CADTH Reimbursement Review

Andexanet Alfa (Ondexxya)

Sponsor: AstraZeneca Canada Inc.

Therapeutic area: Reversal of FXa inhibitor anticoagulant effects

This multi-part report includes:

Clinical Review

Pharmacoeconomic Review

Stakeholder Input

Clinical Review

Executive Summary

An overview of the submission details for the drug under review is provided in Table 1.

Table 1: Background Information of Application Submitted for Review

FXa = factor Xa; NOC/c = Notice of Compliance with Conditions.

Introduction

Direct oral anticoagulants (DOACs), such as apixaban and rivaroxaban, are a class of anticoagulant commonly used for the prevention of stroke and embolisms in atrial fibrillation (AF) and venous thromboembolism (VTE).^2,3 However, serious bleeding is a major risk associated with anticoagulant treatment and could manifest as gastrointestinal (GI) bleeding (30% to 45% risk), intracranial hemorrhage (ICH) (10% to 25%), and/or bleeding at a critical site (intraocular, pericardial, intraspinal, intra-articular).^4-7 It is estimated that, in Canada, approximately 900,000 patients are on rivaroxaban or apixaban with an increase year over year of approximately 4%,⁸ and the annual risk of major bleeding is estimated to range from 2% to 4% in these patients.^6,7,9 Factor Xa (FXa) inhibitor–related major bleeding events are associated with an increased risk of death; 30-day mortality estimates after FXa inhibitor–related major bleeding range from 9% to 45%.^5,10-15

Reversal drugs for FXa inhibitors support the management of FXa inhibitor–related bleeding, in addition to cessation of anticoagulant therapy, initiation of resuscitation, application of local hemostatic measures, and administration of supporting therapy with blood products. To date, there is no available antidote in Canada to specifically reverse the anticoagulant effect of FXa inhibitors in the event of major bleeding. The usual care in the management of major bleeding associated with FXa inhibitors is the off-label use of prothrombin complex concentrate (PCC), a plasma-derived product that provides nonspecific supplementation of exogenous coagulation factors, although it does not directly reverse FXa inhibitor activity. Clinical trial evidence for PCC in treating FXa inhibitor–related bleeding is lacking and its use is primarily informed by the results of observational studies.¹⁶

The objective of this report is to review and critically appraise the evidence submitted by the sponsor on the beneficial and harmful effects of andexanet alfa in adult patients treated with FXa inhibitors (rivaroxaban or apixaban) when rapid reversal of anticoagulation is needed due to acute major bleeding, including life-threatening bleeds.

Stakeholder Perspectives

The information in this section is a summary of input provided by the patient and clinician groups that responded to CADTH’s call for input and from the clinical experts consulted by CADTH for the purpose of this review.

Patient Input

Two patient groups, the Canadian Venous Thromboembolism Research Network (CanVECTOR) and the HeartLife Foundation, submitted input for this review. The input by CanVECTOR was informed by 33 interviews with patients with lived experience with VTE (deep vein thrombosis [DVT], n = 15; pulmonary embolism [PE], n = 6; both DVT and PE, n = 12) and varied treatment experience, including with warfarin, DOACs, and low-molecular-weight heparin.

According to the input by CanVECTOR, patients receiving treatment for VTE have to find a balance between the risk of another clot and managing the side effects of treatment. Patients expressed that bleeding is the most concerning potential side effect of VTE treatments and they may live with a fear of bleeding that can impact life choices and daily activities, quality of life, and mental health. The input noted that a variety of personal factors affect a patient’s treatment preferences. Many patients described the burden of treatment with warfarin (the inconvenience of blood monitoring, restrictions on diet and alcohol intake), or with low-molecular-weight heparin injections (pain, bruising, discomfort with self-injections) and some had later switched to a DOAC. The input noted fewer burdens with DOACs, but the lack of a reversal drug was mentioned by a few patients, for example, “However much I have my misgivings on Warfarin, it had one benefit. It was reversible. It had all the other things that were not attractive to it, but if you could make some of those things more attractive, or certainly if the new drug that would come on the market to replace or be an alternative to apixaban, if it had a reversibility aspect to it, I think would be attractive.” The HeartLife Foundation noted that the use of an FXa inhibitor can increase the risk of bleeding; as well, patients with heart failure may require invasive surgeries, which put them at risk of surgical bleeds. The group expressed a need for a treatment that can rapidly and effectively reverse the anticoagulant effect of an FXa inhibitor to prevent further bleeding and ensure the best possible outcome in patients who require urgent surgery or are experiencing life-threatening bleeds. This was echoed by the input from CanVECTOR, which highlighted that patients were concerned that the absence of a reversal drug for DOACs could prevent them from undergoing a surgery in case of an emergency: “If an accident happened and the surgery is almost immediate, a reversal agent can help me … [go] into the emergency surgery safely.” CanVECTOR also highlighted that some patients with a high risk of another clot will be prescribed blood thinners for the rest of their lives, and the experience could be daunting for a young person with decades of treatment ahead.

Clinician Input

Input From the Clinical Experts Consulted by CADTH

The clinical experts consulted by CADTH noted there is an unmet need for a specific reversal drug for FXa inhibitor–related bleeding that can achieve rapid reversal of anticoagulation in patients with time-sensitive major bleeding. They also noted that PCC and blood products are not universally available in all clinical settings. As well, the risk of thrombosis carried by existing treatments is not well-characterized. The clinical experts expected that andexanet alfa would have the same place in therapy as PCC in the treatment of FXa inhibitor–related major bleeding. However, they were uncertain whether andexanet alfa or PCC should be the preferred therapy, and suggested this should be based on clinical presentation.

The clinical experts noted that patients receiving an FXa inhibitor who have major bleeding are potential candidates for andexanet alfa treatment. As well, patients who require emergency surgery while receiving an FXa inhibitor are also reasonable candidates in the clinical experts’ opinion, although the use of andexanet alfa was not studied in this patient population. The clinical experts noted that treatment candidates are identified based on clinical judgment with consideration of the patient’s clinical status (severity, bleed location, and response to nonspecific supportive measures) primarily, along with laboratory markers; clinicians in Canada most commonly use the International Society on Thrombosis and Hemostasis (ISTH) definition of major bleed. The clinical experts noted that patients with a minor bleed would not be suitable for andexanet alfa treatment since minor bleeds are expected to resolve on their own without the need for a reversal drug and with minimal patient harm.

The clinical experts noted that, in general, response to treatment is assessed based on hemostatic stability, hemostatic control (i.e., cessation of bleeding), need for blood transfusion, survival, and health-related quality of life (HRQoL). They considered it reasonable to allow prescribing by clinicians who have expertise in the management of acute major bleeds in a hospital setting. They noted that redosing with andexanet alfa is expected to be rare and should be done only in exceptional circumstances under the guidance of a transfusion medicine specialist, thrombosis physician, or hematologist, given the increased risk of thrombosis.

Clinician Group Input

The clinician group input was obtained from 5 clinician groups: Thrombosis Canada (represented by 2 clinicians), members of the Thrombosis and Anticoagulation Team at Dalhousie University (represented by 3 clinicians), faculty members at McMaster University in hematology and/or thromboembolism (represented by 5 clinicians), the Canadian Stroke Consortium (represented by 5 clinicians), and a journal club comprising local emergency department physicians in Peel Region (represented by 5 clinicians).

The clinician groups noted that current reversal treatments for bleeding associated with DOACs in patients who are receiving oral FXa inhibitors are nonspecific. The groups noted that PCCs, which are currently the treatment of choice for oral FXa reversal, have no effect on eliminating the active inhibitory action of oral FXa inhibitors and their use is considered off-label in the absence of robust clinical trial data supporting their efficacy or safety in this population. They noted that PCCs may promote a prothrombotic state and could potentially not be used safely in patients who have a history of heparin-induced thrombocytopenia or thrombosis, since a small amount of heparin may be present in PCC.

The clinician groups noted that andexanet alfa, being the only specific reversal drug available to patients who are on an FXa inhibitor, would be used as a first-line treatment in patients who require urgent anticoagulant reversal in the setting of serious, life-threatening bleeding or need for urgent surgery. One clinician group noted that PCCs would likely be used for an “average” patient who needs DOAC reversal, whereas andexanet alfa would be used in select patients with life-threatening bleeding that does not respond to supportive management (i.e., fluids, packed red blood cells [RBCs]), or patients with critical site bleeding (intracranial, spinal, pericardial) or in need of emergency (within 6 to 8 hours) or urgent (within 12 to 24 hours) surgery. They noted there is no established threshold for clinically significant hemostatic impairment and that most treatment centres do not have FXa inhibitor drug assays available, so treatment is usually considered based on the timing of the last dose, drug half-life, patient’s kidney or liver function, and examination findings (e.g., hypotension), radiographic findings (e.g., CT scans), and clinician judgment. They also noted that the patients who would be less suitable would be those who: last took their dose of FXa inhibitor more than 1 to 2 days ago (in the presence of normal renal function), have bleeding that is not life-threatening, or who can have their surgery delayed for 1 to 2 days after their last dose of an FXa inhibitor.

Outcomes to assess response to treatment deemed important by the clinician groups included: achievement of excellent or good hemostatic efficacy, thrombotic events, decreased mortality, decrease in hemoglobin or hematocrit by less than 20% compared with baseline, improvement of symptoms, reduction in hematoma expansion in ICH, survival to discharge, disability score on discharge, and measurement of anti-FXa levels before and after drug administration (not widely available). They noted that length of hospital stay may also be a surrogate marker. The clinical groups noted that criteria for discontinuation would include unexpected allergic or infusion reactions or thromboembolic events.

The clinical groups also supported the use of andexanet alfa in the hospital setting, including the emergency department, critical care unit, operating room, tertiary trauma centre, and stroke or neurosurgical referral centre, and administered by anesthesiologists or by surgical or other specialists such as those in emergency or internal medicine, depending on the site of the bleeding (e.g., neurologist or neurosurgeon for ICH, gastroenterologists for GI bleeding).

Drug Program Input

The drug programs expressed interest in understanding whether PCC is an appropriate comparator for andexanet alfa and the comparative efficacy and safety of andexanet alfa versus PCC, based on the submitted evidence. They also sought clinical expert input on the treatment approach for patients whose condition is not adequately managed by currently available treatments.

Other questions from the drug programs relate to: the initiation criteria, appropriateness of using andexanet alfa for edoxaban- and enoxaparin-related bleeding, sequencing of andexanet alfa relative to PCC, redosing, implementation issues related to idarucizumab, treatment eligibility in specific patient populations (pediatric patients, pregnant patients, and patients requiring emergency surgery in the absence of a life-threatening bleed), time required for reconstitution and administration of andexanet alfa relative to PCC, overdose management, requirement for laboratory tests for follow-up assessments, the logistics of dose assessment and accessing of drug products from transfusion medicine laboratories, assumptions of the pharmacoeconomic model, and drug utilization at the hospital level.

The clinical experts consulted by CADTH provided advice on the potential implementation issues raised by the drug programs. Refer to Table 4 for more details.

Clinical Evidence

Pivotal Studies and Randomized Controlled Trial Evidence

Description of Studies

One pivotal phase IIIb and IV, open-label, single-arm trial, ANNEXA-4 (N = 477),¹⁷ was included in the sponsor’s submission. The study assessed change in anti-FXa activity from baseline and whether hemostatic efficacy was reached with andexanet alfa treatment in adult patients with acute major bleeding while receiving FXa inhibitor treatment. Patients with features of severe ICH (i.e., Glasgow Coma Scale [GCS] score of less than 7 or an intracerebral hematoma volume above 60 cc) and those with an expected survival of less than 1 month were excluded from the study. Percent change from baseline in anti-FXa activity to on-treatment nadir and achievement of hemostatic efficacy at 12 hours postinfusion (coprimary end points), rebleeding, use of non–study-prescribed blood products and/or hemostatic drugs, RBC transfusion, change in neurologic status scores in patients with ICH (exploratory end points), and mortality (safety end point) were assessed. At baseline, patients (safety analysis set) had a mean age of 77.9 years (standard deviation [SD] = 10.66), with approximately half of them being male (54.3%) and the majority being white (86.8%). Most patients received apixaban (51.4%) or rivaroxaban (36.5%) anticoagulation and had an ICH (69.0%) or GI bleed (22.9%).

The sponsor submitted 2 pivotal phase III, double-blind, randomized controlled trials (RCTs), ANNEXA-A (N = 68) and ANNEXA-R (N = 80),¹⁸ comparing andexanet alfa and placebo with respect to the reversal of apixaban- (ANNEXA-A) or rivaroxaban-related (ANNEXA-R) anticoagulation in healthy volunteers; however, given that the study populations do not align with the indicated population for andexanet alfa, they were considered to have limited relevance to this review by the CADTH review team, in consultation with the clinical experts. Refer to the main body of the report for a brief summary of the study design and results.

Efficacy Results: ANNEXA-4

The efficacy end points that were noted to be important to patients and clinicians based on stakeholder input are summarized in Table 2.

All efficacy end points were assessed in the efficacy population except for rebleeding, use of non–study-prescribed blood products and hemostatic drugs, and RBC transfusion, which were assessed in the safety population. The efficacy population consisted of 347 of the 477 enrolled patients (73%) who met the ISTH-based criteria for bleeding severity and had baseline anti-FXa activity above the prespecified threshold (at least 75 ng/mL for apixaban- and rivaroxaban-treated patients, at least 40 ng/mL for edoxaban-treated patients, or at least 0.25 IU/mL for enoxaparin-treated patients). The safety population consisted of all 477 enrolled patients who received andexanet alfa treatment.

Table 2: Summary of Key Results From Pivotal Studies and RCT Evidence

CI = confidence interval; EOI = end of infusion; FXa = factor Xa; GCS = Glasgow Coma Scale; ICH = intracranial hemorrhage; ISTH = International Society on Thrombosis and Hemostasis; NIHSS = National Institutes of Health Stroke Scale; RBC = red blood cell; RCT = randomized controlled trial; SD = standard deviation; TEAE = treatment-emergent adverse event.

Note: Outcomes summarized in this table were noted to be important to patients and clinicians based on input received from patient groups, clinician groups, and the clinical experts consulted by CADTH. Unless otherwise specified, safety data were presented by Medical Dictionary for Regulatory Activities (MedDRA) preferred term.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aResults were presented by FXa inhibitor received. Results of edoxaban- and enoxaparin-treated patients were not of interest to this review and, hence, not summarized.

^bThe efficacy population consisted of 347 patients (73%) who met the ISTH-based criteria for bleeding severity and with baseline anti-FXa activity above the prespecified threshold (at least 75 ng/mL for apixaban- and rivaroxaban-treated patients, at least 40 ng/mL for edoxaban-treated patients, or at least 0.25 IU/mL for enoxaparin-treated patients). The safety population consisted of all 477 enrolled patients who received andexanet alfa treatment.

^cThe on-treatment nadir was defined as the minimum value observed from the start of andexanet alfa administration (30 minutes after bolus) to 30 minutes before EOI. If results of both time points are missing, then it was equal to the baseline value.

^fFor patients with missing values for nadir, change and percent change were imputed as zero. The 95% CI for the median was based on the distribution-free method.

^eAchievement of hemostatic efficacy refers to patients with “excellent” or “good” hemostatic efficacy as adjudicated by the end point adjudication committee. Patients who were adjudicated as non-evaluable for clinical reasons were included in the efficacy population and considered as having “poor or none” hemostatic efficacy. Patients adjudicated as non-evaluable for administrative reasons (n = 7) were excluded.

^fCI based on the exact binomial method.

^gRebleeding was defined as recurrent bleeding from the same (or different) anatomic site in patients within 24 hours of initial andexanet alfa treatment and after achieving initial good or excellent hemostasis, as adjudicated by the end point adjudication committee.

^hThe safety population consisted of all 477 enrolled patients who received andexanet alfa treatment.

ⁱCI based on the Fischer exact method.

^jRefers to the use of these treatments between the start of andexanet treatment and 12 hours after the EOI. They included platelets, 4-factor prothrombin complex concentrate, fresh frozen plasma, antithrombin III, human albumin, plasma, vitamin K, packed RBC transfusion, tranexamic acid, carbazochrome sodium sulfonate hydrate, phenylephrine, and other blood or coagulation treatments.

^kNIHSS testing and the additional GCS assessments were not implemented until Protocol Amendment 4; therefore, the number of patients evaluated was lower for the baseline NIHSS and the later time points for GCS.

^lAs adjudicated by the end point adjudication committee.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Change From Baseline in Anti-FXa Activity (Coprimary End Point)

In the primary analysis, the median percent change from baseline in anti-FXa activity at on-treatment nadir (i.e., minimum value between end of bolus [EOB] and end of infusion) was −93.3% (95% confidence interval [CI], −94.2% to −92.5%) in apixaban-treated patients and −94.1% (95% CI, −95.1% to −93.0%) in rivaroxaban-treated patients in the efficacy population. A subgroup analysis showed results consistent with the primary analysis across bleed types; however, a subgroup analysis by estimated glomerular filtration rate (eGFR) is not available. Results in the safety population showed results consistent with the primary analyses.

Achievement of Hemostatic Efficacy (Coprimary End Point)

In the primary analysis of hemostatic efficacy, 80.0% (95% CI, 75.3% to 84.1%) of the efficacy population achieved “good” or “excellent” hemostatic efficacy. The lower bound of the 95% CI was greater than the predefined threshold of 50%, which is sufficient to reject the null hypothesis at the 0.05 level. Three sensitivity analyses were conducted to assess the effect of including patients rated as “non-evaluable for administrative reasons” among all patients with nonmissing baseline anti-FXa level, and in patients otherwise evaluable but with a baseline anti-FXa level below a prespecified threshold. The results of all sensitivity analyses and estimates within the subgroups of interest (by bleed type and eGFR) were consistent with the primary analysis.

The results of a hemostatic efficacy analysis in patients with ICH with a high risk of hematoma expansion (exploratory end point) were consistent with the whole ICH subpopulation as well as the efficacy population.

Rebleeding, Use of Non–Study-Prescribed Blood Products and Hemostatic Drugs, RBC Transfusion (Exploratory End Points)

In the safety population, rebleeding as adjudicated by the end point adjudication committee (EAC) occurred in 1 of 264 assessed patients (0.4%); 24.7% of patients received non–study-prescribed blood products and/or hemostatic drugs between the start of andexanet alfa treatment and 12 hours after the end of the infusion, and 19.5% (95% CI, 16.0% to 23.3%) of patients received RBC transfusion.

Modified Rankin Scale, National Institutes of Health Stroke Scale, and GCS Scores (Exploratory End Point)

The proportions of patients with ICH in the efficacy population who had a modified Rankin Scale (mRS) score of 0 to 2 at baseline, 1 hour posttreatment, 12 hours posttreatment, and at day 30 were 32.2% (95% CI, 26.4% to 38.5%), 22.6% (95% CI, 16.1% to 30.3%), 23.4% (95% CI, 16.8% to 31.2%), and 35.9% (95% CI, 29.6% to 42.7%), respectively.

The mean change from baseline in National Institutes of Health Stroke Scale (NIHSS) score at 1 hour posttreatment, 12 hours posttreatment, and day 30 was 0.4 (SD = 2.66), 1.0 (SD = 3.58), and −1.0 (SD = 4.96), respectively, in patients with ICH.

The mean change from baseline in GCS score 1 hour posttreatment, 12 hours posttreatment, and day 30 was −0.4 (SD = 1.69), −0.6 (SD = 2.06), and 0.2 (SD = 2.23), respectively, in patients with ICH.

Intensive Care Unit Admission, Hospital Length of Stay, and HRQoL (Other Outcomes of Interest to This Review)

Intensive care unit (ICU) admission and HRQoL were not measured. Length of hospital stay was not an efficacy outcome of the trials but the median was reported to be 10.9 days in the safety population.

Harms Results: ANNEXA-4

A summary of harms in the ANNEXA-4 trial is shown in Table 2.

A treatment-emergent adverse event (TEAE) was reported in 72.5% of patients, with the most common TEAEs being urinary traction infection (10.5%) and pneumonia (8.2%). Serious TEAEs were reported in 41.9% of patients, with the most common being pneumonia (4.2%), respiratory failure (2.5%), and ischemic stroke (2.1%). Four patients (0.8%) discontinued treatment due to TEAEs. Death occurred in 17.0% of patients.

A thromboembolic event was reported in 10.5% of patients, including cerebrovascular accident (4.6%), DVT (2.5%), myocardial infarction (1.9%), PE (1.0%), and transient ischemic attack (0.4%). Infusion-related reaction was reported in 2 patients (0.4%). There was no report of neutralizing antibodies to FX, FXa, or andexanet alfa.

Critical Appraisal: ANNEXA-4

An important limitation was the noncomparative study design, which precludes conclusions on whether any observed effect could be attributed to andexanet alfa alone due to the lack of consideration for potential confounders. The open-label study design could increase uncertainty in neurologic status outcomes (mRS, NIHSS, and GCS) due to potential reporting bias, given that these scales involve the subjective assessment of clinical parameters by the investigators, although the presence and extent of such bias could not be determined. All end points other than the coprimary end points were exploratory, which precludes definitive conclusions to be drawn from these analyses. Approximately 20% of patients were excluded from the primary efficacy analyses due to baseline anti-FXa activity below the prespecified threshold, and the impact of excluding such patients from neurologic status outcomes is unclear, since no sensitivity analyses were conducted. There is also potential attrition bias in neurologic status outcomes, given the large amount of missing data, although the direction and extent of bias are unclear.

The generalizability of the study population is limited due to the younger patient population and the exclusion of patients with severe ICH (GCS score less than 7, estimated intracerebral hematoma volume above 60 cc) and an expected survival of less than 1 month, which were suggestive of a study population with a better prognosis than in general clinical practice, based on clinical expert input. Patients without major bleeding but requiring FXa inhibitor reversal for emergency surgery, who are also expected to be reasonable candidates for andexanet alfa according to clinical expert input, were not included in the study; the treatment effects in these patients are unknown. According to the clinical experts, the coprimary end points were not as clinically relevant and meaningful to patients and clinicians compared with clinical outcomes, such as mortality, morbidity, and functional status, which were assessed as exploratory end points in the study. They noted that anti-FXa activity is not routinely measured in clinical practice during FXa inhibitor reversal treatment nor widely available; while hemostatic efficacy is a commonly measured outcome in clinical studies of reversal treatments for major bleeds, it is not a patient-important outcome. The absence of direct comparative evidence between andexanet alfa and PCC, the most relevant comparator for andexanet alfa per clinical expert input, represents a gap in evidence.

