Etiology and management of venous thromboembolism recurrence in patients anticoagulated for venous thromboembolism: a single-center study

Abstract

Background

Venous thromboembolism (VTE) recurrence on anticoagulation (ROA) and anticoagulant failure (ACF) are frequently described but poorly defined.

Objectives

To determine the incidence and causes of ROA and ACF in patients with VTE. The secondary aim was to describe the management of these cases, including anticoagulation selection and dosing strategies.

Methods

Patients with acute VTE enrolled in the Mayo Clinic VTE Registry from 2013 to 2022 were included. Charts of patients with VTE recurrence despite anticoagulation were reviewed. ROA was defined as any new thrombotic event in patients prescribed and recently taking anticoagulants. ACF was defined as ROA in the absence of modifiable risk factors.

Results

Of the 4980 patients included, 150 (3.0%) ROA events occurred in 130 patients. Altered absorption (n = 40, 26.7%), medication dosing (n = 35, 23.3%), and recent interruption of therapy (n = 37, 24.7%) were frequently documented reasons for ROA events. Malignancy was present in 107 (71.3%) patients, with 47 (43.9%) having metastatic malignancy. ROA was attributed to a modifiable cause in 75 (50.0%) patients. In 26 (17.3%) patients, there were no identifiable reasons for failure, and these were labeled as ACF. After a recurrent VTE event, 92 (61.3%) patients were transitioned to an alternative anticoagulant, with the most frequent being enoxaparin. However, 27 (18.0%) patients remained on the same medication at an increased dose, and 27 (18.0%) did not have any change in therapy. There were 4 (2.7%) patients with unclear management strategies.

Conclusion

ROA can affect 3% of patients with VTE. Nearly half of ROA events are preventable with careful review of medications and patient characteristics. When not modifiable, the most used strategy for managing patients with ACF is to transition to another anticoagulant, usually enoxaparin. Our practice strategy for evaluating and managing ROA is outlined.

Essentials

• •ACF is poorly defined, leading to uncertainty in management.
• •ACF led to recurrent VTE in 3% of the patients.
• •Half of the recurrent events were attributed to preventable causes.
• •When there is no modifiable reason for ACF, enoxaparin is most frequently used for management.

1. Introduction

Venous thromboembolism (VTE), including pulmonary embolism and deep venous thrombosis, occurs annually in approximately 1 to 2 patients per 1000 individuals and is the third most common cause of vascular-related death in the United States [1,2]. Anticoagulation is the cornerstone of treatment for patients experiencing VTE, with recent preferential use of direct oral anticoagulants (DOACs) over vitamin K antagonists (VKAs) or heparin-based strategies for the management of uncomplicated VTE and cancer-associated VTE [[3], [4], [5], [6], [7], [8], [9]]. All 3 options, DOACs, VKAs, and heparin-based strategies, remain important pillars in the treatment of VTE patients, with anticoagulant choice based on patient-specific factors (weight, financial feasibility, medication interactions, etc.).

VTE recurrence on anticoagulation (ROA) is infrequent overall but requires careful navigation to determine underlying causes and optimal anticoagulation therapy. Prior studies have identified reasons for ROA, including patient nonadherence, medication interactions, reduced absorption (eg, patients with a history of weight-reduction surgeries), and body weight extremes [[10], [11], [12], [13], [14]]. The true prevalence of ROA is unknown. Because of its rarity, little guidance exists on the evaluation or management of patients after a recurrent event.

Patients with preventable, modifiable reasons for ROA should be distinguished from those with true anticoagulant failure (ACF). Patients with ROA have modifiable or preventable reasons for VTE recurrence, whereas those with ACF have nonmodifiable reasons or no clear reason for recurrence. In some cases of ACF, this leads to uncertainty in management, with a lack of clarity for both the provider and the patient. We reviewed a large, single-center VTE registry to characterize these challenging cases and better understand the etiology of ROA, with particular interest in identifying modifiable risk factors for ROA.

Due to minimal data and a paucity of recommendations for evaluating these clinical scenarios, the primary aim of this study was to identify reasons for ROA and ACF. The secondary aim was to investigate management practices of anticoagulation after confirmed recurrent VTE. Lastly, we outlined a strategy for navigating these challenging clinical scenarios.

