The risk of major bleeding was higher with DOACs than LMWH in both the Hokusai-VTE Cancer and SELECT-D trials, although rates of major bleeding were similar in the CARAVAGGIO, ADAM-VTE, and CASTA-DIVA trials (Table 1) ^{[18][19][21][20][22]}[16,17,18,27,28]. Overall, pooled estimates from meta-analyses have reported a non-significantly higher rate of major bleeding complications among patients with CAT receiving a DOAC as compared to LMWH (HR:1.26; 95% CI 0.84–1.90 and RR: 1.36; 95% CI 0.55 to 3.35) ^[22][32][28,38]. Hence, identifying patients at higher risk of bleeding complications might be helpful to tailor anticoagulation in this patient population. In the Hokusai-VTE Cancer trial, the excess bleeding risk was attributable mainly to patients with GI cancer, of whom 12.7% (21/165) in the edoxaban arm experienced major bleeding as compared to 3.6% (5/140) in the dalteparin arm ^[33][39]. Additionally, for most of the edoxaban-treated patients with major bleeding, the site of the bleed was the upper GI tract (16 of 21 cases), with the remaining sites being the lower GI tract, epistaxis, and retro-peritoneum. By contrast, only one of the five cases of major bleeding in the dalteparin-treated patients was at a GI site ^[33][39]. Similarly, in SELECT-D, 45.5% (5/11) of all major bleeding episodes in rivaroxaban-treated patients occurred in the GI tract ^[19][17]. Like Hokusai-VTE Cancer, the SELECT-D trial also showed a signal for a higher risk of bleeding in patients with GI cancer, with the data safety monitoring committee of the SELECT-D trial noting a non-significant increase in major bleeding events in 19 patients with esophageal or gastroesophageal junction cancers after a safety review of the first 220 patients ^[19][17]. Patients with those cancers were subsequently excluded from enrolment. The CARAVAGGIO trial did not report any difference in major bleeding complications between patients receiving apixaban or dalteparin ^[21][34][18,40]. A total of 1.9% (11/576) and 1.7% (10/579) of patients had GI major bleeding complications among those receiving apixaban and dalteparin, respectively ^[34][40]. The reasons for the discrepancy in GI major bleeding are unclear and may be related to differences in baseline characteristics (tumor types, etc.) among the included patients in the different studies or related to the properties of the individual DOACs (once vs. twice a day, topical mucosal anticoagulant effect, etc.). A recent observational study reported that apixaban had a higher rate of major bleeding complications in patients with luminal GI cancers compared to those with non-GI cancers (15.6 vs. 3.7 per 100 person-years, p = 0.004) and compared to enoxaparin in patients with luminal GI cancer (15.6 vs. 3.2, p = 0.04) ^[35][41]. Hence, all DOACs should be use cautiously in patients with GI cancers, especially in those with unresected luminal tumors.

Other patient characteristics are also important for clinicians to consider. Subgroup analyses of major bleeding in the Hokusai-VTE Cancer safety population suggest that, in addition to GI cancer, other features associated with a higher risk of major bleeding include urothelial cancer, renal impairment, thrombocytopenia, intracranial malignancy, regionally advanced or metastatic cancer, recent surgery, and use of antiplatelet agents or bevacizumab ^[18][16]. Analysis of clinically relevant bleeding events in the CATCH trial confirmed that intracranial malignancy increases the risk of bleeding regardless of the type of anticoagulation ^[36][42]. Age > 75 years was also significantly associated with an increased risk of clinically relevant bleeding in this analysis ^[36][42].

The Hokusai-VTE Cancer, SELECT-D, and CARAVAGGIO trials have reported greater proportions of DOAC-treated patients who experienced clinically relevant non-major bleeding (CRNMB) events with HRs of 1.38 (95% CI 0.98 to 1.94), 3.76 (95% CI 1.63 to 8.69), and 1.42 (95% CI 0.88 to 2.30) for edoxaban, rivaroxaban, and apixaban, respectively ^[18][19][21][16,17,18]. In the Hokusai-VTE Cancer trial, CRNMB events were numerically more common for GI, epistaxis, hematuria, or abnormal uterine bleeding in patients receiving edoxaban compared to those receiving dalteparin ^[33][39]. In SELECT-D, the significantly higher rates of CRNMB in patients receiving rivaroxaban were due to GI and GU bleeding, which accounted for 9 and 11 of the 25 CRNMB events, respectively ^[19][17]. In CARAVAGGIO, the numerical increase in CRNMB was largely due to bleeding in the GU and upper airway tracts, which accounted for 20 and 14 cases, respectively, of the 59 CRNMB events in patients receiving apixaban ^[34][40]. Additionally, patients with GI cancer appeared to be at higher risk of bleeding events with DOAC, with 13.2% (19/144) of patients with GI cancer who were treated with apixaban experiencing CRNMB as compared to 4.9% (7/144) in the dalteparin arm ^[34][40]. Overall, GI or GU CRNMB may be more common in patients receiving a DOAC than in those treated with LMWH.