Studies Addressing Gaps in the Pivotal and RCT Evidence

Weighted Comparative Observational Evidence

Description of Studies

In the absence of direct comparative evidence between andexanet alfa and PCC, the sponsor submitted 3 published observational analyses assessing the comparative effect of andexanet alfa versus PCC in patients with major bleeding while receiving apixaban- or rivaroxaban-related anticoagulation, based on individual patient data (IPD) from the ANNEXA-4 trial and 3 observational cohort studies of PCC in real-world clinical practice, including the ORal ANticoagulant aGEnt-Associated Bleeding Events Reporting System (ORANGE) study, the Hartford Health care study (HHCS), and the German-Wide Multicenter Analysis of Oral Anticoagulant-Associated Intracerebral Hemorrhage Part Two (RETRACE-II) study.^21-23

The analysis of the ANNEXA-4 trial versus the ORANGE study was conducted in patients with any bleed type using propensity score matching (PSM). The analysis of the ANNEXA-4 trial versus the HHCS was conducted in patients with ICH using propensity score overlap weighting. The analysis of the ANNEXA-4 trial versus the HHCS was conducted in patients with atraumatic ICH based on a propensity score model using the inverse probability treatment weighting (IPTW) approach. Outcomes assessed included 30-day mortality, in-hospital mortality, hemostatic efficacy (or hematoma expansion, which reflects failure to achieve hemostatic efficacy in ICH), mRS score at discharge or at day 30, and thrombotic event at day 5.

ANNEXA-4 Trial Versus ORANGE Study (Any Bleed Type)

The analysis included 322 patients in the andexanet alfa cohort from the ANNEXA-4 trial and the effective sample size (ESS) of the PCC cohort from the ORANGE study was 88 after PSM. Results of the 30-day mortality analysis estimated results in favour of andexanet alfa versus PCC in both the whole cohort (relative risk = 0.43; 95% CI, 0.29 to 0.63) and ICH cohort (relative risk = 0.31; 95% CI, 0.20 to 0.48). A reduced risk of 30-day mortality in the GI cohort (relative risk = 0.49; 95% CI, 0.21 to 1.16) and an increased risk with andexanet alfa in other bleeds (relative risk = 1.29; 95% CI, 0.17 to 9.55) were shown, but the results did not show a significant difference between the interventions. Harms were not assessed in this study.

ANNEXA-4 Trial Versus the HHCS (ICH Only)

The analysis included 107 patients in the andexanet alfa cohort from the ANNEXA-4 trial and 95 patients in the 4-factor PCC (4F-PCC) cohort from the HHCS. After weighting, results were in favour of andexanet alfa over 4F-PCC with respect to hemostatic efficacy (weighted odds ratio [OR] = 2.733; 95% CI, 1.163 to 6.421) and 30-day all-cause mortality (weighted OR = 0.355; 95% CI, 0.129 to 0.977). Thromboembolism occurred in 2 patients (1.9%) in the andexanet alfa cohort and 0 patients in the 4F-PCC cohort within 5 days posttreatment.

ANNEXA-4 Trial Versus the RETRACE-II Study (Atraumatic ICH Only)

In the subanalysis of patients receiving andexanet alfa (n = 85) or PCC (n = 73), after weighting, results were in favour of andexanet alfa over PCC with respect to hematoma expansion (risk ratio = 0.443; 95% CI, 0.223 to 0.878) but did not show a difference between the interventions in in-hospital mortality (hazard ratio = 0.852; 95% CI, 0.397 to 1.827) and mRS score at discharge or at 30 days (mean difference = −0.517; 95% CI, −1.146 to 0.113). Harms were not assessed in this study.

Critical Appraisal

Several important limitations were common to the 3 weighted comparative analyses that preclude drawing definitive conclusions regarding the comparative efficacy of andexanet alfa versus PCC. This includes a risk of selection bias for studies included in the analyses, given the absence of a systematic literature review (SLR) or the lack of reporting in SLR methods. Furthermore, there was evidence of heterogeneity in the inclusion and exclusion criteria between the included studies (e.g., definition of major bleeding, recent history of blood products, definition of ICH severity, recent history of thromboembolism, expected survival). There was also potential residual confounding due to inadequate adjustment for prognostic factors or treatment-effect modifiers; specifically, the studies were adjusted for covariates inconsistently and the resulting estimates varied widely. Lastly, the comparative efficacy of andexanet alfa and PCC for mental status, ICU admission, length of hospital stay, and HRQoL, all of which were of interest to the stakeholders, were not assessed.

Comparative Observational Evidence

Description of Studies

In the absence of direct comparative evidence between andexanet alfa and PCC, the sponsor submitted 6 studies summarizing the comparative evidence of andexanet alfa versus PCC in real-world clinical practice. All 6 studies were multicentre, retrospective hospital chart audits conducted in the US. The patient population for Coleman et al. (2021),²⁴ Dobesh et al. (2022),²⁵ Fermann et al. (2022),²⁶ and Dobesh et al. (2023)²⁷ were sourced from US Hospital Chart Audit, while the populations for Sutton et al. (2022)²⁸ and Sutton et al. (2023)²⁹ were sourced from US Department of Veterans Affairs databases. In all studies, data were captured from electronic medical records. The eligible patient population comprised adult patients aged 18 years and older hospitalized for FXa inhibitor–related bleeding. Patients were identified through electronic medical records using International Classification of Diseases, Tenth Revision (ICD-10) billing codes,^24-27 or administrative claims and pharmacy dispensation information from the Veterans Affairs Informatics and Computing Infrastructure.^28,29 Outcomes of interest were stratified by bleed type for each reversal or replacement agent (i.e., andexanet alfa and PCC) and included in-hospital mortality,^24-27,29 in-hospital 30-day mortality,²⁸ length of hospital stay,^24,25,29 and ICU length of stay.^24,25,29

Study sample size ranged from 255 to 4,395 patients. The average age of patients across the studies ranged from 65.0 years to 70.1 years. The most commonly used FXa inhibitors documented were apixaban (range, 40% to 84.0%) and rivaroxaban (range, 14.5% to 56%). Exposure to andexanet alfa ranged from 11.3% to 48.8% across the studies, while exposure to 4F-PCC ranged from 24.2% to 80.1%.

In-Hospital Morality

In-hospital mortality for those treated with andexanet alfa was 4% in Coleman et al. (2021), 6% in Dobesh et al. (2022) and (2023), and 10.6% in Sutton et al. (2022). In-hospital mortality for those treated with 4F-PCC was 10% in Coleman et al. (2021), 8% in Dobesh et al. (2022), 10.6% in Dobesh et al. (2023), and 25.3% in Sutton et al. (2022). In Coleman et al. (2021), in-hospital mortality was 11% in those treated with fresh frozen plasma (FFP). In Fermann et al. (2022), in-hospital mortality among patients treated with andexanet alfa was approximately 2.5% among patients with a GI bleed and ranged from 9.8% to 16.8% among patients with ICH; in patients treated with 4F-PCC, in-hospital mortality ranged from 3.2% to 6.0% among patients with a GI bleed and 14.5% to 24.0% in patients with ICH.

In the adjusted analysis performed by Dobesh et al. (2022), treatment with andexanet alfa was associated with a lower likelihood of death compared with treatment with 4F-PCC (OR = 0.69; 95% CI, 0.49 to 0.98). In Dobesh et al. (2023), treatment with andexanet alfa was also associated with a lower likelihood of in-hospital mortality compared with 4F-PCC across all bleeds (OR = 0.50; 95% CI, 0.39 to 0.65), GI bleeds (OR = 0.49; 95% CI, 0.29 to 0.81), and ICH (OR = 0.55; 95% CI, 0.39 to 0.76). In the adjusted analysis performed by Fermann et al. (2022), treatment with andexanet alfa was associated with lower odds of in-hospital mortality compared with treatment with 4F-PCC (OR = 0.67; 95% CI, 0.48 to 0.94). In Sutton et al. (2022), treatment with andexanet alfa was associated with a lower hazard of in-hospital mortality compared with treatment with 4F-PCC (hazard ratio [HR] = 0.31; 95% CI, 0.14 to 0.71).

30-Day Hospital Mortality

Thirty-day hospital mortality was explored in Sutton et al. (2022) only. The 30-day mortality rate was 20.0% and 32.4% in patients treated with andexanet alfa and 4F-PCC, respectively. Treatment with andexanet alfa was associated with a lower hazard of 30-day mortality compared with treatment with 4F-PCC (HR = 0.54; 95% CI, 0.30 to 0.98).

ICU Length of Stay

For patients who received andexanet alfa, the overall median ICU length of stay was 2 days (interquartile range [IQR] = 1 to 4) in Coleman et al. (2021), 2 days (IQR = 3) in Dobesh et al. (2022), and 1 day (IQR = 0 to 4) in Sutton et al. (2022). In Sutton et al. (2023), the mean lCU length of stay was 4.0 days (SD = 7.1). For patients who received 4F-PCC, the median ICU stay was 3 days (IQR = 2 to 5) in Coleman et al. (2021), 2 days (IQR = 3) in Dobesh et al. (2022), and 2 days (IQR = 0 to 5) in Sutton et al. (2022). In Sutton et al. (2023), the mean length of ICU stay was 5.0 days (SD = 8.6). In Coleman et al. (2021), the overall median ICU stay for patients who were treated with FFP was 3 days (IQR = 2 to 5).

Hospital Length of Stay

For patients who received andexanet alfa, the median hospital length of stay was 5 days (IQR = 3 to 6) in Coleman et al. (2021), 5 days (IQR = 5) in Dobesh et al. (2022), and 6 days (IQR = 3 to 10) in the Sutton et al. (2022) study. In Sutton et al. (2023), the mean length of stay was 11.3 days (SD = 22.8) while, in Fermann et al. (2022), the mean length of stay ranged from 6.4 (SD = 4.2) to 8.9 (SD = 6.2) days. For patients who received 4F-PCC, the median hospital stay was 5 days (IQR = 4 to 7) in Coleman et al. (2021), 5 days (IQR = 5) in Dobesh et al. (2022), and 7 days (IQR = 4 to 18) in Sutton et al. (2022). In Sutton et al. (2023), the mean length of hospital stay was 12.2 days (SD = 17.5) while, in Fermann et al. (2022), the mean length of hospital stay ranged from 6.3 (SD = 4.1) to 8.3 (SD = 6.2) days. In Coleman et al. (2021) the median length of hospital stay for patients who were treated with FFP was 5 days (IQR = 4 to 8).

Critical Appraisal

The nonrandomized comparison of the 6 real-world evidence (RWE) studies makes interpretation of the efficacy of andexanet alfa compared with 4F-PCC and other reversal or replacement agents challenging. The retrospective nature of the studies meant that data collection and entry were dependent on the personnel at each medical institution; thus, the quality and accuracy of data were dependent on the quality and accuracy of routine documentation. Moreover, the use of ICD-10 codes is associated with limitation for the purpose of clinical research.^30,31 The influence of an important patient-level factor (e.g., advanced directives, thrombotic complications after administration of a reversal agent) — which could not be ascertained from medical records — on the treatment effect of andexanet alfa and 4F-PCC could not be ruled out. Although Dobesh et al. (2022), Dobesh et al. (2023), Fermann et al. (2022), and Sutton et al. (2022) employed covariate adjustment, there is a possibility of selection bias or residual confounding due to the nonrandomized nature of the study designs and a lack of any comprehensive literature to assess the relevant confounding variables. As a result, there is uncertainty around the comparative treatment effects of andexanet alfa compared with 4F-PCC due to selection bias and unmeasured and residual confounding that cannot be entirely ruled out.

Conclusions

One pivotal single-arm, open-label, phase IIIb and IV trial (ANNEXA-4) provided evidence regarding the efficacy and safety of andexanet alfa in the treatment of patients with acute major bleeding while receiving FXa anticoagulation. Results of clinically relevant outcomes, including 30-day mortality, neurologic status, rebleeding, need for blood products and hemostatic drugs, and length of hospital stay, were generally in line with the clinical experts’ clinical experience with PCC. Analyses of hemostatic efficacy and anti-FXa activity also lent support to the ability of andexanet alfa to achieve hemostasis and reduce the anticoagulant activity of FXa inhibitors. However, there was uncertainty in the magnitude of clinical benefit attributable to andexanet alfa, given the inherent limitations of a noncomparative study design. The generalizability of study findings was also limited by the younger study population compared with clinical practice, and the exclusion of patients with severe ICH and patients with expected survival of less than 1 month. Three weighted comparative observational studies and 6 comparative observational studies assessed the comparative efficacy of andexanet alfa versus PCC in real-world clinical practice; however, the results were inconclusive due to important methodological limitations (e.g., selection bias, inadequate adjustment for prognostic factors and effect modifiers) of these studies. The safety profile of andexanet alfa in the pivotal trial appeared to be consistent with PCC, per clinical expert input. One weighted comparative analysis suggested the frequency of thrombotic events was similarly low for both andexanet alfa and PCC within 5 days posttreatment, although results are uncertain due to the aforementioned limitations of the analysis.

Introduction

The objective of this report is to review and critically appraise the evidence submitted by the sponsor on the beneficial and harmful effects of andexanet alfa 200 mg powder solution for IV infusion in adult patients treated with FXa inhibitors (rivaroxaban or apixaban) when rapid reversal of anticoagulation is needed due to acute major bleeding.

Disease Background

The contents of this section were informed by materials submitted by the sponsor and clinical expert input. The following was summarized and validated by the CADTH review team.

DOACs, also known as non–vitamin K antagonist oral anticoagulants, are a class of anticoagulant that directly inhibit specific enzymes within the coagulation cascade.^32,33 DOACs include oral FXa inhibitors (i.e., apixaban, rivaroxaban, edoxaban) that work by selectively blocking the active site of FXa. These drugs are used for a number of different indications, including the prevention of strokes and embolisms in individuals with AF and VTE.^2,3 However, unanticipated, serious bleeding is a major risk associated with anticoagulant treatment that may occur spontaneously or following trauma, complications from invasive procedures, or other illnesses and conditions. It is estimated that, in Canada, approximately 900,000 patients are on rivaroxaban and apixaban, with an increase year over year of approximately 4%.⁸ According to the literature, FXa inhibitors consistently show that the annual risk of major bleeding ranges from 2% to 4%.^6,7,9 Major FXa inhibitor–associated bleeds may occur in several critical sites, including intracranial, intraocular, pericardial, intraspinal, and intra-articular, or bleeding in the GI region. The most frequent type of major bleed related to FXa inhibitor use are GI bleeds, which account for 30% to 45% of all major bleeds, and ICHs, which account for between 10% to 25% of major bleeding events.^4-7 FXa inhibitor–related major bleeding events are associated with an increased risk of death; in fact, 30-day mortality after FXa inhibitor–related major bleeding can range from 9% to 45%.^5,10-15 In particular, a retrospective cohort study conducted in Ontario found that, in patients with ICH, 30-day mortality rates were approximately 34.7%.¹⁵ Alternative sources report that the 30-day mortality rate for ICH may be as high as 45%.^10,13,14 The ARISTOTLE trial reported that all-cause mortality within 30 days of a GI bleed was approximately 9.2%,⁵ with additional trials from the US reporting 30-day mortality rates of approximately 14%.³⁴ Major bleeds are managed through hemostatic measures or interventions to stop the bleed at its source, and additional supportive measures like fluid replacement or transfusion products can further help stabilize the patient.^16,35 As anticoagulants exacerbate bleeding episodes, specific reversal drugs are required to stop the activity of the anticoagulant to further stabilize the patient or attempt to save the patient’s life.^11,36 To date, there is no available antidote in Canada to specifically reverse the anticoagulant effects of FXa inhibitors in the event of major bleeding.

Several patient-specific risk factors have been identified that increase the risk for bleeding in patients receiving FXa inhibitors, including active or metastatic cancer, increasing age (e.g., > 65 years), genetic factors affecting vitamin K antagonist metabolism and antithrombotic effect, comorbidities (e.g., hypertension, diabetes, renal insufficiency, liver disease, chronic obstructive pulmonary disease, cardiomyopathy, peripheral vascular disease), concomitant medications (e.g., nonsteroidal anti-inflammatory drugs, antiplatelets), history of bleeding, previous stroke, recent surgery, thrombocytopenia, anemia, and alcohol abuse.^11,12 Regrettably, some of these factors are expected to increase the use of FXa in Canada as the population ages, increasing the risk for thromboembolic conditions and thus the need for anticoagulation therapies. Recently, prescriptions for rivaroxaban and apixaban have increased in Canada following their addition to provincial formularies for stroke prevention in patients with AF.

Standards of Therapy

The contents of this section were informed by materials submitted by the sponsor and clinical expert input. The following was summarized and validated by the CADTH review team.

PCC, an off-label, plasma-derived, prohemostatic agent, is the usual care in Canada for managing major bleeding associated with FXa inhibitors. Clinical practice guidelines from bodies such as Thrombosis Canada, Canadian Cardiovascular Society, and Canadian Heart Rhythm Society support the use of PCC if a specific antidote is not available.^16,37 The American College of Gastroenterology and Canadian Association of Gastroenterology have noted that the selective use of PCC may be clinically justifiable in some patients who have taken DOACs within the 24 hours preceding a life-threatening GI bleed; however, they could not recommend routine use of PCC for patients with GI bleeding receiving DOAC due to the uncertainty of the available evidence.³⁸

PCC helps control bleeding by providing nonspecific supplementation of exogenous coagulation factors but does not directly reverse FXa inhibitor activity. Clinical trial evidence for PCC in treating FXa inhibitor–related bleeding is lacking and its use is primarily informed by results of small observational studies.¹⁶ Adjunctive use of anti-fibrinolytic treatment, most commonly tranexamic acid in Canada, may also be considered in managing FXa inhibitor–associated bleeding in select patients.¹⁶ None of the existing treatments have a specific mechanism of action targeting the reversal of FXa inhibitors.

In addition to PCC and adjunctive tranexamic treatments, major bleeding is managed by pausing anticoagulant therapy, initiating resuscitation, applying local hemostatic measures (including surgical measures), and administering supportive therapy with blood products (e.g., RBC, platelets, fibrinogen concentrate, cryoprecipitate, and plasma).¹⁶

The key goals of treatment in patients with an acute major bleed include improving survival, achieving hemodynamic stability, and avoiding thrombotic complications, according to the clinical experts consulted by CADTH. In patients with ICH, additional important treatment goals related to neurologic recovery are reducing morbidity and improving function, enabling an ability to return to independent living, and HRQoL. Avoidance of transfusion and ICU admission can also be patient-important outcomes.

Drug Under Review

The key characteristics of andexanet alfa and 4F-PCC for rapid reversal of anticoagulation are summarized in Table 3.

Andexanet alfa is a recombinant modified human FXa protein that lacks procoagulant or anticoagulant activity and acts as an antidote for apixaban and rivaroxaban by binding these FXa inhibitors in the plasma, thereby freeing endogenous FXa to resume its normal function in hemostasis.¹ In addition to reversing the anticoagulant effect of drugs that target FXa, andexanet alfa binds to tissue factor pathway inhibitor (TFPI), an endogenous, naturally occurring anticoagulant that normally circulates in very low concentrations in plasma. TFPI binds reversibly to FXa, and the resulting TFPI–FXa complex inhibits the tissue factor–factor VIIa complex, which plays a key role in the activation of the tissue factor pathway, leading to thrombin generation. When andexanet alfa binds to TFPI, circulating TFPI concentrations are reduced, which may lead to increased thrombin generation.

This is the first review for andexanet alfa by CADTH. Andexanet alfa was granted a Health Canada Notice of Compliance with Conditions on June 16, 2023. The reimbursement request aligns with the Health Canada indication, which is for adult patients treated with FXa inhibitors (rivaroxaban or apixaban) when rapid reversal of anticoagulation is needed due to life-threatening or uncontrolled bleeding. Andexanet alfa was approved by the FDA on May 3, 2018, for patients treated with rivaroxaban and apixaban when reversal of anticoagulation is needed due to life-threatening or uncontrolled bleeding.³⁹ It was approved by the European Medicines Agency on April 26, 2019, and by the Medicines and Health care products Regulatory Agency in the UK on May 13, 2022, for the same indication. In its guidance issued on May 12, 2021, the National Institute for Health and Care Excellence (NICE) andexanet alfa is recommended as an option for reversing anticoagulation from apixaban or rivaroxaban in adults with life-threatening or uncontrolled bleeding only if the bleed is in the GI tract and the company supplying the andexanet alfa provides it according to a commercial arrangement. For ICH, NICE recommended that andexanet alfa be used only in research.

Andexanet alfa is administered as an IV bolus with a target rate of 30 mg/minute followed by continuous infusion for up to 120 minutes. Dosage is based on the specific oral FXa inhibitor, the dose, and the time since the patient’s last dose of the inhibitor.¹ A low dose with an initial IV bolus of 400 mg and a follow-on infusion of 4 mg/minute for 120 minutes (480 mg) is indicated for a rivaroxaban dose of 10 mg or less, or an apixaban dose of 5 mg or less where the last dose was less than 8 hours ago or when the time of the last dose is unknown. A low dose is also indicated when the timing of the last dose was 8 hours or more, regardless of the last dose of the FXa inhibitor. A high dose with an initial IV bolus of 800 mg and a follow-on infusion of 8 mg/minute for 120 minutes (960 mg) is indicated for a rivaroxaban dose that is greater than 10 mg or unknown or an apixaban dose that is greater than 5 mg or unknown, where the last dose was less than 8 hours ago or unknown.¹ The administration of andexanet alfa is to be conducted and/or monitored by a health care practitioner and, due to thromboembolic and ischemic risks, resumption of anticoagulant therapy should be considered as soon as medically appropriate after completion of treatment.¹

Table 3: Key Characteristics of Andexanet Alfa and 4F-PCC

4F-PCC = 4-factor prothrombin complex concentrate; DOAC = direct oral anticoagulant; FXa = factor Xa.

^aNot a selective reversal agent and its effect is debatable in managing FXa inhibitor–related major bleeds with a lack of robust clinical and scientifically rigorous evidence.

^bHealth Canada–approved indication.

Sources: Andexanet alfa product monograph (draft), Beriplex product monograph, Octaplex product monograph, and DOACs: Management of Bleeding (Thrombosis Canada).^1,16,40,41

Stakeholder Perspectives

Patient Group Input

This section was prepared by the CADTH review team based on the input provided by patient groups. The full original patient inputs received by CADTH have been included in the stakeholder section at the end of this report.