2. Methods

Consecutive patients receiving anticoagulation for an index VTE event were prospectively enrolled in the Mayo Clinic VTE Registry (ClinicalTrials.gov: NCT03504007) from March 1, 2013, through November 1, 2022. All 4980 enrolled patients were prospectively followed for VTE recurrence and all-cause mortality after evaluation by our thrombosis clinic. Some patients were initiated on anticoagulation by primary care, hospitalists, or emergency department providers prior to referral and evaluation by the thrombosis clinic. Follow-up included in-person visits at predefined intervals (3 months, 6 months, and annually). If a face-to-face encounter was not feasible, a questionnaire was used; patients received this either by email, letter, or a scripted telephone interview. All events were adjudicated by a physician committee using previously outlined study criteria [15].

Patients experiencing recurrent VTE were identified using prospective registry data. For those with a reported recurrence, an in-depth review of their medical records for details of anticoagulation therapy, medical comorbidities, medication review, documented reasons for ROA, recent surgical procedures, and anticoagulation interruptions was completed retrospectively by a thrombosis specialist (N.B.), with a second reviewer (A.I.C.) involved as needed. Similar events were grouped into categories for further analysis.

The study was approved by the Mayo Clinic Institutional Review Board.

2.1. Outcome measures and study definitions

In our cohort, VTE recurrence was defined as any new VTE event (either a new location or recurrence at the same location, verified by interval imaging). VTE events included deep venous thrombosis and pulmonary embolism, as well as venous thrombosis in unusual locations (splanchnic, cerebral sinuses, etc.). All events were adjudicated through review of medical records by 3 thrombosis specialists using previously defined study criteria. Discrepancies between reviewers were resolved through a repeat review of the medical record and discussion between the adjudicators until consensus was reached.

VTE ROA was defined as recurrent VTE despite uninterrupted anticoagulation at the time of recurrence. Patients with preventable, modifiable reasons for ROA were distinguished from those with true ACF. ACF was defined as patients without modifiable reasons for failure or no clear reason for VTE recurrence. We defined anticoagulation interruption for warfarin therapy as >5 days since the last warfarin dose or an international normalized ratio (INR) < 2.0. For patients receiving DOAC therapy or low-molecular-weight heparin (LMWH), interruption was defined as >48 hours since the last dose. To further clarify, all patients with ROA had received anticoagulation therapy in the days before VTE recurrence (warfarin in the prior 5 days, INR > 2.0, or had received DOAC/LMWH within the preceding 48 hours). Patients with ROA in the absence of documented modifiable or preventable reasons were labeled as ACF.

The secondary aim was to investigate clinician management of anticoagulation after ROA was identified. This included a review of anticoagulant choice and dosing strategies.

Baseline data were reported as percentages of the total or as medians with IQRs. Descriptive statistics were utilized in this study.

3. Results

3.1. Patient demographics

Of the 4980 patients enrolled in the Mayo Clinic VTE Registry, we identified 150 (3.0%) ROA events in 130 (2.6%) patients, which was the focus of our study. Of these, 117 (90.0%) patients experienced 1 ROA event. A total of 13 (8.7%) patients had more than 1 ROA: 9 (6.9%) with 2 documented events, 3 (2.3%) with 3 events, and 1 (0.8%) with 6 events. The median patient age was 63 years; 61 (46.9%) were female, with an average patient weight of 83.1 kg (IQR, 33.1). Baseline laboratory data demonstrated normal renal function (median creatinine, 0.9 mg/dL; IQR, 0.4) and normal platelet counts in 105 (70.0%) patients. Mild to moderate thrombocytopenia was identified in 35 (23.4%) patients, with severe thrombocytopenia (platelets < 50,000/μL) in 4 (2.7%). The average time between the initial VTE event and recurrence was 97 days (IQR, 157). Only 1 (0.7%) ROA event occurred less than 7 days after anticoagulation initiation (Table 1).

Table 1.

Demographics of patients with venous thromboembolism recurrence on anticoagulation.

