Evaluation of Rivaroxaban- and Dabigatran-Associated Hemorrhagic Events Using the FDA-Adverse Event Reporting System Database (FAERS)

Background Rivaroxaban and dabigatran are non-vitamin K antagonist oral anticoagulants that are Objective To estimate and compare hemorrhagic events reported for rivaroxaban and dabigatran. Setting FDA Adverse Event Reporting System (FAERS) database. Methods The reporting odds ratio (ROR) was used to analyze the reporting of hemorrhagic events. Main outcome measure The overall hemorrhagic events and hemorrhagic events in different physiological systems and indications.


Introduction
Non-vitamin K antagonist oral anticoagulants (NOACs) are increasingly considered as the preferred anticoagulation treatment to prevent stroke for patients with nonvalvular atrial brillation (NVAF) and to prevent and treat deep vein thrombosis (DVT) and pulmonary embolism (PE), particularly for anticoagulant-naive patients [1][2][3]. NOACs have a better risk-bene t pro le than vitamin K antagonists (VKAs) for management of NVAF and are better than low-molecular-weight heparin (LMWH) for treatment and prevention of DVT and PE [4,5]. However, the extensive clinical application of NOACs has raised concerns on bleeding risk. Pivotal e cacy randomized controlled trials (RCTs) have documented an increased risk of gastrointestinal bleeding and a decreased risk of intracranial hemorrhage for NOACs compared with warfarin [6,7].
Rivaroxaban and dabigatran are the most commonly used NOACs, and they have different anticoagulation mechanisms. Rivaroxaban binds reversibly to the factor Xa (the common factor between intrinsic and extrinsic coagulation pathways). A major bene t of using direct factor Xa inhibitors is that they can inactivate prothrombinase activity in addition to clot bound or free factor Xa. Dabigatran can competitively bind to the active site of thrombin (downstream of the coagulation pathway). In doing so, dabigatran prevents thrombin from converting brinogen to its active form of brin, and thereby inhibits thrombus development [8]. Because they act on different coagulation pathway sites, rivaroxban and dabigatran may have different levels of anticoagulant effects and consequently may confer different risks for adverse events (e.g. major bleeding). Estimation of such differences requires an appropriate assessment of the drug exposure and the anticoagulant effect. The current evidence base for the safety of NOACs does not include any head-to-head trials that directly compare the different NOACs [9][10][11].
The U.S. Food and Drug Administration (FDA) Adverse Event Reporting System Database (FAERS) is a spontaneous reporting system and common pharmacovigilance tool that collects case reports of suspected adverse drug reactions. With disproportionality analysis methods, FAERS can be used to assess relative differences in the reporting of common adverse events between case reports listing rivaroxban and reports listing dabigatran. FAERS has been used to evaluate hemorrhage risk associated with dabigatran and/or rivaroxaban [12][13][14]. However, most studies focused on comparing hemorrhagic reporting between warfarin and NOACs. To our knowledge, no FAERS studies have directly compared hemorrhagic reporting between different NOACs. However, this is an important consideration for the clinical choice between NOACs.

Aim of the study
The objective of this study was to compare hemorrhagic event reporting between case reports listing rivaroxaban and reports listing dabigatran in the FAERS database and analyze the reporting of hemorrhagic events overall and hemorrhagic events in different physiological systems and indications.

Ethics approval
The ethics approval for this study was not applicable. However, necessary permission for utilization of FAERS data was obtained.

Outcome
The outcome of interest was hemorrhagic event. The primary outcome was overall hemorrhagic event.
The secondary outcomes were hemorrhagic events in cases affecting the top 5 physiological systems: gastrointestinal (GI), nervous, renal and urinary, skin and subcutaneous tissue, and eye system. At the same time, secondary outcomes also included hemorrhagic events in different indications. The hemorrhagic events were identi ed from case reports listing the interventions of interest by searching the following preferred terms: "bleed", "bleeding", "haemorrhagic", "haematoma", "hemorrhage" and "haemorrhage". Hemorrhagic events in speci c systems were identi ed using MedDRA classi cations of the side effect.

