Treatment of refractory bleeding after cardiac operations with low-dose recombinant activated factor VII (NovoSeven ®): a propensity score analysis

Treatment of refractory bleeding after cardiac operations with low-doserecombinant activated factor VII (NovoSevenW): a propensity score analysis

Sandro Gelsomino a,*, Roberto Lorusso b, Stefano Romagnoli c, Sergio Bevilacqua a,Giuseppe De Cicco b, Giuseppe Bille c, Pierluigi Stefano c, Gian Franco Gensini d

a Experimental Surgery Unit, Department of Heart and Vessels, Careggi Hospital, Florence, ItalybCardiac Surgery Unit, Civic Hospital, Brescia, Italy

cCardiac Surgery Unit, Department of Heart and Vessels, Careggi Hospital, Florence, ItalydUniversity of Florence, Florence, Italy

Received 19 July 2007; received in revised form 6 October 2007; accepted 9 October 2007; Available online 8 November 2007

Abstract

Background: Recombinant activated factor VII (rFVIIa) has been increasingly used to stop life-threatening bleeding following cardiacoperations. Nonetheless, the issue of dosing, given the expense and potential for thrombotic complications, is still of major concern. Wereport our experience with small-dose rFVIIa in patients with refractory bleeding after cardiac surgery. Methods and results: From September2005 to June 2007, 40 patients (mean age 70.1 � 9.2 years, 52.5 males) received a low dose of rFVIIa (median: 18 mg/kg, interquartile range: 9—16 mg/kg) for refractory bleeding after cardiac surgery. Forty propensity score-based greedymatched controls were compared to the study group.Low dose of rFVIIa significantly reduced the 24-h blood loss: 1610 ml [ 1285—1800 ml] versus 3171 ml [2725—3760 ml] in the study and controlgroups, respectively (p < 0.001). Thus, hourly bleeding was 51.1 ml [34.7—65.4 ml] in patients receiving rFVIIa and 196.2 ml/h [142.1—202.9 ml]in controls (p < 0.001). Furthermore, patients receiving rFVIIa showed a lower length of stay in the intensive care unit (p < 0.001) and shortermechanical ventilation time (p < 0.001). In addition, the use of rFVIIa was associated with reduction of transfusion requirements of red bloodcells, fresh frozen plasma and platelets (all, p < 0.001). Finally, treated patients showed improved hemostasis with rapid normalization ofcoagulation variables (partial thromboplastin time, international normalized ratio, platelet count, p < 0.001). In contrast, activated pro-thrombin time and fibrinogen did not differ between groups ( p = ns). No thromboembolic-related event was detected in our cohort. Conclusions:In our experience low-dose rFVIIa was associated with reduced blood loss, improvement of coagulation variables and decreased need fortransfusions. Our findings need to be confirmed by further larger studies.# 2007 European Association for Cardio-Thoracic Surgery. Published by Elsevier B.V. All rights reserved.

www.elsevier.com/locate/ejctsEuropean Journal of Cardio-thoracic Surgery 33 (2008) 64—71

Keywords: Hemorrhage; Plasma; Platelet-derived factors; Platelets

1. Introduction

There has recently been a great deal of interest in the useof recombinant activated factor VII (rFVIIa, NovoSeven,NovoNordisk, Bagsvaerd, Denmark), for control of refractorybleeding after cardiac operations [1]. Recombinant activatedfactor VII was licensed in 1999 by the United States Food andDrug Administration (FDA) for its use in the management ofpatients with hemophilia with inhibitors to factor VIII or IX[2]. In 2005 it was further approved for surgical procedures inhemophiliacs and for patients with factor VII deficiency [3].Then, the off-label use of this agent was reported in patientswith life-threatening bleeding after trauma or majorabdominal surgery [4]. However, the efficacy in reducing

* Corresponding author. Address: Experimental Surgery Unit, Department ofHeart and Vessels, Careggi Hospital, Viale Morgagni 85, 50134 Florence, Italy.Tel.: +39 055 794 7467; fax: +39 055 794 7628.

E-mail address: [email protected] (S. Gelsomino).

1010-7940/$ — see front matter # 2007 European Association for Cardio-Thoracic Sdoi:10.1016/j.ejcts.2007.10.004

blood loss and transfusion requirements in patients under-going cardiac surgery is debated yet. Furthermore, the issueof dosing (given the expense and potential for thromboticcomplications) is still of major concern.

We report our experience with small-dose rFVIIa used asrescue therapy in patients with refractory bleeding aftercardiac surgery.

2. Materials and methods

2.1. Patients and definitions

Starting from September 2005 to June 2007, 2398consecutive patients underwent cardiac surgery at ourinstitution (Cardiac Surgery, Careggi Hospital, Florence,IT). Of them, 298 (12.4%) suffered from significant bleedingdefined as bleeding that compromises hemodynamics and/orbleeding �500 ml/h during the first postoperative hour, or

urgery. Published by Elsevier B.V. All rights reserved.

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—71 65

bleeding �300 ml/h for three consecutive hours after chestclosure, or bleeding�1200 ml/h after the fifth postoperativehour [5]. Data were collected by research assistants whowere blinded to the nature of the study and were analyzed bya surgeon (GB) and an anesthesiologist (SB) who identified228 patients with refractory bleeding (Fig. 1). Blood loss wasconsidered refractory when a surgical bleeding source wasexcluded at surgical re-exploration after a completetransfusion protocol in patients who received a secondcomplete transfusion protocol.

Among patients with refractory bleeding, 17.5% (n = 40)were treated with small-dose rFVIIa as rescue therapy. Thefinal decision to employ rFVIIa was made by the attendinganesthesiologist and it was based on a risk/benefit evaluationand/or his/her personal preferences. No formal exclusioncriteria were applied.

