Objectives: Recombinant human activated factor VIIa has been used prophylactically to mitigate requirements for transfusion in liver transplant. We explored its effectiveness and risks among liver transplant recipients at high risk for massive transfusion.
Materials and Methods: We performed a retrospective study of recipients who underwent liver transplant from 2012 to 2015. Patients considered at risk for massive transfusion received up to two 20 μg/kg doses of recombinant human activated factor VIIa, with rescue use permitted for other patients. We used propensity matching to determine the average treatment effects on patients who received recombinant human activated factor VIIa prophylactically to prevent massive transfusion. We determined thromboembolic events from medical record review.
Results: Of 234 liver transplant recipients, 38 received prophylactic and 2 received rescue recombinant human activated factor VIIa. We used a prediction model to readily identify those who would receive prophylactic recombinant human activated factor VIIa (C statistic = 0.885; 95% CI, 0.835-0.935). Propensity matching achieved balance, particularly for massive transfusion. Twenty-three of 38 patients (60.5%) who received recombinant human activated factor VIIa and 47 of 76 matched controls (61.8%) experienced massive transfusion. The coefficient for the average treatment effect of prophylactic administration was -0.013 (95% CI, -0.260 to 0.233; P = .92). The cohorts exhibited no difference in number of thromboembolic events (P > .99), although fatal events occurred in 1 patient who had prophylactic and 1 patient who had rescue recombinant human activated factor VIIa.
Conclusions: Prophylactic recombinant human activated factor VIIa use in patients at elevated risk of massive transfusion did not affect incidence of massive transfusion and was not associated with an increase in thromboembolic events overall. The lack of clinical benefit and the potential for fatal throm-boembolic events observed with recombinant human activated factor VIIa precluded its prophylactic use in liver transplant recipients.
Key words : Bleeding, rFVIIa administration, Thromboembolic events
Orthotopic liver transplant (LT) is frequently associated with extensive blood loss that requires patients to undergo massive transfusion (MT).1 In recent years, blood loss has been reduced through modifications of surgical technique, improved graft preservation, and appropriate supplementation of coagulation factors and platelets.2-4 Recombinant human factor VIIa (rFVIIa), which can improve hemostasis,5 has been proposed as a possible treatment to mitigate blood loss during LT.6 However, it continues to be expensive,7 its clinical effectiveness remains undetermined,8 and its administration may increase the risk of thromboembolic complications.9,10
The literature regarding rFVIIa administration during LT is inconclusive because only a few small trials have been conducted.2,8,11 Some have suggested that prophylactic rFVIIa administration would be particularly useful in those at greatest risk of extensive blood loss,12 but existing studies have not adequately addressed this question. Additionally, the potential risks of thromboembolic events associated with rFVIIa administration remain a concern.9,10
To better appreciate the risks and benefits of rFVIIa administration during LT, particularly for those thought to be at high risk for MT, we analyzed a 3-year period of a prospectively maintained LT database when rFVIIa was administered prophylactically to patients at elevated risk of MT. Because treatment assignment in this cohort was not randomized and could confound any benefit of prophylactic rFVIIa administration, we performed propensity matching to achieve greater balance. In addition, we determined the associated thromboembolic events by review of medical records. In this article, we investigated whether administration of rFVIIa to LT patients at high risk of MT reduces that risk.
Materials and Methods
The Johns Hopkins Medicine institutional review board approved this study and waived the requirement for written informed consent.
All patients were identified through the LT database, which is maintained prospectively on site as part of the Scientific Registry of Transplant Recipients. We included patients undergoing simultaneous liver and kidney transplantation (SLK) as their perioperative course is more similar to LT than to kidney transplant.13 All transplants were from deceased donors.
Our patient population was previously described.14,15 We extracted data from the blood bank database, the laboratory medicine database, the hospital billing database, the anesthesia information management system (Metavision; iMDsoft), and the electronic patient records from Sunrise (Altera Digital Health) and Epic (Epic Systems Corporation). We extracted all data from the electronic medical records, which were stored and analyzed using the SAFE (Secure Analytic Framework Environment) Desktop and its associated tools, including Microsoft Excel and Stata, in compliance with the data management requirements set by the Johns Hopkins Medical Institute institutional review board. No separate data control group was created. We included all patients who received LTs between June 6, 2012, and August 1, 2015, when low-dose rFVIIa was administered prophylactically to patients at high risk of MT.
