Supplementary Online Content

Transcription

1 Supplementary Online Content Lip GYH, Skjøth F, Nielsen PB, Kjaeldgaard JN, Larsen TB. Effectiveness and safety of standarddose nonvitamin K antagonist anticoagulants and wafarin among patients with atrial fibrillation with a single stroke factor: a nationwide cohort study. JAMA Cardiol. Published online June 14, doi: /jamacardio emethods. Extended Methods. ereferences. etable 1. Definitions on Comorbidity and Concomitant Medication According to ICD-10 Codes and ATC-Codes. etable 2. Risk Score Definitions etable 3. Number of Events, and Weighted Event Rates According to Treatment. AF Hospitalized Cohort Sensitivity Analysis. etable 4. Number of Events, and Weighted Event Rates According to Treatment. Low Mortality Cohort Sensitivity Analysis. etable 5. Number of Events, and Weighted Event Rates According to Treatment. Patients in Age Between 65 and 75 years Supplementary Analysis. etable 6. Number of Events, and Weighted Event Rates According to Treatment. Patients With Hypertension Supplementary Analysis. etable 7. Number of Events, and Weighted Event Rates According to Treatment. Male Patients Supplementary Analysis. etable 8. Number of Events, and Weighted Event Rates According to Treatment. Female Patients Supplementary analysis. etable 9. Number of Events, and Weighted Event Rates According to Treatment. Falsification Analysis\ efigure 1. Flowchart Diagram of Study Population Selection. efigure 2. Plot of Maximum of Pairwise Standardized Differences for Cohort Baseline Characteristics Before and After Applying Propensity Weights (IPTW). Main Analysis. efigure 3. Propensity Score Densities for the Treatment Options. Main Analysis. efigure 4. Propensity Score Densities for the Treatment Options. AF Hospitalized Cohort. efigure 5. Propensity Score Densities for the Treatment Options. Low Mortality Cohort. efigure 6. Plot of Maximum of Pairwise Standardized Differences for Cohort Baseline Characteristics Before and After Applying Propensity Weights (IPTW). AF-Hospitalised Cohort.

2 efigure 7. Plot of Maximum of Pairwise Standardized Differences for Cohort Baseline Characteristics Before and After Applying Propensity Weights (IPTW). Low Mortality Cohort. efigure 8. Propensity Weighted (IPTW) Cox Hazard Ratios for 2.5 Years Follow-Up for NOACs Compared With Warfarin for Stroke and Death End Points. This supplementary material has been provided by the authors to give readers additional information about their work.

3 emethods. Extended Methods We conducted an observational cohort study of Danish citizens initiating OAC treatment for stroke prevention in AF with 1 stroke risk factor. To align with the RCTs we focused on standard dose NOAC agents only (apixaban 5mg bid, dabigatran 150mg bid, rivaroxaban 20mg od). In Denmark all citizens hold a unique identification number being used throughout all nationwide databases. Danish legislation allows linkage between databases for research purposes. For this study we used data from three Danish nationwide databases: (i) the Danish National Prescription Registry[1] which holds detailed information on every prescription withdrawal since 1994; (ii) the Danish National Patient Register[2] established in 1977, which includes data for >99% of somatic hospital admissions; and (iii) the Danish Civil Registration System[3] with demographic information. Study population The study was based on new users of OAC with no hospital information for treatment for other indications than AF. We identified patients with first-time purchases of each NOAC from their respective dates of approval for AF: apixaban (December 10, 2012), dabigatran (August 10, 2011), rivaroxaban (February 1, 2012), as well as patients who initiated warfarin treatment (since August 10, 2011). Patient inclusion was terminated by February 28, For this analysis, we only recruited patients with one non-gender stroke risk factor assigned 1 point in the CHA 2 DS 2 -VASc score, ie: congestive heart failure or left ventricular disease, hypertension, 65yr age <75yr, diabetes or vascular diseases (myocardial infarction, aortic plaque or peripheral aortic disease), thus excluding patients with prior stroke, systemic embolism or transient ischemic attack or age >75yr. As the randomized trials were primarily focused on standard doses, and anticipating that the study population was only in minor degree eligible for reduced dose treatment, all NOACs were restricted to

