      • Crowley J.
      • Storer B.E.
      Estimation of failure probabilities in the presence of competing risks: new representations of old estimators.
      for assessing stroke risk, and a HAS-BLED (Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio [INR], Elderly, Drugs/Alcohol) bleeding risk score
      • Dentali F.
      • Riva N.
      • Crowther M.
      • Turpie A.G.G.
      • Lip G.Y.H.
      • Ageno W.
      Efficacy and safety of the novel oral anticoagulants in atrial fibrillation: a systematic review and meta-analysis of the literature.
      (definitions in Supplementary Table 2, available online).
      Table 1Participant Characteristics at Baseline, According to VKA-experience Stratum and Current Treatment
      Vitamin K Antagonist-naïveVitamin K Antagonist-experienced
      WarfarinDabigatran

      110 mg
      Dabigatran

      150 mg
      WarfarinDabigatran

      110 mg
      Dabigatran

      150 mg
      Subjects, no.18257936461918547412
      Age, years
       Median768369758270
       Interquartile range69-8278-8764-7469-8278-8665-74
      Age ≥65 years, %84.996.170.085.597.374.8
      Age ≥75 years, %52.982.818.651.384.820.9
      Female sex, %44.056.235.938.755.038.8
      Risk scores at baseline
       CHADS2, mean (SD)
      CHADS2: score ranging from 0-5, which reflects the stroke risk in atrial fibrillation patients not in anticoagulant therapy (see Supplementary Table 2).
      3.15 (0.90)3.41 (0.79)2.64 (0.74)3.19 (0.92)3.51 (0.79)2.91 (0.92)
       CHA2DS2VASc, mean (SD)
      CHA2DS2VASc: score ranging from 0-9, which reflects the risk of stroke in atrial fibrillation patients not in anticoagulant therapy (see Supplementary Table 2).
      4.76 (1.40)5.22 (1.14)3.87 (1.19)4.71 (1.33)5.35 (1.11)4.34 (1.31)
       HAS-BLED, mean (SD)
      HAS-BLED: score ranging from 0-9, which reflects the risk of bleeding in atrial fibrillation patients undergoing anticoagulant therapy (see Supplementary Table 2).
      3.11 (1.04)3.18 (0.89)2.72 (1.00)2.81 (0.96)2.99 (0.95)2.72 (0.99)
      Comorbidities at baseline
       Prior ischemic stroke75.381.274.975.782.176.5
       Prior transient ischemic attack36.332.035.837.232.434.7
       Prior bleeding, %16.220.913.019.224.519.7
       Diabetes, %16.115.413.018.014.120.6
       Hypertension, %36.433.029.637.736.738.1
       Abnormal renal function, %9.53.90.96.03.33.2
       Abnormal hepatic function, %0.40.80.00.10.50.0
       Prior congestive heart failure, %9.810.22.911.815.511.7
       Prior myocardial infarction, unstable angina, or cardiac arrest, %17.617.58.419.825.022.1
      Medications at baseline
       Aspirin, %43.042.734.823.025.621.8
       Clopidogrel, %21.420.120.33.06.45.8
       NSAID, %5.24.24.84.34.44.9
       Clopidogrel and aspirin/NSAID, %7.76.25.00.42.01.5
      Abbreviations: CHADS2 = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, and previous stroke or TIA; CHA2DS2VASc = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category (female sex); HAS-BLED = Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile INR, Elderly, Drugs/Alcohol; NSAID = nonsteroidal antiinflammatory drugs; SD = between-subjects standard deviation; VKA = vitamin K antagonist.
      CHADS2: score ranging from 0-5, which reflects the stroke risk in atrial fibrillation patients not in anticoagulant therapy (see Supplementary Table 2).
      CHA2DS2VASc: score ranging from 0-9, which reflects the risk of stroke in atrial fibrillation patients not in anticoagulant therapy (see Supplementary Table 2).
      HAS-BLED: score ranging from 0-9, which reflects the risk of bleeding in atrial fibrillation patients undergoing anticoagulant therapy (see Supplementary Table 2).

