Prophylactic tranexamic acid in patients with hematologic malignancy: a placebo-controlled, randomized clinical trial

Hematologic malignancies accounted for approximately 10% of new cancers in the United States in 2020.¹ Ten-year survival rates have improved with aggressive therapy but require substantial transfusion support during prolonged marrow hypoplasia. As early as 1966, prophylactic platelet transfusion was proven to decrease bleeding incidence in acute leukemia.² More than 2 million platelet doses and 11 million red cell units were transfused in 2013, with hematology and/or oncology services accounting for 43% of platelets and almost 20% of red cells transfused.³ 

Despite prophylactic platelet transfusion therapy, the TOPPS (Trial of Prophylactic Platelets) study⁴ reported that 43% of patients experienced WHO grade 2 bleeding (supplemental File 2, Appendix A, available on the Blood Web site). In the PLADO (Platelet Dose Study), WHO grade 2 bleeding occurred in 70% of the 1272 enrolled patients and 79% of patients undergoing allogeneic hematopoietic stem cell transplantation.⁵ WHO grades 3 and 4 bleeding events occurred in 10% of enrolled patients.

Tranexamic acid (TXA) and ε aminocaproic acid inhibit fibrinolysis by blocking lysine binding sites on plasminogen. Both decrease bleeding and mortality in orthopedic,⁶ cardiothoracic,⁷ and organ transplantation surgery,⁸ in trauma,⁹ and in obstetrical bleeding.¹⁰ Antifibrinolytics are commonly used to prevent and treat bleeding in patients with platelet function disorders and inherited coagulopathies.^11,12 Platelets are the largest cellular source of plasminogen activator inhibitors. Clots formed in the presence of thrombocytopenia are friable, lyse rapidly, and result in a poor hemostatic plug,¹³ possibly due to plasminogen activator inhibitor deficiency. Therefore, TXA might be beneficial in thrombocytopenic bleeding. Reports of antifibrinolytic therapy in small numbers of thrombocytopenic patients with hematologic malignancies suggest efficacy in decreasing bleeding, platelet, and red cell transfusion requirements.^14,15

To determine whether TXA could safely reduce bleeding incidence and transfusion requirements in patients undergoing treatment for hematologic malignancies, A-TREAT (American Trial Using Tranexamic Acid in Thrombocytopenia), a multicenter, double-blinded placebo-controlled randomized clinical trial, was initiated. Study results are reported here.

Eligible patients were enrolled at the University of Washington, the University of Pittsburgh, and the University of North Carolina. The human subjects review board at each participating institution approved the study, which was conducted according to the principles of the Declaration of Helsinki and the International Conference on Harmonization Good Clinical Practice guidelines. All patients provided written informed consent. An independent data and safety monitoring board, appointed by the National Heart, Lung, and Blood Institute, regularly evaluated the conduct and outcome of the study and patient safety. S.P.B., S.S., and S.E. analyzed the data, and all authors had access to primary clinical trial data. Study Protocol and Statistical Analysis Plan are available in supplemental File 1. The trial was registered with Clinicaltrials.gov (NCT02578901).

Patients with a diagnosis of hematologic malignancy or aplasia, undergoing chemotherapy, immunotherapy, or hematopoietic stem cell transplant, and anticipated to have hypoproliferative thrombocytopenia resulting in a platelet count of ≤10 000 per µL for ≥5 days were eligible for this study. In addition, participants could not have a history of thromboembolism or cardiovascular thrombotic events, be on anticoagulant therapy, or be considered to be at increased risk of thrombosis. Disseminated intravascular coagulation, immune thrombocytopenic purpura, acute promyelocytic leukemia, elevated platelet transfusion threshold, severe uremia, or dialysis were all exclusion criteria. Patients were not refractory to platelets at the time of enrollment. For a complete list of inclusion and exclusion criteria, see supplemental File 2, Appendix B.

