NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Yank V, Tuohy CV, Logan AC, et al. Comparative Effectiveness of In-Hospital Use of Recombinant Factor VIIa for Off-Label Indications vs. Usual Care [Internet]. Rockville (MD): Agency for Healthcare Research and Quality (US); 2010 May. (Comparative Effectiveness Reviews, No. 21.)

Cover of Comparative Effectiveness of In-Hospital Use of Recombinant Factor VIIa for Off-Label Indications vs. Usual Care

Comparative Effectiveness of In-Hospital Use of Recombinant Factor VIIa for Off-Label Indications vs. Usual Care [Internet].

Show details

Summary and Discussion


The use of rFVIIa is increasingly frequent for a range of indications not directly related to its FDA approved use in hemophilia. The drug’s biology suggests the potential for hemostatic benefits in uncontrolled bleeding and a risk of thromboembolic adverse effects. Wide in-hospital diffusion of this expensive medication (approximately $10,000 for a single 90 μg/kg dose in a 70 kg patient) has occurred despite limited comparative information.

Based on our evaluation of the Premier database, off-label in-hospital use was estimated to be 125 “cases” (defined as any rFVIIa use during hospitalization) in 2000, underwent a slow increase until 2005 when use became more frequent and was estimated to be 11,057 cases, and by 2008 was estimated to be 17,813 cases (97 percent of all of the estimated 18,311 in-hospital cases), although the rate of increase may be plateauing for many indications. In 2008, the most frequent off-label in-hospital indications for rFVIIa use were adult cardiac surgery, trauma (both at the body and brain), and intracranial hemorrhage, which together account for 68 percent of such use. In contrast, uses for the indications of liver transplantation and prostatectomy were rare. The mean in-hospital mortality rate of 27 percent among cases suggests that rFVIIa is being used in high-risk patient populations that may be more uniformly sick than patients enrolled in studies of rFVIIa. We chose to analyze use by case because it captures the medical decision-making component of care about whether to use or not use rFVIIa for a given patient. Analyses by dosing, rather than cases of use, could have different findings.

Evaluation of the comparative effectiveness of rFVIIa versus usual care for specific indications is limited by a narrow evidence base. We present a summary of our findings in Figure 34. Seventy-four articles met our inclusion criteria for review. Seventeen (23 percent) evaluated off-label indications not included in our subsequent comparative effectiveness review of intracranial hemorrhage, trauma, liver transplantation, cardiac surgery, and prostatectomy. The most frequent indications for rFVIIa use among these were bleeding related to liver disease other than transplantation, obstetrics, neurosurgical procedures, and hematologic malignancies.

Figure 34 depicts a study summary of the use of rFVIIa for indications not directly related to its FDA approved use in hemophilia. As in previous figures, there was no readily apparent pattern of association.

Figure 34

Study summary.

Based on the Premier database analysis, these additional off-label indications that are frequently present in the evaluative literature but are not assessed for effectiveness in this report, accounted for eight percent of in -hospital community practice use in 2008. There were other off-label indications that were prominent in the Premier database but not in the literature —including gastrointestinal bleeding not related to liver disease and primary and secondary clotting disorders, which together account for 12 percent of off-label rFVIIa use in clinical practice. Finally, mortality rates among patients in the Premier database were often higher than the rates among patients who received rFVIIa for similar indications in our effectiveness review, which may suggest that rFVIIa is administered more frequently for end-stage use in the community than it is in published studies (where it is more frequently employed for treatment use).

For the indications whose efficacy we evaluated, there remained 57 studies: 14 randomized controlled trials, 24 comparative observational studies, and 19 non-comparative reports from registries or cohorts (with cohorts limited to those with at least 15 patients). Of the comparative observational studies, the 10 with the highest quality were used for detailed analyses (of effectiveness and harms), while the remaining 12 were used primarily for the harms analyses.

