Drug interactions with apixaban: A systematic review of the literature and an analysis of VigiBase, the World Health Organization database of spontaneous safety reports

Abstract Apixaban, a direct oral anticoagulant, has emerged over the past few years because it is considered to have a low risk of drug‐drug interactions compared to vitamin K antagonists. To better characterize these interactions, we systematically reviewed studies evaluating the drug‐drug interactions involving apixaban and analyzed the drug‐drug interactions resulting in an adverse drug reaction reported in case reports and VigiBase. We systematically searched Medline, Embase, and Google Scholar up to 20 August 2018 for articles that investigated the occurrence of an adverse drug reaction due to a potential drug interacting with apixaban. Data from VigiBase came from case reports retrieved up to the 2 January 2018, where identification of potential interactions is performed in terms of two drugs, one adverse drug reaction triplet and potential signal detection using Omega, a three‐way measure of disproportionality. We identified 15 studies and 10 case reports. Studies showed significant variations in the area under the curve for apixaban and case reports highlighted an increased risk of hemorrhage or thromboembolic events due to a drug‐drug interaction. From VigiBase, a total of 1617 two drugs and one adverse drug reaction triplet were analyzed. The most reported triplet were apixaban—aspirin—gastrointestinal hemorrhage. Sixty‐seven percent of the drug‐drug interactions reported in VigiBase were not described or understood. In the remaining 34%, the majority were pharmacodynamic drug‐drug interactions. These data suggest that apixaban has significant potential for drug‐drug interactions, either with CYP3A/P‐gp modulators or with drugs that may impair hemostasis. The most described adverse drug reactions were adverse drug reactions related to hemorrhage or thrombosis, mostly through pharmacodynamic interactions. Pharmacokinetic drug‐drug interactions seem to be poorly detected.


| INTRODUC TI ON
Direct oral anticoagulants (DOACs) act by direct inhibition of coagulation factor II (thrombin) or factor Xa, 1,2 in contrast with heparin or vitamin K antagonists (VKAs). DOACs have emerged over the past few years from the need for a new generation of oral anticoagulants with a more predictable and safer pharmacological profile and more suitable for long-term use. They have become an alternative to VKAs, the only drugs available for long-term anticoagulation for decades.
DOACs have several advantages over other types of anticoagulants: rapid onset and offset of action, a wide therapeutic window and a predictable anticoagulant response that allows fixed doses and eliminates the need for routine monitoring. Moreover, they are considered to be at low risk of drug-drug interactions (DDIs) and fooddrug interactions compared to VKAs. 2,3 Concerning safety, DOACs have been associated with a lower risk of intracranial hemorrhage compared to VKAs and to sequential treatment with low-molecular-weight heparin (LMWH) and VKAs, regardless of their therapeutic indication. 4 There is evidence suggesting a lower mortality risk after suffering a major hemorrhage in patients under DOACs than in patients taking VKAs or LMWH-VKAs, 5,6 but conversely, DOACs are associated with a higher risk of gastrointestinal hemorrhage. 7,8 Currently, there are five DOACs approved for use worldwide: an oral direct thrombin inhibitor, dabigatran, 9 and four oral direct factor Xa inhibitors: rivaroxaban, apixaban, edoxaban, and betrixaban. 10 Apixaban is used for the prevention of atrial thromboembolic events in patients with nonvalvular atrial fibrillation and venous thromboembolism (VTE) recurrence and prevention in major orthopedic surgery and for the treatment of acute VTE. 11 In patients with atrial fibrillation (AF), apixaban was superior to warfarin in the prevention of stroke or systemic embolism. 12 For the treatment of acute VTA, apixaban was noninferior to enoxaparin combined with warfarin. 13 Overall, the results from the three ADVANCE trials showed a higher efficacy of apixaban than enoxaparin in the prevention of VTE after total hip or knee replacement. [14][15][16] Small to modest effects in the pharmacokinetic/pharmacodynamic (PK/PD) profile of apixaban were observed in relation to sex and age, thus considered of no clinical relevance. No dose adjustments are therefore recommended for apixaban regarding sex or age alone. 11,17 Apixaban exposure increased by 30% in the low-body-weight group and decreased by 20% in the high body weight group when compared with a reference weight group. The magnitude of these changes was not considered clinically meaningful either, and no dose adjustment based on body weight alone is recommended. 18 However, a dose reduction is recommended for patients with a body weight < 60 kg and age > 80 years or serum creatinine > 1.5 mg/dL. 11 Likewise, apixaban exposure was not significantly modified by mild and moderate hepatic impairment (Child-Pugh A and B, respectively), but apixaban is contraindicated The half-life of apixaban is 8-15 h and it is metabolized by cytochrome P450 (CYP) 3A and is a P-glycoprotein (P-gp) substrate.
Apixaban is therefore at risk of DDIs with CYP3A/P-gp inhibitors and inducers. 19,20 The overall objective of this study was to evaluate DDIs involving apixaban by a review of the current published data available in the literature and by a real-life assessment of the data on apixaban interactions from VigiBase, the WHO (World Health Organization) global database of individual case safety reports (https://www.whoumc.org). 21

