Bleeding outcomes in thrombocytopenic acute leukemic patients with venous thromboembolism

Abstract Cancer‐associated thrombosis in acute leukemia patients with severe thrombocytopenia (platelets ≤50 × 109/L) poses a management challenge due to competing risks of bleeding and recurrent thrombosis. A retrospective analysis was conducted to determine the occurrence of clinically relevant bleeding (CRB) rates during treatment for acute venous thromboembolic events (VTE) in thrombocytopenic acute leukemic patients. A cohort of 74 patients were subgrouped into three VTE‐treatment interventions: anticoagulation (n = 24), inferior vena cava filter placement (n = 22), and observation (n = 28). Multivariate analysis found a significant correlation between CRB occurrence and quantity of overall blood transfusions, chemotherapy administration, and relapsed leukemia presentation. There was no difference in the occurrence of CRB between VTE‐treatment subgroups, regardless of initial platelet count at the time of VTE diagnosis. Regarding the hematologic parameters, only the velocity of the platelet count recovery was associated with the risk of bleeding. From this analysis, it appears the trajectory of the platelet count and the factors associated with a slower recovery of it, are the main determinants for the occurrence of hemorrhagic complications during VTE treatment in acute leukemia.


INTRODUCTION
Compared to patients without malignancy, cancer patients carry an increased risk of both bleeding events, as well as venous thromboembolic events (VTE) during the course of their disease and treatment [1,2]. Bleeding in those patients can occur as a result of many intrinsic and extrinsic factors including anemia, thrombocytopenia, vascular integrity, coagulopathy, malignancy, chemotherapy, and infection [3]. The risk of developing a VTE in patients with malignancy is seven times higher than that of the general population; 28 times higher in patients with hematologic cancers [4]. Patients with hematologic malignancy often develop thrombosis as a result of a combination of multiple characteristics such as tumor type, obesity, immobility, complete blood count variables (ie, platelets, hemoglobin, leukocytes), central venous device use, and antineoplastic chemotherapy [5]. It is estimated that 20 to 30% of VTE are associated with malignancy [6] and are one of the leading contributors to mortality in cancer patients [7].
Incidence of VTE in patients with acute leukemia has been recorded between 2% and 12%, often within the first month of leukemia diagnosis [8]. Management of cancer-associated thrombosis (CAT) is therefore especially prudent in this population. Many patients with hematologic malignancies also frequently develop severe thrombocytopenia (platelets ≤50 × 10 9 /L) during the course of their disease and treatment [9]. In thrombocytopenic patients, platelet count has an imprecise association with increased risk of bleeding. In the PLADO trial, the risk of bleeding was elevated in those with platelet counts ≤5 × 10 9 /L compared to those with platelet counts ≥81 × 10 9 /L, although there was otherwise no clear correlation of decreased bleeding risk with increased platelet counts [10]. Though thrombocytopenia may be associated with bleeding, there is no platelet count threshold at which the risk of bleeding cannot be accounted for [3]. The rates of bleeding in prior studies of patients with CAT and prolonged thrombocytopenia have ranged between 7% and 33% [11,12]. Prolonged thrombocytopenia for greater than 30 days in patients with CAT not only correlates with an increased risk of bleeding but has also been associated with increased risk of recurrent VTE [6,12]. Thrombosis and bleeding events negatively impact the quality of life of patients with malignancy and can possibly postpone or halt treatment for their cancer. Development of CAT in patients with thrombocytopenia therefore poses a unique management challenge due to competing risks of bleeding events and recurrent thrombosis.

Main outcome measures
To examine the rates of CRB by the VTE-treatment intervention group in acute leukemia patients with CAT and thrombocytopenia. Association between the CRB events and clinical and laboratory characteristics was analyzed. Outcomes determined within 12 months from initial VTE diagnosis included recurrent VTE and location, CRB occurrence, location, and other complications of therapy.  [14]. Clinically relevant nonmajor bleeding (CRNMB) was defined as any sign or symptom of bleeding which did not meet the ISTH criteria for major bleeding and prompted face-to-face evaluation, medical intervention, hospitalization, or increased level of medical care [15].