The trial evaluated the efficacy and safety of antifibrinolytic therapy with TXA in preventing bleeding in patients who were thrombocytopenic due to primary bone marrow disorders or chemotherapy, immunotherapy, and/or radiation therapy. When platelet counts fell below 50 000 per μL, patients were randomized to either TXA or placebo, preferably administered orally, unless it was not tolerated. Randomization was stratified according to clinical site and disease group (allogeneic transplant, autologous transplant, or chemotherapy for hematologic malignancy). At the time of randomization, study staff contacted a centralized website, which pulled the randomization assignment from a list based on a computer-generated random number and notified the site pharmacy by e-mail. All study staff except the investigational drug services pharmacists and the unblinded statisticians remained blinded to the patient treatment group during both treatment and data abstraction. When the platelet count fell below 30 000 per µL, patients who had not developed contraindications to treatment were activated to the treatment phase and received TXA (either oral 1.3 g or IV 1.0 g every 8 hours) or matched placebo in a double-blind fashion. Study treatment was initiated as soon as possible following activation, and those patients were considered evaluable. Criteria for permanent or temporary discontinuation of study treatment are provided in supplemental File 2, Appendix C.

All patients received care under the direction of their physician in addition to receiving the study drug. Patients were treated with a prophylactic platelet transfusion once their platelet count fell below 10 000 or at clinician discretion. Patients received daily hemostatic assessments by a study team member while hospitalized and completed self-reported daily hemostatic assessments while in the outpatient setting.

Outcomes and safety data were collected for 30 days after activation or discontinuation (whichever was longer; for study assessments, see study protocol in supplemental File 1). The study staff abstracting the data were blinded to treatment group assignments. Laboratory values and transfusion data were collected daily while inpatient and twice-weekly while outpatient; bleeding assessments (adapted from the WHO bleeding scale)¹⁶ and a thrombotic review were completed daily by an observer on inpatients and through diaries on outpatients. Bleeding was identified by research coordinators trained to administer a questionnaire. Data entered from this questionnaire were verified by blinded auditors from the Data Coordinating Center (see supplemental File 1, Protocol for additional details). Thrombotic events were reported based on data collected from the patient’s electronic medical record 30 days after discontinuation of the study drug and 120 days after activation. When available, coordinators confirmed the reported event through diagnostic imaging reports. Thromboses were identified per attending physician notations. Weekly testing was performed for visual acuity using the Snellen Eye Chart, central visual field using the Amsler Grid, and color perception using the Ishihara color plates up to 2 weeks after discontinuation of the study drug. Race and ethnicity were obtained per electronic medical record documentation and were reported according to local hospital policies.

The primary outcome was the incidence of bleeding of WHO grade ≥2 during the first 30 days after activation. Secondary outcomes were the number of platelet transfusions per subject during the first 30 days after activation and the number of days alive and without WHO grade ≥2 bleeding during the first 30 days after activation.

Outcomes that were assessed for exploratory purposes as alternative measures of bleeding included: the number of days with bleeding or thrombocytopenia during the first 30 days after activation; time from activation to the first episode of bleeding of WHO grade ≥2 or death within 30 days after activation; highest grade of bleeding experienced during the first 30 days after activation; and death ascribed to bleeding during the first 30 days after activation.

Outcomes that were assessed for exploratory purposes for alternative measures of transfusion frequency included: the number of platelet transfusions per subject by thrombocytopenic day (≤10 000 per µL) during the first 30 days after activation; the number of red cell transfusions per subject and by thrombocytopenic day during the first 30 days after activation; survival ≥30 days after activation without a platelet transfusion; survival ≥30 days after activation without a red cell transfusion; and the number of patients with platelet count >10 000 per µL over the study period of 30 days from activation.

Safety endpoints included incidence of serious adverse events, adverse events, thrombotic events, veno-occlusive disease, all-cause mortality within 30 days after study drug discontinuation, and deaths ascribed to thrombosis within 120 days after activation.