| Literature search
To select relevant publications, we applied the eligibility criteria described in Table 1 Finally, in Google Scholar, the keywords rivaroxaban OR apixaban OR dabigatran OR edoxaban AND interaction OR interactions AND « case report » were applied.
The reference managing software Zotero® (version 5.0.47) removed duplicates, and two reviewers screened the title and abstract of the remaining records for potential relevance. If more than one article described a single study and each presented the same data, the most recent one was included. Articles were split into two groups: interaction studies and case reports.
The verification process was performed by reviewing the SmPC (Summary of Product Characteristics), 11 UpToDate-Lexicomp, 23 the For interaction studies, the types of interactions assessed were PK interactions mediated by CYP3A and P-gp modulators or gastric pH modifiers and PD interactions mediated by other antithrombotic agents and nonsteroidal anti-inflammatory drugs (NSAIDs).
Interactions not matching any of the previous categories were pooled into an additional category called "other drugs".
Data from these study were classified into in vitro/animal studies or phase I to phase IV human studies. Each study was reviewed and described individually. Moreover, each DDI described in an included study was compared with those described in the SmPC. This post hoc analysis allowed us to assess if some DDI were missing and if the SmPC included all data described in the literature.
For case reports, information collected (when available) was the following: patient characteristics, information on apixaban (dosage, start and end of treatment, duration of treatment) and potential interacting drugs, ADR description, and list of additional medication.
A review of the list of potential interacting drugs was then performed by checking the SmPC, UpToDate-Lexicomp,the table of cytochrome P450 substrates and the table of inhibitors and inducers of cytochrome P450 and P-gp. 11,23,24