A small-dose (1.2 mg) rFVIIa was given as a slowintravenous bolus. This dose was chosen empirically afterexamination of published reports [1]. Doses ranged from 10.9to 19.3 mg/kg (median: 18 mg/kg; interquartile range: 9—16 mg/kg). The infusion was repeated in case of persistentbleeding defined as bleeding �150 ml/h for three consecu-tive hours after the treatment. If bleeding persisted patientsunderwent further surgical re-exploration.

Fig. 1. Study design.

2.2. Institutional protocol

Before the initiation of cardiopulmonary bypass (CPB),all patients received porcine heparin at an initial dose of300 U/kg, injected IV before cannulation of the aorta. Anadditional dose of 5000 U of heparin is administered whenthe kaolin activated clotting time (ACT) is �400 s [6]. Anadditional dose of 2500 U was given if the ACT declinedabove 300 s. After complete weaning from CPB heparin isneutralized by endovenous infusion of protamine hydro-chloride at the dose of 0.6 mg per 100 U of heparinadministered [6]. Heparin neutralization is consideredadequate if post-protamine ACT value is within 10% ofthe pre-heparin value.

Red blood cells (RBCs) are given to maintain thehemoglobin concentration �7 g/dl during CPB. All patientsreceive antifibrinolytics (10 mg/kg bolus of Tranexamic acidfollowed by an infusion of 1 mg kg�1 hm+ for the duration ofthe procedure and, in case of reintervention or acute aorticdissection, 2.0 million KIU aprotinin before CPB followed by2.0 million KIU in CPB-prime and 500,000 KIU/h for theduration of the procedure).

All patients with refractory bleeding were operated on bythree surgeons. The intensive care unit (ICU) staff (physi-cians) was the same and anesthetic management wasidentical in the entire cohort. Surgery was performedthrough a median sternotomy on CPB with antegrade/retrograde cold blood cardioplegia. The lowest temperatureachieved was 34 8C (21 8C in patients who underwent deephypothermic circulatory arrest).

Subjects with significant bleeding undergo 1000 IU ofprothrombin complex concentrates (PCC) [7] associated tothe following transfusion protocol: [8,9]

- R

- 1

- O

2.3. End points

End points were measured in traded patients and controlsfollowing the second transfusional protocol (T0). Primaryoutcome was the effectiveness in terms of blood loss,transfusion requirements, changes in coagulation laboratoryfindings and clinical findings (death, complications, time ofmechanical ventilation [MV] and intensive care unit length ofstay [ICU LoS]). Blood loss was explored either as hourlybleeding or as 1-, 3-, 5-, 10-, 15-, 20-, or 24-h bleeding. Theneed for RBCs, FFP, and PLT transfusion in study patients afterthe administration of rFVIIa was compared with controls.Laboratory evaluation included partial thromboplastin time(PT), international normalized ratio (INR), activated pro-thrombin time (aPTT), fibrinogen (FBG) and platelet (PLT)count.

Screening for thromboembolic events was performed byphysical examination. If a thromboembolic complication wassuspected, color Doppler sonography, transesophageal echo-cardiography, computed tomography scan and laboratorytests were performed to confirm the diagnosis.

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—7166

Table 1Prematching characteristics

Study Controls p

Age (years) 70.1 � 9.2 75.8 � 9.8 0.052Male 21 (52.5) 95 (41.6) 0.2BSA 1.7 � 0.1 1.7 � 0.2 >0.9NYHA class 3 [3—4] 3[3—4] 0.6Atrial fibrillation 5 (12.5) 44 (16.0) 0.3Congestive heart failure 2 (5.0) 3 (1.3) 0.09Endocarditis 5 (12.5) 11 (4.8) 0.07LVEF � 35% 5 (12.5) 21 (9.2) 0.5

EuroScoreAdditive 8.3 [5—10] 10.4 [9—14] 0.3Logistic 15 [11—19] 18 [15—20] 0.2

Hypertensiona 8 (20.0) 57 (25.0) 0.2Diabetesb 7 (17.5) 51 (22.3) 0.2Chronic obstructivepulmonary diseasec

6 (15.0) 44 (19.2) 0.6

Renal insufficiencyd 3 (7.5) 2 (0.8) 0.02Creatinine (mg/dl) 1.2 [1.1—1.4] 1.1 [1.0—1.2] 0.4Cerebral vascular disease 2 (5.0) 10 (4.3) 0.5Peripheral vascular disease 1 (2.5) 11 (4.8) 0.9

Baseline coagulation findingsPT 86 [75—94] 109 [80—125] 0.03INR 1.2 [0.9—1.4] 1.0 [0.6—1.3] 0.04aPTT 46 [41—54] 25 [22—36] 0.02FBG 375 [322—397] 410 [335—411] 0.07PLT 171 [140—191] 182 [150—196] 0.1

Baseline Hb 12 [10—13] 13 [12—14] 0.051Antiplatelet therapy 7 (17.5) 45 (19.7) 0.7Anticoagulant therapy 5 (12.5) 21 (9.2) 0.5Urgency/emergency 7 (17.5) 41 (18.0) 0.9Redo surgery 4 (10.0) 15 (6.5) 0.4

Surgical proceduresCABG 17 (42.5) 114 (50.0)Aortic valve surgery 4 (10.0) 11 (4.8) 0.5MV surgery 4 (10.0) 13 (5.8)CABG + valve surgery 4 (10.0) 30 (13.1)Multiple valve 2 (5.0) 19 (8.3)Ascending aorta replacement 5 (12.5) 23 (10.1)Aortic arch replacement 4 (10.0) 18 (7.9)