      Statistical Analysis

      Time-to-event analysis was used to compare the risk of stroke/transient ischemic attack between treatment groups within the 2 VKA-experience strata (naive/experienced), measuring risk time from baseline and until the relevant event, emigration, death, or July 31, 2013, whichever came first.
      Crude cumulative incidences of stroke/transient ischemic attack were calculated with the Aalen-Johansen method under competing risk of death.
      • Gooley T.A.
      • Leisenring W.
      • Crowley J.
      • Storer B.E.
      Estimation of failure probabilities in the presence of competing risks: new representations of old estimators.
      Cox regression was used to contrast event rates between the dabigatran users and warfarin controls within each of the VKA-experience strata. To address confounding by indication of treatment, regression analyses were adjusted for the baseline values of the following indications: age (continuous; cubic spline); components of CHA2DS2VASc and HAS-BLED (binary); and months since August 2011 (continuous; cubic spline). In the VKA-experienced stratum, we also adjusted for time since initiation of VKA therapy (continuous; cubic spline).
      We estimated 3-month persistence probabilities,
      • Gooley T.A.
      • Leisenring W.
      • Crowley J.
      • Storer B.E.
      Estimation of failure probabilities in the presence of competing risks: new representations of old estimators.
      defining time of nonpersistence as the time of treatment switching or >30 days discontinuation (ascertained from previous package sizes and a standard daily dose).
      A number of sensitivity/supplementary analyses were carried out. First, we repeated regression analyses after censoring individuals at the time of nonpersistence in order to quantify the effect of continuous treatment (implicitly assuming censoring to be noninformative conditionally on baseline covariates). Second, to assess robustness to a more stringent endpoint definition (with presumably higher validity), regression analyses were also repeated when requiring endpoints to have been registered as the primary diagnosis in connection with hospitalization for at least one night. Lastly, we repeated a subset of the main analyses in the primary prevention group, that is, the analogously defined 2 VKA-experience strata based on the subset of the warfarin/dabigatran purchase data that excluded subjects with a prior diagnosis of stroke/transient ischemic attack.
      Stata/MP version 12.1 was used for the statistical analysis (StataCorp LP, College Station, TX). A 2-sided P value < .05 was considered as statistically significant.

      Results

      Study Population Characteristics

      A flow chart of the study population is shown in Figure 1. In the VKA-naive stratum, we included 1439 patients with atrial fibrillation and a history of stroke/transient ischemic attack making a first-time dabigatran purchase, alongside 1825 patients making a first-time warfarin purchase (controls). In the VKA-experienced stratum, 959 dabigatran switchers were matched to 1918 warfarin controls (selected among 11,159 unique subjects with a total of 76,553 purchases).
      Figure thumbnail gr1
      Figure 1Flow chart showing how the study population was obtained from prescription purchase data. Boxes show number of purchases with number of unique subjects in parentheses. The final study population with prior stroke is displayed in dashed boxes. In the vitamin K antagonist (VKA)-experienced stratum, warfarin purchasers were sampled in a 2:1 ratio for each switcher to dabigatran (according to calendar month of purchase and duration of VKA-experience).
      Baseline information is shown in Table 1. Within each of the VKA-experience strata, D110 users were older, with a median age over 80 years (compared with median ages below 76 and 70 years for warfarin and D150 users, respectively). Accordingly, D110 users also had a higher stroke and bleeding risk (according to CHADS2/CHA2DS2VASc and HAS-BLED scores). There were slightly more females than males in the D110 group (55% to 56%) but not in the D150 and the warfarin group (36% to 44% females). Comparing between VKA-experience strata, the age and sex distribution was similar for each of the 3 treatments. Patients in the VKA-experienced stratum generally had higher CHADS2/CHA2DS2VASc scores but a similar or slightly decreased HAS-BLED score. There were substantially fewer clopidogrel users in the VKA-experienced stratum (3.0% to 5.8%) compared with the VKA-naive stratum (20.1% to 21.4%), and also fewer aspirin users (21.8% to 23.0%) than in the VKA-naive stratum (34.8% to 43.0%).