4 standard doses: apixaban 5mg bid, dabigatran 150mg bid and rivaroxaban 20mg od. Warfarin is only available in 2.5mg dose tablets in Denmark. To establish an OAC naïve cohort, we excluded patients with prior experience of any OAC inclusive doses approved for other indications within one year. Eventually, we excluded patients with prior hospital diagnoses indicating valvular AF (mitral stenosis or mechanical heart valves) or venous thromboembolism (pulmonary embolism or deep venous thromboembolism). Also patients with contra-indication for standard NOAC dose regimens due to renal impairment were excluded. This population formed the main cohort for the analyses. To increase certainty for the indication of nonvalvular AF sensitivity analyses were reported on the subgroup with prior hospital diagnose of AF. To increase homogeneity[4] across patient groups, due to potential selective prescribing related to comorbidity with high mortality, we also performed a sensitivity analysis excluding patients with prior heart failure, chronic pulmonary disease (CPD) or cancer. Endpoints and variable definitions Clinical endpoints were extracted from hospital discharge codes in the Danish National Patient Register using the 10th Revision of ICD codes (see Supplementary Table 1 for specific codes) with follow-up until April 30, Primary and secondary discharge codes from non-emergency ward were extracted. To minimize the risk of interpreting historical codes as new outcomes we only analysed endpoints obtained as discharge codes from hospitalizations, thus not including codes from ambulatory visits. Effectiveness was investigated by the occurrence of ischemic stroke or systemic embolism (SE). All-cause death, obtained from the Danish Civil Registration System, was included as a lone endpoint. Safety of treatment was investigated by the following bleeding events that have been the recent focus in relation to the NOACs[5,6]: intracranial, gastro-intestinal, traumatic intracranial or clinically relevant nonmajor bleeding reported in total as any bleeding and specifically for intracranial (incl. traumatic) or gastrointestinal bleedings (see supplementary Table 1 for ICD-10 discharge codes).

5 Patient s baseline comorbidities and co-medications at treatment initiation (as listed in Table 1) were ascertained from the Danish National Patient Registry and the Danish National Prescription Registry (for code definitions, see Supplementary Table 1). The overall risk of bleeding was assessed by the HAS-BLED score[7] (see score definition in Supplementary Table 2). The Danish registries are well-validated with sufficiently high positive predictive value (>80%) of included comorbidities and outcomes; the registries have been deemed as a valuable resource of epidemiological investigations.[8] Statistical analysis Time-to-event analysis was used to compare the risk of an endpoint between treatment groups, measuring risk time from initial prescription and until the relevant event, emigration, death, or end of follow-up, whichever came first. An intention to treat approach was applied for the main analyses. This was supplemented by a continuous treatment analysis by censoring follow-up if the patient were prescribed another treatment than that initiated. incidence s were calculated as number of events divided by person-time. Cox regression with robust variance estimator was used to compare event s between the treatment groups with warfarin as the primary reference. To address confounding by indication of treatment an inverse probability of treatment weighted (IPTW) analysis was applied. Such approach is suitable in the situations with several treatment alternatives.[9,10] Generalized boosted models, based on 10,000 regression trees, where used to calculate weights for optimal balance between the treatment populations[11]. We used weights that enabled estimates representing population average treatment effects. The underlying propensity models included the following treatment predictors: age (continuous); binary indicators for sex, prior bleeding, vascular disease, hypertension, diabetes, renal disease, chronic pulmonary disease, heart failure, cancer, recent prescription of aspirin, beta blockers, NSAIDs, statins, loop diuretics, and the HAS-BLED score. The treatments should be contrasted on comparable populations and any patient must have positive probability for any treatment, hence substantial overlap between the scores for each treatment should be