Complex surgery 15 (37.5) 110 (48.2) 0.2CPB time 140 [80—188] 175 [120—210] 0.06CCL time 80 [60—120] 91 [80—151] 0.1DHCA 6 (15.0) 28 (12.2) >0.9DHCA time 40 [18—38] 34 [16—77] 0.6ACP 5 (12.5) 33 (14.4) 0.7ACP time 20 [11—41] 18 [15—34] 0.8Nadir Hct on CPB 22 [18—24] 20 [16—20] 0.02Tranexamic acid 40 (100) 228 (100) N/AAprotinin 9 (22.5) 51 (22.3) >0.9Bleedinge 1685

[1590—1770]1290[970—1510]

0.1

Transfusions in ORe

RBC 5 [1—4] 2 [0—3] 0.01PLT 2 [0—3] 4 [2—7] 0.04FFP 6 [2—8] 3 [1—4] 0.03

Transfusions in ICUe

RBC 9 [4—12] 2 [1—7] <0.001PTL 2 [1—4] 4 [2—6] 0.06FFP 4 [2—7] 7 [4—10] 0.051PCC (1000 IU) 40 (100) 228 (100) N/A

Continuous variables are presented as means � 1 SD; discrete variables arepresented as percentage (parentheses). Non-parametric data are shown asmedians [interquartile range].Abbreviations: ns: not significant; N/A: not applicable; BSA: body surface area(kg/m2); NYHA: New York Heart Association functional class; LVEF: left ven-tricular ejection fraction; PT: prothrombin time (%); INR: international normal-

2.4. Ethical issues

Following the World Medical Association guidelinesconcerning ethical principles for medical research involvinghuman subjects [11], the Institutional Ethics Board approvedthe study. Our institution approved the use of rFVIIa in thesetting of life-threatening post-cardiac surgery hemorrhage.Specific patient or patient family consent of this use was notobtained because of the emergency nature of the circum-stance, although all patients had given informed consent tothe procedure and they were aware that hemorrhage was apossible and significant complication. Nonetheless, ethicsboard approval was obtained to review records of patientsreceiving rFVIIa as well as the records of all patientsundergoing cardiac surgery to identify controls.

The authors had full access to the data and takeresponsibility for its integrity. All authors have read andagreed to the manuscript as written.

2.5. Statistical analysis

Pre-matched baseline characteristics of patients withrefractory bleeding are shown in Table 1. Data werecompared with Pearson chi-square and Wilcoxon rank-sumtest for categorical and continuous variables, respectively.

To allow an unbiased comparison between patients withrefractory bleeding receiving rFVIIa and controls, thepropensity score analysis was employed [10].

The logistic procedure allowed us to calculate thepredicted probability (propensity score) of receiving therFVIIa [11]. Variables considered for inclusion in this modelincluded: endocarditis, congestive heart failure, baselinehemoglobin, baseline platelets count, baseline PT, baselineINR, baseline aPTT, baseline FBG, baseline creatinine,antiplatelet therapy, anticoagulation therapy, urgency/emer-gency surgery, redo surgery, complex operation (ascendingaorta and/or arch operation, multiple valve or associatedvalve and coronary artery bypass grafting procedures), CPBtime, hypothermic circulatory arrest (HCA time), bleeding atT0, Nadir hematocrit (Hct) during CPB, use of aprotinin(tranexamic acid was received by all patients) and bloodproducts transfused either in the intensive care unit (ICU) or inthe operative room (OR) before T0. Themodel’s reliability andits predictiveabilitywere testedwithHosmer—Lemeshow (HL)goodness-of-fit test and the C-index, respectively [12]. Thismodel did not show evidence of lack of fit based on the HLstatistic (p = 0.27) and confirmed high discriminative ability(C-test = 0.94).

ized ratio; aPTT: activated prothrombin time (s); FBG: fibrinogen (mg/dl); PLT:platelets count (n � 109/l); Hb: hemoglobin (g/dl); CABG: coronary arterybypass grafting; AVR: aortic valve replacement; MV: mitral valve; CPB: car-diopulmonary bypass; CCL: cross-clamp; DHCA: deep hypothermic circulatoryarrest; ACP: antegrade cerebral perfusion; Hct: hematocrit; OR: operativeroom; ICU: intensive care unit; RBC: red blood cells; FFP: fresh frozen plasma.

a Blood pressure>140/90 or a history of high blood pressure or the need forantihypertensive medications.

b Symptoms of diabetes mellitus plus casual plasma glucose concentration�200 mg/dl or fasting plasma glucose level �126 mg/dl or 2-h postprandialplasma glucose �200 mg/dl after a 75-g glucose load.

c FEV1 < 40%.d Serum creatinine >2-mg/dl compared with normal 0.6 to 1.2 mg/dl.e At T0.

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—71 67

Table 2Greedy 5! 1 digit match

Study Matched controls p

Age (years) 70.1 � 9.2 73.2 � 7.8 0.09Male 21 (52.5) 15 (37.5) 0.051BSA 1.7 � 0.1 1.7 � 0.2 >0.9NYHA class 3 [3—4] 3[3—4] 0.6Atrial fibrillation 5 (12.5) 6 (15.0) >0.9Congestive heart failure 2 (5.0) 1 (2.5) >0.9Endocarditis 5 (12.5) 6 (15.0) >0.9LVEF � 35% 5 (12.5) 11 (27.5) 0.16

EuroScoreAdditive 8.3 [5—10] 9 [6—11] 0.08Logistic 15 [11—19] 16 [10—18] 0.1

Hypertensiona 8 (20.0) 10 (25.0) 0.78Diabetesb 7 (17.5) 3 (7.5) 0.1Chronic obstructivepulmonaryc disease

6 (15.0) 3 (7.5) 0.1

Renal insufficiencyd 3 (7.5) 2 (5.0) >0.9Creatinine (mg/dl) 1.2 [1.1—1.4] 1.1 [1.0—1.4] 0.7Cerebral vascular disease 2 (5.0) 5 (12.5) 0.43Peripheral vascular disease 1 (2.5) 4 (10.0) 0.35

