Association Between Anticoagulant-Related Bleeding and Mortality in Patients With Solid Tumors and Cancer-Associated Venous Thromboembolism

Venous thromboembolism (VTE) is a significant cause of morbidity and mortality in cancer patients. Managing VTE in cancer patients involves balancing elevated risks of VTE recurrence with anticoagulant (AC)-related bleeding. Risk of AC-related bleeding is exacerbated in cancer patients due to older age, thrombocytopenia, frailty, comorbid disease, and tumor invasion [1]. Limited data exist on the association between AC-related bleeding and survival in patients with cancer-associated VTE. In one meta-analysis of patients with cancer, the case fatality rate for AC-related major bleeding was nearly 1 in 10 patients [2]. In another study, patients with cancer-associated VTE had a 2.7-fold increased rate of bleeding-related mortality compared to patients with VTE without cancer [3]. However, fatality in respect to site of bleed was not reported. This study aims to quantify the relationship between clinically significant bleeding events and death in patients with cancer-associated VTE newly initiated on AC therapy, stratified by site of bleeding.

Utilizing data from the nationwide US Veterans Affairs (VA) healthcare system between 2012 and 2020, we retrospectively identified patients with solid tumor using international classification of diseases, 9th and 10th Revisions (ICD-9/10) codes. We identified new cancer-associated VTE using a validated algorithm [4] that used the combination of ICD-9/10 codes (from both inpatient and outpatient encounters) and initiation of AC therapy. A VTE was considered as occurring in a patient with active cancer if diagnosed within 3 months prior to cancer diagnosis or 6 months after cancer diagnosis or at any time after a diagnosis of metastatic cancer. Patients with cancer-associated VTE were included if they received a prescription for at least 30 days of direct oral anticoagulants (DOACs), low molecular-weight-heparin (LMWH), or vitamin K antagonists (VKA) within 30 days of VTE diagnosis. To ensure a clear temporal relationship between AC initiation and VTE diagnosis, patients with outpatient AC prescriptions within 6 months preceding the VTE diagnosis were excluded. Using the VA Informatics and Computing Infrastructure platform, data were obtained via the VA Corporate Data Warehouse. We extracted baseline demographic data at the start of AC therapy. All variables considered for the study are defined in the Table S1. We calculated the VA-Frailty Index (FI) as previously described except that we omitted bleeding-related codes.

The primary outcome of interest was death within 12-month of AC therapy initiation. We identified date of death using the VA Vital Status File and survival time was calculated as the number of days from the date of initiating AC therapy to the date of death. The primary exposure of interest was clinically significant AC-related bleeding requiring hospitalization within 12-month of AC therapy initiation. Hospitalized bleeding events were identified based on previously validated ICD-9/10 codes and categorized by anatomical site (Tables S1 and S2). Demographic and clinical characteristics were compared between patients with and without bleeding within 12 months of AC initiation using chi-square, Student's t, or Cochran–Mantel–Haenszel tests as appropriate. The association between any bleeds and survival was assessed using Cox proportional hazards regression, adjusting for potential confounding variables. Bleeding was analyzed as a time-varying covariate to account for immortal time bias. The analysis was repeated stratifying bleeds by site (GI, GU, ICH, or Other). We used SAS 9.4 (SAS Inc. Cary, NC, USA) for all analyses. All tests were two-tailed, with p < 0.05 determined as statistically significant. This study was approved by the Saint Louis VHA Medical Center (IRB 1606326-15) and Washington University School of Medicine (IRB 202208139) institutional review boards.

A total of 9326 patients with solid tumor and newly diagnosed VTE were included (Figure S1), with median time from cancer diagnosis to VTE of 95 days. Median age was 68 years and 21.7% of the cohort were black (Table S3). Lung (27.8%) was the most frequent cancer type, followed by prostate (13.3%), and lower GI (11.2%). A total of 3245 (34.8% of the cohort) received chemotherapy. The median time from cancer diagnosis to start of AC therapy was 90 days. Most patients received AC therapy with LMWH (52.8%) or a DOAC (24%). In our cohort, 746 (8.0%) patients developed bleeding with median time from start of AC therapy to bleed of 59 days and 12-month cumulative incidence of 9.1% (95% confidence interval [CI] 8.4–9.7). Bleeding was more prevalent in patients with a history of alcohol abuse, anemia, previous bleeding, frailty, liver disease, metastatic disease, stroke history, thrombocytopenia, and uncontrolled hypertension. Among patients with bleeding, GI was the most frequent bleeding site (56.2%), followed by GU (20.1%) and ICH (10.9%) (Table S4).