We collected demographic data, including the calculated Model for End-Stage Liver Disease (MELD) and MELD-sodium (MELD-Na) score at the time of transplant; presence of ascites, varices, and variceal bleeding; presence of hepatocellular carcinoma; operative characteristics of retransplant or SLK transplant; preoperative laboratory values, including hemoglobin concentration, platelet concentration, international normalized ratio (INR), fibrinogen, and ionized calcium; kaolin-activated thromboelastogram (TEG) values (R interval, K interval, angle, maximum amplitude, and lysis at 30 min); and creatinine level, all measured before intraoperative administration of blood products. The maximum MELD and MELD-Na scores were capped at 40, and all status 1 patients were assigned MELD scores of 40. To further describe liver disease, we used the 5-stage classification of cirrhosis, which improves prediction of liver-related mortality in patients with low MELD and describes complications of portal hypertension16,17: (1) no varices or ascites, (2) varices only, (3) variceal bleeding, (4) ascites with or without varices, and (5) ascites with variceal bleeding. For the purpose of classification, any mention of varices or variceal bleeding in the patient record was considered positive for these indices. Presence of ascites was defined by computed tomography to confirm large-volume ascites, ascites requiring serial therapeutic paracentesis preoperatively, or intraoperative evacuation of ≥3 L of ascitic fluid. Thromboelastogram was performed with the thromboelastograph hemostasis analyzer system (Haemonetics).
The most common surgical approach involves hepatectomy with caval preservation, partial clamping of the inferior vena cava, and the “piggy-back” anastomosis technique, with occasional complete clamping of the inferior vena cava.18 Veno-venous bypass and portosystemic shunts are not routinely used at our institution.
After induction of general anesthesia and tracheal intubation, we place arterial and central venous catheters. Transesophageal echocardiography or pulmonary arterial catheters are used at the discretion of the attending anesthesiologist. Anesthesia is maintained with volatile agents in an oxygen-air mixture and supplemented with intravenous analgesics and muscle relaxants. Norepinephrine and occasionally vasopressin are used for blood pressure support. When necessary, epinephrine is used for inotropic support. A combination of body and fluid warmers are used to maintain core temperature at 36 to 37 ºC.
Arterial blood gases, sodium, potassium, ionized calcium, glucose, lactate, and hemoglobin and platelet concentrations, INR, activated partial thromboplastin time, fibrinogen, and TEG are followed on an hourly or every-other-hour basis. During the anhepatic phase, these laboratory values are determined every 30 minutes. A rapid infuser system (Belmont Instrument Corp) is used for blood product administration. Patients are extubated at the conclusion of the procedure at the discretion of the attending anesthesiologist if appropriate clinical criteria are met. All patients recover in the surgical intensive care unit immediately after surgery.
To optimize patient blood management, we developed a standardized approach to the transfusion of blood products that depends on the extent of ongoing hemorrhage guided by the Johns Hopkins Hospital interdisciplinary policy PAT064. We utilize a restrictive fluid strategy with a central venous pressure goal of 5 to 10 mm Hg for all patients. Transfusion of packed red blood cells (pRBCs) is guided by the patient’s hemodynamic state, hemoglobin concentration, and rate of bleeding. In general, a hemoglobin concentration of 7 g/dL is used as a transfusion trigger; however, if the patient is actively hemorrhaging, a hemoglobin concentration of 8 g/dL prompts transfusion of pRBCs. For patients who have slow ongoing bleeding and do not require MT, laboratory values of platelet concentration (less than 50 × 109/dL) and ongoing bleeding serve as a guide for the transfusion of fresh frozen plasma (FFP) and platelets. In general, FFP is transfused in a 1:1 ratio with pRBCs because we are unable to obtain real-time laboratory data. Apheresis platelets are transfused in a 6:5:1 ratio of pRBCs:FFP:platelets. Apheresis platelet preparations are equivalent to approximately 6 U of whole blood-derived platelets suspended in 1 U of plasma. As a result, the ratio of pRBCs-to-FFP-to-platelets of 6:5:1 is equivalent to a 1:1:1 ratio at our institution. Regardless of the rate of hemorrhage, cryoprecipitate is transfused when the fibrinogen concentration is ≤100 to 150 mg/dL in the presence of an abnormal angle on the TEG. When MT is required, transfusion ratios are further adjusted based on TEG values. Independent of hemorrhage severity, tranexamic acid is administered at 10 mg/kg over 10 minutes in a bolus dose and then infused at 1 mg/kg/h until the end of the case if fibrinolysis is evident on the TEG. Intraoperative cell salvage is used routinely for all patients without malignancy, and 1 unit of FFP is given for each 500 mL of salvaged blood. Factor concentrates are not routinely used at our institution.