Baseline coagulation findingsPT 86 [75—94] 88 [70—98] 0.73INR 1.2 [0.9—1.4] 1.2 [0.7—13] 0.2apt 46 [41—54] 32 [25—37] 0.051FBG 375 [322—397] 381 [335—411] 0.8PLT 171 [140—191] 189 [153—196] 0.09

Baseline Hb 12 [11—13] 12 [11—13] 0.8Antiplatelet therapy 7 (17.5) 8 (20.0) >0.9Anticoagulant therapy 5 (12.5) 4 (10.0) >0.9Urgency/emergency 7 (17.5) 11 (27.5) 0.3Redo surgery 4 (10.0) 1 (2.5) 0.35

Surgical proceduresCABG 17 (42.5) 19 (47.5)Aortic valve surgery 4 (10.0) 2 (5.0) 0.65MV surgery 4 (10.0) 2 (5.0)CABG + valve surgery 4 (10.0) 6 (15.0)Multiple valve 2 (5.0) 3 (7.5)Ascending aorta replacement 5 (12.5) 5 (12.5)Aortic arch replacement 4 (10.0) 3 (7.5)

Complex surgery 15 (37.5) 17 (42.5) 0.81CPB time 140 [80—188] 170 [110—192] 0.06CCL time 80 [60—120] 97 [84—163] 0.053DHCA 6 (15.0) 5 (12.5) >0.9DHCA time 40 [18—38] 38 [20—88] 0.8ACP 5 (12.5) 5 (12.5) >0.9ACP time 20 [11—41] 16 [15—44] 0.7Nadir Hct on CPB 22 [18—24] 22 [18—24] >0.9Tranexamic acid 40 (100%) 40 (100%) N/AAprotinin 9 (22.5) 8 (20.0) >0.9Bleedinge 1685

[1590—1770]1570[1110—2250]

0.3

Transfusions in ORe

RBC 5 [1—4] 4 [2—3] 0.2PLT 2 [0—3] 2 [1—3] 0.8FFP 6 [2—8] 5 [3—7] 0.08

Transfusions in ICUe

RBC 9 [4—12] 5 [2—14] 0.09PTL 2 [1—4] 3 [2—5] 0.1FFP 4 [2—7] 5 [3—9] 0.23PCC (1000 IU) 40 (100) 40 (100) N/A

Continuous variables are presented as means � 1 SD; discrete variables arepresented as percentage (parentheses). Non-parametric data are shown asmedians [interquartile range].Abbreviations: ns: not significant; N/A: not applicable; BSA: body surface area(kg/m2); NYHA: New York Heart Association functional class; LVEF: left ven-tricular ejection fraction; PT: prothrombin time (%); INR: international normal-

The SAS Greedy 5! 1digit match macro was used toidentify a matched control for each treated patient [13]according to their propensity score.1

Adequacy of covariate balance in the matched sample wasassessed with Mc Nemar or Wilcoxon’s signed rank test, whichwere also employed to assess the differences betweentreated patients and matched controls for outcome andadverse events as well as for pre—post comparison withingroups.

Coagulation measures were presented as median andinterquartile range. The propensity score-based greedymatching algorithm successfully matched all the 40 treatedpatients. Adequacy of covariate balance in the matchedsample is shown in Table 2.

A sensitivity analysis was performed to assess thesensitivity of the findings of this case—control study to theassumption about unmeasured confounders using the follow-ing formula [14]:

R ¼ R�A; A ¼ G 1P1 þ ð1� P1Þ

G 0P0 þ ð1� P0Þ; G 0 ¼ G 1 ¼ G

where R and R* are, respectively, the apparent and truerelative risks of bleeding associated with the treatment. Ais the adjustment factor. G0 and G1 are the relative risksassociated with the unmeasured confounder and P0 and P1the prevalences of the unmeasured confounder in treatedversus control patients, respectively.

In Appendix 1 the point and the interval estimates for theodds ratio of interest after adjustment for an unmeasuredconfounder are displayed.

Based on 24-h blood loss as a primary outcome, this studyhad 80% power to detect a 10 ml difference in 24-h bleeding.

We considered a p-value of <0.05 to be statisticallysignificant. Analyses were performed using SAS StatisticalPackage, release 9.1 (SAS Institute, Cary, NC).

3. Results

3.1. Blood loss and transfusions requirement

Blood loss after rVIIa was significantly lower than beforethe therapy ( p < 0.001). Median 24-h blood loss (Fig. 2A) was1610 ml [interquartile range: 1285—1800 ml] and 3171 ml[2725—3760 ml] in the study and control groups, respectively

ized ratio; aPTT: activated prothrombin time (s); FBG: fibrinogen (mg/dl); PLT:platelets count (n � 109/l); Hb: hemoglobin (g/dl); CABG: coronary arterybypass grafting; CPB: cardiopulmonary bypass; CCL: cross-clamp; DHCA: deephypothermic circulatory arrest; ACP: antegrade cerebral perfusion. Hct:hematocrit; OR: operative room; ICU: intensive care unit; RBC: red bloodcells; FFP: fresh frozen plasma.

a Blood pressure>140/90 or a history of high blood pressure or the need forantihypertensive medications.

b Symptoms of diabetes mellitus plus casual plasma glucose concentration�200 mg/dl or fasting plasma glucose level �126 mg/dl or 2-h postprandialplasma glucose �200 mg/dl after a 75-g glucose load.

c FEV1 < 40%.d Serum creatinine >2 mg/dl compared with normal 0.6—1.2 mg/dl.e Bleeding at T0 (ml).

1 http://www2.sas.com/proceedings, last access 20/06/2007.