Two patient groups, CanVECTOR and the HeartLife Foundation, submitted input for this review. The input from CanVECTOR was informed by 33 interviews with patients with VTE lived experience (DVT, n = 15; PE, n = 6; both DVT and PE, n = 12) and varied treatment experience, including warfarin, DOACs, and low-molecular-weight heparin for anywhere from 1 to 6 months to long-term treatment (more than 3 years, n = 16).

According to the input by CanVECTOR, patients receiving treatment for a VTE have to find a balance between the risk of another clot and the actual or potential side effects of treatment. Patients expressed that bleeding is the most concerning potential side effect of VTE treatments and they may live with a fear of bleeding that can impact life choices and daily activities, quality of life, and mental health. The input noted that a variety of personal factors affect a patient’s treatment preferences. Many patients described the burden of treatment with warfarin (the inconvenience of blood monitoring, restrictions on diet and alcohol intake) or with low-molecular-weight heparin injections (pain, bruising, discomfort with self-injections), and some had later switched to a DOAC. The input noted there were fewer burdens described with DOACs, but the lack of a reversal drug was mentioned by a few patients, for example, “However much I have my misgivings on Warfarin, it had one benefit. It was reversible. It had all the other things that were not attractive to it, but if you could make some of those things more attractive, or certainly if the new drug that would come on the market to replace or be an alternative to apixaban, if it had a reversibility aspect to it, I think would be attractive.” The HeartLife Foundation noted that the use of an FXa inhibitor can increase the risk of bleeding; as well, patients with heart failure may require invasive surgeries, which put them at risk of surgical bleeds. The group expressed a need for a treatment that can rapidly and effectively reverse the anticoagulant effect of FXa inhibitors to prevent further bleeding and ensure the best possible outcome in patients who require urgent surgery or are experiencing life-threatening bleeds. This was echoed by the input by CanVECTOR, which highlighted that patients were concerned that the absence of a reversal drug for DOACs could prevent them from undergoing a surgery in case of an emergency: “If an accident happened and the surgery is almost immediate, a reversal agent can help me … [go] into the emergency surgery safely.” The CanVECTOR also highlighted that some patients with a high risk of another clot will be prescribed blood thinners for the rest of their lives, and the experience could be daunting for a young person with decades of treatment ahead.

Clinician Input

Input From Clinical Experts Consulted by CADTH

All CADTH review teams include at least 1 clinical specialist with expertise regarding the diagnosis and management of the condition for which the drug is indicated. Clinical experts are a critical part of the review team and are involved in all phases of the review process (e.g., providing guidance on the development of the review protocol, assisting in the critical appraisal of clinical evidence, interpreting the clinical relevance of the results, and providing guidance on the potential place in therapy). The following input was provided by 2 clinical specialists with expertise in the diagnosis and management of acute major bleeding in patients receiving an FXa inhibitor.

Unmet Needs

The clinical experts noted there is currently a lack of a specific reversal drug for FXa-related bleeding. They noted that currently available hemostatic agent (i.e., PCC) do not always achieve a rapid response; this can be a concern in patients with time-sensitive major bleeding, such as patients with bleeding within the central nervous system, high-risk GI bleeding, a trauma in need of urgent operative intervention and, in rare instances, patients with acute kidney injury who experience a prolonged anticoagulant effect from an FXa inhibitor. The clinical experts also noted that PCC and blood products are not universally available in all clinical settings. In addition, the risk of thrombosis with existing treatments is unclear, according to the clinical experts.

Place in Therapy

The clinical experts noted that andexanet alfa is a reversal drug specifically for FXa inhibitors and may act more rapidly than the other reversal agents currently available. The clinical experts anticipated that andexanet alfa would have the same place in therapy as PCC (i.e., as an alternative to PCC) in the treatment of acute major bleeding in patients receiving an FXa inhibitor; however, based on available evidence, it is unclear whether andexanet alfa or PCC should be the preferred therapy, and suggested this should be based on clinical presentation. The clinical experts noted that in addition to clinical evidence, practical factors such as treatment access (e.g., through hospital transfusion medicine laboratories versus hospital pharmacies) and cost could also affect which treatment is used; therefore, it is challenging to conclude with certainty whether andexanet alfa has the potential to cause a shift in the current treatment paradigm.

Patient Population

The clinical experts noted that patients receiving an FXa inhibitor who have acute major bleeding are potential candidates for andexanet alfa treatment. According to the clinical experts, such patient populations would most commonly include older patients who are compliant with their FXa inhibitor treatment and present to the emergency department with brisk nonvariceal upper GI bleeds or head trauma resulting in intracranial bleeding that is believed to be worsening or at risk of worsening. The clinical experts also noted that patients who are being treated with an FXa inhibitor and require urgent surgery (e.g., patients with a hip fracture who could benefit from early operative intervention) may also be candidates for andexanet alfa, although this patient population was not studied in the pivotal trial.

The clinical experts noted that the definition of major bleed by the ISTH⁴² is the 1 most commonly used by clinicians in Canada. According to the clinical experts, patients with major bleeding who require reversal of the anticoagulant effect of an FXa inhibitor would be identified in clinical practice primarily based on assessments of patients’ clinical status (severity, bleed location, and response to nonspecific supportive measures). Laboratory markers such as prothrombin time are not sensitive measures of anticoagulant activity; establishing when the last dose of anticoagulant was taken would be a more reliable indicator of whether anti-FXa activity is still present, according to the clinical experts. The clinical experts noted that FXa assays could help inform treatment decisions if done in a timely fashion (turnaround time of less than 1 hour); however, this is not realistic outside of a small number of treatment centres in Canada.

The clinical experts noted that patients with minor bleeding would not be suitable for andexanet alfa treatment since non-major bleeds are expected to resolve on their own without the need for a reversal agent and with minimal patient harm.

Assessing the Response to Treatment

The clinical experts noted that, in general, response to treatment is assessed based on hemostatic stability, hemostatic control (i.e., cessation of bleeding), need for blood transfusion, survival, and HRQoL. Stabilization of vital signs, improvement or normalization in laboratory markers (e.g., serial hemoglobin and lactate measures, coagulation profile), and cessation of bleeding (e.g., based on endoscopy findings) are typical indicators of achievement of hemostatic control, according to the clinical experts. The clinical experts noted that continuous monitoring of hemostatic control, transfusion needs, and survival is typically done for the first 72 hours after admission. An absence of thrombosis (ideally assessed at 5 days and 1 month posttreatment) and an improvement in HRQoL (ideally assessed at 1 month posttreatment) would also be indicative of a positive response to treatment, although the timing of the assessment could differ between patients.

Prescribing Considerations

The clinical experts noted that the vast majority of emergency department physicians would be comfortable prescribing andexanet alfa in consultation with a hematologist, thrombosis physician, or transfusion medicine specialist; however, the clinical experts indicated that access to any of these specialists is limited in remote and rural areas and that lack of such access could potentially be a barrier to the timely administration of a reversal drug in an emergency situation. Therefore, the clinical experts noted it may be reasonable to allow prescribing by clinicians who have expertise in the management of acute major bleeds.

The clinical experts noted it would be appropriate to prescribe andexanet alfa in a hospital setting (e.g., emergency department or an inpatient or operating room).

The clinical experts noted that redosing of a reversal treatment is rare in clinical practice, and the pharmacology of andexanet alfa suggests that redosing is not necessary. Redosing is associated with increased thrombotic risk and should be done only in exceptional circumstances under the guidance of a transfusion medicine specialist, thrombosis physician, or hematologist.

Clinician Group Input

This section was prepared by the CADTH review team based on the input provided by the clinician groups. The full original clinician group inputs received by CADTH have been included in the stakeholder section at the end of this report.

Clinician group input was obtained from 5 clinician groups, including Thrombosis Canada (represented by 2 clinicians), members of the Thrombosis and Anticoagulation team at Dalhousie University (represented by 3 clinicians), faculty members at McMaster University in hematology and/or thromboembolism (represented by 5 clinicians), the Canadian Stroke Consortium (represented by 5 clinicians), and a journal club comprising local emergency medicine physicians in Peel Region (represented by 5 clinicians).

The clinician groups noted that PCC, which is currently the reversal treatment of choice for DOAC-related major bleeding, has a nonspecific mechanism of action and is an off-label drug with no robust clinical data supporting its efficacy or safety. They noted that PCC may promote a prothrombotic state and could potentially not be safely used in patients who have a history of heparin-induced thrombocytopenia or thrombosis, since a small amount of heparin may be present in PCC.

The clinician groups noted that andexanet alfa, being the only specific reversal treatment available to patients who are on an FXa inhibitor, would be used as a first-line treatment in patients who require urgent anticoagulant reversal in the setting of serious, life-threatening bleeding or a need for urgent surgery. One clinician group noted that PCCs would likely be used for an “average” patient who needs DOAC reversal, whereas andexanet alfa would be used in select patients with:

• life-threatening bleeding that does not respond to supportive management (i.e., fluids, packed RBCs)

• critical site bleeding (intracranial, spinal, pericardial)

• a need for emergency (within 6 to 8 hours) or urgent (within 12 to 24 hours) surgery.

They noted there is no established threshold for clinically significant hemostatic impairment, and most centres do not have FXa inhibitor drug assays available, so treatment is usually considered based on the timing of the last dose, drug half-life, patient’s kidney or liver function, and examination findings (e.g., hypotension), radiographic findings (e.g., CT scans), and clinician judgment. They also noted that patients who would be less suitable would be those who took their last dose of FXa inhibitor more than 1 to 2 days ago (in the presence of normal renal function), who have bleeding that is not life-threatening, or whose surgery can be delayed for 1 to 2 days after their last dose of an FXa inhibitor.

Outcomes to assess response to treatment deemed important by the clinician groups included achievement of excellent or good hemostatic efficacy, thrombotic events, decreased mortality, decrease in hemoglobin or hematocrit by less than 20% compared with baseline, improvement in symptoms, reduction in hematoma expansion in ICH, survival to discharge, disability score on discharge, and measurement of anti-FXa levels before and after drug administration (not widely available). They noted that length of hospital stay may also be a surrogate marker. The clinical groups noted the criteria for discontinuation would include unexpected allergic or infusion reactions or thromboembolic events.

The clinical groups also supported the use of andexanet alfa in the hospital setting, including the emergency department, critical care unit, operating room, tertiary trauma centre, or stroke or neurosurgical referral centre, and administered by anesthesiologists or by surgical or other specialists such as those in emergency or internal medicine, depending on the site of the bleeding (e.g., neurologist or neurosurgeon for ICH, or gastroenterologists for GI bleeding).

Drug Program Input

The drug programs provide input on each drug being reviewed through CADTH’s reimbursement review processes by identifying issues that may impact their ability to implement a recommendation. The implementation questions and corresponding responses from the clinical experts consulted by CADTH are summarized in Table 4.

Table 4: Summary of Drug Plan Input and Clinical Expert Response

CBS = Canadian Blood Services; CPEC = Canadian Plasma Protein Product Expert Committee; DOAC = direct oral anticoagulant; FXa = factor Xa; GCS = Glasgow Coma Scale; GI = gastrointestinal; GU = genitourinary; ICH = intracranial hemorrhage; ICU = intensive care unit; ITC = indirect treatment comparison; PCC = prothrombin complex concentrate; PPRP = Plasma Protein and Related Products.

Clinical Evidence

The objective of CADTH’s Clinical Review Report is to review and critically appraise the clinical evidence submitted by the sponsor on the beneficial and harmful effects of andexanet alfa 200 mg powder for solution for IV infusion for rapid reversal of anticoagulation due to acute major bleeding, including life-threatening bleeds, in adult patients treated with FXa inhibitors (rivaroxaban or apixaban). The focus will be placed on comparing andexanet alfa with relevant comparators and identifying gaps in the current evidence.

A summary of the clinical evidence included by the sponsor in the review of andexanet alfa is presented in 2 sections, and CADTH’s critical appraisal of the evidence is included after each section. The first section includes pivotal studies and RCTs that were selected according to the sponsor’s systematic review protocol. The second section includes additional studies that were considered by the sponsor to address important gaps in the pivotal and RCT evidence.

Included Studies

Clinical evidence from the following is included in the CADTH review and appraised in this document:

• Three pivotal studies (ANNEXA-4, ANNEXA-A, and ANNEXA-R)^17,18

• Three weighted comparative observational studies (a PSM analysis, a propensity score overlap–weighted analysis, and a propensity score-weighted analysis using IPTW)^21-23

• Six comparative observational studies: Coleman et al. (2021), Dobesh et al. (2022), Fermann et al. (2022), Dobesh et al. (2023), Sutton et al. (2022), Sutton et al. (2023).^24-29

Pivotal Studies and RCT Evidence

The contents of this section were informed by materials submitted by the sponsor. The following was summarized and validated by the CADTH review team.

One pivotal study, ANNEXA-4,¹⁷ met the inclusion criteria for the systematic review conducted by the sponsor and was presented in this report. The sponsor submitted 2 additional pivotal trials of andexanet alfa conducted in healthy participants (ANNEXA-A and ANNEEXA-R),¹⁸ neither of which met the inclusion criteria for the sponsor’s systematic review. A brief summary of the ANNEXA-A and ANNEXA-R trials is also provided at the end of this section. The characteristics of the 3 pivotal studies are summarized in Table 5.

Table 5: Details of Pivotal Studies and RCT Evidence Identified by the Sponsor

DVT = deep vein thrombosis; EAC = end point adjudication committee; EOI = end of infusion; FXa = factor Xa; Hb = hemoglobin; ICH = intracranial hemorrhage; PCC = prothrombin complex concentrate; PE = pulmonary embolism; RCT = randomized controlled trial; VTE = venous thromboembolism.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aFor example, severe hypotension, poor skin perfusion, mental confusion, or low urine output that could not be otherwise explained.

^bThe dose of enoxaparin had to be at least 1 mg/kg/day.

^cAs determined by the investigator based on medical history, full physical examination (including blood pressure and pulse rate measurement), 12-lead electrocardiogram, and clinical laboratory tests.

^dFor example, controlled hypertension, noninsulin-dependent diabetes, osteoarthritis, hypothyroidism.

^eUsing the Cockcroft-Gault equation.

^fAs assessed by CT or MRI.

^gIncluding venous thromboembolism (e.g., DVT, PE, cerebral venous thrombosis), myocardial infarction, including an isolated troponin level, disseminated intravascular coagulation, cerebral vascular accident, transient ischemic attack, unstable angina pectoris hospitalization, or severe peripheral vascular disease.

^hAdministration of platelets or packed red blood cells was not an exclusion criterion.

ⁱPatients with planned minimally invasive surgery or procedure (e.g., endoscopy, bronchoscopy, central line, bur hole) could be enrolled into the study.

^jFor example, patients with a known or suspected hypercoagulable state, history of VTE, DVT, stroke, myocardial infarction, cancer (other than nonmelanoma skin cancer), atrial fibrillation, congestive heart failure, cardiomyopathy, phlebitis, lower extremity edema, major surgery or trauma within 2 months of study day −1, airplane travel ≥ 2 hours during 4 weeks before study day −1, or general immobility were excluded.

^kPrior to initiation of andexanet alfa or placebo, patients received apixaban 5 mg orally every 12 hours for 3.5 days.

^lPrior to initiation of andexanet alfa or placebo, patients received rivaroxaban 20 mg orally every 24 hours for 4 days.

Sources: Clinical Study Reports for the ANNEXA-4, ANNEXA-A, and ANNEXA-R trials.^19,49,50

Description of ANNEXA-4 Trial

The ANNEXA-4 study was a phase IIIb and IV, multicentre, open-label, single-arm trial with primary objectives to demonstrate the decrease in anti-FXa activity following andexanet alfa treatment and to evaluate the hemostatic efficacy of andexanet alfa treatment in patients receiving an FXa inhibitor who are experiencing an acute major bleed (N = 477). The secondary objective was to assess the relationship between a decrease in anti-FXa activity and achievement of hemostatic efficacy in the study population. The study was conducted at 85 sites spread across North America (including 2 in Canada), Europe, and Asia. Patients were enrolled between April 2015 and March 2020 and consisted of 3 study periods: a screening period (day 1), where study eligibility was assessed; a treatment period (day 1), where all enrolled patients received andexanet alfa treatment (bolus followed by continuous infusion); and a safety evaluation period (day 1 to day 30), where patients were monitored for safety outcomes. The study design of ANNEXA-4 is presented in Figure 1.

Figure 1: ANNEXA-4 Study Design

echo = echocardiogram; Ei = end of infusion; EOB = end of bolus; FXa = factor Xa; h = hour; ICH = intracranial hemorrhage; min = minutes.

Note: This figure outlines the study design in the treatment phase. The screening phase (not shown), which occurred just before the treatment phase, involved study eligibility assessments. The follow-up phase (not shown), which occurred after the treatment phase, involved safety monitoring for up to 30 days.

^# Redosing procedures began before the start of the second bolus for measurement of anti-FXa activity.

** CT/MRI was required for ICH and intraspinal patients. The Ei + 1 hour scan (window Ei + 4 hours) is required for ICH only.

^¥ Pericardial bleeds only.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Population

Inclusion and Exclusion Criteria

The key inclusion and exclusion criteria of the ANNEXA-4 trial are presented in Table 5.

Eligible patients in the ANNEXA-4 trial were at least 18 years of age, presenting with acute major bleeding, and had (or were believed to have) received 1 of: apixaban, rivaroxaban, edoxaban, or enoxaparin within the past 18 hours.¹⁷ Acute major bleeding was defined (using the modified ISTH definition) as bleeding with 1 or more of the following features: potentially life-threatening bleeding with signs or symptoms of hemodynamic compromise, bleeding associated with a decrease in the hemoglobin level of at least 2 g/dL or a hemoglobin level of 8 g/dL or lower if baseline level was unavailable, or bleeding in a critical area or organ. Key exclusion criteria were planned surgery within 12 hours after andexanet alfa treatment (with the exception of minimally invasive procedures), ICH in a patient with a score of less than 7 on the GCS scale (scores range from 15 [normal] to 3 [deep coma]) or an estimated intracerebral hematoma volume of more than 60 cc, expected survival of less than 1 month, history of thrombotic event in the past 2 weeks, or use of a vitamin K antagonist, dabigatran, PCC, recombinant factor VIIa, whole blood, or plasma in the past 7 days.

Interventions

Patients in the ANNEXA-4 trial received either high-dose (800 mg bolus over approximately 15 to 30 minutes then 8 mg/minute continuous infusion for 120 minutes, i.e., 960 mg) or low-dose (400 mg bolus over approximately 15 to 30 minutes, then 4 mg/minute continuous infusion for 120 minutes, i.e., 480 mg) andexanet alfa in an inpatient setting based on the specific anticoagulant taken and the timing and amount of the last dose, as depicted in Table 6. If the timing of the last dose of FXa inhibitor was known, the start of the andexanet alfa bolus began within 18 hours following the last dose. If the timing of the last dose of FXa inhibitor was unknown, the andexanet alfa bolus began as soon as possible but no later than 3 hours after the completion of pretreatment procedures.

Platelet transfusions, systemic antifibrinolytic (e.g., aminocaproic acid and tranexamic acid), other systemic hemostatic agent, local hemostatic agents (e.g., microfibrillar collagen, chitosan-containing products), and topical vasoconstrictors were allowed. Packed RBC transfusions were allowed if the hemoglobin level was equal to or lower than 8.0 g/dL (plus or minus 1 g/dL). Procoagulant factor infusions (e.g., 3- or 4-factor PCC or activated PCC, recombinant factor VIIa) and whole blood were avoided until after the 12-hour hemostatic efficacy evaluation unless absolutely required, and the use of such blood products resulted in hemostatic efficacy with andexanet alfa being rated as “poor or none” for that patient. The use of anticoagulant and antiplatelet drugs was prohibited until after the 12-hour hemostatic efficacy evaluation.

One additional andexanet treatment was allowed in patients with rebleeding (i.e., recurrent bleeding from the same or another anatomic site, or new bleeding from a different anatomic site, within 24 hours of initial andexanet treatment and after achieving initial good or excellent hemostasis) based on predefined criteria, and investigators were required to document the clinical justification in the case report forms.

No blinding of patients, investigators, or study personnel took place in the trial, with the exception of the end point adjudication committee, which was blinded to all anti-FXa activity data during the adjudication process, and the core imaging laboratory that read the CT scans, which was blinded to anti-FXa activity data and the clinical outcomes for each patient.

Outcomes

A list of the efficacy end points assessed in this Clinical Review Report is provided in Table 7. Summarized end points are based on those included in the sponsor’s summary of clinical evidence as well as any identified as important to this review according to stakeholders, for example, the clinical experts, clinician groups, or patient groups.

Table 6: Andexanet Alfa Dosing Regimens — ANNEXA-4 Trial

FXa = factor Xa.

Note: Low dose = 400 mg bolus plus 4 mg/minute continuous infusion for 120 minutes; high dose = 800 mg bolus plus 8 mg/minute continuous infusion for 120 minutes.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Table 7: Outcomes Summarized From the ANNEXA-4 Trial

EOB = end of bolus; EOI = end of infusion; FXa = factor Xa; HRQoL = health-related quality of life; ICH = intracranial hemorrhage; ICU = intensive care unit; NA = not applicable; RBC = red blood cell.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aThe on-treatment nadir was defined as the minimum value observed during treatment administration (i.e., minimum of EOB and EOI).

^bStatistical testing for these end points was adjusted for multiple comparisons (e.g., hierarchal testing).

^cAs per input from the clinical experts and clinician groups, ICU admission, length of stay in the hospital, and HRQoL are important outcomes to consider when assessing the efficacy of andexanet alfa; however, such outcomes were not measured as efficacy outcomes in the pivotal studies.

^dHospital length of stay was not an efficacy end point, but it was reported in the Clinical Study Report.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Efficacy Outcomes

Assessment time points included baseline; EOB; EOI; 1, 4, 8, 12, 18, 24, 48, and 72 hours after EOI; and day 30. Baseline measurement was defined as the last value obtained before the start of the andexanet alfa bolus.

Anti-FXa Activity

The percent change from baseline in anti-FXa activity to on-treatment nadir was the first coprimary efficacy end point. Anti-FXa activity was measured at baseline; EOB; EOI; and 4, 8, and 12 hours after EOI. On-treatment nadir was defined as the minimum value between EOB and EOI.

Anti-FXa activity refers to the concentration of an FXa inhibitor and was used in the study to assess the ability of andexanet alfa to reverse the anticoagulation effect of the FXa inhibitor. It was measured with a modified chromogenic assay at a central laboratory using plasma samples. As summarized in Table 8, no relevant studies assessing the psychometric properties and no established minimal important difference (MID) of anti-FXa activity were identified in the sponsor’s submission.