      Stroke and Transient Ischemic Attacks

      The average follow-up time was 12.6 months (SD 4.5 months). Plots of crude cumulative incidences for the composite endpoint of stroke/transient ischemic attack (Figure 2) showed that, in the VKA-naive stratum, D110 users had the lowest risk and D150 the highest risk, with the risk increasing more rapidly in the early follow-up period. The stroke risk in the VKA-experienced stratum was lower overall compared with the VKA-naive stratum: within this stratum, warfarin users had the lowest and D110 users the highest risk.
      Figure thumbnail gr2
      Figure 2Crude cumulative incidence of the composite endpoint of stroke or transient ischemic attack, according to Vitamin K antagonist-experience and current treatment.
      In the VKA-naive stratum (“new starters”), crude annual event rates of stroke/transient ischemic attack ranged from 14.0% to 20.0% (Table 2). For the composite endpoint stroke/TIA, adjusted event rates showed a significant 36% reduction among D110 users compared with warfarin (hazard ratio [HR] 0.64; 95% confidence interval [CI], 0.50-0.80), but similar rates for D150 users (HR 0.92; 95% CI, 0.73-1.15). When investigating stroke and transient ischemic attack separately, similar results were found, with slightly larger rate reductions relative to warfarin for the endpoint transient ischemic attack. Crude annual rates of fatal stroke/transient ischemic attack (not shown in Table 2) were generally low, ranging from 0.3% to 1.2%.
      Table 2Event Rates and Cox Hazard Ratios for the Various Endpoints, According to VKA-experience Stratum and Current Treatment
      No. EventsRate, % per YearCrude Hazard Ratio (95% CI)Adjusted
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      Hazard Ratio (95% CI)
      Vitamin K antagonist-naïve stratum
       Stroke or transient ischemic attack
      Warfarin30018.01 (reference)1 (reference)
      Dabigatran 110 mg9314.00.74 (0.59-0.93)0.67 (0.52-0.86)
      Dabigatran 150 mg10218.20.99 (0.78-1.24)1.02 (0.80-1.30)
       Stroke
      Warfarin23113.41 (reference)1 (reference)
      Dabigatran 110 mg8212.20.85 (0.67-1.10)0.74 (0.56-0.97)
      Dabigatran 150 mg8214.11.02 (0.79-1.31)1.10 (0.83-1.44)
       Transient ischemic attack
      Warfarin844.61 (reference)1 (reference)
      Dabigatran 110 mg172.40.49 (0.29-0.83)0.48 (0.28-0.83)
      Dabigatran 150 mg254.00.85 (0.55-1.34)0.79 (0.49-1.28)
      Vitamin K antagonist-experienced stratum
       Stroke or transient ischemic attack
      Warfarin1335.91 (reference)1 (reference)
      Dabigatran 110 mg5910.81.77 (1.30-2.40)1.54 (1.11-2.13)
      Dabigatran 150 mg459.21.57 (1.12-2.19)1.79 (1.25-2.56)
       Stroke
      Warfarin1024.51 (reference)1 (reference)
      Dabigatran 110 mg529.42.03 (1.45-2.84)1.73 (1.21-2.47)
      Dabigatran 150 mg336.61.49 (1.00-2.20)1.79 (1.18-2.72)
       Transient ischemic attack
      Warfarin391.71 (reference)1 (reference)
      Dabigatran 110 mg142.41.42 (0.77-2.62)1.30 (0.68-2.5+)
      Dabigatran 150 mg152.91.73 (0.95-3.14)1.72 (0.92-3.22)
      Abbreviations: CHA2DS2VASc = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category (female sex); CI = confidence interval; HAS-BLED = Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio (INR), Elderly, Drugs/Alcohol; VKA = vitamin K antagonist.
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      Turning to the VKA-experienced stratum (“switchers”), crude annual event rates of stroke/transient ischemic attack were overall substantially lower than in the VKA-naive stratum, ranging from 4.8% to 10.8% (Table 2). For stroke/TIA, we saw a significant doubling of the event rate for both D110 and D150, compared with warfarin (D110 HR 1.99; 95% CI, 1.42-2.78; D150 HR 2.34; 95% CI, 1.60-3.41). When considering stroke and transient ischemic attack separately, there were comparably sized increases in adjusted rates for dabigatran relative to warfarin. There were again few fatal strokes/transient ischemic attacks (not shown in Table 2), with crude annual rates ranging from 0.2% for D150 (0.3% for warfarin) to 2.3% for D110.

      Treatment Persistence and Supplementary Analyses

      Persistence probabilities at 3 months were 87% (warfarin), 79% (D110), and 82% (D150) in the VKA-naive stratum, and 91% (warfarin), 81% (D110), and 84% (D150) in the VKA-experienced stratum.
      The persistence-adjusted Cox regression gave similar results as the main analysis (Supplementary Table 3, available online); the increased rate of stroke/transient ischemic attack with dabigatran relative to warfarin in the VKA-experienced stratum was slightly more pronounced in this analysis (D110 HR 2.33; 95% CI, 1.53-3.55; D150 HR 2.48; 95% CI, 1.53-4.02). Upon censoring patients at time of nonpersistence, crude event rates in the VKA-naive stratum were increased compared with the main analysis. The results of the second sensitivity analysis, featuring a more restrictive endpoint definition (Supplementary Table 4, available online), were also consistent with the main analysis, although events rates were low.
      Repeating regression analyses in the primary prevention group (Supplementary Table 5, available online) led to similar rates of stroke/transient ischemic attack across strata in the adjusted comparison of D150 versus warfarin (VKA-naive HR 0.98; 95% CI, 0.72-1; VKA-experienced HR 0.83; 95% CI, 0.51-1.35). In contrast, the rate for D110 was increased compared with warfarin in the VKA-experienced stratum, but not in the VKA-naive (HR 1.86; 95% CI, 1.30-2.67 and HR 1.01; 95% CI, 0.74-1.37, respectively). Of note, the primary prevention group was substantially younger than the secondary prevention group; and with lower stroke and bleeding risks, according to the CHADS2/CHA2DS2VASc and HAS-BLED scores, respectively (Supplementary Table 5, available online).

