PubMed Health. 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.)

Results

Summary of Studies and Data on In-Hospital, Off-Label rFVIIa Use

Table A in the Executive Summary provides an overall “snapshot” of the data and results for each of the Key Questions, as well as the conclusions of our effectiveness review.

Summary of Included and Excluded Studies

Our searches for studies on off-label use of rFVIIa identified 5,668 potentially relevant articles of which 1,326 merited full-text review (Figure 2). A total of 74 articles met our inclusion criteria for this review (Table 11). Seventeen of these only met our inclusion criteria for the first part of this report, the section on delineating the breadth of use of rFVIIa and the comparative studies of its use for the clinical indications not covered in Key Questions 2-4. An additional 57 articles met our inclusion criteria for the comparative effectiveness review, of which 14 were RCTs, 24 were comparative observational studies, and 19 were non-comparative reports from registries or cohorts (cohorts limited to those with at least 15 participants). Table 11 gives summary data on the characteristics of included studies. Most of the studies had small to moderate sample sizes, and there was great variability in the doses of rFVIIa administered. Tables 12 and 13 summarize the distribution of included studies by indication, design, and how they are used in this report. While the majority of studies reported data on mortality and thromboembolic events, almost none were powered to distinguish important differences in the rates of these between treatment groups (Appendix C, Appendix Table 2). Instead, most studies used indirect endpoints (e.g., change in hematoma volume or transfusions requirements) as their primary outcomes (Appendix C, Appendix Table 2).

Figure 2 shows that a search for studies on off-label use of rFVIIa
identified 5,668 potentially relevant articles, of which 1,326 merited
full-text review. The figure follows the citations from the 5,668 unique
citations to the 1,326 requiring full text review to the 74 unique citations
included in the analysis. These 74 are divided at the bottom of the figure
into three boxes: RCT has 24, comparative observational has 31, and
non-comparative observational has 19.

Figure 2

Search results for included studies. “Small sample size” applies only to non-comparative studies with less than 15 patients. “Duplicate publication” includes studies with overlapping patient populations.

Table 11. General characteristics of all included studies by clinical indication.

Table 11

General characteristics of all included studies by clinical indication.

Table 12. Clinical indication and study type of included studies.

Table 12

Clinical indication and study type of included studies.

Table 13. Study categories by study design and use in analysis.

Table 13

Study categories by study design and use in analysis.

Unpublished Studies in the Grey Literature

Studies identified via online databases where clinical trials are registered but without a subsequent publication. Our search of online resources (e.g., ClinicalTrials.gov or the manufacturer’s website) for ongoing or completed RCTs that were registered on an online database identified 29 trials with and without subsequent publications, either in the form of an abstract or published article (Appendix Table 4). Of the completed RCTs on Key Questions 2-4, 11 of 16 (69%) have been published: 71% (5/7) on intracranial hemorrhage, 50% (2/4) on body trauma, 50% (1/2) on adult cardiac surgery, and 100% on brain trauma (1/1), liver transplantation (1/1), and pediatric cardiac surgery (1/1). Three of five of the unpublished studies (one on intracranial hemorrhage and two in body trauma) were sponsored by the manufacturer.

Abstracts identified without subsequent publications. Our literature search identified 76 abstracts for which we were unable to find a subsequent full publication. All but three of these were excluded based on our exclusion criteria. Of these three abstracts, one described an RCT of rFVIIa use in liver transplantation with 12 patients in each of the treatment and placebo arms.83 The remaining two abstracts described case series, containing 17 and 24 adult cardiac surgery patients, respectively.84,85

Non-English language studies. We found 134 articles written in 15 different languages: 45 potentially addressing Key Questions 2-4 indications (intracranial hemorrhage, body and brain trauma, liver transplantation, adult and pediatric cardiac surgery, and prostatectomy) and 89 addressing other indications. Because we were not in a position to translate all of the different languages, we relied on English language abstracts and information in indexing databases to determine if articles met our inclusion criteria. Of the 45 articles on Key Questions 2-4 indications, 37 (82%) were excluded because they met other exclusion criteria. The remaining 8 (17%) were excluded solely on the basis of being published in a non-English language (Appendix Table 5). Of these, two were comparative studies. One was an RCT on adults who underwent cardiac surgery in China (11 patients each in the rFVIIa and placebo groups); this article was used for sensitivity analyses in the adult cardiac surgery section of Key Question 3. The other was a comparative cohort study of brain trauma with seven patients in the treatment group. The remaining six articles were case series (range of 15 to 34 patients). Of the 89 articles on other indications, 83 (93%) were excluded because they met other exclusion criteria; of the remaining six that were excluded on the basis of non-English language, four were case series, one was a comparative observational study and for one the study design was unclear.

Quality of Studies Included in Key Questions 2-4

The assigned grades of the two independent assessors never differed by more than one level of the categorical grading schema (good, fair, or poor). Of the RCTs assessed, there was initial agreement on the grade assignment in 13 of 14 (92%). Of the comparative observational studies assessed, there was initial agreement on the grade assignment in 22 of 24 (91%). All disagreements related to whether to categorize the given study as “poor” or “fair.” These were successfully resolved by discussion.

Overall, the published literature contained relatively few fair or good quality comparative studies within any given clinical indication (Table 14). Three of 14 (21 percent) RCTs were determined to be poor quality. In all cases, there was little or no description of the methods of randomization or allocation concealment or of the blinding of subjects, providers, or outcomes assessors, along with other methodologic shortcomings. Poor quality RCTs were still included in the evaluations of effectiveness for each clinical indication. In contrast, 14 of 24 (58 percent) comparative observational studies were judged to be of poor quality and were not included in the comparative effectiveness review but were used for qualitative sensitivity testing, and their data on harm (in the patients who received rFVIIa) were included in the results section on non-comparative evidence of harm. In all cases where comparative observational studies were determined to be of poor quality, there were study designs that were inappropriate for generating comparability of groups, inadequate methods to control for confounding, lack of blinding of outcomes assessors, or differential follow-up/high rates of missing data, among other methodological shortcomings.

Table 14. Study quality assessments.

Table 14

Study quality assessments.

Summary of Data from the Premier Database

The search of the Premier database identified nationally representative information on patterns of inpatient use of rFVIIa from 235 U.S. hospitals of the 615 hospitals in the database (38 percent). From 2000 through 2008, the database identified a total of 12,644 hospital cases (any use during a patient hospitalization) where rFVIIa use was reported. Our results indicate d that real-world application of rFVIIa was concentrated among three of the key clinical indications that are the focus of this report: intracranial hemorrhage, body trauma, and cardiac surgery. Application to these indications encompassed 55 percent of hospital cases of rFVIIa use from 2000 to 2008, and 68 percent of hospital casesin 2008 alone. There was variability in use for other indications, as discussed below under Key Question 1.

Key Question 1. Overview of off-label rFVIIa use in-hospital and comparative studies

Overview of Trends in Factor VIIa Use in United States Hospitals, 2000–2008

The majority of use of rFVIIa occurs in the outpatient setting, and the majority of outpatient use is for on-label indications related to hemophilia. Nonetheless, in-hospital U.S. sales of rFVIIa in 2007 are estimated to have been $138.5 million.8 Data from the Premier sample of 615 U.S. hospitals provide nationally representative information (via weighted estimates) about patterns of inpatient rFVIIa use. From 2000 through 2008, there were an estimated 73,747 hospital discharges in the U.S. where rFVIIa use was reported. Over this period, there was growth in in-hospital use of rFVIIa where the unit of analysis was a “case” of use (any application during a patient hospitalization). While most cases were limited to hemophilia and several related hematologic conditions from 2000–2001, such use has leveled over time. In contrast, recent years have witnessed more frequent off-label in-hospital cases. Off-label in-hospital rFVIIa use was estimated to be 125 cases in 2000, underwent a moderate 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(Figures 3 and 4). The most prominent and rapidly increasing indications are cardiac surgery and traumatic bleeding, with cardiac surgery demonstrating the most sustained increase in use. More modest use is associated with non-traumatic intracranial hemorrhage, liver disease, gastrointestinal bleeding, and aortic aneurysm (Table 15 and Figures 3 and 4).

Figure 3 depicts the growth of in-hospital, off-label uses of rFVIIa
in Premier database for hemophilia, ICH, trauma, liver transplant, CV
surgery, prostatectomy, and other uses. The x axis starts at 0
hospitalizations and ends at 8,000. The y axis shows years from 2000 to
2008. The line graph shows the most change in all other off label
uses.

Figure 3

Growth of in-hospital, off-label vs. on-label uses of rFVIIa in Premier database.

Figure 4 depicts the growth of in-hospital, off-label vs. on-label
uses of rFVIIa in Premier database from 2000-2008. It examines the following
hospitalizations: hemophilia, ICH, trauma, CV surgery and all other
off-label use. It shows how over time cases of use for hemophilia increased
3.7 fold since 2000 but then plateaued; how CV surgery was the most frequent
and most rapidly rising indication, although the frequency level may be
leveling off; how use of rFVIIa for trauma grew continuously between 2002
and 2007 but leveled off in 2008; and how ICH reached sizable scale only in
2005, growing rapidly until 2008. Figures 3 and 4 present similar
information. However, “all other” in Figure 4 includes liver transplantation
and prostatectomy, as well as all of the other indications.

Figure 4

Growth of in-hospital, off-label vs. on-label uses of rFVIIa in Premier database, 2000–2008. Figures 3 and 4 present similar information. However, “all other” in Figure 4 includes liver transplantation and prostatectomy, as well (more...)

Table 15. Summary data from Premier database according to clinical indication for rFVIIa use.

Table 15

Summary data from Premier database according to clinical indication for rFVIIa use.

Hemophilia A and B, and related conditions. Initial use of rFVIIa was limited to the FDA-approved indications, including hemophilia A and B, as well as related conditions, such as other factor deficiencies, von Willebrand Disease and Glanzmann’s thrombasthenia. Over time, cases of use for hemophilia A and B increased 3.7-fold since 2000 (131 during 2000 versus 498 during 2008) but then plateaued, while there has be en a 7.4-fold increase in cases for related conditions (107 during 2000 versus 792 during in 2008). These two groups remained the most frequent cluster of indications from 2000 through 2003. Together, they accounted for 10 percent of all reported cases, but their representation among indications for in-hospital use fell over time, from 92.9 percent of cases in 2000 to just 7.1 percent of cases in 2008 (498 cases). Hemophilia A and B, by themselves, accounted for only 2.7 percent of cases in 2008.

Cardiovascular surgery. rFVIIa use in cardiac surgery was initially observed in 2002, and by 2008 was the most frequent (29% of cases, along with trauma) and most rapidly rising indication. Use in pediatric cardiac surgery, largely for rep air of congenital anomalies, increased only modestly, accounting for 2.3 percent of cases overall and 2.1 percent in 2008. On the other hand, rFVIIa use in adult cardiac surgery, largely for aortic valve, mitral valve and CABG procedures, rapidly increased over time, accounting for 16.4 percent of cases overall and 2 6.6 percent in 2008. By 2008, use in all cardiac surgery (5,250 cases) was nearly four times higher than in 2005 ( 1,375 cases), indicating rapid adoption of rFVIIa use for this indication, although the frequency of use may be leveling off.

Trauma. Traumatic bleeding represents the first area of major rFVIIa usage beyond hemophilia and related conditions. Sizable use began in 2002 and it remained the dominant indication for off-labelin-hospital cases of rFVIIa use until it was matched by cardiac surgery (at 29% of casess) in 2008. Use of rFVIIa for trauma grew continuously between 2002 and 2007 but leveled off in 2008, the first indication of stabilization in usage patterns. Nevertheless, trauma other than brain trauma remained the second most frequent indication for rFVIIa use and accounted for 15.9 percent of overall 2000–08 cases and 17.6 percent of cases in 2008 (3,214 cases). Use in brain trauma, particularly traumatic subdural hematoma, grew over time to constitute 9.7 percent of overall 2000–08 cases and 11.1 percent of cases in 2008 (2,033 cases).

Intracranial hemorrhage. rFVIIa use in non-traumatic intracranial hemorrhage, particularly intracerebral hemorrhage(ICH), reached sizable scale only in 2005. Use for this indication then grew rapidly, with cases in 2008 (2,005 cases) nearly 8-fold higher than was reported in 2004 (253 cases). Notably, however, use of rFVIIa for intracranial hemorrhage fell slightly from 2007 to 2008. Intracranial hemorrhage accounted for 10.5 percent of rFVIIa cases overall and 11 percent in 2008.

Liver disease. A range of indications related to liver disease collectively constitute another cluster of modest rFVIIa use. Overall, these uses accounted for six percent of all cases, including liver transplant (0.2 percent), liver biopsy (1.2 percent), variceal bleeding (1.2 percent), and other liver indications (3.3 percent). There were 954 cases estimated for 2008, which was down from a peak of 1023 cases in 2007.

Other conditions. Other gastrointestinal bleeding, particularly hemorrhage from the colon, accounted for 5.3 percent of cases overall. Management of aortic aneurysm, in the presence and absence of surgical intervention, contributed modestly and stably to overall use of rFVIIa, with 1.7 percent of cases overall and in 2008. Other vascular surgery accounted for 2.1 percent of cases overall and in 2008. A range of other conditions contributed minimally to rFVIIa use, including pulmonary indications (1.5 percent of cases overall, particularly biopsy and lung transplant), cancer-related conditions ( 1.5 percent, particularly leukemia), neonatal use (1.0 percent), and obstetrical conditions( 0.9 percent, particularly post-partum hemorrhage). A variety of other hematologic conditions were associated with rFVIIa use (5.1 percent of cases overall, 4.2 percent in 2008), particularly secondary thrombocytopenias and complications of warfarin anticoagulation. rFVIIa use also was associated with a wide variety of other surgical procedures, although none are individually prominent. Together, these procedures account for 5.3 percent of cases overall and 2.2 percent in 2008. Brain surgery, as the most frequent procedure among these, constituted 0.6 percent of cases overall and 0.7 percent in 2008. Of note, despite its prominence in this report, prostate surgery was an exceedingly rare indication for rFVIIa use, comprising only an estimated 120 cases nationally from 2000 through 2008 with no cases noted in 2008. As a whole, procedural use of rFVIIa declined between 2006 and 2008.

Age and gender distribution. Age and gender distribution. Overall, about 26 percent of in-hospital rFVIIa cases were in patients under the age of 45 years of age. Consistent with the growth of off-label indications, there was a significant increase in the mean age of patients from 3 years in 2000 to 59 years in 2008. The age distribution of rFVIIa use varied enormously by indication. Use for Hemophilia A and B was predominantly in patients 25 years of age and younger (73 percent). At the other extreme, 58 percent of rFVIIa cases in intracranial hemorrhage were for patients 65 years of age and older, with 36 percent in those 75 years and older. Beyond intracranial hemorrhage, other conditions where use in the elderly (>65 years) was prominent included aortic aneurysm (82 percent of cases), prostatectomy (66 percent), brain trauma (58 percent), adult cardiac surgery (57 percent), and gastrointestinal bleeding (57 percent) (Table 16).

Table 16. Mean age and disposition of patients who received rFVIIa during hospitalizations from 2000–2008.

Table 16

Mean age and disposition of patients who received rFVIIa during hospitalizations from 2000–2008.

In-hospital use of rFVIIa in the early 2000s was almost exclusively in males (98 percent of cases in 2000), as is expected from the inheritance pattern of Hemophilia A and B. Given the expansion of use into off-label indications this differential has diminished over time. A male predominance persisted (63 percent in 2008) largely due to a preponderance of men treated for the most frequent indications of adult cardiac surgery (68 percent) and body trauma (68 percent).

In-hospital mortality. Overall, in-hospital mortality was substantial among patients receiving rFVIIa, with 27 percent of patients dying while hospitalized (Table 16). Only 43 percent of patients were discharged directly home. A small percentage of patients receiving rFVIIa were discharged to hospice (2 percent). Most of the remaining patients were transferred to other facilities, including nursing homes, rehabilitation hospitals, and other acute care facilities (29 percent). Mortality increased substantially over time from five percent in 2000 to a peak of 31 percent in 2004, before it declined to 27 percent in 2008. Across all of the reported indications, mortality was infrequent only for rFVIIa use in Hemophilia A and B (4 percent). The most substantial mortality rates were associated with aortic aneurysm (54 percent), neonatal use (47 percent), variceal bleeding (39 percent), other liver disease (40 percent), liver biopsy (36 percent), vascular procedures (39 percent), intracranial hemorrhage (34 percent), brain trauma (33 percent), body trauma (33 percent), and gastrointestinal bleeding (30 percent). The populations receiving rFVIIa for adult and pediatric cardiac surgery experienced 23 and 22 percent in-hospital mortality rates, respectively, in contrast to patients undergoing prostatectomy who had a mortality rate of zero.

Hospital characteristics. rFVIIa use was reported in 235 of the 615 hospitals (38 percent) represented in the Premier database. Most of these hospitals had minimal and sporadic use of rFVIIa, whereas the ten hospitals with the highest number of uses by discharge accounted for 46 percent of all rFVIIa cases of use. These same hospitals had a particularly large share of pediatric cardiac surgery. They had a much smaller share of adult cardiac surgery, consistent with the wider diffusion of rFVIIa use to other hospitals for this indication. A majority of rFVIIa use occured in non-teaching hospitals (68 percent of cases). Over time, the proportion of use of rFVIIa in non-teaching hospitals grew from just 10.6 percent of cases in 2000 to a peak of 73 percent in 2005 and a similar but lower proportion of 67 percent in 2008. The majority of cases of rFVIIa use for each indication occured in non-teaching hospitals, with the exception of hemophilia (41 percent) and liver transplantation (10 percent). Fifty-six percent of cases occurred in hospitals with less than 500 beds.

Geographically, a majority of cases of rFVIIa use occured in the South (52 percent), with much smaller shares in the West (25 percent), Midwest (12 percent), and Northeast (11 percent). While the South is the most populous region in the U.S., the cases of rFVIIa use were disproportionate to its share of population (36 percent), total hospitalizations (38 percent), and hospital procedures (40 percent). In contrast, the Midwest and Northeast together (with 23 percent of cases) comprised comparable shares of population (40 percent), hospitalizations (41 percent), and hospital procedures (39 percent). These regional variations were present for most indications, including the FDA-approved indication of hemophilia, where 47 percent of cases occured in the South and only 7 percent in the West.68

Sensitivity analyses. We also conducted several sensitivity analyses of the coding scheme we used to define clinical indications for rFVIIa. In general, moving indications up or down in the hierarchy did not greatly change their reported frequency. While still modest, the greatest change occurred when trauma was moved to near the bottom of the hierarchy, reflecting the co-occurrence of other indications in patients with trauma (e.g., when a patient with liver disease experiences a traumatic injury). Nonetheless, we believe that it is reasonable to give trauma priority as a diagnosis in these instances, as it was likely trauma and not the associated diagnoses which instigated the use of rFVIIa.

Overview of All Identified Comparative Studies of Off-Label rFVIIa Use

Our literature search identified 55 comparative studies on any off-label indication of rFVIIa use. Seventeen (31 percent) of these addressed clinical indications not assessed in Key Questions 2–4 of this review (Table 17).