Hemostatic Efficacy

The proportion of patients with excellent or good hemostatic efficacy in stopping an ongoing major bleed at 12 hours from the EOI was the second coprimary end point in the ANNEXA-4 trial. The achievement of hemostatic efficacy in patients with ICH at high risk of hematoma expansion (i.e., the high-risk ICH population) was an exploratory efficacy end point. Hemostatic efficacy (“excellent,” “good,” “poor or none,” “not evaluable due to clinical reasons,” “not evaluable due to administrative reasons”) was adjudicated by the independent EAC that was blinded to anti-FXa activity levels based on predefined criteria. Patients who were adjudicated as non-evaluable for clinical reasons were included in the primary analysis of hemostatic efficacy, which was rated as “poor or none” in these patients. Patients who were adjudicated as non-evaluable due to administrative reasons (e.g., missing the 12-hour assessment due to investigator oversight, withdrawing consent before the 12-hour assessment) were excluded from the primary analysis.

According to the sponsor, the efficacy variable of “effective hemostasis” chosen in the study was an end point developed in collaboration with the FDA and agreed to by the European Medicines Agency’s Committee for Medicinal Products for Human Use as part of the scientific advice the sponsor received in 2014. The criteria for this end point were originally based on those used in the study by Sarode et al.,⁵¹ which was a registration trial for a reversal agent for warfarin-induced bleeding. The scoring system for effective hemostasis allows for the assessment of ICH, pericardial bleeding, intraspinal bleeding, and nonvisible bleeding based on objective measures, such as CT, MRI, echocardiogram, and corrected hemoglobin or hematocrit level. A description of the hemostatic efficacy rating system and psychometric properties are summarized in Table 8.

Rebleeding

The occurrence of rebleeding following andexanet alfa treatment was an exploratory efficacy end point. Rebleeding was defined as recurrent bleeding from the same (or different) anatomic site in patients within 24 hours of initial andexanet alfa treatment and after achieving initial good or excellent hemostasis. Rebleeding was adjudicated by the EAC.

Use of Blood Products and Hemostatic Drugs

The use of non–study-prescribed blood products and/or hemostatic drugs, the proportion of patients who received an RBC transfusion, and the number of RBC units transfused per patient from the start of the andexanet alfa bolus through 12 hours after the EOI were exploratory outcomes.

mRS, NIHSS, and GCS

The mRS, NIHSS, and GCS were evaluated for patients with ICH only as exploratory efficacy end points. The assessments were performed at baseline, at 1 hour, 12 post the end of andexanet alfa infusion, and at day 30 follow-up.

The mRS is a clinician-administered assessment that describes “global disability” after a stroke. The mRS scale score ranges from 0 (perfect health without symptoms) to 6 (death). A single point change on the scale is regarded as being clinically relevant, according to Broderick et al. (2017).⁵²

The NIHSS is a clinician-administered rating scale for neurologic deficit and is composed of 11 items, each of which scores a specific ability between 0 and 4. For each item, a score of 0 typically indicates normal function in that ability, while a higher score is indicative of some level of impairment. The individual scores from each item are summed to calculate a patient’s total NIHSS. A score of 0, 1 to 4, 5 to 15, 16 to 20, or 21 to 42 indicates no stroke symptoms, minor stroke, moderate stroke, moderate to severe stroke, and severe stroke, respectively. A change of 2 points or more has been suggested as being clinically relevant, based on Harrison et al. (2013).⁵³ In patients with an ICH following a treatment for thrombolysis or ischemic stroke, a change in 2 points or more in an individual parameter, or a change in 4 points or more overall, is considered clinically relevant.⁵⁴

The GCS is a commonly used clinician-administered scale that describes the level of consciousness in all patients with acute medical conditions or trauma. In the trial, each patient was rated for all 3 GCS domains (i.e., eye opening, verbal response, and best motor response). The individual scores for each domain were summed to calculate a patient’s total GCS score. The maximum possible score was 15 (normal) and the minimum score was 3 (deep coma). No MID has been identified in patients with ICH.

ICU Admission, Length of Hospital Stay, and HRQoL

These outcomes were identified to be important to patients and clinicians based on stakeholder input; however, they were not assessed as efficacy end points in the pivotal trials. Hospital length of stay in the safety population was reported in the Clinical Study Report.

Table 8: Summary of Outcome Measures and Their Measurement Properties

CI = confidence interval; EOI = end of infusion; FXa = factor Xa; ICH = intracranial hemorrhage; LMWH = low-molecular-weight heparin; MID = minimal important difference; RBC = red blood cell; UH = unfractionated heparin.

Statistical Analysis

A summary of the statistical analysis methods of the ANNEXA-4 trial is provided in Table 9.

Sample Size and Power Calculation

The study originally aimed to include 162 efficacy-evaluable patients, which would provide 80% power for detecting a response rate of 61% and a 95% CI completely above 50% with respect to the coprimary end point of effective hemostasis. The sample size was further increased to 250 based on the assumption that approximately 30% of the patients in the safety population would have a baseline anti-FXa activity of less than 75 ng/mL (or 0.25 IU/mL for patients receiving enoxaparin) and, thus, would not be included in the primary analysis, as well as the assumption that 5% of patients would not be evaluable for reasons unrelated to andexanet alfa. The final sample size was adjusted to 500 patients to allow enrolment of a minimum of 110 evaluable patients with ICH, as per a request from the FDA, and more patients receiving edoxaban or enoxaparin as well as patients of Japanese descent.

The sponsor noted that the threshold for a response rate of 50% was informed by unpublished observations at the April 2014 Cardiac Safety Research Consortium and by expert opinion noting that the ability to help 50% or more patients on FXa inhibitors with an acute bleed to achieve effective hemostasis would represent an important advancement relative to the current standard of care for such patients. The sponsor also noted that a 50% efficacy rate is similar to the lower bound of the 95% CI for effective hemostasis in the study by Sarode et al.,⁵¹ in which the same effective hemostasis end point was used in a population of vitamin K antagonist–treated patients with acute major bleeding.

Primary Efficacy End Points

Statistical Tests

The primary coprimary end point of percent change from baseline in anti-FXa activity was evaluated in the efficacy population using distribution-free 95% CIs for the median using the order statistics. If the CI for the median excluded 0, the first coprimary end point was considered to have been met.

The secondary coprimary end point of proportion of patients in the efficacy population with effective hemostasis (excellent or good) was considered met if the proportion was statistically significantly higher than 50% (at the 0.05 threshold). The significance test was calculated using the 1-sided exact binomial test.

Acknowledging the potential for error due to multiplicity in testing, a conclusion on the secondary coprimary end point (hemostatic efficacy) could be drawn only if the primary coprimary end point (percent change from baseline in anti-FXa activity) was met.

Handling of Missing Data

When both the EOB and EOI assessments of anti-FXa activity were missing, the maximum change was replaced with zero (i.e., no change from baseline) and the nadir was replaced with the baseline value (baseline observation carried forward). When either the EOB or EOI value were missing, the available sample was used for the imputation (last observation carried forward).

In the primary analysis of hemostatic efficacy, patients who were rated as “non-evaluable due to administrative reasons” were excluded and were considered to have experienced treatment failure.

Sensitivity and Subgroup Analyses

Sensitivity analyses were performed for both of the coprimary efficacy end points in patients who were otherwise efficacy-evaluable but who had a baseline anti-FXa level of less than 75 ng/mL for apixaban and rivaroxaban, less than 40 ng/mL for edoxaban, and less than 0.25 IU/mL for patients receiving enoxaparin. In addition, the primary efficacy analysis was performed for all patients in the safety population with a baseline anti-FXa level, including those below the evaluability threshold. For the coprimary end point of hemostatic efficacy, an additional sensitivity analysis was conducted in which all patient rated as “non-evaluable due to administrative reasons” were included and considered as having “poor or none” hemostatic efficacy.

Subgroup analyses were performed as exploratory analyses for the 2 coprimary end points. Two subgroups were of interest to this review based on input from the clinical experts consulted by CADTH: bleed type (GI, ICH, other) and renal function (eGFR of less than 50 mL/minute, eGFR of at least 50 mL/minute). No adjustment for type I error was made, nor was an interaction-by-subgroup test performed in the 2 subgroup analyses.

Secondary Efficacy End Points

There were no secondary efficacy end points in this study; however, a logistic regression model was used to estimate the OR of success of hemostatic efficacy (rated as excellent or good) by anti-FXa level activity as a covariate. This analysis was not believed by the CADTH review team to describe a meaningful effect, based on input from the clinical experts consulted by CADTH; therefore, the methods are not summarized in this report.

Exploratory Efficacy End Points

Occurrence of rebleeding, use of non–study-prescribed bleed products and/or hemostatic drugs, and change from baseline in NIHSS and GCS scores were analyzed using descriptive statistics. For the mRS analysis, the proportions of patients with an mRS score of 0 to 2 (i.e., qualified as success) and those with a score greater than 2 (i.e., qualified as failure) were presented at each assessment time point and the corresponding 95% Fisher exact CIs were presented. The proportion of patients who received an RBC transfusion and the Fisher exact 95% CI were presented. No quantitative analysis on the number of RBC units transfused per patient was reported.

Analysis Populations

The analysis populations of the ANNEXA-4 trial are summarized in Table 10.

Results

Patient Disposition

The patient disposition from the ANNEXA-4 trial is summarized in Table 11. The number of screening failures was not reported. In the ANNEXA-4 trial, 477 patients were enrolled, 17.2% of whom discontinued from the study, most commonly as a result of death (16.4%). The safety population consisted of all enrolled patients. The efficacy population consisted of 347 patients (72.7%). The most common reason for exclusion from the efficacy population was having a baseline anti-FXa level below the prespecified threshold (n = 94; 19.7%).

Table 9: Statistical Analysis of Efficacy End Points — ANNEXA-4 Trial

BOCF = baseline observation carried forward; CI = confidence interval; EOB = end of bolus; EOI = end of infusion; FXa = factor Xa; ICH = intracranial hemorrhage; LOCF = last observation carried forward; NA = not applicable; RBC = red blood cell.

Note: No adjustment factors were applied to any of the analyses. The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aDescriptive statistics refers to, for example, mean (and standard deviation) or median (and range).

Source: ANNEXA-4 Clinical Study Report.¹⁹

Table 10: Analysis Populations — ANNEXA-4 Trial

EAC = end point adjudication committee; FXa = factor Xa; Hb = hemoglobin; ICH = intracranial hemorrhage; ISTH = International Society on Thrombosis and Hemostasis.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.

^aThe patients who met the ISTH-based bleeding inclusion criteria were those who had an acute overt major bleeding episode requiring urgent reversal of anticoagulation, which was defined as either: acute overt bleeding that is potentially life-threatening (e.g., with signs or symptoms of hemodynamic compromise); acute overt bleeding associated with a fall in Hb level by ≥ 2 g/dL, or an Hb ≤ 8 g/dL if no baseline Hb was available; or acute bleeding in a critical area or organ, such as a pericardial, intracranial, or intraspinal bleed.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Table 11: Summary of Patient Disposition — ANNEXA-4 Trial

FXa = factor Xa.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.

^aEarly withdrawal refers to patients who withdrew after day 1 and before day 30.

^bWithdrawal of consent refers to patients who withdrew at hour 12.

^cReasons for exclusion from the efficacy population in the ANNEXA-4 trial included patients who: had a baseline anti-FXa level below the efficacy threshold (i.e., baseline anti-FXa levels of < 75 ng/mL for apixaban and rivaroxaban, < 40 ng/mL for edoxaban, or < 0.25 IU/mL for enoxaparin) (n = 94), were missing a baseline anti-FXa level (n = 17), did not meet the clinical bleeding inclusion criteria (n = 14), did not meet the clinical bleeding inclusion criteria (refer to Table 4), or had a baseline anti-FXa level below the efficacy threshold (n = 5).

Source: ANNEXA-4 Clinical Study Report.¹⁹

Baseline Characteristics

A summary of baseline patient characteristics in the ANNEXA-4 trial is shown in Table 12. The baseline characteristics outlined in the table are limited to those that are most relevant to this review or were felt to impact the outcomes or interpretation of the study results.

The efficacy population had a mean age of 77.8 years (SD = 10.6). Fifty-three percent of patients were male. The majority of patients were white (86.5%). Most patients had a creatinine clearance of 30 mL/minute or above (84.7%) and received apixaban (49.6%) or rivaroxaban (37.5%). Bleeding was predominantly intracranial (71.2%) or GI (22.5%). Patients were hospitalized for a mean time of 3.4 hours (SD = not reported) before starting andexanet alfa treatment. At the time of informed consent for participation in the study, 76.9% of patients were in the emergency department and 15.9% of patients were in the ICU. No notable differences in the baseline characteristics were noted between the safety population and efficacy population.

Exposure to Study Treatments

In the ANNEXA-4 trial, 79.0% of the patients in the efficacy population received the low-dose regimen for andexanet alfa and 21.0% of patients received the high-dose regimen. The majority of patients (91.6%) completed andexanet alfa treatment without interruption, while others had study-drug interruptions (6.1%), modifications (1.4%), or a discontinuation (1.4%).

Table 12: Summary of Baseline Characteristics — ANNEXA-4 Trial

BMI = body mass index; FXa = factor Xa; GI = gastrointestinal; ICH = intracranial hemorrhage; SD = standard deviation.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.

^aCreatinine clearance was estimated according to the Cockcroft-Gault formula.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Concomitant Medications

The use of concomitant medication, specifically coagulation factors, packed RBCs, non-RBC agents, and hemostatic agents, between the start of andexanet alfa bolus and 12 hours after EOI, was measured in the safety population and is summarized in Table 13. In the safety population, 24.7% received non–study-prescribed blood products and/or hemostatic drugs during this study period. The proportions of patients who received coagulation factor transfusions, hemostatic treatments, other coagulation products, or packed RBCs were 4.4%, 1.3%, 1.0%, and 20.3%, respectively.

Table 13: Concomitant Medications of Interest — ANNEXA-4 Trial (Safety Population)

EOI = end of infusion; PCC = prothrombin complex concentrate; RBC = red blood cell.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Protocol Deviations

Major protocol deviations were reported in 59.1% of patients in the safety population. Major protocol deviations were most commonly related to missing or out-of-window laboratory data (35.4%) (e.g., anti-FXa data, hemoglobin level) and andexanet dosing (28.9%) (e.g., incorrect bolus or infusion rate, infusion dose, infusion started later than acceptable).

Efficacy

The efficacy results for the ANNEXA-4 trial are shown in Table 14, unless otherwise specified.

Anti-FXa Activity

Change From Baseline in Anti-FXa Activity (Coprimary End Point)

In the primary analysis, the median percent change from baseline in anti-FXa activity at on-treatment nadir (i.e., minimum value between EOB and EOI) in the efficacy population was −93.3% (95% CI, −94.2% to −92.5%) in apixaban-treated patients and −94.1% (95% CI, −95.1% to −93.0%) in rivaroxaban-treated patients. Subgroup analyses showed results consistent with the primary analysis across bleed types (ICH, GI, other bleeds) (Table 31 in Appendix 1); however, a subgroup analysis of eGFR level is not available.

A sensitivity analysis in patients who were otherwise efficacy-evaluable but who had a baseline anti-FXa level below the prespecified threshold (i.e., 75 ng/mL for apixaban and rivaroxaban, 40 ng/mL for edoxaban, and 0.25 IU/mL for enoxaparin) was planned but not performed. The results in the safety population (which included patients with low baseline anti-FXa levels) were consistent with the primary analyses, with a median percent change from baseline in anti-FXa activity at on-treatment nadir in apixaban-treated patients of −92.6% (95% CI, −93.1% to −91.9%) and −92.9% (95% CI, −93.7% to −90.7%) in rivaroxaban-treated patients.

Hemostatic Efficacy

Achievement of Hemostatic Efficacy (Coprimary End Point)

In the primary analysis of hemostatic efficacy, 80.0% (95% CI, 75.3 to 84.1) of the efficacy population achieved good or excellent hemostatic efficacy. The results of the 3 sensitivity analyses (which assessed the effect of including patients rated as “non-evaluable for administrative reasons” in the safety population with a baseline anti-FXa level, and in patients otherwise efficacy-evaluable but whose baseline anti-FXa level was below the prespecified threshold) were consistent with the primary analysis, as shown in Table 32 in Appendix 1. The results of the subgroup analyses were also consistent with the primary analysis across the 2 subgroups of interest (bleed type and eGFR), as shown in Figure 2 in Appendix 1.

Achievement of hemostatic efficacy in patients with ICH with a high risk of hematoma expansion (i.e., in the high-risk ICH population) was an exploratory end point. Results in the high-risk ICH population were consistent with the ICH subpopulation as well as the overall efficacy population.

Rebleeding (Exploratory End Point)

Of the 264 patients in the safety population after the implementation of Protocol Amendment 4 (when rebleeding was added as an exploratory outcome), rebleeding as adjudicated by the EAC (exploratory end point) occurred in 1 patient (0.4%).

Use of Blood Products and Hemostatic Drugs

Use of Non–Study-Prescribed Blood Products and Hemostatic Drugs (Exploratory End Point)

In the safety population, 24.7% received non–study-prescribed blood products and/or hemostatic drugs between the start of andexanet alfa treatment and 12 hours after the end of the infusion. The proportions of patients who received coagulation factor transfusion, hemostatic treatments, other coagulation products, and packed RBC transfusion were 4.4%, 1.3%, 1.0%, and 20.3%, respectively (Table 13).

Number of RBC Units Transfused (Exploratory End Point)

The number of RBC units transfused was recorded on the case report form; however, no quantitative analysis was conducted on the data.

Proportion of Patients Who Received 1 or More RBC Transfusions (Exploratory End Point)

In the safety population, 19.5% (95% CI, 16.0% to 23.3%) of patients received 1 or more RBC transfusions.

Clinical Neurologic Status in ICH Patients

Change From Baseline in mRS Score (Exploratory End Point)

The proportions of patients with ICH in the efficacy population who had an mRS score of 0 to 2 at baseline, 1 hour posttreatment, 12 hours posttreatment, and day 30 were 32.2% (95% CI, 26.4% to 38.5%), 22.6% (95% CI, 16.1% to 30.3%), 23.4% (95% CI, 16.8% to 31.2%), and 35.9% (95% CI, 29.6% to 42.7%), respectively.

Change From Baseline in NIHSS Score (Exploratory End Point)

The mean change from baseline in NIHSS score at 1 hour posttreatment, 12 hours posttreatment, and day 30 was 0.4 (SD = 2.66), 1.0 (SD = 3.58), and −1.0 (SD = 4.96), respectively, in the patients with ICH in the efficacy population.

Change From Baseline in GCS Score (Exploratory End Point)

The mean change from baseline in GCS score at 1 hour posttreatment, 12 hours posttreatment, and day 30 was −0.4 (SD = 1.69), −0.6 (SD = 2.06), and 0.2 (SD = 2.23), respectively, in the patients with ICH in the efficacy population.

ICU Admission, Hospital Length of Stay, HRQoL

ICU admission and duration and HRQoL were not measured in the ANNEXA-4 trial. Hospital length of stay was not an efficacy outcome in the trials but the median was reported to be 10.9 days in the safety population.

Table 14: Key Efficacy Results — ANNEXA-4 Trial

CI = confidence interval; EOI = end of infusion; FXa = factor Xa; ICH = intracranial hemorrhage; NIHSS = National Institutes of Health Stroke Scale; RBC = red blood cell; SD = standard deviation.

Note: Outcomes summarized in this table were noted to be important to patients and clinicians based on input received from patient groups, clinician groups, and the clinical experts consulted by CADTH.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aResults were presented by FXa inhibitor received. Results of edoxaban- and enoxaparin-treated patients were not of interest to this review and, hence, not summarized.

^bThe on-treatment nadir was defined as the minimum value observed from the start of andexanet alfa administration (30 minutes after bolus) to 30 minutes before EOI. If the results of both time points were missing, then it was equal to the baseline value.

^cFor patients with missing values for nadir, change and percent change were imputed as zero. The 95% CI for the median was based on the distribution-free method.

^dAchievement of hemostatic efficacy refers to patients with an “excellent” or “good” hemostatic efficacy result as adjudicated by the end point adjudication committee. Patients adjudicated as non-evaluable for clinical reasons are included in the efficacy population and were considered as having a “poor or none” hemostatic efficacy result. Patients adjudicated as non-evaluable for administrative reasons (n = 7) were excluded.

^eCI based on the exact binomial method.

^fRebleeding was defined as recurrent bleeding from the same (or different) anatomic site in patients within 24 hours of initial andexanet alfa treatment and after achieving initial good or excellent hemostasis, as adjudicated by the end point adjudication committee.

^gCI based on the Fischer exact method.

^hRefers to the use of these treatments between the start of andexanet treatment and 12 hours after the EOI. Treatments included platelets, 4-factor prothrombin complex concentrate, fresh frozen plasma, antithrombin III, human albumin, plasma, vitamin K, packed RBC transfusion, tranexamic acid, carbazochrome sodium sulfonate hydrate, phenylephrine, and other blood or coagulation treatments.

^INIHSS testing and the additional Glasgow Coma Scale assessments were not implemented until Protocol Amendment 4; therefore, the number of patients evaluated was lower for the baseline NIHSS and the later time points for the Glasgow Coma Scale.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Harms

A summary of harms in the ANNEXA-4 trial is shown in Table 15.

Adverse Events

A TEAE was reported in 72.5% of patients, with the most common TEAEs being urinary traction infection (10.5%) and pneumonia (8.2%).

Serious Adverse Events

Serious TEAEs were reported in 41.9% of patients, with the most common being pneumonia (4.2%), respiratory failure (2.5%), and ischemic stroke (2.1%).

Withdrawal Due to Adverse Events

Four patients (0.8%) discontinued treatment due to TEAEs (i.e., sudden death, cerebral vascular accident, acute myocardial infarction, and infusion-related reaction; 1 of each).

Mortality

Death occurred in 17.0% of patients. Cardiovascular and noncardiovascular-related deaths were reported in 12.8% and 3.1% of patients, respectively.

Notable Harms

Thromboembolic events were reported in 10.5% of patients, including cerebrovascular accident (4.6%), DVT (2.5%), myocardial infarction (1.9%), PE (1.0%), and transient ischemic attack (0.4%). Infusion-related reaction was reported in 2 patients (0.4%). There was no report of neutralizing antibodies to FX, FXa, or andexanet alfa.