      Discussion

      In this large register-based observational study of secondary stroke prevention among atrial fibrillation patients, we found similar effectiveness of dabigatran relative to warfarin for secondary prevention of stroke/transient ischemic attack among “new starters” on anticoagulant therapy. In contrast, we found a doubling of the stroke/transient ischemic attack rate among “switchers” to both dabigatran doses compared with persisters on warfarin. The overall stroke risk was generally higher in “new starters” on anticoagulant therapy regardless of therapy. A supplementary analysis in the primary stroke prevention group indicated no differences in stroke risk relative to warfarin for both “new starters” and “switchers” on dabigatran, except for switchers to D110 who were at an increased risk of stroke/TIA.
      Altogether, our study suggests that caution and vigilance is needed when switching prior VKA-experienced atrial fibrillation patients to dabigatran, especially for patients who are able to consistently maintain a high average time in therapeutic range. The extent to which the increased risk with dabigatran can be attributed to the intervention of switching is, however, not clear, particularly in light of the findings from randomized trials. A predefined sub-study of RE-LY among participants with prior stroke/transient ischemic attack found no significant difference between both dabigatran doses and warfarin in relation to the risk of stroke or systemic embolism.
      • Dentali F.
      • Riva N.
      • Crowther M.
      • Turpie A.G.G.
      • Lip G.Y.H.
      • Ageno W.
      Efficacy and safety of the novel oral anticoagulants in atrial fibrillation: a systematic review and meta-analysis of the literature.
      Of note, there were several key differences from the present study: for example, prior VKA experience was not explicitly accounted for, participants were much younger, and recurrent stroke rates were lower. However, in the phase III trial on rivaroxaban versus warfarin,
      • Patel M.R.
      • Mahaffey K.W.
      • Garg J.
      • et al.
      Rivaroxaban versus warfarin in nonvalvular atrial fibrillation.
      • Mahaffey K.W.
      • Hellkamp A.S.
      • Patel M.R.
      • et al.
      End of study transition from study drug to open-label vitamin K antagonist therapy: the ROCKET AF experience.
      a marked increase in cardiovascular events were seen during the transition to open-label therapy for patients switching therapy at the end of the trial. Meta-analyses and indirect comparisons have thus far confirmed the noninferiority findings from the RE-LY subgroup studies.
      • Rasmussen L.H.
      • Larsen T.B.
      • Graungaard T.
      • Skjøth F.
      • Lip G.Y.H.
      Primary and secondary prevention with new oral anticoagulant drugs for stroke prevention in atrial fibrillation: indirect comparison analysis.
      • Schneeweiss S.
      • Gagne J.J.
      • Patrick A.R.
      • Choudhry N.K.
      • Avorn J.
      Comparative efficacy and safety of new oral anticoagulants in patients with atrial fibrillation.
      Divergent findings in an observational study necessitate caution because design limitations can easily induce spurious results. Bias due to residual or unmeasured confounding by indication is a serious concern: an apparent treatment effect may simply reflect unaccounted indications for which the treatment was administered. In the present study, clinicians may have been more likely to switch patients with difficulties, maintaining a high time in therapeutic range. Because INR measurements are not available from the registries, we were unable to investigate this. While confounding bias alone is a viable explanation for our findings, it does raise new questions about the sources of such confounding. Indeed, in our analyses, we adjusted for variables that a priori would be expected to be strong confounders, yet we observed only a modest change in the crude point estimates. If confounding bias were to completely explain our findings, we speculate that there would either have to be many unmeasured confounders with a moderate effect or a few very strong unmeasured confounders. As a crude, quantitative assessment, a binary unmeasured confounder with a prevalence of 80% among switchers and 20% among warfarin persisters would only completely explain a Cox model HR of 1.7 if the confounder triples the rate of stroke/transient ischemic attack.
      • Lin D.Y.
      • Psaty B.M.
      • Kronmal R.A.
      Assessing the sensitivity of regression results to unmeasured confounders in observational studies.
      A combination of confounding bias and (causal) effects due to treatment switching also may explain our divergent findings. Patients may be less protected in the early period following a switch from warfarin, for example, because of “latent” strokes previously prevented by warfarin and appearing after warfarin discontinuation – or simply because the patient is less familiar with the new treatment. Indeed, plots of crude incidences lend some support to the observation that the early period after switching is a higher-risk period. Lastly, irrespective of whether or not our findings are partly attributable to drug-related causal effects, they still convey the important message that real-world prior stroke/transient ischemic attack “switchers” from warfarin to dabigatran represent a high-risk group. Our study may offer a valuable insight for a future randomized trial of outcomes after treatment switching.
      We conclude by commenting on the overall stroke risk differences between VKA-naive and VKA-experienced users, which are clearly visible in Figure 2. These are likely to be attributable to the “healthy user” selection inherent in the definition of the VKA-experienced stratum: to be included in the VKA-experienced stratum, one has to survive the time from treatment initiation to study inclusion with, at most, one stroke. The differences apparent in Figure 2 emphasize the appropriateness of treating the VKA-naive and VKA-experienced stratum separately.
      Limitations of the present observational study in terms of confounding bias have already been discussed, with the lack of INR measurements being a particular concern. Also, the accuracy of the proposed “intention-to-treat approach” relies on patients actually taking their drugs, as is the case for all comparative effectiveness studies. While the validity of an ischemic stroke diagnosis is high in the Danish Registry of Patients,
      • Krarup L.-H.
      • Boysen G.
      • Janjua H.
      • Prescott E.
      • Truelsen T.
      Validity of stroke diagnoses in a National Register of Patients.
      misclassification is also a relevant limitation (reassuringly, a more stringent definition of endpoints leads to results consistent with our main analysis). Lastly, one should be aware of the limitations of a comparative effectiveness study of a newly marketed drug such as dabigatran; these can include channeling of selected patient groups toward the new drug, as well as time-varying user population characteristics.
      • Schneeweiss S.
      • Gagne J.J.
      • Glynn R.J.
      • Ruhl M.
      • Rassen J.A.
      Assessing the comparative effectiveness of newly marketed medications: methodological challenges and implications for drug development.
      Strengths of the study include the use of large “real world” atrial fibrillation population, the long follow-up period, and the completeness of the registries.