| Analysis of data from spontaneous reports in VigiBase
To explore DDIs between apixaban and other drugs, we used sponta-  27 Therefore, the Ω value may increase or decrease as new reports enter VigiBase. Ω 0,25 is used as a threshold in the screening of potential DDIs because it is the lower limit of a 95% credibility interval for Ω. Prior to analysis, the dataset was cleaned, first by removing all DDAs with Ω 0,25 less than or equal to 0. Then, some non-relevant MedDRA preferred terms were excluded, such as "condition aggravated" because they are not real ADRs. Similarly, some non-relevant drug names were also excluded, such as "placebo" or "drug name/s under assessment for WHO-DD". Finally, all rows with drugs reported as "concomitant" were removed from the file, therefore only drugs reported as "interacting" or "suspected" were kept. For analysis of the seriousness and the outcome, each ICSR was summarized to only one line, according to the column with the outcomes. We chose to keep the line with the worst outcome (Fatal > not recovered/not resolved > recovering/resolving > recovered/ resolved with sequelae > recovered/resolved > unknown) and seriousness (death > life-threatening > caused/prolonged hospitalization > disabling/incapacitating > congenital anomaly/birth defect > other).
The search and extraction from VigiBase of ICSRs related to apixaban and DDIs was performed by the UMC on 24 April 2018 from a database freeze conducted on the 2 January 2018.
We considered the number of DDA triplets related to each MedDRA system organ class (SOC), the number of DDA triplets for apixaban and one specific ADR and the number of combinations for apixaban-one specific suspected/interacting drug in the DDA triplet. The data for the outcome and the seriousness were extracted and their number was calculated.
We classified the DDIs as linked to the PK or PD mechanism: PK DDIs were further classified as due to absorption (PKA), distribution (PKD), metabolism (PKM), or excretion (PKE) and PD DDIs according to the direct effect at receptor function (PD1), interference with a biological or physiological control process (PD2) or additive/opposed pharmacological effect (PD3). When a DDI was verified for the two mechanisms, they were counted in both. These DDIs were classified according to SmPC, UpToDate, and PubMed.
When more than one mechanism was found, all were listed.
Due to the large quantity of data extracted with the VigiBase analysis, this article focuses on apixaban only.

| Literature
The literature search retrieved 15 interaction studies, some investigating several drugs, and 10 case reports (from nine published articles).
The selection process is illustrated in the PRISMA flowchart ( Figure 1) and Table 2 summarizes the interaction studies.

In vitro
In an in vitro study performed by Sayani et al, apixaban did not interact with tacrolimus when combined into citrated plasma. 28 In another in vitro study performed by Margelidon-Cozzolino et al, three PDE5 inhibitors (sildenafil, tadalafil, and vardenafil) strongly inhibited apixaban efflux by P-gp suggesting potential clinically relevant DDI. 29 The maximal inhibition was higher with vardenafil and sildenafil than with tadalafil. 29 Phase I studies In healthy volunteers, ketoconazole increased apixaban AUC and Cmax by 2-fold and 1.6-fold, respectively. 30 31 Administration of tacrolimus led to a 13% and a 22% decrease in the Cmax and the AUC of apixaban, respectively, but it did not reach statistical significance. 31 Finally, administration of clarithromycin to healthy volunteers led to an increase in the Cmax and the AUC of 30% and 60%, respectively, compared to administration of apixaban alone. 32

Phase III studies
Flaker et al analyzed the influence of amiodarone on the outcomes of the ARISTOTLE trial, which compared apixaban and warfarin for the prevention of stroke or systemic embolism in AF patients. 33 Statistical analysis performed in their study only compared apixaban versus warfarin. Thus, there is no head-to-head comparison for each anticoagulant with or without amiodarone.
Nevertheless, the observed rates for safety endpoints seem to indicate that, in the ARISTOTLE trial, there were no significant differences concerning the incidence of hemorrhagic events for apixaban with or without amiodarone (eg, the major hemorrhage rate for apixaban with amiodarone is 1.86%/year and without amiodarone is 2.18%/year). 33

Phase I studies
In healthy subjects, rifampicin reduced the AUC of apixaban by 54% and the Cmax by 42%. 34

Interaction tested
Reference Type of study Effect observed

| CYP3A and P-gp substrates
Phase I studies The digoxin PK profile was not affected by apixaban co-administration. 35

| Other antithrombotic agents and NSAIDs
Phase I studies A phase I study carried out by Barrett et al showed that enoxaparin did not modify the PK of apixaban. Nevertheless, enoxaparin was associated with an additive increase in the anti-factor Xa activity of apixaban. 36 Combined administration of apixaban and naproxen increased apixaban exposure (54% increase in AUC, 61% increase in Cmax), but led to no clinically relevant prolongation of the bleeding time. 37

