Predictors of Hematoma Expansion and Response to Andexanet in Patients With Intracerebral Hemorrhage: Secondary Analyses of the ANNEXA-I Randomized Clinical Trial

Abstract

BACKGROUND:

Andexanet improves hemostatic efficacy in FXa (factor Xa) inhibitor–associated intracranial hemorrhage but carries a risk of thrombotic events. These secondary analyses of the ANNEXA-I trial (A Randomized Clinical Trial of Andexanet Alfa in Acute Intracranial Hemorrhage in Patients Receiving an Oral Factor Xa Inhibitor) examined determinants of hematoma expansion (HE) and the corresponding response to andexanet.

METHODS:

ANNEXA-I enrolled patients aged ≥18 years with acute FXa inhibitor–associated intracranial hemorrhage within 6 hours of onset between June 6, 2019, and May 27, 2023, at 131 sites in 23 countries. Participants were randomized (1:1) to andexanet or usual care. Among 530 participants, 459 (87%) had a qualifying intracerebral hemorrhage and adequate brain imaging for these analyses. The primary outcome was HE (≥12.5-mL or ≥35% increase in baseline volume) at 12 hours. We report differences in the proportion of HE between treatment groups stratified by variables significantly associated with HE in regression models.

RESULTS:

HE occurred in 149 of 459 participants (32.5%). Symptom onset-to-treatment time (adjusted odds ratio per hour, 0.73 [95% CI, 0.63–0.85]), baseline hematoma volume (adjusted odds ratio per 10 mL, 1.11 [95% CI, 1.01–1.23]), and diastolic blood pressure (adjusted odds ratio per 10 mm Hg, 1.15 [95% CI, 1.02–1.29]) were associated with HE but not thrombotic events. In a separate multivariable model replacing volume and onset-to-treatment time with prescan hematoma growth rate, growth rate was also associated with HE (adjusted odds ratio per mL/h, 1.02 [95% CI, 1.01–1.04]). Patients in the highest risk quartiles (baseline volume >22.4 mL, growth rate >11.4 mL/h, diastolic blood pressure >95.0 mm Hg, and onset-to-treatment time ≤3.3 hours) had ≈50% to 60% absolute risk of HE with usual care. Numerically greater absolute risk reductions with andexanet (≈25%; number needed to treat: 4) were observed in the highest quartiles of baseline volume and growth rate.

CONCLUSIONS:

Shorter onset-to-treatment time, larger baseline hematoma volume, higher diastolic blood pressure, and higher prescan hematoma growth rate predict HE but not thrombotic events in FXa inhibitor–associated intracerebral hemorrhage. Andexanet benefit is observed across these ranges and may be amplified through patient selection using these metrics.

REGISTRATION:

URL: https://www.clinicaltrials.gov; Unique identifier: NCT03661528.

Treatment with FXa (factor Xa) inhibitors is associated with a high risk of hematoma expansion (HE) in patients who experience intracerebral hemorrhage (ICH).¹ Final hematoma volume is a major determinant of the extent of primary and secondary injury after ICH.² By extension, HE is a significant predictor of poor outcome in patients with acute ICH.^3,4 It is estimated that for each 1-mL and 10% increase in baseline hematoma volume, patients are 6% and 16% more likely, respectively, to experience a 1-point increase in the modified Rankin Scale score, indicating greater disability, at 90 days.⁵ Reducing hematoma growth is thus an important mechanistic target to improve clinical outcomes in treatment trials of patients with acute ICH.

Andexanet rapidly reverses the anticoagulant effects of FXa inhibitors.^6,7 In the randomized controlled ANNEXA-I trial (A Randomized Clinical Trial of Andexanet Alfa in Acute Intracranial Hemorrhage in Patients Receiving an Oral Factor Xa Inhibitor), andexanet increased hemostatic efficacy compared with usual care (adjusted difference per 100 patients, 13.4 [95% CI, 4.6–22.2]).⁸ However, this benefit was partially offset by increased thrombotic events (TE) in patients receiving andexanet (adjusted difference per 100 patients, 4.6 [95% CI, 0.1–9.2]), underscoring the importance of patient selection for optimal treatment outcomes.