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—7168

Fig. 3. Units of red blood cells (A), fresh frozen plasma (B) and platelets (C)transfused in study and control groups starting from T0 (see text). Data arepresented as median [interquartile range].

Fig. 2. (A) Blood loss in study and control groups starting from T0 (see text).Data are presented as medians (dotted lines) and 25th and 75th percentiles(solid lines). (B) Blood loss/h. Data are presented as median [interquartilerange]. *Significance versus study group.

( p < 0.001). Thus, hourly bleeding was 51.1 ml [34.7—65.4 ml] in patients receiving rFVIIa and 196.2 ml/h[142.1—202.9 ml] in controls ( p < 0.001). A further compar-ison of the degree of bleeding in the two groups (Fig. 2B)showed that it was reduced significantly in the study group at3 h ( p < 0.001), 15 h ( p < 0.001), 20 h ( p = 0.02), and 24 h,remaining unchanged at 5 h and 10 h ( p = ns). In contrast, incontrols, blood loss did not significantly change at 3, 5, 10,15, 20, and 24 h and it was, at all times, constantly higherthan in the study group ( p < 0.001). The result of sensitivityanalysis (Appendix 1) strongly support the conclusion thatrVIIa decreased the risk of persistent bleeding.

Transfusion requirements for RBCs ( p < 0.001), FFP( p < 0.001) and PLTs ( p < 0.001) were lower in the studygroup patients than in controls (Fig. 3A—C).

Blood product usage after rVIIa therapy was significantlylower than before it ( p < 0.001).

3.2. Coagulation laboratory findings

Following the administration of rFVIIa, PT increasedsignificantly only in the study group ( p < 0.001, Table 3).The INR showed a significant reduction in both groups( p < 0.001 and p = 0.03 in the study and matched controls,respectively) and it was significantly lower in the study group

( p < 0.001). The aPTT improved in treated patients( p = 0.001) and controls ( p = 0.03) without significantdifference between the groups. Contrastingly, FBG increasedsignificantly only in patients receiving rFVIIa ( p < 0.001).Finally, the platelet count increased in the study group( p < 0.001) while it reduced significantly in controls( p < 0.001).

3.3. Clinical outcomes

Clinical outcomes are shown in Table 4. Death occurred in3 patients (7.5%) in the control group versus 2 (5.0%) in thestudy group ( p > 0.9). Complication rates were 20.0% (n = 8)and 17.5% (n = 7) in the study and control groups,respectively ( p > 0.9). Time of MV ( p < 0.001) as well as

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—71 69

Table 3Coagulation findings

Study Matched controls

T0 24 h T0 24 h

PT 66.5 [50.5—84] 131 [104—168] a 77.2 [41—73] 95.5 [99—119] b

INR 1.5 [1.2—1.9] 0.9 [0.2—1.4]a 1.5 [1.2—1.8] 1.2 [0.6—1.5]a, b

aPTT 59.3 [41.2—72] 39.1 [32.2—42.6] a 60.6 [38.2—73.4] 36.1 [40.2—66.5] a

FBG 193 [160—254] 250 [214—328] a 193.5 [157—269] 203 [217—346]b

PLT 86 [64.5—112] 115 [87.5—154] a 84 [66—102] 60.9 [37—92]a, b

Variables are presented as median [interquartile range].Abbreviations: PT: prothrombin time (%); INR: international normalized ratio; aPTT: activated prothrombin time (s); FBG: fibrinogen (mg/dl); PLT: platelets count(n � 109/l).

a Significance versus T0.b Significance versus study group.

Table 4Clinical outcome

Study group Matched controls p

ICU LoS (h) 151 [93—176] 443 [388—576] <0.001Time of MV (h) 13 [6—30] 73 [20—80] <0.001

Complications 8 (20) 7 (17.5) >0.9Stroke 2 (5.0) —Renal failure 1 (2.5) 2 (5.0)Respiratory failure 2 (5.0) 2 (5.0)Low-output syndrome 1 (2.5) 1 (2.5)Infection 2 (5.0) 2 (5)

Deaths >0.9Low-output syndrome 1 (2.5) 1 (2.5)Respiratory failure 1 (2.5) 1 (2.5)Multiorgan failure (MOF) — 1 (2.5)

Discrete variables are presented as percentage (parentheses). Non-parametricdata are shown as medians [interquartile range]. Abbreviations: ICU: intensivecare unit; LoS: length of stay; MV: mechanical ventilation (time).

ICU LoS ( p < 0.001) was significantly lower in patientsreceiving rFVIIa than in controls.

Thirty-five patients in the control group (87.5%) under-went a second surgical re-exploration for persistent bleedingand in only 2 cases was a surgical bleeding site identified. Incontrast, 3 patients (7.5%) receiving rFVIIa (x2 = 8.3,p < 0.001) underwent second surgical re-exploration dueto bleeding persisting after a second dose of VIIa. A surgicalsource of blood loss was identified in all these 3 patients.

Two treated patients had postoperative stroke; in both apredisposing factor for cerebrovascular accident was clearlyidentified: the first patient had preoperative transientischemic attacks followed by stroke occurring 8 monthsbefore surgery, resulting in complete functional recovery.The second underwent prolonged deep hypothermic circu-latory arrest with postoperative hypoperfusion resulting inmultiple ischemic injures shown by postoperative CTscan. Inthe remaining patients, the use of rFVIIa caused nothromboembolic complication at clinical examination,laboratory tests, and transesophageal echocardiography.

4. Comment

Undetermined effectiveness, safety and high cost are thereasons why rFVIIa is mainly used as the very last therapeutic

option in absence of response to transfusion and conventionalprocoagulant agents. To our knowledge, only 4 studies havereported the utilization of rFVIIa as primary therapy [15] or asa prophylactic hemostatic agent [16]. The aim of this studywas to determine the effectiveness of rFVIIa employed atvery low doses in reducing blood loss and transfusionrequirements after cardiac surgery. To assess the efficacyof VIIa recombinant, we compared the study group with 40propensity score-based greedy matched controls.