Variable	Value (N = 150)
Age (y), median (IQR)	63.0 (21)
Female sex, n (%)	61 (46.9)
Weight (kg), median (IQR)	83.1 (33.1)
Contributing factors, n (%)
Active malignancy	107 (71.3)
Recent surgery	59 (39.3)
Known thrombophilia	10 (6.7)
Laboratory data
Platelet count/μ, n (%)
<50	4 (2.7)
50-100	10 (6.7)
100-150	25 (16.7)
>150	105 (70.0)
Creatinine, mg/dL, median (IQR)	0.9 (0.4)
Time to recurrence (d), median (IQR)	97 (157)
Interruption of anticoagulation within the prior 30 days, n (%)	37 (24.7)
Anticoagulation in the Mayo Clinic VTE Registry
LMWH	1836 (39.3)
Heparin	1229 (26.3)
Apixaban	988 (21.2)
Rivaroxaban	507 (10.9)
Warfarin	89 (1.9)
Dabigatran	14 (0.3)
Other	8 (0.2)
Missing	55 (1.1)
Initial anticoagulation for patients with documented failure
Apixaban	68 (45.3)
LMWH	38 (25.3)
Rivaroxaban	28 (18.7)
Warfarin	14 (9.3)
Dabigatran	2 (1.3)
Unknown	1 (0.7)

IQR, interquartile range; LMWH, low-molecular-weight heparin; VTE, venous thromboembolism.

In patients with ROA, underlying malignancy was identified in 107 (82.3%) patients. The most common types of malignancy included hepatobiliary (n = 27, 25.0%), gynecologic (n = 19, 17.6%), hematologic (n = 17, 15.7%), and gastrointestinal (GI; n = 14, 13.0%). Other less common malignancy subtypes included pulmonary, urologic, breast, and other (Table 2).

Table 2.

Documented reasons for recurrence on anticoagulation.

Variable	n (%)
	(N = 150)
Identified reason for ACF	124 (82.7)
Multiple reasons documented (2 or more)	39 (26.0)
No identifiable cause for failure	26 (17.3)
Modifiable causes for failure	72 (52.0)
Aggressive or recurrent malignancy	47 (43.9)
Hepatobiliary	15 (31.9)
Gynecologic	8 (17.0)
Gastronitestinal	6 (12.7)
Lung	4 (8.5)
Hematologic	3 (15.7)
Breast	2 (4.3)
Urologic	1 (2.1)
Othera	6 (12.7)
Altered absorption	40 (26.7)
Distal bowel resection	17 (42.5)
Proximal bowel resection	14 (35.0)
Rivaroxaban fasting	6 (15.0)
History of bariatric surgery	1 (2.5)
Poor oral intake	1 (2.5)
Other	1 (2.5)
Recent surgery (30 d)	24 (16.0)
Insufficient anticoagulant dosing	34 (22.7)
Dose reduction due to perceived bleeding riskb	15 (44.1)
Dose reduction after 6 mo of therapy, found to have new cancer	5 (14.7)
Insufficient due to patient misunderstanding of dosing	4 (11.8)
Dose insufficient related to weight changes	3 (8.8)
Dosing errorsc	3 (8.8)
Dose reduction for medication interaction	2 (5.8)
Medication interaction leading to insufficient dosing	2 (5.8)
Active bleeding resulting in interruption	13 (8.7)
Nonadherence	11 (7.3)
Otherd	29 (19.3)

ACF, anticoagulant failure.

^aOther cancers include spindle cell, adrenocortical carcinoma, glioblastoma multiforme, thyroid, and a combination of 2 malignancies.

^bPerceived bleeding risk either due to thrombocytopenia or determined by the primary team.

^cDosing errors: missed bridging on warfarin or a script error.

^dOther causes include structural causes (May–Thurner syndrome or tumor compression), direct vessel injury due to surgery or radiation, critical illness, or acute infection with COVID-19.

3.2. Reasons for VTE ROA

Documented reasons for ROA were grouped into categories based on similarities. These included recent anticoagulation interruption for surgery or due to bleeding risk, patient nonadherence, medication error, altered drug absorption, medication interactions, major hypercoagulable state in advanced malignancy, or other. Other causes included COVID-19 infection, structural etiology such as May–Thurner syndrome or other mechanical venous compression syndrome, and direct vessel injury. Of the 150 events, 39 (26.0%) had multiple reasons contributing to ROA. There were 26 (19.7%) events without an identifiable etiology.