2.4.1. Features Consistent with a High Risk of GI Bleeding

Given that bleeding rates appear to be higher with DOACs in patients with GI tumors or those on treatments such as bevacizumab that are associated with tumor necrosis and bleeding ^{[18][19][33][34]}[16,17,39,40], the committee suggests considering the use of LMWH for patients with these or other features that are associated with a high risk of GI bleeding, such as angiodysplasia, GI lesion, previous variceal bleed, or treatment-associated mucosal toxicity. The risk of GI perforation and/or hemorrhage associated with a patient’s anticancer therapies should be taken into consideration regardless of which anticoagulant is selected.

2.4.2. Thrombocytopenia

Thrombocytopenia increases the risk of bleeding complications in patients with CAT ^[37][43]. Unfortunately, there is limited evidence to guide management in patients with platelet counts <50,000 platelets/mL. The CLOT trial excluded patients with baseline platelet counts <75,000 platelets/mL, while the Hokusai-VTE Cancer, CARAVAGGIO, and SELECT-D trials excluded patients with baseline platelet counts of less than 50,000, 75,000, and 100,000 platelets/mL, respectively ^{[14][18][19][21]}[8,16,17,18]. Guidance from the SSC of the ISTH suggests that therapeutic dose of anticoagulation can be used for patients with platelet count of ≥50,000 platelets/mL ^[38][44]. In patients with platelet counts of less than 50,000 platelets/mL, 50% or prophylactic dose LMWH may be used or full-dose anticoagulation with platelet transfusion support may be considered ^[38][44]. The Canadian consensus committee suggests that LMWH is preferred in these patients but recommends seeking an expert opinion from a specialized physician when initiating anticoagulation in the setting of severe thrombocytopenia (i.e., platelet counts <50,000/mL). In cases of transient thrombocytopenia due to anticancer therapies, clinical judgment should be used to determine whether the anticoagulant needs to be dose-reduced or temporarily held until platelet levels recover to ≥50,000 platelets/mL.

2.4.3. Intracranial Lesions

As there were few patients with intracranial tumors (primary brain tumor or brain metastasis) included in the DOAC trials (none in CARAVAGGIO, 7% of patients (74/1046) in Hokusai VTE Cancer, and only 1% of patients in SELECT-D), there are limited data regarding the safety of this anticoagulant class in these patients ^[18][19][21][16,17,18]. Some reassurance may be provided by a retrospective cohort study of the cumulative incidence of intracranial hemorrhage (ICH) with DOAC vs. LMWH in patients with brain tumors and VTE ^[39][45]. In this study, no ICH was noted among 20 patients with primary brain tumors treated with DOACs, while the cumulative incidence among the 47 patients treated with LMWH was 37%. Among 105 patients with brain metastases, the cumulative incidence of ICH was 11% among those treated with DOAC and 18% in those treated with LMWH ^[39][45]. Similarly, an international two-center study suggested comparable safety of LMWH and DOACs in patients with brain metastases. The 12-month cumulative incidence of major ICH was 5.1% in DOAC-treated patients and 11.1% in those treated with LMWH (HR: 0.45; 95% CI 0.09 to 2.21) ^[40][46]. When anticoagulation was analyzed as a time-varying covariate, the risk of any ICH did not differ between DOAC- and LMWH-treated patients (HR: 0.98; 95% CI 0.28 to 3.40) ^[40][46]. Finally, a single-center retrospective chart review of 125 patients with primary and metastatic brain tumors on anticoagulation reported rates of major bleeding of 26% and 9.6% in patients receiving LMWH or DOAC, respectively ^[41][47]. Patients receiving DOAC also had a lower rate of ICH compared to those receiving LMWH (5.8% vs. 15%) ^[41][47]. Nevertheless, given the small numbers and the limitations of retrospective studies, as well the shorter half-life of LMWH, the consensus committee suggests considering the initial use of LMWH for patients with CAT and high-risk intracranial lesions (e.g., glioma).