The optimal dosage of rFVIIa in cardiac surgery remainsunclear and satisfactory coagulation was yielded with dosesranging from 13 to 192 mg/kg [17]. Furthermore, recurrentinjections were sometimes necessary due to the short half-life of rFVIIa [18]. In the present study, a small dose of rVIIa(10.9—19.3 mg/kg [median: 18 mg/kg, interquartile range:9—16 mg/kg]) was employed as rescue therapy in life-threatening post-cardiac surgery refractory hemorrhage. Asingle dose of rFVIIa significantly reduced the postoperativebleeding in 92.5% of patients. Three patients showedpersistent bleeding after a second dose and they underwentsurgical re-exploration, which showed a surgical siteresponsible for blood loss in all. In contrast, 35 patientsin the control group (87.5%) underwent a second surgical re-exploration for persistent bleeding and in only two caseswas a surgical bleeding site identified. Furthermore,according to Karkouti et al. [18], patients receiving rFVIIashowed a lower ICU LoS and shorter time of mechanicalventilation. In addition, the use of rFVIIa was associatedwith a reduction of RBCs, FFP, and PLT transfusionrequirements. Finally, treated patients showed improvedhemostasis with rapid normalization of coagulation vari-ables: PT was increased in patients who received rFVIIa( p < 0.001), and, accordingly, INR was significantly reducedin the study group ( p < 0.001). Finally, PLT count wassignificantly higher in patients treated with rFVIIa( p < 0.001), whereas aPTT and FBG did not differ betweengroups.

4.1. Thrombotic risk

Recombinant factor VIIa is a drug generally well tolerated.Nonetheless, the occurrence of thrombosis is of primaryconcern with the administration of rFVIIa [19]. The tissuefactor (TF) is localized in the deeper layers of vascular wallsand, normally, it is not exposed to the circulating blood. TFbinds to FVII at a site of vascular injury and TF—VIIa complex

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—7170

allows the conversion of FVII to FVIIa, activates factors IX andX and enhances thrombin generation. Cardiopulmonarybypass (CPB) may up-regulate TF systematically as well asat the site of surgical injury [20] and this may lead to systemicor local thrombosis.

Nevertheless, despite the use of rFVIIa in 304 cardiacsurgery patients with refractory hemorrhage, the incidenceof this serious adverse event remains very low (4.6%) [1].This can be theoretically explained by the presence of somepotential protective mechanisms which work againstthrombotic complications in these patients: (1) the plasmaconcentration of tissue factor pathway inhibitor (TFPI), astrong inhibitor of the enzymatic activity of the TF—FVIIacomplex, rises, still remaining high at the end of CPB; [21](2) antifibrinolytic agents are normally used in conjunctionwith rFVIIa. Tranexamic acid inhibits binding of plasmino-gen to fibrin and aprotinin is a non-specific inhibitorof serine proteases, although its relative affinity is highestfor plasmin [22]. However, even though published datahave shown a good level of safety in cardiac surgerypatients, these subjects must be considered at high risk forthrombotic complications [19]. Particularly, the influenceof rFVIIa on graft patency after coronary artery bypassgrafting (CABG) is still unknown and many authors believethe use of rFVIIa to be contraindicated in these subjects. Inthe present report, among patients in the study group, 20(50%) underwent isolated or combined CABG and they didnot show either clinical, or electrocardiographic andechocardiographic signs of graft occlusion. Two studypatients had postoperative stroke and in both a predispos-ing factor was clearly identified; in the remaining patientsno thromboembolic complication was detected.

4.2. Study limitations

Our study findings should be viewed in light of someinherent limitations. This study was neither randomized norprospective. Nonetheless, in the absence of randomization,we employed the propensity score to ensure an unbiasedcomparison for treatment effectiveness. However, thepropensity score matching can only adjust for observedpotential confounders, thus we employed a sensitivityanalysis for persistent bleeding to account for unmeasuredconfounders.

PCC has been shown to be more effective than rVIIa toreverse melagatran anticoagulation [23] and coumarinanticoagulation [24] and that manipulation of prothrombinconcentration with PCC significantly impacts the effective-ness of rVIIa in a cell model of hemophilia [25] suggesting thatthe combination of drugs may offer advantages over the useof rVIIa alone and this could explain why such minor doses ofrVIIa were, in our experience, effective.

4.3. Strength of the study

The patients in our study were treated in a singleinstitution according to standardized clinical guidelines.Moreover, our patient population was more homogeneousthan in other published studies. Additionally, the dose ofrFVIIa given and the timing of administration were morestandardized than in other reports. Finally, the dosage

employed was, as far as we know, the lowest reported in theEnglish literature.

5. Conclusions

Even with the above-mentioned limitations, we canconclude that low doses resulted to be effective in refractoryhemorrhage associated with cardiac surgery. Further larger,randomized, multicenter, controlled trials are necessary toconfirm the efficacy of rFVIIa in this setting.

Acknowledgments

We gratefully acknowledge Dr Judith Wilson for theEnglish revision of the manuscript. We thank Dr OrlandoParise for the statistical analysis.

References

[1] Warren O, Mandal K, Hadjianastassiou V, Knowlton L, Panesar S, John K,Darzi A, Athanasiou T. Recombinant activated factor VII in cardiacsurgery: a systematic review. Ann Thorac Surg 2007;83(2):707—14.

[2] Hedner U. Recombinant factor VIIa: its background, development andclinical use. Curr Opin Hematol 2007;14(3):225—9.