Prior anticoagulation interruption for surgery or bleeding in the preceding 30 days occurred in 37 (24.7%) ROA: 24 (16.0%) for surgery and 13 (8.7%) due to recent bleeding. The median time from anticoagulation interruption to diagnosis of VTE recurrence was 18 days. In all of them, anticoagulation was resumed at the time of VTE diagnosis.

Possible altered drug absorption was noted in 40 (26.7%) ROA events. This included patients with a history of distal bowel resection (n = 16, 40%), proximal bowel resection or bariatric surgery (n = 15, 37.5%), taking rivaroxaban without food (n = 6, 15.0%), poor oral intake (n = 1, 2.5%), or other absorption issues (n = 2, 5.0%). We do not routinely test for DOAC levels, so these events could not be confirmed.

Medication interactions and underdosing occurred in 34 (22.7%) patients. Medication interactions known to decrease the efficacy of anticoagulation, such as concurrent use of dabigatran and phenytoin, occurred in 2 (5.8%) patients. In other instances, prescribers empirically reduced anticoagulation dosing due to concern for potential interaction, eg, dose-reducing apixaban while prescribing itraconazole. This occurred in 2 (5.8%) additional patients. Anticoagulation dose reduction was noted in 15 (44.1%) patients due to an elevated bleeding risk; 8 (23.5%) had concurrent thrombocytopenia with a median platelet count of 91 (range, 20-199). The other 6 (17.6%) patients had their dosing reduced because of a perceived elevated bleeding risk (due to a history of bleeding, baseline coagulopathy, etc.). Insufficient dosing also occurred in 15 (44.1%) patients, with 5 (14.7%) receiving a dose reduction after 6 months of completing therapeutic anticoagulation, then being found to have a new malignancy leading to ROA. Another 4 (11.8%) patients were underdosed due to patients’ misunderstanding of dosing instructions, 3 (8.8%) due to weight changes, and 3 (8.8%) due to dosing errors (a lack of bridging with subtherapeutic INR and a single prescription error).

Nonadherence to the prescribed anticoagulation regimen was reported in 11 (7.3%) ROA. Some examples of medical nonadherence included the inability to swallow medication, patient-directed dose reduction due to concerns about bleeding, lapse in refilling medications, and intermittent forgetting doses (never for >48 contiguous hours).

Patients were also reviewed for underlying thrombophilias. A total of 56 (43.1%) patients underwent at least partial thrombophilia testing, including evaluation for antiphospholipid syndrome (APS). Of these, a small portion (n = 10, 7.7%) were found to have an underlying thrombophilia, with the most commonly identified being a prothrombin gene mutation (n = 5, 50%) followed by a factor (F)V Leiden mutation (n = 4, 40%). Antithrombin deficiency was diagnosed in 1 patient. None of the patients screened for APS were positive. While thrombophilia testing should be considered in all VTE recurrences, the population in this cohort had a high rate of known prothrombotic malignancies or other clear reasons for thrombosis recurrence. Our vascular experts therefore did not offer thrombophilia testing in patients of older age with clear triggers for recurrence, but instead navigated these recurrences based on individual risk profiles and shared decision-making with patients.

While 107 patients in our cohort had known malignancy, 47 (43.9%) were documented to have a recurrent or “aggressive malignancy” identified as a contributing etiology to their ROA. “Aggressive malignancy” represented advanced-stage cancers (typically stage IV) with progression despite therapy or multiple simultaneous malignancies. The most frequently represented malignancies were hepatobiliary (n = 17, 48.6%), gynecologic (n = 8, 22.9%), GI (n = 6, 17.1%), pulmonary (n = 4, 11.4%), hematologic (n = 3, 8.6%), and breast (n = 2, 5.7 %). Other sites of cancer were infrequent (n = 6, 17.1%); see Table 2. In 2 (5.7%) patients, 2 separate malignancies were identified. The first patient had GI and prostate cancer, and the second had both a pancreatic neuroendocrine tumor and concurrent urothelial cancer.

Other documented reasons (n = 29, 19.3%) for ROA included structural causes, direct vessel injury, critical illness with hospitalization, and acute infection with COVID-19. May–Thurner syndrome was reported in 10 (34.5%) events and was the most frequent structural reason documented, followed by direct venous compression by a tumor. Examples of direct vessel injury included injury from recent surgery, line-associated injury, or radiation therapy.