2.4.4. Hepatic and Renal Impairment

Patients with functional hepatic impairment may have reduced ability to metabolize DOACs, all of which are at least partially metabolized by cytochrome P450 (CYP) enzymes ^[42][43][44][48,49,50]. Patients with significant liver disease are thus considered to be at higher risk of bleeding when treated with DOACs and were excluded from clinical trials. For these reasons, none of the DOACs are recommended for use in patients meeting criteria for Child-Pugh class C ^[42][43][44][48,49,50]. Rivaroxaban is contraindicated in patients with hepatic disease (including Child-Pugh class B and C) associated with coagulopathy and having clinically relevant bleeding risk ^[42][48]. Apixaban should be used with caution in patients with mild or moderate hepatic impairment (Child-Pugh class A or B) ^[43][49]. With edoxaban, patients with Child-Pugh class A or B exhibited comparable pharmacokinetics and pharmacodynamics to healthy controls ^[44][50].

The previous iteration of the Canadian expert consensus treatment algorithm suggested that LMWH might be preferable to DOAC in patients with CAT and a creatinine clearance of 30–50 mL/min, especially if additional risk factors for bleeding were present ^[8][19]. This recommendation was based on the limited evidence available at the time suggesting a potentially elevated risk of bleeding with edoxaban in these patients ^[18][8][16,19]. However, this recommendation was not supported by the CARAVAGGIO trial, which found no significant between-treatment differences in rates of major bleeding in patients with creatinine clearance of 30–80 mL/min treated with apixaban or LMWH ^[21][18]. Thus, the consensus committee currently recommends that clinicians follow product monograph recommendations for contraindications and dose adjustment of anticoagulants in patients with impaired renal function (Table 2).

Table 2.

 Product monograph dosing recommendations according to creatinine clearance.

Anticoagulant	Creatinine Clearance (mL/min)
	<15 or Dialysis	15–29	30–50	>50
LMWH
Dalteparin [45]	Dalteparin [51]	Dose reduction should be considered	a	Dose reduction should be considered	a	200 IU/kg once daily for 1 month, and then 150 IU/kg	200 IU/kg once daily for 1 month, and then 150 IU/kg
Enoxaparin [46]	Enoxaparin [52]	100 IU/kg once daily	100 IU/kg once daily	100 IU/kg twice daily	100 IU/kg twice daily
Tinzaparin [47]	Tinzaparin [53]	175 IU/kg once daily	a	175 IU/kg once daily	a	175 IU/kg once daily	175 IU/kg once daily
DOAC
Apixaban [43]	Apixaban [49]	Not recommended	10 mg twice daily for 7 days, and then 5 mg twice daily	b	10 mg twice daily for 7 days, and then 5 mg twice daily	b	10 mg twice daily for 7 days, and then 5 mg twice daily	b
Edoxaban [44]	Edoxaban [50]	Not recommended	Not recommended	30 mg once daily (following initial 5–10 days of LMWH)	60 mg once daily (following initial 5–10 days of LMWH)
Rivaroxaban [42]	Rivaroxaban [48]	Not recommended	15 mg twice daily for 3 weeks, and then 20 mg once daily	b	15 mg twice daily for 3 weeks, and then 20 mg once daily	b	15 mg twice daily for 3 weeks, and then 20 mg once daily	b

^a Use with caution when treating patients with creatinine clearance <30 mL/min; see product monograph for dosing in hemodialysis and hemofiltration. ^b Must be used with caution in patients with creatinine clearance 15–29 mL/min due to potentially higher bleeding risks. DOAC = direct-acting oral anticoagulant; LMWH = low-molecular-weight heparin.

2.4.5. Use of Antiplatelet Agents

The use of either dual antiplatelet therapy or higher doses of acetylsalicylic acid (ASA) was not permitted in the DOAC vs. LMWH RCTs, with concomitant ASA at doses >75 mg, >100 mg, and >165 mg daily being exclusion criteria in the SELECT-D, Hokusai-VTE Cancer, and CARAVAGGIO trials, respectively ^[18][19][21][16,17,18]. However, even at low doses, ASA is known to increase the risk of upper GI bleeds, a risk which appears to be increased when it is used in conjunction with oral anticoagulants ^[48][49][54,55]. This was confirmed by subgroup analysis of data from the Hokusai-VTE cancer trial, which showed a numerical increase in the risk of major bleeding in DOAC-treated patients on concomitant antiplatelet therapy ^[18][16]. Similarly, in the CARAVAGGIO trial, 22.7% (5/22) of patients on concomitant antiplatelet therapy treated with apixaban experienced major bleeding as compared to 11.8% (68/576) of apixaban-treated patients without antiplatelet therapy ^[34][40]. No major bleeding events were reported among the 23 patients in the LMWH arm on concomitant antiplatelet therapy, while 12.8% (74/579) of those without antiplatelet therapy had a major bleed ^[34][40]. Given this, the consensus committee recommends that the indication for antiplatelet agents be reassessed, and discontinuation should be considered in the absence of a strong indication in patients with new diagnosis of CAT. Shared decision-making with other health care providers would be warranted in these circumstances.

