Markus Pihusch, Abteilung für Haematologie und Internistische Onkologie, Klinikum der Universitaet Regensburg, Franz-Josef-Strauss-Allee 11, 93053 Regensburg, Germany. Tel.: +49 941 944 5510; fax: +49 941 944 5511; e-mail: email@example.com
Summary. Background: Bleeding is a common complication following hematopoietic stem cell transplantation (HSCT) and standard hemostatic treatment is often ineffective. We conducted a multicentre, randomized trial of the efficacy and safety of activated recombinant factor VII (rFVIIa, NovoSeven®) in the treatment of bleeding following HSCT. Methods: 100 patients with moderate or severe bleeding (52 gastrointestinal; 26 hemorrhagic cystitis; seven pulmonary; one cerebral; 14 other) were included from days +2 to +180 post-transplant (97 allogeneic; three autologous) to receive seven doses of rFVIIa (40, 80 or 160 μg kg−1) or placebo every 6 h. The primary efficacy endpoint was the change in bleeding score between the first administration and 38 h. Results: No significant effect of increasing rFVIIa dose was observed on the primary endpoint. A post hoc analysis comparing each rFVIIa dose with placebo showed that 80 μg kg−1 rFVIIa improved the bleeding score at the 38 h time point (81% vs. 57%, P = 0.021). This effect was not seen at 160 μg kg−1. There were no differences in transfusion requirements across dose groups. There was no trend in the type or number of severe adverse events observed. Six thromboembolic events were observed in the active treatment groups: three during, and three following the 96-h observation period. Conclusions: Despite no overall effect of rFVIIa treatment on primary endpoint, post hoc analysis showed an improvement in the control of bleeding for 80 μg kg−1 rFVIIa vs. standard hemostatic treatment. The heterogeneity of the population may have contributed to the lack of an increasing effect with increased dose. Further trials should focus upon identifying the patient populations that may benefit from treatment with rFVIIa.
Hematopoietic stem cell transplantation (HSCT) is a commonly used treatment modality for a number of life-threatening malignant and non-malignant diseases. Many transplantation-associated complications predispose to bleeding complications . These include prolonged aplasia, septicemia, graft-vs.-host disease (GVHD) and thrombotic microangiopathy (TM) and affect up to 32% of the patients, resulting in substantial morbidity and mortality .
Management of bleeding episodes in HSCT is complex and depends on the causal circumstances. In general, systemic hemostatic treatment aims to correct platelet and coagulation abnormalities by the transfusion of platelets, the application of antifibrinolytic agents, and the use of fresh frozen plasma (FFP). While these measures may attenuate bleeding, severe bleeding often may not be controlled despite additional interventional, endoscopic or surgical action.
Activated recombinant factor VII (rFVIIa/NovoSeven®, Novo Nordisk, Bagsvaerd, Denmark) is a novel and unique hemostatic agent with the potential for broad-spectrum applications in bleeding patients with congenital and acquired coagulopathies due to a variety of etiologies . The hemostatic efficacy of rFVIIa is thought to be the result of binding to the surface of activated platelets with subsequent activation of Factor X and generation of thrombin leading to the formation of a hemostatic plug. The action of rFVIIa is localized to the area of vascular injury, where tissue factor (TF) is expressed and activated platelets are found .
We conducted a multi-center, randomized, placebo-controlled trial to evaluate the efficacy and safety of rFVIIa in the treatment of bleeding complications following allogeneic and autologous HSCT.
Patients and methods
Between April 2001 and October 2003, patients ≥12 years of age receiving allogeneic hematopoietic stem cell grafts were included in this prospective phase II trial. Patients were admitted for a variety of hematological and oncological disorders at various stages of their disease. In August 2002, recruitment was extended to include patients receiving autologous stem cell transplants. The study was conducted in accordance with the Declaration of Helsinki, and signed informed consent was obtained.