Overall study quality was fair to poor and the strength of evidence low, with the exception of randomized controlled trials for ICH and certain meta-analyses of adult cardiac surgery outcomes. These determinations were made on the basis, most frequently, of the presence of a high risk of bias and imprecise estimates of effect. In addition, clinical efficacy was often defined via indirect/surrogate outcomes, such as transfusion requirements, rather than through direct endpoints such as mortality or functional outcome. In cases where we had evidence regarding both indirect and direct outcomes, a close link between improvements in intermediate outcomes (e.g., cessation of bleeding) and improvements in direct outcomes (e.g., mortality) was not substantiated by the evidence. Possible explanations for this are that use of rFVIIa, despite its reversal of bleeding, cannot reverse more systemic derangements that ultimately influence direct outcomes or that potential harms produced by the administration of rFVIIa (e.g., thromboemboli or other harms not yet identified) outweigh the benefits observed for some selected intermediate outcomes.

There was considerable heterogeneity around the dose and administration of rFVIIa, which made comparisons across studies difficult. Most randomized controlled trials evaluated the effectiveness of a single injection of rFVIIa, albeit with wide dosing range. In contrast, most comparative cohort studies evaluated the use of rFVIIa with more flexible dosing strategies: those that allowed for repeated doses of rFVIIa and determinations by the physicians regarding the exact dose infused. For certain indications, rFVIIa was studied solely as a prophylactic agent (e.g., in liver transplantation) or a treatment agent (e.g., intracranial hemorrhage and traumatic bleeding), whereas in still others it was evaluated, in separate studies, as both a prophylactic and therapeutic agent (e.g., cardiac surgery). The tendency, again not uniform, was for the mean dose of rFVIIa to be lower in prophylaxis trials than in therapeutic trials, and lower for repeated doses in studies of therapeutic efficacy, particularly those that were observational.

The potential for publication bias is an important consideration in any systematic review. The drug’s manufacturer, Novo Nordisk, has played a substantial role in sponsoring, designing, directing, analyzing, and publishing much of the RCT evidence available on rFVIIa. The potential conflict inherent in this role warrants special attention to several forms of bias that may affect the validity of research findings. In particular, publication bias may affect validity both through the failure to publish unfavorable results and delay in publishing results. We have found no definitive instances of publication bias regarding rFVIIa. We note, however, that the results of one intracranial hemorrhage and two body trauma RCTs—directed by Novo Nordisk and completed in 2007 and 2008, respectively—have yet to be published. The small number of studies available on rFVIIa precludes the use of standard meta-analytic tools, such as funnel plots, that would allow for the evaluation of publication bias across the aggregate of studies available. Beyond publication bias, multiple steps in the process of designing, implementing and reporting on RCTs are potentially susceptible to bias. We have found no definitive instances of decisions in clinical trial design and conduct that reflect such bias. We also note, however, that the available studies rely heavily on surrogate/indirect outcomes that could be more likely to yield favorable findings, but may not fully reflect the drug’s effect on direct outcomes. Likewise, the selection of usual care as the comparator, rather than a pre-existing hemostatic agent such as aPCC, could increase the likelihood of favorable findings. These choices do not imply direct or indirect bias, however, and are easily justified as pragmatic decisions seeking to make efficient use of limited research resources.

As evidenced by our harms analyses, there was variability in estimates of mortality and thromboembolic events among the different data sources: the Premier database, RCTs, and comparative and non-comparative observational studies. The higher mortality rates in the Premier database patients may be explained by the real-world application of rFVIIa to patients in extremis, as a last attempt to control hemorrhage, versus what one might imagine is more controlled use in study settings where patient inclusion and exclusion criteria are designed to give the studied agent a realistic chance of performing better than usual care. The harms analyses also indicate, in general, that observational studies reported higher thromboembolic event rates than did RCTs. This is not surprising and, indeed, is part of the reason for including observational studies in any evaluations of harm.

Among the specific clinical indications, there were four RCTs and one observational study of intracranial hemorrhage that evaluated 968 rFVIIa patients. These provided a moderate strength of evidence for all outcomes. The findings are most applicable to patients in their 60s without prior anticoagulation who present for spontaneous intracerebral hemorrhage. In such cases, rFVIIa limits expansion of hematoma volumes but also increases the risk of arterial thromboembolic events when compared to usual care, without having a significant impact on mortality or functional outcome. The currently available evidence thus suggests that neither benefits nor harms substantially exceed each other, which argues against use of rFVIIa in most patients.