There are several clinical variables that are established to increase the risk of HE in patients presenting to the hospital with an acute ICH and that may help identify patients more likely to benefit from andexanet treatment. These variables include shorter time from symptom onset to presentation, larger baseline ICH volumes, and the potency of antithrombotic treatment.⁹ Two of these variables, time from symptom onset and baseline ICH volume, can be combined into the metric of prescan hematoma growth rate (also termed ultraearly hematoma growth).¹⁰ Prescan hematoma growth rate is strongly correlated with risk of HE, and higher growth rates have been reported to identify patients with ICH who have greater benefit from acute intensive blood pressure lowering.^10–12

We hypothesized that patients with FXa inhibitor–associated ICH who are at increased risk for HE^9,10 are likely to benefit most from andexanet, and we aimed to characterize predictors of HE and response to treatment in ANNEXA-I participants.

Methods

Data Sharing Statement

Data will not be shared within 1 year of publication. Please contact the corresponding authors with any proposals for analysis, which will be reviewed by the study Publication Committee for novelty, feasibility, and scientific value. Data underlying the findings described in this manuscript may be obtained after this period in accordance with AstraZeneca’s data sharing policy described at https://astrazenecagrouptrials.pharmacm.com/ST/Submission/Disclosure. Data for studies directly listed on Vivli can be requested through Vivli at www.vivli.org. Data for studies not listed on Vivli could be requested through Vivli at https://vivli.org/members/enquiries-about-studies-not-listed-on-the-vivli-platform/. AstraZeneca Vivli member page is also available outlining further details: https://vivli.org/ourmember/astrazeneca/

Study Design

ANNEXa-I was a phase IV, multicenter, prospective, randomized, open-label, blinded-end point trial in patients with acute ICH treated with FXa inhibitors. The design and main results of the ANNEXA-I trial have been previously reported.⁸ Approval for the trial protocol was obtained from pertinent health authorities, as well as institutional review boards, for all participating sites. Written informed or emergency consent was obtained from participants or their legally authorized representative. Further details are found in the supplement. We performed exploratory secondary analyses of ANNEXa-I focused on determining predictors of HE (observational analyses) and potential subgroup heterogeneity in response to randomized intervention (andexanet compared with usual care; preserving randomization) according to identified predictors of HE in trial participants with index ICH.

This trial was coordinated by the Population Health Research Institute, a joint institute of McMaster University and Hamilton Health Sciences, and was initially funded by Portola, which was purchased by Alexion AstraZeneca Rare Disease. Portola, followed by Alexion, was the trial sponsor and provided the andexanet free of charge. Enrollment occurred at 131 sites in 23 countries. The statistical analysis plan for these secondary analyses was developed by study authors (A.S., S.C., M.S., and L.X.) with input from the Steering Committee and sponsor representatives. The statistical analysis was performed by an author (L.X.) who is a statistician at the Population Health Research Institute. Sponsor approval was not required for submission.

Study Participants

The trial initially enrolled patients with acute FXa inhibitor–associated intracranial hemorrhage with an intracranial bleed volume of up to ≤60 mL, all presenting for a baseline imaging scan within 12 hours of symptom onset. After protocol amendment, eligibility criteria were refined to include only patients with the following: ICH (intraparenchymal or intraventricular bleeds of ≥0.5 to ≤60 mL) as the primary site of bleeding, a maximum National Institutes of Health Stroke Scale (NIHSS) score of 35, and a baseline scan obtained within 6 hours of symptom onset. Hemorrhages needed to be visually confirmed on neuroimaging within 2 hours before randomization, and the last dose of FXa inhibitor had to be taken within 15 hours before randomization. Exclusion criteria included a Glasgow Coma Scale score of <7 at the time of randomization, planned surgery to evacuate the hematoma within 12 hours, or a history of TE within the past 2 weeks. This study followed CONSORT guidelines (Consolidated Standards of Reporting Trials; Supplemental Material).¹³

Intervention

Patients were randomly assigned in a 1:1 ratio to receive either andexanet or usual care, with stratification based on the intention to use prothrombin complex concentrate if allocated to usual care and time elapsed from symptom onset to baseline scan (<3 versus ≥3 hours).