Table 15: Summary of Harms — ANNEXA-4 Trial (Safety Population)

FX = factor X; FXa = factor Xa; TEAE = treatment-emergent adverse event.

Note: Unless otherwise specified, safety data were presented by MedDRA preferred term.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aPresent in ≥ 4% of patients in the ANNEXA-4 trial.

^bPresent in ≥ 2% of patients.

^cThe survival status of 6 patients could not be determined (3 patients were lost to follow-up, 1 patient withdrew early before study day 30, and 1 patient withdrew consent at hour 12); these patients are included in the analysis and considered to be still alive.

^dThe reasons for the deaths that occurred during the 30-day safety follow-up period were adjudicated by the end point adjudication committee.

^eRefers to the patients who were not adjudicated.

^fAs adjudicated by the end point adjudication committee.

^gOne patient experienced serious infusion-related reactions.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Critical Appraisal

Internal Validity

The open-label single-arm design of the trial is a key limitation to interpreting the estimates of the study. The lack of randomization precludes conclusions on whether any observed effect could be attributed to andexanet alfa alone.⁵⁶ Further, the open-label study design could increase uncertainty in neurologic status outcomes (mRS, NIHSS, and GCS) due to potential reporting bias, given that these scales involve subjective assessment of clinical parameters by the investigators. However, the presence and extent of such bias could not be determined from the trial data alone. The risk of bias due to the open-label design was low for objective outcomes, including anti-FXa activity, need for RBC transfusion (determined based on a study-defined, hemoglobin-based criterion), and need for other blood products and hemostatic drugs (administered as per institutional or local protocol). The concern was similarly low for hemostatic efficacy and rebleeding, which were adjudicated by an independent EAC that was blinded to anti-FXa activity levels of patients based on predefined criteria of objective measures (e.g., CT, MRI, echocardiogram, hemoglobin, hematocrit).

A large number of protocol deviations related to missing or out-of-window laboratory data (35.4%) and andexanet dosing (28.9%) was reported; however, it is unclear whether such deviations had any impact on the study findings.

The study was adequately powered to detect a response rate that has a 95% CI completely above a prespecified threshold of 50% with respect to the coprimary end point of effective hemostasis. The clinical experts consulted by CADTH noted that response rate could vary considerably by bleed type, although they considered the threshold of 50% assumed by the sponsor to be acceptable in the given study population. All end points other than the coprimary end points were exploratory, which preclude definitive conclusions to be drawn from these analyses. Subgroup analyses were performed on both coprimary end points; however, the findings should be considered exploratory due to the lack of sample size consideration, control for multiplicity, and treatment-by-subgroup interaction analysis.

The primary analyses of the coprimary end points were not assessed in the full trial population and instead were conducted in the efficacy population, including patients who had a baseline anti-FXa level above the prespecified threshold and who met the ISTH-based bleeding criteria for inclusion. A total of 130 of 477 enrolled patients (27.3%) were excluded from the efficacy population, most due to a baseline anti-FXa level below the prespecified threshold (19.7%). The sponsor noted that anti-FXa levels below the threshold are considered subtherapeutic and the exclusion of patients with such baseline anti-FXa levels from the efficacy population was to ensure that patients included in the analysis had at least therapeutic anti-FXa levels. The selection of patients with higher baseline anti-FXa levels introduces the potential for bias due to regression to the mean and likely produces an overestimation of the change in anti-FXa levels posttreatment.⁵⁷ The impact of excluding such patients from the primary analysis of hemostatic efficacy was assessed in 2 sensitivity analyses (i.e., in otherwise efficacy-evaluable patients with a baseline anti-FXa level below the threshold, and in the safety population of patients with a baseline anti-FXa level), both of which showed results consistent with the primary analysis. No sensitivity analysis was conducted for the other end points studied in the efficacy population (i.e., mRS, NIHSS, GCS); therefore, the impact of excluding patients with a low baseline anti-FXa level from the primary analysis of these exploratory outcomes is unclear.

Study discontinuation occurred in 17.2% of patients, most commonly as a result of death. The baseline observation carried forward and last observation carried forward approaches that were used for imputation of missing data in the analysis of percent change from baseline in anti-FXa activity describe a hypothetical scenario in which patients would have survived to day 30. The target estimand is unclear, but these approaches likely produced a conservative estimate of percent change in anti-FXa activity that would be observed in clinical practice. There is uncertainty in the neurologic status outcomes, given that there were a lot of missing data at the posttreatment assessment time points (between 12.1% and 56.7% for the mRS and GCS analyses at all time points; 31.5% on day 30 for the NIHSS analysis). No data imputation was performed to account for the missing data. There is a risk of attrition bias, since the patients who remained in the trial were more likely to be those who had a better treatment response, although the magnitude of bias is unclear.

There is some evidence for the reliability of the hemostatic efficacy rating system, although evidence for validity and responsiveness was not identified. Exploratory analyses of the ANNEXA-4 trial suggested there was a correlation between achieving hemostatic efficacy and lower mortality, but not between anti-FXa activity and mortality in andexanet alfa–treated patients;¹⁷ however, these correlations do not describe a causal surrogate mechanism and can be largely explained by patients at high risk of death having poor prognostic factors. As noted in the FDA Guidance for Industry: Expedited Programs for Serious Conditions — Drugs and Biologics, evidence of pharmacological activity alone is insufficient to support a conclusion that a relationship of an effect on a surrogate end point to the effect on a clinical outcome is reasonably likely.⁵⁸ Nonetheless, in the FDA Summary Basis for Regulatory Action for andexanet alfa, the FDA considered reduction of anti-FXa activity by andexanet alfa to be reasonably likely to predict clinical benefits in morbidity and mortality upon consideration of biological plausibility and the magnitude of anti-FXa activity reduction in the studies of healthy volunteers.³⁹

External Validity

The clinical experts consulted by CADTH noted that the inclusion and exclusion criteria of the study in general align with the criteria for selecting candidates for reversal treatment of major bleeding related to FXa inhibition although, in their opinion, the exclusion of patients with a severe ICH (GCS score less than 7, estimated intracerebral hematoma volume above 60 cc) and an expected survival of less than 1 month could result in a study population with a better prognosis than seen in clinical practice. In addition, the clinical experts noted that the study population is younger compared with clinical practice, and it is reasonable to expect that younger patients would have better clinical outcomes. Patients without major bleeding but who require emergency surgery are also expected to be reasonable candidates for andexanet alfa, according to clinical expert input; however, such patients were not included in the study and the treatment effect within this population is unknown.

The dosing criteria of andexanet alfa in the study were consistent with the product monograph. The use of blood products and hemostatic drugs (e.g., RBC transfusion, antifibrinolytics) was generally in line with clinical practice, according to the clinical experts.

The clinical experts noted that while hemostatic efficacy, as assessed using modified ISTH criteria, is a commonly measured outcome in clinical trials of reversal agents for major bleeds, the criteria were adjudicated based on imaging and laboratory findings, which are not patient-important outcomes. The clinical experts expressed that the relevance of anti-FXa activity level is limited from a clinician’s perspective. They noted that anti-FXa level is not routinely measured in clinical practice and anti-FXa level is not a known surrogate marker for bleeding-related outcomes in patients receiving a DOAC. As such, this would favour interpretation of the results from the safety population rather than the efficacy population, which removed patients based on anti-FXa levels. The clinical experts noted that clinical outcomes such as mental and functional statuses and mortality are much more meaningful to patients and clinicians and have more weight in the overall assessment of treatment response. The neurologic status scales used in the study (i.e., GCS, NIHSS, mRS) are routinely conducted in clinical practice to assess the functional status of patients with an ICH and were considered to be relevant, per the clinical experts. The outcomes of interest to the clinical experts, which were not assessed, are ICU admission, hospital length of stay, and HRQoL; hence, the effect of andexanet alfa on these outcomes is unknown. The clinical experts considered the assessment time frame of hemostatic efficacy (12 hours postinfusion), neurologic status scales, and safety (up to 30 days for both) to be informative for describing the treatment effects of andexanet alfa.

This is the only phase III trial to date assessing the efficacy and safety of andexanet alfa in the reversal of an acute major bleed related to treatment with an FXa inhibitor. The absence of direct comparative evidence between andexanet alfa and PCC (the most relevant comparator for andexanet alfa, according to the clinical experts consulted by CADTH) represents a gap in the evidence.

Summary of ANNEXA-A and ANNEXA-R Trials

The CADTH review team, in consultation with the clinical experts, considered the relevance of the pivotal ANNEXA-A and ANNEXA-R trials to be limited, given that the study population does not align with the indicated population of andexanet alfa. The study design and findings are briefly summarized subsequently.

Description of Studies

ANNEXA-A and ANNEXA-R were similarly designed phase III, randomized, double-blind, placebo-controlled trials that aimed to compare andexanet alfa and placebo with respect to reversal of apixaban (ANNEXA-A trial) or rivaroxaban (ANNEXA-R trial) anticoagulation in healthy individuals aged 50 to 75 years with well-controlled, chronic, stable medical conditions who were not actively bleeding or at low risk of abnormal bleeding or thrombosis based on laboratory tests, risk factors, and medical history (ANNEXA-A trial, N = 68; ANNEXA-R trial, N = 80). Patients received either apixaban (ANNEXA-A trial) or rivaroxaban (ANNEXA-R trial) treatment for 4 days before being randomized to either andexanet alfa or placebo. Patients were divided into 2 cohorts (part 1 and part 2) that followed different dosing regimens for andexanet alfa (IV bolus only in part 1; bolus followed by continuous IV infusion in part 2). Note that the results of part 1 are not of interest to this review, since the dose in that cohort does not align with the recommended dose in the product monograph.

Results: Part 2

Efficacy

In part 2, the enrolled patients had a mean age of 59.4 (SD = 7.5) years in the ANNEXA-A trial and 57.3 (SD = 5.16) years in the ANNEXA-R trial. In both trials, the majority of patients were male and white.

The key efficacy results for part 2 of the trials are summarized in Table 16. In both trials, the analyses of the between-arm difference with respect to mean percent change from baseline in anti-FXa activity at nadir after the EOI (primary end point) and mean percent change from baseline in anti-FXa activity at nadir after the EOB (secondary end point) were both in favour of andexanet alfa. As well, all patients in the andexanet alfa arms and no patient in the placebo arms had at least an 80% reduction in anti-FXa activity from baseline at nadir after the EOI (secondary end point).

Harms

In the ANNEXA-A trial, a TEAE was reported in 10 patients (41.7%) in the andexanet alfa arm and 4 patients (50.0%) in the placebo arm. In the ANNEXA-R trial, a TEAE was reported in 11 patients (42.3%) in the andexanet arm and 6 patients (46.2%) in the placebo arm. There were no reports of any serious TEAEs or death in either trial. Study treatment discontinuation due to an adverse event was reported in 1 patient in each arm of the ANNEXA-A trial and in none of the patients in the ANNEXA-R trial. There were no reported deaths in either trial.

Table 16: Key Efficacy Results of ANNEXA-A and ANNEXA-R Trials — Part 2 (Efficacy Analysis Population)

CI = confidence interval; EOB = end of bolus; EOI = end of infusion; FXa = factor Xa; SD = standard deviation.

Note: Baseline is the last assessment obtained before the bolus dose of andexanet alfa or placebo. Only patients with both a baseline and postbaseline assessment are presented in the table. The outcomes summarized in this table were noted to be important to patients and clinicians based on input received from patient groups, clinician groups, and the clinical experts consulted by CADTH.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aFor part 2, nadir after the EOI was defined as the smallest value for anti-FXa activity between the 110-minute time point (10 minutes before the end of the infusion) and the 5-minute time point after the completion of the infusion for each patient.

^bBetween-arm difference was calculated using the Hodges-Lehman estimate of shift method.

^cP value obtained from a 2-sided exact Wilcoxon rank sum test.

^dThe end point has been adjusted for multiplicity.

^eFor part 2, the nadir after the EOB was defined as the smaller value for anti-FXa activity at the + 2 minute or + 5 minute time point after the completion of the andexanet alfa bolus.

^fComparison between arms was conducted using a Fisher exact test.

Sources: ANNEXA-A and ANNEXA-R Clinical Study Reports.^49,50

Infusion-related reaction was reported in 4 patients (16.7%) in the andexanet alfa arm and 2 patients (25.0%) in the placebo arm in the ANNEXA-A trial, and none of the patients in the ANNEXA-R trial. There were no reported thromboembolic events or neutralizing antibodies to FX, FXa, or andexanet alfa in either trial.

Long-Term Extension Studies

No long-term extension studies were identified by the sponsor.

Indirect Evidence

No indirect evidence in the form of a network meta-analysis or matched-adjusted analysis were submitted for this review. Three weighted comparative analyses involving comparisons of IPD were submitted by the sponsor and are presented in the following section.

Studies Addressing Gaps in the Pivotal and RCT Evidence

The contents of this section were informed by materials submitted by the sponsor. The following was summarized and validated by the CADTH review team.

The sponsor identified 3 gaps in the pivotal and RCT evidence:

• lack of comparative evidence versus the treatment used in current clinical practice (i.e., 4F-PCC)

• lack of comparative evidence in the intended patient population

• lack of evidence on health care resource use and mortality outcomes

To address these gaps, the sponsor submitted 3 studies presenting weighted comparative evidence for andexanet alfa in the pivotal trial (ANNEXA-4) versus PCC in real-world clinical practice, and 6 studies summarizing comparative observational evidence of andexanet alfa versus PCC in real-world clinical practice.

Weighted Comparative Observational Evidence: Andexanet Alfa in the Pivotal Trial Versus PCC in Real-World Clinical Practice

Description of Studies

Due to the lack of evidence from RCTs comparing andexanet alfa with PCC, the sponsor submitted 3 published observational analyses comparing IPD between the ANNEXA-4 trial and 3 external cohorts that recorded the use of PCC in real-world clinical practice, including the ORANGE study in the UK, the HHCS in the US, and the German-Wide Multicenter Analysis of Oral Anticoagulant-Associated Intracerebral Hemorrhage Part Two (RETRACE-II) study.^21-23,59,60

All analyses were conducted in patients with major bleeding while receiving apixaban or rivaroxaban; in particular, patients with ICH in the analyses of the ANNEXA-4 trial versus the HHCS and the ANNEXA-4 trial versus the RETRACE-II study. Outcomes of interest to this review included 30-day mortality (included in all studies), hemostatic efficacy, thrombotic events (assessed in the analysis of the ANNEXA-4 trial versus the HHCS), and hematoma expansion, change in ICH lesion volume, 30-day in-hospital mortality, and mRS score (assessed in the analysis of the ANNEXA-4 trial versus the RETRACE-II study). PSM or weighting was used to adjust for potential confounding variables due to the absence of investigator control of treatment assignment in these studies. The key study design features and analysis methods of these studies are summarized in Table 17.

ANNEXA-4 Trial Versus the ORANGE Study

Objective

The objective of this study was to compare the effectiveness of andexanet alfa versus PCC in apixaban- or rivaroxaban-treated adults with a life-threatening or uncontrollable major bleeding event.

Study Selection: ANNEXA-4 Trial Versus the ORANGE Study

IPD from 2 sources were included in the sponsor’s study: the ANNEXA-4 trial and the ORANGE study, the latter of which enrolled patients receiving a range of treatments, including PCC. The ORANGE study was a 3-year (2013 to 2016) prospective cohort trial that collected information from multiple UK hospitals on adult patients who were admitted for a major bleeding episode while on an oral anticoagulant therapy. The results of the ANNEXA-4 trial were based on patients enrolled between April 2015 and May 2018.

Table 17: Key Study Design Features and Analysis Methods Used for Studies Submitted by the Sponsor

4F-PCC = 4-factor prothrombin complex concentrate; AF = atrial fibrillation; BMI = body mass index; BP = blood pressure; DOAC = direct oral anticoagulant; eGFR = estimated glomerular filtration rate; FXa = factor Xa; GI = gastrointestinal; HF = heart failure; HHC = Hartford Health care; HHCS = Hartford Health care study; ICH = intracranial hemorrhage; MI = myocardial infraction; mRS = modified Rankin Scale; SBP = systolic blood pressure; vs. = versus.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aThis outcome is not of interest to this review and the results have not been summarized in this report.

^bResults for 30-day mortality since the analysis was done after failure of the statistical hierarchy (at 30-day in-hospital mortality).

^cThese covariates were selected for inclusion from a list of predefined covariates based on predefined selection methods to include only the most important variable into the propensity score model. The list of predefined covariates considered for inclusion included age, gender, race, weight, body temperature, mean arterial pressure, systolic blood pressure, diastolic blood pressure, apixaban use, rivaroxaban use, reduce dose and last intake of DOAC, indication of DOAC, comorbidities (e.g., hypertension, diabetes dyslipidemia, prior myocardial infarction, congestive heart failure), laboratory values (hemoglobin, sodium, creatinine), Glasgow Coma Scale, National Institutes of Health Stroke Scale, ICH score, CHADS₂ score, HAS-BLED score, initial intracranial hemorrhage volume, and presence of intraventricular hemorrhage and infratentorial bleeding.

Sources: Technical reports for the ANNEXA-4 trial vs. the ORANGE study and the ANNEXA trial vs. the RETRACE-II study,^59,60 and Costa et al. (2022).²²

The ORANGE trial was identified in an SLR and was included in the sponsor’s review because the population of PCC-treated patients had a distribution of bleed type and temporal concordance in recruitment time frame that was similar to the ANNEXA-4 trial, and the standard of care for anticoagulation treatment and reversal or replacement was comparable between the UK and the population of the ANNEXA-4 trial.

To increase comparability across the 2 study populations and reduce the potential heterogeneity of the treatment effects, only IPD from patients treated with rivaroxaban or apixaban and who had available data for all baseline characteristics of interest and 30-day mortality were included; in the ORANGE study, only IPD from patients treated with PCC were included. Thirty-day mortality was the primary outcome of interest in the analysis.

An assessment of similarity between the included studies is presented in Table 18.

Table 18: Assessment of Homogeneity for the ANNEXA-4 Trial Versus the ORANGE Study

FXa = factor Xa; GCS = Glasgow Coma Scale; PCC = prothrombin complex concentrate.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

Source: ANNEXA-4 trial vs. ORANGE study technical report.⁵⁹

Analysis Methods: ANNEXA-4 Trial Versus ORANGE Study

A feasibility assessment was first performed among all patients in the patient cohort and in the ICH, GI bleed, and other major bleeds subgroups to determine whether PSM would be a numerically stable approach. A logistic regression model was used to contrast binary treatment assignment (andexanet alfa or PCC).⁶¹ The model covariates considered to be potential confounders for 30-day mortality and the treatment cohorts included: age, bleed site, and a medical history of AF, hypertension, diabetes, cancer, renal dysfunction, stroke, DVT, PE, coronary artery disease, or transient ischemic attack. A medical history of myocardial infarction or ischemic heart disease was expected to have an effect on 30-day mortality, but neither was included in the model due to insufficient data. Patients in the andexanet alfa group were matched to patients in the PCC group one-to-one using nearest-neighbour matching with replacement. The magnitude of the difference between the baseline characteristics of the 2 groups was calculated before and after matching to determine whether matching improved similarity. Balance between groups was considered successful if the absolute differences between the groups after matching were less than 10.⁶² Propensity scores were not trimmed, as there was overlap in all regions of the propensity score range.

Thirty-day mortality rates for patients receiving andexanet alfa or a PCC were calculated before and after PSM for each treatment group and each bleed type subgroup (ICH, GI bleed, and other major bleeds). Relative risk of 30-day mortality and 95% CIs were calculated for the 2 treatment groups after PSM adjustment. Lastly, because of potential differences in the severity of bleeds within the ICH subgroup, a sensitivity analysis was conducted where such patients were further matched by intracranial compartment: intracerebral, subarachnoid, and subdural bleeds in the ANNEXA-4 trial (no patients had epidural intracranial bleeds), and intracerebral, subarachnoid, subdural, and epidural bleeds in the ORANGE study.

Results: ANNEXA-4 Trial Versus ORANGE Study

After applying additional exclusion criteria to promote comparability between the studies, 322 of the 352 patients (91.5%) from the ANNEXA-4 trial and 145 of the 2,192 patients (6.6%) from the ORANGE study were included in the analysis.

The baseline patient characteristics before and after matching in the analysis of the ANNEXA-4 trial versus the ORANGE study are presented in Table 19 (whole cohort) and Table 33 in Appendix 1 (by bleed type subgroups). After matching, the ESS for the ORANGE cohort was reduced by 39.3%, from 145 to 88. There is no notable imbalance between the cohorts with respect to the covariates selected for inclusion in the propensity score model.

A notable reduction in the ESS of the ORANGE cohort was observed after matching across all bleed type subgroups (reduction ranged from 35.6% to 52.9%). Notable between-cohort differences in history of hypertension and cancer were consistently observed in all subgroups.

Table 19: Baseline Patient Characteristics for Whole Cohort, Before and After Matching — ANNEXA-4 Trial Versus ORANGE Study

AA = andexanet alfa; GI = gastrointestinal; ICH = intracranial hemorrhage; PCC = prothrombin complex concentrate.

Note: The results of the ANNEXA-4 trial were based on patients enrolled between April 2015 and May 2018.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

Source: ANNEXA-4 trial vs. ORANGE study technical report.⁵⁹

Table 20 presents the 30-day mortality before and after PSM in the analysis of the ANNEXA-4 trial versus the ORANGE study in the whole cohort and by bleed type subgroup. The relative risk of 30-day mortality between the ANNEXA-4 cohort and the ORANGE cohort was 0.43 (95% CI, 0.29 to 0.63) in the whole cohort, 0.31 (95% CI, 0.20 to 0.48) in the ICH subgroup, and 0.49 (95% CI, 0.21 to 1.16) in the GI bleed subgroup. In the “other bleeds” subgroup, the relative risk was 1.29 (95% CI, 0.17 to 9.55), although it was based on a low sample size or ESS in both cohorts. Results for the sensitivity analyses, which included further matching by ICH bleed compartments (intracerebral, subarachnoid, and subdural or epidural hemorrhage), were similar to the primary analysis (whole cohort).

Table 20: 30-Day Mortality Before and After Matching — ANNEXA-4 Trial Versus ORANGE Study

AA = andexanet alfa; CI = confidence interval; ESS = effective sample size; GI = gastrointestinal; ICH = intracranial hemorrhage; NA = not applicable; PCC = prothrombin complex concentrate.