Trial design and procedures
This was a multicentre, randomized, double-blind, parallel group, placebo-controlled study evaluating the efficacy and safety of rFVIIa in the treatment of bleeding complications following HSCT. After screening, eligible patients were randomized to receive i.v. rFVIIa at the dose of 40, 80 or 160 μg kg−1 body weight or placebo, administered every 6 h for 36 h in addition to standard management procedures. Patients were followed-up for 96 h following the initial dose (Fig. 1). The randomization was computer-generated and was performed in center blocks with an equal allocation ratio between the treatment groups. The drug was supplied as sterile, freeze-dried powder in single-use vials to be reconstituted with sterile water for bolus injection. To ensure blinding of the investigator to dose allocation, all doses were reconstituted to ensure that an identical volume/kg body weight was administered. A preplanned interim analysis was scheduled following recruitment of 48 patients.
Patients with mild bleeding (score 2) for more than three full consecutive days, or with severe (score 3) or serious (score 4) bleeding episodes according to predefined criteria (Table 1) were included if bleeding occurred between days 2 and 120 after HSCT. The bleeding site was identified at trial entry and was scored at baseline, and subsequently reassessed at 24, 38, 48, 72, and 96 h after the first dose of study medication. If multiple bleeding sites were identified, the site with the highest bleeding score was chosen as the primary site. In cases where the scores were equal, the investigator determined the primary bleeding site. In April 2002 eligibility was extended to bleeding episodes occurring up to 180 days after HSCT provided that patients had a platelet count <50 × 109 L−1.
Table 1. Scoring system for assessing severity of bleeding 
Occult blood in body secretions (detected by heme positive in dipstick), mild petechiae, or minimal vaginal spotting
Minor bleedings that do not require RBC transfusions over the routine transfusion needs (epistaxis, vaginal bleeding, mild hematemesis, melena, and mild hematuria)
Hemorrhage causing rapid decrease in hematocrit level necessitating one or more units of RBC per day over the routine transfusion needs
Life-threatening hemorrhage, defined as either massive bleeding causing severe hemodynamic compromise or bleeding into a vital organ (e.g. intracranial hemorrhage, pericardial, or diffuse alveolar hemorrhage DAH)
Patients were ineligible for the trial if they had any history of active atherosclerotic disease, stroke or deep vein thrombosis during the previous 3 months, disseminated intravascular coagulation (DIC), moderate or severe TM (increased lactate dehydrogenase and >4.8% fragmented cells), severe hepatic venocclusive disease (VOD), active acute myeloid leukemia (AML) FAB types M3, M4, and M5 or had received granulocyte transfusion within the previous 24 h.
Adverse events (AE) and serious adverse events (SAE) were recorded throughout the study. Acute and chronic GVHD (aGVHD, cGVHD) was graded using established clinical and biochemical criteria [5,6] and confirmed by biopsy in a minority of patients. Hepatic VOD was diagnosed and classified according to established signs and symptoms . Cases of thromboembolism were identified using laboratory markers, clinical signs, electrocardiography, ultrasonic examination, phlebography, or computer tomography scan.
Red blood cell (RBC) transfusion was recommended when the hemoglobin level was below 80 g L−1. Platelet concentrates (PC) were to be administered to treat and prevent hemorrhage when the platelet count was below 20 × 109 L−1 and to treat HC and diffuse alveolar hemorrhage when the platelet count was below 75 × 109 L−1. Resistance to platelet transfusion was defined as no detectable increase in platelet count within 24 h after transfusion of two units of platelets (defined as PC derived from a single donor, or a pool of four to six donors, according to local practice)  and/or known presence of platelet antibodies. In addition, the use of topical and systemic antifibrinolytic agents or non-steroidal anti-inflammatory drugs (NSAID) concomitantly with rFVIIa was not recommended.
Blood sampling and assessments
Blood samples were obtained prior to rFVIIa dosing and at regular intervals (at least every 6 h during the first 36 h, then every 12 h) up to 96 h after the first infusion. The blood count was analyzed locally at the participating hospitals using the local standard techniques.
Coagulation parameters and their plasma levels were analyzed centrally (MDS Pharma Services, Hamburg, Germany) by commercial tests (coagulometric assay respectively turbidimetry; Dade Behring, Schwalbach, Germany): antithrombin (AT), fibrinogen, D-dimers, prothrombin fragments (F1+2), activated partial thrombin time (APTT), and prothrombin time (PT). FVII:C levels were determined at baseline, then at 15 min, and 1 h after the first infusion, and at regular intervals during the 96-h follow-up (Fig. 2). Assessments were performed immediately following each scheduled rFVIIa administration. FVII:C levels were determined at a central laboratory (Capio Diagnostics, Copenhagen, Denmark) by the FVII:C assay, a one-stage clotting assay using recombinant human TF (Innovin®, Dade Behring) and immunodepleted FVII-deficient plasma .