The studies of massive bleeding due to body trauma evaluated 267 treated patients in two RCTs and two cohort studies. These suggested no difference in mortality and a possible reduced rate of ARDS (most likely to be present in blunt trauma patients) in patients treated with rFVIIa compared to usual care. There was little evidence of increased risk of thromboembolic events with treatment. But the findings were complicated by the exclusion of patients with early mortality from both of the RCTs and one of the cohort studies. Given that the risks of thromboembolic events in this patient population appear low, the risk-benefit profile of rFVIIa therapy may weigh in favor of its use for body trauma, but this assessment is based on a low strength of evidence that does not permit definitive conclusions. Thus, current evidence of low strength suggests the potential for benefit and little evidence of increased harm

Trials of hemorrhage secondary to brain trauma were limited to one RCT and a sub-set of one cohort study, which included a total of 79 patients treated with rFVIIa. There was no evidence of treatment effect on either mortality or thromboembolic events. Therefore, current evidence of low strength is too limited to compare harms and benefits.

There were four RCTs and one comparative observational study with 215 patients who received prophylactic rFVIIa at initiation of liver transplantation. There was no evidence of treatment effect on either mortality or thromboembolic events. These studies yielded a current evidence of low strength that is too limited to compare harms and benefits. Findings are of questionable relevance given the limited use of rFVIIa reported for this indication.

For the indication of bleeding secondary to cardiac surgery in adults, 251 patients who received rFVIIa were assessed in two RCTs and four cohort studies. From the meta-analyses of the RCTs and good quality observational studies, there was a moderate strength of evidence to suggest an increase in thromboembolic events with use of rFVIIa compared to usual care but a low strength of evidence for the finding of no effect on mortality. The risk-benefit profile of rFVIIa therefore remains unclear.

In pediatric patients undergoing congenital heart defect repair, the one study, an RCT, provided insufficient evidence to determine the effects of prophylactic use of rFVIIa on mortality and morbidity or for comparing harms and benefits. The subgroup of patients requiring ECMO was not evaluated in the RCT included in the comparative effectiveness evaluation.

Finally, the findings in the one study on prostatectomy, an RCT, have been made obsolete by the evolution of care away from the surgeries examined and toward less invasive and less bloody interventions.

In summary, available evidence on off-label rFVIIa use is limited across a wide spectrum of off-label indications. Considering the evidence as a whole, off-label rFVIIa may provide some benefit for certain clinical indications, but this conclusion is largely based on indirect outcomes that have an uncertain relationship to patient survival or functional status. Of the indications we studied, the benefit-to-risk ratio may be more favorable for body trauma than for other indications because its use may reduce the occurrence of ARDS; however, the strength of evidence is low for this as well as most other outcomes, which precludes definitive conclusions. Available evidence does not indicate that use of off-label rFVIIa reduces mortality or improves other direct outcomes for the indications we studied. Thromboembolic events are increased by use of rFVIIa in intracranial hemorrhage and adult cardiac surgery. Despite this state of evidence, in-hospital off-label cases of rFVIIa use have increased in the last decade, particularly for cardiac surgery, trauma, and intracranial hemorrhage.