The dosage of andexanet was individualized according to the type of FXa inhibitor used, in conjunction with the amount and timing of the last dose. Usual care, under the discretion of local physicians, excluded andexanet but could include prothrombin complex concentrate.⁸

Baseline Characteristics

Baseline demographics, medical history, medication use, stroke severity, and hematoma characteristics were collected at study entry. Prescan hematoma growth rate (mL/h) was calculated as the baseline hematoma volume divided by the time from symptom onset to the baseline scan.¹⁰

Outcomes

The primary efficacy end point of ANNEXa-I was hemostatic efficacy, defined by expansion of the hematoma volume by 35% or less at 12 hours after baseline, an increase in the NIHSS score of <7 points, and no receipt of rescue therapy between 3 and 12 hours. An end point adjudication committee masked to treatment assignment reviewed all potential TE. All outcomes were assessed from the date and time of randomization. To determine the change in hematoma volume, a volumetric analysis of the hemorrhage using Quantomo software (Cybertrials Inc, Calgary, Canada)¹⁴ was performed at a dedicated imaging core laboratory by trained physicians who were masked to treatment assignment.

Current Secondary Analyses

For these exploratory secondary analyses, patients enrolled into ANNEXa-I with a qualifying ICH (excluding patients with subdural, epidural, or subarachnoid hemorrhages) and who had an interpretable baseline imaging scan and subsequent scan at 12 hours after randomization were eligible. The primary outcome for these analyses is HE, defined as an increase in hematoma volume of ≥35% or ≥12.5 mL between brain imaging obtained at baseline and 12 hours after randomization. The 12.5-mL severe absolute hematoma volume growth threshold was included in the primary end point due to increasing evidence indicating its strong association with a poor outcome on the modified Rankin Scale score at 90 days in patients with acute ICH.^3,4 The safety outcome was TE within 30 days, where patients who died within 30 days without a TE were considered as not having had a TE. A time-to-event sensitivity analysis, censoring for death, was also performed.

Statistical Analysis

Baseline characteristics were summarized and compared between participants with and without HE. Continuous variables were summarized as mean±SD or median (interquartile range [IQR]) and compared between groups by using either a t test for the means or a Wilcoxon rank-sum test for the medians. Categorical variables were reported as counts (percentages) and compared between groups using a Pearson χ² test. Missing values in time from symptom onset to treatment (n=33) were imputed with the median value of this variable in the corresponding treatment arm.

Multivariate logistic regression was performed in an observational manner using the dichotomized HE as the outcome and risk factors identified from the baseline univariate analysis (P<0.10), along with treatment arm, as independent variables. Odds ratios (ORs) and 95% CIs obtained from the logistic regressions were used to assess the association between the risk factors and the outcome of HE. Because the indication for high-dose andexanet is highly correlated with baseline anti-FXa levels, only the baseline anti-FXa level variable (but not the high-dose andexanet indication variable) was included in the model to avoid colinearity. Similarly, as ICH volume is in the causal pathway of—and colinear with—higher NIHSS scores and has greater biological importance for predicting risk of further ICH growth, ICH volume was included in the model, and NIHSS score was not. Furthermore, the anti-FXa level has greater biological plausibility for a causal relationship with HE. To avoid colinearity between prescan hematoma growth rate (mL/h) and baseline ICH volume (mL) or time from symptom onset to treatment (hours), these variables were separately considered in 2 different regression models. The final models were subsequently applied to the safety outcome of TE within 30 days.