With 330 patients activated (165 per group), the trial was estimated to have 88% power to detect a 30% relative reduction in WHO grade ≥2 bleeding rates (57% in the placebo and 40% in the treatment group based on a 1:1 randomization, and a 1-sided level of significance of 0.025). The analysis population included all activated patients; patients were analyzed according to their randomized treatment group regardless of the amount of study drug received. The safety population was limited to patients who received any doses of the study drug. A 1-sided O’Brien-Fleming futility boundary was used for 3 interim formal analyses. No efficacy boundary was used. The primary analysis was based on the score statistic from logistic regression adjusted for site and disease group. Estimates were based on multiply imputed data for patients missing the primary outcome and adjusted for sequential monitoring. The primary outcome was missing for patients who, before experiencing a grade ≥2 bleed, failed to complete their bleeding diaries, withdrew from the study, or died within 30 days of activation. Multiple imputations were based on hazard models, stratified by treatment group, and included site and time-varying absolute neutrophil count as covariates. Details of the imputation model are provided in supplemental File 2, Appendix D. Secondary endpoints and subgroups were analyzed using multiply imputed data in regression analyses adjusting for site and disease group. Statistical inference was based on Wald statistics computed using robust sandwich estimates of standard errors. None of the secondary or subgroup analyses were adjusted for sequential testing or other multiple comparisons, and the results are thus exploratory. Separate logistic regression analyses in patients with and without bleeding at randomization and patients within the disease group (allogeneic and autologous transplant and leukemia) were prespecified subgroups. Clinical importance was explored in post hoc linear regression analyses of the risk difference in the proportion of participants with WHO grade ≥2 bleeding, as well as the proportion of patients with WHO grade ≥3 bleeding.

The inference is based on a 2-sided significance threshold of 0.05; confidence intervals (CIs) are correspondingly two-sided at 95%. Analyses were performed in R version 4.0.2 (R Foundation for Statistical Computing) and SAS 9.4 (SAS Institute Inc.).

Deidentified individual participant data, along with other study documentation such as the protocol, manual of operations, data collection forms, and data dictionary, will be available at the BioLINCC repository (https://biolincc.nhlbi.nih.gov/home/).

The first and last patients were activated on 16 June 2016 and 12 February 2020. The end of the follow-up occurred on 11 June 2020. Of 3120 patients assessed for eligibility, 2764 were excluded, with 1002 not meeting inclusion criteria and 1058 meeting exclusion criteria. In addition, 816 patients did not consent. The most common reason for exclusion was a history of arterial or venous thromboembolism (n = 727), followed by current anticoagulant use (n = 242) (Figure 1). Of 356 patients enrolled, 337 were randomized, and 330 were activated, with 165 activated to each of the 2 treatment groups. Two patients in each of the study groups were activated but never received the study drug. Of the 330 activated patients, 145 patients in the TXA group and 144 patients in the placebo group had complete primary outcome data. One TXA patient and no placebo patients experienced the competing risk of death before grade ≥2 bleeding, and an additional 19 and 21 patients were missing primary outcome data due to failure to keep diaries or withdrawal from the study, respectively. For those 20 and 21 patients with missing primary outcome data in the TXA and placebo groups, respectively, the primary outcome data were multiply imputed (supplemental File 2, Appendix D). Baseline characteristics were relatively well balanced across treatment groups (Table 1). In the placebo group, the study drug was administered for a mean of 14.7 days (standard deviation, 8.4); in the TXA group, the mean duration of treatment was 14.3 days (standard deviation, 7.8) (supplemental File 2, Appendix F). There were no missing thrombotic assessments while in the study. Among patients who received ≥1 study drug dose and neither withdrew nor died, 93% (240 of 258) had at least a baseline, 1 on-drug, and 1 after drug visual assessment. Further details regarding patient disposition, study drug administration, and protocol deviations are provided in supplemental File 2, Appendix F.

The primary outcome of WHO grade ≥2 bleeding during the first 30 days after activation was observed for 73 out of 145 (50.3%) and 78 out of 144 (54.2%) patients in the TXA and placebo groups, respectively. Multiple imputation was used for the 20 TXA and 21 placebo patients missing the primary outcome, leading to an odds ratio of WHO grade ≥2 bleeding of 0.83 (95% CI, 0.50-1.34; P = .44) (Table 2). A graphical representation of the timing of bleeds by group is presented in Figure 2.

The mean number of platelet transfusions per patient during the first 30 days after study activation (secondary outcome) among those with complete data were 7.7 and 7.6 in the TXA and placebo groups, respectively, yielding an imputed mean difference of 0.07 transfusions (95% CI, −1.90 to 2.04; P = .95). The mean number of days alive without grade ≥2 bleeding (secondary outcome) was 28.1 and 27.7 days in the TXA and placebo groups, respectively, among patients with complete data, yielding an imputed mean difference of 0.79 days (95% CI, −0.40 to 1.98; P = .19). The mean number of red cell transfusions per thrombocytopenic day (≤10 000 per µL; exploratory outcome) was 0.4 and 0.4 in the TXA and placebo groups, respectively, yielding a mean difference of 0.03 (95% CI, −0.06 to 0.12; P = .50).