In general, our systematic review differs from most previous reviews by including information about real-world patterns of use, incorporating data from comparative observational studies (not just RCTs) in the effectiveness review, using non-comparative observational studies in the evaluation of harms, and assessing the impact of dosing level of rFVIIa on outcomes when warranted. Our use of the Premier database accomplished two objectives. First, it provided a method of gauging the relevance of past studies. This allowed us to reinforce the importance of examining rFVIIa use in cardiac surgery, trauma, and intracranial hemorrhage, at the same time pointing to the debatable relevance of investigating rFVIIa use in liver transplantation and prostatectomy. Second, high in-hospital mortality and the limited number of patients discharged home allowed us to conclude that patients in real-world practice who receive rFVIIa are likely more ill than are patients in most studies. Our use of comparative observational studies in the effectiveness review and non-comparative observational studies for a closer examination of the potential harm of rFVIIa corresponds to a growing skepticism of relying on RCTs alone for the assessment of harms.1 Despite increasing the complexity of our literature review, it allowed us to highlight that, for many indications, the mortality and thromboembolic event rates associated with rFVIIa use were higher in the observational studies than in the RCTs. We also assessed the potential need for evaluation of the differential impact of rFVIIa by dosing level. For intracranial hemorrhage, we found evidence that this might be the case and so, a priori, made the decision to evaluate the RCT evidence by low, medium, and high dose use of rFVIIa (see Key Question 2 above for details). Finally, our analysis also included several key trials completed and/or published only recently. Particularly, in the areas of intracranial hemorrhage and cardiac surgery, these trials added important new evidence that might have altered assessment of the harms and benefits of rFVIIa in prior reviews. Despite these potential differences, we discuss below that our findings are similar to those of most other meta-analyses and systematic reviews.

Reviews of rFVIIa use across off-label indications by Cochrane Collaboration researchers and by Squizzato and colleagues did not come to definitive conclusions regarding safety or efficacy,27,63 although the Cochrane review did note a “trend against rFVIIa for increased thromboembolic adverse events.” Hsia an d colleagues found trends toward decreased transfusion requirements and possibly mortality but also toward increased arterial thromboembolic events.230 These results are consistent with our findings of significantly increased rates of arterial and total thromboembolic events with rFVIIa use in studies of intracranial hemorrhage and adult cardiac surgery, respectively. Finally, a recently released technology report from the Canadian Agency for Drugs and Technologies in Health231 performed separate comprehensive assessments of rFVIIa use in RCTs on ICH, trauma (both at the body and brain), gastrointestinal bleeding, and spinal surgery. Investigators for that report found no statistically significant evidence of effect on mortality or thromboembolic events for any indication.

Other reviews and meta-analyses have confined themselves to more narrow clinical indications, but, again, primarily evaluated RCTs. Regarding ICH, a recent meta-analysis on the topic did not incorporate data from the Mayer 2008 trial,88 so that its findings are outdated.34 A recently-updated Cochrane review that included the Mayer 2008 trial evaluated RCTs of “hemostatic drug therapies” for ICH, including rFVIIa. All but two of the 975 participants in the intervention arms of the included trials received rFVIIa, while the two outliers received epsilon-aminocaproic acid (EACA). The review concluded that, with rFVIIa treatment, there was no reduction in mortality but that there was a “trend” toward increased thromboembolic serious adverse events.232 These findings are consistent with ours, with the exception that we focused on all arterial thromboembolic events and found evidence of significantly higher rates with rFVIIa use, but nonetheless found no evidence of differences in outcome by dosing level (whether low, medium, or high).

We identified two systematic reviews of rFVIIa use in body trauma.233,234 Both evaluated RCTs, comparative observational studies, and non-comparative observational studies and cited evidence from the Boffard blunt trauma trial96 that rFVIIa use is associated with decreased RBC transfusion requirements. But, both also concluded that there was no evidence to suggest decreased mortality with treatment in either blunt or penetrating trauma. These findings are generally consistent with those of our effectiveness review, which determined that the risk-benefit profile of rFVIIa therapy may favor its use for this indication, but that this assessment is based on a low strength of evidence that does not permit definitive conclusions.