Phase II studies
Apixaban was associated with a dose-dependent increase in clinically relevant hemorrhagic events during the APPRAISE trial, a phase II study in patients with recent acute coronary syndrome (ACS) receiving antiplatelet therapy (aspirin alone or with clopidogrel). This increase was more pronounced in patients receiving dual antiplatelet agents than aspirin alone with apixaban. 38

Phase III studies
In the APPRAISE-2 trial, coadministration of apixaban with antiplatelet therapy (aspirin alone or aspirin plus clopidogrel) significantly increased major hemorrhagic events, including fatal and intracranial hemorrhages in high-risk ACS patients. This increase was not associated with a significant decrease in recurrent ischemic events, which is why the trial was terminated prematurely. 39,40 In AF patients, the concomitant use of aspirin and apixaban or warfarin (ARISTOTLE trial) was associated with a higher hemorrhage risk in both groups. However, a similar benefit/risk profile of apixaban vs warfarin remained regardless of concomitant aspirin use. 41

Phase I studies
In healthy subjects, the H 2 antagonist famotidine had no impact on apixaban's PK. 42

| Other drugs
In vitro studies

| Case reports
Ten case reports in nine publications relating to apixaban were found in the literature. [43][44][45][46][47][48][49][50][51] Cases concerned mainly men except for three cases, and the age range was 43-88 years old. Apixaban indication was AF in all cases. Additional pathophysiological factors contributing to the development of the ADR were reported in several cases, the most relevant being renal impairment.
With regard to the mechanism of DDI, five cases were PK interactions, three cases were PD interaction, and two involved both PK and PD interactions. Concerning the PK interaction, two cases were treated with CYP3A and/or P-gp inhibitors and three cases were treated with P-gp and/or CYP3A inducers. For CYP3A/P-gp inhibitors, both case led to a hemorrhage, but one case involved an interaction with diltiazem and the second involved an interaction with diltiazem and amiodarone. 43,44 For the CYP3A and/or P-gp inducers, in the first case, an interaction with carbamazepine was deemed possible, but the apixaban concentrations were still lower than expected after discontinuation of carbamazepine. 47 In another case, apixaban plasma concentration increased fourfold (89 ng/mL to 361 ng/mL) after phenobarbital discontinuation. 51 In the last case of induction, the co-administration of efavirenz with apixaban led to a pulmonary embolism. 50 Two case reports described cardiac tamponade after apixaban and ibrutinib co-administration, caused by a PD interaction. 48,49 The last PD interaction involved an SSRI alone. 46 For the PK/PD interactions, one case involved a selective serotonin reuptake inhibitor (SSRI) that induced platelet dysfunction and CYP34 inhibition ( 45 ) and one case involved both an SSRI (platelet dysfunction) and a CYP3A/P-gp inhibitor. 46  of the cases, the seriousness was reported as "other" (those belonging to none of the other categories) (Figure 2A).

| VigiBase
As illustrated in Figure 2B, the outcome was unknown in a large proportion of the ICSRs (38.7%. n = 528). Ten percent of the cases of the patients were deemed as recovering when the case was reported to VigiBase ( Figure 2B).
Of the 263 DDA triplets reviewed, 179 DDIs were not described in the literature. For the others, a total of 12 PK DDIs, 68 PD DDIs, and 4 PK/PD DDIs were described in the literature. The most common PK DDIs was inhibition of drug metabolism, and the most common PD DDIs was additive pharmacological effect.
Regarding PK DDIs, inhibitors of CYP3A and P-gp were the most reported drugs, and hemorrhagic events were the most reported ADR (Table 3). For PD DDIs, antithrombotic agents and NSAIDs were the most reported drugs, and hemorrhage was the most reported ADR.
Regarding hemorrhage, the most reported site was gastro-intestinal hemorrhage (Table 3).