Primary efficacy endpoint was the change in bleeding score from baseline to 38 h after the initial dose; 2 h (approximate half-life of rFVIIa) after the last dose of trial medication at 36 h. Secondary endpoints comprised changes in the bleeding score from time of initial dose to 24, 48, 72, and 96 h after initial dose as well as amounts of RBC, platelets, and FFP transfused during the 96-h trial period. Safety assessments included all AEs and SAEs recorded during the 96-h follow-up, as well as changes in coagulation parameters.
Based on simulation with assumptions on the change in bleeding score for placebo (Stopped: 0%, Decreased: 20%, Unchanged: 60%, and Worsened: 20%) and rFVIIa treated patients (160 μg kg−1 group: Stopped: 45%, Decreased: 20%, Unchanged: 30%, and Worsened: 5%), a sample size of 25 patients per treatment arm was planned to provide a statistical power of 85%. The sample size calculation also included an assumption that the expected proportion of patients with HC recruited to the study was 27% . A preplanned interim efficacy analysis of change in bleeding score was performed after inclusion of the first 48 patients and the significance level for analyses of effects on bleeding was adjusted to maintain an overall type I error of 5% (2.9% for interim and final analysis, Pocock's method). Correspondingly, the confidence intervals (CI) presented are the 97.1% CI. The significance level for all other analyses was 5%. All analyses were performed for the intention-to-treat (ITT) population and the per protocol (PP) population which comprised 100 and 91 patients, respectively. Given the exploratory nature of the study and as results based on the ITT and PP population were comparable, only results for the ITT population will be presented.
A one-sided Jonckheere-Terpstra trend test was used to analyse the hypothesis of decreasing bleeding scores with increasing rFVIIa dose. A two-sided Jonckheere-Terpstra was used to analyze the requirements of RBC, platelets, and FFP. Changes in coagulation-related variables were compared between treatment groups by analysis of variance using two-sided F-tests. The proportion of subjects having one or more thromboembolic AE was compared between treatment groups using Fishers exact test.
In order to compare the categorical outcome, namely the change in bleeding score from 0 to 38 h while adjusting for effects of covariates, a proportional odds/cumulative logit model was also developed (an extension of the logistic regression model). The outcome levels were bleeding stopped, decreased, unchanged or worsened. The outcomes unchanged and worsened were grouped as only one patient (placebo group) had worsened bleeding at 38 h.
All covariates thought to have an effect on the bleeding outcome were tested in models including treatment and one covariate effect at a time. The covariates that were significant at the 10% level were included in a final model. In the final model the four treatment groups were compared by the likelihood ratio test. Due to the interim analysis performed halfway through the study, each active group was compared with placebo at a significance level of 0.029.
The statistical analyses were performed using SAS® (Version 8.2, SAS Institute, Cary, NC, USA).
In total, 100 patients from 21 transplant centers were enrolled and randomized. Baseline data and transplant characteristics are summarized in Table 2. Appearance and severity of GVHD, the incidence of infectious complications, and vital signs did not differ between the groups. There was some variability across treatment groups with regard to primary bleeding site. Only two patients in the 80 μg kg−1 dose group suffered from HC, compared with a total of 24 patients with HC in the other three groups. Five patients did not complete the trial, two due to progressive respiratory failure and two due to treatment failure (HC and epistaxis), although all four were included in the primary efficacy analysis. One patient withdrew for personal reasons, and was not included in the primary efficacy analysis. One further patient was not included in the primary efficacy analysis, due to a missing bleeding evaluation at 38 h. Although in the protocol it was recommended not to use medication that can interfere with hemostasis, 34 patients received this kind of medication (heparin: 22 patients; defibrotide: 15 patients; NSAIDs: five patients) during the trial period. At the interim analysis, it was found that the 48 patients recruited included a higher than expected proportion of patients with hemorrhagic cystitis (HC) (n = 22; 45.8%). Further patients with HC (as single site bleeding) were therefore subsequently excluded in order to avoid reducing the power of the study.