Table 17. General characteristics of all comparative studies on off-label rFVIIa use for clinical indications not assessed in Key Questions 2–4.

Table 17

General characteristics of all comparative studies on off-label rFVIIa use for clinical indications not assessed in Key Questions 2–4.

Indications and populations for which rFVIIa has been studied. Comparative studies available for rFVIIa use in cardiac surgery (12), trauma (9), intracranial hemorrhage (8), liver transplantation (8), and prostatectomy(1) accounted for 69 percent (38/55) of all such studies. Additional indications included other liver disease (liver resection, liver biopsy, variceal bleeding, and all other liver-related indications), skin grafting, a variety of cranial and spinal neurosurgical procedures, orthopedic surgery, dengue hemorrhagic fever, and hematopoietic stem cell transplantation and other cancer treatment-related conditions.

While the leading off-label uses of rFVIIa (cardiac surgery, trauma, and intracranial hemorrhage) were each represented by a number of studies, beyond these indications there were different types of mismatches between patterns of in-hospital community practice use and the availability of comparative studies. There were prominent community uses that lacked studies, such as primary clotting disorders other than hemophilia, secondary clotting disorders, and gastrointestinal bleeding not related to liver disease. Other indications with no studies included aortic aneurysm, other vascular procedures, and neonatal use (beyond cardiac surgery). In contrast, there were indications that had been studied but where community use was limited. According to the Premier database, from 2005 to 2008 there were fewer than 30 annual cases of rFVIIa use in prostatectomy and liver transplantation. Use in pelvic fracture, skin grafting, and spine surgery also was limited. Finally, many studies examined only the mode of prophylactic rFVIIa use for clinical indications where treatment or end-stage use may also be frequent.

In general, study patients were younger and had lower clinical acuity than did patients in community practice (Table 16). This was particularly true for body trauma where the mean age in the community was 53 years compared to 32 years in the trauma RCTs. In addition, mortality rates in community practice (Table 16) generally were higher than those noted in the comparative studies. For example, in the intracranial hemorrhage RCTs, the reported 90 day mortality was 20 percent compared to 34 percent in-hospital mortality in the Premier database.

In summary, some study populations overlap with the major off-label indications for in-hospital rFVIIa use in community practice, while others represent indications with minimal use. Anticipated applicability of existing studies may be compromised by study of less acutely ill patients and a mismatch between indications studied and patterns of community use by indication and mode of use (i.e., prophylaxis versus treatment versus end -stage therapy).

Characteristics of comparative studies. The comparative studies of rFVIIa use compared this product to the usual care of hospitalized patients. There were 24 RCTs and 31 comparative observational studies (38 on Key Question 2–4 indications of which 22 (57.8 percent) had fair or better quality (Table 14)) available on rFVIIa use across a wide variety of clinical indications. Because the clinical circumstances of rFVIIa use differ so fundamentally between different indications, the suitability of pooling information across clinical indications is controversial.

Study sample sizes in existing comparative studies were generally small. Across all studies, the median and mean number of patients receiving rFVIIa were 24 and 58, respectively. For RCTs, the median and mean were 62.5 and 95, respectively. In addition, where multiple doses were studied it was typical to have fewer than 20 patients in each dosage arm. Evaluations of intracranial hemorrhage were the exception to the general rule of relatively low sample sizes, with the two largest RCTs together containing information on almost 900 patients treated with rFVIIa.

rFVIIa dosages varied enormously from 5 to 956 mcg/kg of patient weight for the studies examined in the effectiveness review s. The 5 mcg/kg dose was reported in several dose ranging studies. The large doses were used in studies where multiple doses were given over an extended time period and, given the 2.5 hour half-life of rFVIIa, the aggregate effect of these sequential doses may not have been greater than the effect of the single doses of 160–200 mcg/kg given in other studies.

The primary outcomes of most of the comparative studies were indirect outcomes that assessed surrogate markers or components of the health care process rather than patient -centered outcomes. Common outcomes included RBC transfusion requirements, blood loss, and duration of surgery or ICU stay. While these outcomes may be correlated with direct clinical outcomes, this correlation is incomplete and there are many instances where discrepancies are expected. Many studies included direct outcomes as secondary end-points. These included mortality, thromboembolic events, functional status, and other clinical events such as acute respiratory distress syndrome. Unfortunately, most studies were individually underpowered to evaluate these direct outcomes, although some studies of intracranial hemorrhage were an important exception.

It is also notable that most clinical trial research on rFVIIa was sponsored by Novo Nordisk, the product’s manufacturer. In many instances, clinical trials were directed and their data analyzed and reported with the involvement of the manufacturer.

Based on the characteristics of the published studies reviewed above, it is anticipated that the strength of evidence of these studies maybe compromised by small study size, inconsistent study quality, heterogeneity by dosage and indication, use of indirect outcomes, and potential conflicts of interest.

Key Question 2. Intracranial hemorrhage and comparative effectiveness of rFVIIa

Background

All of the RCTs evaluated in this section focus on intracerebral hemorrhage (ICH), rather than other forms of intracranial bleeding (e.g., subarachnoid or subdural hemorrhage). In the one comparative observational study, half of the patients had ICH while the other half had isolated subdural hematomas. Intracerebral hemorrhage is associated with high levels of mortality and functional disability. Over one third of patients die within one month, 50 percent have poor functional status at time of discharge, and 20 percent remain institutionalized at three months.159,160 Early hematoma growth occurs even in the absence of detectable systemic coagulopathy and is an important independent predictor of mortality and morbidity.33,161 There are no proven therapies for ICH.162 The purpose of this section is to describe the comparative studies of rFVIIa versus usual care for the treatment of intracranial hemorrhage, but the section necessarily focuses primarily on ICH because the majority of studies focused on this form of hemorrhage.

Usual care during the time frame of included studies. While there remains wide variation in practice patterns, usual care for ICH has evolved over the time span during which the studies included in this effectiveness review were conducted (February 2000 to February 2007). The most notable change in practice is the trend toward less tolerance of “permissive hypertension” because of the suggestion in recent studies, albeit far from conclusive, of increased hematoma expansion with higher blood pressures and potentially better outcomes with the control of blood pressure within six hours. 163–165 The largest and highest quality studies discussed below are those RCTs that were conducted most recently, between May 2005 and February 2007. The included trials did not specify hypertension management.

General Characteristics of Studies of Intracranial Hemorrhage

In the area of ICH, we identified four published RCTs (two good quality,23,88 two fair quality86,87 (Table 14)) and one comparative observational study 89 (fair quality) that examined treatment use of rFVIIa in 968 intervention patients. The 944 patients who received rFVIIa in the RCTs were not on anticoagulation therapy and had experienced intracerebral hemorrhage (ICH), a subset of intracranial hemorrhage. All of the trials were performed by the same study group, were sponsored by the manufacturer of rFVIIa, and had at least one author or biostatistician who was an employee of the manufacturer. We identified an additional RCT on the Novo Nordisk site that appears to have been performed by a different research group and also sponsored by the manufacturer,166 but this study has not yet been published in any form, enrollment is small (45 patients in treatment and control groups, respectively), and the online posting gives only summary details. In the comparative observational study by Ilyas 2008,89 half of the patients had experienced ICH and the other half had experienced subdural hematomas as their manifestation of intracranial hemorrhage.

Place of studies within analytic framework. All of the included studies of intracranial hemorrhage evaluated rFVIIa for treatment use (versus prophylaxis or end-stage use, which are other potential uses, as outlined in our Analytic Framework (Figure 1)). The RCTs had well-documented approaches to data collection and analysis of direct outcomes such as mortality, thromboembolic events, and functional outcome, although only one of them was powered to analyze direct outcomes as its primary outcome, and it used a combined endpoint of “severe disability or death.”88 The Ilyas observational study examined time to correction of INR as its primary endpoint.

Qualitative considerations of heterogeneity. Intracranial hemorrhage can occur inside the hemispheres of the brain (i.e., intracerebral hemorrhage, or ICH) or around the hemispheres (e.g., sudural or subarachnoid bleeding). Because the underlying causes and associated risks of these types of bleeding can vary, we considered results for them separately. Second, baseline coagulopathy is another source of potiential heterogeneity, because patients who are coagulopathic—generally from oral anticoagulation therapy with warfarin —may differ in important ways from those who are not coagulopathic. Again, we considered these patient groups separately in our discussion. Third, because there were indications in the literature regarding a possible dose response relationship between rFVIIa and certain outcomes (e.g., thromboembolic events) and multiple doses of rFVI were analyzed in each RCT, we chose to analyze the data according to low, medium, and high dose rFVIIa groups, as described in our methods and defined as less than or equal to 40 μg/kg, greater than 40 but less 120 μg/kg, and at least 120 μg/kg, respectively. Finally, three of the RCTs defined “poor functional outcome” as a modified Rankin scale score of 4–6, whereas the fourth RCT defined it as a a score of 5–6 but also included a graphical representation of the scores which allowed us to perform our own calculation of the proportion of patients with scores of 4–6, such that the data could be combined with the other RCTs.

Comparison to studies on other indications. The RCTs on ICH had the longest and most comprehensive follow-up of all of the indications. The largest of them also enrolled more patients than studies in any other indication. On average, the patients tended to be older than those in most other indications, but were similar to those in adult cardiac surgery and prostatectomy. The dose range of rFVIIa was from the low-to-middle range of doses used across indications.

Patient Characteristics and Study Design

RCTs. Table 18 summarizes important characteristics of each of the trials, all of which had similar patient populations and methodologies. The two large RCTs randomized a total of 399 and 841 patients respectively,23,88 whereas the smaller trials randomized 47 and 40 patients, respectively.86,87 In all of the trials, important inclusion criteria were that patients were required to have received a baseline CT scan within 3 hours of symptoms onset and study drug within 4 hours. Similarly, important shared exclusion criteria were that a patient was known to be taking oral anticoagulants, was in a deep coma, or was anticipated to need surgical evacuation within 24 hours. Overall, the groups randomized to rFVIIa versus usual care were similar on key baseline characteristics such as age, location of hemorrhage, Glasgow Coma Scale (GCS) score (a scale of neurological function), blood pressure, and time to treatment with rFVIIa or placebo. A possible exception to the groups being well-matched occurred in the largest and most recent RCT, which identified a higher rate of baseline intraventricular hemorrhage in the group that received 80 mcg/kg of rFVIIa compared to the usual care group.88 The mean age of patients in the aggregate treatment and usual care groups in every trial was between 61 and 68 years, with standard deviations in the range of 12–15. In general, the RCTs used appropriate methods of blinding, in particular with the blinding of the two radiologists who independently measured the hematoma size and growth on all of the head computed tomography (CT) scans. In addition, the RCTs had appropriately long follow-up periods of 90 days, had very low rates of loss to follow-up, applied an intention-to-treat approach, and used appropriate statistical analyses.

Table 18. General characteristics of comparative studies on off-label rFVIIa use for intracranial hemorrhage.

Table 18

General characteristics of comparative studies on off-label rFVIIa use for intracranial hemorrhage.

Comparative observational study. Unlike the above RCTs, the one comparative observational study by Ilyas89 included patients on oral anticoagulation therapy with warfarin (INR>1.4) and with subdural hematomas or ICH (50 percent apiece). It retrospectively compared 24 patients treated with rFVIIa to 30 usual care patients.

Intervention Characteristics

RCTs. The rFVIIa dose was administered similarly in all of the studies: packaged in identical vials to placebo and administered as a single dose within one hour of baseline head CT scan and four hours of symptoms onset. The doses varied widely, however. The smaller dose -finding studies had the widest variation, with doses of 10, 20, 40, 80, 120, and 160 μg/kg,86 and 5, 20, 40, and 80 μg/kg,87 respectively, with equal numbers of patients per dose tier in a given study. The larger studies had less variation in dose range, with 20 and 80 μg/kg,23 and 40, 80, and 160 μg/kg,88 respectively, and also had similar numbers of patients within each dose tier.

Comparative observational study. There was a wide dose range of 10–100 μg/kg among treated patients in the Ilyas study.89

Outcomes

As explained in our methods, for the intracranial hemorrhage indication there were special statistical considerations, such that we made several a priori decisions regarding statistical analyses. Because there were indications in the literature regarding a possible dose response relationship between rFVIIa and certain outcomes (e.g., thromboembolic events and hematoma volume) and multiple doses of rFVI were analyzed in each RCT, we chose to analyze the data according to low, medium, and high dose rFVIIa groups, defined as less than or equal to 40 μg/kg, greater than 40 but less 120 μg/kg, and at least 120 μg/kg, respectively. However, in all of the RCTs, the different levels of treatment dosage were compared to a common control. In addition, some studies did not contain all levels of the treatment dosage. Because of these complexities, we applied meta-analytic methods developed by Olkin et al82 to analyze this kind of data when generating the summary effect sizes. Second, because there were suggestions in the literature of a possible association between rFVIIa and arterial thromboembolic events, but not venous events, and both types of data were available to us from the ICH RCTs, we chose to analyze arterial and venous thromboembolic events separately for this indication. Third, while the summary effect sizes for all analyses (included in both the text and figures) are indeed accurate, their graphical representation using forest plots is complicated by their use of a common control for the different treatments dosages, so should be considered an aide to interpretation rather than a strict representation of the underlying metrics employed. Fourth, among the studies included in the risk difference meta-analyses, assessments of the significance and magnitude of heterogeneity by the Q and I2 statistics did not identify significant heterogeneity for any outcome. Finally, meta-analytic results using the arcsine metric were consistent in all cases with those described below for the risk difference (Appendix C, Appendix Figures 19).

Direct (patient-centered) outcomes. There were no dose-dependent trends in the outcomes reported. Because the RCTs had similar patient populations and measured comparable outcomes, we were able to perform meta-analyses of their data for all major outcomes. The direct (patient-centered) outcomes are presented in Tables 19 and 20 and in Figures 518. We also plotted mean differences in mortality and thromboembolic event rates for each comparative study and according to each rFVIIa indication using circle charts, with the area of each circle approximating the total sample size of its respective study (Figures 5 and 6).

Table 19. Mortality and poor outcome on modified Rankin Scale score in comparative studies of rFVIIa use in intracranial hemorrhage.

Table 19

Mortality and poor outcome on modified Rankin Scale score in comparative studies of rFVIIa use in intracranial hemorrhage.

Table 20. Thromboembolic events (arterial and venous) in comparative studies of rFVIIa use in intracranial hemorrhage.

Table 20

Thromboembolic events (arterial and venous) in comparative studies of rFVIIa use in intracranial hemorrhage.

Figure 5 illustrates the meta-analytic evaluation of differences in
mortality rates in the context of the comparable findings for several
clinical indications and by comparative study. In the figure, several
circles are plotted where each circle represents a study; larger circles
correspond to larger studies; shaded circles represent studies on treatment
use of rFVIIa, and white circles represent studies on prophylactic use of
rFVIIa. Trauma and Intracranial hemorrhage have five shaded circles.
Traumatic brain injury has two shaded circles. Liver Transplantation has
four circles that are not shaded. Adult Cardiac Surgery has five circles -
four are shaded; one is not. The risk difference summary statistics are
reported below Figure D. In sum, there was no effect of rFVIIa on
mortality.

Figure 5

Mean differences in mortality rates, by study and rFVIIa indication (rFVIIa minus usual care).

Figure 6 illustrated the meta-analytic evaluation of thromboembolic
event risk differences rates in the context of the comparable findings for
several clinical indications and by comparative study. In the figure,
several circles are plotted where each circle represents a study; larger
circles correspond to larger studies; shaded circles represent studies on
treatment use of rFVIIa, and white circles represent studies on prophylactic
use of rFVIIa. The risk difference summary statistics are reported below
Figure D. In sum, there was an increased rate of arterial thromboembolic
events with rFVIIa use vs. usual care for the medium- and high-dose
groups.

Figure 6

Mean differences in rates of thromboembolic events, by study and rFVIIa indication (rFVIIa minus usual care). ICH=intracranial hemorrhage here—although in rest of report “ICH” indicates a subset of intracranial hemorrhage, namely (more...)

Figure 7 depicts a random effects model of mortality in ICH with low
rFVIIa dose. The summary statistics are as follows: Risk difference summary
effect size -0.031 (95%CI: -0.086 to 0.024); P value for Q statistic: 0.248;
I-Squared=22.73

Figure 7

Mortality in ICH (low rFVIIa dose).

Figure 8 depicts a random effects model of mortality in ICH with
medium rFVIIa dose. The summary statistics are as follows: Risk difference
summary effect size -0.02 (95%CI: -0.076 to 0.036); P value for Q statistic:
0.248; I-Squared=22.73

Figure 8

Mortality in ICH (medium rFVIIa dose).

Figure 9 depicts a random effects model of mortality in ICH with high
rFVIIa dose. The summary statistics are as follows: Risk difference summary
effect size -0.027 (95%CI: -0.121 to 0.068) ; P value for Q statistic:
0.248; I-Squared=22.73

Figure 9

Mortality in ICH (high rFVIIa dose).

Figure 10 depicts a random effects model of poor modified Rankin score
in ICH with low rFVIIa dose. The summary statistics are as follows: Risk
difference summary effect size -0.024 (95%CI: -0.092 to 0.045); P value for
Q statistic: 0.088; I-Squared=43.53

Figure 10

Poor modified Rankin score in ICH (low rFVIIa dose).

Figure 11 depicts a random effects model of poor modified Rankin score
in ICH with medium rFVIIa dose. The summary statistics are as follows: Risk
difference summary effect size -0.029 (95%CI: -0.099 to 0.041); P value for
Q statistic: 0.088; I-Squared=43.53

Figure 11

Poor modified Rankin score in ICH (medium rFVIIa dose).

Figure 12 depicts a random effects model of poor modified Rankin score
in ICH with high rFVIIa dose. The summary statistics are as follows: Risk
difference summary effect size -0.04 (95%CI: -0.154 to 0.075) P-value for Q
statistic: 0.088; I-Squared=43.53

Figure 12

Poor modified Rankin score in ICH (high rFVIIa dose).

Figure 13 depicts a random effects model of arterial TE events in ICH
with low rFVIIa dose. The summary statistics are as follows: Risk difference
summary effect size 0.025 (95%CI: -0.004 to 0.053) P-value for Q statistic:
0.277; I-Squared=19.29

Figure 13

Arterial TE events in ICH (low rFVIIa dose).

Figure 14 depicts a random effects model of arterial TE events with
medium rFVIIa dose. The summary statistics are as follows: Risk difference
summary effect size 0.035 (95%CI: 0.008 to 0.062) P-value for Q statistic:
0.277; I-Squared=19.29

Figure 14

Arterial TE events in ICH (medium rFVIIa dose).

Figure 15 depicts a random effects model of arterial TE events with
high rFVIIa dose. The summary statistics are as follows: Risk difference
summary effect size 0.063 (95%CI: 0.011 to 0.114) P-value for Q statistic:
0.277; I-Squared=19.29

Figure 15

Arterial TE events in ICH (high rFVIIa dose).