Other bleeding- or transfusion-related exploratory endpoints were similar across the 2 treatment groups. Most prespecified safety endpoints were also similar across treatment groups (Table 3).

In the prespecified subgroups based on disease groups, the odds ratio of grade ≥2 bleeding comparing patients who received TXA to patients who received placebo was 0.96 (95% CI, 0.46-2.00) among patients receiving an allogeneic transplant, 0.62 (95% CI, 0.19-2.08) among patients receiving an autologous transplant, and 0.84 (95% CI, 0.40-1.76) among patients receiving chemotherapy (Table 2). Only 6 patients had bleeding on the day before or the day of randomization, so we were not able to evaluate the prespecified subgroup evaluation of patients with bleeding at randomization.

In a post hoc analysis, the grade ≥2 bleeding risk difference was estimated to be −3.5% (95% CI, −13.6% to 6.5%). Additionally, there were 12 and 9 patients with grade ≥3 bleeding in the TXA and placebo groups, respectively (risk difference, 1.8%; 95% CI, −3.4% to 7.1%; P = .50).

The proportion of patients experiencing serious adverse events and adverse events were similar across groups (Table 3). There was no significant difference in thrombotic events between study groups (6 events [3.7%] in the TXA and 9 events [5.5%] in the placebo group). Veno-occlusive disease was infrequent in both groups (3 [1.8%] in the TXA and 2 [1.2%] in the placebo group). The TXA group had more central line occlusions than the placebo group (27 [16.6%] in the TXA and 11 [6.7%] in the placebo group; 95% CI of the difference, 2.3-17.3%). While visual changes were commonly reported (5% of participants reported new colorblindness and 10% newly abnormal Amsler grid exams), there were no substantive differences in visual changes between the groups (supplemental File 2, Appendix I).

In this study of patients with hematologic malignancy undergoing chemotherapy or hematopoietic stem cell transplantation receiving routine platelet transfusions, the addition of prophylactic TXA did not significantly reduce the risk of the primary endpoint, WHO grade ≥2 bleeding, compared with placebo. WHO grade ≥2 bleeding has not been established as prognostic of more severe bleeding; however, its wide use in clinical trials of platelet transfusion has shown it to be a consistent and reproducible measure of bleeding and, along with the comparison of transfusion, needs are clinically relevant.¹⁷ 

As in other studies, we observed that patients who have longer periods of thrombocytopenia (ie, undergoing allogeneic transplant or chemotherapy) were at a higher risk of bleeding than autologous transplant patients (Table 2). This may account for at least some of this difference in bleeding rates between disease groups.^5,17

TXA is an effective intervention to reduce bleeding in a variety of clinical scenarios, including cardiac surgery,^1,8  trauma,^9,19 postpartum hemorrhage,¹⁰ and joint replacement.²⁰ It is licensed for the prevention of bleeding in hemophilia,²¹ congenital platelet dysfunction disorders (including von Willebrand disease),²² and menorrhagia.²³ However, TXA failed to reduce bleeding in patients with gastrointestinal hemorrhage²⁴ or improve outcomes in subarachnoid hemorrhage,²⁵ suggesting that the mechanism, site, and/or severity of bleeding may be critical determinants of efficacy. Furthermore, even for a given indication, the efficacy of TXA may be limited to a specific time window within a complex physiologic response to vascular injury. For example, in both obstetric bleeding and major trauma, including traumatic brain injury,¹⁹ TXA was effective only if administered within the first 3 hours of onset, with a potential worsening of outcomes beyond this time point. This therapeutic window may coincide with a transient phase of fibrinolytic activation, as has been proposed in trauma-induced coagulopathy.²⁶ Further mechanistic studies are required to define and document whether the presence of fibrinolytic activation determines the clinical efficacy of antifibrinolytic agents in various bleeding scenarios.

There was no significant difference in rates of platelet or red blood cell transfusion between the 2 treatment groups, suggesting no difference in occult bleeding. Previous retrospective and small prospective studies have shown a decrease in red blood cell and platelet transfusions in thrombocytopenic patients receiving antifibrinolytic therapy.^14,15 One hypothesis is that occult microvascular blood loss might be decreased by antifibrinolytic therapy; however, no significant differences in blood product usage were seen in this study.