There have been several systematic reviews of rFVIIa use in cardiac surgery. A meta-analysis on this indication observed no effect on mortality, a non-significant reduction in the rate of surgical re-exploration, and a trend toward an increase in the rate of perioperative stroke,235 but was published prior to the publication of the Gill RCT.119 These findings are consistent with our determination that rFVIIa has no effect on mortality but does differ from our finding of a significantly increased rate of thromboembolic events with rFVIIa use. The other reviews, which did not include meta-analytic evaluations, were perhaps more supportive of rFVIIa use. One found that adequate evidence of efficacy was lacking but that rFVIIa use appeared “promising and relatively safe.” This review calculated an associated thromboembolic event rate of 5.3 percent,51 which is lower, in general, than the rates in the studies we evaluated, but which was calculated using data from studies in pediatric, as well as adult, patients, while we evaluated pediatric patients separately. A Canadian Consensus Conference on application of rFVIIa in cardiac surgery reviewed the literature and then provided expert consensus opinion—namely, that rFVIIa not be used as prophylaxis, given a pattern of higher thromboembolic events with use, but that it might reasonably be used as rescue therapy.26 The Audit and Guidelines Committee of the European Association for Cardio-Thoracic Surgery (EACTS) reached a similar conclusion that “after cardiac surgery, intractable bleeding refractory to conventional haemostatic intervention may be treated successfully with factor VIIa, but there is a small risk of serious or fatal thrombotic complications (Grade C recommendation).”236

Interestingly, reviews of rFVIIa use in other types of surgery have reached similar conclusions and have generally highlighted a lack of evidence of increased thomboembolism. A meta-analysis of case series and RCTs on patients receiving abdominal, vascular, and urologic surgery noted a reduction in bleeding and no increased risk of thromboembolism,50 although a similar study of vascular surgeries alone identified three cases of arterial thrombosis.25 Another meta-analysis of patients undergoing major surgical procedures found a reduction in transfusion requirement and, again, no increased risk of thromboembolism.45 A meta -analysis of applications of rFVIIa to vascular surgery concluded that rFVIIa may reduce hemorrhage,50 but that more study was needed.

Specific to the topic of harm, O’Connell and colleagues published an important evaluation of data on post-marketing adverse events associated with off-label rFVIIa use reported to the FDA’s Adverse Event Reporting System (AERS).237 They documented many instances of arterial and thromboembolic events, which often resulted in serious morbidity and mortality. A subsequent review of the safety profile of rFVIIa for patients with coagulopathy due to anticoagulation, cirrhosis, or trauma found a six percent thromboembolism rate.43 Such studies raised specific concerns regarding the safety of the drug for off-label indications. Since that time, many of the studies, including the meta-analyses noted above, have emphasized safety evaluations, but few have found definitive evidence of increased risk. In our analyses, which had the benefit of recently published trials and of evaluating comparative observational studies in addition to RCTs, we note an increased risk of arterial thromboembolism in use of rFVIIa for ICH and cardiac surgery. However, we found little evidence of increased risk of such events for the other clinical indications; albeit the strength of evidence for these is low or insufficient.

Limitations of the Premier Database Analysis

There are several limitations that must be acknowledged regarding our analysis of in-hospital rFVIIa prescribing practices. Most relevant are those limitations connected to the Premier database itself and our hierarchical definition of indications for rFVIIa use.

First, the Premier database is largely a convenience sample of non-federal U.S. hospitals that are willing to provide detailed cost and billing information. Although the sample is nationally representative in terms of hospital characteristics, there may be important differences between participating and non-participating U.S. hospitals. It provides no information about use in federal hospitals or outside the U.S. The database also captures only in-hospital use of rFVIIa. Second, it fails to include office-based rFVIIa administration, which is certainly pertinent to on-label use in patients with hemophilia and inhibitors, and may be considerable. Third, the data elements available are collected for administrative purposes, rather than designed to capture all clinically relevant information. For example, the data do not allow use as prophylaxis versus treatment versus end-stage use to be distinguished. In addition, we are unable to identify cases where rFVIIa use is linked to important harms outcomes, such as thromboembolic events. Fourth, there may be overlap between the patients identified in the Premier database and the U.S. patients reported in the research studies analyzed. Given the relatively few American patients included in the largest clinical trials, however, this overlap is likely quite limited. In fact, the experience represented by the Premier database is several-fold larger than the aggregate international experience reported in research studies. Fifth, we are limited to the calculation of in-hospital mortality. Given the substantial fraction of rFVIIa patients discharged to skilled nursing facilities and rehabilitation hospitals, in-hospital mortality may substantially underestimate total mortality attributable to patients’ underlying indications or to treatment with rFVIIa itself. Finally, but importantly, our ability to define cohesive indication categories is limited by the relatively sparse clinical information available. Our hierarchy of clinical indications thus may not precisely define the reason for rFVIIa use in all cases. For example, by prioritizing trauma above many other indications we run the risk of misclassifying patients into this category who have minor trauma but also have another more direct indication for rFVIIa (e.g., gastrointestinal bleeding from a non-traumatic cause). Our hierarchical scheme for assigning a single indication may fail to recognize patients with more than one indication (e.g., ICH in a patient with liver disease).