Factor VIIa administration
From June 6, 2012, until August 1, 2015, low-dose rFVIIa was administered prophylactically at the discretion of both the transplant anesthesiologist and the transplant surgeon to patients with a MELD score >20 and a preoperative INR ≥2 based on a previous report that used the same parameters to assess risk in patients undergoing LT.12 In such cases, 20 μg/kg rFVIIa was administered within 30 minutes of the incision and repeated once 2 hours later. Low-dose rFVIIa was chosen to minimize the potential for thromboembolic complications.9,10 During the same study period, rFVIIa could be administered as a rescue medication to patients who did not initially receive prophylactic rFVIIa but in whom rapid hemorrhage became a concern. Rescue use of rFVIIa consisted of one dose of 40 to 80 μg/kg. Coagulation abnormalities were normalized before adminis-tration of prophylactic rFVIIa.
Our primary intraoperative outcome was whether the recipient experienced MT, defined as an intraoperative transfusion of more than 10 U of pRBCs obtained from the blood bank. Salvaged RBCs were not included in the calculation. Postoperatively, thromboembolic events during the initial admission were determined and included clinically identifiable graft venous or arterial thrombosis, pulmonary embolism, deep vein thrombosis, and other throm-boembolic events.
We initially used Fisher exact test to evaluate the association between prophylactic rFVIIa admi-nistration and MT. We then used logistic regression to develop a prediction model for prophylactic rFVIIa administration. We also determined the receiver operating characteristic curve and related C statistic for this model. The variables from this model were combined with any additional variables previously identified to be associated with MT in a prior study.14,15
We then used these variables to perform 2:1 propensity matching with replacement, wherein each subject receiving prophylactic rFVIIa was matched with 2 control subjects, where replacement was permitted. Thus, duplicates could appear in the matched controls. We then used these data to determine the potential benefit of rFVIIa administration in preventing MT for those who did receive rFVIIa and the average treatment effect on the treated (ATET). Next, we determined the absolute standardized mean differences for those who did and did not receive rFVIIa appropriate for the continuous and dichotomous variables for the original and propensity-matched data to better characterize the quality of the propensity match. Finally, we performed a confirmatory analysis using inverse probability weighting with regression adjustment, where the previously identified variables associated with MT and prophylactic rFVIIa use were input separately into the model.
All dichotomous variables are reported as a frequency (%) and analyzed with Fisher exact test. All continuous variables are reported as a median (interquartile range [IQR]) and analyzed with the Mann-Whitney U test or linear regression as appropriate. Differences were considered statistically significant at P < .05, and all reported significance values are from 2-sided tests. We used Stata 16.1 (StataCorp) for analyses.
The patients in our liver transplant database were described previously.14,15 From these, we identified a cohort of 234 who underwent transplant between June 6, 2012, and August 1, 2015, when rFVIIa was administered prophylactically to those identified by the transplant team as high risk for MT. Within this cohort, 38 patients received prophylactic rFVIIa treatment. During the study period, 2 participants who did not receive prophylactic rFVIIa subsequently received a rescue dose. We included these patients in the portion of the cohort that did not receive prophylactic rFVIIa. Among the 196 patients who did not receive prophylactic rFVIIa, 52 (26.5%) experienced MT. Among the 38 patients who received prophylactic rFVIIa, 23 (60.5%) experienced MT (Table 1). Compared with patients who did not receive prophylactic rFVIIa, patients who received prophylactic rFVIIa had higher MELD and MELD-Na scores, higher INR, and higher serum creatinine concentration but lower hemoglobin, platelet, and fibrinogen concentrations and more advanced cirrhosis stage (Table 1). On TEG, the group that received rFVIIa had a prolonged K time and lower angle and maximum amplitude, consistent with a hypocoagulable state (Table 1).