Figure 16 depicts a random effects model of relative change in
hematoma volume for ICH with low rFVIIa dose. The summary statistics are as
follows: Standardized mean difference -0.157 (95%CI: -0.302 to -0.012)
P-value for Q statistic: 0.653; I-Squared=0

Figure 16

Relative change in hematoma volume for ICH (low rFVIIa dose).

Figure 17 depicts a random effects model of relative change in
hematoma volume for ICH with medium rFVIIa dose. The summary statistics are
as follows: Standardized mean difference -0.293 (95%CI: -0.439 to -0.148)
P-value for Q statistic: 0.653; I-Squared=0

Figure 17

Relative change in hematoma volume for ICH (medium rFVIIa dose).

Figure 18 depicts a random effects model of relative change in
hematoma volume for ICH with high rFVIIa dose. The summary statistics are as
follows: Standardized mean difference -0.304 (95%CI: -0.549 to -0.06)
P-value for Q statistic: 0.653; I-Squared=0

Figure 18

Relative change in hematoma volume for ICH (high rFVIIa dose).

Mortality. Meta-analytic evaluation of mortality rate indicated no difference between the aggregate rFVIIa patients and the usual care patients (Figures 79)( risk difference: low dose −0.031 (95 percent CI −0.086 to 0.024), medium dose −0.020 (95 percent CI −0.076 to 0.036), high dose −0.027 (95 percent CI −0.121 to 0.068); P value of the Q statistic for all risk differences 0.248). The mean mortality rate differences between rFVIIa and usual care groups by study are shown and placed in the context of other indications in Figure 5. The Ilyas observational study of patients on warfarin also found no difference in mortality between groups (Table 19).

Poor modified Rankin Scale. Poor modified Rankin Scale (mRS) score is the most widely accepted, validated measure of functional outcome in ICH. Note that other functional outcome measurements were assessed in most of the ICH studies and had similar outcomes to those reported for the mRS. Meta-analytic evaluation of poor mRS score indicate d no difference in outcomes between the aggregate rFVIIa patients and the usual care patients (Figures 1012)( risk difference: low dose −0.024 (95 percent CI −0.093 to 0.045), medium dose −0.029 (95 percent CI −0.099 to 0.041), high dose −0.040 (95 percent CI −0.154 to 0.075); P value of the Q statistic for all risk differences 0.088).

Thromboembolic events. Meta-analytic evaluation of arterial thromboembolic events identified significantly higher rates with rFVIIa use compared to usual care for the medium and high dose groups and a similar, but non-significant, finding for the low dose group (Figures 1315) (risk difference: low dose 0.025 (95 percent CI −0.004 to 0.053), medium dose 0.035 (95 percent CI 0.008 to 0.062), high dose 0.063 (95 percent CI 0.011 to 0.063); P value of the Q statistic for all risk differences 0.277). These results suggest that there is an increase in arterial thromboembolic events with rFVIIa use versus usual care. There were no differences between groups in venous thromboembolic events (risk difference: low dose 0.010 (95 percent CI −0.018 to 0.038), medium dose −0.004 (95 percent CI −0.030 to 0.022), high dose −0.012 (95 percent CI −0.049 to 0.026); P value of the Q statistic for all risk differences 0.935). The Ilyas observational study of patients on warfarin noted only one thromboembolic event in any group, a myocardial infarction in a rFVIIa patient (Table 20). Figure 6 displays for each study, and also in the context of studies of the other indications, the mean rate differences for all thromboembolic events (venous and arterial) between the rFVIIa and usual care groups.

Indirect (surrogate) outcomes. Relative hematoma expansion. Meta-analytic evaluation of relative hematoma expansion demonstrated significant reductions in the rFVIIa group compared to the usual care group at all dosing levels (standardized mean difference: low dose −0.146 (95 percent CI −0.291 to −0.001), medium dose −0.240 (95 percent CI −0.385 to −0.095), high dose −0.334 (95 percent CI −0.579 to −0.090); P value of the Q statistic for all standardized mean differences 0.840) (Figures 1618). While the large Mayer 2005a21 study reported a significant dose-response effect of reduced hematoma growth with higher doses of rFVIIa, statistical tests for differences between dosing levels in our meta-analyses found no significant dose effect. The Ilyas study did not report on this outcome(Table 21).

Table 21. Indirect outcomes in comparative studies of rFVIIa use in intracranial hemorrhage.

Table 21

Indirect outcomes in comparative studies of rFVIIa use in intracranial hemorrhage.

Consideration of poor quality comparative observational studies. In the poor quality comparative observational studies by Pickard,90 Brody,91 and Hallevi,92 the findings on mortality, poor functional outcome, and thromboembolic events are within the range of those described above (Tables 19 and 20).Other outcomes were not reported.

Other Considerations

Timing of rFVIIa and Changes in ICH Volume

Background. Until recently, increases in ICH volume were thought to be completed within minutes of onset, but recent studies have shown continued growth of the hemorrhage up to several hours after symptoms onset.167 Additionally, this late growth of the ICH has been associated with neurological deterioration and poor clinical outcome.33,167 A pooled meta-analysis of individual patient data from the earlier RCTs23,86,87 and a study by Brott et al. 167 showed that for each 10 percent increase in ICH between the baseline CT scan (performed within 3 hours of symptoms onset) and follow up CT performed 24 hours after the baseline, the hazard rate of dying increased by five percent. Similarly, the hazard ratio of increasing the modified Rankin Scale by one point (toward worse functional outcomes), was 16 percent for each 10 percent increase in ICH growth.33 In this section we explore the association between timing of rFVIIa and hemorrhage growth.

Results. Table 22 shows the published data from the largest of the earlier 2 ICH RCTs, Mayer 2005a,23 but compares those patients treated within three hours of symptoms onset to those treated after three hours from symptoms onset, but still within the four-hour time window established by the study as an inclusion criterion. The data suggest diminished hematoma expansion in the group that received earlier treatment. The subsequent large RCT, Mayer 2008,88 reported hematoma expansion occurring in “under 2 hours,” “under 3 hours,” and “at any time (i.e., within the four hour protocol of the study),” such that the data are not directly comparable to those of the earlier study. However, the findings suggest a similar pattern to the earlier RCT, one of diminished hematoma expansion when rFVIIa is administered earlier.

Table 22. Post-hoc evaluations of rFVIIa use in ICH before versus after 3 hours from time of symptoms onset.

Table 22

Post-hoc evaluations of rFVIIa use in ICH before versus after 3 hours from time of symptoms onset.

Discussion. Relative to later treatment, earlier treatment with rFVIIa may reduce hematoma growth. One drawback to our analysis is that the patients received rFVIIa within the relatively short time window of four hours. Using timing of treatment as a predictor of hematoma growth also presents analytical problems. Although it is now known that hematoma growth can continue for several hours after symptoms onset, much of ICH growth occurs prior to the baseline CT scan. Therefore, “growth” of the ICH as documented by sequential CT scans may be highly confounded by the timing of the CT scans relative to symptom onset and whether the baseline scans occur in the earliest, highest growth phase of hematoma formation or in the later, more modest growth phase. Existing studies have not explored this potential source of confounding in great detail.

Post-hoc analyses of age and other factors. Post-hoc analyses in the Mayer 2008 study posited improved functional outcomes with rFVIIa therapy when patients had a combination of younger age, lower baseline hematoma volume, and earlier rFVIIa administration. Additional post hoc analyses identified increased age and previous use of an antiplatelet agent as possible independent risk factors for thromboembolic events.88

Comparison with Premier Database

Study patients were approximately the same age (mid 60s)and had lower mortality rates than the mortality rate of 0.34 among patients in the Premier database. Application of rFVIIa to intracranial hemorrhagein the Premier database was very low prior to 2005 but experienced a notable increase in use that year, the same year in which the first Mayer RCT23 was published (Figures 3 and 4).

Strength of Evidence

We judged the strength of evidence grade to be moderate for all outcomes based on having four RCTs with a low risk of bias and nearly 900 patients on rFVIIa therapy. However, it is important to note that effect size estimates for all outcomes—other than arterial thromboembolic events and changes in hematoma volume—were imprecise, which precludes definitive conclusions about their effects. Another frequent component to limit the strength of evidence was inconsistency, which was due to both the variability of effect sizes and differences in direction of effect. We found no clear evidence (across any dosing level) that rFVIIa has an effect on mortality, poor modified Rankin Scale score, or venous thromboembolic events, but did find that it is associated with an increased rate of arterial thromboembolic events and a decrease in relative hematoma expansion (Table 23).

Table 23. Strength of evidence for rFVIIa use in intracranial hemorrhage.

Table 23

Strength of evidence for rFVIIa use in intracranial hemorrhage.

Applicability

The overall applicability of the evidence for this indication was good for treatment use in the population targeted by the RCTs—adult patients with intracerebral hemorrhage who were not on anticoagulation (Table 24). For instance, the evidence had good duration and intensity of follow-up and also encompassed a range of outcomes that included important measures of functional ability, morbidity, and mortality. The evidence is less applicable to patients on anticoagulation therapy (e.g., warfarin), who had isolated subdural or subarachnoid bleeding, or for whom surgical interventions were planned, because such patients were excluded from the RCTs. The only study to include patients on anticoagulation therapy (with subdural hemorrhage or ICH) was a small comparative observational study of fair quality, which limits applicability.

Table 24. Applicability assessment of intracranial hemorrhage studies.

Table 24

Applicability assessment of intracranial hemorrhage studies.

Conclusions

For patients of mean age 65 without anticoagulation use who present for spontaneous ICH (a subset of intracerebral hemorrhage), current evidence of moderate strength suggests that neither benefits nor harms substantially exceed each other. Use of rFVIIa compared to usual care appears to attenuate hematoma growth but also increase the risk of arterial thromboembolic events without having a significant impact on mortality or functional outcome. Notably, these patients have lower rates of mortality than do patients in the Premier database. Whether patients who are on oral anticoagulation therapy, have other forms of intracranial hemorrhage (e.g., subdural or subarachnoid hemorrhage), are treated earlier with rFVIIa, are younger, have lower baseline hematoma volumes, or have no prior use of antiplatelet agents may experience better outcomes (relative to the populations already studied) remains unclear.

Key Question 3.a. Acquired, coagulopathic massive bleeding from body trauma and comparative effectiveness of rFVIIa

Background

Trauma is the leading cause of death in young men between the ages of 15 and 40. Hemorrhage is the leading cause of early death (within 24–48 hours) in trauma and second only to traumatic brain injury (TBI) as the overall cause of mortality.168 Hemorrhage after traumatic injury is associated with an acquired coagulopthy known as the “acute coagulopathy of trauma.”169–171 The coagulopathy develops when there is tissue injury in combination with hypotension. The severity of coagulopathy increases with increasing injury severity and is associated with worse outcomes.172 Resuscitation of these patients can worsen the coagulopathy. The dilution of blood due to rapid infusion of crystalloid, the development of hypothermia, and persistent acidosis occur during resuscitation and are known together as the “lethal triad,” which conspires to impede coagulation. Not surprisingly, the conditions which lead to the development of lethal triad are worse in cases of severe hemorrhage, particularly in those cases that require massive transfusions (most frequently defined as the use of 10 or more units of packed red blood cells (RBCs) within 24 hours of injury). This acquired coagulopathy potentiates further bleeding, which in turn leads to further physiologic derangements, increased morbidity, and increased mortality. Unfortunately, blood products that are used to replace lost blood and to treat coagulopathy can carry risks of their own. In particular, studies have highlighted increased risks of acute respiratory distress syndrome (ARDS), multiorgan failure (MOF), and sepsis with higher levels of blood product transfusions.173–175 rFVIIa has been investigated in trauma as an adjunct to control bleeding and thereby reduce the above risks.

Usual care during the time frame of included studies. As noted above, there is growing evidence of an association between coagulopathy and poor outcomes after trauma. A recent trend in the “usual care” of traumatic bleeding has been to address coagulopathy early in cases of a massive transfusion. This has been achieved by using fresh frozen plasma (FFP) earlier during the resuscitation and in greater amounts. Recent data from civilian databases and the Iraq war suggest that a 1:1 ratio of RBCs:FFP may have a survival advantage when compared to ratios closer to 4:1. In addition, some experts advocate for early transfusion of platelets as well, with a RBCs:FFP:platelet ratio of 1:1:1.176 As transfusing large quantities o f blood products and maintaining a high ratio of FFP to RBCs is challenging, many institutions have implemented “massive transfusion protocols.”177 These protocols faciliate coordination between the physicians and the blood bank, have been associated with improved survival, and may include include rFVIIa as part of the protocol. The studies evaluated in this section contain data on patients treated during the period of flux. Both plasma-rich transfusion practices and massive transfusion protocols were being implemented contemporaneously and after the published studies included in this section. This means that comparison to “usual care” is limited and does not necessarily reflect current practice. Two published RCTs (published in one paper) were conducted from March 2002 to September 2003,96 while the three comparative observational studies 97–99 evaluated data on patients treated from January 2000 to February 2005, December 2003 to October 2005, and April 2006 to August 2008, respectively. Another RCT, as of yet unpublished, was conducted from October 2005 to June 2008.

General Characteristics of Studies of Traumatic Bleeding

For this indication, among the published literature, we identified two RCTs (both fair quality) and three comparative observational studies (all fair quality) with 267 patients who received rFVIIa for treatment use. The RCTs were two trials conducted in parallel by the same investigators and published in the same article. One of the trials enrolled patients with blunt trauma and the other with penetrating trauma.96 The two trials are evaluated separately in this review. The trials were sponsored by Novo Nordisk, and the trial statistician was an employee of the company. Another RCT was completed in June 2008 and presented at an international meeting in March 2009, but it has not yet been published.178,179 It also evaluates rFVIIa in both blunt and penetrating trauma and is sponsored by the manufacturer. Three restrospective cohort studies studies were also analyzed. They were not sponsored by the manufacturer, and the authors or statisticians were not employed by Novo Nordisk.

Place of studies within analytic framework. All of the included studies evaluated rFVIIa for treatment use (versus prophylaxis or end-stage use), as outlined in our Analytic Framework (Figure 1). However, the censoring of deaths within the first 48 hours in the randomized controlled trials96 and within 24 hours in one of the observational studies 98 raises the question of whether some end-stage use might have occurred. The RCTs reported direct outcomes (e.g., mortality, thromboembolic events) but were underpowered to detect a difference in these. Therefore, they have as their primary outcome the surrogate endpoint of RBC transfusion. The comparative observational studies all examined direct outcomes as their primary outcomes. Mortality was the primary outcome in the Rizoli and Spinella studies,97,98 and limb revascularization is the primary outcome in the Fox cohort.99

Qualitative considerations of heterogeneity. The RCTs and observational trials included both blunt and penetrating mechanisms and were from both civilian and military populations. Different anatomic mechanisms of injury often result in the final common pathways of severe tissue injury and hypotension, and these conditions are thought to drive the coagulopathy of trauma. Therefore, despite the differences in injury mechanisms, injuries of sufficient severity do share physiologic characteristics. The role of rFVIIa is to act upon these physiologic disturbances. For this reason, despite the heterogeneic mechanisms, we felt it appropriate to assess the patient populations together in this analysis.

Comparison to studies on other indications. These studies had younger patients, on average, than studies in any other adult indication. The RCTs evaluated the highest dose of rFVIIa used in any indication, although the observational studies applied much lower doses.

Patient Characteristics and Study Design

RCTs. The two double-blind RCTs by Boffard (both published in a single paper and both fair quality (Table 14)) enrolled patients from 32 hospitals in eight countries that did not include the U.S.(namely, Australia, Canada, France, Germany, Israel, Singapore, South Africa, and the United Kingdom) and were of modest sample sizes (Table 25). Approximately 70 patients were enrolled in the rFVIIa group in each trial. Inclusion criteria included hemorrhage from a blunt or penetrating traumatic injury requiring at least six units of RBCs within four hours of hospitalization. Exclusion criteria included low GCS score (less than eight), severe base deficit (over 15 mEq/L) or acidosis (pH under 7.0). There was good baseline similarity between treatment groups. A limitation of the trials was that there were instances of missing outcomes data. For example, the primary outcome of the studies was the number of RBC units transfused. The numbers of patients for whom data was reported was lower than the number of patients who were said to have completed the trials. Another important con textual factor was the investigators’ choice to exclude from further analysis those patients who died within the first 48 hours of hospitalization, which comprised nearly 20 percent of patients in each treatment group.

Table 25. General characteristics of comparative studies of off-label rFVIIa use for massive bleeding due to body trauma.

Table 25

General characteristics of comparative studies of off-label rFVIIa use for massive bleeding due to body trauma.

The more recent but unpublished RCT was conducted at 100 sites in 26 countries, including 23 sites in the United States.178 It was terminated prematurely after a pre -planned interim futility analysis performed by the Data and Safety Monitoring Committee indicated that “the mortality observed in all enrolled patients was lower than expected during study design, meaning that the study as a whole would be underpowered to assess the primary endpoint [of mortality].” Therefore, “the sponsor chose to close the study.”180 Prior to terminations, the trial enrolled 468 patients with blunt trauma (221 rFVIIa, 247 placebo) and 86 with penetrating trauma (46 rFVIIa, 40 placebos). Limited data were obtained on the Clinical Trials. gov and manufacturer’s websites. These data were used for qualitative sensitivity analyses but not included in the primary analyses.

Withdrawal of deferred informed consent. The FDA rules for the study of emergency therapies (and similar policies in other countries) allow for deferral of informed consent.64,65 The publication of the Boffard RCTs states in the methods section, “Because of the emergency conditions and the possible absence of relatives at enrollment into the trial, waived informed consent was authorized by the ethics committees. However, whenever a patient was included without written informed consent, such consent was promptly sought from the legally authorized representative and subsequently from the patient. Adequate confirmation of consent was not obtained for six patients, and their data were excluded from analysis.”96 The publication reported that the withdrawals of deferred consent —described as “consent not confirmed” following “waived informed consent”—occurred after randomization and dosing (whether of rFVIIa or placebo), but did not note the treatment arm of the patients who withdrew. Five patients in the blunt trauma trial were withdrawn for this reason versus one patient in the penetrating trauma trial.

Comparative observational studies. The Spinella retrospective cohort study98 assessed data from a military database, the Joint Theatre Trauma Registry. Patients in this registry sustained wartime injuries in Iraq. Patients were included in the registry if they had suffered severe trauma (defined by Injury Severity Score over 15) and required transfusion of 10 or more units of RBCs within the first 24 hours of hospitalization. The time period of data collection was 2003 to 2005. Similar to the RCTs, this study excluded from further analyses those patients who had early in hospital death, although in this case, early death was defined as occurring within the first 24 hours. Among this cohort, investigators identified 49 patients who received rFVIIa and 75 who did not. The groups were relatively well matched at baseline. The authors evaluated for confounding by testing for interactions between rFVIIa use and other variables when evaluating the primary outcome of mortality at 30 days.