Statistical analysis
Disproportionality analysis was conducted using the reporting odds ratio (ROR) and 95% con dence intervals (CIs) to compare the proportional reporting of hemorrhagic events for rivaroxaban and dabigatran. The ROR was calculated by dividing the odds of hemorrhagic event reporting for the drug of interest by the odds of hemorrhagic event reporting for the comparison drug. The odds of hemorrhagic event reporting for a drug were calculated by dividing the number of drug case reports listing "hemorrhagic events" by the number of drug case reports listing all other adverse events. In this study, rivaroxaban was the drug of interest and dabigatran was the comparison drug. A signal was de ned as ROR ≥ 2.0 [15]. ROR between 1 and 2 is considered as a moderate signal.
Subgroup analyses were performed for the primary outcome by gender, age, year of reporting, and severity of adverse event. The severity of adverse events was classi ed as death, life-threatening, disabling, hospitalization, requiring intervention and others. We additionally evaluated nervous system hemorrhagic events in different reporting years to assess deviations and heterogeneity.

Hemorrhagic event by indications
The RORs (95% CI) for hemorrhagic events in patients with atrial brillation, PE or DVT comparing rivaroxaban with dabigatran were 1.85 (1.79-1.91), 2.02 (1.67-2.47) and 2.17 (1.82-2.59). There were signals for the risk of hemorrhage with the use of rivaroxaban compared with dabigatran in patients with PE or DVT (Fig. 3). However, no signal of increased bleeding risk were found in GI and nervous system under different indications.
Subgroup analysis for hemorrhagic events Across gender and age groups, there were no strong signals of increased hemorrhagic reporting comparing rivaroxaban with dabigatran (Fig. 1). In the subgroup analyses by severity of adverse events, there were signals of increased hemorrhagic event reporting among case reports associated with hospitalization (2.08, 2.01-2.12) and other severity events (3.03, 2.82-3.28).
In the subgroup analysis by year of reporting, compared with dabigatran, rivaroxaban showed a signal of increased hemorrhagic event reporting in 2016 (3.54, 3.31-3.79) (Fig. 1). In order to assess whether the case report year in uenced the reporting of rivaroxaban-and dabigatran-associated neurological hemorrhagic events, secondary analyses of neurological hemorrhagic events were conducted by the year of reporting (Fig. 4). The only ROR (95% CI) greater than 1.0 was for 2014 (1.79, 1.59-2.01). No signals were detected for years 2015 to 2019, which was consistent with the secondary analyses in nervous system.