[3] Connell KA, Wood JJ, Wise RP, Lozier JN, Braun MM. Thromboembolicadverse events after use of recombinant human coagulation factor VIIa.JAMA 2006;295:293—8.

[4] Lodge JP, Jonas S, Oussoultzoglou E, Malago M, Jayr C, Cherqui D,Anthuber M, Mirza DF, Kuhlman L, Bechstein WO, Dıaz JC, Tartiere J,Eyraud D, Fridberg M, Erhardtsen E, Mimoz O. Recombinant coagulationfactor VIIa in major liver resection: a randomized, placebo-controlled,double-blind clinical trial. Anesthesiology 2005;102:269—75.

[5] Kirklin JW, Barrat Boyes B. Morphology, diagnostic criteria, naturalhistory, techniques, result and indications. In: Kirklin JW, Barrat BoyesB, editors. Cardiac surgery. Philadelphia: Churchill Livingstone; 1997. p.195—253.

[6] Hensley FA, Martin DE, Gravlee GP. A practical approach to cardiacanesthesia, 3rd ed., Philadelphia: Lippincott Williams & Wilkins; 2003 .pp. 495—498.

[7] Mayer SA, Brun NC, Begtrup K, Broderick J, Davis S, Diringer MN, SkolnickBE, Steiner T, Recombinant Activated Factor VII Intracerebral Hemor-rhage Trial Investigators. Recombinant activated factor VII for acuteintracerebral hemorrhage. N Engl J Med 2005;352:777—85.

[8] Yasaka M, Sakata T, Naritomi H, Minematsu K. Optimal dose of prothrom-bin complex concentrate for acute reversal of oral anticoagulation.Thrombosis Res 2005;115:455—9.

[9] Nuttall GA, Stehling LC, Beighley CM, Faust RJ, American Society ofAnesthesiologists Committee on Transfusion Medicine. Current transfu-sion practices of members of the American society of anesthesiologists: asurvey. Anesthesiology 2003;99(6):1433—43.

[10] Rosembaum PR, Rubin DB. The central role of the Propensity Score inobservational studies for causal effects. Biometrika 1983;70:41—55.

[11] Parsons, LS. Using SAS Software to Perform a Case-Control Match onPropensity Score in an Observational Study. In: Proceedings of the Twenty-Fifth Annual SAS Users Group International Conference, Cary, NC: SASInstitute Inc. 2000; 1166—1171.

[12] Feinstein AR. Multiple logistic regression. In: Feinstein AR, editor.Multivariable analysis: an introduction. New Haven: Yale UniversityPress; 1996 . p. 297—330.

[13] Parsons LS. Reducing bias in a propensity score match-pair sample usinggreedy matching techniques. In: Proceedings of the 26th Annual SAS UsersGroup International Conference. Cary, NC: SAS Institute Inc. 2001;214—26.

[14] Lin DY, Psaty BM, Kronmal RA. Assessing the sensitivity of regressionresults to unmeasured confounders in observational studies. Biometrics1998;54:948—63.

S. Gelsomino et al. / European Journal of Cardio-thoracic Surgery 33 (2008) 64—71 71

[15] Tanaka KA, Waly AA, Cooper WA, Levy JH. Treatment of excessivebleeding in Jehovah’s Witness patients after cardiac surgery with recom-binant factor VIIa (NovoSeven). Anesthesiology 2003;98:1513—5.

[16] Tobias JD, Simsic JM, Weinstein S, Schechter W, Kartha V, Michler R.Recombinant factor VIIa to control excessive bleeding following surgeryfor congenital heart disease in pediatric patients. J Inten Care Med2004;19:270—3.

[17] Raivio P, Suojaranta-Ylinen R, Kuitunen AH. Recombinant factor VIIa inthe treatment of postoperative hemorrhage after cardiac surgery. AnnThorac Surg 2005;80:66—71.

[18] Karkouti K, Beattie WS, Dattilo KM, McCluskey SA, Ghannam M, Hamdy A,Wijeysundera DN, Fedorko L, Yau TM. Recombinant factor VIIa forintractable blood loss after cardiac surgery: a propensity score-matchedcase control analysis. Transfusion 2005;45:26—34.

[19] O’Connel KA, Wood JJ, Wise RP, Lozier JN, Braun MM. Thromboembolicadverse events after use of recombinant human coagulation factor VIIa.JAMA 2006;295:293—8.

Appendix A. Sensitivity analysis

G P1 P0

0.0 0.1 0.2

20.0 12.42 (4.25—36.26)0.2 10.35 (3.54—30.11) 11.38 (3.89—33.13) 12.42 (4.25—36.20.4 8.87 (3.03—25.81) 9.75 (3.34—28.39) 10.64 (3.64—30.90.6 7.76 (2.65—22.58) 8.53 (2.92—24.84) 9.31 (3.19—27.100.8 6.9 (2.36—20.07) 7.59 (2.59—22.08) 8.28 (2.83—24.091.0 6.21 (2.12—18.07) 6.83 (2.33—19.87) 7.45 (2.55—21.68

40.0 12.42 (4.25—36.26)0.2 7.76 (2.65—22.58) 10.09 (3.45—29.36) 12.42 (4.25—36.20.4 5.64 (1.93—16.04) 7.33 (2.51—21.35) 9.03 (3.09—26.280.6 4.43 (1.51—12.09) 5.76 (1.97—16.78) 7.09 (2.43—20.650.8 3.65 (1.25—10.63) 4.74 (1.62—13.81) 5.84 (2.00—17.001.0 3.10 (1.03—9.03) 4.03 (1.38—11.74) 4.96 (1.70—14.45