Table 2. Epidemiologic and transplantation characteristics of the patients studied
Number of patients (ITT population)
Total (n = 100)
Placebo (n = 23)
40 μg kg−1 (n = 20)
80 μg kg−1 (n = 26)
160 μg kg−1 (n = 31)
Case numbers (percentages) are given. Full intensity: Conditioning regimen containing a dose of 12 Gy total body irradiation or conditioning regimen containing Busulfan (orally administered 4 mg kg−1 BW for 4 days). Reduced intensity: Conditioning regimen containing a dose < 12 Gy total body irradiation and containing no Busulfan.
Age (years [median (range)])
Day post-transplantation [median (range)]
Acute lymphoblastic leukemia
Chronic myeloid leukemia
Stem cell source
Autologous peripheral blood
Allogeneic conditioning regimen
Busulfan containing regimen
Cyclophosphamide containing regimen
Irradiation containing regimen
Bleeding sites at inclusion
Primary bleeding sites The location of the primary bleeding site and severity of bleeding episodes at inclusion are summarized in Table 3. The majority of patients (n = 77; 77%) were bleeding from a single site, while the lower-gastrointestinal (l-GI) tract was the most frequently identified bleeding site (n = 38; 38.0%). Most patients included in the study had a bleeding score of 3 (n = 46; 46.0%) or 2 (n = 41; 41.0%), with upper or lower GI bleeding more likely to be associated with a higher bleeding score.
Table 3. Bleeding events of the patients at inclusion time
Number of patients (ITT population)
Total (n = 100)
Placebo (n = 23)
40 μg kg−1 (n = 20)
80 μg kg−1 (n = 26)
160 μg kg−1 (n = 31)
The table gives the number (percentage) of patients.
No overall significant effect of increasing rFVIIa dose on the change in bleeding score at 38 h post-treatment was observed. A post hoc analysis comparing each dose with placebo showed that the 80 μg kg−1 dose (but not 40 and 160 μg kg−1) resulted in significant improvement in the change in bleeding score relative to placebo [P = 0.0213; cum OR = 4.20; 97.1% CI (1.05; 16.84); Table 4].
Table 4. Bleeding status 38 h after initial dosing
Number of patients (ITT population)*
Total (n = 98)
Placebo (n = 22)
40 μg kg−1 (n = 20)
80 μg kg−1 (n = 26)
160 μg kg−1 (n = 30)
The model includes the following covariates: HC as primary bleeding site, GVHD at baseline and use of antiocagulants.
*Information for two patients was unavailable.
Stopped indicates a score of 0 at 38 h. Decreased indicates a decrease of the bleeding score at 38 h. Unchanged/worsened indicates no change or an increase of bleeding score at 38 h.
Values in parenthesis are percentage.
38 h post initial dosing
Unchanged or worsened
Predictors for change in bleeding score Patients with HC as primary bleeding site, and the use of anticoagulants had a negative impact on the bleeding outcome (P = 0.0173; cum OR = 0.31; and P = 0.026; cum OR = 0.35, respectively). Other variables, including underlying disease (AML, ALL), infections at baseline, conditioning regimen, engraftment, platelet resistance, and the use of hemostatic agents (FXIII, fibrinogen, and antifibrinolytics), had no significant effect on bleeding status, irrespective of treatment group.
Secondary efficacy evaluation
There was no overall significant effect of treatment on bleeding score in patients at 24, 48, 72, and 96 h following initial administration, and there was no significant difference in the proportion of patients responding to treatment (bleeding stopped at 38 h after initial dose) between each treatment group and placebo. There was no overall significant trend toward reduced RBC, PC, or FFP transfusion requirements with increasing dose in actively bleeding patients, or in patients with HC or moderate or severe bleeding, within 96 h after initial administration. Notably, there was an overall significant trend toward higher amounts of RBC and PC transfused with increasing dose in the patient group excluding HC (RBC: P = 0.041; platelets: P = 0.022).
