Association between anticoagulant-related bleeding and mortality in patients with hematological malignancies and cancer-associated venous thromboembolism

Study population

We included patients in the VA healthcare system with hematological malignancies from 2012 to 2020 and newly diagnosed cancer-associated VTE plus a new prescription for AC therapy. We defined cancer-associated VTE if the VTE occurred within 3 months before up to 6 months after a new hematologic malignancy, or at any time after a hematologic malignancy that was considered incurable. Incurable malignancies consisted of hematological malignancies that are typically considered incurable: chronic lymphocytic leukemia, chronic myelogenous leukemia, low grade lymphomas (e.g., follicular lymphoma) myelodysplastic syndrome, myeloproliferative neoplasms, and plasma cell neoplasms. Hematological malignancy was identified using the International Classification of Diseases (ICD), 9^th and 10^th revisions codes. VTE was identified using a previously validated algorithm [11]; however, patients were only included if they had a new prescription for AC therapy for at least 30 days within 30 days of VTE diagnosis [11]. Candidate AC therapy included direct oral anticoagulant (DOAC), low-molecular-weight heparin (LMWH), or a vitamin K antagonist (VKA). To minimize misclassification, patients were excluded if they had an outpatient prescription for AC therapy within 6 months prior to VTE diagnosis for any indication.

Data collection and variables

Data were obtained from the VA Corporate Data Warehouse by using the VA Informatics and Computing Infrastructure platform. Baseline demographic data (Supplementary Table 1) were collected at the time when AC therapy was started (index date). The VA-Frailty Index (FI) was calculated as previously described [12] but with the exclusion of bleeding-related ICD codes. The VA-FI was selected over alternative measures of comorbidity given the validation with VA data and its association with mortality in this population [13].

Clinically relevant bleeding (exposure of interest)

We identified clinically relevant bleeding events within 12 months after start of AC therapy, defined as bleeding associated with admission to the hospital using a previously validated algorithm [14] (Supplementary Table 2). For intracranial hemorrhage (ICH), gastrointestinal (GI), epistaxis, and gross hematuria, ICD codes in any position for inpatient admissions were included, given their high positive predictive value (PPV) [14]. For all other bleeds, coding was limited to the primary position. We stratified site of bleeding into four categories: intracranial hemorrhage (ICH), gastrointestinal (GI), genitourinary (GU), and other bleeding sites.

Mortality (primary outcome)

The primary outcome was death within 12 months after the start of AC therapy. Date of death was identified using the VA Vital Status File, which combines the death dates from multiple sources including the VA Beneficiary Identification and Records Locator Subsystem Death File, the Social Security Administration Death Master File, the Medicare Vital Status File, and the Medical SAS Inpatient Dataset. Survival was also measured from the date of a clinically relevant bleeding to the date of death as a secondary outcome.

Statistical analysis

Continuous variables were summarized by mean (standard deviation [SD]) or median (inter-quarter range [IQR]), as appropriate, and categorical variables by count and percentage. The baseline demographics and clinical characteristics between patient groups with and without bleeding events were compared using chi-square, student’s t test, or Cochran-Mantel-Haenszel tests as appropriate. We evaluated the association between clinically relevant bleeding events and death within 12 months of AC initiation using multivariate Cox proportional hazards regression, accounting for multiple pre-defined potential confounders with known associations with mortality (age, race, body mass index [BMI], cancer type, cancer therapy, chronic liver disease, kidney function (by estimated glomerular filtration rate [eGFR], extranodal involvement, and frailty). To account for immortal time bias, AC therapy was a time-varying covariate. We repeated the analysis stratified by site of bleeding. We performed a sensitivity analysis limiting the cohort to patients with cancer-associated VTE within 90 days after diagnosis of hematological malignancy). 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.

The Saint Louis VHA Medical Center and Washington University School of Medicine institutional review boards approved the study prior to cohort assembly and analysis.