Discussion
We compared hemorrhagic event reporting between case reports listing dabigatran and reports listing rivaroxaban in the FAERS database including 53,085 hemorrhagic event reports for rivaroxaban and 13,151 hemorrhagic event reports for dabigatran from 2014 to 2019. We used disproportionality analysis to assess the reporting of hemorrhagic events overall and hemorrhagic events in different physiological systems. We found there was a moderate signal for hemorrhage associated with rivaroxaban use compared with dabigatran (ROR 1.58, 1.54-1.62) (Fig. 1). In adverse event cases involving gastrointestinal systems, nervous systems, renal and urinary systems, skin and subcutaneous tissue, or eye systems, hemorrhagic events were not have a higher risk in case reports of rivaroxaban compared to dabigatran (Fig. 2). It is worth noting that the RORs were increased for the risk of hemorrhage with the use of rivaroxaban compared with dabigatran in patients with PE or DVT (Fig. 3).
Hemorrhage is the most common complication of anticoagulants. There are no head-to-head randomized clinical trials to investigate the hemorrhage risk of different NOACs. Up to now, several studies have been conducted to assess the bleeding risk of NOACs using some spontaneous reporting systems (14,(16)(17).
It should be noted that they all use warfarin or all other drugs present in the database for the reference drug to disproportionality analysis. In this study, we directly use rivaroxaban and dabigatran for comparison. Thus making the baseline characteristics of the study population more consistent.
We compared hemorrhagic event reporting between rivaroxaban and dabigatran by different physiological systems. GI hemorrhage and intracranial hemorrhage are the two main adverse reactions that clinicians focus on. A meta-analysis showed rivaroxaban, but not dabigatran, was associated with an increased risk for major GI hemorrhage (RR 1.39; 95% CI, 1.17-1.65 and HR 1.14; 95% CI, 1.04-1.23) [20]. In our study, compared with hemorrhagic event reporting for dabigatran, there was a moderate signal of hemorrhage for rivaroxaban with the ROR (95% CI) in GI was 1.38 (1.34-1.42). For intracranial hemorrhage, some large clinical trials have reported the bleeding event rates per. Rates of intracranial bleeding with the 110-mg dose of dabigatran was 0.23% and the 150-mg dose of dabigatran was 0.30% [6], rates of intracranial bleeding with the rivaroxaban was 0.50% [7]. Since it is not a head-to-head study, no clinical conclusion can be drawn from it. We evaluated nervous system hemorrhage as intracranial hemorrhage. Intracranial hemorrhage consisted of hemorrhagic stroke and subdural or subarachnoid hemorrhage. There was not a signal of hemorrhage for rivaroxaban in nervous system compared with dabigatran with the ROR (95% CI) in nervous system was 0.94 (0.90-0.98). In adverse event cases involving GI systems, nervous systems, renal and urinary systems, skin and subcutaneous tissue, or eye systems, hemorrhagic events were not have a higher risk in case reports of rivaroxaban compared to dabigatran (Fig. 2).
In different indications, we found a increased risk of hemorrhage with the use of rivaroxaban compared with dabigatran in patients with PE or DVT (ROR 2.02, 95% CI: 1.67-2.47 in PE and ROR 2.17, 95% CI: 1.82-2.59 in DVT) (Fig. 3). No direct head-to-head comparisons are available for NOACs in venous thrombo-embolism (VTE) treatment. A network meta-analysis, on the basis of indirect comparison of NOACs for the treatment of VTE, revealed a relative risk for a major or clinically relevant nonmajor bleeding 1.50 (95% CI: 1.17-1.92, p = 0.001) for rivaroxaban versus dabigatran [21]. Those ndings suggest that VTE treatment including DVT or PE with rivaroxaban is associated with more bleeding compared to administration of dabigatran.
Furthermore, we performed subgroup analyses for the primary outcome by conducting the disproportionality analysis by gender, age, year of reporting, and severity of adverse event. For the secondary analyses, there was a moderate signal for rivaroxaban-associated hemorrhage among case reports of patients over 65 years old (ROR 1.95, 1.89-2.02). Our results are consistent with a previous study on elder patients (> 80 years) with nonvalvular atrial brillation [22] which showed dabigatran was associated with a lower risk of major bleeding (HR 0.77, 0.67-0.90) compared with rivaroxaban. For subgroup analyses by reporting year (Fig. 1), the strikingly low ROR for 2014 may be partly due to the increasing number of patients using rivaroxaban. A study based on IMS Health National Disease and Therapeutic Index found that rivaroxaban treatment visits increased from 2011 Q4 to 2014 Q4, while dabigatran treatment visits remained relatively stable since 2011 Q4 [23].
The major strength of this study was to provide a head-to-head assessment of hemorrhagic event reporting between rivaroxaban and dabigatran in large pharmacovigilance database and assessed the association in different physiological systems and indications. Certainly there are several limitations.
First, FAERS data is prone to reporting biases and missing data, and we were unable to fully control for confounding. Second, because the data lacks a meaningful denominator for causal analyses, we cannot assess incidence and make causal inferences from FAERS data. Third, drug-drug interactions documented in "Reaction" variable of FAERS case reports may not represent all potential interactions as drug-drug interaction may not be reported as an adverse event. And we are not restrict to hemorrhagic drug-drug interaction case reports that have drugs of interest listed as primary or secondary suspect. Lastly, we did not consider the effect of drugs dosage due to missing data, and this could be an important factor in bleeding events.

Conclusions
In conclusion, this study found a moderate signal for hemorrhage associated with rivaroxaban compared to dabigatran. The signals for bleeding events varied in subgroup analyses. In reported cases involving GI systems, nervous systems, renal and urinary systems, skin and subcutaneous tissue, or eye systems, we saw no signal of increased bleeding risk in rivaroxaban compared to dabigatran. But when treating patients with PE or DVT, rivaroxaban had a higher bleeding risk compared with dabigatran. Patients with rivaroxaban to treat VTE or PE who experience symptoms related to hemorrhage should be closely monitored and advised to adhere to an appropriate care plan.