The second retrospective cohort study97 by Rizoli identified all trauma patients who were managed at a level I trauma center in Canada and were transfused with eight units of RBCs within the first 12 hours of hospitalization. Among this cohort, investigators identified 38 patients who received rFVIIa and 204 who did not. Groups were relatively well matched at baseline, except that the rFVIIa group received higher numbers of transfusions in the first six hours of admission. The mean time from admission to rFVIIa administration was six hours. The primary outcome was mortality at 24 hours and the secondary outcome was in-hospital mortality. Investigators used univariate analysis followed by multivariable logistic regression analysis to address issues of confounding. Importantly, investigators noted that five of the 38 patients in their cohort who received rFVIIa were also included in one of the above trauma RCTs, although the particular trial (blunt versus penetrating) was not specified.

The final retrospective cohort study by Fox,99 like the Spinella study, examined data from the Joint Theatre Trauma Registry but for patients treated during a later timeframe—from 2006 to 2007. Patients were included in the study if they suffered life-threatening vascular injuries and required transfusion of over four units of RBCs. Patients were not censored for early in-hospital death. Investigators identified 41 patients who received rFVIIa and 12 patient who did not. Groups were relatively well matched at baseline, with the possible exception of somewhat higher mean Injury Severity scores in the treatment group versus controls (29 versus 22, respectively), although this difference was not statistically significant. The authors did not evaluate for confounding.

Intervention Characteristics

RCTs. The Boffard RCTs both used the same dosing schema consisting of three sequential infusions of rFVIIa (200,100, and 100 μ/kg) or placebo. The first dose was administered after the eighth unit of RBCs, the second one hour later, and the third three hours later. rFVIIa patients in the unpublished RCT178 received the same regimen as above, except that the first dose was given earlier during the resuscitation—after the fourth unit of RBCs. As noted above, the unpublished trial was used only for sensitivity analysis.

Comparative observational studies. In the observational studies, the exact dose of rFVIIa administered was not clear. In the Spinella study, it was noted that military guidelines recommended a dose of 120 μg/kg, but the actual drug dose administered was not reported.98 The Rizoli study97 did not report the doses of drug administered except to say that initial practice at their center was to start with a dose of 17.1 μg/kg, which was liberalized over time to higher doses, and that repeat doses were common. The Fox study99 described the doses of rFVIIa as “typically 90–120 μg/kg.” The range of rFVIIa doses administered across studies thus appears to be potentially quite broad (from 17.1 μg/kg to 120 μg/kg). In all cases, drug administration was triggered by the patient reaching a transfusion threshold that was variably defined but considered to be massive or life-threatening bleeding.

Outcomes

Direct (patient-centered) outcomes. Results are summarized in Table 26. Dose-reponse relationships were not apparent for these outcomes.

Table 26. Mortality, thromboembolic events, and ARDS in comparative studies of rFVIIa use in body trauma.

Table 26

Mortality, thromboembolic events, and ARDS in comparative studies of rFVIIa use in body trauma.

Mortality. All of the studies reported on mortality, but none were powered to detect differences in mortality. There were differences between studies in the exclusion from analysis of those patients who experienced “early” in -hospital death (defined to be death within 48 hours in the RCTs and 24 hours in the Spinella study), so comparisons across studies must be interpreted with caution. Among patients who survived at least 48 hours, the RCTs noted no difference in 30-day mortality, although mortality was lower in both rFVIIa groups compared to controls. In the unpublished RCT, there was also no difference in mortality rate between groups, whether measured for blunt or penetrating trauma or at 30 or 90 days.178 The Spinella study noted a significant decrease in 30-day mortality for patients who received rFVIIa and survived at least 24 hours. The Rizoli study identified a significant improvement in 24-hour mortality with rFVIIa. In-house mortality was lower after rFVIIa but was not significant. The Fox study noted no difference in mortality between groups at 24 hours. To place the mortality differences in the context of the comparable findings for the other clinical indications, also see Figure 5 (in the Key Question section on intracranial hemorrhage). (Note that the Rizoli study is not included in this figure because its findings were reported as an odds ratio rather than event rate.)

Thromboembolic events. Of the published studies, the Boffard RCTs and two of the retrospective trials (Spinella and Fox) evaluated thromboembolic complications. While the retrospective study by Rizoli did not report on thromboembolic complications, an extension of the study by Nascimento181 did report this. None of the studies found any difference between the rFVIIa and usual care groups for such events. However, the absolute number of events was low, so the studies were likely underpowered to detect any difference between groups. The unpublished RCT also found no differences between groups in thromboembolic event rates.178 To place the differences in thromboembolic event rates in the context of the comparable findings for the other clinical indications, see Figure 6 (in the Key Question section on intracranial hemorrhage).

Acute respiratory distress syndrome (ARDS). Only the Boffard RCTs and Spinella study evaluated the rate of ARDS at 30 days. The blunt trauma RCT identified a significantly lower rate of ARDS in the rFVIIa group compared to the usual care group, while the penetrating trauma RCT and Spinella study together suggested a trend in the same direction. Event rates for ARDS were low, which again raises concern that the studies did not have adequate power to detect a difference for this outcome. Of note, we did not evaluate the outcome of multi-organ failure (MOF), because it was not clear how MOF was defined in the study or if these cases overlapped with cases diagnosed with ARDS.

Indirect (surrogate) outcomes. Results are summarized in Table 27. Dose -response relationships are not apparent for these outcomes.

Table 27. RBC transfusion in comparative studies of rFVIIa use in body trauma.

Table 27

RBC transfusion in comparative studies of rFVIIa use in body trauma.

Red Blood Cell (RBC)transfusion. All of the studies reported on some aspect of RBC transfusion, although comparisons across studies are difficult for this outcome because of the variable censoring of patients who experienced early in-hospital death. For patients who survived to 48 hours, the blunt trauma RCT reported a significant reduction in the 48 -hour RBC transfusion rate for patients who received rFVIIa versus controls. In the penetrating trauma RCT, there was a similar but non-significant finding of reduced RBC transfusions in the rFVIIa group. In the unpublished RCT, there were comparable findings: a significant reduction in RBC transfusions for rFVIIa patients in blunt trauma and a non-significant reduction in penetrating trauma.178 In contrast, the Spinella study identified a significant increase in transfusion requirements with rFVIIa. The Rizoli and Fox studies did not independently assess this outcome.

Consideration of poor quality comparative observational studies. Two studies which were considered to be of “poor quality” were compared with the other studies. In the studies by Dutton100 and Harrison,101 the findings on mortality, thromboembolic events, and RBC transfusions were consistent with those described above (Tables 26 and 27). Other outcomes were not reported.

Other Considerations

Differences in baseline mortality rate. Both the Rizoli97 and Spinella98 studies suggest that the higher baseline mortality rates in the patients they studied account for the greater and significant decreases in mortality they observed compared to the Boffard RCTs.96 They also cite higher rates of markers of injury severity. For instance, both cohort studies cite higher transfusion requirements within the first 24 hours, and the Spinella study describes lower admission SBP and pH among their patients versus the Boffard RCT patients.

Possible interactions with acidosis, thrombocytopenia, and timing and era of administation. In a post-hoc subgroup analysis of the patients who received rFVIIa, the Rizoli study identified an association between patients with higher baseline pH and platelet counts and increased survival. The Spinella study found a similar post-hoc association between shorter time to intervention and increased survival. Of note, in the unpublished RCT178 patients were administered rFVIIa earlier in the resuscitation than in the published RCTs (after 4 units of RBCs versus 8 units). In addition, the trial was conducted during the time period when “usual care” more often included the use of matched transfusion ratios of RBCs to FFP (1:1) and massive transfusion protocols. This may explain why the baseline mortality rate among controls appears to be lower than it was for the Boffard RCTs.

Comparison to Premier Database

Study patients, mean ages 28–41, were younger by 20 years, on average, than patients in the Premier database (mean age 53 years). However, mortality rates were comparable to the Premier mortality rate of 0.33. Use of rFVIIa in the Premier patients started slowly in the early 2000s, but increased sharply in 2005, the same year the Boffard RCTs were published (Figures 3 and 4).

Strength of Evidence

The overall strength of evidence was low for all of the outcomes evaluated, including the outcome for which rFVIIa was weakly but non-significantly favored, namely ARDS. These determinations were based primarily on weaknesses in two strength of evidence domains: the “risk of bias” domain, which had ratings of a “medium” or “high” overall risk of bias for all outcomes, and the “precision” domain, which had ratings of “imprecise” for all outcomes. The rating of “imprecise” was based on wide confidence intervals for the major outcomes which in turn were due to low event rates in studies that were underpowered to detect mortality and major morbidity outcomes (Table 28).

Table 28. Strength of evidence for rFVIIa use in body trauma.

Table 28

Strength of evidence for rFVIIa use in body trauma.

Applicability

The overall applicability of the evidence was fair for treatment use in the population targeted—adult patients with blunt or penetrating trauma who were not censored for early in hospital death (defined as 24 hours or 48 hours, depending on the study). Specifically, the types of trauma and practice settings represented by the evidence likely have good applicability to major trauma centers in the U.S., as well as combat settings experienced by U.S. troops. On the other hand, the “usual care” of such patients has likely changed at least moderately since the studies were conducted (e.g., with the introduction of massive transfusion protocols and 1:1 transfusion ratios for RBCs:FFP). The applicability of the mortality outcomes is also difficult to interpret given the censoring of patients who experienced early in-hospital mortality. Finally, the follow-up of 30 days may be insufficient to judge applicability to general practice (Table 29).

Table 29. Applicability assessments of studies of body trauma.

Table 29

Applicability assessments of studies of body trauma.

Conclusions

The available evidence has low strength and does not permit definitive conclusions regarding the impact of rFVIIa use compared to usual care. The two RCTs, in blunt and penetrating trauma and with mean ages of 29 to 33 years old, reported no improvement in mortality with rFVIIa. However, the conclusions which can be drawn from this are limited by their lack of power for evaluating mortality and the censoring of deaths which occcured within the first 48 hours. One observational study found a significantly reduced death rate with treatment, while the other had non-significant findings in the same direction. For acute respiratory distress syndrome, the blunt trauma RCT demonstrated a significant reduction for rFVIIa patients, while across the remaining two studies that evaluated this outcome(the penetrating trauma RCT and one observational study)there was a trend in the same direction. There was no evidence of an increase in thromboembolic harm with treatment, but again the strength of evidence for this outcome was low. Thus, current evidence of low strength suggests the potential for benefit and no evidence of harm. Importantly, the patient s were younger in the studies than in the Premier database, but nonetheless had similar mortality rates. The importance and nature of interactions between outcomes and the era of rFVIIa administration remain unclear, as does the relationship between outcomes and the timing of rFVIIa administration or baseline levels of acidosis or thrombocytopenia.

Key Question 3.b. Bleeding from brain trauma and comparative effectiveness of rFVIIa

Background

Traumatic brain injury (TBI) is the leading cause of overall death and disability after trauma.183 In severe cases, mortality rates are as high as 30–50 percent.184 Bleeding within the brain parenchyma is a significant contributor to these poor outcomes. The mechanisms are complex and poorly understood but differ in important ways from spontaneous ICH, which typically consist of more localized areas of bleeding at sites of prior atherosclerotic or thrombotic injury. In the presence of the coagulopathy of trauma,155 it is more likely that injured parenchyma, and even non-injured parenchyma, will become ischemic. Medical management includes optimization of cerebral blood flow and oxygen delivery. Surgical evacuation for large hemorrhagic lesions secondary to TBI is the only existing treatment for large bleeds (when decompression must be performed to relieve increased intracranial pressure), but often can not be attempted expeditiously secondary to the presence of concurrent coagulopathy, whether due to the trauma itself or the need for resuscitation fluids for hemorrhage at other regions of the body. Of note, TBI is a well-established risk factor for the development of deep vein thrombosis (DVT).

Usual care during the time frame of included studies. The changes in approach to treatment of coagulopathy noted in the above section on body trauma also apply to patients with TBI. Patients with bleeding due to TBI are now recognized to have similar problems with the coagulopathy of trauma.185 There is growing understanding within the field regarding the need to address this coagulopathy early in the course of therapy. In addition, new evidence suggests that patients with bleeding from TBI experience significant hematoma expansion within the first 24 hours of injury and that this expansion contributes to subsequent morbidity and mortality.186,187 However, specific interventions to ameliorate this expansion have not been systematically studied in the TBI literature or have not been shown to help. For example, an RCT of plasma (FFP) use in TBI found that treatment was associated with increased mortality.188 For large hematomas, the only widely available current therapy is neurosurgical evacuation. But coagulopathy must be corrected prior to surgery, which is accomplished by plasma infusions that may be high in volume and poorly tolerated by older patients or those with cardiac dysfunction. 107 Thus, any therapy that can address TBI-associated coagulopathy and allow for earlier neurosurgical intervention—while also limiting or stopping hemorrhage—is eagerly sought.

General Characteristics of Studies of Traumatic Bleeding

On this topic, we identified one RCT (fair quality) and one comparative observational study (fair quality), which together had 79 patients who received rFVIIa. The RCT, Narayan 2008, both was sponsored and had one author employed by the manufacturer of rFVIIa.106 The comparative observational study selected for inclusion was the earlier of two published by Stein and colleagues with overlapping patient populations. While the latter cohort study189 included more patients, these patients were more likely to have significant concurrent trauma elsewhere at the body and were less well matched to controls at baseline. The retrospective cohort chosen for inclusion, Stein 2008, was not sponsored and had no authors or statisticians employed by the manufacturer.107 The full Stein 2008 cohort study included some patients with traumatic injuries at areas of the body other than the brain, but also performed subgroup analyses on patients with isolated TBI (Table 30). We evaluated the data on the patients with isolated TBI -associated bleeding, and when evaluated in this light, the study quality was determined to be fair, with good baseline matching of groups.

Table 30. General characteristics of comparative studies on off-label rFVIIa use for bleeding due to brain trauma.

Table 30

General characteristics of comparative studies on off-label rFVIIa use for bleeding due to brain trauma.

Place of studies within analytic framework. The two included studies of brain trauma evaluated rFVIIa for treatment use (versus prophylaxis or end-stage use, which are other potential uses, as outlined in our Analytic Framework (Figure 1)). While the Narayan RCT106 did not define its primary outcome, it reported on important direct outcomes (e.g., mortality, thromboembolic events) but was likely underpowered for these. It also reported on the indirect outcome of change in hematoma volume. The Stein comparative observational study107 did not define a primary outcome but did examine a number of different direct (e.g., mortality, thromboembolic events) and surrogate (e.g., time to neurosurgical intervention, ICU length of stay) endpoints.

Qualitative considerations of heterogeneity. As noted above, we identified one important source of potential heterogeneity among studies as the inclusion of patients with significant body trauma (as well as brain trauma) among the study population. We addressed this concern by confining our assessment to studies or sub-groups of studies that effectively excluded such patients by focusing primarily on hemorrhage from isolated TBI.

Comparison to studies on other indications. The populations in these studies were young compared to most of the non-trauma studies in adults for other indications, with mean age in the Narayan RCT of 51 in both groups and 45–50 in the Stein cohort study. The rFVIIa dose range of 40–200 μg/kg in the RCT and 8–140 μg/kg in the cohort study were in the low to middle range of those used for rescue treatment across indications.

Patient Characteristics and Study Design

RCTs. The Narayan 2008 RCT was a fair quality dose-finding and safety trial with small sample sizes: 36 patients in the usual care group and 61 patients in the aggregate rFVIIa group (divided relatively evenly between five dosing groups containing an average of 12 patients per group).106 It was conducted at 38 hospitals in 10 countries, but not the U.S. Inclusion criteria included intracranial hematoma volume of at least two mL, baseline CT scan within six hours of injury, and administration of trial drug within 2.5 hours of baseline CT scan. Important exclusion criteria included known vitamin K antagonist use and planned neurosurgical intervention within 24 hours of admission. There was acceptable baseline similarity between groups among the patients, blinding of the two radiologists who independently evaluated the CT scans for hematoma size and growth, and no loss to follow-up at day 15, the endpoint of the trial. The most problematic aspect of study design was the extremely low enrollment rate of four percent of those screened, which might have been even lower if the investigators had not amended their protocol mid-way to change the minimum lesion volume size required for enrollment from five mL to two mL. The primary outcome was not defined, but the study states that its focus was on “safety endpoints,” namely those of thromboembolic complications (including DVT) and mortality at 15 days.

Comparative observational study. The Stein 2008 retrospective cohort study evaluated patients from the trauma registry of the Shock Trauma Center at the University of Maryland.107 Patients were included if they were found to have “severe” TBI (defined at a GCS score under nine and an abbreviated Injury Severity Score of four or five) and coagulopathy (defined as INR greater than or equal to 1.4), and were “deemed appropriate for immediate neurosurgical intervention” at the time of hospital admission. Patients who received vitamin K antagonists such as warfarin were included in the study. Among this cohort’s subgroup of patients with isolated TBI, which is the focus of our evaluation, there were small sample sizes: 18 patients received rFVIIa and 17 received usual care. These groups were well-matched at baseline, with the possible exception of a higher rate of pre-injury warfarin use in the rFVIIa group. Investigators did not perform any statistical analyses to evaluate or adjust for residual confounding. One limitation of the study design relates to the determination at the time of admission of appropriateness for immediate neurosurgical intervention, which was not further defined in the paper. The primary outcome was time to neurosurgical intervention, the only widely available definitive therapy for large bleeds. The paper also reported on mortality and thromboembolic events.

Intervention Characteristics

RCT. The RCT intervention group received single doses of rFVIIa (40, 80, 120, 160, or 120 μg/kg), while controls received placebo, delivered within 2.5 hours of the baseline CT scan. No repeat doses were given.

Comparative observational study. The vast majority of patients who received rFVIIa were administered a single dose but the doses varied widely from 8 to 140 μg/kg. The dose was apparently determined by the choice to administer the entire 1.2 mg vial in which the rFVIIa is packaged, rather than by a weight-based calculation for a given patient. It is unclear at what time in the hospital course rFVIIa was delivered, but as neurosurgical intervention occurred within a mean of 185 minutes from arrival, rFVIIa was clearly given within a mean time of less than that amount.

Outcomes

Direct (patient-centered) outcomes. Summary results are reported in Table 31.

Table 31. Mortality, thromboembolic events, and absolute change in hemotoma volume in comparative studies of rFVIIa use in brain trauma.

Table 31

Mortality, thromboembolic events, and absolute change in hemotoma volume in comparative studies of rFVIIa use in brain trauma.

Mortality. The Narayan 2008 RCT reported no difference between groups in mortality at 15 days, with rates of 11 percent in each group (seven of 61 in the combined rFVIIa group, four of 36 in the usual care group). The Stein 2008 cohort study did not clearly define the length of follow-up for the patients, but this is likely equal to the hospital length of stay (LOS) given the retrospective nature of the study. The mean hospital LOS was 14.6 days and 19.1 days for the rFVIIa and usual care groups, respectively, which is roughly equivalent to the 15-day follow-up time of the RCT. Among patients with isolated TBI, the Stein cohort identified a reduced mortality with rFVIIa compared to usual care but this was a non-significant finding: 33.3 percent (6 of 18) versus 52.9 percent (9 of 17) in the two groups, respectively. The findings in both studies are limited by low event rates. To place the mortality differences in the context of the comparable findings for the other clinical indications see Figure 5 above (in the Key Question section on intracranial hemorrhage).

