Comparative Stroke, Bleeding, and Mortality

ARTICLE IN PRESS ( » 12 hours). We censored for admission to a skilled nursing facility or nursing home due to concerns about incomplete capture of outcomes in these settings, and for transfer to hos- pice care because deaths in these patients were expected and therefore unlikely to be anticoagulant related. Study outcomes were hospitalized thromboembolic stroke, intracranial hemorrhage, major extracranial bleed- ing, and all-cause mortality. To facilitate comparison with other studies, results for major gastrointestinal bleeding, a large component of major extracranial bleeding, are also presented in the Supplement (available online). Outcomes were defined using ICD-9 codes from the first hospital dis- charge diagnosis position ( Supplementary Table 2 , avail- able online). Thromboembolic stroke was defined using a validated algorithm with a positive predictive value (PPV) of 88%-95% . 18-20 Intracranial hemorrhage was defined using a validated algorithm for nontraumatic intracranial hemorrhage (PPV 89%-97%) , 18-20 to which we added codes for intracranial hemorrhage accompanied by nonpenetrat- ing head trauma to capture events that might have been pre- ceded by a fall. Major extracranial bleeding was based on a validated algorithm for hospitalized bleeding by Cunning- ham et al (PPV 87%) , 21 with a requirement that the bleed- ing event was treated with red blood cell or whole blood transfusion, or involved a critical site (ie, intraarticular, pericardial, retroperitoneal) , 21 or resulted in death. Mortal- ity was ascertained by linkage to Social Security files, which capture over 95% of deaths for persons 65 years or older in the United States . 22 Our death outcome included deaths occurring as the first study outcome or within 30 days after hospitalization for another outcome event. Weighted Kaplan-Meier cumulative incidence plots were generated to characterize risk over time . 23 For each outcome, a single weighted Cox proportional hazards model with robust estimation was used to estimate hazard ratios (HR) and 95% confidence intervals (CI) for all NOAC − warfarin and NOAC − NOAC pairwise comparisons. Adjusted inci- dence rates and incidence rate differences were also esti- mated. For thromboembolic stroke, intracranial hemorrhage, and major extracranial bleeding, the adjusted 30-day case fatality rate was determined. Prespecified subgroup analyses were performed for all out- comes in categories defined by age, sex, use of antiplatelet agents , and CHA 2 DS 2 -VASc and HAS-BLED scores. Prespe- cified sensitivity analyses included increasing the gap between anticoagulant prescriptions from 3 to 14 days; restricting anal- ysis to patients with at least 2 dispensings of their study drug; restriction to patients initiating an anticoagulant on or after apixaban’s approval date (December 28, 2012) to examine for potential time-period bias; and repeating the analysis using Cox regression without and with multivariable adjustment. We also performed several post hoc sensitivity analyses ( Appendix , available online). This study was classified as public health surveillance by the Food and Drug Administration (FDA) and was exempt from review by its Research in Human Subjects Committee. Analyses were performed using R 3.3.2 (R Foundation for Statistical Computing, Vienna, Austria) and SAS v. 9.4 (SAS Institute Inc., Cary, NC). RESULTS A total of 448,944 anticoagulant initiators contributed 159,927 person-years of on-treatment follow-up (mean dura- tion 130 days). The mean age across cohorts was 75.4 years, of whom 47.4% were female. Prior to adjustment, there were minor differences for age, kidney failure, obesity, smoking, cardioversion, emergency department visits, pre- scriber specialty, and use of injectable anticoagulants, digoxin, and loop diuretics ( Table , Supplementary Table 3 , available online) After inverse probability of treatment weights adjustment, cohorts were closely balanced for all covariates ( Supplementary Table 4 , available online). During follow-up, there were 11,263 outcome events ( Supplementary Table 5 , available online). Adjusted inci- dence rates for thromboembolic stroke, intracranial hem- orrhage, and all-cause mortality were highest with warfarin, while major extracranial bleeding rates were highest with rivaroxaban ( Figure 1 ) . Compared with warfarin, each NOAC was associated with significantly lower HRs for thromboembolic stroke (20%-29% reduction; P = .002 [dabigatran], P < .001 [rivar- oxaban, apixaban]), intracranial hemorrhage (35%-62% reduction; P < .001 [each NOAC]), and mortality (19%- 34% reduction; P < .001 [each NOAC]) ( Figure 2 , Supple- mentary Table 6 , available online). Major extracranial bleeding risk was increased with rivar- oxaban, decreased with apixaban, and was similar to warfa- rin for dabigatran. Major gastrointestinal bleeding, which accounted for 82% of major extracranial bleeding, was increased with dabigatran and rivaroxaban and decreased with apixaban, compared with warfarin ( Supplementary Table 6 ) . Although the proportional hazards assumption was violated for major extracranial and major gastrointestinal bleeding, a time-dependent analysis met the proportional hazards assumption and generally confirmed results from the primary analysis, with the most important difference being that risks for both bleeding outcomes were significantly increased with dabigatran compared with warfarin only after more than 60 days of treatment ( Supplementary Table 7 , available online). The NOACs were similar to each other for thromboem- bolic stroke, but rivaroxaban was associated with significantly increased risks of intracranial hemorrhage (vs dabigatran: HR 1.71; 95% CI, 1.35-2.17), major extracranial bleeding (vs dabigatran: HR 1.32; 95% CI, 1.21-1.45; vs apixaban: HR 2.70; 95% CI, 2.38-3.05), and death (vs dabigatran: HR 1.12; 95% CI, 1.01-1.24; vs apixaban: HR 1.23; 95% CI, 1.09-1.38) ( Figure 2 , Supplementary Table 6 ) . Dabigatran was associated with significantly reduced risk of intracranial hemorrhage (HR 0.70; 95% CI, 0.53-0.94) and with significantly increased risk of major extracranial bleeding (HR 2.04; 95% CI, 1.78- 2.32) compared with apixaban. Adjusted Kaplan-Meier plots ( Figure 3 ) and adjusted incidence rate differences Graham et al Comparative Safety and Effectiveness of Anticoagulants in Atrial Fibrillation 3