For continuous variables that were identified as significantly associated with HE in the above regression model, the difference in the proportion of participants with HE between the andexanet and usual care groups across the quartiles of these variables, along with the 95% CIs, was estimated using the Wald confidence limit method. Predicted probability of HE, stratified by treatment arm at various levels for each independent variable, was additionally estimated from fitting each of the corresponding logistic regression models in which 2 independent variables were included (the randomized treatment arm and the baseline clinical variable; Figure 1). These analyses preserved randomization and followed the intention-to-treat paradigm. The number needed to treat (NNT) and number needed to harm were calculated by dividing 100 by the absolute change in HE and TE, respectively, with andexanet per 100 patients treated.

Figure 1.

Independent predictors of hematoma expansion (HE) and probability of HE. Predicted probability of HE (defined as an increase in baseline hematoma volume of ≥12.5 mL or ≥35%) by (A) baseline volume (variables in the model included treatment and baseline hematoma volume [mL]); (B) symptom onset–to-treatment time (variables in the model included treatment and time from symptom onset to treatment [h]); (C) baseline anti-FXa (factor Xa) levels (variables in the model included treatment and baseline anti-FXa level [ng/mL]); (D) prescan growth rate (variables in the model included treatment and prescan hematoma growth rate [mL/h]); and (E) diastolic blood pressure (DBP; variables in the model included treatment and baseline DBP [mm Hg]).

P values are 2-sided. Statistical significance was accepted at the.05 level. Statistical analyses were performed using SAS software (release 9.4).

Results

Of 530 patients enrolled between June 6, 2019, and May 27, 2023, 459 had baseline ICH or intraventricular hemorrhage and adequate baseline and follow-up imaging and were included in these analyses (Figure S1). The mean (±SD) age was 79.1 (±8.3) years, and 46.2% (n=212) of patients were female. The majority (n=455 [99.1%]) of patients had an intraparenchymal hemorrhage, and only 4 had primary intraventricular hemorrhage (0.9%). In 30 (6.5%) cases, the ICH was preceded by trauma. The median (IQR) baseline hematoma volume was 9.2 (IQR, 3.5–22.4) mL, and the NIHSS score was 9.0 (IQR, 5.0–16.0). The median (IQR) time from symptom onset to baseline scan was 2.3 (IQR, 1.4–3.8) hours, and the median time from symptom onset to treatment was 4.0 (IQR, 3.2–5.5) hours. Most patients (86.4%) assigned to the usual care group received prothrombin complex concentrate.

In a univariate analysis, higher baseline NIHSS score, anti-FXa level, hematoma volume, and prescan hematoma growth rate were associated with HE (Table S1). Patients with HE also had higher baseline diastolic blood pressure (DBP), shorter time from symptom onset to treatment, and a greater likelihood to be eligible for high-dose andexanet.

In the first multivariable regression model (Table 1; model 1), shorter time from symptom onset to treatment (OR per 1-hour increase, 0.73 [95% CI, 0.63–0.85]; P<0.001), larger baseline hematoma volume (OR per 10 mL, 1.11 [95% CI, 1.01–1.23]; P=0.04), and higher DBP (OR per 10 mm Hg, 1.15 [95% CI, 1.02–1.29]; P=0.02) were independently associated with HE. In the second model, prescan hematoma growth rate (OR per mL/h, 1.03 [95% CI, 1.01–1.04]; P<0.001) and DBP (OR per 10 mm Hg, 1.18 [95% CI, 1.05–1.33]; P<0.01) were associated with HE.

Table 1.

Multivariable Models for Risk of HE* at 12 Hours

The effect of andexanet versus usual care on HE was consistent across the range of hematoma volumes, time to treatment, and hematoma growth rates (Figure 1).

The rate of HE in the usual care arm increased from 27.9% to 53.6% between participants within the first (≤3.5 mL) and fourth (>22.4 mL) quartiles of baseline ICH volume. Corresponding rates of HE between the lowest and highest quartiles were 48.8% (≤3.3 hours) and 24.6% (>5.4 hours), respectively, for time from symptom onset to treatment, 20.0% (≤1.2 mL/h) and 59.3% (>11.4 mL/h) for prescan hematoma growth rate, and 29.5% (≤73.0 mm Hg) and 51.8% (>95.0 mm Hg) for baseline DBP (Table 2). With andexanet, the NNT to prevent HE in this total sample of participants was 7. In patients within the highest risk quartiles of baseline hematoma volume (>22.4 mL), time from symptom onset to treatment (≤3.3 hours), prescan hematoma growth rate (>11.4 mL/h), and DBP (>95 mm Hg), the NNTs were 4, 7, 4, and 6, respectively (Table 2). Baseline hematoma volume, time from symptom onset to treatment, DBP, and prescan hematoma growth rate were not predictive of TE (Table 3). A sensitivity time-to- event analysis for this end point using the Cox proportional hazard model, in which patients who died within 30 days without a TE were censored, yielded similar results.