Thromboembolic events. Event rates are also low for thromboembolic outcomes. Across the studies there was a trend toward higher rates of thromboembolic events in the rFVIIa group compared to usual care group. In the Narayan RCT the event rates were 16.4 percent (10 of 61) in the rFVIIa group versus 5.6 percent (2 of 36) in the usual care group. Five of the 10 events in the rFVIIa group consisted of DVTs, all of which were symptomatic. In the Stein cohort study the event rates were 22.2 percent (4 of 18) in the rFVIIa group versus 17.6 percent (3 of 17) in the usual care group. The one DVT among these occurred in the usual care group. To place the differences in thromboembolic event rates in the context of the comparable findings for the other clinical indications, see Figure 6 above (in the Key Question section on intracranial hemorrhage).

Indirect (surrogate) outcomes. Summary results are reported in Table 31.

Hematoma expansion. Only the Narayan 2008 RCT reported on hematoma change at 24 hours, which it defined as the mean volume change in hematoma size from baseline. Of note, baseline hematoma volume was well-matched at 11.3 mL (SD 10.9) in the combined rFVIIa group and 10.4 mL (SD 10.8) in the usual care group. The study noted a non-significant reduction in expansion in the rFVIIa group compared to controls (7.0 mL (SD 12.9) versus 10.4 mL (25.0), respectively).

Time to neurosurgical intervention. Only the Stein cohort study evaluated this outcome, which was its primary outcome. The study found that, with treatment, there was a significant decrease in absolute minutes to neurosurgical intervention: 185.3 (SD 219.9) for the rFVIIa group versus 518.6 (SD 409.8) for the usual care group (p=0.005).

Consideration of poor quality comparative observational study. In the poor quality comparative observational study by Tawil,108 the findings on thromboembolic events are consistent with those described above (Table 31). Other outcomes were not reported.

Other Considerations

Possible interactions with coagulopathy, CHF, or blunt trauma injuries to the cerebral vasculature. The Stein cohort study enrolled only patients with baseline derangements in laboratory markers of coagulopathy, defined as INR equal to or greater than 1.4. While the Narayan RCT had no comparable inclusion criteria, it performed post hoc analyses of the subgroup of patients with platelets less than 100K, PT > 14, aPTT > 45 seconds, or INR > 1.2 (13 rFVIIa, 8 placebo). These analyses note a difference between groups, favoring rFVIIa, in the degree of hematoma expansion but are based on small sample sizes. Another subgroup that the Stein study proposed may have the potential for greater benefit from rFVIIa are those patients who have poor toleration of large volume infusions (e.g., the elderly and those with congestive heart failure (CHF)) and yet require rapid correction of coagulopathy to allow for neurosurgical intervention.107

Other authors have raised concerns regarding a subgroup of patients who might have the potential for increased harm with rFVIIa administration.108 These are patients who have experienced blunt trauma to the cerebral vessels and thus may already be at increased risk for post-traumatic cerebral infarction, a well recognized complication of TBI.190 Limited cohort study data suggest that rFVIIa administration to this group of patients may increase the risk of post-traumatic cerebral infarction even further.108

Comparison to Premier Database

Study patients, with mean ages 36–52, were younger by 15–20 years, on average, than patients in the Premier database (mean age 63 years). Those in the Stein cohort study experienced a mortality rate comparable to the rate of 0.34 in the Premier database, but the RCT patients had a lower mortality rate than either of these. The Premier database indicated a low level of rFVIIa use for brain trauma in the early 2000s but, similar to the intracranial hemorrhage and body trauma indications, there was an increase in use in 2005 and a possible leveling off more recently (Figure 19).

Figure 19 depicts the Premier database rFVIIa rise in intracranial
hemorrhage and trauma of the body and brain, 2000- 2008. The Premier
database indicated a low level of rFVIIa use for brain trauma in the early
2000s but, similar to the intracranial hemorrhage and body trauma
indications, there was an increase in use in 2005 and a possible leveling
off more recently.

Figure 19

Premier database rFVIIa use in intracranial hemorrhage and trauma of the body and brain.

Strength of Evidence

The strength of evidence assigned to all outcomes was low, on the basis of determinations of a medium to high risk of bias for all study types (driven by the uniformly “fair” quality scores) and imprecise estimates of effect. The imprecision of effects for the direct morbidity and mortality outcomes was driven by the low event rates for these outcomes—or in the case of change in hematoma volume, a small absolute effect size—among small patient populations, suggesting that these studies were underpowered for these outcomes (Table 32).

Table 32. Strength of evidence for rFVIIa use in brain trauma.

Table 32

Strength of evidence for rFVIIa use in brain trauma.

Applicability

The overall applicability is fair for treatment use in the population targeted —patients with intracranial hemorrhage seconday to TBI, most of whom were not on anticoagulation (Table 33). The population applicability of both included studies is only fair. The Narayan RCT was limited by the small percent of patients screened who were ultimately enrolled (four percent), just as the data we evaluated from the Stein cohort was limited, in our analysis, by the inclusion of only those patients with isolated TBI rather than those with both TBI and body trauma, the latter of which is a more common pattern of injury. Another important limitation to applicability derives from the small sample sizes, which preclude meaningful determinations regarding the most direct measures of morbidity and mortality. Similarly, while the studies made admirable attempts at ascertainment of thromboembolic harms, their follow-up of 15 days is likely insufficient to make important determinations regarding the long-term health ramifications of rFVIIa therapy. Finally, the study settings at specialized trauma centers in multiple countries are likely comparable to trauma centers in the U.S.

Table 33. Applicability assessment of studies on brain trauma.

Table 33

Applicability assessment of studies on brain trauma.

Conclusions

Current evidence of low strength is too limited to compare the harms and benefits of rFVIIa versus usual care for intracranial bleeding due to brain trauma. The lone RCT identified no difference in mortality, while the one cohort study identified a non-significant reduction with rFVIIa treatment. There was a trend across the two studies toward increased thromboembolic events with therapy. The cohort study found a significant reduction in time to neurosurgical intervention. Across studies, event rates for mortality and thromboembolism were low, and there was low strength of evidence for all effect estimates. Study patients were younger than those in the Premier database, and compared to Premier patients, those in the RCT had a lower mortality rate, while those in the cohort study had a comparable mortality rate. The importance and nature of interactions between rFVIIa administration and coagulopathy, congestive heart failure, or blunt trauma injuries to the cerebral vasculature remain unclear.

Key Question 4.a. Liver transplantation and comparative effectiveness of rFVIIa

Background

Liver transplantation is associated with considerable, and in some cases massive, blood loss, not the least because the patients who undergo surgery have significant chronic, acquired coagulopathies related to their advanced liver disease. They often require intraoperative transfusions to correct for the coagulopathies and surgical blood loss. Such transfusions are associated with increased rates of post-operative mortality, multi-organ dysfunction, and infection, as well as reduced graft survival.191–194 The use of prophylactic rFVIIa at the initiation of surgery was investigated for its potential to lower the number of intraoperative transfusions required, thereby ameliorating some of these risks. However, there were concerns about rFVIIa increasing the risk of clotting at unforeseen and unwelcome sites. For instance, thrombosis of the hepatic vessels is a well-recognized complication of liver transplantation that can be devastating.195 Hepatic artery thombosis occurs in approximately five percent of transplantations, generally early after transplantation, and results in graft failure.

Usual care during the time frame of included studies. Over the past 10 years, the average transfusion requirement during liver transplantation has decreased dramatically.196 Much of this change has been attributed to advances in surgical and anesthetic technique, improved understanding and management of coagulopathy, and recognition of the significant risks associated with transfusions that are described above.197 Although wide variation between centers still exists, the current median RBC transfusion requirement per transplantation is estimated to be less than five units. This compares to a median of 20 or more units in the past. A small but growing proportion of patients is now surviving liver transplantation without the transfusion of any blood products.198

Jehovah’s Witnesses as a special population. For religious reasons, Jehovah’s Witnesses refuse the transfusion of blood products (whole blood, red blood cells, white blood cells, platelets, and plasma). However, the religion enjoins its members to make personal determinations regarding the appropriateness of infusion of blood fractions such as cryoprecipitate, recirculated autologous blood, cell-saved blood, albumin, and recombinant products, which some Jehovah’s Witness patients will accept. The first liver transplantation in a Jehovah’s Witness patient occurred in 1994.199 But there have been a growing number since that time, and the challenge of these are discussed in the transplantation literature.200,201 Usual care in these patients can include pre-operative use of recombinant erythropoietin and intra-operative use of recirculated autologous and cell-saved blood, aprotinin, cryoprecipitate, and albumin. Products that minimize blood loss in such patients are eagerly sought and may change their risk-benefit ratio of transplantation.

General Characteristics of Studies of Bleeding in Liver Transplantation

We identified four RCTs (two fair quality, two poor quality) and one comparative retrospective cohort study (fair quality) that examined the prophylactic use of rFVIIa in 215 liver transplant recipients (Table 34). The RCTs by Planinsic111 and Lodge 110 both had authors employed by the manufacturer of rFVIIa. The RCTs by Pugliese112 and Liu 113 reported no financial ties to the manufacturer. The comparative observational study by Hendriks 114 had one author employed by the manufacturer.

Table 34. General characteristics of comparative studies on off-label rFVIIa use for liver transplantation.

Table 34

General characteristics of comparative studies on off-label rFVIIa use for liver transplantation.

Place of studies within analytic framework. Liver transplantation is the one indication for which all of the included studies evaluated rFVIIa for prophylactic use (versus treatment or end-stage use, which are other potential uses, as outlined in our Analytic Framework (Figure 1)). Among the four RCTs, those by Lodge,110 Planinsic,111 and Pugliese 112 reported at least minimal comparative data on the direct outcomes of mortality and thromboembolic events and the indirect outcomes of RBC transfusion requirements, operating room (OR) time, and ICU length of stay, wheres the Liu RCT113 only reported comparative data on OR time. Furthermore, only the Lodge and Planinsic studies explicitly identified a “primary outcome” (in this case, two for both)—RBC transfusion requirements during the procedure itself and during the first 24 hours post-peratively. The Hendriks observational study did not identify a primary outcome, but reported on the direct endpoints of thromboembolic events and the surrogate endpoints of transfusion requirements, blood loss, and operating time.

Comparison to studies on other indications. The studies of rFVIIa use in liver transplantation all evaluated its prophylactic application in comparison to the treatment use to which it was applied in most of the other indications. The mean age of patients was approximately 50 years, which is similar to those in the studies of brain trauma and younger than those in the studies of intracranial hemorrhage or adult cardiac surgery. The dose of rFVIIa was broad, with an approximate range of 20–360 μg/kg. All but one study evaluated a single dose of rFVIIa.

Patient Characteristics and Study Design

None of the studies described having enrolled patients who were Jehovah’s Witnesses.

RCTs. We identified four RCTs (two fair quality, two poor quality (Table 14)). The first, by Lodge,110 was a double -blind study with modest sample sizes: 61 patients in the usual care group and 121 patients in the aggregate rFVIIa group (divided relatively evenly between two dosing groups). It was conducted in 14 hospitals in Europe between August 2001 and September 2003. Inclusion criteria targeted adults with end-stage liver disease (defined as Child-Pugh class B or C) requiring liver transplantation. Important exclusion criteria included previous liver transplantation and multiorgan transplantation, both of which are associated with higher rates of bleeding during surgery. The groups were well-matched at baseline, with the possible exception of the rFVIIa patients having a slightly higher rate of B scores on the Child -Pugh scale (i.e., they may have been slightly less sick) than the usual care group.

The Planinsic RCT111 was a double -blind study with small sample sizes: 19 patients in the usual care group and 64 patients in the aggregate rFVIIa group (divided relatively evenly between three dosing groups). It was conducted in nine hospitals in the U.S. and Europe between February and September 2000. The inclusion and exclusion criteria, baseline matching, and primary outcomes were essentially the same as those described for the Lodge RCT above.

The primary outcomes reported were the number of RBC units transfused both intraoperatively and in the first post-operative 24 hours.

The Pugliese RCT112 was a double -blind study with very small sample sizes: 10 patients in the usual care group and 10 in the rFVIIa group. It was conducted at a single Italian center from November 2003 to July 2004, and had an extremely short six hour time frame of data collection and follow-up. Inclusion criteria included patients “who underwent OLT [orthotopic liver transplantation]” (Child-Pugh class not specified) and who also had hemoglobin greater than eight mg/dL, INR greater than 1.5, and fibrinogen greater than 100 mg/dL. Exclusion criteria and the primary outcome were not specified.

The Liu RCT113 contained no description of blinding and was also small, with 14 patients in both the treatment and usual care groups. It was conducted at a single center in China from March 2003 to July 2006 and does not designate the timeframe of follow-up. The sole inclusion criterion appears to have been need for liver transplantation, and the exclusion criteria and primary outcome were not specified.

Comparative observational study. The fair quality Hendriks cohort study evaluated patients treated at a single hospital in the Netherlands.114 It evaluated only 12 patients who received usual care matched in a 2:1 ratio with six patients who received rFVIIa. Inclusion criteria were similar to those of the Lodge and Planinsic RCTs, namely adult patients with end-stage liver disease (defined as Child-Pugh class B or C) requiring liver transplantation. Controls were chosen from the hospital database of patients who had received a liver transplantation. They were matched in a 2-to-1 ratio with rFVIIa patients on the characteristics of year of transplantation, Child-Pugh score, blood urea nitrogen levels, and cholestatic versus non-cholestatic liver disease. The groups were well matched at baseline, with the possible exception of longer cold ischemia time for the livers transplanted into the usual care group. The primary outcome was not specified.

Intervention Characteristics

RCTs. Patients in the four RCTs received infusions of rFVIIa or placebo at the initiation of surgery. In the Lodge study,110 treated patients received one dose of either 60 or 120 μg/kg at the initiation of surgery, repeat doses in the same amount every two hours while still in the OR, and a final dose in the same amount at the time of wound closure. Most patients in the study received a total of three doses for approximate cumulative doses of 120 and 360 μg/kg, respectively. In the remainder of the studies, treated patients received a single dose of rFVIIa: in the Planinsic study111 a dose of 20, 40, or 80 μg/kg; in the Pugliese study112 a dose of 40 μg/kg; and in the Liu study113 a dose of “70–80 μg/kg.”

Comparative observational study. Treated patients in the Hendriks cohort study 114 received a single 80 μg/kg dose of rFVIIa at initiation of surgery.

Outcomes

Direct (patient-centered) outcomes. There were no dose -dependent trends in the outcomes reported and the events rates for these outcomes were low. Therefore, the event rates in the text are given for the aggregate rFVIIa group compared to the usual care group. Table 35 also summarizes these findings.

Table 35. Mortality and thromboembolic events in comparative studies of rFVIIa use in liver transplantation.

Table 35

Mortality and thromboembolic events in comparative studies of rFVIIa use in liver transplantation.

Mortality. The Liu RCT 113 did not report on mortality. The remaining three RCTs failed to define explicitly the time to follow-up for the mortality endpoint. However, all of them reported on the number of transfusions within the first 24 -hour post-operative period and/or the subsequent ICU length of stay, such that the minimum follow-up time for mortality may reasonably be expected to be 24 hours. There were no differences between groups in mortality rates, although the studies were not powered for this outcome. The Lodge study110 noted “6 deaths” in the text but provided further details on only four of these in its associated table, for minimum mortality rates of two percent (3 of 121) in the aggregate rFVIIa group and two percent (1 of 61) in the usual care group. The Planinsic study111 stated that “seven deaths occurred during the study period,” but described only five deaths for the following minimum mortality rates: six percent (4 of 64) in the aggregate rFVIIa group and five percent (1 of 19) in the placebo group. The Pugliese study112 reported no deaths in either group. The Hendriks cohort study114 also did not report on mortality. To place the mortality differences in the context of the comparable findings for the other clinical indications, also see Figure 5 above (in the Key Question section on intracranial hemorrhage).

Thromboembolic events. There were no differences between groups noted by any of the studies for this outcome. The Lodge RCT110 noted 16 percent (19 of 121) in the aggregate rFVIIa group and 10 percent (6 of 61) in the usual care group. The Planinsic RCT111 described thromboembolic event rates of 13 percent (8 of 64, comprised of five arterial thromboses, one myocardial infarction, and two episodes of thrombophlebitis) in the aggregate rFVIIa group and 16 percent (3 of 19, comprised of two arterial thromboses and one episode of thrombophlebitis) in the usual care group. The Pugliese112 RCT stated that no thromboembolic events occurred in either group. The Liu RCT113 only stated that no events occurred in the treatment group, without commenting on the control group. The Hendriks cohort study114 reported only one thromboembolic event in the study—thrombosis of the hepatic artery in a patient who received rFVIIa: an event rate of 17 percent (1 of 6), versus 0 of 12 in the usual care group. To place the differences in thromboembolic event rates in the context of the comparable findings for the other clinical indications, also see Figure 6 above (in the Key Question section on intracranial hemorrhage).

Indirect (surrogate) outcomes. There were no dose-dependent trends noted for the indirect outcomes described in detail below and summarized in Table 36.

Table 36. Indirect outcomes in comparative studies of rFVIIa use in liver transplantation.

Table 36

Indirect outcomes in comparative studies of rFVIIa use in liver transplantation.

RBC transfusion during the 24-hour post-operative period. Among the RCTs, the Lodge study identified a non-significant reduction and the Pugliese study identified a significant reduction in the 24-hour post-operative RBC transfusion requirements with treatment, whereas the Planinsic study found no difference between groups. The Liu study did not report this outcome, but did note a significant reduction in blood loss. In the Hendriks cohort study114 the treatment group had significantly lower RBC transfusion requirements.

OR time. Only the Liu RCT identified a significant difference between groups in OR time, with transplantation taking over 200 minutes less in the rFVIIa group compared to placebo. None of the other studies identified a difference between groups. Both the Lodge110 and Planinsic111 RCTs reported “no notable difference between study groups” for this outcome but gave no further details. While the OR times were slightly shorter for the rFVIIa patients compared to controls in both the Pugliese RCT and Hendriks cohort study (in minutes, 408 versus 432 and 554 versus 598, respectively), these differences were not significant.

ICU length of stay (LOS). None of the RCTs reporting on this outcome found a difference between groups in ICU LOS. The Planinsic RCT111 stated only that “the mean number of intensive care unit days was comparable between study groups,” whereas the Lodge and Pugliese RCTs did report duration for each group. The Liu RCT and Hendriks cohort did not report on this outcome.

Consideration of poor quality comparative observational studies. In the poor quality comparative observational studies by De Gasperi,115 Kalicinski, 116 and Niemann, 117 the findings on mortality, thromboembolic events, and RBC transfusions requirements were consistent with those described above (Tables 35 and 36). Other outcomes were not reported.

Comparison to Premier Database

The mean age of study patients of approximately 50 years was comparable to the mean age of 52 years for Premier patients. Mortality rates among patients in all of the comparative studies were considerably lower than the mortality rate of 0.38 for patients who underwent liver transplantation in the Premier database.