Thrombotic events were not significantly increased in the group receiving TXA. History of thrombosis was a relatively common exclusion criterion (22% of excluded patients) that may have affected the generalizability of these findings. Furthermore, the 3 academic centers that participated in this trial may not be representative of the breadth of facilities that treat patients with hematological malignancies. The need to clear central venous lines with the use of TPA was significantly higher in the TXA arm. Fibrin sheaths commonly occur within central venous lines without associated venous thrombosis, as was the case in this study.²⁷ This study corroborates the established safety data of antifibrinolytic therapy at these doses. A meta-analysis of >49 000 nonsurgical patients in 22 studies showed a reduction in all-cause mortality without increased risk of venous or arterial thrombotic complications.²⁸ 

This study was not powered to evaluate whether there may be subgroups of this patient population that might still benefit from tranexamic acid to prevent or treat active bleeding in patients at higher risk due to infection, mucositis, graft-versus-host disease, prolonged periods of thrombocytopenia, platelet transfusion refractoriness, or during procedures. Two ongoing trials, TREATT (TRial to EvaluAte Tranexamic acid therapy in Thrombocytopenia)²⁹ and PATH (Platelet Transfusion Requirements in Hematopoietic Transplantation Pilot Study),³⁰ will evaluate additional groups of patients and whether TXA could substitute for routine prophylactic platelet transfusion. Patients with acute promyelocytic leukemia were excluded from our study as bleeding and thrombosis are more common due to a unique pathophysiology.³¹ 

Among patients with hematologic malignancy undergoing chemotherapy or hematopoietic stem cell transplantation, prophylactic treatment with TXA compared with placebo did not significantly reduce the risk of WHO grade ≥2 bleeding.

The authors are thankful for the patients and their families and caregivers who participated in this trial and for staff members who contributed to the success of this trial. Research coordinators: Devereux Fitzgerald-Laverdure, Rachel DeWitt, Colette Norby-Slycord, Elizabeth Robertson, Brett Phillips, Amy Brightman, Pam D'Andrea, and Rosemary Bolinger. Statistical support: David Whitney. Database and website support: Jonas Carson, Winnie Kirdpoo, and Sherrill Slichter for her counsel. Except for Sherril Slichter, all other nonauthor contributors received salary support from the National Heart, Lung, and Blood Institute (NHLBI) for their effort in the trial.

The trial was funded by NHLBI grants HL122894 and HL122272. The trial was conducted under a cooperative agreement with the NHLBI having representation on the steering committee that oversaw the design and conduct of the study, collection, management, analysis, and interpretation of the data, preparation, review, approval of the manuscript, and decision to submit the manuscript for publication. The content of this work is solely the responsibility of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the funder (NHLBI).

Contribution: The investigators designed (T.B.G., D.J.T., N.S.K., N.E.K., and S.E.) and conducted (T.B.G., S.P.B., D.J.T., N.S.K., H.H., S.E., and S.M.) the trial; gathered (T.B.G., D.J.T., N.S.K., J.N.P., M.B., B.N.R., and M.B.P.), analyzed (S.P.B., S.S., S.E., and S.M.), and interpreted (all authors) the data; wrote the manuscript draft (T.B.G., S.P.B., H.H., and S.M.) and made the decision to submit it for publication (all authors). S.P.B. and S.M. had full access to all trial data and take responsibility for their integrity, analytic accuracy, completeness, and the fidelity of this report to the trial protocol which is available with the full text of this article in supplemental File 1; and T.B.G., S.P.B., H.H., and S.M. jointly drafted the first version of the manuscript.

Conflict-of-interest disclosure: T.B.G. reports consultancies with Amgen, Principia, Sanofi, Dova, Cellphire, Rigel, Novartis, Platelet Disorder Support Association; and research funding from Rigel, Principa, and NHLBI (this trial). J.N.P. reports consultancy with TeraImmune. M.B. is employed by Genentech, and with the exception of manuscript review, all work related to the study was performed while at the University of Pittsburgh. The remaining authors declare no competing financial interests.

The current affiliation for J.N.P. is Department of Medicine and Department of Pathology, University of Vermont, Burlington, VT.

Correspondence: Terry B. Gernsheimer, Division of Hematology, University of Washington School of Medicine, Box 356330, 1959 NE Pacific St, Seattle, WA 98195; e-mail: bldbuddy@uw.edu.