- Email: [email protected]
Development of a Predictive Model for Blood Transfusions and Bleeding During Liver Transplantation: An Observational Cohort Study
Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Contents lists available at ScienceDirect
journal homepage: www.jcvaonline.com
Original Article
Development of a Predictive Model for Blood Transfusions and Bleeding During Liver Transplantation: An Observational Cohort Study Luc Massicotte, MDn,1, François Martin Carrier, MD†, André Y. Denault, MD, PhD†, Pierre Karakiewicz, MD, MSc‡, Zoltan Hevesi, MD§, Mickael McCormack, MD‡, Lynda Thibeault, MD, MSc||, Anna Nozza¶, Zhe Tiannn, Michel Dagenais, MD, MSc††, André Roy, MD†† n
Anesthesiology Department, Centre Hospitalier de l’Université de Montréal (CHUM), Hôpital St-Luc, Montreal, QC, Canada † Anesthesiology Department and Critical Care Division, Centre hospitalier de l’Université de Montréal (CHUM), Hôpital St-Luc, Montreal, QC, Canada ‡ Urology Division, Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada § Anesthesiology Department, University of Wisconsin || Epidemiology Department ¶ Montreal Health Innovation Coordinating Center (MHICC), Montreal, QC, Canada nn Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada †† Hepato-biliary Division, Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada
Objective: Orthotopic liver transplantation (OLT) frequently is associated with major blood loss and considerable transfusion requirements. The goal of this study was to define the risk factors for multiple transfusions and major bleeding during OLT and to help identify higher risk patients that could benefit from targeted interventions. Design: OLTs were studied for this observational cohort study. Setting: Community hospital. Participants: A total of 800 consecutive OLTs were studied. Intervention: No intervention. Measurements and Main Results: Baseline and intraoperative data were gathered. Multivariate logistic regression analyses were performed to find variables associated with 2 outcomes: transfusion of more than 2 units of red blood cells (RBC) and bleeding Z900 mL. Two nomograms were developed to predict individual risks. The overall intraoperative RBC transfusion was 0.6 7 1.4 units on average, and 61 surgeries (7.6%) received more than 2 units of RBC (4.5 7 1.9). Some variables were associated with the outcomes: 5 were associated with transfusion of more than 2 units of RBC (patient’s height, starting hemoglobin concentration, starting bilirubin value, the use of a phlebotomy, and central venous pressure [CVP] at the time of vena cava clamping) and 3 with blood loss of Z900 mL (starting hemoglobin value, Child-Turcotte-Pugh score, and CVP at the time of vena cava clamping). Preclamping CVP showed the strongest association with both outcomes. Nomograms were developed to predict the individual OLT recipients’ risk of requiring more than 2 units RBC and suffering from major bleeding. Among the variables associated with multiple RBC transfusions and major bleeding, 3 can lead to interventions: baseline hemoglobin value, the use of a phlebotomy, and the preclamping CVP.
1 Address reprint requests to Dr Luc Massicotte, CHUM- Hôpital St-Luc, 1058, St-Denis, Montreal, QC H2X 3J4, Canada. E-mail address: [email protected] (L. Massicotte).
http://dx.doi.org/10.1053/j.jvca.2017.10.011 1053-0770/& 2017 Elsevier Inc. All rights reserved.
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
2
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Conclusion: Some variables were able to predict the risk of multiple transfusions and major bleeding in this low bleeding liver transplantation population. Further studies based on these variables should be done to better define the role of targeted interventions in higher risk liver transplant recipients. & 2017 Elsevier Inc. All rights reserved. Key Words: liver transplantation; bleeding; transfusion; coagulation
ORTHOTOPIC LIVER TRANSPLANTATION (OLT) often is associated with major blood loss and a need for massive blood product transfusions.1 During the past 2 decades, a significant decrease in perioperative blood loss and blood product requirements has been observed during OLT.2 Although some centers used very limited amount of transfusions, major intraoperative bleeding was still common in many other programs.3–5 Hemostasis in cirrhotic patients is a complex state characterized by a reduced level of both pro- and anticoagulant proteins. A low platelet count counterbalanced by an elevated von Willebrand factor and a low plasma levels of ADAMST13 (a von Willebrand factor cleaving protease) are other characteristics of this state.6,7 Hyperfibrinolysis also is very common.6,7 With such an understanding of these hemostatic characteristics, some clinicians adopted a wait-and-see approach instead of a prophylactic one, using transfusion of blood products only when active, nonsurgical bleeding occurs rather than to “correct” abnormal hemostasis laboratory values.8,9 Such an approach likely has contributed to the observed decrease in the volume of blood transfusions associated with OLT today.2,8 However, some studies still report populations with high amounts of RBC transfusions (mean Z 6 RBC units, Z8 RBC units, or 20-30 RBC units) and have tried to predict transfusions with different models based on cohorts between 150 and 800 OLTs.3–5 They obtained conflicting results for many potential predictive variables (age, starting hemoglobin value, international normalized ratio [INR], platelets count, creatinine, albumin, and second OLT). A substantial and growing body of evidence suggests that the excessive use of blood products is associated with an increased morbidity and mortality during OLT.9–16 Because it is likely that decreased blood loss could prevent the deleterious effects of transfusions and affect outcomes, interventions to reduce bleeding in liver transplant recipients are needed. However, few high-quality evidence supports interventions that decrease blood product exposure and improve outcomes in this population, except the use of antifibrinolytic drugs.17–20 To help target interventions that reduce OLT patients’ exposure to blood products, predictive models to identify higher risk patients for bleeding and transfusions should be developed. Previous studies suggested that it is possible to develop such predictive models, even in a liver transplantation center with a low transfusion rate.2,21 Therefore, the authors completed a longitudinal observational cohort study to further define the risk factors for major bleeding and multiple transfusions in their OLT population and to help identify higher risk patients. The primary objectives of this study were