Strength of Evidence

The strength of evidence assigned to all outcomes was low (Table 37). These assessments were made, in part, on the basis of the poor to fair quality scores of the included studies, which prompted a determination that the studies had a medium or high risk of bias in all cases. In addition, there was little certainty regarding the effect size estimates, leading to uniform imprecision on that determinant. Low event rates for all of the morbidity and mortality outcomes contributed to the imprecision for these outcomes, while wide confidence intervals contributed to the imprecision for the surrogate outcomes.

Table 37. Strength of evidence grade for rFVIIa use in liver transplantation.

Table 37

Strength of evidence grade for rFVIIa use in liver transplantation.

Applicability

The overall applicability was fair for prophylactic use in the population targeted—adult patients with cirrhosis of Child’s class B or C. Such patients represent the usual population requiring liver transplantation for cirrhosis. The range of rFVIIa doses administered limits applicability in terms of choice of prophylactic dose. Another limitation is the emphasis on indirect (surrogate and process) outcomes, rather than direct measures of mortality and morbidity, which would have required much larger studies. Follow-up times were short and poorly defined, and in most cases there were limited descriptions of ascertainment for harms, which reduces study applicability in these areas. Finally, the setting of the studies in regional referral centers in the U.S. and abroad has good applicability to U.S. centers where liver transplantations are performed (Table 38). The evidence is not applicable to treatment or end -stage use of rFVIIa for this indication or to patients who require liver transplantation for indications other than Child’s B or C cirrhosis, because such patients were excluded.

Table 38. Applicability assessment of studies on liver transplantation.

Table 38

Applicability assessment of studies on liver transplantation.

Conclusions

The available evidence of low strength is too limited to compare the benefits and harms of prophylactic use of rFVIIa compared to usual care in liver transplantation. There was no evidence of impact on mortality or thromboembolic events but event rates were low. There was a weak trend toward reduced RBC transfusion requirements in the studies as a whole, but no differences between groups for the other indirect outcomes (OR time and ICU length of stay). Study patients were comparable in age to those in the Premier database but had lower mortality rates. The impact of rFVIIa on the care of Jehovah’s Witness patients remains unclear, in the absence of comparative study data in this population.

Key Question 4.b.i. Adult cardiac surgery and comparative effectiveness of rFVIIa

Background

Despite advances in methods to control blood loss during and after cardiac surgery, perioperative blood transfusions are required in up to 80 percent of adult patients, and 3–5 percent of these patients require post-operative transfusions of over 10 RBC units. 202–204 Mediastinal exploration for ongoing post-operative bleeding is necessary in 3–10 percent of patients.205,206 Post-operative bleeding that is refractory to surgical re-exploration or conventional hemostatic therapy is felt to be multifactorial, with contributions from the use of antiplatelet agents prior to surgery and various causes of coagulopathy triggered by the surgery itself: residual heparin effect after cardiopulmonary bypass (CPB), hypothermia, hemodilution causing both thrombocytopenia and dilutional coagulopathy, consumption of coagulation factors, hyperfibrinolysis, inflammatory cascade activation, and platelet consumption and dysfunction.207,208

Because anticoagulation is necessary during the period on CPB, the optimal time period for potential use of rFVIIa is some time after discontinuation of CPB and reversal of the anticoagulation. For “prophylaxis” of post-CPB bleeding, rFVIIa is administered immediately after conclusion of CPB. For “treatment” of post-CPB bleeding, rFVIIa is given at such time (variably defined) when excessive bleeding is identified and felt to require treatment. One group of patients known to be at increased risk for excessive bleeding includes those who have undergone complex cardiac surgeries: repeat surgeries, surgeries involving more than one procedure (e.g., multiple valve replacements or repairs), aortic root or arch replacements, and surgeries for aortic dissection or endocarditis.202

Usual care during the time frame of included studies. Ten years ago, 25–95 percent of adult cardiac surgery patients in the U.S. and U.K. received at least one unit of RBCs post-operatively, with wide variations in local practice. But recent studies have exposed the potential detrimental effects of post-operative transfusion.209 Murphy and Angelini evaluated studies published from 1996 to 2006 and found consistent evidence that transfusions increased the likelihood of infection, stroke, renal failure, prolonged ventilation, and both short-and long-term mortality.210 Thus, usual care appears to be shifting toward limiting the practice of post-operative transfusion, whenever possible. All of the included studies were completed by November 2007, when the antifibrinolytic drug aprotinin—commonly used during cardiac surgeries until that time—was removed from the U.S. market due to safety concerns. Since then, tranexamic acid has become the antifibrinolytic agent of choice during cardiac procedures.

General Characteristics of Studies of Adult Cardiac Surgery

We identified for inclusion two RCTs (one good quality, one fair quality) and four comparative observational studies (two good quality, two fair quality) with 252 patients given rFVIIa as prophylaxis or treatment use following completion of CPB. One RCT examined the efficacy of rFVIIa as prophylaxis for bleeding and given immediately after termination of CPB in complex, non-coronary artery bypass grafting (non-CABG) surgery (Table 39). It stated that it was not sponsored by Novo Nordisk.118 A second RCT evaluated rFVIIa use following any cardiac surgery requiring CPB, including isolated CABG, as treatment for excessive post-operative bleeding in the ICU, similar to the comparative observational studies described below. This latter study was sponsored by the manufacturer. A third RCT, Ma 2006, 211 was published in Chinese but contained an abstract and data tables in English. It examined prophylactic administration of rFVIIa immediately following CPB. It is not included in our primary analyses but is used for sensitivity analyses.

Table 39. General characteristics of comparative studies on off-label rFVIIa use for adult cardiac surgery.

Table 39

General characteristics of comparative studies on off-label rFVIIa use for adult cardiac surgery.

The remainder of the studies, one prospective cohort212 and four retrospective cohorts, 120–123 also evaluated rFVIIa used as treatment for excessive post-operative bleeding, typically in patients with non-CABG surgery. A certain number of patients receiving isolated CABG were included in some of the studies,121,123 most notably in the Gelsomino cohort, of which approximately 45 percent received an isolated CAGB.121 The definition of “excessive bleeding” across studies was variable, but generally included some combination of the following criteria: bleeding that compromised hemodynamics, prevented chest closure, or crossed a certain threshold in the first post-operative hour (range 100–500 mL/h) or for a certain number of consecutive hours thereafter (range 100–300 mL/h). The majority of the patients in each study, and in some cases all of them, received the study drug in the ICU. One of these studies made no mention of manufacturer sponsorship, 121 while the other three explicitly stated that they had no such sponsorship.120,122,123

Place of studies within analytic framework. Adult cardiac surgery was the only indication for which there were comparative studies on more than one type of use of rFVIIa, namely prophylactic and treatment dosing (but not end-stage use) (see our Analytic Framework (Figure 1)). The Diprose RCT118 examined prophylactic use, while the Gill RCT119 examined treatment use, although in both cases, rFVIIa was given after termination of CPB. The remaining four studies, 120–123 all of which were observational, examined treatment use. All of the studies provided comparative data of some sort on the direct outcomes of mortality and thromboembolic events and the surrogate outcomes of RBC (or in the case of the Gill RCT, total) transfusion requirements, as well as ICU length of stay.

Qualitative considerations of heterogeneity. One obvious source of potential heterogeneity among the adult cardiac surgery studies is the distinction between prophylactic and treatment use. However, in both types of use, rFVIIa is administered after CPB has concluded, when the effects of the many potential causes of coagulopathy discussed above may already be in full force—including the effects of residual heparin from CPB, hypothermia, hemodilution, platelet consumption and dysfunction, and the like. For this reason, we determined that patients in the one study on prophylactic rFVIIa use, the Diprose RCT, might be expected to respond to the drug similarly to the patients in the remaining studies of treatment use, and we therefore chose to analyze the studies together.

Comparison to studies on other indications. Like the Diprose RCT (and the Ma RCT in Chinese used for sensitivity analyses), the other indications that assessed prophylactic use of rFVIIa are those of liver transplantation, pediatric cardiac surgery, and prostatectomy. In contrast, the Gill RCT and all of the cohort studies evaluated treatment use of rFVIIa for bleeding during or after surgery, similar to the ICH and trauma indications. The mean age in the studies ranged from the mid to high 60s, which is comparable to the mean ages in the intracranial hemorrhage and prostatectomy studies. The rFVIIa dose was on the lower end of those studied (17–90 μg/kg) and was typically only given in the form of 1–2 infusions, rather than the multiple infusions seen in some of the other indications.

Patient Characteristics and Study Design

RCTs. The fair quality Diprose RCT 118 was a small double blind study of prophylactic use of rFVIIa conducted in one U.K. hospital with 10 patients in the usual care group and 10 patients in the rFVIIa group, and with the administration of the study drug or placebo at the conclusion of CPB. To qualify for inclusion, patients had to have undergone complex cardiac surgery, defined as repeat non-coronary surgery, multiple valve surgeries, surgery at the aortic root or arch or for aortic dissection, or surgery for endocarditis. All patients in the study received a set dose of aprotinin during surgery. While the study identified no single primary outcome, it reported on various transfusion requirements for the intention-to-treat population. However, the outcomes of mortality, thromboembolic events, and ICU length of stay were reported only for the per-protocol population, which excluded one patient from the rFVIIa group for incurring “multiple transfusion protocol violations” after being unblinded at the request of the surgeon.

The good quality Gill RCT119 evaluated treatment use of rFVIIa for excessive post-operative bleeding in the ICU, was conducted in the U.S. and 12 other countries (in Africa, Asia, Europe, and South America), and enrolled 172 patients total, including 68 patients in the usual care group and 104 patients in the rFVIIa group. The description on the ClinicalTrials.gov website indicates that the trial was terminated in November 2007 “without proceeding to the highest dosing cohort [of 160 μk/kg] as this no longer reflects common clinical practice.”213 To qualify for inclusion, patients had to have undergone cardiac surgery that required CPB, which could include simple CABG (12 to 14 percent of patients in each group received only CABG), and also have reached a prespecified bleeding rate (>200 mL/h or >2 mL/kg/h for 2 consecutive hours) after a 30 minute stabilization period in the ICU. Patients were excluded if they were determined to require urgent re-operation or had a history of stroke or non-coronary thrombotic disorder. In the publication there is no mention of whether patients received aprotinin or tranexamic acid for antifibrinolysis during surgery. The primary outcome was the incidence of critical serious adverse events at 30 days, which could include death, myocardial infarction (MI), stroke, pulmonary embolus (PE), or other symptomatic thromboembolic events. Other outcomes reported in the publication included total transfusion volume and blood loss, but not ICU length of stay.

The Ma RCT211 published in Chinese and used for sensitivity analyses evaluated prophylactic rFVIIa use immediately following conclusion of CPB in 11 usual care patients versus 11 treatment patients. All patients underwent valve repair surgery, with a sub-set undergoing a double valve repair (27 percent rFVII group, 36 percent placebo); there were no other types of surgery performed. Because we have limited access to methodological information and outcomes data—based only on the English-language abstract and results tables—we used the data for sensitivity analyses but did not include them in the primary analyses.

Comparative observational studies. We identified four comparative observational studies (two good quality, two fair quality (Table 14)). Again, all of these studies assessed the treatment use of rFVIIa compared to usual care for ongoing bleeding after discontinuation of CPB. The two highest quality comparative observational studies were the Karkouti and Gelsomino cohorts, both of which earned “good” scores and were subsequently included with the Diprose and RCTs in the meta-analyses of direct outcomes.

The Karkouti cohort study assessed the first 51 patients treated with rFVIIa for cardiac surgery bleeding at a single Canadian institution between November 2002 and February 2004.120 The institutional policy required that a consultant hematologist approve the release of rFVIIa and that bleeding be “massive and refractory,” as defined in the paper. The attending physician could then choose between two doses of rFVIIa which were based on drug vial contents rather than patient weight: a 4.8 mg vial (an approximate 70 μg/kg bolus) or a 2.4 mg vial (an approximate 35 μg/kg bolus). Data were abstracted by a blinded research nurse and research assistant. Each patient who received rFVIIa was matched to a single control selected from the hospital’s database of patients who had undergone cardiac surgery in a similar time frame and using a propensity score that modeled for massive perioperative blood loss. Patients were well matched at baseline on most characteristics, such as complex surgery and mean CPB time, but did have significant differences on baseline rates of re-exploration, blood loss, and transfusion products, all of which were higher in the rFVIIa group. No single primary outcome was identified, but the study reported on outcomes of RBC transfusion requirements, mortality, stroke, MI, PE, deep vein thrombosis (DVT), and ICU LOS, amongst others.

The Gelsomino cohort study121 assessed 40 patients treated at one Italian center, between September 2005 and June 2007, who had all types of cardiac surgeries, including isolated CABG, and received rFVIIa for “significant and refractory bleeding,” as defined in the paper. The dose of rFVIIa was the lowest administered for this indication, consisting of a uniform dose of 1.2 mg (a median dose of 18 μg/kg for included patients) that could be repeated once for continued bleeding. rFVIIa patients were matched 1:1 with 40 controls who received cardiac surgery at the hospital over the same time frame using a propensity score. Patients were well matched at baseline for important characteristics, such as the proportion undergoing complex cardiac surgery (approximately 40 percent in each group) and baseline transfusion requirements. The one possible exception to successful matching was the 30 minute longer CPB time in the usual care group (p=0.06). No single primary outcome was identified, but the study reported on outcomes of RBC transfusion requirements, mortality, stroke, respiratory failure, and ICU LOS, amongst others.

The von Heymann cohort123 identified 26 patients who received 1–3 doses of 60 μg/kg rFVIIa at a median of 14 hours following CPB at one German institution between June 2000 and March 2003. The majority of patients received a single dose. The study subsequently excluded two patients because they died during the initial 24-hours after treatment. The remaining 24 treated patients were matched 1:1 with controls from a similar time frame and the same hospital. The controls had to have experienced blood loss over 1000 mL in the first 14 hours after CPB, and were also matched to rFVIIa patients based on the “complexity” of the surgery (defined as single procedure (e.g., isolated CABG or single valve) or combined procedure (e.g., CABG plus valve replacement)). While groups were well-matched at baseline on APACHE II scores, there were possibly important differences in rates of emergency or redo surgeries (higher in the usual care group) and liver failure and endocarditis (higher in the rFVIIa group). No single primary outcome was identified, but the study reported on outcomes of RBC transfusion requirements, need for reexploration, mortality, thromboembolic events, and ICU LOS, amongst others.

The Tritapepe cohort122 was the only study to report on patients who received a single type of surgery, in this case repair of proximal type A aortic dissections, a high-risk, complex cardiac surgery. It identified 23 consecutive patients who received at least one 70 μg/kg dose of rFVIIa for refractory bleeding between January 2000 and March 2006. The majority of patients received a single dose. Usual care patients were matched 1:1 with rFVIIa patients based on a propensity score for “use of rFVIIa,” the modeling of which did not include a variable for baseline bleeding or transfusion requirements. The groups were otherwise well-matched on baseline characteristics, with the possible exception of higher rates of Bentall procedures in the rFVIIa group. No single primary endpoint was identified, but the study reported on outcomes or RBC transfusion requirements, mortality, stroke, and ICU LOS, among others.

Intervention Characteristics

RCTs. In the Diprose RCT118 the patients received a single, prophylactic bolus of 90 μg/kg of study drug or placebo immediately following the conclusion of CPB. rFVIIa patients in the Gill RCT119 received one of two possible single doses after a 30-minute stabilization period in the ICU: 35 received 40 μg/kg and 68 received 80 μg/kg. In the Ma RCT211 published in Chinese, which we are using for sensitivity analyses, a single, prophylactic dose of 40 μg/kg was administered immediately after CPB.

Comparative observational studies. In the cohort studies, rFVIIa was administered as treatment for excessive post-CPB bleeding. Compared to the doses used for treatment (rather than prophylaxis) in other clinical indications, these doses were on average much lower: first-time median doses ranged from 18–70 μg/kg, with the majority of patients receiving only a single dose. In the Karkouti cohort,120 attending physicians could choose between two doses of rFVIIa which were based on drug vial content rather than patient weight: a 4.8 mg vial (an approximate 62 μg/kg bolus) or a 2.4 mg vial (an approximate 37 μg/kg bolus). The majority of patients received just one dose. Patients in the Gelsomino cohort121 received the lowest doses of rFVIIa in all of the cohort studies, consisting of a uniform dose of 1.2 mg (which was equal to a median doses of 18 μg/kg) that could be repeated once for continued bleeding—but a repeat dose was required in only three patients. Patients in the von Heymann cohort123 could receive 1–3 doses of 60 μg/kg rFVIIa a median of 14 hours following CPB, but the majority required only a single dose. The Tritapepe cohort122 used an initial treatment dose of 70 μg/kg dose rFVIIa, and, again, the majority of patients received only a single dose.

Outcomes

Direct (patient-centered) outcomes. Overall, mortality and thromboembolic event rates were consistently reported for the included RCTs and cohort studies. Table 40 summarizes the morbidity and mortality outcomes for the adult cardiac surgery studies. Given the limitations of the data, we were not able to evaluate for a dose-response relationship for these outcomes. As described further below, among the studies included in the meta-analyses, assessments of the significance and magnitude of heterogeneity by the Q and I2 statistics did not identify significant heterogeneity.

Table 40. Mortality and thromboembolic events in comparative studies of rFVIIa use in adult cardiac surgery.

Table 40

Mortality and thromboembolic events in comparative studies of rFVIIa use in adult cardiac surgery.

Mortality. No study reported a significantly higher mortality rate in the rFVIIa group. Among the RCTs and good quality cohort studies, there were non-significant reductions in mortality with rFVIIa treatment compared to usual care in the Diprose RCT and Gelsomino cohort, exactly equal mortality rates between groups in the Karkouki cohort, and a non-significant increase in mortality in the Gill RCT. The remaining cohort studies, deemed fairon quality scoring,122,123 demonstrated equal mortality rates with rFVIIa use versus usual care. The meta-analysis of mortality for the RCT and good quality cohort studies found no difference in mortality with rFVIIa versus usual care (risk difference 0.007; 95 percent CI −0.049 to 0.063; P value for the Q statistic 0.63) (Figure 21), a finding that remained when sensitivity analyses were performed using the data from the Ma RCT211 published in Chinese ( risk difference 0.006; 95 percent CI −0.046 to 0.059; P value for the Q statistic 0.78) (Figure 22). To place the mortality differences in the context of the comparable findings for the other clinical indications, see Figure 5 above (in the Key Question section on intracranial hemorrhage).

Figure 21 depicts a random effects model of mortality in adults from
cardiac surgery. The summary statistics are as follows: Risk difference
summary effect size 0.007 (95% CI: -0.0487; 0.063) P value for Q statistic:
0.63; I-Squared=0%

Figure 21

Mortality in adult cardiac surgery.

Figure 22 depicts a random effects model of a sensitivity analysis of
mortality in adult cardiac surgery that incorporates Chinese RCT data. The
summary statistics are as follows: Risk difference summary effect size 0.006
(95% CI: -0.046; 0.059) P value for Q statistic: 0.78;
I-Squared=0%

Figure 22

Sensitivity analysis: mortality in adult cardiac surgery that incorporates Chinese RCT data.