Laboratory data and pharmacokinetics
The plasma concentration-time profile of FVII:C demonstrated a dose dependent increase (Fig. 2). There were no significant differences with regard to baseline values of PT for all dose groups. There were significant changes from baseline vs. placebo in both PT and APTT for all dose groups up to 36 h. Once rFVIIa administration stopped after 36 h, both PT and APTT declined to baseline levels (data not shown).
Adverse events A total of 149 AEs were experienced by 63 (63.0%) patients during the observation period (Table 5). The majority were assessed as mild or moderate in intensity, and 16 were evaluated as being possibly or probably related to the trial product. A total of 41 serious AEs (SAEs) were reported, equally distributed across the four treatment groups. Of these, 13 occurred during the 96-h observation period. No apparent drug or dose-related effect was evident for those SAEs occurring during the 96-h follow-up. Also no significant difference in the number of thromboembolic events was observed (P = 0.41). In the active treatment group, there were two episodes of deep venous thrombosis and one patient was diagnosed with mild TM. Despite the absence of a clinically confirmed thrombotic episode, this episode was also classified as a thromboembolic event. One thrombophlebitis in the active group occurred 2 days after the 96-h follow-up period.
Table 5. Adverse events reported (*within the 96-h study period, **beyond the 96-h study period)
Number of patients (ITT population)
Total (n = 100)
Placebo (n = 23)
40 μg kg−1 (n = 20)
80 μg kg−1 (n = 26)
160 μg kg−1 (n = 31)
Case numbers (percentages) are given. Serious AE: Death, threat to life of patient, in-patient hospitalization or prolongation of existing hospitalization, persistent or significant disability or incapacity; important medical events that may not result in death, be life-threatening, or require hospitalization may be considered a SAE when, based upon appropriate medical judgement, they may jeopardize the patient or subject and may require medical or surgical intervention to prevent one of the outcomes. Thromboembolic events: Events proven as predefined. Thrombotic events in detail:athrombosis of left jugular vein [2 days post last dosing (DPLD)], bvenous catheter thrombosis right axillar vein (2 DPLD), cTM (2 DPLD), dcerebral infarction (4 DPLD), emyocardial infarction (4 DPLD), fthrombophlebitis (4 DPLD).
Adverse events [patients (%)/events]*
Although the use of heparin during the trial was not recommended in the trial protocol, a total of 22 patients received heparin during the 96-h period of the trial. It should be noted that in this setting heparin was used in prophylactic doses (2500–10 000 units per day), and only 15 patients were dosed simultaneously with rFVIIa and heparin. One patient treated with heparin (and 160 μg kg−1 rFVIIa) developed a thromboembolic event, corresponding to an incidence of 4.5% (1/22). Thromboembolic events were reported in 6.4% (5/78) of patients not treated concomitantly with heparin. Based on this numbers, the incidence of thrombotic events does not appear to be different in heparin-treated vs. non-heparin treated patients. Due to the small sample size, a formal statistical analysis was not found to be valid.
There were 31 deaths, occurring at a median of 61 days (range 3–369 days) after initial administration. Two deaths occurred during the 96-h follow-up period, one in the placebo group (due to respiratory failure) and one in the 160 μg kg−1 group (due to acute respiratory distress syndrome). Both were considered unrelated to study medication. Of the remaining 29 deaths occurring outside the 96-h follow-up period, two were assessed as possibly or probably related to study medication (fatal thromboembolic complications – one myocardial infarction and one cerebral infarction) (Table 5).
The first fatal thromboembolic event beyond the observation period was the cerebral infarction 4 days after last dosing in the rFVIIa 80 μg kg−1 group. Upon review of the SAEs, the initial clinical exclusion of TM was revised, based on the examination of peripheral blood smear. The second fatal thromboembolic event was a myocardial infarction. This patient had previously recovered from intracerebral hemorrhage (ICH) in the rFVIIa 160 μg kg−1 group and then received prophylactic off-label administrations of rFVIIa at 80 μg kg−1 up to five times after the trial end. She then experienced her first myocardial infarction and died from a second myocardial infarction. This patient suffered from severe GVHD.