      Conclusion

      In this nationwide cohort study, we found that for patients with a history of stroke/transient ischemic attack who are VKA-naive, both dabigatran doses provided similar protection to warfarin against recurrent stroke/TIA. Among VKA-experienced patients, the risk of recurrent stroke/transient ischemic attack was significantly increased compared with continued warfarin usage. Although clinical implications from observational data must be drawn carefully, our findings stress the importance of caution and vigilance when switching prior VKA-experienced atrial fibrillation patients to dabigatran, especially for patients with prior good-quality anticoagulation as reflected by a high time in therapeutic range.

      Appendix

      Supplementary Table 1ICD-10/ATC Codes Used to Identify Comorbid Conditions and Co-medication
      International Classification of Diseases 10th Revision (ICD-10) CodeAnatomical Therapeutic Chemical (ATC) Code
      Condition
       Congestive heart failureI11.0; I13.0; I13.2; I42.0; I50and C03C
       Left ventricular dysfunctionI50.1; I50.9
       HypertensionSee specified definition
      We identified subjects with hypertension from combination treatment with at least 2 of the following classes of antihypertensive drugs:I.Alpha adrenergic blockers (C02A, C02B, C02C)II.Non-loop diuretics (C02DA, C02L, C03A, C03B, C03D, C03E, C03X, C07C, C07D, C08G, C09BA, C09DA, C09XA52)III.Vasodilators (C02DB, C02DD, C02DG, C04, C05)IV.Beta-blockers (C07)V.Calcium channel blockers (C07F, C08, C09BB, C09DB)VI.Renin-angiotensin system inhibitors (C09).
       Diabetes mellitusE10.0; E10.1; E10.9; E11.0; E11.1; E11.9or A10
       Ischemic strokeI63; I64.9
       Systemic embolismI74
       Transient ischemic attackG45
       Aortic plaqueI70.0
       Peripheral arterial diseaseI70.2-I70.9; I71; I73.9
       Myocardial infarctionI21-I23
       Unstable anginaI20
       Cardiac arrestI46
       Abnormal renal functionI12; I13; N00-N05; N07; N11; N14; N17-N19; Q61
       Abnormal hepatic functionB15.0; B16.0; B16.2; B19.0; K70.4; K72; K76.6; I85
       BleedingI60-I62; D62; J94.2; H11.3; H35.6; H43.1; N02; N95; R04; R31; R58; K25.0; K26.0; K27.0; K28.0; K29.0; S06.3C; S06.4; S06.5; S06.6
       Alcohol intakeE22.4; E52.9A; F10; G31.2; G62.1; G72.1; I42.6; K29.2; K70; K86.0; L27.8A; O35.4M; T51; Z71.4; Z72.1
       Atrial fibrillationI48
      Medication
       DabigatranB01AE07
       WarfarinB01AA03
       AspirinB01AC06
       ClopidogrelB01AC04
       PhenprocoumonB01AA04
       Nonsteroidal antiinflammatory drugsM01A
      We identified subjects with hypertension from combination treatment with at least 2 of the following classes of antihypertensive drugs:
      • I.
        Alpha adrenergic blockers (C02A, C02B, C02C)
      • II.
        Non-loop diuretics (C02DA, C02L, C03A, C03B, C03D, C03E, C03X, C07C, C07D, C08G, C09BA, C09DA, C09XA52)
      • III.
        Vasodilators (C02DB, C02DD, C02DG, C04, C05)
      • IV.
        Beta-blockers (C07)
      • V.
        Calcium channel blockers (C07F, C08, C09BB, C09DB)
      • VI.
        Renin-angiotensin system inhibitors (C09).
      Supplementary Table 2Definition of Risk Scores
      Risk ScorePoints, if Present
      CHADS2
      Reflects stroke risk in atrial fibrillation patients not in anticoagulant therapy (Gage BF, van Walraven C, Pearce L, et al. Selecting patients with atrial fibrillation for anticoagulation: stroke risk stratification in patients taking aspirin. Circulation. 2004;110(16):2287-92).
       Congestive heart failure1
       Hypertension1
       Age ≥65 years1
       Diabetes mellitus1
       Stroke (ischemic stroke or transient ischemic attack,)2
      CHA2DS2VASc
      Reflects stroke risk in atrial fibrillation patients not in anticoagulant therapy (Lip GYH, Nieuwlaat R, Pisters R, Lane DA, Crijns HJGM. Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the euro heart survey on atrial fibrillation. Chest. 2010;137(2):263-72).
       Congestive heart failure or left ventricular dysfunction1
       Hypertension1
       Age ≥65 years1
       Age ≥75 years1
       Diabetes mellitus1
       Stroke (ischemic stroke, transient ischemic disease or systemic embolism)2
       Vascular disease (myocardial infarction, peripheral arterial disease, or aortic plaque)1
       Sex category (female)1
      HAS-BLED
      Reflects bleeding risk in atrial fibrillation patients undergoing anticoagulant therapy (Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJGM, Lip GYH. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest. 2010;138(5):1093-100).
       Hypertension1
       Abnormal renal function1
       Abnormal hepatic function1
       Stroke (ischemic stroke or transient ischemic attack)1
       Bleeding1
       Labile international normalized ratio
      Not included due to unavailable information.
      1
       Elderly age (≥65 years)1
       Drugs (aspirin, clopidogrel, or non-steroidal anti-inflammatory drugs)1
       Alcohol intake1
      Abbreviations: CHADS2 = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, and previous stroke or TIA; CHA2DS2VASc = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category (female sex); HAS-BLED = Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio (INR), Elderly, Drugs/Alcohol.
