Patients with cancer are at increased risk of thrombosis, particularly those with a central venous device [10,11,12,13]. The incidence of UEDVT is increasing due to the widespread use of PICCs and improved survival of patients with cancer. Symptomatic catheter-associated thrombosis occurs in 3.0–5.0% of patients with cancer requiring venous access, which may increase to as much as 30.0% when including asymptomatic cases [12, 14].

However, the management of PICC-related UEDVT in cancer patients may be more complicated than in the general population given the higher risk of recurrent thrombosis and anticoagulation-associated bleeding experienced by cancer patients [15,16,17,18]. In the noncancer population, the ACCP 2016 guidelines recommended direct oral anticoagulants (DOACs) over vitamin K antagonist (VKA) therapy for VTE based on the greater convenience and abundant evidence that DOACs have similar efficacy in noncancer patients with fewer adverse events, such as life-threatening bleeding [19,20,21,22].

LMWH was recommended as the first-line choice of anticoagulant for cancer patients with VTE in major guidelines. From 2018 to 2020, several high-quality trials, including the Hokusai VTE, SELECT-D, ADAM VTE and Caravaggio studies, were published to compare DOACs with LMWH for the treatment of VTE in patients with cancer [2,3,4,5]. These head-to-head studies and subsequent meta-analyses indicated that DOACs were associated with lower VTE recurrence. In Caravaggio and ADAM studies, apixaban demonstrated similar or even lower incidence of bleeding events compared to LMWH. However, results from SELECT-D and Hokusai studies raised concern about the increased risk of bleeding in patients treated with rivaroxaban and edoxaban. Based on these results, contemporary guidelines, including the most recent NCCN guidelines, cite DOACs as an acceptable option for VTE treatment in cancer patients [6, 7, 19].

In contrast to classic LMWH and VKA, DOACs need no daily injection and routine monitoring, and the extensive experience in treating PE and lower extremity DVT using DOACs in cancer patients indicated that DOACs may be an alternative option for the long-term management of UEDVT in cancer patients. However, there are only sporadic reports about the possible use of DOACs in cancer patients with PICC-related UEDVT. Limited randomized controlled trials have focused on the management of this topic, and most recommendations are based upon observational studies or extrapolation from studies of noncatheter-related lower-extremity deep vein thrombosis (LEDVT). The optimal choice and duration of PICC-related UEDVT in cancer patients are still unclear.

In our study, the 180-day cumulative risk of recurrent VTE in the cancer patients receiving rivaroxaban was comparable with those receiving nadroparin (1.7% vs 2.0%, p = 0.777). Compared to the recurrence of 7–11.1% demonstrated in previous studies of LMWH treatment in cancer-associated thrombosis (CAT), the recurrent VTE rate at 180 days in our study was at a relatively low level of 1.8%, which was consistent with the results of studies focusing on UEDVT treatment in cancer patients. In the catheter 2 study that assessed rivaroxaban monotherapy in cancer patients with UEDVT, seventy cancer patients were included with a mean age of 54 years, and the most common malignancy was breast cancer (41%). All patients were treated with rivaroxaban for 12 weeks; the recurrent VTE at 12 weeks was 1.43% [23]. Similar low recurrence rate was also seen in the Catheter study and a recent meta-analysis [24]. In a meta-analysis published in 2021, a pooled analysis from 7 trials with 100% cancer patients and an indwelling catheter showed that the recurrent VTE rate was 1.7% (8/468) [1, 25].

UEDVT is usually excluded from large clinical trials of anticoagulants for VTE treatment. In comparison to usual site VTE, PICC-related UEDVT is often provoked by unique risk factors determining incidence and recurrence. The presence of an indwelling catheter in the upper arms represents a local and transient additional thrombotic risk factor in cancer patients because of vessel wall mechanical damage and blood coagulation activation from infused medications. This may partially explain why the outcome from our analysis was different from other data focusing on cancer patients treated for usual site VTE in the literature.

Another important objective of our study was to assess the rates of all bleeding events during anticoagulation.

In our study, although there was no major bleeding event observed at 180 days of anticoagulation, Kaplan‒Meier analysis demonstrated a significant difference in CRNMB rates at 180 days. Adjusted for baseline characteristics, the Cox proportional hazard model showed a higher CRNMB risk in the rivaroxaban group than in the nadroparin group (HR 3.30, 95% CI 1.15–9.50, p = 0.03).

In the two previous cited studies evaluating DOACs for PICC-related UEDVT in cancer patients, the bleeding results were similar to our study [23, 25]. In the first study comparing 44 patients treated with rivaroxaban to 40 patients treated with enoxaparin/warfarin, nonmajor bleeding events occurred in 7.3% and 11.4% of the rivaroxaban- and enoxaparin/warfarin-treated patients, respectively. In the Catheter 2 study, all bleeding events occurred in 13% (7 major bleeding and 4 CRNMB) [23].

Several limitations in our study should be acknowledged. First, given the inherent limitation of the retrospective and nonrandomized design of this study, loss of follow-up and selection bias could have been introduced into our cohort. Our study focused on PICC-related UEDVT in cancer patients, the included patients were homogeneous in terms of baseline characteristics, underlying UEDVT risk factors, anticoagulant treatment and duration of follow-up. As we mentioned before, the data of this cohort were derived from a prospective electronic registry database designed for central venous access administration, accompanied as an important part of anticancer therapy, and only 4.7% (11/233) of patients were lost to follow-up.

Second, our study had a relatively small sample size. Although catheter-related UEDVT accounts for 50–90% of all UEDVT cases, recent studies reported that the incidence of UEDVT was just 4–10% of all newly diagnosed VTE cases [16]. Available data on DOACs for UEDVT in cancer patients are still very scarce. In a recently published meta-analysis assessing anticoagulation for UEDVT, 1473 patients from 20 studies were included, and the average number of enrolled patients in each study was no more than 80 (range from 10 to 210) [1].

Third, we included only symptomatic UEDVT patients, which may have underestimated the incidence of all UEDVT cases. However, many impressive studies for VTE treatment (such as SELECT-D, Hokusai-VTE, Caravaggio studies) set symptomatic or incidental VTE as inclusion criteria, and routine ultrasound surveillance for DVT is also not recommended in major guidelines unless there are clinical signs for VTE [3,4,5]. In our Venous Assess Center, consultations for suspected UEDVT by specialist PICC nurses were provided when patients came to our center for weekly dressing changes. Further imaging examination was arranged once any clinical signs were present.