Table 2.

HE* Rates at 12 Hours, Stratified by Predictors of HE and Treatment Group

Table 3.

Multivariable Models for the Odds of TE Within 30 Days

The number needed to harm for an excess TE with andexanet compared with usual care was ≈26 (the increase with andexanet per 100 patients treated was 3.9 [95% CI, –1.2 to 8.9]) in this sample of patients with ICH.

Discussion

In these analyses, we show that there are 3 primary factors that independently predict the risk of HE: higher baseline hematoma volume, shorter time from symptom onset to treatment (or scan), and higher baseline DBP. Two of these, volume and time, can be combined in the metric of prescan hematoma growth rate. These findings are consistent with previous analyses of predictors of HE.^9,10 We have also shown that the treatment effect of andexanet in preventing HE is highly consistent with similar relative treatment effects observed in patients predicted to be at low or high risk of expansion. As there is an urgent need to restore normal coagulation in patients with FXa inhibitor–associated acute ICH, these variables can be used to rapidly triage benefit risk (Figure 2).

Figure 2.

Proposed 11 to 1 rule for individualized andexanet treatment decisions in patients with acute FXa (factor Xa) inhibitor–associated intracerebral hemorrhage (ICH). Prescan hematoma growth rate quartiles 2 and 3 from Table 3 were combined for the intermediate category. The absolute risk of thrombotic event (TE) with andexanet compared with usual care was greater in patients with a prior history of ischemic stroke or myocardial infarction (MI; 10.1% with andexanet vs 8.1% with usual care; adjusted difference per 100 patients treated, 2.0 [95% CI, −4.3 to 8.3]; number needed to harm [NNH], 50) compared with those without a prior history of stroke or MI (10.8% with andexanet vs 2.7% with usual care; adjusted difference per 100 patients treated, 8.1 [95% CI, −0.3 to 16.5]; NNH, 12). The small total number of TE negates further stratification of NNH by prescan hematoma growth rate, but there was no association between prescan hematoma growth rate and the risk of TE observed. Elevated blood pressure additionally increases the risk of hematoma expansion (HE) and should be considered in individualized treatment decisions. NNT indicates number needed to treat.

Absolute risk reductions in HE and corresponding NNT were robust across all quartiles of these variables (NNT was generally <10), with the greatest absolute risk reduction being observed in patients within the highest quartiles of baseline hematoma volume (>22.4 mL) and prescan hematoma growth rate (>11.4 mL/h), where the NNT to prevent HE was 4. Similarly, prescan hematoma growth rate has been previously demonstrated to identify patients with acute ICH who benefit the most from intensive blood pressure lowering in the ATACH-2 trial.¹² However, due to the exploratory nature of our analyses and insufficient power, we are unable to demonstrate statistically significant treatment interactions for greater absolute benefit in patients at the highest risk of HE, and the numerical trends suggesting this should be considered hypothesis-generating. Moreover, due to the limited number of TE and the short follow-up of 30 days, our analyses may have been underpowered to detect associations between TE and these variables.

These data suggest that we may be able to optimize the benefit-to-risk ratio of andexanet treatment through patient selection. Rates of HE with usual care were ≈50% in patients within the highest quartiles of baseline hematoma volume and DBP, as well as those within the lowest quartile of time from symptom onset to treatment. The rate of HE was greatest and nearing 60% (59.3%) in patients within the highest quartile of prescan hematoma growth rate. Notably, the rate of HE with usual care remained clinically significant (ranging between 20% and 30%), and the NNT to reduce HE remained robust (<10), even in patients within the lowest risk quartiles of these variables. None of these variables was associated with TE. Thus, by selecting patients at a higher risk of HE using these variables (particularly prescan hematoma growth rate), treatment effects for HE may be maximized while maintaining the same level of thrombotic risk.