In 75 (50.0%) patients, VTE recurrence could have been prevented by selecting a different anticoagulant or improving adherence. These included patients with known issues with absorption, medication nonadherence, underdosing, and medication interactions.

3.3. Anticoagulation trends and management after ROA

Among patients with identified ROA, apixaban (n = 68, 45.3%) was the most frequently used anticoagulant, followed by LMWH (n = 38, 25.3%), rivaroxaban (n = 28, 18.7%), and warfarin (n = 14, 9.3%; see Table 1). Dabigatran was infrequently prescribed (n = 2, 1.3%). These findings were consistent with the most common anticoagulation strategies prescribed in our registry.

After ROA, 92 (61.3%) patients were transitioned to an alternative anticoagulant (Figure 1, Table 3). The most common agent used after ROA was enoxaparin (n = 63, 68.5%). Other agents chosen included apixaban (n = 13, 14.3%), rivaroxaban (n = 8, 8.8%), warfarin (n = 3, 3.3%), unfractionated heparin (n = 3, 3.3%), dalteparin (n = 1, 1.1%), and fondaparinux (n = 1, 1.1%). In 25 (18.3%) ROA events, the same anticoagulation therapy was continued at an increased dose, and 25 (18.3%) did not have any change in anticoagulation therapy following ROA. Of the patients without change in anticoagulation, the majority had “modifiable” reasons for ROA, including dosing issues, patient nonadherence, or recent interruption due to surgical procedures.

Figure 1.

Anticoagulant selection before and after ROA.

Table 3.

Described management for patients with recurrence on anticoagulation.

Variable	n (%)
	(N = 150)
Anticoagulation management
Transitioned to alternative medication	92 (61.3)
Enoxaparin	9 (6.0)
DOAC	71 (47.3)
Warfarin	12 (8.0)
Same anticoagulant, increased dose	24 (16.0)
Enoxaparin	20 (13.3)
Prophylactic dosing increased	4 (2.7)
Therapeutic dosing increased	16 (10.7)
DOAC (prophylactic dosing to therapeutic)	4 (2.7)
No anticoagulation change	25 (18.3)
Anticoagulation change
Enoxaparin	63 (68.5)
Apixaban	13 (14.3)
Rivaroxaban	8 (8.8)
Warfarin	3 (3.3)
Unfractionated heparin	3 (3.3)
Dalteparin	1 (1.1)
Fondaparinux	1 (1.1)

DOAC, direct oral anticoagulant.

4. Discussion

4.1. A stepwise approach to navigating recurrence in anticoagulation

To navigate ROA, our team used a multistep approach when evaluating patients. After confirming VTE recurrence, a 3-part comprehensive review of patients was employed (Figures 2 and 3). This reflects the most common reasons for ROA: interruptions in therapy, issues inherent to the selected anticoagulant (treatment-related), or a hypercoagulable substrate (patient-related).

Figure 2.

Suggested algorithm for the identification and management of anticoagulant failure in venous thromboembolism. AC, anticoagulation; aPTT, activated partial thromboplastin time; INR, international normalized ratio.

Figure 3.

Suggested algorithm for the identification and management of ROA in venous thromboembolism (VTE). APS, antiphospholipid syndrome; HIT, heparin-induced thrombocytopenia; INR, international normalized ratio; MGTS, monoclonal gammopathy of thrombotic/thrombocytopenic significance; VKA, vitamin K antagonist.

4.1.1. ROA related to interruptions

First, a review of recent interruptions in anticoagulation therapy was completed [16]. In patients with a history of unprovoked or cancer-associated VTE, the annual risk of VTE recurrence is high, ranging from 3.7% to 10.3% [3,6,17]. Furthermore, if treatment is interrupted for surgery (a prothrombotic state), nonadherence, or other reasons (such as bleeding), it may result in additional risk of VTE recurrence, particularly in the 3 months following the index event [18]. In our cohort, 37 (24.7%) patients experienced anticoagulation interruption in the preceding 30 days, either due to surgery or a recent bleeding event. Additionally, 11 (7.3%) patients reported nonadherence, defined as missing >2 doses since the previous visit but not exceeding >48 consecutive hours of missed therapy. This is similar to prior studies reporting a 6.5% to 20.7% nonadherence rate at 1 year for patients prescribed DOAC therapy for atrial fibrillation [[19], [20], [21]]. There is limited data available on DOAC persistence for VTE treatment, but it is likely that the barriers to DOAC adherence are similar between patients anticoagulated for atrial fibrillation and for VTE and include patient fear of side effects, prohibitive cost, or lack of active symptoms to support the need for medication [20,22].

