Clinical Characteristics and Outcomes Associated With Oral Anticoagulant Use Among Patients Hospitalized With Intracerebral Hemorrhage

Key Points Question What is the association between prior oral anticoagulant use (factor Xa [FXa] inhibitors, warfarin, or none) and in-hospital outcomes among patients with nontraumatic intracerebral hemorrhage (ICH)? Findings In this registry-based cohort study of 219 701 patients with ICH, prior use of FXa inhibitors was associated with higher in-hospital mortality vs no prior anticoagulant use, although FXa inhibitors were associated with lower in-hospital mortality than warfarin. Meaning These findings suggest that patients with FXa inhibitor–associated ICH have a higher risk of mortality than those not taking an oral anticoagulant but better outcomes than those with warfarin-related ICH.


Introduction
Intracerebral hemorrhage (ICH) is the most feared complication of oral anticoagulant (OAC) therapy and is associated with high mortality and morbidity. Compared with warfarin, factor Xa (FXa) inhibitors, such as apixaban, rivaroxaban, and edoxaban, reduce the risk of ICH by 52%. 1 Despite their favorable safety profiles, as many as 0.5% of patients taking FXa inhibitors will still experience an intracranial bleeding event with each year of therapy. [2][3][4] Although the use of FXa inhibitors has increased substantially over the past decade, prior studies [5][6][7][8] of FXa inhibitor-associated ICH are limited in size and scope. We have characterized ICH in patients taking non-vitamin K antagonist OACs, including direct thrombin inhibitor (dabigatran) and FXa inhibitors as a group. 9 Given their differences in mechanism of action and targets at the coagulation cascade, there remains limited experience with FXa inhibitor-associated ICH. In addition, many patients were taking both anticoagulant and antiplatelet therapy before ICH, but the potential incremental risk associated with concomitant therapy remains unknown.

Using data from the American Heart Association and American Stroke Association Get With The
Guidelines-Stroke (GWTG-Stroke) registry, we sought to evaluate the characteristics of patients who experienced a nontraumatic ICH with preceding use of FXa inhibitors compared with no OAC or with warfarin, and to determine the risk of mortality and disability according to the type of anticoagulants, and any incremental risk associated with concomitant antiplatelet therapy in nationwide clinical practice.

Data Source
The GWTG-Stroke program is an ongoing, voluntary, national stroke registry and quality improvement initiative sponsored by the American Heart Association and American Stroke Association. Details of GWTG-Stroke registry data collection and variable definitions have been described elsewhere. 10 Standardized data collection includes patient demographic characteristics, medical history, medications taken before admission, diagnostic testing, brain imaging, in-hospital treatment, and outcomes. The validity and reliability of data collection have been reported in previous research. 11 Each participating hospital received either human research approval to enroll patients without individual patient consent under the Common Rule (45 CFR §46) or a waiver of authorization and exemption from subsequent review by their institutional review board. IQVIA, Inc serves as the data collection and coordination center. The Duke Clinical Research Institute serves as the data analysis center and has an agreement to analyze the aggregate deidentified data for research purposes. This study was approved by the institutional review board of Duke University.
This report follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.
inhibitor) were not included in the study population. Because andexanet-α, a specific reversal agent for rivaroxaban and apixaban, may affect outcomes in ICH and because the GWTG-Stroke registry did not previously collect the name of the reversal agent, we limited our study period until May 3, 2018, before the US Food and Drug Administration's approval of andexanet-α. Prior use of OACs was defined as documentation of patients taking an OAC within 7 days before hospital arrival. On the basis of this information, we categorized patients into 3 mutually exclusive groups: direct FXa inhibitors, including rivaroxaban, apixaban, or edoxaban; warfarin; and no OAC use before ICH. The Figure shows details of inclusion and exclusion criteria. Briefly, we excluded patients who had in-hospital stroke onset, transferred out to another hospital, had discharge information missing, left against medical advice, had a prosthetic heart valve, or took anticoagulants other than warfarin or FXa inhibitors before their stroke. After these exclusions, the final study population consisted of 219 701 patients with nontraumatic ICH from 1870 GWTG-Stroke hospitals in the US. Because the National Institute of Health Stroke Scale (NIHSS) score (range, 0-42, with a higher score indicating greater stroke severity) is a critical factor associated with outcomes, we further created a subgroup of individuals with complete NIHSS data (143 340 patients [65.2%]) for a sensitivity analysis.
The primary outcome was in-hospital mortality (yes or no). Secondary outcomes included a composite measure of in-hospital mortality or discharge to hospice (yes or no), discharge disposition (home vs other), ambulatory status at discharge (able to ambulate independently vs not), and modified Rankin Scale (mRS) score at discharge (range, 0 [no symptoms] to 6 [death]). Patients with an mRS score of 0 or 1 were classified as free from substantial disability, and those with an mRS score of 0 to 2 were classified as functionally independent.