      Reflects stroke risk in atrial fibrillation patients not in anticoagulant therapy (Gage BF, van Walraven C, Pearce L, et al. Selecting patients with atrial fibrillation for anticoagulation: stroke risk stratification in patients taking aspirin. Circulation. 2004;110(16):2287-92).
      Reflects stroke risk in atrial fibrillation patients not in anticoagulant therapy (Lip GYH, Nieuwlaat R, Pisters R, Lane DA, Crijns HJGM. Refining clinical risk stratification for predicting stroke and thromboembolism in atrial fibrillation using a novel risk factor-based approach: the euro heart survey on atrial fibrillation. Chest. 2010;137(2):263-72).
      Reflects bleeding risk in atrial fibrillation patients undergoing anticoagulant therapy (Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJGM, Lip GYH. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial fibrillation: the Euro Heart Survey. Chest. 2010;138(5):1093-100).
      § Not included due to unavailable information.
      Supplementary Table 3Persistence-adjusted (Censoring Participants at the Time of Nonpersistence) Event Rates and Cox Hazard Ratios for the Various Endpoints, According to VKA-experience Stratum and Current Treatment
      No. EventsRate, % per YearCrude Hazard Ratio (95% CI)Adjusted
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      Hazard Ratio (95% CI)
      Vitamin K antagonist-naive stratum
       Stroke or transient ischemic attack
      Warfarin22323.21 (reference)1 (reference)
      Dabigatran 110 mg6917.20.73 (0.56-0.95)0.69 (0.52-0.92)
      Dabigatran 150 mg8423.91.08 (0.84-1.39)1.13 (0.86-1.49)
       Stroke
      Warfarin17417.71 (reference)1 (reference)
      Dabigatran 110 mg6014.80.82 (0.61-1.09)0.75 (0.55-1.03)
      Dabigatran 150 mg6618.11.08 (0.81-1.43)1.19 (0.87-1.61)
       Transient ischemic attack
      Warfarin555.31 (reference)1 (reference)
      Dabigatran 110 mg112.60.48 (0.25-0.91)0.50 (0.26-0.96)
      Dabigatran 150 mg215.41.06 (0.64-1.77)0.94 (0.55-1.63)
      Vitamin K antagonist-experienced stratum
       Stroke or transient ischemic attack
      Warfarin956.81 (reference)1 (reference)
      Dabigatran 110 mg4112.91.82 (1.26-2.63)1.5 (1.01-2.23)
      Dabigatran 150 mg2810.31.49 (0.98-2.27)1.67 (1.07-2.61)
       Stroke
      Warfarin735.21 (reference)1 (reference)
      Dabigatran 110 mg3511.02.03 (1.36-3.04)1.59 (1.03-2.45)
      Dabigatran 150 mg217.61.44 (0.89-2.34)1.81 (1.08-3.04)
       Transient ischemic attack
      Warfarin231.61 (reference)1 (reference)
      Dabigatran 110 mg92.71.64 (0.76-3.55)1.59 (0.69-3.69)
      Dabigatran 150 mg82.81.71 (0.76-3.81)1.41 (0.61-3.26)
      Abbreviations: CHA2DS2VASc = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category (female sex); CI = confidence interval; HAS-BLED = Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio (INR), Elderly, Drugs/Alcohol; VKA = vitamin K antagonist.
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      Supplementary Table 4Event Rates and Cox Hazard Ratios for the Various Endpoints, Requiring Endpoints to Have Been Registered as the Primary Diagnosis in Connection with Hospitalization for at Least One Night, According to VKA-experience Stratum and Current Treatment
      No. EventsRate, % per YearCrude Hazard Ratio (95% CI)Adjusted
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      Hazard Ratio (95% CI)
      Vitamin K antagonist-naive stratum
       Stroke or transient ischemic attack
      Warfarin975.31 (reference)1 (reference)
      Dabigatran 110 mg334.60.84 (0.56-1.25)0.69 (0.46-1.03)
      Dabigatran 150 mg233.60.67 (0.43-1.06)0.77 (0.48-1.25)
       Stroke
      Warfarin663.51 (reference)1 (reference)
      Dabigatran 110 mg263.60.98 (0.62-1.54)0.71 (0.45-1.13)
      Dabigatran 150 mg121.90.51 (0.28-0.95)0.62 (0.32-1.19)
       Transient ischemic attack
      Warfarin361.91 (reference)1 (reference)
      Dabigatran 110 mg91.20.62 (0.30-1.27)0.61 (0.30-1.24)
      Dabigatran 150 mg121.90.95 (0.49-1.83)0.99 (0.50-1.98)
      Vitamin K antagonist-experienced stratum
       Stroke or transient ischemic attack
      Warfarin592.61 (reference)1 (reference)
      Dabigatran 110 mg315.52.10 (1.36-3.25)1.49 (0.94-2.36)
      Dabigatran 150 mg203.91.53 (0.92-2.53)2.08 (1.20-3.59)
       Stroke
      Warfarin321.41 (reference)1 (reference)
      Dabigatran 110 mg234.02.85 (1.67-4.88)1.83 (1.04-3.21)
      Dabigatran 150 mg112.11.55 (0.78-3.07)2.71 (1.28-5.77)
       Transient ischemic attack
      Warfarin301.31 (reference)1 (reference)
      Dabigatran 110 mg91.61.20 (0.57-2.53)1.04 (0.47-2.30)
      Dabigatran 150 mg101.91.48 (0.72-3.03)1.50 (0.71-3.19)
      Abbreviations: CHA2DS2VASc = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category (female sex); CI = confidence interval; HAS-BLED = Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio (INR), Elderly, Drugs/Alcohol; VKA = vitamin K antagonist.
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      Supplementary Table 5Event Rates and Cox Hazard Ratios for the Various Endpoints, According to VKA-experience Stratum and Current Treatment, among Atrial Fibrillation Patients with No History of Stroke or Transient Ischemic Attack
      Vitamin K Antagonist-naïveVitamin K Antagonist-experienced
      WarfarinDabigatran