Median overall survival (OS) was 14.1 months. There were 2003 deaths (21.5%) within 12 months from AC therapy initiation. For patients with clinically significant bleeding, the median OS was 10.0 months versus 14.8 months (p value = < 0.001) in patients without AC-related bleeding. Following a bleed, the 30-day cumulative incidence of mortality was 19.4%. ICH events had the highest incidence of 30-day mortality of 32.1%, followed by GI bleeds (20.9%), GU bleeds (13.1%), and all other bleeding events (12.6%). After adjusting for potential confounders, AC-related bleeding was associated with a significantly increased risk of mortality (adjusted hazards ratio [aHR] 2.91, 95% CI 2.48–3.42) at 12 months (Figure 1). Other covariates that were associated with mortality (Figure 1) were: older age, BMI < 18.5, certain cancer types (brain tumors, upper GI, non-prostate GU, lung), newly diagnosed cancer, eGFR < 30 mL/min, moderate frailty, liver disease, and metastatic disease. Black race (compared to white), BMI ≥ 25, and treatment with chemotherapy were associated with lower of mortality. Analysis of mortality by site of bleed found the aHR (95% CI) for mortality 5.91 (3.97–8.80) after an ICH, 2.80 (2.29–3.44) after a GI bleed, 1.66 (1.13–2.43) after a GU bleed, and 1.91 (1.22–2.97) after bleeding at other sites (Figure S2). When restricting the cohort to patients diagnosed with VTE within 90 days after cancer diagnosis (n = 4827), the aHR (95% CI) between any bleed and mortality was 2.66 (2.05–3.46) (Figure S3). When analyzed by bleeding site, ICH remained associated with the highest mortality (aHR 4.75, 95% CI 2.53–8.92), followed by GI bleeding (aHR 2.47, 95% CI 1.75–3.48), other bleeding sites (aHR 1.82, 95% CI 0.93–3.57) (Figure S4).

FIGURE 1

Open in figure viewerPowerPoint

Multivariable analysis, association between anticoagulant-related bleeding and mortality in patients with cancer-associated venous thromboembolism (n = 9326).

Our population-based study highlights the high incidence of clinically significant bleeding and the significant association of AC-related bleeding with increased mortality in patients with solid tumor and cancer-associated VTE. In our cohort of 9326 patients, AC-related bleeding was associated with a 2.9-fold increased risk of mortality after adjusting for multiple important confounders. Several studies have highlighted the significant impact of bleeding on mortality in anticoagulated patients, but limited data exist for cancer-associated VTE. Wee et al. found higher bleeding-related mortality (HR 2.66, 95% CI, 1.05–6.73) in patients with cancer-associated VTE compared to non-cancer patients with VTE [3]. A recent analysis of 791 cancer patients (15.2% on AC therapy) by Englisch et al. [5] found that beyond immediate bleeding-related mortality, the occurrence of clinically relevant bleeding was associated with an independent increase in risk of all-cause mortality (multivariable THR 5.80; 95% CI: 4.53–7.43). Our study addresses this gap by providing further evidence and characterization of the significant association between AC-related bleeding and mortality in patients with cancer-associated VTE.

In our cohort, the incident of bleeding (12-month cumulative incidence 9.1%) was comparable to rates reported in prior cohort studies (12-month incidence ~10%) [2, 6]. The effect of bleeding site on mortality was significant. ICH was associated with the highest risk of mortality with a 30-day mortality of 32.1% and a 5.91-fold increased risk of mortality. 30-day mortality following a GI bleeding was 20.9% with a 2.80-fold increased risk of mortality. Other studies have demonstrated the association between major bleeding and mortality risk [3].

This study has several strengths, including a large, nationwide cohort and adjusting for confounders that could affect bleeding and/or mortality. The application of time-varying adjustment for bleeding events allows for a more accurate reflection of the temporal relationship between AC therapy, bleeding events, and mortality, by avoiding an immortal period [7]. This study also has several limitations. We adjusted for cancer type, chemotherapy, new cancer diagnosis, and metastatic cancer, but residual confounding remains possible such as more advanced malignancy predisposing to both bleeding and death. While bleeding was included as a time-varying covariate, only baseline confounders were accounted for in the analysis, which may not reflect the dynamic changes in the clinical profile of cancer patients. Due to limitations of ICD codes in capturing bleeding events, potential underreporting of non-ICH bleeding events can occur. The study population predominantly consisted of male US Veterans, which may limit generalizability to other populations. Additionally, we could not identify the cause of death after the bleeds.

In conclusion, our study demonstrates a significant association between AC-related bleeding and increased mortality in patients with solid tumor and cancer-associated VTE, particularly after ICH and GI bleeding. In the absence of reliable bleeding risk assessment models in cancer patients, the findings underscore the importance of individualized risk assessment and vigilant monitoring in managing AC therapy in this frail population.