Statistical analysis
Demographic data and clinical and laboratory parameters were collected (Table 1), and bleeding events retrospectively captured (Table 2).
Multivariate regression models compared the association between CRB and multiple clinical variables (Table 3)

RESULTS
There were 74 patients out of a cohort of 2705 with acute leukemia and VTE who met inclusion criteria for analysis. Reasons for exclusion have been described in a previous publication of this same cohort [13]. Demographics, clinical, and laboratory profiles of patients are included in Table 1  Bleeding severity classification and location, along with blood product transfusion profiles, are listed in Table 2. Characteristics associated with bleeding events in the univariate analysis are described in Table 3. There was no difference between those who bled and those who did not in terms of gender, ECOG performance status, history of prior/concurrent malignancy, history of VTE, or histopathology type. Patients with relapsed or refractory leukemia bled more than those with de novo leukemia (P = .0294). The use of AC did not increase the occurrence of CRB (P = .6035). Patients who received chemotherapy showed a trend to bleed more than those who did not receive chemotherapy (P = .0654). There was no difference in bleeding events between groups based on treatment with hematopoietic stem cell transplantation (HSCT), immunotherapy, or tyrosine kinase inhibitor (TKI) therapy. Patients with CRB received more transfusions of platelets (P = .0012), packed red blood cells (pRBC) (P = .0017), and fresh frozen plasma (FFP) (P = .0573) compared to those who did not have a CRB event. There was no difference in serum creatinine, prothrombin time, or partial thromboplastin time between those who suffered CRB and those who did not. There was a statistically significant difference in the median of fibrinogen levels between groups (P = .0155), but all median fibrinogen levels between groups were >300 mg/dL (Table 3).
We evaluated the platelet count change over time by bleeding group within 30 days after the index VTE ( Figure 1