Thromboembolic events. The rFVIIa and control groups in the Diprose RCT had the same rates of thromboembolic events, whereas the Gill RCT had higher rates in the treatment group. The two good quality cohort studies together demonstrated a weak trend toward increased events with rFVIIa. The fair quality cohorts had low but equal rates in both groups. The meta-analysis of the RCTs and good quality cohort studies identified a higher rate of thromboembolic events in the rFVIIa group compared to controls(risk difference 0.053; 95 percent CI 0.01 to 0.096; P value for the Q statistic 0.99) (Figure 23). This finding was replicated on the sensitivity analyses that incorporated data from the Ma RCT 211 published in Chinese (risk difference 0.049; 95 percent CI 0.008 to 0.091) (Figure 24). To place the different thromboembolic event rates in the context of the comparable findings for the other clinical indications, see Figure 6 above (in the Key Question section on intracranial hemorrhage).

Figure 23 depicts a random effects model of all thromboembolic events
in adult cardiac surgery. The summary statistics are as follows: Risk
difference summary effect size 0.053 (95% CI: 0.01 to 0.096) P value for Q
statistic: 0.99; I-Squared=0%

Figure 23

All thromboembolic events in adult cardiac surgery.

Figure 24 depicts a random effects model of a sensitivity analysis of
all thromboembolic events in adult cardiac surgery that incorporates Chinese
RCT data. The summary statistics are as follows: Risk difference summary
effect size 0.049 (95% CI: 0.008 to 0.091) P value for Q statistic: 0.98;
I-Squared=0%

Figure 24

Sensitivity analysis: all thromboembolic events in adult cardiac surgery that incorporates Chinese RCT data.

Indirect (surrogate) outcomes. These results are summarized in Table 41.

Table 41. Indirect outcomes in comparative studies of rFVIIa use in adult cardiac surgery.

Table 41

Indirect outcomes in comparative studies of rFVIIa use in adult cardiac surgery.

RBC transfusion during the 24-hour post-operative period. The Karkouti study did not report comparative data on this outcome. The Diprose RCT and the only good quality cohort study to report on the outcome, the Gelsomino cohort, respectively demonstrated a non-significant and significant difference between groups that favored a reduction with rFVIIa use (p values 0.11 and <0.001, respectively). While the Gill RCT did not report on isolated RBC transfusions, it did report significant differences between the two treatment groups (40 and 80 μg/kg) and controls in total blood transfusion volume. The Ma RCT (published in Chinese and used only for sensitivity analyses) reported a significant reduction in units of RBCs transfused (rFVIIa 3.5 (SD 2.2) versus control 6.3 (SD 3.1) (p<0.01)).211 There were inconsistent findings among the remaining studies—namely, the fair quality cohorts. The von Heymann study found no difference between groups, and the Tritapepe study found a significant effect against the rFVIIa group (i.e., increased transfusions with treatment).

ICU length of stay. The Diprose RCT found no significant difference between groups, although the absolute difference in means indicated an increased ICU LOS within the rFVIIa group. The publication of the Gill RCT119 does not comment on ICU LOS, but the synopsis on the manufacturer’s website states that there was “no difference” between groups for this outcome.214 The Ma RCT (published in Chinese and used only for sensitivity analyses) reported a significantly shorter ICU LOS for the rFVIIa group (2.7 day (SD 0.5) versus 3.3 days (SD 0.7) for controls (p<0.05)).211 There were inconsistent findings among the cohort studies as well. The good quality Karkouti cohort study identified a significantly longer ICU LOS for rFVIIa patients compared to controls. However, the other good quality cohort study by Gelsomino had significant findings in the opposite direction—in favor of reduced time for the treatment group. The Tritatpepe study had non-significant findings in the same direction, whereas the remaining cohort study by von Heymann found no difference between groups.

Consideration of poor quality comparative observational studies. In the poor quality comparative observational studies by Bowman124 and Trowbridge,125 the findings on mortality, thromboembolic events, and RBC transfusions were generally consistent with those described above (Tables 40 and 41). Other outcomes were not reported.

Other Considerations

Timing of administration. The Karkouti cohort study120 overlapped substantially with another study by the same group,215 which was excluded from subsequent analysis on the basis of that overlap and because it focused on “determinants of complications” rather than the comparison of rFVIIa with usual care and had methodological limitations. Nevertheless, it highlighted an important question regarding the timing of treatment use of rFVIIa in cardiac surgery patients.

The study findings should be interpreted with caution, but did identify a non-significant increased risk of adverse events when rFVIIa was given “late” in the resuscitation—with “late” defined as infusion after nine or more units of RBCs had already been administered. None of the included studies performed a similar evaluation of the impact of timing of rFVIIa administration on outcomes.

Comparison to Premier Database

The age of study patients (mid 60s) was comparable to the mean age of 65 years for the Premier patients. The mortality rate of 0.23 among the Premier patients was higher than the rate found for rFVIIa patients in the RCTs but relatively comparable to those identified for the cohort studies. According to the Premier database, rates of use for the adult cardiac surgery indication have continued to increase in the community (with a much slower rate of increase followed by a recent leveling off for pediatric cardiac surgery patients who are discussed further in the subsequent section)(Figure 20).

Figure 20 depicts the Premier database rFVIIa rise in use in adult and
pediatric cardiac surgery, 2000-2008. According to the Premier database,
rates of use for the adult cardiac surgery indication have continued to
increase in the community (with a much slower rate of increase followed by a
recent leveling off for pediatric cardiac surgery patients who are discussed
further in the subsequent section).

Figure 20

Premier database rFVIIa use in adult and pediatric cardiac surgery.

Strength of Evidence

The strength of evidence assigned to the thromboembolic event outcome was moderate, based on the consistency and precision of findings on the meta-analyses of the higher quality studies for these outcomes (Table 42). The remainder of the outcomes were assigned a low strength of evidence based primarily on weaknesses in two strength of evidence domains: the risk of bias domain—which had ratings of a medium or high—and the precision domain—which had ratings of imprecise. The rating of imprecise was based on wide confidence intervals for the major outcomes which in turn were due to low event rates in studies that were underpowered to detect mortality and major morbidity outcomes.

Table 42. Strength of evidence for rFVIIa use in adult cardiac surgery.

Table 42

Strength of evidence for rFVIIa use in adult cardiac surgery.

Applicability

The overall applicability of the evidence for this indication is fair for both prophylactic and treatment use in the population targeted—adult patients undergoing cardiac surgery, including more straightforward procedures (e.g., isolated CABG) and complex surgeries (e.g., ascending aortic dissection repair) (Table 43). Such study populations are diverse, giving them good applicability. The relatively narrow rFVIIa dose range that was studied may also be relevant to practice. However, the variability among studies in use of rFVIIa for prophylaxis versus treatment may be a limitation. The follow-up time was no longer than one month in any study, and the methods of ascertainment for harm was generally not well described, which limits applicability. The study settings were primarily academic centers, which may be more applicable to regional referral centers than community hospitals.

Table 43. Applicability assessment of studies on adult cardiac surgery.

Table 43

Applicability assessment of studies on adult cardiac surgery.

Conclusions

There is evidence of moderate strength to suggest that the use of rFVIIa in adult cardiac surgery increases the rate of thromboembolic events compared to usual care. The strength of evidence was low for the remainder of outcomes, including for the finding of no effect of rFVIIa use on mortality. Among the RCTs and higher quality cohort studies, there was a trend toward reduced transfusion requirements with therapy, but no difference in ICU length of stay. Thus, current evidence of moderate strength (for thromboembolic events) or low strength (for all other outcomes) suggests that neither benefits nor harms substantially exceed each other. Study patients were similar in age to those in the Premier database. Compared to Premier patients, those in the cohort studies had comparable mortality rates, whereas those in the RCTs had somewhat lower mortality rates. The importance and nature of interactions between the timing of treatment use of rFVIIa and important clinical outcomes remain uncertain.

Key Question 4.b.ii. Pediatric cardiac surgery and comparative effectiveness of rFVIIa

Background

The infant population that requires cardiac surgery to correct congenital heart defects is particularly susceptible to dilutional coagulopathies that can cause excessive bleeding. RBC or whole blood transfusions are frequently required during operative or post-operative periods. No hemostatic agents have been found to be consistently useful in such surgeries. While a meta-analysis of aprotinin compared to usual care demonstrated reduced need for transfusions,216 these findings are not uniform217 and the drug is no longer available in the U.S. due to safety concerns. Multiple studies have shown that DDAVP does not reduce transfusion requirements during surgery for children with congenital heart disease.218–220 Pediatric cardiac surgery is further complicated by the frequent need post-surgery for the infant to remain on extracorporeal membrane oxygenation (ECMO) machines, despite the well recognized increased risk of thromboembolism with ECMO use.221

Usual care during the time frame of the included study. Pediatric cardiac surgeons differ in their preference for packed RBCs versus whole blood for transfusion support during surgery based on conflicting evidence in recent trials.222,223 Furthermore, recent studies have begun to explore whether transfusion use in pediatric cardiac surgery is associated with adverse effects, such as an increased risk of infection, similar to what has been suggested in the adult cardiac surgery literature.224

RCT Design and Intervention

We found only one study on prophylactic use of rFVIIa in pediatric cardiac surgery, a poor quality RCT. This small, double-blind, placebo-controlled trial by Ekert did not report any Novo Nordisk sponsorship.137 It initially enrolled 82 patients, but six patients (four in the usual care group and two in the rFVIIa group) “did not receive trial medication” and were subsequently excluded from the intention-to-treat analysis, which left 36 control and 40 rFVIIa patients to be evaluated. Treatment with 40 μg/kg of rFVIIa (or placebo) was administered immediately after termination of CPB in children under one year of age who were undergoing repair of complex congenital heart defects. A second dose could be administered 20 minutes later in the operating room at the discretion of the surgeons. A third dose does not appear to have been required in any case. The mean cumulative dose administered was 63 μg/kg. Groups were similar at baseline in age, weight, type of surgery, and CPB time. The primary outcome was time to chest closure. The study also reported on outcomes of transfusion requirements of “blood ” (which could include packed cells or whole blood) and thromboembolic events. It did not explicitly comment on mortality. There was no mention of ECMO use in the study(Table 44). While attempts at a six-week follow-up were made, these were largely unsuccessful because of considerable missing data.

Table 44. General characteristics of comparative studies on off-label rFVIIa use for pediatric cardiac surgery.

Table 44

General characteristics of comparative studies on off-label rFVIIa use for pediatric cardiac surgery.

Place of studies within analytic framework. The Ekert RCT evaluated prophylactic use of rFVIIa in pediatric cardiac surgery (versus treatment or end-stage use, which are other potential uses, as outlined in our Analytic Framework (Figure 1)).

Comparison to studies on other indications. This was the only included study on a solely pediatric patient population, in this case infants. The mean rFVIIa dose of 63 μg/kg was similar to the doses used in adults studies of prophylactic use.

Outcomes

The study identified no thromboembolic events and did not explicitly report mortality (Table 45). With respect to the primary endpoint, rFVIIa patients had a longer time to chest closure than did controls (98.8 minutes (SE 27.3) versus 58.3 minutes (SE 29.2), p=0.026). Nonetheless, there was a non-significant decrease in transfusion requirement for RBCs and/or whole blood in the rFVIIa group (Table 46).

Table 45. Mortality and thromboembolic events in RCT on rFVIIa use in pediatric cardiac surgery.

Table 45

Mortality and thromboembolic events in RCT on rFVIIa use in pediatric cardiac surgery.

Table 46. Indirect outcomes in RCT on rFVIIa use in pediatric surgery.

Table 46

Indirect outcomes in RCT on rFVIIa use in pediatric surgery.

Consideration of poor quality comparative observational studies. Unlike the RCT, the poor quality comparative observational studies by Tobias,139 Agarwal, 138 and Niles 140 evaluated treatment use of rFVIIa. Nonetheless, their findings on thromboembolic events and RBC transfusions were consistent with those described above (Tables 44 and 45), with the possible exception of the Agarwal study, which noted a higher rate of thromboembolic events in patients who received rFVIIa versus usual care (6 of 24 (25 percent) and 0 of 22, respectively) and also raised the possibility of an increased risk of severe events among the subgroup of patients on ECMO, which is discussed in the section immediately below. Among these studies, only the one by Tobias reported mortality data, noting no deaths in either group. Other outcomes were not reported.

Other Considerations

Risk of thromboembolic complications in the setting of ECMO use. A comparative cohort study by Agarwal et al. that did not meet criteria for inclusion in the effectiveness review due to methodological limitations nonetheless raises important questions about a subgroup of patients who might experience increased risk for harm within the population of infants undergoing congenital heart surgery.138 The Agarwal study evaluated a subgroup of 12 patients on ECMO who received therapeutic rFVIIa for bleeding after CPB completion, two of whom experienced life-threatening thromboembolic adverse events. In the first, the ECMO circuit clotted to the point where blood flow was compromised and the infant required emergent resuscitation and exchange of the circuit. The second patient, who underwent unsuccessful placement of a femoral arterial line after surgery and then received rFVIIa, developed an ipsalateral ischemic lower extremity, which required amputation, as well as both atrial and pericardial thrombi, which required surgical evacuation. In contrast, none of the 15 control patients on ECMO who received usual care were noted to have thromboembolic events of any kind. Given methodological limitations and small sample sizes, these findings should be interpreted with due caution.

Comparison to Premier Database

The Premier database indicates a low but steady level of use of rFVIIa among pediatric cardiac surgery patients(Figure 20). The mortality rate of 0.22 in the Premier database cannot be comparated to any mortality rate in the RCT, because the RCT did not report this outcome. The mean age of study patients (3.9–4.0 months) was lower than the mean age of Premier patients (2.6 years). Whereas study patients were receiving their first surgical intervention at the time of rFVIIa administration, the Premier database may include patients undergoing repeat cardiac procedures, because the full repair of congenital heart defects often involves staged surgeries performed over time, which may explain their older age compared to study patients.

Strength of Evidence

The strength of evidence was insufficient for all outcomes. This determination was made on the basis of having only one small, poor quality study for this indication, which put it at high risk for bias and limited its precision (Table 47).

Table 47. Strength of evidence for rFVIIa use in pediatric cardiac surgery.

Table 47

Strength of evidence for rFVIIa use in pediatric cardiac surgery.

Applicability

The overall applicability of the evidence for this indication is fair for prophylactic use in the population targeted—infant patients with congenital heart defects requiring cardiac surgery for repair. This population, while limited in absolute number, has good applicability to similar populations at specialized referral centers like the institution that was the setting for the RCT. The low dose of rFVIIa is likely applicable as an appropriate amount for prophylaxis. The emphasis on indirect outcomes rather than mortality and morbidity outcomes is a limitation to applicability. While attempts at adequate follow-up were made, these were unsuccessful, and the method of ascertainment of harms was not described, making the follow-up applicability poor (Table 48).

Table 48. Applicability assessment of study of pediatric cardiac surgery.

Table 48

Applicability assessment of study of pediatric cardiac surgery.

Conclusions

Current evidence is insufficient for comparing the harms and benefits of rFVIIa use in infant patients undergoing cardiac surgery for congenital heart defect repair. The importance and nature of interactions between rFVIIa administration, ECMO use, and the risk of thromboembolic events remain uncertain.

Key Question 4.c. Prostatectomy and comparative effectiveness of rFVIIa

Background and Changes in Usual Care

The usual care of patients who require prostatectomy has changed considerably over the time period encompassing and since the performance of the Friederich RCT141 on rFVIIa use in prostatectomy (the one included study for this indication). Whereas radical retropubic prostatectomy and the Millin prostatectomy were frequent in decades past, they have been supplanted in most centers by laparoscopic approaches, which are associated with significantly less bleeding and fewer transfusions.225 For example, the authors of a letter to the editor on the Friederich RCT noted that they do not routinely prepare blood for transfusion prior to their procedures, because in most surgeries they have no transfusion requirements,226 which has been the case for others as well.227 Other changes in surgical approaches, including the use of robotics, have greatly reduced blood loss and hence transfusion requirements. These changes in practice likely account for the negligible use of rFVIIa noted in the Premier database.

RCT Design and Intervention Characteristics

The only identified study of rFVIIa use in prostatectomy was the small Friederich RCT 141 that did not report any Novo Nordisk sponsorship and was deemed to be of fair quality (Table 49). It evaluated placebo versus prophylactic use rFVIIa in patients who were not on anticoagulation and were undergoing one of two possible procedures: radical retropubic prostatectomy for prostate cancer or the Millin procedure for benign prostatic hypertrophy. Study drug or placebo was administered after lymph node dissection in the former and placement of guiding sutures in the latter. The RCT enrolled 12 usual care patients and an aggregate of 24 rFVIIa patients, of whom eight received 20 μg/kg and 16 received 40 μg/kg, with the patients who received the different procedures being well balanced between groups. Groups also appeared well balanced at baseline on age and weight, but there was no other information given on other potentially relevant characteristics. Analyses were intention-to-treat and follow-up occurred at 10 days. There was no single primary outcome identified, but the study reported on RBC transfusion requirements, mortality, thromboembolic events, and the duration of operation (OR time).

Table 49. General characteristics of comparative studies on off-label rFVIIa use for prostatectomy.

Table 49

General characteristics of comparative studies on off-label rFVIIa use for prostatectomy.

Place of studies within analytic framework. The Friederich RCT examined prophylactic use of rFVIIa in prostatectomy (versus treatment or end-stage use, which are other potential uses, as outlined in our Analytic Framework (Figure 1)).

Comparison to studies on other indications. Studies of adult cardiac and liver transplantation also evaluated prophylactic use of rFVIIa. The mean age of the prostatectomy patients (61–64) was comparable to that of the intracranial hemorrhage and adult cardiac surgery patients. The single-dose infusion of either 20 or 40 μg of rFVIIa encompassed the lowest mean dose of any indication.

Outcomes

There were no deaths in either study group (Table 50). One patient in the 20 μg/kg dose group experienced a myocardial infarction at day 14, the only thromboembolic event identified in the study. The RBC transfusion requirements were significantly reduced in the rFVIIa group, as was the OR time (Table 51).

Table 50. Mortality and thromboembolic events in RCT on rFVIIa use in prostatectomy.

Table 50

Mortality and thromboembolic events in RCT on rFVIIa use in prostatectomy.

Table 51. Indirect outcomes in RCT on rFVIIa use in prostatectomy.

Table 51

Indirect outcomes in RCT on rFVIIa use in prostatectomy.

Comparison to Premier Database

The very few Premier database patients who received rFVIIa were older (mean age, 69 years) than the study patients (mean age, 61–64 years). There were no deaths in either sample.

Strength of Evidence

The strength of evidence was insufficient for all outcomes (Table 52). While the RCT was fair in quality, the patient sample size was low, which limits certainty regarding effect size estimates. In addition, the low event rates for thromboembolic events and mortality outcomes contributed to the determination of imprecision for these outcomes.