Statistical Analysis
Medians (25th to 75th percentiles) and percentages were used to describe the distribution of continuous and categorical variables, respectively. Baseline characteristics were compared across the 3 prior anticoagulation treatment groups using the Pearson χ 2 test for categorical variables and Kruskal-Wallis test for continuous variables. Multivariable logistic regression models were performed to investigate the associations of prior anticoagulation therapies with each clinical outcome. Because it is inappropriate to impute outcome measures, complete case analyses were performed for death (100% of cases), discharge disposition (100% of cases), ambulatory status (67.6% of cases), and mRS (67.7% of cases) models.
The prior anticoagulation treatment was included as an independent variable, with no OAC or warfarin as the reference groups. A generalized estimation equations modeling approach was used to account for within-hospital clustering of patients. These analyses adjusted for baseline patient demographic and clinical factors before the index event and hospital characteristics that are expected to be associated with outcome and have been used in prior GWTG-Stroke analyses. 9,12

Figure. Study Population Flowchart
Covariates included age, sex, race/ethnicity (admission staff, medical staff, or both recorded the patient's self-reported race/ethnicity, usually during the registration), insurance, medical history (atrial fibrillation or flutter, coronary artery disease or prior myocardial infarction, prior stroke, prior transient ischemic attack, carotid stenosis, heart failure, hypertension, peripheral vascular disease, diabetes, dyslipidemia, obesity, renal insufficiency, smoking status, and drug or alcohol abuse), arrival and admission information (emergency medical services arrival vs private transportation, transfer-in, and arrived off-hours), medication taken before admission (antihypertensive, lipid-lowering medication, diabetic medication, single-antiplatelet agent [aspirin, clopidogrel, prasugrel, ticagrelor, or ticlopidine], dual-antiplatelet therapy [aspirin plus dipyridamole or aspirin plus clopidogrel, prasugrel, ticagrelor, or ticlopidine], other antiplatelet or combination), rural hospital, hospital number of beds, academic center, geographic region, primary stroke center, and comprehensive stroke center.
NIHSS score was not included in the primary analysis because NIHSS score may lie somewhere between OAC and outcomes, reflecting more rapid hematoma growth before hospital arrival.
Instead, the same analyses were replicated in patients with documented NIHSS scores while adjusting for NIHSS score along with other original covariates (143 340 patients [65.2%]) in the sensitivity analysis. The Glasgow Coma Scale (GCS) and ICH scores are commonly used risk stratification scales for ICH. Use of anticoagulation reversal treatment during the hospitalization may have been associated with outcomes. However, these data elements are available for only 86 800 patients (39.5%) hospitalized in centers that opt to report comprehensive stroke center data element. These variables have high missing rates (GCS score, 29.1%; ICH score, 38.1%; and reversal treatment, 72.2%). Therefore, these data were provided for information purpose only and were not included in the risk-adjustment model.
To evaluate the incremental risk of mortality and disability with the concomitant prior antiplatelet therapy, separate multivariable logistic regression models with generalized estimation equations were performed in each OAC group (FXa inhibitors, warfarin, and no OAC), respectively, with antiplatelet treatment category (no antiplatelet, single-antiplatelet agent, or dual-antiplatelet agents) as the independent variable. Patients who were not in any of these 3 antiplatelet groups were excluded from the analysis.
All statistical analyses were performed using SAS statistical software version 9.4 (SAS Institute).
All P values are 2-sided, with P < .05 considered significant. The original analyses were performed in December 2019.

Results
Baseline characteristics specific to anticoagulants used are shown in Table 1

JAMA Network Open | Neurology
Outcomes Associated With Oral Anticoagulant Use Among Patients Hospitalized With ICH were associated with greater odds of in-hospital mortality compared with no OAC (Table 2). Similarly,

Incremental Risk of Mortality and Disability With Concomitant Anticoagulant and Antiplatelet Therapy
Among patients with ICH with prior use of FXa inhibitors, there were no significant incremental risks of mortality or functional outcomes at discharge with either single-antiplatelet or dual-antiplatelet agents compared with patients without concomitant antiplatelet use ( Table 3). By contrast, in patients with prior use of warfarin, both single-antiplatelet and dual-antiplatelet agents were

JAMA Network Open | Neurology
Outcomes Associated With Oral Anticoagulant Use Among Patients Hospitalized With ICH associated with higher odds of in-hospital mortality (eg, dual-antiplatelet agent: aOR, 2.07; 95% CI, 1.72-2.50; P < .001) and death or discharge to hospice (eg, dual-antiplatelet agent: aOR, 1.86; 95% CI, 1.54-2.26; P < .001). In addition, these patients were less likely to be discharged home or have better mRS score at discharge, although some of these differences were not significant. Among patients without prior use of OAC, only dual-antiplatelet agents were associated with worse outcomes at discharge. In contrast, patients taking single-antiplatelet agents were more likely to be discharged to home, ambulate independently, and have mRS scores of 0 to 2 at discharge.