      110 mg
      Dabigatran

      150 mg
      WarfarinDabigatran

      110 mg
      Dabigatran

      150 mg
      Subjects, no921622363363601814841797
       Age, years
      Median728167748168
      Interquartile range65-7976-8662-7267-8076-8564-73
       Female sex, %41.454.736.737.95434.4
      Risk scores at baseline
       HAS-BLED, mean (SD)0.96 (0.88)1.38 (0.82)0.61 (0.74)1.16 (0.90)1.54 (0.87)0.91 (0.86)
       CHADS2, mean (SD)2.41 (1.43)3.15 (1.17)1.78 (1.19)2.61 (1.33)3.35 (1.20)2.19 (1.28)
       CHA2DS2VASc, mean (SD)1.73 (1.06)2.01 (0.90)1.50 (1.02)1.66 (0.91)1.94 (0.87)1.63 (1.00)
      Stroke or transient ischemic attack during follow-up
       No. events1945655984921
       Rate, % per year2.02.71.71.23.10.9
      Crude estimates
      Hazard ratio1 (reference)1.310.831 (reference)2.470.75
      95% confidence interval0.97-1.770.61-1.131.75-3.480.47-1.2
       Adjusted
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).
      estimates
      Hazard ratio1 (reference)0.991.051 (reference)1.731.04
      95% confidence interval0.72-1.360.76-1.451.2-2.480.64-1.7
      CHADS2: score ranging from 0-5 that reflect the stroke risk in atrial fibrillation patients not in anticoagulant therapy (see Supplementary Table 2).
      CHA2DS2VASc: score ranging from 0-9 that reflect the risk of stroke in atrial fibrillation patients not in anticoagulant therapy (see Supplementary Table 2).
      HAS-BLED: score ranging from 0-9 that reflect the risk of bleeding in atrial fibrillation patients undergoing anticoagulant therapy (see Supplementary Table 2).
      Abbreviations: CHADS2 = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, and previous stroke or TIA; CHA2DS2VASc = Congestive heart failure, Hypertension, Age 75 years, Diabetes mellitus, previous Stroke/transient ischemic attack, Vascular disease, Age 65-74 years, and Sex category (female sex); HAS-BLED = Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio (INR), Elderly, Drugs/Alcohol; SD = between-subjects standard deviation.
      Estimates adjusted for: CHA2DS2VASc and HAS-BLED components (binary); age (cubic spline); months since August 2011 (cubic spline); time since initiation of VKA therapy (if applicable; cubic spline).