2.4.5. Drug–Drug Interactions

Polypharmacy is common in patients with cancer, who are often treated with multiple anticancer and supportive therapies. It is thus important to evaluate the potential for drug–drug interactions when selecting the appropriate anticoagulant therapy for CAT. All DOACs are substrates of P-glycoprotein, and apixaban and rivaroxaban are also substrates of CYP3A4, so therapies that affect P-glycoprotein or CYP3A4 metabolism have the potential to interact with DOACs [56]. Numerous anticancer therapies are inhibitors or inducers of the P-glycoprotein and/or CYP3A4 pathways, with the potential to interact with DOACs [57]. Anticancer therapies for which the potential for drug–drug interactions with DOACs should be considered include abiraterone, acalabrutinib, afatinib, ceritinib, cyclosporine, cobimetinib, crizotinib, dabrafenib, dasatinib, dexamethasone, doxorubicin, enzalutamide, erdafitinib, ibrutinib, idelalisib, imatinib, ipilimumab, lapatinib, mitotane, neratinib, nilotinib, nintedanib, niraparib, olaparib, panobinostat, ponatinib, ribociclib, sunitinib, tacrolimus, tamoxifen, trametinib, trastuzumab emtansine, vandetanib, vemurafenib, and vinblastine [57]. However, assessing the potential for clinically significant interactions is complex as not all potential interactions appear to be clinically important [58]. In fact, sub-analysis of patients treated concomitantly with anticancer agents and anticoagulants in the CARAVAGGIO trial found no significant differences in rates of major bleeding, recurrent VTE, or death between the DOAC and LMWH arms [59]. Table 3 lists drug–drug interactions with DOACs that have been shown to have clinical relevance. Notably, a recent registry of the ISTH including 202 patients receiving concurrent DOACs and targeted anticancer therapies has reported a high rate of bleeding complications in patients receiving Bruton’s tyrosine kinase (BTK) inhibitors [60]. A recent observational study has also reported a higher risk of bleeding in patients receiving concurrent vascular endothelial growth factor receptor (VEGFR) tyrosine kinase inhibitors (TKIs) and LMWH [61]. The study sample size was inadequate for between-treatment comparisons with concurrent TKIs and DOACs. Given the complexity of the therapeutic regimens used to treat many patients with cancer, the consensus committee recommends that patients with CAT be referred for a pharmacist-led drug interaction evaluation, which should be repeated if cancer management changes. Alternatively, online drug–drug interaction applications or websites can be helpful, although previous publications have highlighted important differences in the accuracy and quality of these tools [62,63]. However, when using such tools, clinicians must keep in mind that the majority of reported interactions are theoretical rather than having been proven to be associated with decreased drug levels (and thus thrombosis) or increased drug levels (and thus bleeding).

Table 3. Clinically significant drug-drug interactions with direct-acting oral anticoagulants [58,60].

Interacting Drug	Outcome	Proposed Mechanism of Interaction
Acalabrutinib	↑ bleeding risk	Weak CYP3A4 inhibitor/antiplatelet effect
Amiodarone	↑ bleeding risk	Weak CYP3A4/P-gp inhibitor
Carbamazepine	↓ antithrombotic efficacy	Strong CYP3A4/P-gp inducer
Clarithromycin	↑ bleeding risk	Strong CYP3A4/P-gp inhibitor
Cyclosporine	↑ bleeding risk	Weak CYP3A4/P-gp inhibitor
Diltiazem	↑ bleeding risk	Moderate CYP3A4/P-gp inhibitor
Efavirenz	↓ antithrombotic efficacy	Moderate CYP3A4 inducer
Fluconazole	↑ bleeding risk	Moderate CYP3A4 inhibitor
Ibrutinib	↑ bleeding risk	Weak CYP3A4/P-gp inhibitor/antiplatelet effect
Loperamide	↑ bleeding risk	Mechanism unclear
Miconazole (topical)	↑ bleeding risk	Mechanism unclear
Nevirapine	↓ antithrombotic efficacy	Weak CYP3A4 inducer
Oxcarbazepine	↓ antithrombotic efficacy	Weak CYP3A4 inducer
Phenobarbital	↓ antithrombotic efficacy	Strong CYP3A4 inducer
Phenytoin	↓ antithrombotic efficacy	Strong CYP3A4/P-gp inducer
Quinidine	↑ bleeding risk	Moderate P-gp inhibitor
Rifampicin	↓ antithrombotic efficacy	Strong CYP3A4/P-gp inducer
Ritonavir	↑ bleeding risk	Strong CYP3A4/P-gp inhibitor
Tocilizumab	↓ antithrombotic efficacy	Indirect P-gp inducer
Verapamil	↑ bleeding risk	Moderate CYP3A4/P-gp inhibitor