4.1.2. ROA related to treatment

Second, we evaluated issues inherent in the anticoagulant, such as medication interactions, underdosing, body weight extremes, or potential absorption issues. Our ROA group with medication issues comprised primarily of patients on DOAC therapy. While infrequent, prescribers must be aware that medications that induce cytochrome P450 may lead to a decline in serum DOAC levels [14]. Commonly overlooked medications that can lead to subtherapeutic DOAC levels include antiseizure medications (carbamazepine, phenytoin, levetiracetam, and phenobarbital), rifampin, and herbal supplements, such as St. John’s wort [14,23]. Interactions in our cohort included those with phenytoin and levetiracetam.

Our study also identified potential inherent issues with anticoagulant absorption. This primarily occurred in patients on DOAC therapy. Rivaroxaban was the most frequent DOAC for which absorption issues were identified. For apixaban, most absorption occurs independently of food intake in the small intestine and ascending colon [24]. Little data exist on the impact of GI surgery on apixaban levels. For rivaroxaban, absorption is primarily gastric, and bioavailability depends on food intake [10,25]. A prior study [12] has demonstrated a substantial decline in rivaroxaban drug levels in patients with prior bariatric surgery and fasting drug use. There is a theoretical risk of subtherapeutic anticoagulant levels in all DOAC patients with a history of GI surgery; therefore, this should be explored in all patients presenting with ROA. In our cohort, potential causes of altered drug absorption were identified retrospectively by experts in 40 (26.7%) ROA events; however, providers managing ROA acutely cited it as the cause of ROA in only 11 (7.3%) events. Potential for altered absorption included patients with a history of distal bowel resection (n = 17, 42.5%), proximal bowel resection or bariatric surgery (n = 15, 37.5%), fasting intake of rivaroxaban (n = 6, 15.0%), poor oral intake (n = 1, 2.5%), or other absorption issues such as autoimmune gastritis (n = 1, 2.5%). If possible, drug levels should be obtained in these instances to better understand drug bioavailability. This should be considered particularly in cancer patients with poor oral intake or prior proximal GI surgery and highlights the importance of considering these factors before prescribing specific anticoagulants.

While medication interactions and absorption issues often occur with the use of VKAs due to numerous interactions, these were not frequent in our cohort, likely due to provider awareness of interactions and close follow-up of INR by anticoagulation clinics.

Other anticoagulant-dependent reasons for ROA included dosing errors or empiric anticoagulation dose reductions in 35 (23.3%) ROA. While dosing oversights due to the prescriber were infrequent (n = 9, 6.0%), these represent a preventable cause of VTE recurrence. Other examples of anticoagulant-dependent ROA included empiric dose reduction at the discretion of the care team, either due to potential medication interactions (such as concurrent use of a cytochrome P450 inhibitor inhibitor) or due to perceived bleeding risk. Interestingly, for those on anticoagulant dose reduction due to concern about bleeding, only 1 patient had severe thrombocytopenia (defined as platelets < 50,000) for which dose reduction was indicated. Current recommendations are to continue therapeutic-dose anticoagulation in patients with mild to moderate thrombocytopenia (platelets > 50,000/μL) [26]; we identified 8 instances in which this guideline was not followed. Since our cohort has a high coincidence of malignancy, this likely represents the suboptimal treatment of a high thrombotic risk group [26]. There were 2 (1.3%) events on prophylactic dosing after 6 months of guideline-directed treatment; however, both recurrences occurred in the context of newly identified malignancy [27].