Study cohort

A total of 1825 patients with hematological malignancies and cancer-associated VTE prescribed AC therapy were analyzed (Figure 1). Most cancer-associated VTE included pulmonary embolism (N=761, 41.7%) followed by lower extremity deep vein thrombosis (N=733, 40.2%), while 12.4% (N=227) presented with an isolated upper extremity deep vein thrombosis (Supplementary Table 3). Of the 1825 patients, 1008 (55.2%) were hospitalized at VTE diagnosis. Their mean age was 69 years (SD 10.9), 95.8% (N=1748) were male, and 22.8% (N=416) were Black (Table 1). The most common type of cancer was lymphoma (N=976, 53.5%), followed by plasma cell neoplasm (N=478, 26.2%) and myeloproliferative neoplasm (MPN, N=151, 8.3%). Half of the patients (N=913) were receiving anticancer therapy at the time of AC start. AC therapy included DOAC (N=712, 39.0%), LMWH (N=668, 36.6%), and warfarin (N=445, 24.4%). Concurrent antiplatelet therapy (aspirin and/or clopidogrel or ticagrelor) was used in 355 (19.5%) patients. The median time from cancer diagnosis to start of AC therapy was 251 days.

One hundred twenty-three clinically relevant bleeding events occurred within 12 months of starting AC (cumulative incidence 7.3%, 95% confidence interval [CI] 6.1%-8.6%), with the median time from start of AC therapy to first bleeding event of 56 days. The most frequent sites of bleeding were GI (n=71, 57.7%), GU (n=17, 13.8%) and ICH (n=12, 9.8%) (Table 2). Patients who developed bleeding were more likely to have baseline anemia, extranodal involvement of their malignancies, a history of bleeding, and moderate-severe frailty (Table 1). They were also more likely to be receiving aspirin and anticancer therapy.

The median overall survival was 48.3 months, with 162 deaths (8.9%) occurring within 12 months after AC therapy start (Figure 2). The median overall survival was 24.9 months in patients who had a clinically relevant bleed vs. 50.2 months in those without bleeding (p = <0.0001). When stratified by site of bleeding, ICH had the highest 30-day mortality at 41.7%, followed by GI bleeds (9.9%), bleeding at other sites (8.7%), and GU bleeding (5.9%). Of the 123 patients who developed bleeding in the study, 73 (59.3%) had a subsequent prescription for anticoagulant therapy within 90 days. Most subsequent prescriptions (N=53, 72.6%) occurred within the first 30 days after bleeding. Of the patients who resumed anticoagulant therapy, 8 (11.0%) died within 90 days. Of the 50 patients who did not resume anticoagulant therapy, 19 (15.4%) died and 31 (62%) survived to 90 days.

In multivariable analysis, AC-related bleeding was associated with tripled risks of death at 12 months compared to those without bleeding (adjusted Hazard Ratio [aHR] 3.26, 95% CI 1.96-5.45) (Table 3A). Additionally, age (per year increase), anemia (hemoglobin <10 g/dL), chronic liver disease, and extranodal involvement were associated with an increased risk of death (Table 3A). A BMI ≥25 was associated with a decrease in the risk of death. When stratified by site of bleeding, ICH was associated with the highest mortality (aHR 13.41, 95% CI 4.13-43.48) followed by GI bleeding (HR 4.55, 95% CI 2.48-8.35) (Table 3B). Bleeding at other sites, including the GU tract, was not associated with an increased risk of death.

Sensitivity analyses

When limiting the cohort to patients with VTE within 90 days of hematological malignancies diagnosis (N=616), results were similar with bleeding associated with tripling the risk of death at 12 months (aHR 3.22; 95% CI: 1.44-7.19). Furthermore, ICH and GI bleeding remained to be associated with the highest risks of death (aHR 8.64, 95% CI 1.08-69.16 and aHR 5.36, 95% CI 2.13-13.47, respectively).