6 present. This was assessed by graphical inspection of the weight distributions.[10] Balance between treatment-populations were evaluated by standardized differences of all baseline covariates, using a threshold of 0.1 to indicate imbalance.[12] Further model control was done by applying the propensity weighted cohort in the analyses on (falsification) endpoints that a-priori should be expected not to be associated with the effects of treatment [13]. For this study, we considered pneumonia, hip-fractures, cancer, and urinary tract infections. Successful balancing of the propensity weighing should lead to non-differential treatment effects apart from random fluctuations[14]. Sensitivity of inclusion criteria and applied analytic methods were analyzed: we repeated the analyses on the subgroup with hospital discharge diagnosis for AF either before or within 30 days since first prescription. In addition, as there may be confounding due to differences in the patient s health status, we also report analyses where patients with high mortality conditions (heart failure, cancer, or chronic pulmonary disorder) at treatment initiation were excluded. This allowed for an evaluation of potential bias resulting from patients with severe comorbidities could be more likely to discontinue treatment especially for conditions with relative few symptoms (i.e. AF) [15], as well as potential bias due to selective treatment allocation. Eventually, the results of the IPTW analysis were compared with trimmed analysis removing 5% of the extreme weights as well as with an ordinary crude and adjusted analysis (data not shown). The analyses on the entire population were supplemented by stratified analyses on the populations with one of the frequent risk factors respectively older than 65, and patients with hypertension; additionally, the analyses were reported for men and females sepaly. Stata/MP version 14 and R version was used for the statistical analysis. A two-sided P value <0.05 was considered statistically significant.

7 Ethical Considerations No ethical approval is required for anonymous register studies in Denmark. The study was approved by the Danish Data Protection Agency (J. No. File No ). ereferences 1 Kildemoes HW, Sørensen HT, Hallas J. The Danish National Prescription Registry. Scand J Public Health 2011;39: doi: / Lynge E, Sandegaard JL, Rebolj M. The Danish National Patient Register. Scand J Public Health 2011;39:30 3. doi: / Pedersen CB. The Danish Civil Registration System. Scand J Public Health 2011;39: Shrank WH, Patrick AR, Brookhart MA. Healthy user and related biases in observational studies of preventive interventions: A primer for physicians. J Gen Intern Med 2011;26: doi: /s Larsen TB, Gorst-Rasmussen A, Rasmussen LH, et al. Bleeding events among new starters and switchers to dabigatran compared with warfarin in atrial fibrillation. Am J Med 2014;127: e5. doi: /j.amjmed Holster IL, Valkhoff VE, Kuipers EJ, et al. New oral anticoagulants increase risk for gastrointestinal bleeding: a systematic review and meta-analysis. Gastroenterology 2013;145: e15. doi: /j.gastro Pisters R, Lane DA, Nieuwlaat R, et al. 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: doi: /chest Schmidt M, Schmidt SSAJ, Sandegaard JLJ, et al. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015;7: doi: /clep.s Robins JM, Hernán MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology 2000;11: Stürmer T, Wyss R, Glynn RJ, et al. Propensity scores for confounder adjustment when assessing the

8 effects of medical interventions using nonexperimental study designs. J Intern Med 2014;275: doi: /joim McCaffrey DF, Griffin BA, Almirall D, et al. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med 2013;32: doi: /sim Austin PC. Some Methods of Propensity-Score Matching had Superior Performance to Others: Results of an Empirical Investigation and Monte Carlo simulations. Biometrical J 2009;51: doi: /bimj Lipsitch M, Tchetgen E, Cohen T. Negative Controls: A Tool for Detecting Confounding and Bias in Observational Studies. Epidemiology 2011;21: doi: /ede.0b013e3181d61eeb.negative 14 Prasad V, Jena A. Prespecified Falsification End Points: Can they validate true observational associations? Jama 2013;309: doi: /s Glynn RJ, Knight EL, Levin R, et al. Paradoxical relations of drug treatment with mortality in older persons. Epidemiology 2001;12:

9 etable 1: Definitions on comorbidity and concomitant medication according to ICD-10 codes and ATCcodes. Conditions marked with was used in the calculation of the CHA 2 DS 2 -VASc score. Conditions marked with # was used in the calculation of the HAS-BLED score. International Classification of Diseases 10th revision (ICD-10) code Anatomical Therapeutic Chemical (ATC) code Condition Congestive heart failure I110 I130 I132 I420 I50 CO3C and C09 Left ventricular dysfunction I501 I509 #Hypertension Diabetes mellitus #Ischemic stroke Systemic embolism #Transient ischemic disease E100 E101 E109 E110 E111 E119 I63 I64 I74 G45 See specified definition* A10 Aortic plaque I70.0 Peripheral arterial disease Myocardial infarction #Mode/severe renal disease #Mode/Severe liver disease Cancer Chronic pulmonary disorder Mitral stenosis Mechanical heart valve #Haemorrhagic stroke intercranial bleeding #Extracranial or unclassified major bleeding #Gastrointestinal bleeding I702-I709 I71 I739 I21-I23 I12 I13 N00 N01 N02 N03 N04 N05 N07 N11 N14 N17 N18 N19 Q61 B150 B160 B162 B190 K704 K72 K766 I85 C J40 J41 J42 J43 J44 J45 J46 J47 J60 J61 J62 J63 J64 J65 J67 J684 J701 J703 J841 J920 J921 J982 J983 I05 Z952 Z953 Z954 I60 I61 I62 D62 J942 H113 H356 H431 N02 R04 R31 R58 K250 K252 K254 K260 K262 K264 K270 K272 K274 K280

10 K282 K290 K291 #Traumatic intercranial bleeding #Alcohol Pulmonary embolism Deep venous thromboembolism Atrial fibrillation Hip fracture Pneumonia Urinary tract infection S063C S064 S065 S066 E224 E529A F10 G312 G621 G721 I426 K292 K70 K860 L278A O354 T51 Z714 Z721 I26 I801 I802 I803 I808 I809 I819 I636 I676 I822 I823 I829 I48 S72 J12 J13 J14 J15 J16 J17 J18 A481 A709 N30 Medication Apixaban Dabigatran Rivaroxaban Warfarin Phenprocoumon #Aspirin #Clopidogrel Beta-blockers Statins #NSAIDs Loop diuretics ACE/ARB inhibitors B01AF02 B01AE07 B01AF01 B01AA03 B01AA04 B01AC06 B01AC04 C07 C10 M01A C03C C09 NSAIDs: Non Steroidal Anti Inflammatory Drugs. * We identified subjects with hypertension from combination treatment with at least two 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)

11 VI Renin-angiotensin system inhibitors (C09) As well as combination drugs: C07B, C09BB04, C09DA, C09DB, C09DX01, C09DX04 Danish hospital discharge codes have been validated in extensively in the literature, please consult Schmidt et al. Clin Epid 2015;7: for a comprehensive overview and Sundbøll et al. BMJ Open 2016;6(11).

12 etable 2: Risk score definitions Risk score CHA 2 DS 2 VASc a Points if present Congestive heart failure or Left Ventricular Dysfunction 1 Hypertension 1 Age 65 years 1 Age 75 years 1 Diabetes mellitus 1 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 b Hypertension 1 Abnormal renal function 1 Abnormal hepatic function 1 Stroke (ischemic stroke or transient ischemic attack) 1 Bleeding 1 Labile international normalized ratio c 1 Elderly age ( 65 years) 1 Drugs (aspirin, clopidogrel, or non-steroidal anti-inflammatory drugs) 1 Alcohol intake 1 a Reflects stroke risk in atrial fibrillation patients not in anticoagulant therapy (Lip GYH, et al. Chest 2010;137(2):263-72) b Reflects bleeding risk in atrial fibrillation patients undergoing anticoagulant therapy (Pisters R, et al. Chest 2010;138(5): ) c Not included due to unavailable information

13 etable 3: Number of events, crude and weighted event s according to treatment. AF hospitalized cohort Sensitivity analysis. s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. Apixaban Dabigatran Rivaroxaban Warfarin One Year follow-up Events Weighte d Events Weighte d Events Weighted Events Weighted Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding < < years follow-up Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding <

14 etable 4: Number of events, crude and weighted event s according to treatment. Low mortality cohort Sensitivity analysis. s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. Apixaban Dabigatran Rivaroxaban Warfarin One Year follow-up Events Weighte d Events Weighte d Events Weighted Events Weighted Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding < < years follow-up Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding <

15 etable 5: Number of events, crude and weighted event s according to treatment. Patients aged between 65 and 75 years Supplementary analysis. s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. Apixaban Dabigatran Rivaroxaban Warfarin One Year follow-up Events Weighte d Events Weighte d Events Weighted Events Weighted Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding < < < years follow-up Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding < <

16 etable 6: Number of events, crude and weighted event s according to treatment. Patients with hypertension Supplementary analysis. s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. Apixaban Dabigatran Rivaroxaban Warfarin One Year follow-up Events Weighte d Events Weighte d Events Weighted Events Weighted Ischemic stroke/se < < All-cause death < < Any Bleeding < < Intracranial / Gastrointestinal bleeding < < years follow-up Ischemic stroke/se < < All-cause death < Any Bleeding < Intracranial / Gastrointestinal bleeding < <

17 etable 7: Number of events, crude and weighted event s according to treatment. Male patients Supplementary analysis. s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. Apixaban Dabigatran Rivaroxaban Warfarin One Year follow-up Events Weighte d Events Weighte d Events Weighted Events Weighted Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding < < < years follow-up Ischemic stroke/se All-cause death Any Bleeding Intracranial / Gastrointestinal bleeding < <

18 etable 8: Number of events, crude and weighted event s according to treatment. Female patients Supplementary analysis. s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. Apixaban Dabigatran Rivaroxaban Warfarin One Year follow-up Events Weighte d Events Weighte d Events Weighted Events Weighted Ischemic stroke/se < All-cause death Any Bleeding < < Intracranial / Gastrointestinal bleeding < < years follow-up Ischemic stroke/se All-cause death Any Bleeding < Intracranial / Gastrointestinal bleeding < <

19 etable 9: Number of events, crude and weighted event s according to treatment. Falsification analysis s are events divided by person time pr 100 years, weighted s are based on IPTW population and express population average treatment s pr 100 years. One year follow-up. Apixaban Dabigatran Rivaroxaban Warfarin Main analysis cohort Events Weighte d Events Weighte d Events Weighted Events Weighted Cancer Hip fracture Pneumonia Urinary tract infection AF hospitalised cohort Cancer Hip fracture Pneumonia Urinary tract infection Low mortality cohort Cancer Hip fracture Pneumonia Urinary tract infection Falsification outcomes were identified as primary or secondary discharge codes obtained during hospitalisation. The following ICD-10 codes were used: Cancer: C; Hip (femur) fracture: S72; Pneunomia: J12-J18 A481 A709; Urinary tract infection: N30.

20 efigure 1. Flowchart diagram of study population selection.

21 efigure 2: Plot of maximum of pairwise standardized differences for cohort baseline characteristics before ( ) and after ( ) applying propensity weights (IPTW). Main analysis. The vertical dashed line indicates a threshold of 0.01; all points below this threshold suggest good covariate balance across groups

22 efigure 3: Propensity score densities for the treatment options. Main analysis. Vertical bars below the presented curves represent individual data points.

23

24

25 efigure 4: Propensity score densities for the treatment options. AF hospitalised cohort. Vertical bars below the presented curves represent individual data points.

26

27 efigure 5: Propensity score densities for the treatment options. Low-mortality cohort. Vertical bars below the presented curves represent individual data points.

28

29

30 efigure 6: Plot of maximum of pairwise standardized differences for cohort baseline characteristics before ( ) and after ( ) applying propensity weights (IPTW). AF-hospitalized cohort. The vertical dashed line indicates a threshold of 0.01; all points below this threshold suggest good covariate balance across groups.

31

32 efigure 7: Plot of maximum of pairwise standardized differences for cohort baseline characteristics before ( ) and after ( ) applying propensity weights (IPTW). Low-mortality cohort. The vertical dashed line indicates a threshold of 0.01; all points below this threshold suggest good covariate balance across groups.

33

34 efigure 8: Propensity weighted (IPTW) Cox hazard ratios for 2.5 year follow-up for NOACs compared to warfarin for effectiveness, safety, and death endpoints.

35