Weight extremes are another previously described reason for ROA. The International Society on Thrombosis and Haemostasis (ISTH) defines weight extremes as patients with a body mass index (BMI) of ≥40 or an absolute body weight of ≥120 kg [10]. In our ROA cohort, 13 (10.0%) patients had a reported body weight of ≥120 kg. While DOAC therapy is reasonable at weight extremes, there is limited data to strongly support efficacy. Based on the available data from phase 4 studies, the ISTH supports standard dosing of rivaroxaban and apixaban in patients regardless of BMI or weight [10]. However, there is a true paucity of safety data in patients with a BMI ≥ 50 or an absolute weight of ≥150 kg. In our ROA group, 3 (2.3%) patients weighed more than 150 kg. Only 1 patient was prescribed low-dose rivaroxaban, possibly contributing to failure, while the others were prescribed enoxaparin and warfarin. The patient on enoxaparin had a dose reduction by the prescribing provider (prescribed 130 mg twice daily for a weight of 170 kg), indicating a 0.8 mg/kg twice-daily dosing strategy [28]. This individual had anti-Xa testing, which demonstrated subtherapeutic anti-Xa levels. The warfarin patient was considered ACF, as the INR at the time of recurrence was 2.1 and was thought to be structurally related to a brace on the lower extremity complicated by morbid obesity.

4.1.3. ROA related to patient characteristics

Third, an in-depth examination of patient-centric factors contributing to hypercoagulability or venous compression was completed (Figure 2). Hypercoagulability can be due to heritable or acquired thrombophilia, as well as malignancy, heparin-induced thrombocytopenia, and, rarely, monoclonal gammopathy of thrombotic thrombocytopenic significance [29]. Our study examined whether patients with ROA underwent thrombophilia evaluation (including heritable causes or antiphospholipid antibodies). In 56 (37.3%) events, patients underwent at least a partial thrombophilia workup, with 10 (6.7%) positive for inherited thrombophilias, most commonly the FV Leiden mutation or a prothrombin gene mutation. It is well established that while thrombophilias increase the risk of first-time VTE, only a few cause recurrent events or are considered to explain ROA due to severe hypercoagulability [30,31]. These mild thrombophilias identified in our cohort are unlikely to cause ROA, a finding supported by prior studies, but are merely additive risk factors in patients with other significant risks of VTE recurrence [31]. Of potential concern would be “high-risk” thrombophilias and DOAC dose reduction, which are not as well studied. We found patients with protein C, protein S, or antithrombin deficiency in our recurrent cohort. Possible explanations include that our cohort is slightly older (median age, 63 years) than expected for the initial presentation of high-risk thrombophilias, that high-risk thrombophilias are uncommon in the population, and that malignancy had a high prevalence in our cohort. Malignancy is a strong underlying hypercoagulable state and risk factor for a first thrombotic event and ROA. As supported by guidelines, we do not test for thrombophilia in patients with malignancy or other strong provoking factors. While the timing of thrombophilia testing was not specifically reported for our cohort, our team uses genetic-based testing for FV Leiden and prothrombin gene mutations. If protein C, protein S, or antithrombin deficiencies are suspected and patients meet appropriate clinical criteria, then testing is offered after the initial thrombotic event.

APS is another well-established risk for thrombotic recurrence (6.3% on warfarin) and DOAC failure [[32], [33], [34]]. It is our clinical practice to treat all APS patients with heparin-based therapy or VKAs due to the high risk of thrombotic recurrence; this may be a reason why we have a small representation of APS patients in our ROA cohort.

Our group predominantly consisted of a high percentage of patients with underlying, aggressive malignancy. Patients with malignancy carry a 9-fold increased risk of VTE and a 3-fold risk of VTE recurrence compared with patients without cancer [9,35]. It is important to recognize that the thrombotic risk associated with malignancy is polymorphic, varying depending on the malignancy subtype and extent of metastatic disease; therefore, the role of newer DOAC therapy in aggressive cancer subtypes is unclear [35]. A recent study by Agnelli et al. [35], which further analyzed cancer patients in the Caravaggio study, demonstrated the highest rate of VTE recurrence in patients with gynecologic and GI cancers, with recurrence rates as high as 16.9% and 9.1%, respectively [6]. Patient age, prior history of VTE, advanced malignancy, and use of certain anticancer therapies are established risk factors for VTE recurrence in cancer-associated VTE [35,36]. Our findings support prior data of 21 (14%) ROA events in patients with hepatobiliary and GI cancers. These findings are similar to the overall distribution of cancer-associated VTE in our registry, with GI cancers accounting for 19.9% of our cohort [37].