| D ISCUSS I ON
The arrival of apixaban into routine clinical practice was a major step in anticoagulation therapy due to its alleged favorable profile, which translates into undeniable benefits for patients, especially regarding its ease of use. One of the most relevant aspects of apixaban is its theoretically low potential for interactions with other medications, food, and herbal products. However, phase IV We also compared the ADRs reported in the case reports included in our literature search with those reported in apixaban's SmPC. 11 Hemopericardium and gluteal artery hemorrhage were identified in our case reports but were not specifically mentioned in apixaban TA B L E 3 Drug reported as interacting with apixaban in VigiBase with interaction mechanism and most frequently reported adverse effect Several suspected/interacting drugs were excluded from our analysis of the ICSRs, as they were not documented or understood from a pharmacological perspective as associated with DDIs with DOACs. Additionally, in many DDA triplets, the reported ADR did not seem to correlate with the drug pair, irrespective of whether the drug pair did or did not have an established DDI, such as apixaban-tamsulosin-memory impairment or apixaban-dofetilide-thirst.
We found that the proportion of PD DDIs was higher than the proportion of PK DDIs, suggesting that apixaban might be at higher-risk of interacting with drugs with the same pharmacological profile than with CYP3A4/P-gp inhibitors or inducers. However, this may be a bias, as VigiBase is a database that is dependent on spontaneous ADR reports, and healthcare professionals often know better of PD DDIs. In a study that used this same database, there were more PD DDIs (41%) than PK DDIs (25%). 61 ADR reporting databases, such as VigiBase, have inherent limitations. The two first limitations to mention are underreporting and selective reporting. Another limitation in these databases is the lack of a denominator that allows estimating a risk. Additionally, the available dataset did not allow us to find a plausible explanation for the DDIs. They could be attributed to the heterogeneity of the data stored in VigiBase, which comes from regulatory and voluntary sources and, in some cases, may lack a proper causality assessment, since not all national pharmacovigilance centers contributing to VigiBase perform causality assessments of their ICSRs. 62 Finally, the quality and information contained in an ICSR is limited by the way this ICSR was coded into the database, with crucial data, such as the start or stop date of the drug, often missing. Information available in free text in original reports would also be important because it often contains additional relevant clinical details. 63 This approach entails a detailed case-by-case analysis of ICSRs and largely depends on the completeness of each report because it relies on fields that are not mandatory to be fulfilled for reports to be accepted in VigiBase. 64 To improve drug interaction surveillance in VigiBase, the UMC suggests the use of certain reporting patterns as indicators of DDIs in addition to a positive Ω 0,25 . 65 Other information useful in identifying suspected adverse drug interactions from ICSRs would be a plausible time course, a positive dechallenge and alternative causes of the reaction. 63 Our results have to be interpreted in this light.

| CON CLUS ION
Our analysis shows that apixaban has significant potential for DDIs with other drugs, mostly CYP3A/P-gp inhibitors, CYP3A/P-gp inducers and drugs that may impair hemostasis, such as ASA and NSAIDs, and therefore, a significant number of DDIs with apixaban must be considered by clinicians and patients.
This review of the literature, especially the analysis of reports from VigiBase, notes that pharmacodynamic interactions that occur through the known properties of the drug and that are predictable are widely known and reported. On the other hand, the data analysis shows that the detection and reporting of pharmacokinetic interactions that occur through cytochromes or transporters are sparse because they are badly recognized.
This should motivate clinicians to stay alert on every adverse drug reaction encountered in a patient and to always consider that this adverse drug reaction could also be due to a drug-drug interaction and can be at least partly avoidable.

ACK N OWLED G EM ENTS
The information in this article does not represent the opinion of the Uppsala Monitoring Centre (UMC) or the World Health Organization.
The authors are very grateful to Camilla Westerberg, research pharmacist and Lucie Gattepaille, data scientist, for extracting and providing data from VigiBase. We thank the UMC (Uppsala, Sweden) for its valuable support. This research did not receive any specific grant from funding agencies in the public, commercial, or not-forprofit sectors.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.