Table 52. Strength of evidence for rFVIIa use in prostatectomy.

Table 52

Strength of evidence for rFVIIa use in prostatectomy.

Applicability

Overall applicability of the evidence is poor for prophylactic use in the populations targeted—patients undergoing retropubic prostatectomy for prostate cancer or benign hyperplasia but not on anticoagulation (Table 53). The baseline transfusion requirements and amount of bleeding are much higher in the study patient population than in the general population of patients undergoing prostatectomy and study patients may be younger as well, both of which make the population applicability poor. In addition, the “usual care” approach to prostatectomy has evolved into something very different, in most cases, from the surgeries evaluated in the included RCT, thereby making the applicability of the comparator also poor. Specifically, surgeons now favor laparoscopic or robotic approaches and alternative surgeries that, on average, incur much less blood loss than their predecessor procedures, thus altering the context of rFVIIa administration. Regarding other criteria of applicability, the dose of rFVIIa was on the lower end of those observed, but was relatively narrow, granting it fair applicability. Other limitations to applicability include the emphasis on surrogate measures, the short follow-up time (10 days), and the setting of the study being an academic rather than community hospital, the latter of which is more common in practice.

Table 53. Applicability assessment of study on prostatectomy.

Table 53

Applicability assessment of study on prostatectomy.

Conclusions

Current evidence is insufficient for comparing the harms and benefits of rFVIIa use in prostatectomy. In addition, the usual care for prostatectomy has likely evolved far beyond the standard of care employed in the RCT, making its relevance to current practice uncertain.

Evaluation of Harms in Patients Who Received rFVIIa: Comparing the Premier, RCT, and Observational Study Data

Given concerns about the potential harms of rFVIIa, we sought to compare and contrast the absolute rates of morbidity and mortality associated with the use of rFVIIa across different types of evidence—that is, comparative and non-comparative data. We therefore examine mortality and thromboembolic event outcomes for the Premier database, RCTs, and observational studies. Additionally, we sought to evaluate the association of key predictors (study design, clinical indication for the use of rFVIIa, mean patient age, and total dose of rFVIIa) with these outcomes. For these analyses, we included data from the intervention arms of the RCTs and comparative observational studies, and from those non-comparative observational studies that reported relevant data for 15 or more patients (Table 54).

Table 54. Harms noted in the analysis of rFVIIa use.

Table 54

Harms noted in the analysis of rFVIIa use.

In prior sections of the report only comparative observational studies of fair or good quality were reviewed in detail. However, in this section all comparative observational studies are included, along with their non-comparative counterparts. Also note that we are unable to report on thromboembolic events following rFVIIa use in the Premier database.

Findings

Across clinical indications, the mortality rate ranged extremely widely from 0 to 0.87 and the thromboembolic event rate ranged from 0 to 0.39, depending on the clinical indication for use of rFVIIa, age, dose, and study design. Figure 25 presents the adjusted mortality rates by age and indication for the rFVIIa groups in the RCTs—it shows the enormous heterogeneity of effects across indications. Adjusted thromboembolic rates varied similarly (Figure 26).

Figure 25 presents the adjusted mortality rates by age and indication
for the rFVIIa groups in the RCTs—it shows the enormous heterogeneity of
effects across indications namely ICH, body trauma, brain trauma, liver transplantation, adult cardiac surgery, pediatric cardiac surgery, and
prostatectomy. On the left, adjusted mortality rate (by control) ranges from
-0.4 to 0.4. The bottom has ages, starting at 0. There are symbols for the
following: ICH, body trauma, brain trauma, liver transplantation, adult
cardiac surgery, pediatric cardiac surgery, and prostatectomy. With the
exception of two points for liver transplantation, all other points are
after about age 45. The greatest mortality rate is approximately just below
0.3 and is for ICH. All other points are between -0.2 and 0.2.

Figure 25

Adjusted mortality rate by age from the RCTs.

Figure 26 depicts the adjusted thromboembolic event rates by age and
indication for the rFVIIa groups in the RCTs—it shows the enormous
heterogeneity of effects across indications ICH, body trauma, brain trauma,
liver transplantation, adult cardiac surgery, pediatric cardiac surgery, and
prostatectomy. On the left, adjusted mortality rate (by control) ranges from
-0.4 to 0.4. The bottom has ages, starting at 0. There are three points on
the graph before the age of 40. There is an x for pediatric cardiac surgery
at approximately 1 year of age and 0.0; the other two points are at
approximately 23 and 37 near 0.0. The rest of the points are after
approximately age 50. The highest point on the scatter plot is for ICH and
is between 0.2 and 0.4 after age 60 and the lowest point on the scatter plot
is for ICH and is just above --0.4 after age 60.

Figure 26

Adjusted all thromboembolic event rate by age from the RCTs.

For each indication, Figures 27 and 28 present the mean mortality and thromboembolic event rates, respectively, by study design. In general, the mortality rates were lowest in the RCTs and highest in the Premier database. Similarly, in general, the thromboembolic rates were generally lowest in the RCTs and highest in the observational studies.

Figure 27 depicts the weighted mean mortality (95 percent CI) by study
design for each indicator: ICH, body trauma, brain trauma, liver transplantation, adult cardiac surgery, pediatric cardiac surgery, and
prostatectomy. There are seven scatter graphs in this figure, one for each
indicator. Each looks at a weighted mean mortality from 0.0 to 0.5 on the
left side and RCT, Comp Obs, Non Comp Obs, and Premier Data along the
bottom. In general, the mortality rates were lowest in the RCTs and highest
in the Premier database.

Figure 27

Weighted mean mortality (95 percent CI) by study design for each indication.

Figure 28 depicts the weighted mean total thromboembolic event rate
(95 percent CI) by study design for each indicator: ICH, body trauma, brain
trauma, liver transplantation, adult cardiac surgery, pediatric cardiac
surgery, and prostatectomy. There are seven scatter graphs in this figure,
one for each indicator. Each looks at a weighted mean mortality from 0.0 to
0.25 on the left side and RCT, Comp Obs, and Non Comp Obs along the bottom.
In general, the thromboembolic rates were generally lowest in the RCTs and
highest in the observational studies.

Figure 28

Weighted mean total thromboembolic event rate (95 percent CI) by study design for each indication.

Figures 29 through 32 present the mortality and thromboembolic rates by mean age and total rFVIIa dose. In general, there was no apparent pattern relating age and mortality or age and thromboembolic events, although it is notable that the ages tended to cluster for a given indication. There was also no apparent pattern between dose and mortality or dose and thromboembolic events (with the possible exception of brain trauma, which is discussed further below). Figure 33 displays the mortality by thromboembolic event rate, and, again, there was no readily apparent pattern of association.

Figure 29 depicts mortality by mean age per indication for ICH, body
trauma, brain trauma, liver transplantation, adult cardiac surgery,
pediatric cardiac surgery, and prostatectomy. There are seven scatter graphs
in this figure. On the left mortality ranges from 0.00 to 1.00. At the
bottom, mean age begins at 0 and goes through 80. Data points in this figure
include only the intervention arms (i.e. rFVIIa patients) from comparative
studies. A data point for each dosing arm is presented for RCTs with
multiple dosing arms; thus, the number of studies may not equal the number
of data points. In general, there was no apparent pattern relating age and
mortality, although it is notable that the ages tended to cluster for a
given indication. There was also no apparent pattern between dose and
mortality or dose and thromboembolic events (with the possible exception of
brain trauma, which is discussed separately.

Figure 29

Mortality by mean age per indication. Data points in this figure include only the intervention arms (i.e. rFVIIa patients) from comparative studies. A data point for each dosing arm is presented for RCTs with multiple dosing arms; thus, the number of studies (more...)

Figure 30 depicts mortality by rFVIIa dose per indication for ICH,
body trauma, brain trauma, liver transplantation, adult cardiac surgery,
pediatric cardiac surgery, and prostatectomy. There are seven scatter graphs
in this figure. On the left is a mortality scale from 0.0 to 1.0 and at the
bottom, rFVIIa Dose ranges from 1 to 400. Data points in this figure include
only the intervention arms (i.e. rFVIIa patients) from comparative studies.
A data point for each dosing arm is presented for RCTs with multiple dosing
arms; thus, the number of studies may not equal the number of data points.
In general, there was no apparent pattern relating age and mortality or age,
although it is notable that the ages tended to cluster for a given
indication. There was also no apparent pattern between dose and mortality or
dose and thromboembolic events (with the possible exception of brain trauma,
which is discussed separately.

Figure 30

Mortality by rFVIIa dose per indication. Data points in this figure include only the intervention arms (i.e. rFVIIa patients) from comparative studies. A data point for each dosing arm is presented for RCTs with multiple dosing arms; thus, the number (more...)

Figure 31 depicts thromboembolic events by mean age per indication for
ICH, body trauma, brain trauma, liver transplantation, adult cardiac
surgery, pediatric cardiac surgery, and prostatectomy. There are seven
scatter graphs in this figure. On the left are thromboembolic event rates
from 0.0 to .5 and at the bottom, mean ages from 0 to 80. Data points in
this figure include only the intervention arms (i.e. rFVIIa patients) from
comparative studies. A data point for each dosing arm is presented for RCTs
with multiple dosing arms; thus, the number of studies may not equal the
number of data points. In general, there was no apparent pattern relating
age and thromboembolic events, although it is notable that the ages tended
to cluster for a given indication. There was also no apparent pattern
between dose and mortality or dose and thromboembolic events (with the
possible exception of brain trauma, which is discussed
separately.

Figure 31

Thromboembolic events by mean age per indication. Data points in this figure include only the intervention arms (i.e. rFVIIa patients) from comparative studies. A data point for each dosing arm is presented for RCTs with multiple dosing arms; thus, the (more...)

Figure 32 depicts thromboembolic events by rFVIIa dose per indication
for ICH, body trauma, brain trauma, liver transplantation, adult cardiac
surgery, pediatric cardiac surgery, and prostatectomy. There are seven mini
scatter graphs in this figure. At the left is a thromboembolic event rate
from 0.0 to 0.5, and at the bottom are rFVIIa Doses from 0 to 400. Data
points in this figure include only the intervention arms (i.e. rFVIIa
patients) from comparative studies. A data point for each dosing arm is
presented for RCTs with multiple dosing arms; thus, the number of studies
may not equal the number of data points. In general, there was no apparent
pattern relating age and thromboembolic events, although it is notable that
the ages tended to cluster for a given indication. There was also no
apparent pattern between dose and mortality or dose and thromboembolic
events (with the possible exception of brain trauma, which is discussed
separately.

Figure 32

Thromboembolic events by rFVIIa dose per indication. Data points in this figure include only the intervention arms (i.e. rFVIIa patients) from comparative studies. A data point for each dosing arm is presented for RCTs with multiple dosing arms; thus, (more...)

Figure 33 depicts the association of mortality and thromboembolic
event rates for ICH, body trauma, brain trauma, liver transplantation, adult
cardiac surgery, pediatric cardiac surgery, and prostatectomy. There are
seven scatter graphs in this figure with mortality on the left and ranging
from 0.0 to 1.0 and thromboembolic event rates across the bottom and ranging
from 0.0 to 0.5. Data points in this figure include only the intervention
arms (i.e. rFVIIa patients) from comparative studies. A data point for each
dosing arm is presented for RCTs with multiple dosing arms; thus, the number
of studies may not equal the number of data points.

Figure 33

Association of mortality and thromboembolic event rate. Data points in this figure include only the intervention arms (i.e. rFVIIa patients) from comparative studies. A data point for each dosing arm is presented for RCTs with multiple dosing arms; thus, (more...)

Intracranial Hemorrhage. In contrast to other clinical indications, most of the harms data for rFVIIa use in intracranial hemorrhage comes from RCTs rather than comparative observational studies (Figures 27 and 28 and Table 54). Four RCTs evaluated harms for ICH: in their various dosing arms mortality ranged from 0 to 0.375 and total thromboembolic events ranged from 0 to 0.33. Four comparative observational studies reported harms for ICH and, in a small number of patients, other forms of intracranial hemorrhage. Mortality ranged from 0.13 to 0.43, and total thromboembolic events ranged from 0.04 to 0.2. Mortality in the Premier database was higher (0.34) than in the aggregate RCTs and observational studies. Thromboembolic events were higher in the observational studies than in the RCTs.

The RCTs excluded patients on anticoagulants, whereas these patients were included in the observational studies and the Premier database. One observational study91 had a significantly higher mortality than the other studies (0.42). It specifically analyzed patients with warfarin-related ICH and other patients who were considered to be at “especially high risk” for hematoma growth.

Body trauma. Across study designs, mortality rates among patients receiving rFVIIa for body trauma were among the highest of all the clinical indications, although, again, the range was wide. The two RCTs reported the following event rates: mortality 0.25 and thromboembolism 0.03 in the blunt trauma trial and mortality 0.24 and thromboembolism 0.06 in the penetrating trauma trial. Twelve observational studies reported mortality and/or thromboembolic event rates among trauma patients. Rates for mortality ranged from 0.07 to 0.58 and thromboembolism ranged from 0 to 0.12. Mortality in the Premier database (0.33) was somewhat higher than that of the RCTs (Figures 27 and 28 and Table 54).

The wide range of mortality across study types may be due to differences in inclusion criteria. The RCT96 excluded patients over 65 years, those receiving greater than eight units of RBCs prior to arrival at the hospital, and those with severe acidosis or base deficit. Another study98, 182 of combat trauma patients reported that only 28 percent of patients would have met inclusion for the RCT.96 Several of the observational studies included patients on warfarin or other anticoagulants. As noted earlier in the report, we are aware of a recently concluded phase III trial of rFVIIa in trauma that also uses restrictive inclusion criteria—only two percent of patients from the largest Australian registry102 would have met its inclusion criteria.

The rate of thromboembolic events in trauma was somewhat lower than for other clinical indications. This may be the case because trauma patients were generally younger and healthier than patients receiving rFVIIa for other indications. It may also reflect differences in harms ascertainment, given the high rate of mortality in trauma (i.e., if patients die early, other harms may not be attributed to them). One observational study228 with a protocol for identifying thromboembolic complications among trauma patients who received rFVIIa (N=242) identified a thromboembolic event rate of 0.11—which was higher than the rates identified for the RCTs but still lower than the rates reported for other rFVIIa indications, particularly cardiac surgery.

Brain trauma. One RCT evaluated harms among rFVIIa patients with bleeding due to TBI. This study had five arms, each of which evaluated a different rFVIIa dose. Mortality in this study ranged from 0 to 0.33 between dosing arms and may be linear with dose, although the study was quite small, so that no definitive conclusions can be made. Interestingly, total thromboembolic events ranged from 0 to 0.21 and did not appear to be associated with dose. The comparative observational study and non-comparative observational study reported quite different mortality and thromboembolic rates from each other (0.34 and 0.07 for the former and 0.21 and 0 for the latter). Patients included in the Premier database had higher mortality (0.33) and were generally older than the patients in the RCTs (Figures 27 and 28 and Table 54).

Mortality in the comparative observational study107 was significantly higher than in the RCT.106 This may be due to differences in inclusion criteria: the RCT included patients with a Glasgow Coma Scale (GCS) of 4 to 14 and excluded patients with planned neurosurgery, whereas the observational study limited rFVIIa use to coagulopathic patient with possibly more severe TBI (GCS less than 9, INR greater than 1.4) who required neurosurgical management of their hemorrhage.

Liver Transplantation. Four RCTs evaluated liver transplantation patients: their reported mortality ranged from 0.02 to 0.08. However, two RCTs110, 111 did not report mortality by arm (treatment versus placebo); thus, this range underestimates the true mortality rate among patients receiving rFVIIa in RCTs. Rates of thromboembolic events ranged from 0 to 0.19. Again, this range underestimates the total thromboembolic event rate for patients in RCTs because one study111 did not report venous events by arm (treatment versus placebo). Four observational studies evaluated harm in liver transplantation: mortality ranged from 0 to 0.07, and the rate of thromboembolic events ranged from 0 to 0.17. Mortality was substantially higher in patients from the Premier database (0.38) than in other studies (Figures 27 and 28 and Table 54).

The treatment dose administered in the RCT by Lodge110 was substantially higher than in the other RCTs,111–113 and Lodge 110 reported a significantly higher rate of thromboembolic events—although it may be that the other studies simply under reported such harms. All studies of liver transplant recipients provided rFVIIa prophylactically and, in most cases, at lower doses (with the exception of the Lodge RCT) than in other indications. This may account, in part, for the relatively lower rates of mortality and thromboembolic harms among liver transplant recipients compared to patients who received rFVIIa for other indications.

Adult Cardiac Surgery. Two RCTs reported data on adult cardiac surgery. The reported mortality rates in the rFVIIa arms of the RCTs ranged from 0 to 0.11 and total thromboembolic event rates ranged from 0.06 to 0.22. Mortality in the Premier database (0.23) was high compared with the RCTs and comparative observational studies (Figures 27 and 28 and Table 54).

The mortality rates in the observational studies of cardiac surgery were generally higher than those reported in the RCTs (mortality, range 0 to 0.87). This may have to do with differences in inclusion criteria—namely, that the RCTs typically excluded the most severely ill patients, whereas observational studies often included such patients. For instance, one observational study 127 included only patients undergoing orthotopic heart transplant or LVAD implantation, and another122 enrolled patients undergoing aortic dissection repair.

Pediatric Cardiac Surgery. Harms were reported in the intervention groups of the one RCT and three observational studies of pediatric cardiac surgery. Overall, the rates of thromboembolic events were lower in the pediatric cardiac surgery studies than in the adult studies (Table 54). Mortality was not reported in the RCT, but was lower in the observational studies than in the adult studies or the Premier database.

Prostatectomy. There was one RCT of prostatectomy which reported no mortality and one myocardial infarction (Table 54).

Conclusions

Because this analysis includes data from observational studies and the Premier database, as well as from RCTs, it usefully highlights areas of consistency and contrast among different data sources. Across all indications except for prostatectomy (where no deaths were reported in any data set), mortality rates among patients in the Premier database were uniformly higher than the mortality rates in the RCTs, emphasizing that patients receiving rFVIIa in practice may differ in important ways from those in the included trials. Such distinctions may alter the risk-benefit profile of rFVIIa administration for real-world populations. Because of the limitations of our data, we can not comment further on how to interpret these differences, except to say that extrapolation from our comparative effectiveness review to real-world contexts, where the patients appear generally to be older and sicker, should be undertaken with appropriate caution.

Image methodsf1
Image appendixes.app3f1appendixes.app3f2appendixes.app3f3appendixes.app3f4appendixes.app3f5appendixes.app3f6appendixes.app3f7appendixes.app3f8appendixes.app3f9
Image appendixes.app3f2
Image appendixes.app3f3
Image appendixes.app3f4
Image appendixes.app3f5
Image appendixes.app3f6
Image appendixes.app3f7
Image appendixes.app3f8
Image appendixes.app3f9
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].
Comparative Effectiveness Reviews, No. 21.
Yank V, Tuohy CV, Logan AC, et al.

Download

AHRQ (US Agency for Healthcare Research and Quality)

PubMed Health Blog...

read all...

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...