60.0 12.42 (4.25—36.26)0.2 6.21 (2.12—18.07) 9.31 (3.19—27.10) 12.42 (4.25—36.20.4 4.14 (1.41—12.04) 5.81 (2.01—16.88) 8.28 (2.83—24.090.6 3.10 (1.06—9.03) 4.65 (1.59—13.55) 6.21 (2.12—18.070.8 2.48 (0.85—7.22) 3.72 (1.27—10.84) 4.96 (1.70—14.451.0 2.07 (0.70—2.02) 3.10 (1.06—9.03) 4.14 (1.41—12.04

80.0 12.42 (4.25—36.26)0.2 5.17 (1.77—15.05) 8.79 (3.01—25.60) 12.42 (4.25—36.20.4 3.26 (1.11—9.51) 5.55 (1.9—16.16) 7.48 (2.68—22.820.6 2.38 (0.81—6.95) 4.06 (1.39—11.81) 5.73 (1.96—16—60.8 1.88 (0.64—5.47) 3.19 (1.09—9.30) 4.51 (1.54—13.141.0 1.55 (0.53—4.51) 2.63 (1.00—7.68) 3.72 (1.27—10.84

100.0 12.42 (4.25—36.26)0.2 4.43 (1.51—12.90) 8.42 (2.88—24.52) 12.42 (4.25—36.20.4 3.70 (1.92—7.85) 5.13 (1.75—14.92) 7.56 (2.58—21.990.6 2.94 (1.66—5.64) 3.68 (1.56—10.72) 5.43 (1.86—15.810.8 1.51 (0.51—4.40) 2.87 (1.26—8.37) 4.24 (1.45—12.341.0 1.24 (0.42—3.61) 2.35 (1.08—6.86) 3.47 (1.19—10.11

The table shows point estimates and 95% confidence interval for the odds ratio of pe

for definitions) as binary outcome with adjustment for an unmeasured binary co

confounder. We assumed [16] that G0 = G1 = G. P0 and P1are the prevalences of

displayed only the estimates for P1 � P0. The estimates will be inflated upward

conclusion that the use of rVIIa reduces the risk of persistent bleeding. In fact, the p

95% confidence intervals are greater than 1 except for six entries with G � 6 and wi

influence the rVIIa effect on persistent bleeding, should have an imbalance in p

[20] Chung JH, Gikakis N, Rao AK, Drake TA, Colman RW, Edmunds Jr LH.Pericardial blood activates the extrinsic coagulation pathway duringclinical cardiopulmonary bypass. Circulation 1996;93:2014—8.

[21] Donahue BS, Gailani D, Mast AE. Disposition of tissue factor pathwayinhibitor during cardiopulmonary bypass. J Thromb Haemost2006;4(5):1011—6.

[22] Bevan DH. Cardiac bypass haemostasis: putting blood through the mill. BrJ Haematol 1999;104(2):208—19.

[23] Sørensen B, Ingerslev J. A direct thrombin inhibitor studied by dynamicwhole blood clot formation. Thromb and Haemost 2006;96:446—53.

[24] Dickneite G. Prothrombin complex concentrate versus recombinant fac-tor VIIa for reversal of coumarin anticoagulation. Thromb Res2007;119:643—51.

[25] Geoffrey A, Allen GA, Hoffman M, Roberts HR, Monroe DM. Manipulationof prothrombin concentration improves response to high-dose factorVIIa in a cell-based model of haemophilia. Br J Haematol 2006;134:314—9.

0.3 0.4 0.5

6)7) 11.53 (3.94—33.56) 12.42 (4.25—36.26)) 10.09 (3.45—29.36) 10.86 (3.72—31.62) 11.64 (3.98—33.8)) 8.97 (3.07—26.10) 9.66 (3.30—28.11) 10.35 (3.54—30.11)) 8.07 (2.76—23.49) 8.69 (2.97—25.29) 9.31 (3.19—27.1)

6)) 10.72 (3.67—31.21) 12.42 (4.25—36. 26)) 8.42 (2.88—24.52) 9.75 (3.34—28.39) 11.08 (3.79—32.26)) 6.94 (2.37—20.19) 8.03 (2.75—23.38) 9.13 (3.12—26.57)) 5.89 (2.02—17.16) 6.83 (2.33—19.87) 7.76 (2.65—22.58)

6)) 10.35 (3.54—30.11) 12.42 (4.25—36.26)) 7.76 (2.65—22.58) 9.31 (3.19—27.10) 10.86 (3.72—31.62)) 6.21 (2.12—18.07) 7.45 (2.55—21.6) 8.69 (2.97—25.29)) 5.17 (1.77—15.05) 6.21 (2.12—18.07) 7.24 (2.48—21.08)

6)) 10.13 (3.46—29.48) 12.42 (4.25—36.26)8) 7.40 (2.53—21.54) 9.07 (3.10—26.41) 10.74 (3.68—31.27)) 5.83 (1.99—16.97) 7.15 (2.44—20.80) 8.46 (2.90—24.64)) 4.81 (1.64—14.00) 5.89 (2.02—17.16) 6.98 (2.39—20.32)

6)) 9.99 (3.42—29.07) 12.42 (4.25—36.26)) 7.18 (2.45—20.89) 8.92 (3.05—25.97) 10.67 (3.65—31.05)) 5.60 (1.91—16.30) 6.96 (2.38—20.27) 8.33 (2.85—24.24)) 4.59 (1.57—13.37) 5.71 (1.95—16.62) 6.83 (2.39—19.87)

rsistent bleeding after rVIIa use in patients with refractory bleeding (see text

nfounder. G0 and G1 are the relative risk associated with the unmeasured

the unmeasured confounder in treated vs control patients, respectively. We

if P1 � P0.The results of this sensitivity analysis significantly support the

oint estimates are always much higher than 1, whereas the lower limits of the

th P0 = 0 and P10.8 � 1. In other words, an unmeasured binary confounder, to

revalence between control and treated group �80% (odds ratio �6).