Note: Results of the ANNEXA-4 trial were based on patients enrolled between April 2015 and May 2018.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aRefers to the ESS of the PCC (ORANGE) cohort after propensity score matching.

Source: ANNEXA-4 trial vs. ORANGE study technical report.⁵⁹

ANNEXA-4 Trial Versus the HHCS (Only Patients With ICH)

The objective of this study was to compare the effectiveness of andexanet alfa versus PCC in apixaban- or rivaroxaban-treated adults with ICH.

Study Selection

IPD from 2 sources were included in the propensity score overlap–weighting analysis: the ANNEXA-4 trial and the electronic health record data accessed across multiple acute care hospitals within a single health care system (Hartford Health care) in the northeastern US. The selection of the HHCS was not informed by an SLR.

To promote similarity among the patients from the 2 studies that were included in the analysis, the analysis included only patients experiencing an ICH while being treated with apixaban or rivaroxaban and whose ICH was managed with either andexanet alfa (from the ANNEXA-4 study) or 4F-PCC (from the HHCS data gathered from December 1, 2016, to August 30, 2020). In both studies, only adult patients admitted to a US hospital for a radiographically confirmed (CT or MRI) acute ICH, defined as a spontaneous or traumatic bleed in an intracerebral, subdural, or subarachnoid space, who had taken apixaban or rivaroxaban within 24 hours of the bleed were included in the analysis. Exclusion criteria for this analysis included a GCS score of less than 7 upon admission, an intracerebral bleed volume above 60 mL upon index CT or MRI scan, or planned surgery within 12 hours of the index scan.

An assessment of homogeneity between the ANNEXA-4 trial and the HHCS is presented in Table 21.

Table 21: Assessment of Homogeneity — ANNEXA-4 Trial Versus the HHCS

4F-PCC = 4-factor prothrombin complex concentrate; DNR = do not resuscitate; FXa = factor Xa; GCS = Glasgow Coma Scale; GI = gastrointestinal; HHCS = Hartford Health care study; ICH = intracranial hemorrhage; PCC = prothrombin complex concentrate.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

Source: Costa et al. (2022).²²

Analysis Methods

Potential adjustment factors were identified based upon commonly used variables and accepted preselected risk factors for differential andexanet alfa or 4F-PCC exposure and development of the outcome of interest, including demographics, comorbidities, concurrent medications, ICH characteristics, vital signs, neurologic assessments, and laboratory values. The most impactful variables for inclusion in the propensity score model were selected if the variable was at least moderately associated with the outcome (absolute standardized difference [ASD] above 0.2) and not balanced between treatment cohorts, and less than 10% of data were missing. Propensity scores were calculated based on multivariable logistic regression.⁶³ The covariates selected for inclusion in the propensity score model included baseline age; sex; BMI; creatinine clearance; AF as the indication for oral anticoagulation; average systolic blood pressures at admission (above 160 mm hg); medical history of heart failure, diabetes, myocardial infarction, or stroke; concomitant antiplatelet use; time from index scan to reversal agent initiation; time from end of reversal agent administration to repeat scan; traumatic versus spontaneous bleeding; infratentorial region involvement; single- versus multi-compartment bleeding; bleeding in the intracerebral, intraventricular, subdural, or subarachnoid space on index scan and a bleed size greater than 10 mL in volume or 10 mm in thickness on index scan. Since 100% of the patients in this study who received 4F-PCC were treated in the US, “country” could not be included in our propensity score model (nor weighted for). Consequently, estimated propensity scores were used to weight patients for analysis using an overlap-weighting approach. Overlap weighting assigned weights to patients that were proportional to their probability of belonging to the opposing treatment cohort.^64,65 Propensity score model–eligible variables with less than 10% missing data had missing values imputed using a multiple imputation approach based on a fully conditional specification linear regression model utilizing all other available covariates and outcomes.⁶⁶ ASDs were calculated for each variable before propensity score overlap weighting to illustrate the magnitude of the imbalance between cohorts at baseline (an ASD greater than 0.1 was considered to represent a relevant difference).⁶³

The coprimary outcomes for this study were hemostatic effectiveness (“excellent” or “good” versus “poor or none”) and 30-day all-cause mortality. Occurrence of thrombotic events during the first 5 days after administration of the reversal agent was the secondary end point of interest in this review. Excellent or good hemostasis was defined as no greater than a 35% increase in hematoma size from index to the repeat scan taken approximately 12 hours after reversal administration. The repeat scan closest to 12 hours was utilized whenever possible. If no repeat scan was available within 12 hours (plus or minus 5 hours) of reversal administration, then the worst scan within 24 hours was used. If no repeat scan was available within 24 hours, the patient was assumed to have had poor hemostatic effectiveness. Patients whose index or repeat scan could not be accessed due to administrative reasons were excluded from this study. Adjudicated hemostatic effectiveness determinations from the ANNEXA-4 trial were used for all patients included in this study who received andexanet alfa. Patients who received 4F-PCC had their index and repeat scans read by 2 independent clinicians. Hemostatic effectiveness for patients who received 4F-PCC was then established based on the criteria for the ANNEXA-4 trial and after reaching investigator consensus. For the hemostatic effectiveness and mortality and thromboembolism outcomes, the propensity score overlap–weighted ORs were estimated with accompanying 95% CIs. Binomial (logit-link) generalized estimating equations using a robust sandwich estimator were used to calculate ORs.

A sensitivity analysis was performed in which attainment of hemostatic effectiveness (or lack thereof) for patients without a repeat scan within 24 hours was adjudicated based on the clinical judgment of 2 independent investigators after a review of outcomes (e.g., need for unplanned surgery, subsequent administration of reversal agent, mortality), rather than being assumed to represent poor or no effectiveness. Propensity scores were recalculated for the sensitivity analysis. A subgroup analysis restricted to patients with an intracerebral and/or intraventricular hemorrhage was also performed that evaluated the abovementioned outcomes and the absolute change in hematoma volume in millilitres from index to repeat scan. The subgroup analysis used propensity scores based on a logistic regression model, where total bleed volume in millilitres on index scan (continuous) was substituted for bleed sizes above 10 mL/mm. A propensity score overlap–weighted linear generalized estimating equation was used to determine the mean difference between andexanet alfa and 4F-PCC in hematoma volume from index to repeat scan.

Results

A total of 107 of 477 patients (22.4%) from the ANNEXA-4 trial and 95 of 385 patients (24.7%) from the HHCS were included in the analysis. Baseline characteristics before and after weighting in the analysis of the ANNEXA-4 trial versus the HHCS are presented in Table 22. Following propensity score overlap weighting, no notable imbalances were noted in any recorded covariates between the 2 cohorts.

Incidence of hemostatic efficacy, 30-day mortality, and thrombotic events within 5 days of reversal treatment is summarized in Table 23. After weighting, the adjusted OR between andexanet and 4F-PCC was 2.733 (95% CI, 1.163 to 6.421) with respect to achievement of excellent or good hemostatic efficacy, and 0.355 (95% CI, 0.129 to 0.977) with respect to 30-day mortality, both in favour of andexanet alfa. Thromboembolism occurred in 2 patients (1.9%) in the andexanet alfa cohort and 0 patients in the 4F-PCC cohort within 5 days of reversal agent administration.

Table 22: Baseline Patient Characteristics Before and After Propensity Score Overlap Weighting — ANNEXA-4 Trial Versus the HHCS

4F-PCC = 4-factor prothrombin complex concentrate; ASD = absolute standardized difference; HHCS = Hartford Health care study; NA = not applicable; SD = standard deviation.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

Source: Costa et al. (2022).²²

In the sensitivity analysis (wherein patients without a repeat scan within 24 hours had their hemostatic effectiveness adjudicated based on clinical judgment rather than being assumed to represent poor or no effectiveness), the OR in hemostatic efficacy between the ANNEXA-4 cohort and HHCS cohort was 2.19 (95% CI, 0.91 to 5.25).

In the subgroup analysis in patients with a single compartment intracerebral and/or intraventricular hemorrhage, the results for hemostatic effectiveness and 30-day mortality did not suggest a significant difference between the ANNEXA-4 cohort and the HHCS cohort in this subgroup, but estimates did favour andexanet alfa, with a weighted OR of 1.71 (95% CI, 0.36 to 8.11) for hemostatic efficacy and 0.22 (95% CI, 0.04 to 1.41) for 30-day mortality. There were no thromboembolic events within 5 days of reversal treatment administration in either treatment cohort.

Table 23: Comparative Hemostatic Efficacy, 30-Day Mortality, Thrombotic Events After Propensity Score Overlap Weighting — ANNEXA-4 Trial Versus the HHCS

4F-PCC = 4-factor prothrombin complex concentrate; CI = confidence interval; HHCS = Hartford Health care study; NA = not applicable.

Note: 30-day vascular mortality was an outcome of the study but was not of interest to this review and, thus, not summarized.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

Source: Costa et al. (2022).²²

ANNEXA-4 Trial Versus RETRACE-II Study (Only Patients With ICH)

The objective of this study was to compare the effectiveness of andexanet alfa versus PCC in apixaban- or rivaroxaban-treated adults with atraumatic ICH.

Study Selection

IPD from 2 sources were included in the propensity score analysis: the ANNEXA-4 trial and the RETRACE-II study, which is a multicentre retrospective cohort study of patients with ICH treated with oral anticoagulants at 19 tertiary care centres across Germany between January 2011 and December 2015. Selection of the RETRACE-II study was not informed by an SLR.

A subcohort of patients with atraumatic FXa inhibitor–related ICH was created from the ANNEXA-4 trial that included patients with bleeds in locations other than the parenchyma and excluded patients taking edoxaban. Patients from the RETRACE-II study who had been treated with dabigatran and patients with an ICH related to vitamin K antagonists were excluded. Further, patients were excluded if: their last intake of rivaroxaban or apixaban was more than 18 hours before admission, their ICH was secondary to trauma or hemorrhagic transformation of malignancies, their GCS score on admission was less than 7, their initial ICH volume was above 60 mL, or information about abnormal liver function or alcohol abuse was missing. Results of the ANNEXA-4 trial were based on patients enrolled between April 2015 and May 2018.

An assessment of homogeneity between the ANNEXA-4 trial and the RETRACE-II study is presented in Table 24.

Table 24: Assessment of Homogeneity — ANNEXA-4 Trial Versus RETRACE-II Study

GCS = Glasgow Coma Scale; GI = gastrointestinal; ICH = intracranial hemorrhage; mRS = modified Rankin Scale; PCC = prothrombin complex concentrate;

SoC = standard of care.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

Source: ANNEXA-4 trial vs. RETRACE-II study technical report.⁶⁰

Analysis Methods

A collection of potential adjustment covariates included demographics, vital signs, type of DOACs and indication for oral anticoagulants, comorbidities, concomitant medications, laboratory values, and ICH disease characteristics. These variables were identified based on the largest available intersection of variables between both datasets. Only the most important variables that were at least moderately associated with the outcome and not balanced between treatment groups (ASD above 0.1) with less than 10% of data missing were included. A propensity score model was conducted using logistic regression analysis, with treatment as the dependent variable and the selected covariates as independent variables to model the probability of receiving each treatment. Propensity score–adjusted analyses were performed using the IPTW approach. Weights were stabilized by multiplying with the marginal probability of receiving treatment in the respective dataset. Distribution of weights in the treatment groups was evaluated in terms of mean, median, minimum, and maximum to identify indicators of outliers with subsequent trimming at a minimum of 0.1 and a maximum of 10. The balance of variables included in PSM was assessed by calculating the ASDs for the weighted cohorts.

The covariates included in the propensity score model were as follows:

• Hematoma expansion analysis: Prior use of antiplatelets, DOAC indication classified as “other,” initial ICH volume, GCS score, presence of infratentorial bleeding, and presence of intraventricular hemorrhage.

• Mortality analysis: Diastolic blood pressure, history of diabetes, eGFR level, ICH score, GCS score, HAS-BLED score, presence of infratentorial bleeding, and presence of intraventricular hemorrhage.

• mRS analysis: Systolic blood pressure, history of dyslipidemia, eGFR level, ICH score, HAS-BLED score, CHA₂DS₂-VASc score, and hemoglobin level.

The primary outcome was defined as the proportion of patients with hematoma expansion 12 hours postbaseline in the ANNEXA-4 trial and at the first follow-up imaging in the RETRACE-II study. Hematoma expansion was defined as an increase in the ICH lesion volume of greater than 35% from baseline. Secondary outcomes of interest to this review included in-hospital mortality and functional outcome at discharge or at day 30. Data analyses to estimate treatment effects were performed as unadjusted analyses in the unweighted dataset and as IPTW analyses.

Inference of potential treatment effects was based on the weighted analyses. Hazard ratios for in-hospital mortality were estimated from IPTW Cox proportional hazard regression models. In these models, “person-time at risk” ended at the end of hospitalization, day 30, or death, whichever came first. Change in baseline hematoma volume was estimated using IPTW least squares means from a mixed model for repeated measures that included time, treatment, and time by treatment as independent variables.

Preplanned sensitivity analyses were conducted to evaluate the robustness of the analysis, including full PSM. Specifically, a sensitivity analysis was performed that included only patients treated with PCC in the “usual care” (standard of care) group. The sensitivity analyses included: all covariates for the propensity score model, patients with available baseline anti-FXa inhibitor levels or NIHSS values only, all follow-up radiographic measurements (independent of their time post baseline), and time between initial and follow-up imaging. A prespecified subgroup analysis was conducted for the subgroup of patients with PCC treatment in the RETRACE-II study. It was planned to address a secondary objective of 30-day mortality only if statistically significant superiority of andexanet alfa was observed for the primary outcome of 30-day in-hospital mortality (hierarchical testing). Finally, patients who died between hospital discharge and day 30 were assigned a value of 6 (death) in the ANNEXA-4 trial while, in the RETRACE-II study, these patients were analyzed using their mRS at discharge (which is necessarily less than 6). To address this, a sensitivity analysis was performed that used the last available assessment at hospital discharge (instead of a value of 6) in these patients.

Results

A total of 85 out of 352 patients (24.1%) from the ANNEXA-4 trial and 97 out of 168 patients (57.7%) from the RETRACE-II study were included in the analysis. Baseline characteristics (included in the propensity score model) before and after application of IPTW in the overall cohort are presented in Table 25. No notable imbalances were noted in these covariates between the cohorts after IPTW was applied.

The analysis estimates are summarized in Table 26. The relative risk of hematoma expansion between andexanet alfa and standard of care was estimated to be 0.40 (95% CI, 0.20 to 0.78) in favour of andexanet alfa. The estimated HR between andexanet alfa and standard of care for in-hospital mortality was 0.49 (95% CI, 0.24 to 1.04), the between-arm difference in mean mRS score at discharge was −0.47 (95% CI, −1.02 to 0.08), and the relative risk for an mRS score of 4 or above at discharge was 0.910 (95% CI, 0.728 to 1.138). Note that results of 30-day mortality were not reported because the study planned to only investigate 30-day mortality in detail if a significant effect was observed for in-hospital mortality during the first 30 days.

Table 25: Covariates Included in the Propensity Score Model, Before and After Application of IPTW — ANNEXA-4 Trial Versus the RETRACE-II Study

ASD = absolute standardized difference; BP = blood pressure; DOAC = direct oral anticoagulant; eGFR = estimated glomerular filtration rate; ICH = intracranial hemorrhage; IPTW = inverse probability treatment weighting; mRS = modified Rankin Scale; SD = standard deviation; SoC = standard of care.

Note: For categorical variables, P values were based on a chi-square test or Fisher exact test (50% or more of the cells have expected counts of less than 5).

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aDifferences between treatment groups were reported as ASDs.

Source: ANNEXA-4 trial vs. RETRACE-II study technical report.⁶⁰

Results of sensitivity analyses of in-hospital mortality, hematoma expansion, and proportion of patients with an mRS score of 4 or above at discharge were, in general, consistent with the primary analyses (Table 34), except for the sensitivity analysis of hematoma expansion (only patients from the ANNEXA-4 trial with a baseline NIHSS were included), which did not suggest a difference in hematoma change between andexanet alfa and PCC.

Table 26: Outcomes in the Overall Cohort — ANNEXA-4 Trial Versus the RETRACE-II Study

CI = confidence interval; ICH = intracranial hemorrhage; IPTW = inverse probability of treatment weighting; mRS = modified Rankin Scale; SoC = standard of care.

Note: Results of 30-day mortality were not reported because the study plan was to investigate 30-day mortality in detail only if a significant effect was observed on in-hospital mortality during the first 30 days.

The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aHematoma expansion was defined as an increase in the initial ICH lesion volume of greater than 35%.

Source: ANNEXA-4 trial vs. RETRACE-II study technical report.⁶⁰

The results for an a priori–defined subgroup analysis comparing patients treated with andexanet alfa (n = 85) with those treated with PCC (n = 73) are shown in Table 27. IPTW-adjusted analyses showed results consistent with the whole cohort (in favour of andexanet with respect to hematoma expansion, no difference between groups with respect to in-hospital mortality and mRS at discharge or 30 days).

Table 27: Subanalysis of Patients Receiving Andexanet Alfa or PCC — ANNEXA-4 Trial Versus RETRACE-II Study

CI = confidence interval; ICH = intracranial hemorrhage; IPTW = inverse probability of treatment weighting; mRS = modified Rankin Scale; PPC = prothrombin complex concentrate.

Note: The details in the table were taken from the sponsor’s summary of clinical evidence.²⁰

^aHematoma expansion was defined as an increase of the initial ICH lesion volume of greater than 35%.

^bTreatment effect was estimated using the Cochrane-Mantel-Haenszel method.

^cTreatment effect was estimated using Cox proportional hazard regression.

Source: ANNEXA-4 trial vs. RETRACE-II study technical report.⁶⁰

Critical Appraisal

ANNEXA-4 Trial Versus ORANGE Study

The comparator ORANGE study was identified based on an SLR; however, it is unclear whether the SLR was conducted using appropriate study-selection methods (e.g., a priori selection criteria, duplicate independent reviewers for article screening and data extraction, risk of bias assessment) since there was no report on how the SLR was conducted. A justification for selecting the ORANGE study was provided by the sponsor; however, there was no mention of what other studies were identified in the literature search and the reasons for excluding them. Given the limited disclosure on how the SLR was conducted, it is unclear whether the ORANGE study was indeed systematically selected and, thus, there is a risk of selection bias.

Despite applying additional exclusion criteria to promote similarity among the patients included in the analyses, the analyses could not account for several differences with respect to the inclusion and exclusion criteria between studies. First, the definition of major bleeding differs between studies: in the ANNEXA-4 trial, patients with bleeding who had hemodynamic compromise were considered to have met the criteria for a major bleed, but this was not an inclusion criterion in the ORANGE study. Further, while the ANNEXA-4 trial required the bleed to be uncontrolled or life-threatening for enrolment, it was not a requirement in the ORANGE study, although the clinical experts consulted by CADTH commented that the sponsor’s assumption that PCC could be used as a proxy for the severity of the bleed (i.e., any patients who had received PCC were considered to have experienced a life-threatening or uncontrolled bleeding event) was reasonable. Second, patients who received a transfusion of RBCs or FFP were eligible for inclusion in the ORANGE study but not the ANNEXA-4 trial. The differences in these inclusion criteria could mean that the study population differed between the included studies, although it is unclear to what extent. Important differences in the exclusion criteria of the included studies were also not accounted for in the analysis, specifically, patients with an ICH who had a GCS score of less than 7, a hematoma volume above 60 cc, and an expected survival of less than 1 month, were excluded from the ANNEXA-4 trial but not from the ORANGE study. Since patients who meet such criteria typically have a worse prognosis than those who do not, as per input from the clinical experts, it is plausible that the patients in the ANNEXA-4 trial were generally healthier than those in the ORANGE study.

In addition to heterogeneity in the study population, there is some temporal discordance in patient recruitment between studies, and differences between the trials in the study site locations (UK only in the ORANGE study; North America, Europe, and Japan in the ANNEXA-4 trial). However, the impact of these differences on the study results is likely minimal, since the standard of care in these locations is not known to differ considerably during the study time frame, according to the clinical experts.

Covariates included in the propensity score model were balanced between the study cohorts after matching. However, differences in the baseline size or volume of bleeds, blood pressure reading, ventricular involvement for ICH, time from ICH symptoms to CT, medical history of ischemic heart disease and myocardial infarction — all of which were noted by the authors to be major determinants of mortality — were not adjusted for in the model due to insufficient data. Sex, weight, and creatinine clearance level were excluded from adjustment; however, they are considered by the clinical experts consulted by CADTH to be prognostic factors and were considered relevant in the other studies discussed. The lack of adjustment for these potential prognostic factors or treatment-effect modifiers could bias the estimated treatment effect due to unmeasured confounding across treatment groups. It is also unclear if there is any between-cohort difference in exposure to supportive treatments, such as blood transfusions and tranexamic acid, since such information was only available for the ANNEXA-4 cohort.

In the whole-cohort analysis, the ESS of the ORANGE cohort was notably reduced (by 39.3%) after matching, which suggests poor population overlap. Similarly, there was a notable reduction in the ESS of the ICH, GI bleed, and other bleed subgroups (by 35.6%, 49.1%, and 52.9%, respectively). A significant reduction in sample size can contribute to imprecision, increasing uncertainty of the results.

The study assessed 30-day mortality, which is an outcome that is important to patients and clinicians. The comparative efficacy of andexanet alfa and PCC in other relevant clinical outcomes such as mental and functional statuses, ICU admission, length of stay at hospital, and HRQoL were not investigated.

ANNEXA-4 Trial Versus the HHCS

The comparator trial, the HHCS, was not identified based on an SLR and justification for its selection was not provided. Thus, there is a risk of selection bias, given that studies included in the analysis were not identified using a systematic approach.

Patients were selected from the HHCS to mirror the inclusion criteria of the ANNEXA-4 trial; however, patients with an expected survival of less than 1 month and a history of thromboembolism in the past 2 weeks were not excluded from the HHCS, which could potentially result in a non-transportable patient cohort, with a better prognosis and lower risk of thrombosis in the HHCS compared with the ANNEXA-4 trial. Furthermore, anti-FXa assays were not available in most patients in the HHCS cohort, since obtaining these levels was not standard practice. As a result, anti-FXa levels could not be adjusted for or used to assess effectiveness in the present study, and it is not known whether the HHCS cohort and ANNEXA-4 cohort had similar levels of anticoagulation. Differential use of supportive treatments (e.g., RBC transfusion, antifibrinolytics) could also potentially impact clinical outcomes, although it is unclear what proportion of patients in the HHCS received these treatments. No significant between-study difference was identified with respect to outcome definition and assessment. There is a considerable overlap in study time frame between the ANNEXA-4 trial and the HHCS; as well, the ANNEXA-4 cohort included patients in the US only to reduce the potential for bias resulting from differences in standard of care across different countries between the HHCS and ANNEXA-4 cohorts.