Discussion
In this cohort study of more than 200 000 patients with nontraumatic ICH, 4.2% were receiving FXa inhibitors and 9.8% were receiving warfarin before stroke. Although patients receiving FXa inhibitors were more likely to have favorable outcomes than those taking warfarin in terms of mortality and functional outcomes, prior use of FXa inhibitors was associated with increased odds of mortality and death or discharge to hospice compared with those not taking OAC, with more than 1 in 4 patients still dying in the hospital. These findings highlight the need to identify optimal strategies to care for these complex but increasingly common clinical challenges. In addition, concomitant antiplatelet use was common in patients with ICHs associated with use of FXa inhibitors and warfarin. Although both single-antiplatelet and dual-antiplatelet therapy were associated with increased odds of worse outcome among patients taking warfarin, such significant differences were not observed among patients with ICH taking concomitant FXa inhibitors and antiplatelet agents.
The current US and European guidelines recommend FXa inhibitors and direct thrombin inhibitor over warfarin for stroke prevention in high-risk patients with atrial fibrillation. 13,14 FXa

Limitations
There are limitations that should be kept in mind when interpreting these data. First, this was an observational analysis. Treatment selection and residual or unmeasured confounding could affect the validity of study findings and potentially account for the differences in risk-adjusted outcomes.
Because treatment was given before the ICH, it is impossible to randomize patients with different OAC regimens. Despite the observational nature, our study still provides important clinical insights to FXa inhibitor-associated ICH, especially given the low incidence rates in pivotal trials. Second, prior use of OAC was defined as documentation of patients taking an OAC within 7 days, yet timing and dose of last OAC intake were not available. It is possible that some patients may have received a lower dose or stopped taking OAC, which consequently influenced ICH outcomes. However, it could be argued that patients are more likely to experience an ischemic stroke rather than an ICH in such scenarios.
Third, some patients were excluded from the ambulatory status or functional outcome models because of missing outcomes. Although this approach may bias the results of this analysis, it is unlikely that physicians will report outcome measures differently according to the type of OAC taken before the stroke. Fourth, ICH volume and location and the presence of hematoma expansion were not available in the registry, which may explain the differences in outcomes between ICH patients with or without OAC use. Fifth, platelet transfusion may have influenced outcomes in patients taking concomitant antiplatelet and anticoagulant therapy. However, such data were not collected in the registry. Sixth, data are obtained from hospitals participating in the GWTG-Stroke program and may not be able to be extrapolated to patients treated in hospitals outside the registry or to patients in other countries. Notwithstanding these limitations, GWTG-Stroke is the largest stroke registry in the US, covering approximately three-fourths of the US population. Furthermore, ICH cases tend to be concentrated at or transferred to tertiary hospitals. Given the higher representation of high-volume, academic, and stroke centers in the GWTG-Stroke registry, the study population of this investigation is potentially more representative of OAC-associated ICH in the US.

Conclusions
In conclusion, FXa inhibitor-associated ICH remains a devastating complication of OAC therapy, with 27.0% of patients in this study dying in hospital. Although associated with better outcomes than warfarin, this study found that preceding use of FXa inhibitors was associated with higher risk of mortality and death or discharge to hospice than ICH without prior use of OACs. Further study is warranted to identify the best treatment strategies for FXa inhibitor-associated ICH.
serving as volunteer chair of the American Heart Association/American Stroke Association Get With the Guidelines (GWTG)-Stroke Clinical Work Group and as a consultant to Coverdell Grant. No other disclosures were reported.
Funding/Support: The GWTG-Stroke registry is organized by the American Heart Association and the American Stroke Association. GWTG-Stroke is sponsored, in part, by Novartis, Boehringer Ingelheim, Lilly, Novo Nordisk, Sanofi, AstraZeneca, Bayer, and Portola Pharmaceuticals. This analysis was supported by a grant from Portola Pharmaceuticals, now a subsidiary of Alexion Pharmaceuticals.

Role of the Funder/Sponsor:
The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Meeting Presentation: This manuscript was presented at the 2020 Virtual Conference of the European Stroke Organisation and World Stroke Organization; November 7-9, 2020.