Patients with cancer are known to be at higher risk of venous thromboembolism (VTE) than the general population (1, 2). In recent years, several studies have shown that direct-acting oral anticoagulants (DOACs) edoxaban, rivaroxaban, and apixaban are as effective as low-molecular-weight heparin (LMWH) in prevention of recurrent thrombosis. DOACs were generally found to be non-inferior to LMWH (3–6), and are currently an alternative to LMWH for cancer patients (7). Apixaban is a substrate of cytochrome P450 3A4 (CYP3A4) (8) and permeability glycoprotein 1 (P-gp) (9), which makes it susceptible to pharmacokinetic interactions with inducers or inhibitors of these pathways. Such interaction can reduce or increase the plasma concentration of apixaban, respectively. Accordingly, DOACs are not recommended for cancer patients taking anticancer therapies that significantly affect these pathways (10). Less is, however, known about how changes in the plasma concentration translate into clinical outcomes, and how real-world combinations of both inducers and inhibitors of CYP3A4 and/or P-gp affect the efficacy of apixaban.

In addition to drugs with pharmacokinetic interactions, many drugs commonly taken by cancer patients can increase the risk of bleeding or thrombosis in apixaban users through pharmacodynamic interactions, by directly influencing the hemostasis. Examples are the increased risk of bleeding seen with non-steroid anti-inflammatory drugs (NSAIDs) or platelet inhibitors, and the increased thrombosis risk associated with antihormonal treatments like tamoxifen (11, 12).

A systematic review of drug-drug interactions with DOACs found that interactions between amiodarone and dabigatran were the most frequently associated with bleeding events (13). Amiodarone, fluconazole, rifampicin, and phenytoin were found to increase the risk of major bleeding when taken in combination with one of the DOACs dabigatran, rivaroxaban, or apixaban (14). Hanigan et al. reported a trend towards increased bleeding risk in patients taking rivaroxaban or apixaban combined P-gp and CYP3A4 inhibitors like amiodarone, dronedarone, diltiazem, and verapamil (15).

In a post-hoc analysis of the Caravaggio study, which compared dalteparin with apixaban in cancer patients with VTE, no association was found between anti-cancer treatment and the risk of recurrent VTE or major bleeding (16). Likewise, drug-drug interactions did now influence recurrent VTE or major bleeding in a cohort of patients treated for cancer (17).

The aim of the current report is to describe how drugs expected to interact with apixaban influenced recurrent VTE and bleedings in cancer patients participating in a multicenter single-arm interventional study of apixaban (CAP study) (18).

A post-hoc analysis was be performed on data collected from the CAP study, which was originally designed to investigate the safety and efficacy of apixaban for treating VTE in cancer patients. The study was carried out from April 2016 until May 2018, and included 298 patients across nine Norwegian hospitals. The inclusion and exclusion criteria for the study have previously been reported (18). In short, the patients were 18 years or older and either had a diagnosis of cancer or were treated for cancers other than basal cell or squamous cell carcinoma of the skin, had an objectively verified VTE, and did not use strong inhibitors or inducers of CYP3A4 or P-gp. Interacting drugs that were not allowed in the study according to the protocol were voriconazole, posaconazole, itraconazole, idelalisib, several drugs against HIV and hepatitis C, phenytoin, carbamazepine, phenobarbital, primidone, enzalutamide, dabrafenib, and rifampicin. Drugs that were considered moderate interactors or with an unknown degree of interaction were allowed. The patients received apixaban 10 mg twice daily the first week, then 5 mg twice daily for 6 months. If there were indication of continued anticoagulation after 6 months, they received apixaban 2.5 mg twice daily. Clinical endpoints and a complete list of medications taken were registered at inclusion, after 3, 6, and 12 months, and then every 12^th month for a total of 36 months. Ethics approval was obtained from the Norwegian Regional Committees for Medical and Health Research Ethics and the local data protection officer at each participating hospital. The study was carried out in accordance with the principles of the Declaration of Helsinki and Good Clinical Practice.

More than half of the cancer patients treated with apixaban for VTE in the current study used interacting drugs expected to increase the risk of thrombosis as well as bleeding. None of the potentially interacting drugs, or groups of drugs with similar interactions, were associated with recurrent VTE or clinically relevant bleeding. Although fluconazole perhaps increased the risk of CRNMB. Thus, a number of drugs expected to increase the risk of bleeding or thrombosis did not do so in our study. This included NSAIDs, platelet inhibitors, glucocorticoids, and bevacizumab. We found a high level of polypharmacy, with each patient taking a median of 13 different drugs at some time during the 36 month period.

Other studies of interaction with apixaban and other DOACs have focused on specific drugs or families of drugs, and most studies are done in the non-cancer population. In contrast, the current study investigated the clinical effects of interactions with all potentially interacting drugs used by cancer patients. A recent study by Wang et al. (17) looked at drug-drug interactions with DOACs or LMWH in cancer patients using LexiComp® data, grouping patients as either using interacting drugs or not. Like us, they found that many cancer patients used drugs interacting with anticoagulant treatment (42%), but there were no differences in recurrent VTE or bleeding events between patients with or without drug-drug interactions.