Overall, before propensity matching, those who ultimately experienced MT were more likely to have received prophylactic rFVIIa (P < .001). In anti-cipation of performing propensity matching, we developed a prediction model of prophylactic rFVIIa administration that ultimately included cirrhosis stage, SLK, MELD score, hepatocellular carcinoma, retransplantation, hemoglobin concentration, platelet concentration, and TEG R time. As indicated by C statistics of 0.885 (95% CI, 0.835-0.935) for the associated receiver operating curve (Figure 1), we found that the prediction model of prophylactic rFVIIa administration discriminated well between the control and treatment groups. To determine whether prophylactic rFVIIa administration reduced MT, we performed 2:1 propensity matching using the variables from the prediction model for prophylactic rFVIIa administration. These variables were previously found to be useful for predicting MT.14,15 The propensity matching results from the 2:1 model are summarized in Table 2 along with the corresponding absolute standardized mean differences. The changes in the standardized mean differences achieved with the 2:1 propensity matching are summarized in Figure 2 for the relevant variables. In the 2:1 propensity matching model, the incidence of MT was 61.8% for the matched controls and 60.5% for those who received prophylactic rFVIIa. The coefficient for the ATET in the 2:1 propensity model was -0.013 (95% CI, -0.260 to 0.233; P = .92). This finding was confirmed with inverse probability weighted regression adjustment analysis where the ATET was 0.036 (95% CI, -0.186 to 0.258; P = .75).
The number of thromboembolic events for each arm of the cohort is summarized in Table 3. One patient who received prophylactic rFVIIa and one who received rescue rFVIIa experienced intraoperative catastrophic thromboembolic events that led to death. One patient who received a prophylactic dose developed intracardiac and intravascular thrombosis 90 minutes after administration of rFVIIa, as documented by a visualization of the clot on a transesophageal echo. The patient who received a rescue dose died from intraoperative cardiac arrest, and this was classified as a possible rFVIIa-associated event. The incidence of thromboembolic complica-tions, including pulmonary embolus/deep vein thrombosis, hepatic artery thrombosis, portal vein thrombosis, and other catastrophic complications, was 9.2% in the matched controls and 7.9% in those who received prophylactic rFVIIa (P > .99). When those who did or did not receive rFVIIa for any reason, prophylactic or rescue, were considered, 2 of 40 (5%) of those who received rFVIIa died from catastrophic thromboembolic complications, whereas 0 of 194 patients who did not receive rFVIIa experienced this outcome (P = .029).
In this cohort from a prospectively maintained database of LT patients who were eligible to receive prophylactic rFVIIa, we found no benefit of prophylactic rFVIIa for prevention of MT in those thought to be at high risk of MT. Prophylactic rFVIIa was administered at the discretion of the surgical and anesthetic teams guided by the calculated MELD score and INR. Although the prediction model for prophylactic rFVIIa administration ultimately included the MELD score and preoperative measures of coagulation, it also showed that the decision to administer prophylactic rFVIIa over the period of data acquisition was more nuanced. Whereas those who were administered prophylactic rFVIIa did not experience a greater number of thromboembolic complications overall, those who received rFVIIa for any reason, prophylactic or rescue, were more likely to experience a fatal thromboembolic event. Overall, our study provided no evidence that prophylactic rFVIIa reduces MT in patients at elevated risk of MT. Furthermore, it may be that administration of rFVIIa for any reason, prophylactic or rescue, is associated with fatal thromboembolic events.
Our study builds upon results from previous studies of rFVIIa use to limit hemorrhage during LT,2,8,11 which were inconclusive. The earlier studies demonstrated no benefit,19,20 a decrease in requirements for pRBC transfusion,21 or a reduction in the proportion of patients receiving pRBC.22 In some studies, when prophylactic rFVIIa was administered only to patients thought to be at greater risk of bleeding, a reduction in pRBCs transfusion was reported.12,23 In a study of pediatric patients who were given rFVIIa when the surgeons and anesthesiologists agreed that the risk of hemorrhage was elevated, a return to “normal” hemostasis was speculated to have occurred.24 These earlier reports of rFVIIa administration to patients at greater risk of intraoperative bleeding during LT12,23 prompted the prophylactic rFVIIa use in our LT program. In these prior studies, risk stratification included MELD score >20 and INR >1.5 or 2.0, depending upon the study.12,23
Additionally, risk of thromboembolic events remains of concern. Because previous studies of rFVIIa in LT reported no differences in clinical effects across dose ranges, but thromboembolic event risk in all patient populations receiving medium dose (>40 and <120 μg/kg) and high-dose (≥120 μg/kg) rFVIIa was greater than that in those who received low-dose rFVIIa (≤40 μg/kg),9 we selected a lower dose of rFVIIa for prophylactic use in our practice. Consistent with prior studies of prophylactic rFVIIa use during LT, in which thrombotic event rates were reported to be in the range of 0% to 20%,9,11 we did not observe an increased risk of thromboembolic events in our patients who received prophylactic rFVIIa. However, in our cohort, the possibility of a fatal thromboembolic event was increased with rFVIIa use for any reason, prophylactic or rescue. Inability to predict who is at risk for such catastrophic events may stem from the unknown status of the rebalanced coagulation system of patients with cirrhosis at the time of transplant and therefore the unpredictable effect of rFVIIa administration on risk of bleeding and thromboem-bolic events.25 The unpredictability of major thromboembolic events associated with rFVIIa use, both prophylactic and rescue, was sufficient to end its administration after the events described above.