The ANNEXA-I trial was designed to assess hemostatic efficacy with andexanet compared with usual care. The 30-day duration of follow-up in the trial was too brief to detect differences in disability between the treatment arms, and the study was not powered for death or disability.⁸ However, the requisite thresholds of hematoma growth of ≥35% and ≥12.5 mL that were used in our definition of HE have been determined to independently predict poor outcome in patients with acute ICH, with positive predictive values of 70% and 80%, respectively, for a modified Rankin Scale score of 4 to 6 at 90 days.^3,4

Based on these analyses, a simple model that uses prescan hematoma growth rate, DBP, and prior history of arterial TE could be considered for individualized treatment decisions (Figure 2). When balancing the observed NNT for reducing HE and the number needed to harm for TE, our estimates favor treatment with andexanet over usual care in patients presenting with FXa inhibitor–associated ICH who fulfill the ANNEXA-I trial eligibility, except in patients with prior arterial TE who present with low-to-intermediate prescan hematoma growth rates (<11 mL/h; Figure 2). These estimates assume similar clinical consequences from HE and TE, as previously reported in the ANNEXa-I data set.¹⁵

Our findings highlight the importance of blood pressure as a predictor of expansion in FXa inhibitor–associated ICH and further support the importance of bundles of care in patients with acute ICH that prioritize timely normalization of coagulation and lowering of blood pressure, among other acute interventions.^16–18 Prior evaluations of HE in patients with acute ICH have identified systolic blood pressure as a stronger determinant of HE than DBP. The biological underpinning of our observation that DBP is a stronger predictor of HE in FXa inhibitor–associated ICH is uncertain. One could hypothesize that in circumstances of impaired coagulation, trough (DBP) rather than peak (systolic blood pressure) vessel transmural pressures are of greater importance for hemostatic clot formation and stability. In this regard, it is interesting that the ROCKET-AF trial (Rivaroxaban Once Daily, Oral, Direct Factor Xa Inhibition Compared With Vitamin K Antagonism for Prevention of Stroke and Embolism Trial in Atrial Fibrillation) investigators also identified DBP to be a stronger determinant of anticoagulant-associated ICH incidence in patients with atrial fibrillation treated with rivaroxaban or warfarin. Additional planned secondary analyses of ANNEXA-I will further explore the relationship between systolic blood pressure, DBP, and HE, as well as potential treatment effect modification between acute blood pressure control and andexanet treatment in achieving hemostatic efficacy.

To conclude, higher prescan hematoma growth rate, a composite of time from symptom onset to baseline scan and baseline hematoma volume, and higher DBP predict HE, but not TE, in patients with FXa inhibitor–associated ICH. The benefit of andexanet is robust across the range of these variables and may be further amplified by patient selection using these metrics. These results can provide preliminary guidance for clinical practice and inform the design of future trials.

ARTICLE INFORMATION

Acknowledgments

Additional Information: Coauthor Stuart J. Connolly, MD, died June 2, 2024.