Adjustment factors were selected from a list of commonly used variables and accepted preselected risk factors for differential andexanet alfa or 4F-PCC exposure, and the development of the outcome of interest was based on a multivariable analysis. Prespecified criteria were used to narrow down the covariates for inclusion in the propensity score model to those with the most impact on the effect estimates. No notable imbalance in the selected covariates was noted between the cohorts after weighting. The collection of variables included in the adjusted analysis agreed with those discussed by the clinical experts consulted. Nonetheless, there remains potential residual confounding from the less “impactful” covariates that were excluded from adjustment and the unmeasured heterogeneity of patient characteristics between the studies.

The results of the primary analysis of hemostatic efficacy and 30-day mortality were in favour of andexanet alfa; however, a sensitivity analysis of hemostatic efficacy that allows for adjudication of hemostatic efficacy in these patients based on clinical judgment (rather than assuming them to have “poor or none” hemostatic efficacy in the primary efficacy result) showed imprecise results. This leads to uncertainty in the ability to draw a conclusion on the relative effect of andexanet alfa and PCC with respect to hemostatic efficacy. As well, for both hemostatic efficacy and 30-day mortality outcomes, imprecision was similarly noted in the subgroup analyses of patients with an intracerebral and/or intraventricular hemorrhage.

It is noteworthy that change (and percent change) in ICH size from index scan and change in GCS score at 24 hours post reversal and at discharge were prespecified outcomes, in addition to the published outcomes, as per the study protocol. However, these outcomes were not published and no justification was provided; therefore, there is a risk of publication bias.

A limitation with respect to generalizability is that a low dose of 25 units/kg of 4F-PCC was used in the majority of patients (79.3%) in the HHCS cohort, which does not align with standard practice in Canada, according to the clinical experts consulted by CADTH. The clinical experts noted that a higher dose of PCC (minimum 2,000 unit per dose) is commonly used in Canada, and that the use of low-dose PCC has been shown to associate with worse clinical outcomes.

ANNEXA-4 Trial Versus RETRACE-II Study

Similar to the ANNEXA-4 trial versus HHCS analysis, there is a risk of selection bias, given that studies included in this analysis were not selected using a systematic approach.

The study applied additional exclusion criteria to promote heterogeneity in the patients included in the analysis although, unlike the ANNEXA-4 trial, patients with an expected survival of less than 1 month, recent history of thromboembolism, and planned surgery within 12 hours of admission were not excluded from the RETRACE-II study. The recruitment period and study locations differ between the studies (RETRACE-II study, January 2011 to December 2015 in Germany; ANNEXA-4 trial, April 2015 to March 2020 in the US, Europe, and Japan); however, the clinical experts noted that the standard of care has largely remained unchanged in this time frame and is not known to differ significantly between the countries where the studies took place. With respect to end point assessment, the time between the initial and follow-up imaging (at 12 hours posttreatment in the ANNEXA-4 trial; up to 36 hours posttreatment in the RETRACE-II study) and the mRS posttreatment assessment time point (at 30 days in the ANNEXA-4 trial; at discharge in the RETRACE-II study) were different between the studies, although the authors conducted descriptive analyses and linear regression analyses that suggested there was no meaningful relationship between the timing of these posttreatment assessments and the outcomes.

Prespecified criteria were used to identify the most important variable for inclusion in the propensity score model. Some variables were excluded from adjustment due to a low strength of association with the outcome or treatment assignment or due to having more than 10% of the data missing in 1 of the treatments groups, although the results of the sensitivity analyses of the full propensity score model were consistent with the primary analyses for all outcomes, increasing certainty of the findings. However, demographic data, such as age, sex, and weight, were not included in the adjustment model; therefore, there is a high risk of bias due to unmeasured confounding.

In both the whole-cohort analysis and PCC subgroup analysis, imprecision was evident in the point estimates (HR) of in-hospital mortality, with a wide 95% CI, which included the possibility of appreciable benefit and/or worse outcomes. This precludes any firm conclusion regarding the relative effect of andexanet alfa and standard of care (i.e., PCC) on in-hospital mortality. For the outcome of mRS after treatment, no baseline assessment was available in the RETRACE-II study. This limited the ability to control for potential baseline differences in this variable. However, other variables probably associated with the magnitude of mRS (e.g., ICH score, GCS score) were included in the propensity score model to work as a proxy for mRS at baseline.

Clinically relevant outcomes were assessed, including mRS (a measure of disability) and in-hospital mortality, while other outcomes that were of interest to stakeholders such as mental status, ICU admission, length of stay at hospital, and HRQoL were not.

Other Observational Evidence: Andexanet Alfa Versus PCC in Real-World Clinical Practice

The contents of this section were informed by materials submitted by the sponsor. The following was summarized and validated by the CADTH review team.

Description of Studies

The sponsor submitted 6 comparative observational studies comparing andexanet alfa with PCC in real-world clinical practice.^24-29 The study design, inclusion and exclusion criteria, and outcomes of the 6 studies are summarized in Table 28.

Briefly, all 6 studies were multicentre, retrospective chart audits conducted in the US. The patient population for Coleman et al. (2021),²⁴ Dobesh et al. (2022),²⁵ Fermann et al. (2022),²⁶ and Dobesh et al. (2023)²⁷ were sourced from audits of US hospital charts, while the populations for Sutton et al. (2022)²⁸ and Sutton et al. (2023)²⁹ were sourced from US Veterans Affairs databases. In all studies, data were captured from electronic medical records. The eligible patient population comprised adult patients aged 18 years and older hospitalized for FXa inhibitor–related bleeding. Patients were identified through electronic medical records using ICD-10 billing codes,^24-27 or administrative claims and pharmacy dispensation information from the Veterans Affairs Informatics and Computing Infrastructure.^28,29 Fermann et al. (2022) estimated propensity scores using logistic regression to create inverse probability of treatment weights to balance the treatment groups with regards to potential confounders, and matching was conducted using full, nearest-neighbour matching.

Outcomes included in-hospital mortality,^24-27,29 in-hospital 30-day mortality,²⁸ length of hospital stay,^24,25,29 and ICU length of stay.^24,25,29

Coleman et al. (2022), Dobesh et al. (2022), Fermann et al. (2022), and Dobesh et al. (2023) provided descriptive analyses of results, stratified by bleed type for each reversal or replacement agent. The remaining studies presented results for each reversal or replacement agent, irrespective of bleed type. Across all studies, mortality outcomes were reported as counts (n) and frequency (%), and length-of-stay outcomes were reported as median days with IQR or mean days with SD. The Dobesh et al. (2022), Dobesh et al. (2023), and Fermann et al. (2022) studies also performed adjusted multivariate logistic regressions assessing the association between reversal or replacement agent and death. The Sutton et al. (2022) study conducted propensity score–weighted Cox models to explore the relationship between reversal or replacement agent and mortality. Of note, due the descriptive nature of most of the studies, the comparisons between andexanet alfa and PCC are descriptive in intent.

Table 28: Details of RWE Studies Addressing Gaps in Pivotal and RCT Evidence

4F-PCC = 4-factor prothrombin complex concentrate; AF = atrial fibrillation; FXa = factor Xa; GI = gastrointestinal; ICD-10 = International Classification of Diseases, Tenth Revision; ICD-10-CM = International Classification of Diseases, Tenth Revision, Clinical Modification; ICH = intracranial hemorrhage; ICU = intensive care unit; LoS = length of stay; NA = not applicable; RCT = randomized controlled trial; RWE = real-world evidence; VT = venous thromboembolism.

^aIn Dobesh et al. (2022), patients who were treated with both agents were categorized into the group for the reversal agent they received first. In Dobesh et al. (2023), patients treated with both andexanet alfa and 4F-PCC during the hospitalization for their bleeding event were excluded.

Sources: Coleman et al. (2021), Dobesh et al. (2022), Dobesh et al. (2023), Fermann et al. (2022), Sutton et al. (2022), and Sutton et al. (2023).^24-29

Patient Population

Participant Characteristics

The study sample size ranged from 255 to 4,395 patients. The average age of patients across the studies ranged from 65.0 years to 70.1 years. The most commonly used FXa inhibitors documented were apixaban (40% to 84.0%) and rivaroxaban (14.5% to 56%); the other FXa inhibitors documented were enoxaparin (1% to 37.5%) and edoxaban (1% to 6%). Bleed types documented across the studies included GI (40% to 59.9%), ICH (11% to 32.0%), critical compartments (4% to 31%), and traumatic (1% to 11.0%). Across all studies, the reporting of bleed types was mutually exclusive.

Exposure to Interventions

Exposures

The exposure to reversal or replacement agents across the 6 studies is summarized in Table 29. Exposure to andexanet alfa ranged from 11.3% to 48.3% across the studies, while exposure to 4F-PCC ranged from 24.2% to 80.1%. Coleman et al. (2021) also documented the proportion of patients exposed to FFP (30.5%), other reversal or replacement therapies (26.2%) and no therapies (14.5%).²⁴

Concomitant Medications

Concomitant use of reversal agents was reported in some studies.^24,25,28,29 The most commonly used concomitant agent reported in the Dobesh et al. (2022) study included IV fluids (14% to 20%), vitamin K (10% to 17%), FFP (9% to 15%), and packed RBCs (8% to 10%). In the 2 Sutton et al. studies, the most commonly used concomitant reversal agents reported were RBCs (32.9% to 43.1%), vitamin K (10.6% to 44.1%), transfusion (8.2% to 15.9%), plasma (8.0% to 15.3%), and platelets (7.0% to 14.3%).^28,29 While concomitant use of reversal agents was allowed in the Coleman et al. (2021) study such that the percentages of patients receiving reversal agents added to more than 100%, further details with respect to concomitant use were not reported.²⁴

Table 29: Exposure to Reversal or Replacement Agents

4F-PCC = 4-factor prothrombin complex concentrate; FFP = fresh frozen plasma; PCC = prothrombin complex concentrate.

^aAll others: 3-factor PCC, recombinant factor VIIa, activated 4F-PCC, tranexamic acid, vitamin K.

Sources: Coleman et al. (2021),²⁴ Dobesh et al. (2022);²⁵ Dobesh et al. (2023),²⁷ Fermann et al. (2022),²⁶ Sutton et al. (2022),²⁸ and Sutton et al. (2023).²⁹

Results

Efficacy

Results for in-hospital mortality, 30-day hospital mortality, and hospital and ICU lengths of stay are summarized in Table 30.

In-Hospital Morality

In-hospital mortality for those treated with andexanet alfa was 4% in Coleman et al. (2021), 6% in Dobesh et al. (2022) and (2023), and 10.6% in Sutton et al. (2022). In-hospital mortality for those treated with 4F-PCC was 10% in Coleman et al. (2021), 8% in Dobesh et al. (2022), 10.6% in Dobesh et al. (2023), and 25.3% in Sutton et al. (2022). In Fermann et al. (2022), in-hospital mortality among patients treated with andexanet alfa was approximately 2.5% among patients with GI bleeds and ranged from 9.8% to 16.8% among patients with an ICH; in patients treated with 4F-PCC, in-hospital mortality ranged from 3.2% to 6.0% among patients with GI bleeds, and from 14.5% to 24.0% in patients with an ICH.

In the adjusted analysis performed by Dobesh et al. (2022), treatment with andexanet alfa was associated with a lower likelihood of death compared with treatment with 4F-PCC (OR = 0.69; 95% CI, 0.49 to 0.98). In Dobesh et al. (2023), treatment with andexanet alfa was also associated with a lower likelihood of in-hospital mortality compared with 4F-PCC across all bleeds (OR = 0.50; 95% CI, 0.39 to 0.65), GI bleeds (OR = 0.49; 95% CI, 0.29 to 0.81), and ICH (OR = 0.55; 95% CI, 0.39 to 0.76). In the adjusted analysis performed by Fermann et al. (2022), treatment with andexanet alfa was associated with lower odds of in-hospital mortality compared with treatment with 4F-PCC (OR = 0.67; 95% CI, 0.48 to 0.94). In Sutton et al. (2022), treatment with andexanet alfa was associated with a lower hazard of in-hospital mortality compared with treatment with 4F-PCC (HR = 0.31; 95% CI, 0.14 to 0.71).

30-Day In-Hospital Mortality

30-day in-hospital mortality was explored in Sutton et al. (2022) only. The 30-day mortality rate was 20.0% and 32.4% in patients treated with andexanet alfa and 4F-PCC, respectively. Treatment with andexanet alfa was associated with a lower hazard of 30-day mortality compared with treatment with 4F-PCC (HR = 0.54; 95% CI, 0.30 to 0.98).

ICU Length of Stay

For patients who received andexanet alfa, overall median ICU length of stay was 2 days (IQR = 1 to 4 days) in Coleman et al. (2021), 2 days (SD = 3) in Dobesh et al. (2022), and 1 day (IQR = 0 to 4 days) in Sutton et al. (2022). In Sutton et al. (2023), the mean ICU length of stay was 4.0 days (SD = 7.1). For patients who received 4F-PCC, the ICU stay was 3 days (IQR = 2 to 5) in Coleman et al. (2021), 2 days (IQR = 3) in Dobesh et al. (2022), and 2 days (IQR = 0 to 5) days in Sutton et al. In Sutton et al. (2023), the mean length of ICU stay was 5.0 days (SD = 8.6).

Hospital Length of Stay

For patients who received andexanet alfa, the median hospital stay was 5 days (IQR = 3 to 6) in Coleman et al. (2021), 5 days (SD = 5) in Dobesh et al. (2022), and 6 days (IQR = 3 to 10) in Sutton et al. (2022); and in Sutton et al. (2023), the mean length of stay was 11.3 days (SD = 22.8) while, in Fermann et al. (2022), the mean length of stay ranged from 6.4 days (SD = 4.2) to 8.9 days (SD = 6.2). For patients who received 4F-PCC, the median hospital stay was 5 days (IQR = 4 to 7) in Coleman et al. (2021), 5 (SD = 5) days in Dobesh et al. (2022), and 7 (IQR = 4 to 18) days in Sutton et al. (2022). In Sutton et al. (2023), the mean length of stay of hospital stay was 12.2 days (SD = 17.5), while in Fermann et al. (2022), the mean length of hospital stay ranged from 6.3 days (SD = 4.1) to 8.3 days (SD = 6.2).

Table 30: In-Hospital Mortality, 30-Day Hospital Mortality, and Hospital and ICU Lengths of Stay Reported Across the 6 RWE Studies

4F-PCC = 4-factor prothrombin complex concentrate; FFP = fresh frozen plasma; GI = gastrointestinal; ICH = intracranial hemorrhage; ICU = intensive care unit; IQR = interquartile range; LoS = length of stay; NE = not evaluated; NR = not reported; RWE = real-world evidence; SD = standard deviation.

^aOther bleeds were not included in the LoS analysis due to low sample size and variability of types of bleeds.

^bLoS excluded mortality.

^cPresented as mean and SD.

Sources: Coleman et al. (2021),²⁴ Dobesh et al. (2022),²⁵ Dobesh et al. (2023),²⁷ Fermann et al. (2022),²⁶ Sutton et al. (2022),²⁸ and Sutton et al. (2023).²⁹

Critical Appraisal

The nonrandomized comparison of the 6 RWE studies makes interpretation of the efficacy of andexanet alfa compared with 4F-PCC challenging. The retrospective nature of the studies meant that the quality and accuracy of data collection was dependent on the procedures of routine documentation at each institution. Moreover, the use of ICD-10 codes is associated with limitation for the purpose of clinical research.^30,31 Although all 6 studies were multicentre studies, each medical centre may have distinct population characteristics and physician decision-making patterns. As a result, heterogeneity across the participating medical centres cannot be ruled out. Further, the point estimates varied between studies for both treatment groups, which increases the uncertainty that these comparisons are sufficient to provide evidence of efficacy. The influence of an important patient-level factor (e.g., advanced directives, thrombotic complications after reversal agent administration) — which could not be ascertained from the medical records — on the treatment effect of andexanet alfa and 4F-PCC could not be ruled out. Although Dobesh et al. (2022), Dobesh et al. (2023), Fermann et al. (2022), and Sutton et al. (2022) employed covariate adjustment and propensity weighting, there is a possibility of selection bias or residual confounding due to the nonrandomized nature of the study designs. As a result, there is uncertainty around the comparative treatment effects of andexanet alfa compared with 4F-PCC due to selection bias and unmeasured confounding that cannot be entirely ruled out.

Discussion

Summary of Available Evidence

This report summarizes the evidence for andexanet alfa in the reversal of the anticoagulant effect of anti-FXa inhibitors (apixaban and rivaroxaban) in patients with an acute major bleed based on 1 open-label single-arm trial, 2 RCTs, 3 weighted observational studies, and 6 comparative observation studies.

One pivotal trial, ANNEXA-4, met the inclusion criteria for the systematic review conducted by the sponsor. The ANNEXA-4 trial (N = 477) was a phase IIIb and IV, open-label, single-arm trial that aimed to demonstrate a decrease in anti-FXa activity and an increase in the achievement of hemostatic efficacy with andexanet alfa treatment in adult patients with acute major bleeding while receiving an FXa inhibitor treatment. Percent change from baseline in anti-FXa activity to on-treatment nadir and achievement of hemostatic efficacy at 12 hours postinfusion (coprimary end points), rebleeding, use of non–study-prescribed blood products and/or hemostatic drugs, RBC transfusion, change in neurologic status scores in patients with ICH (exploratory end points), and mortality (safety end point) were assessed. At baseline, patients (safety analysis set) had a mean age of 77.9 years (SD = 10.66), with approximately half of them being male and the majority being white. Most patients received apixaban (51.4%) or rivaroxaban (36.5%) anticoagulation and had an ICH (69.0%) or GI bleed (22.9%).

ANNEXA-A and ANNEXA-R were pivotal phase III, double-blind, RCTs comparing andexanet alfa and placebo with respect to the reversal of apixaban (ANNEXA-A trial) or rivaroxaban (ANNEXA-R trial) anticoagulation in healthy volunteers. Outcomes related to anti-FXa activity and harms are summarized in this report.

Given the lack of direct comparative evidence between andexanet alfa and PCC, the sponsor submitted 3 observational studies comparing andexanet alfa with real-world clinical practice using IPD from the ANNEXA-4 trial and external cohorts from the ORANGE study (all bleed types), the HHCS (ICH only), and the RETRACE-II study (atraumatic ICH only) using PSM or weighting. Thirty-day mortality, hemostatic efficacy, mRS, and thrombotic events were assessed. As well, 6 RWE comparisons of andexanet alfa and 4F-PCC data were submitted. All 6 studies were multicentre, retrospective chart audits conducted in the US and sourced from audits of US hospital charts and US Veterans Affairs databases. In-hospital mortality, in-hospital 30-day mortality, length of hospital stay, and ICU length of stay were assessed.

Interpretation of Results

Efficacy

In the ANNEXA-4 trial, the study-defined threshold for hemostatic efficacy at 12 hours after andexanet alfa infusion was achieved in the whole cohort (lower limit of the 95% CI above 50%), and across the subgroups of interest (bleed type and eGFR level). The clinical experts consulted by CADTH considered the results to be supportive of the efficacy of andexanet alfa, although they and the CADTH review team agreed it is difficult to determine whether the rate of hemostatic efficacy observed could be attributed to the effect of andexanet alfa alone, since potential confounding factors could not be accounted for in a noncomparative study. No conclusion could be drawn on subgroup effects (by bleed type and eGFR level) due to the lack of sample size consideration, control for multiplicity, and treatment-by-subgroup interaction analysis. Although hemostatic efficacy (adjudicated based on radiological and laboratory findings) is a common end point in clinical studies of reversal agents of anticoagulants, it is not a patient-important outcome.

Mortality and mental and functional statuses are considered to be more meaningful clinical outcomes to patients, clinicians, and the clinical experts consulted by CADTH. Thirty-day mortality was 17.0% in the study, which the clinical experts noted was in line with the mortality rate for PCC reported in the literature. Reduction in disability was assessed in patients with ICH using clinically relevant instruments, including the mRS, NIHSS, and GCS, which are neurologic scales for functional status, severity of stroke symptoms, and mental status, respectively. Results showed that the neurologic status of patients remained stable up to day 30. However, due to the exploratory nature of these neurologic outcomes, the large amount of missing data, and potential reporting bias associated with the open-label design, no definitive conclusion could be drawn. In addition, the analyses were conducted in the efficacy population, which excluded patients who had a baseline anti-FXa below a prespecified threshold and did not meet the bleeding inclusion criteria; the impact of such exclusion is unclear in the absence of sensitivity analyses in the intention-to-treat population (i.e., safety population). Other relevant outcomes, including occurrence of rebleeding, use of RBC transfusion and non–study-prescribed blood products and/or hemostatic drugs, and median length of stay at the hospital, in general, align with the clinical experts’ clinical experience with PCC.

The coprimary end point of percent change in anti-FXa activity from baseline at on-treatment nadir was met (the 95% CI for the median excluded 0). The reduction in the level of anti-FXa activity was more than 90% in rivaroxaban- and apixaban-treated patients. The clinical experts commented that this level of reduction aligned with their expectations, considering the mechanism of action for andexanet alfa is as a decoy protein for FXa inhibitors, although the clinical relevance of an anti-FXa assay is low in their opinion, since it is not routinely measured in clinical practice. The FDA considers reduction of anti-FXa activity by andexanet alfa to be reasonably likely to predict clinical benefits in morbidity and mortality in the indicated population, although there was some disagreement, since no evidence was provided for the mechanism of action beyond correlations. Anti-FXa activity–related end points were also assessed in the ANNEXA-A and ANNEXA-R trials, with the results favouring andexanet alfa compared with placebo; however, the results are not informative for the purpose of this review, given the trials were conducted in healthy volunteers with no major bleeding, as opposed to the indicated population.