As indicated above, an important feature of our study was the administration of prophylactic rFVIIa only to those thought to be at greatest risk of MT. Given any reasonably effective assessment of MT risk by the transplant team, in an observational study, the intervention would be expected to be inexorably linked to MT subsequently experienced by patients who received prophylactic rFVIIa, as we initially demonstrated. The high likelihood of MT in those administered prophylactic rFVIIa could potentially obscure benefits of prophylactic rFVIIa administration. Use of a propensity model to match those receiving prophylactic rFVIIa with those not given rFVIIa but of similar risk for MT reduces these concerns,26 as emphasized by the balance achieved by the propensity model, particularly with respect to MT.
The primary limitation of our study was that it was not a randomized controlled trial. Given its expense and the growing appreciation of the thromboembolic risks of rFVIIa administration, it may not be feasible or ethical to administer prophylactic rFVIIa to all patients presenting for LT. Moreover, because MT requiring more than 10 U of pRBCs occurs in only about 25% of patients who undergo LT, demonstrating a meaningful benefit of prophylactic rFVIIa administration might limit the efficiency of such a study even if it could be shown that blood loss is reduced by prophylactic rFVIIa administration. Although our and previous studies illustrate that it is possible to predict which patients are most likely to require MT,14,15 and those individuals could then be randomized to receive or not receive prophylactic rFVIIa, the growing appreciation of the risk of rFVIIa administration does raise ethical concerns about performing such a study.
Prophylactic administration of rFVIIa to patients with an elevated risk of MT had no impact on the incidence of MT. Prophylactic rFVIIa administration was not associated with an increase in thrombo-embolic events overall, but fatal catastrophic thromboembolic events were observed with rFVIIa use. The lack of clinical benefit and fatal catastrophic thromboembolic events observed with rFVIIa appear to preclude rFVIIa use in LT.
Volume : 20
Issue : 9
Pages : 817 - 825
DOI : 10.6002/ect.2022.0159
From the 1Johns Hopkins Krieger School of Arts and Sciences and the 2Department of Anesthesiology and Critical Care Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA; the 3University of Maryland, College Park, Maryland, USA; the 4Department of Anaesthesiology, Aga Khan University Medical College, Karachi, Pakistan; and the 5Department of Surgery, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
Acknowledgements: The authors have not received any funding or grants in support of the presented research or for the preparation of this work and have no declarations of potential conflicts of interest. The authors are grateful for the editorial assistance of Claire Levine, MS, ELS, Manager, Editorial Services, Department of Anesthesiology and Critical Care Medicine, Johns Hopkins University, Baltimore, MD, USA.
Author contributions: M. Yang and A. Gottschalk participated in research design, writing of the paper, and data analysis. P. Ariyo, B. Perlstein, A. Latif, S. M. Frank, W. T. Merritt, A. Cameron, and B. Philosophe participated in the performance of the research. A. Pustavoitau participated in research design, performance of the research, writing of the paper, and data analysis.
Corresponding author: Muyue Yang, Johns Hopkins Krieger School of Arts and Sciences, Baltimore, MD, USA
Table 1. Characteristics of the 234 Study Patients During the Recombinant Human Factor VII Era
Figure 1. Receiver Operating Characteristic Curve of the Prediction Model for Prophylactic Recombinant Human Activated Factor VIIa Administration
Figure 2. Success of 2:1 Propensity Matching in Achieving Balance in the Data for Both Massive Transfusion and the Variables Used to Predict Prophylactic Recombinant Human Activated Factor VIIa Administration
Table 2. Characteristics of Patients Who Did and Did Not Receive Prophylactic Recombinant Human Factor VII After 2:1 Propensity Matching
Table 3. Summary of Thromboembolic Complications