Disclosures

Dr Shoamanesh has received research grants/contracts from AstraZeneca, Bayer, the Canadian Institutes of Health Research, Daiichi Sankyo Company, the Heart and Stroke Foundation of Canada, the National Institutes of Health, Octapharma USA Inc, and Servier Affaires Medicales; has served on advisory boards for AstraZeneca, Bayer, Daiichi Sankyo Company, and Takeda Pharmaceutical Company; and has served on a data and safety monitoring board for Bayer. Dr Connolly has served as a consultant for AstraZeneca, AstraZeneca AB, AtriCure Inc, Bayer, Bristol Myers Squibb, Daiichi Sankyo Company, Javelin Ventures, and Pfizer Inc. Dr Demchuk has received honoraria for lectures from AstraZeneca Canada and Novo Nordisk AS; has participated on advisory boards for Hoffmann-La Roche Limited and NovaSignal; has served on data and safety monitoring committees for Lumosa and Philips; holds stock and a patent (Patent number: 10373718; 2023039659; 20230277151) for Circle NVI; has received grants/contracts from OroSure Technologies and SFJ Pharmaceuticals. Dr Seiffge has received grants/contracts from and has served as a consultant for AstraZeneca. Dr Sandset has served as a national study coordinator and steering committee member for AstraZeneca, Bayer, and Bristol Myers Squibb; and has received lecture honoraria for Boston Scientific Corporation and has received honoraria as a national coordinator from PORTOLA PHARMACEUTICALS, LLC. Dr Tsigoulis has received grants/contracts from AbbVie, Allergan, Amicus Therapeutics Inc, Bayer, Boehringer Ingelheim, Genesis Pharma, Medtronic, Merck, Novartis, Roche, Shire, and Teva Pharmaceutical Industries. Dr Christensen has served as a consultant on trials for Alexion Pharmaceuticals and Bayer Healthcare and has served as a speaker for Boehringer Ingelheim, Bristol Myers Squibb, Daiichi Sankyo Company, and Medtronic. Dr Beyer-Westendorf has received honoraria for lectures and advisory boards from Alexion Pharmaceuticals, AstraZeneca, Bayer, Daiichi Sankyo Company, and Daiichi Sankyo Europe GmbH; has received grants/contracts from Alexion Pharmaceuticals, Bayer, Daiichi Sankyo Company, and Pfizer Canada Inc; has received support for travel from Alexion Pharmaceuticals, AstraZeneca, Bayer, and Daiichi Sankyo Company; and has served on an end point review committee for Takada Development Center Americas Inc. Dr Coutinho has received research grants/contracts from AstraZeneca, Bayer, and Medtronic; has served as a consultant or steering committee member for Boehringer Ingelheim and Portola Pharmaceuticals, LLC; and is a founder and shareholder of TrianecT BV. Dr Amarenco has served on trial executive or steering committees for AstraZeneca, Bayer Healthcare, Bristol Myers Squibb, Kowa Company, Ltd, and Pfizer; has served on a data and safety monitoring board for FibroGen; and has served on an advisory board or as a speaker for Amgen and Novartis. Dr Veltkamp has received research grants/contracts from Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Daiichi Sankyo Company, and Medtronic; has served on advisory boards for AstraZeneca and Bristol Myers Squibb; has served as a steering committee member for Daiichi Sankyo Company and Javelin Ventures, and has served on an end point review committee for Portola Pharmaceuticals. Dr Middeldorp has received consulting fees from Norgine, VarmX, Pfizer, and AstraZeneca and has received fees from Bayer for data and safety monitoring services. Dr Zini has received funding for speaker honoraria and consulting fees from Boehringer Ingelheim, AstraZeneca, Daiichi Sankyo, and CSL Behring, and for scientific advisory boards from Bayer and AstraZeneca. Dr Himmelmann, Dr Ladenvall, and Dr Knutsson are employees of AstraZeneca. Dr Crowther has served as a consultant for AstraZeneca, Bayer, Eversana, Hemostasis Reference Laboratory, Pfizer Canada Inc, and Syneos Health. Dr Sharma has received research grants/contracts from Bayer and Bristol Myers Squibb; has served as a consultant for Bayer; has received consulting fees from Regeneron, Novartis, AstraZeneca, Anthos Pharmaceuticals, and Janssen Global Services, LLC. The other authors report no conflicts.

Supplemental Material

Table S1

Figure S1

Protocol

CONSORT Checklist

Funding Statement

The ANNEXA-I trial (A Randomized Clinical Trial of Andexanet Alfa in Acute Intracranial Hemorrhage in Patients Receiving an Oral Factor Xa Inhibitor) was supported by Alexion AstraZeneca Rare Disease, AstraZeneca Biopharmaceuticals, and, previously, Portola and Alexion Pharmaceuticals.