A 2013 study by Mueck et al. (23) found a significant increase in AUC and maximum drug concentration in plasma in co-administration of rivaroxaban and CYP3A4/P-gp inhibitors like fluconazole, erythromycin, and clarithromycin (23). In our study, fluconazole seemed to be associated with CRNMB, but not with major bleedings. Interestingly, we see that another study have found the same, apparently contradicting, result between CRNMB and major bleeding (24). We found no convincing effects on clinical outcomes for macrolides. As the number of patients taking macrolides was low, we did not perform analysis on the separate macrolide drugs. Within the group of macrolides, azithromycin is a weak interactor, leading to the possibility that it somewhat masked the interactions of stronger interactors in the macrolide group. A previous study assessed the bleeding risk of older adults taking a DOAC (apixaban, dabigatran or rivaroxaban) and either azithromycin and clarithromycin (25), and found an increased risk of hospital admission because of major bleeding in patients on clarithromycin compared with those taking azithromycin. In our study, none of the 10 patients taking macrolides experienced major bleeding.

There are many previous studies of interaction between DOACs and single drugs, some of those drugs were used by patients in our study. Verapamil and diltiazem have been investigated without finding any interaction with DOACs (15, 26, 27), while our study only had one patient taking each of these drugs, so we were not able to add any new information. The APPRAISE-2 trial found increased bleeding in patients with acute coronary syndromes taking apixaban with aspirin alone or aspirin and clopidogrel compared with placebo (28). The current study had 38 patients taking aspirin and two of those were also taking clopidogrel, but no association with CRNMB or major bleeding was found. A theoretical interaction between aprepitant and DOACs has been described (29), with the potential to cause both increased and decreased effect of anticoagulation, depending on the timing. Our results suggest that, if there is an effect, aprepitant may increase the risk of VTE and reduce the risk of bleeding since no patients taking aprepitant had major bleeding and only two patients had CRNMB, while five patients had recurrent VTE. Bevacizumab has previously been associated with a higher risk of VTE in cancer patients (30), while more recent studies indicate that it can protect against VTE (31). Our results are in line with the latter as none of the 19 patients taking bevacizumab experienced recurrent VTE.

Glucocorticoids are frequently used by cancer patients receiving chemotherapy. They are both associated with increased risk of VTE (32, 33) and with increased gastrointestinal bleeding when taken together with DOAC (34). In our study, 126 of the 298 participants used glucocorticoids. A convincing result was the lack of association with bleeding or recurrent VTE among these patients. The lack of bleeding risk may be related to prophylactic use of proton pump inhibitors.

There are several methodological challenges in investigating clinical effects of drug-drug interactions. One problem is that many drugs are used for short periods of time, e.g., antiemetic drugs, dexamethasone, or antifungal drugs. Thus, the increased risk of an adverse outcome is limited to a few days. We have defined users of interacting drugs as patients who used the interacting drug at any duration of time during the study, not factoring in the time for potential interactions to occur nor the timing in relation to endpoints. A second challenge are the patients who use drugs that can both increase and decrease the efficacy of apixaban and other drugs metabolized through CYP3A4 and P-pg. Thus, the sum of apparently contradicting effects is very difficult to estimate. On the other hand, the data from our study may indicate that polypharmacy could be beneficial, i.e., that the summation of the effects of drugs working opposite on the hemostatic system may reduce the risk of bleeding and recurrent VTE.

The main limitation of the current study is the low number of endpoint events, which resulted in broad CIs of a number of the risk estimates. Thus, there might be drugs increasing the risk of thrombosis or bleeding, where we found no association. Another possible limitation is that there may have been patients in what we have used as the control group, assumed to take no interacting drugs, but still doing so, who we did not consider in our analyses. In addition, we did not have information on intake of food that could possibly interact with apixaban. From the list of potentially interacting drugs, only gemcitabine, dabrafenib, enzalutamide, and tamoxifen were suspected to increase thrombosis risk without any increased bleeding risk. Of the 28 patients taking one or more of these drugs, 20 were also taking one or more drugs that could increase the risk of bleeding, leaving the group of “increased thrombosis risk” with only eight patients. This naturally caused lower confidence in any results from this group, due to the small size.

Apixaban interact with drugs inducing or inhibiting both CYP3A4 and P-gp. It is common among cancer patients to use drugs that potentially interact with apixaban. Despite this concern for drug-drug interaction, drugs with moderate or unknown interacting potential were not found to increase the risk for bleeding or recurrent VTE in cancer patients treated for VTE with apixaban. A possible exception was interaction with fluconazole, which perhaps increases the risk for CRNMB. The results of our study together with the results of previous studies, indicate that drug-drug interactions with drugs of moderate or unknown interacting potential is not a major problem in cancer patients treated with apixaban. We suggest that a possible explanation for this is polypharmacy where drugs with opposite interacting effects on bleeding or thrombosis neutralize each other.