This study represents the first multicentre prospective evaluation of the efficacy and safety of rFVIIa in the treatment of bleeding complications following HSCT. Our major findings can be summarized as follows: An increasing effect on bleeding with increasing rFVIIa dose was not achieved, nor was a significant overall trend toward lower transfusion requirements with increasing rFVIIa dose. HC remains a complex clinical problem not readily treatable with hemostatic agents. Furthermore, prophylactic application of anticoagulants diminishes the chance of achieving control of bleeding. Finally, in terms of safety, it is not possible to completely exclude an increased risk of thromboembolic complications following rFVIIa administration in the HSCT setting. A post hoc analysis comparing each rFVIIa dose with placebo did demonstrate that the 80 μg kg−1 dose was more effective than standard hemostatic treatment in the HSCT setting. However, we were unable to show a significant effect relative to placebo for the higher 160 μg kg−1 rFVIIa dose.
Most patients (n = 86; 86.0%) included were transplanted from an allogeneic hematopoietic stem cell source and entered the trial before days 100 (median days 42, range −1 to 266) post-transplantation. Nearly 60% of these patients suffered from GVHD, confirming recent retrospective data  on the association between GVHD and bleeding. This might be attributed to the thrombocytopenia often associated with GVHD. Nearly two-thirds of the patients had platelet counts <40 × 109 L−1, although only a minority had not achieved neutrophil engraftment. Conversely, GVHD strongly increases the risk of bleeding  by direct local effects. In tissues affected by acute and chronic GVHD, hyperperfusion and proliferation of blood vessels may lead to increased epithelial fragility and subsequent destruction .
The majority of patients included in this study had either moderate or mild score 2 bleeding episodes. The primary bleeding site was most frequently localized in the GI tract, and particularly the lower-GI tract. Bleeding episodes in patients with GVHD most commonly occur in the skin or GI mucosa , and are associated with an increased mortality not readily explained by measurable disturbances in plasma or cellular hemostasis.
In the HSCT setting, ICH is a classical score 4 bleeding, usually associated with a very high acute death rate of over 20% within 1 week after onset, therefore rFVIIa might be an effective treatment. Only one patient was included with ICH, considerably lower than the expected incidence of life-threatening bleeds (>10%). This patient received rFVIIa and bleeding was stopped. Another study also suggested that fungal or bacterial infection may be risk factors for ICH in immunosuppressed patients following HSCT . Our patient, however, had no obvious cerebral infection. A recent multicentre efficacy trial for rFVIIa in ICH has shown that early treatment of ICH (within 4 h of onset and within 1 h of a diagnostic CT scan) significantly reduced hematoma growth, and improved survival .
HC is a common complication in allogeneic HSCT patients, and was the second most frequent bleeding site in this study. It is well known that HC following HSCT usually results from the long-term effects of cyclophosphamide and GVHD , and by superinfection with BK-virus  causing direct injury of the uroepithelium of the bladder wall. HC therefore has a different etiology than the majority of other bleeding episodes seen following HSCT and poses a clinical dilemma for physicians in the transplant setting. The different etiology of HC might explain why we were unable to demonstrate clinical benefit for rFVIIa in terms of the control of bleeding resulting from HC.
Several factors might explain the lack of a dose response in this study. The patients recruited to the study were receiving HSCT as therapy for a range of different conditions. Futhermore, these patients exhibited considerable variability in the primary bleeding site identified at study entry. These differences may have diluted the overall therapeutic effects of rFVIIa. It is worth noting that in the post hoc analysis performed in this study, the only dose to show significant improvement vs. placebo was the 80 μg kg−1 arm, which only included two patients with HC (compared with nine in the 160 μg kg−1 group). This observation suggests that the presence of HC could have been a factor in the apparent lack of a dose response. It was also demonstrated that anticoagulant use had a negative effect on bleeding status in this study, and variability in anticoagulant use may help to account for the lack of effect seen here with increasing doses. However, other factors that were assessed at baseline and that might have affected the response to treatment were shown to have no statistically significant effect. Another important consideration is the choice of dose interval used here. A dose interval of 6 h was chosen to avoid cumulative effects of rFVIIa in the plasma. Given that the documented half-life of rFVIIa is between 2 and 4 h , it is likely that this interval was too long to maintain maximum therapeutic effect. The short half-life of rFVIIa may also mean that the sequential timepoints chosen for evaluating the response of bleeding to rFVIIa might have underestimated the real effect of the drug. The rapid and short-term effects of rFVIIa in other settings, which have been documented and are clinically quite impressive, might therefore have been missed in this trial. Finally, it has been shown that patients with GVHD of the gut have levels of FXIII that are considerably below normal, correlating to the degree of GVHD . FXIII plays an important role in hemostasis and wound healing, and depletion may contribute to the reduced activity of rFVIIa observed here.