Lastly, structural reasons for ROA were considered. Venous compression syndromes, such as May–Thurner syndrome or tumor compression, predispose patients to the development of VTE by increasing venous stasis [38,39]. These represented a minority of patients in our cohort but are an important risk factor for ROA. In severe venous compression syndromes, anticoagulation alone is unlikely to be sufficient regardless of anticoagulant choice, resulting in ACF, and requires consideration for interventions directed at relieving the compression. Other structural reasons include direct vessel injury related to procedures.

4.2. Anticoagulation management after ROA

In patients with isolated, modifiable ROA events, clinicians can continue the previously outlined anticoagulation plan after the underlying factors have been addressed; this occurred in 25 (18.3%) ROA events in our cohort (Figure 3). Examples of modifiable events include interruptions for procedures, patient nonadherence, etc. It is critical to reemphasize the importance of medication adherence for patients and the team to develop a plan to minimize future anticoagulation interruptions. If the modifiable cause of ROA is a medication error (improper dosing, avoidable medication interaction, etc.), this should be addressed promptly. In some circumstances, the same anticoagulant can be continued but at an adjusted, appropriate dose.

However, in some ROA, transitioning to an alternative anticoagulation option is the best step. This most frequently occurs when a nonmodifiable risk for ROA or ACF is identified. LMWHs have long been supported in guidelines as a mainstay therapy in patients with cancer-associated VTE or absorption issues [40,41]. Transition to LMWHs has been a previously described strategy in patients with concern for DOAC failure; however, further studies are required to better understand these mechanisms and to assist in prediction models for patients at high risk of ACF [42]. In our cohort, the most common agent chosen after ROA in patients with nonmodifiable risks was enoxaparin (n = 62, 68.1%), likely reflecting identified absorption issues or the coexistence of malignancy. In patients with ROA or ACF despite enoxaparin use, our group measures heparin-Xa levels to ensure adequate dosing and increases the dose by 20% if warranted [43].

Lastly, there was a high rate of patients with multiple VTE recurrences, with 13/150 (8.7%) having multiple recurrences. This high number of patients with repeated events reflects the clinical difficulties in effectively managing patients at risk of ROA due to modifiable risks, such as nonadherence, or quite challenging hypercoagulable states, such as cancer. For those with severe hypercoagulable states and repeated thrombotic events, our group has outlined potential options for anticoagulation escalation to prevent future recurrence (Figure 3).

In summary, anticoagulation management following ROA is complex, requiring a clear understanding of the mechanism leading to ROA to determine the safest course of action in anticoagulation strategies. After an in-depth analysis of the ROA event (Figure 2), it is important to determine whether the etiology of ROA can be modified to prevent future thrombosis. With this determination, the next steps for anticoagulation can be optimally pursued (Figure 3).

4.3. Study limitations

Limitations of our study include data from a single-center, quaternary referral center. This can lead to referral bias in our registry, resulting in patients with higher medical complexity, which might lead to higher-than-average ROA, particularly given the large proportion of patients with underlying malignancy. Similar studies should be conducted at other centers to ensure the reproducibility of our findings. This is especially necessary to better characterize the risk of ROA/ACF in patients in medical facilities without the support of a vascular medicine/anticoagulation specialty group.

Given that our VTE registry enrolls patients with first-time VTE, several patients our practice evaluates for ROA as a second opinion following a recurrent thrombotic event may have been excluded. A retrospective or multicenter study could be completed to better capture this patient cohort in its entirety.

Our study reflects the practice of our center, which specializes in thrombosis. Although there is homogeneity in practice, there are no preestablished protocols for workup and management of ROA at present.

5. Conclusion

ROA is an infrequent but challenging clinical scenario for care teams, occurring at a rate of only 3.0% in our cohort. Limited data are available to assist clinicians in these difficult clinical scenarios. Based on our study, nearly half of ROA events would have been preventable if a thorough patient review had been conducted at the initial VTE event or a plan had been developed for anticoagulant interruption. After ROA, anticoagulation management required an in-depth evaluation to determine whether the etiology was modifiable. In circumstances where risks were modifiable, such as nonadherence, interruption for surgery, dosing errors, etc., initial anticoagulation was often continued after eliminating ROA factors and developing an anticoagulation plan to minimize future time off therapy. For patients with nonmodifiable events, enoxaparin was the preferred anticoagulant by clinicians in our cohort.