DISCUSSION
Historically, acute VTE in the general population was treated initially with a brief period of unfractionated heparin (UFH) or low-molecular weight heparin (LMWH), followed by long-term oral anticoagulation with a vitamin K antagonist (VKA) [16]. The comparison of Lowmolecular-weight heparin versus oral anticoagulant therapy for the prevention of recurrent venous thromboembolism in patients with cancer (CLOT) trial in 2003 showed that in cancer patients with CAT, dalteparin (LMWH) decreased VTE recurrence rates at 6 months without an increased risk of bleeding or mortality compared to VKA [17]. There was subsequently a similar but nonstatistically significant trend in VTE reduction in the 2015 CATCH trial, comparing tinzaparin (LMWH) to VKA. Tinzaparin showed a significant decrease in rates of CRNMB compared to VKA, but there was no significant difference in overall mortality or MB [18]. Other studies have supported the use of LMWH in the treatment of VTE in cancer patients [19], and adjusted-dose LMWH for those with CAT and thrombocytopenia [20], establishing it as the standard of care for this patient population and currently recommended by international guidelines [21].
New clinical trial evidence for the use of direct oral anticoagulants counts between 30 and 50 × 10 9 /L, with discontinuation of pharmacologic anticoagulation in platelet counts <30 × 10 9 /L [21,30]. Previous analysis has shown that appropriate anticoagulation therapy, as described above, is associated with significantly improved overall survival (OS) without an increased rate of CRB events compared to other therapeutic options for CAT. The risk of CRB episodes has been associated with a number of variables, including history of bleeding, infection, coagulopathy, liver or renal dysfunction, tumor type, and method of anticancer therapy [12].
One prior study of thrombocytopenic cancer patients showed that low platelet counts increased the chance of prophylactic platelet transfusions, but not bleeding events. Bleeding risk in this study was increased with antiplatelet and anticoagulant use, prior hematuria or gastrointestinal bleeding, infection, low hemoglobin, and elevated creatinine and urea nitrogen [31]. Another study in patients with CAT undergoing autologous HSCT revealed that there was no difference in either recurrent VTE or CRB events whether anticoagulation was continued or held temporarily during thrombocytopenia. This study also showed that there was no threshold at which higher platelet counts could predict a decreased risk of bleeding among these patients [32].
Regardless of platelet count, these acute leukemia patients have prothrombotic states associated with VTE occurrence. Original analysis of this cohort of patients at our institution found a worse OS trend with both CAT and thrombocytopenia. There was a significant improvement in VTE recurrence and OS in patients treated with anticoagulation, however, suggesting its effectiveness for use in this patient population [14]. Regardless of VTE-treatment intervention and severity of thrombocytopenia, patients died from similar causes of either recurrent VTE, CRB, progression of cancer, or other causes [14].
Our analysis delved further into examining the risk of recurrent bleeding rates among these patients based on VTE-treatment intervention, and clinical and laboratory parameters, specifically severity and duration of thrombocytopenia. We followed thrombocytopenic patients' consecutive platelet counts for up to 30 days, beyond which time, platelet counts were less consistent between patients. Our findings are similar to prior studies which showed that the risk of both VTE recurrence and CRB events were similar between patients with or without significant thrombocytopenia, treated with standard therapy [11,12,31,32]. There was no significant difference in other laboratory parameters (ie, markers of renal function and coagulation) between groups who bled and did not bleed. There was a variety of bleeding presentations in our cohort among those who developed CRB, ranging on a spectrum of both major and nonmajor bleeding events ( Anticoagulation administration at our institution reflects the most recent society recommendations [29,30], and we use a LMWH "sliding scale" during thrombocytopenia: subcutaneous enoxaparin 1 mg/kg every 12 hours for platelet count ≥50 × 10 9 /L, 0.5 mg/kg every 12 hours for platelet count 25-49 × 10 9 /L, and suspend anticoagulation for platelet counts <25 × 10 9 /L.
Limitations of this study were retrospective analysis nature, very sick patient population, and relatively small subset of patients. Patients were not randomized to VTE-treatment subgroup, which was determined based on clinical judgment at the time of each individual case.
One third of patients received pharmacologic AC, and the other twothirds did not receive any AC for their VTE. Many patients had platelet counts which could not be continually maintained above 50 × 10 9 /L [8], which may have influenced the treating physician in deciding not to use pharmacologic AC at time of VTE. Previous analysis in this cohort showed that patients who were observed without treatment had a significantly lower platelet count than those treated with AC [13]. It also found those treated with AC were more likely in leukemia-remission state, and IVCF was placed in patients with either DVT or DVT/PE [13]. Therefore, the argument could be made that patients at lower risk for recurrent VTE and CRB were treated with AC, while higher risk patients were managed with either IVCF or observed.
In some cases, patients were not able to be analyzed for longer than 30 days from initial VTE index event for recurrence of VTE, recurrent thrombocytopenia, or development of CRB. This made it difficult to create an accurate longitudinal prediction model for the effect of thrombocytopenia upon CRB risk over a longer period of time. Additionally, we could not analyze the effect of posttransfusion platelet counts on the risk of bleeding. Nevertheless, our data analysis included platelet count trajectories that may be more representative than single and cross-sectional measures when assessing the impact of platelet count and bleeding in acute leukemia patients [33].
There was a significant correlation between occurrence of CRB and quantity of overall blood product transfusions, chemotherapy usage, and relapsed leukemia presentation. Prolonged duration of thrombocytopenia in refractory patients receiving antineoplastic therapy often require more frequent blood product transfusions, which offers a plausible relationship for higher risk of bleeding. There was an indirect relationship between median fibrinogen level and bleeding rates among patients (although all median fibrinogen levels between groups were >300 mg/dL). There was no difference in the occurrence of CRB between VTE-treatment subgroups (AC vs IVCF vs observation), regardless of platelet count at time of VTE diagnosis. From this retrospective analysis, it appears that bleeding complication during AC therapy for CAT in severely thrombocytopenic patients with hematologic malignancy is associated with the duration and severity of thrombocytopenia. The AC treatment strategy for CAT in severely thrombocytopenic leukemic patients with reduced-dose LMWH does not seem to represent an additional risk for bleeding occurrence.

AUTHORSHIP CONTRIBUTIONS
CMR and THO conceived the study, performed data analysis,

DATA SHARING STATEMENT
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.