This study failed to demonstrate any evidence for reduced transfusion requirements in patients receiving rFVIIa. Although this finding contradicts recent case studies describing rFVIIa use in the surgical setting , the results from randomized controlled studies in the surgical setting are less emphatic with regards to the effects of rFVIIa on transfusion requirements. The PROSE study, a randomized, controlled trial, which examined rFVIIa use in 36 patients undergoing prostatectomy, demonstrated a significant reduction in RBC transfusion requirements . However, in a more recent study examining the use of rFVIIa in liver resection in 204 patients, no significant benefit on perioperative RBC transfusion requirements was observed . The apparent differences between these results may suggest that transfusion requirement is not a reliable endpoint for such studies. This would be particularly approriate in settings where there is routine use of transfusion with platelets and blood products to maintain acceptable platelet counts and hemoglobin values, for example, in patients receiving HSCT. Routine transfusion use might, in this case, preclude observation of a clear effect for rFVIIa on transfusion requirements.
Patients receiving anticoagulants such as heparin or defibrotide for prophylaxis against thromboembolic phenomena or VOD despite a current bleeding episode did not demonstrate adequate response to the infusion of rFVIIa. As there is considerable controversy as to the appropriate management strategy to use in these cases [20,21], the risk-benefit ratio when using anticoagulants in this setting should be carefully assessed.
The number of patients with thromboembolic complications in this trial was similar to that seen in a historical group of HSCT patients . Despite extensive and varied usage of rFVIIa, the incidence of SAE associated with its use in the literature is <1%. However, concerns remain regarding the potential thrombogenecity of this agent. Thrombotic events reported following rFVIIa treatment have been due primarily to improvements in coagulation parameters rather than rFVIIa treatment per se . It is known that TM following allogeneic HSCT predisposes to thromboembolic events  probably because of activation of platelet glycoproteins and the liberation of thrombocytic microparticles .
While we have demonstrated activity of rFVIIa in patients with bleeding following HSCT, the presence of multiple comorbidities and varied mechanisms of bleeding have made it difficult to demonstrate major clinical impact. The diversity of the hemostatic disturbances, the heterogeneity of the patient population and the rFVIIa administration interval may have all contributed to the failure to observe an increasing effect with increased dose in this study. Further trials with larger numbers of patients should focus on identifying the patient populations most likely to benefit from treatment with rFVIIa in this setting, and on establishing the most appropriate dose.
We thank all investigators for their enthusiastic participation in the trial: Australia: Szer J (Melbourne); Denmark: Jacobsen N (Copenhagen); Finland: Ruutu T (Helsinki); Germany: Bornhaeuser M (Dresden), Zander AR (Hamburg), von Depka Prondzinski M (Hannover), Kiehl M (Idar-Oberstein), Pihusch M, Pihusch R, Rank A (Muenchen), Holler E, Pihusch M (Regensburg), Einsle H (Tuebingen), Schwerdtfeger R (Wiesbaden); Israel: Brenner B (Haifa); Italy: Bosi A, Guidi S (Florence), Bacigalupo A (Genova), Polchi P (Pesaro); Singapore: Goh YT (Singapore); Spain: Balbas CS (Santander), Sanz MA (Valencia); Sweden: Ljungman P (Stockholm); Switzerland: Tsakiris D (Basel); United Kingdom: Chopra R (Manchester).
We also thank the Novo Nordisk A/S, Copenhagen Denmark, for the excellent support in all phases of the trial (Albrektsen N, Begtrup K, Bruce-Winkler V, Ferran JM, Moeller-Soerensen T, Payne-Roejkjaer L, Sallah S).
Disclosure of conflicts of interest
Bettina Gaspar-Blaudschun and Liselotte Hyveled are employees of NovoNordisk A/S, whose product was tested in this trial. The other authors declare that they have no conflicts of interest to disclose regarding this study.