Limitations of the Systematic Review

A weakness of the systematic review component of this report is the limited evidence foundation upon which it builds. There are few large, high quality studies in any clinical indication. The strength of evidence for any particular outcome was almost uniformly low. We encountered further limitations in the areas below.

Non-English Language Articles

Because we were not in a position to translate articles on Key Questions 2–4 that were published in non-English languages, we were unable to fully incorporate these into our systematic review. However, there were only eight such articles that were otherwise eligible for inclusion, and six of these were case series with 35 or fewer patients. The one identified non-English language RCT211 was used for sensitivity analyses for the adult cardiac surgery indication.

Unpublished Studies Identified in the Grey Literature

We identified five clinical trials on Key Questions 2–4 that were registered on online databases (e.g., and completed, but for which we were unable to identify a subsequent publication: two on ICH, two on body trauma, and one on adult cardiac surgery. We did not have the data from these trials, and so were unable to assess them in our effectiveness review or use them to evaluate for evidence of publication bias. We also identified three abstracts reporting on studies that appeared to meet all other inclusion criteria for Key Questions 2–4 but for which were were unable to identify a subsequent full publication. One of these was an RCT on liver transplantation,83 while the others were case series with 24 or fewer patients.

Predecessor Products

As expected, we found no studies comparing use of rFVIIa to predecessor products.

Comparisons to Usual Care

We found evidence of significant evolution in usual care in the area of prostatectomy, enough so as to make the findings of the single RCT on the topic likely obsolete for common practice.141 Similar, although less striking, concerns arose when considering the evolution of practice toward fewer transfusions and more limited blood loss in liver transplantation and adult cardiac surgery. In addition, we found evidence of important changes regarding transfusion practice (including the 1:1 ratio of transfused blood products and institution of massive transfusion protocols) in the treatment of massive hemorrhage in body trauma, which may reduce the marginal benefit of rFVIIa use for this indication.

Manufacturer with Intensive Involvement in Research Studies

As noted above, we found no evidence of bias associated with manufacturer involvement in the majority of RCTs. However, because there were so few comparative studies on any single indication, we were unable to perform formal assessments for publication bias, such as with funnel plots.

Use of Drug in Emergency Situations

At least two of the trials (in body trauma) experienced withdrawal of consent by patients who had been enrolled under some version of an emergency exception to informed consent. In this case, there was no information provided on which treatment arm the patients had been assigned to, so we were unable to assess whether differential withdrawal of consent may have introduced bias, however unavoidable.

Variability in Dose, Repeat Dosing, and Context of rFVIIa Use

Studies varied widely in the doses of rFVIIa used, as well as in their protocols regarding repeat dosing. In addition, within the adult cardiac indication, separate studies evaluated use of rFVIIa for prophylactic and therapeutic purposes, which had the potential to increase the heterogeneity of clinical context and patient characteristics (although this does not appear to have been the the case).

Pediatric Patients

We found evidence of substantial use of rFVIIa among pediatric populations in the Premier database (primarily for cardiac surgery and neonatal applications). Yet we identified only four comparative studies conducted in children and only one that met our inclusion criteria for detailed review, a poor quality RCT on repair of infant congenital heart defects. Although rFVIIa is being used in pediatric patients on ECMO, there were no fair or good quality studies that evaluated such patients.


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (3.0M)

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...