      References

        • Albertsen I.E.
        • Rasmussen L.H.
        • Overvad T.F.
        • Graungaard T.
        • Larsen T.B.
        • Lip G.Y.H.
        Risk of stroke or systemic embolism in atrial fibrillation patients treated with warfarin: a systematic review and meta-analysis.
        Stroke. 2013; 44: 1329-1336
        • Banerjee A.
        • Marín F.
        • Lip G.Y.H.
        A new landscape for stroke prevention in atrial fibrillation: focus on new anticoagulants, antiarrhythmic drugs, and devices.
        Stroke. 2011; 42: 3316-3322
        • Connolly S.J.
        • Ezekowitz M.D.
        • Yusuf S.
        • et al.
        Dabigatran versus warfarin in patients with atrial fibrillation.
        N Engl J Med. 2009; 361: 1139-1151
        • Larsen T.B.
        • Rasmussen L.H.
        • Skjøth F.
        • et al.
        Efficacy and safety of dabigatran etexilate and warfarin in “real-world” patients with atrial fibrillation: a prospective nationwide cohort study.
        J Am Coll Cardiol. 2013; 61: 2264-2273
        • Sørensen R.
        • Gislason G.
        • Torp-Pedersen C.
        • et al.
        Dabigatran use in Danish atrial fibrillation patients in 2011: a nationwide study.
        BMJ Open. 2013; 3: e002758
        • Diener H.-C.
        • Connolly S.J.
        • Ezekowitz M.D.
        • et al.
        Dabigatran compared with warfarin in patients with atrial fibrillation and previous transient ischaemic attack or stroke: a subgroup analysis of the RE-LY trial.
        Lancet Neurol. 2010; 9: 1157-1163
        • Rasmussen L.H.
        • Larsen T.B.
        • Graungaard T.
        • Skjøth F.
        • Lip G.Y.H.
        Primary and secondary prevention with new oral anticoagulant drugs for stroke prevention in atrial fibrillation: indirect comparison analysis.
        BMJ. 2012; 345: e7097
        • Larsen T.B.
        • Rasmussen L.H.
        • Gorst-Rasmussen A.
        • et al.
        Myocardial ischemic events in “real world” patients with atrial fibrillation treated with dabigatran or warfarin.
        Am J Med. 2014; 127: 329-336.e4
        • Larsen T.B.
        • Gorst-Rasmussen A.
        • Rasmussen L.H.
        • Skjøth F.
        • Rosenzweig M.
        • Lip G.Y.H.
        Bleeding events among new starters and switchers to dabigatran compared with warfarin in atrial fibrillation.
        Am J Med. 2014; 127: 650-656
        • Lane D.A.
        • Lip G.Y.H.
        Dabigatran in atrial fibrillation: balancing secondary stroke prevention against bleeding risk.
        Lancet Neurol. 2010; 9: 1140-1142
        • Ezekowitz M.D.
        • Wallentin L.
        • Connolly S.J.
        • et al.
        Dabigatran and warfarin in vitamin K antagonist-naive and -experienced cohorts with atrial fibrillation.
        Circulation. 2010; 122: 2246-2253
        • Kildemoes H.W.
        • Sørensen H.T.
        • Hallas J.
        The Danish National Prescription Registry.
        Scand J Public Health. 2011; 39: 38-41
        • Lynge E.
        • Sandegaard J.L.
        • Rebolj M.
        The Danish National Patient Register.
        Scand J Public Health. 2011; 39: 30-33
        • Pedersen C.B.
        The Danish Civil Registration System.
        Scand J Public Health. 2011; 39: 22-25
        • Gooley T.A.
        • Leisenring W.
        • Crowley J.
        • Storer B.E.
        Estimation of failure probabilities in the presence of competing risks: new representations of old estimators.
        Stat Med. 1999; 18: 695-706
        • Dentali F.
        • Riva N.
        • Crowther M.
        • Turpie A.G.G.
        • Lip G.Y.H.
        • Ageno W.
        Efficacy and safety of the novel oral anticoagulants in atrial fibrillation: a systematic review and meta-analysis of the literature.
        Circulation. 2012; 126: 2381-2391
        • Patel M.R.
        • Mahaffey K.W.
        • Garg J.
        • et al.
        Rivaroxaban versus warfarin in nonvalvular atrial fibrillation.
        N Engl J Med. 2011; 365: 883-891
        • Mahaffey K.W.
        • Hellkamp A.S.
        • Patel M.R.
        • et al.
        End of study transition from study drug to open-label vitamin K antagonist therapy: the ROCKET AF experience.
        Circ Cardiovasc Qual Outcomes. 2013; 6: 470-478
        • Schneeweiss S.
        • Gagne J.J.
        • Patrick A.R.
        • Choudhry N.K.
        • Avorn J.
        Comparative efficacy and safety of new oral anticoagulants in patients with atrial fibrillation.
        Circ Cardiovasc Qual Outcomes. 2012; 5: 480-486
        • Lin D.Y.
        • Psaty B.M.
        • Kronmal R.A.
        Assessing the sensitivity of regression results to unmeasured confounders in observational studies.
        Biometrics. 1998; 54: 948-963
        • Krarup L.-H.
        • Boysen G.
        • Janjua H.
        • Prescott E.
        • Truelsen T.
        Validity of stroke diagnoses in a National Register of Patients.
        Neuroepidemiology. 2007; 28: 150-154
        • Schneeweiss S.
        • Gagne J.J.
        • Glynn R.J.
        • Ruhl M.
        • Rassen J.A.
        Assessing the comparative effectiveness of newly marketed medications: methodological challenges and implications for drug development.
        Clin Pharmacol Ther. 2011; 90: 777-790