It should be noted that the ANNEXA-4 study was conducted in a highly select population, excluding patients with severe ICH (GCS less than 7, estimated intracerebral hematoma volume above 60 cc) and those with an expected survival of less than 1 month. Patients were also generally younger compared with clinical practice, according to clinical expert input. These factors together suggest that the study population may have a better prognosis compared with patients in clinical practice, according to the clinical experts, limiting the generalizability of the study results. It is also unclear whether andexanet alfa could meet the unmet need expressed by the patient groups for a specific reversal agent for patients requiring emergency surgery while on FXa inhibitor treatment, since these patients, which the clinical experts noted were reasonable candidates for andexanet alfa, were not included in the study. ICU admission and HRQoL were important clinical outcomes, according to the clinical expert and clinician group input; however, they were not assessed in this study. Further, this study is the only clinical trial to date of patients experiencing acute major bleeding while receiving andexanet alfa; the absence of a direct comparison with PCC, the usual care for reversal of FXa inhibitor–related bleeding, represents a gap in the evidence.

In an effort to address this evidence gap, the sponsor submitted 3 weighted comparative observational studies, which suggested that andexanet alfa, compared with PCC, was associated with a lower rate of 30-day mortality in patients with ICH; a higher rate of hemostatic effectiveness in patients with ICH, including those related to atraumatic cause; and no difference in mRS at discharge or at 30 days in patients with ICH related to an atraumatic cause. However, these results are uncertain because of the important limitations they share. This includes potential bias in the study-selection process in the absence of an SLR or due to lack of transparency in the SLR methods used. As well, of the available information, there appeared to be heterogeneity in study designs and patient populations between studies that was not adequately accounted for, resulting in uncertainty in the comparative treatment-effect estimates.

Evidence from the 6 retrospective studies exploring the real-world effectiveness of andexanet alfa suggests that andexanet alfa, compared with PCC, was associated with lower 30-day mortality and shorter lengths of total hospital stay and ICU stay. However, the nonrandomized comparison of the 6 RWE studies makes the interpretation of the efficacy of andexanet alfa compared with 4F-PCC challenging. Again, the lack of an SLR makes it difficult to assess whether the studies selected are representative of the literature as a whole, or a select curated collection. Overall, there is uncertainty around the comparative treatment effects of andexanet alfa compared with 4F-PCC due to selection bias and unmeasured and residual confounding that cannot be entirely ruled out across the 6 studies.

The clinical experts noted to the CADTH team that some published meta-analyses of PCC and andexanet alfa for reversal of DOAC-related bleeding suggest there are similar mortality and thrombosis rates between them,^43,44 although considering that they are naive comparisons, these estimated mortality results are also uncertain due to potential confounding.

Many of the limitations of the clinical evidence mentioned previously were similarly noted by NICE in its review of andexanet alfa. Its guidance document highlighted that there is uncertainty in the cost-effectiveness estimates, based on limitations of the clinical evidence, particularly with respect to the extent of the benefits in ICH of andexanet alfa in mortality and long-term disability. NICE therefore recommended andexanet alfa for routine use of reversal of anticoagulation in GI bleeding, but only in research of ICH. Note that a head-to-head comparison of andexanet alfa with usual care in patients with acute ICH is currently assessed in the ongoing phase IV randomized ANNEXA-I trial and results are forthcoming (estimated completion date is July 30, 2024).

Harms

In the ANNEXA-4 trial, 72.5% of patients experienced at least 1 TEAE and 41.9% of these were serious TEAEs at 30 days. Treatment discontinuation due to a TEAE was low, at 0.8%, and death was reported in 17.0% of patients. Thrombotic events were reported in 10.5% of patients. The clinical experts noted that, in general, there is a concern regarding the prothrombotic risk for reversal agents, which typically manifests as thrombosis within 24 hours after administration; however, they noted that, given results are based on a follow-up duration of 30 days, it is not possible to assess whether andexanet alfa has such properties and, given that other factors could also have influenced the thrombosis rate during this extended time frame. The clinical experts also noted that overall, without a control arm, it is challenging to determine whether the observed adverse events could be attributed to andexanet alfa alone. Nonetheless, the clinical experts did not have major concerns with the safety profile of andexanet alfa and noted that the thrombosis rate in this study was in line with that of PCC reported in the literature. The weighted comparative analysis between the reported thrombotic events in ANNEXA-4 and the HHCS suggested a similarly low frequency of thrombosis in both the andexanet alfa (1.9%) and PCC (0%) cohorts within 5 days posttreatment, although no conclusion can be drawn due to the aforementioned limitations of the study. Harms outcomes were not assessed across the 6 comparative observational studies comparing andexanet alfa with PCC in real-world clinical practice.

Conclusion

One pivotal single-arm, open-label, phase IIIb and IV trial (ANNEXA-4) provided evidence regarding the efficacy and safety of andexanet alfa in the treatment of patients with acute major bleeding while receiving FXa anticoagulation. Results of clinically relevant outcomes, including 30-day mortality, neurologic status, rebleeding, need for blood products and hemostatic drugs, and length of hospital stay, were generally in line with the clinical experts’ clinical experience with PCC. The analyses of hemostatic efficacy and anti-FXa activity also lent support to the ability of andexanet alfa to achieve hemostasis and reduce the anticoagulant activity of FXa inhibitors. However, there was uncertainty in the magnitude of the clinical benefit attributable to andexanet alfa, given the inherent limitations of the noncomparative study design. The generalizability of the study findings was also limited by the younger study population compared with clinical practice, and the exclusion of patients with a severe ICH and patients with an expected survival of less than 1 month. Three weighted comparative observational studies and 6 comparative observational studies assessed the comparative efficacy of andexanet alfa and PCC in real-world clinical practice; however, the results were inconclusive due to important methodological limitations (e.g., selection bias, inadequate adjustment for prognostic factors and effect modifiers) of these studies. The safety profile of andexanet alfa in the pivotal trial appeared to be consistent with PCC, as per clinical expert input. One weighted comparative analysis suggested the frequency of thrombotic events was similarly low for both andexanet alfa and PCC within 5 days posttreatment, although the results are uncertain due to the aforementioned limitations of the analysis.

References

1.PrOndexxya® (andexanet alfa): 200 mg powder for solution for intravenous infusion [product monograph]. Missisauga (ON): AstraZeneca Canada Inc; 2023 Jun 16.

2.PrXarelto® (rivaroxaban): 2.5 mg, 10 mg, 15 mg and 20 mg tablets [product monograph]. Mississauga (ON): Bayer Inc.; 2022 Jun 10.

3.PrEliquis® (apixaban) 2.5 mg and 5 mg tablets [product monograph]. Montreal (QC): Bristol-Myers Squibb Canada Co.; 2019 Oct 7.

4.Beyer-Westendorf J, Förster K, Pannach S, et al. Rates, management, and outcome of rivaroxaban bleeding in daily care: results from the Dresden NOAC registry. Blood. 2014;124(6):955-962. PubMed

5.Held C, Hylek EM, Alexander JH, et al. Clinical outcomes and management associated with major bleeding in patients with atrial fibrillation treated with apixaban or warfarin: insights from the ARISTOTLE trial. Eur Heart J. 2015;36(20):1264-1272. PubMed

6.Granger CB, Alexander JH, McMurray JJ, et al. Apixaban versus warfarin in patients with atrial fibrillation. N Engl J Med. 2011;365(11):981-992. PubMed

7.Patel MR, Mahaffey KW, Garg J, et al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med. 2011;365(10):883-891. PubMed

8.PharmaStat Database. Apixaban and rivaroxaban claims, cost, units, and days data. Q3 2021-Q2 2022 [internal sponsor's report]. IQVIA; 2022.

9.Eikelboom JW, Connolly SJ, Bosch J, et al. Rivaroxaban with or without aspirin in stable cardiovascular disease. N Engl J Med. 2017;377(14):1319-1330. PubMed

10.Xian Y, Zhang S, Inohara T, et al. Clinical characteristics and outcomes associated with oral anticoagulant use among patients hospitalized with intracerebral hemorrhage. JAMA Netw Open. 2021;4(2):e2037438. PubMed

11.Summers RL, Sterling SA. Emergent bleeding in patients receiving direct oral anticoagulants. Air Med J. 2016;35(3):148-155. PubMed

12.Qazi JZ, Schnitzer ME, Côté R, Martel MJ, Dorais M, Perreault S. Predicting major bleeding among hospitalized patients using oral anticoagulants for atrial fibrillation after discharge. PLoS One. 2021;16(3):e0246691. PubMed

13.Øie LR, Madsbu MA, Solheim O, et al. Functional outcome and survival following spontaneous intracerebral hemorrhage: a retrospective population-based study. Brain Behav. 2018;8(10):e01113. PubMed

14.Apostolaki-Hansson T, Ullberg T, Norrving B, Petersson J. Prognosis for intracerebral hemorrhage during ongoing oral anticoagulant treatment. Acta Neurol Scand. 2019;139(5):415-421. PubMed

15.Fernando SM, Qureshi D, Talarico R, et al. Intracerebral hemorrhage incidence, mortality, and association with oral anticoagulation use: a population study. Stroke. 2021;52(5):1673-1681. PubMed

16.DOACs: management of bleeding. Whitby (ON): Thrombosis Canada; 2021: https://thrombosiscanada.ca/wp-uploads/uploads/2021/11/23.-DOACs-Management-Bleeding_29November2021-1.pdf. Accessed 2023 May 2.

17.Milling TJ, Jr., Middeldorp S, Xu L, et al. Final study report of andexanet alfa for major bleeding with factor Xa inhibitors. Circulation. 2023. PubMed

18.Siegal DM, Curnutte JT, Connolly SJ, et al. Andexanet alfa for the reversal of factor Xa inhibitor activity. N Engl J Med. 2015;373(25):2413-2424. PubMed

19.Clinical Study Report 14-505 (ANNEXA-4). Prospective, open-label study of andexanet alfa in patients receiving a factor Xa inhibitor who have acute major bleeding (ANNEXA-4) [internal sponsor's report]. San Francisco (CA): Portola Pharmaceuticals, Inc.; 2020.

20.Sponsor's summary of clinical evidence [internal sponsor's report]. In: Drug Reimbursement Review sponsor submission: Ondexxya (andexanet alfa) 200 mg powder for solution for intravenous infusion. Mississauga (ON): AstraZeneca Canada, Inc.; 2023 Mar 2.

21.Cohen AT, Lewis M, Connor A, et al. Thirty-day mortality with andexanet alfa compared with prothrombin complex concentrate therapy for life-threatening direct oral anticoagulant-related bleeding. J Am Coll Emerg Physicians Open. 2022;3(2):e12655. PubMed

22.Costa OS, Connolly SJ, Sharma M, et al. Andexanet alfa versus four-factor prothrombin complex concentrate for the reversal of apixaban- or rivaroxaban-associated intracranial hemorrhage: a propensity score-overlap weighted analysis. Crit Care. 2022;26(1):180. PubMed

23.Huttner HB, Gerner ST, Kuramatsu JB, et al. Hematoma expansion and clinical outcomes in patients with factor-Xa inhibitor-related atraumatic intracerebral hemorrhage treated within the ANNEXA-4 trial versus real-world usual care. Stroke. 2022;53(2):532-543. PubMed

24.Coleman CI, Dobesh PP, Danese S, Ulloa J, Lovelace B. Real-world management of oral factor Xa inhibitor-related bleeds with reversal or replacement agents including andexanet alfa and four-factor prothrombin complex concentrate: a multicenter study. Future Cardiol. 2021;17(1):127-135. PubMed

25.Dobesh PP, Coleman CI, Ulloa J, et al. Andexanet alfa is associated with a significant reduction in in-hospital mortality compared with 4F-pCC in a real-world analysis [abstract]. Presented at: International Society on Thrombosis and Haemostasis 2022 Congress; July 12, 2022; London (UK). 2022.

26.Fermann GJ, Dobesh PP, Danese M, et al. Andexanet alfa is associated with reduced in-hospital mortality compared with 4-factor prothrombin complex concentrate among patients with intracranial or gastrointestinal bleeding [abstract]. Presented at: ACEPT Scientific Assembly 2022; Oct 1-4, 2022; San Francisco (CA).

27.Dobesh PP, Coleman CI, Danese M, et al. Andexanet alfa is associated with lower in-hospital mortality compared with 4-factor prothrombin complex concentrate in patients with factor Xa inhibitor–related major bleeding [abstract]. Presented at: International Society on Thrombosis and Haemostasis 2023 Congress; June 24-28, 2023; Montreal (QC).

28.Sutton S, Maganoli J, Cummings T, Dettling T, Hardin J, Lovelace B. Real-world clinical outcomes among US veterans with oral Factor Xa inhibitor-related major bleeding treated with andexanet alfa or 4-factor prothrombin complex concentrate [abstract]. Presented at: Thrombosis & Hemostasis Summit of North America 2022; Aug 18, 2022; Chicago (IL).

29.Sutton S, Maganoli J, Cummings T, et al. Healthcare resource use among US veterans treated with andexanet alfa or 4-factor prothrombin complex concentrate for factor Xa inhibitor-related major bleeding [abstract]. Presented at: International Society on Thrombosis and Haemostasis 2023 Congress; June 24-28, 2023; Montreal (QC).

30.Mainor AJ, Morden NE, Smith J, Tomlin S, Skinner J. ICD-10 coding will challenge researchers: caution and collaboration may reduce measurement error and improve comparability over time. Med Care. 2019;57(7):e42-e46. PubMed

31.Crabb BT, Lyons A, Bale M, et al. Comparison of international classification of diseases and related health problems, tenth revision codes with electronic medical records among patients with symptoms of coronavirus disease 2019. JAMA Netw Open. 2020;3(8):e2017703. PubMed

32.Sikorska J, Uprichard J. Direct oral anticoagulants: a quick guide. Eur Cardiol. 2017;12(1):40-45. PubMed

33.Garcia D, Crowther M. Management of bleeding in patients receiving direct oral anticoagulants. In: Post TW, ed. UpToDate. Waltham (MA): UpToDate; 2023: http://www.uptodate.com. Accessed 2023 Mar 20.

34.Milling TJ, Jr., Clark CL, Feronti C, et al. Management of Factor Xa inhibitor-associated life-threatening major hemorrhage: a retrospective multi-center analysis. Am J Emerg Med. 2018;36(3):396-402. PubMed

35.Canadian stroke best practices: management of spontaneous intracerebral hemorrhage. Toronto (ON): Heart & Stroke Foundation of Canada; 2020: https://www.strokebestpractices.ca/recommendations/management-of-intracerebral-hemorrhage/emergency-management-of-intracerebral-hemorrhage. Accessed 2022 Sep 2.

36.Milling TJ, Jr., Frontera J. Exploring indications for the use of direct oral anticoagulants and the associated risks of major bleeding. Am J Manag Care. 2017;23(4 Suppl):S67-s80. PubMed

37.Andrade JG, Aguilar M, Atzema C, et al. The 2020 Canadian Cardiovascular Society/Canadian Heart Rhythm Society comprehensive guidelines for the management of atrial fibrillation. Can J Cardiol. 2020;36(12):1847-1948. PubMed

38.Abraham NS, Barkun AN, Sauer BG, et al. American College of Gastroenterology-Canadian Association of Gastroenterology Clinical Practice Guideline: management of anticoagulants and antiplatelets during acute gastrointestinal bleeding and the periendoscopic period. Am J Gastroenterol. 2022;117(4):542-558. PubMed

39.Center for Biologics Evaluation and Research. Summary Basis for Regulatory Action. ANDEXXA (andexanet alfa) IV administration. Company: Astrazeneca. BLA 125586/0. Approval date: 05/03/2018 (FDA approval package). Silver Spring (MD): U.S. Food and Drug Administration (FDA); 2018: https://www.fda.gov/media/113954/download. Accessed 2023 Apr 12.

40.Beriplex® P/N 500 / Beriplex® P/N 1000 (human prothrombin complex): powder and solvent for solution for injection [product monograph]. Ottawa (ON): CSL Behring Canada, Inc; 2019 Oct 23: https://labeling.cslbehring.ca/PM/CA/Beriplex-PN/EN/Beriplex-PN-Product-Monograph.pdf. Accessed 2022 Sep 2.

41.Octaplex® (human prothrombin complex): powder and solvent for solution for injection [product monograph]. Toronto (ON): Octapharma Canada, Inc.; 2017 Nov 3: https://pdf.hres.ca/dpd_pm/00041970.PDF. Accessed 2022 Sep 2.

42.Schulman S, Kearon C. Definition of major bleeding in clinical investigations of antihemostatic medicinal products in non-surgical patients. J Thromb Haemost. 2005;3(4):692-694. PubMed

43.Piran S, Khatib R, Schulman S, et al. Management of direct factor Xa inhibitor-related major bleeding with prothrombin complex concentrate: a meta-analysis. Blood Adv. 2019;3(2):158-167. PubMed

44.Chaudhary R, Singh A, Chaudhary R, et al. Evaluation of direct oral anticoagulant reversal agents in intracranial hemorrhage: a systematic review and meta-analysis. JAMA Netw Open. 2022;5(11):e2240145. PubMed

45.Idarucizumab (Praxbind®) 2.5 g / 50 mL injection. DTC Update Suppl. 2016;Sept(Suppl 1).

46.Connolly SJ, Milling TJ, Jr., Eikelboom JW, et al. Andexanet alfa for acute major bleeding associated with factor Xa inhibitors. N Engl J Med. 2016;375(12):1131-1141. PubMed

47.Connolly SJ, Crowther M, Eikelboom JW, et al. Full study report of andexanet alfa for bleeding associated with factor Xa inhibitors. N Engl J Med. 2019;380(14):1326-1335. PubMed

48.Demchuk AM, Yue P, Zotova E, et al. Hemostatic efficacy and anti-FXa (factor Xa) reversal with andexanet alfa in intracranial hemorrhage: ANNEXA-4 substudy. Stroke. 2021;52(6):2096-2105. PubMed

49.Clinical Study Report 14-503 (ANNEXA-A). A phase 3 randomized, double-blind, placebo-controlled study in older subjects to assess safety and the reversal of apixaban anticoagulation with intravenously administered andexanet alfa [internal sponsor's report]. San Francisco (CA): Portola Pharmaceuticals, Inc.; 2015.

50.Clinical Study Report 14-504 (ANNEXA-R). A phase 3 randomized, double-blind, placebo-controlled study in older subjects to assess safety and the reversal of rivaroxaban anticoagulation with intravenously administered andexanet alfa [internal sponsor's report]. San Francisco (CA): Portola Pharmaceuticals, Inc.; 2015.

51.Sarode R, Milling TJ, Jr., Refaai MA, et al. Efficacy and safety of a 4-factor prothrombin complex concentrate in patients on vitamin K antagonists presenting with major bleeding: a randomized, plasma-controlled, phase IIIb study. Circulation. 2013;128(11):1234-1243. PubMed

52.Broderick JP, Adeoye O, Elm J. Evolution of the modified Rankin Scale and its use in future stroke trials. Stroke. 2017;48(7):2007-2012. PubMed

53.Harrison JK, McArthur KS, Quinn TJ. Assessment scales in stroke: clinimetric and clinical considerations. Clin Interv Aging. 2013;8:201-211. PubMed

54.von Kummer R, Broderick JP, Campbell BC, et al. The Heidelberg bleeding classification: classification of bleeding events after ischemic stroke and reperfusion therapy. Stroke. 2015;46(10):2981-2986. PubMed

55.Abdoellakhan RA, Beyer-Westendorf J, Schulman S, Sarode R, Meijer K, Khorsand N. Method agreement analysis and interobserver reliability of the ISTH proposed definitions for effective hemostasis in management of major bleeding. J Thromb Haemost. 2019;17(3):499-506. PubMed

56.Evans SR. Clinical trial structures. J Exp Stroke Transl Med. 2010;3(1):8-18. PubMed

57.Barnett AG, van der Pols JC, Dobson AJ. Regression to the mean: what it is and how to deal with it. Int J Epidemiol. 2005;34(1):215-220. PubMed

58.Center for Drug Evaluation and Research. Guidance for industry: expedited programs for serious conditions – drugs and biologics. Silver Spring (MD): U.S. Food and Drug Administration (FDA); 2014: https://www.fda.gov/media/86377/download. Accessed 2023 May 17.

59.Propensity score matching technical report: andexanet alfa in the treatment of patients with a life threatnening or uncontrollable major bleeding event. [internal sponsor's report]. St. Albans (UK): Fiecon; 2020.

60.Andexanet alfa versus standard-of-care in patients with intracerebral hemorrhage due to oral anticoagulants: a propensity-score adjusted historical comparison of patients from ANNEXA4 and RETRACE II. Study Report. [internal sponsor's report]. Fort Mill (SC): Xcenda; 2019.

61.Caliendo M, Kopeinig S. Some practical guidance for the implementation of propensity score matching. J Econ Surv. 2008;22:31-72.

62.Mebazaa A, Parissis J, Porcher R, et al. Short-term survival by treatment among patients hospitalized with acute heart failure: the global ALARM-HF registry using propensity scoring methods. Intensive Care Med. 2011;37(2):290-301. PubMed

63.Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46(3):399-424. PubMed

64.Li F, Thomas LE, Li F. Addressing extreme propensity scores via the overlap weights. Am J Epidemiol. 2019;188(1):250-257. PubMed

65.Thomas LE, Li F, Pencina MJ. Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial. JAMA. 2020;323(23):2417-2418. PubMed

66.Sterne JA, White IR, Carlin JB, et al. Multiple imputation for missing data in epidemiological and clinical research: potential and pitfalls. BMJ. 2009;338:b2393. PubMed

Appendix 1: Detailed Outcome Data

Table 31: Percent Change From Baseline in Anti-FXa Activity at On-Treatment Nadir by Bleed Type in the ANNEXA-4 Trial (Efficacy Population)

CI = confidence interval; FXa = factor Xa; GI = gastrointestinal; ICH = intracranial hemorrhage.

Note this table has not been copy-edited.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Table 32: Sensitivity Analyses of Hemostatic Efficacy in ANNEXA-4

CI = confidence interval; EOI = end of infusion; FXa = factor Xa.

^aThe primary analysis was conducted in the efficacy population, which included all patients in the safety population who received any amount of andexanet, met clinical bleeding criteria, and had an anti-FXa level of ≥ 75 ng/mL for apixaban and rivaroxaban, ≥ 40 ng/mL for edoxaban, and ≥ 0.25 IU/mL for enoxaparin. Patients adjudicated as non-evaluable for administrative reasons were excluded.

^bIn this sensitivity analysis, patients adjudicated as non-evaluable for administrative reasons (who were excluded from the primary analysis) were included and imputed as having “poor or none” hemostatic efficacy.

^cPatients in the safety population with baseline anti-FXa level were included in this sensitivity analysis.

Note this table has not been copy-edited.

Source: ANNEXA-4 Clinical Study Report.¹⁹

Table 33: Baseline Patient Characteristics by Bleed Type, Before and After Matching — ANNEXA-4 Trial Versus the ORANGE Study