to describe the authors’ OLT population further with a large cohort study and to develop a model that could predict the risk of using 42 units RBC and major intraoperative bleeding based on baseline and intraoperative characteristics. The secondary objective was to develop a logistic regression-based nomogram that predicts the individual risk of these outcomes. Patients and Methods This is a nonexperimental observational cohort study. The authors report it according to the STROBE statement guidelines for observational studies.22 This study was undertaken at the Centre Hospitalier de l’Université de Montréal (CHUM) (registered on Clinical Trial.gov, NCT: 02548130). After approval by the review ethics board (REB approval #15.113), all consecutive OLTs in adults that took place in the authors’ center between October 2002 and February 2016 were included. No exclusion criteria were applied. Data Collection and Management Baseline population characteristics, including baseline laboratory values, as well as intraoperative data were collected prospectively with a standardized data report form during and after each OLT. Collected intraoperative data included duration of surgery, baseline and preanhepatic CVP, volume of fluid resuscitation, type of fluid used, volume of phlebotomy performed, blood products transfused, volume of cell saver reinfused, total bleeding, and total intraoperative diuresis. Postoperative hemoglobin, creatinine, and INR values, as well as mortality up to 1 year, also were collected. Data were kept in a secured database. Because all OLTs were included, no selection bias was expected. Outcomes The authors’ outcomes of interest were the number of RBC units transfused and the total intraoperative blood lost. Intraoperative blood lost was measured from the blood suctioned through the cell saver minus the irrigating fluid put in the surgical field. Intraoperative Management Protocol A standard monitoring and anesthesia technique were used in all patients, as previously described.2,8,18 All grafts were harvested from cadaveric donors and were transplanted using a total cross-clamping technique with vena cava replacement. The transfusion of procoagulant blood products was based on a “wait and see,” approach and all blood products were
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
transfused based on local guidelines (see Appendix I).2,8 Bedside hemoglobin measurements were used, but no other point-of-care testing was available. Aprotinin was administered for the first 300 OLTs according to the Hammersmith protocol,23 and tranexamic acid was used for the last 500 OLTs according to the BART protocol (Blood conservation using Antifibrinolytics in a Randomized Trial).24 A cell saver device was part of the authors’ routine care for every case, except for the initial 75 OLTs. From the beginning of this cohort and based on a local protocol, anesthesiologists used a restrictive fluid resuscitation strategy before the anhepatic phase. This strategy included a 33% lowering of the central venous pressure (CVP) from baseline, the use of a phlebotomy (withdrawal of 7-10 mL/kg of blood without any infusion of fluid to replace the withdrawn volume), or a combination of both.2 Phlebotomies were performed if patients had both a hemoglobin concentration above 85 g/L and normal renal function. They were interrupted if the mean arterial blood pressure dropped by more than 20% of the baseline value despite vasopressor administration. Phlebotomies were infused back at the end of surgery or earlier if needed clinically. Statistical Analyses The number of RBC units transfused were dichotomized in 2 categories because the authors considered, based on previous data, that transfusing more than 2 units is a more clinically significant outcome than being exposed to 2 units or less.2 Total intraoperative bleeding also was dichotomized at the median ( Z900 mL) for similar statistical analyses. Data are expressed as mean 7 standard deviation for continuous data or as a percentage for proportions. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC). Baseline and intraoperative characteristics were compared between the authors’ dichotomized outcome-based groups using Student t tests and Fisher exact tests. Univariate mixed logistic regression models (PROC GLIMMIX) then were used to assess the association of 16 different variables with transfusion of more than 2 RBCs and with bleeding above the median. To take into account the correlation within subjects (ie, because a subject can have multiple OLTs), a random intercept was included in the model. A mixed logit multivariate model then was used by incorporating the significant factors identified in the univariate analyses. Due to the small number of OLTs in the 4 2 unit RBC group (N ¼ 61), the number of independent variables to put into the final model was limited. The selection for both final models was based on a number of factors: eliminating all variables that were correlated and calculating the discriminant accuracy quantified in terms of the c-index for different combinations of potential independent variables. The regression coefficients from the mixed logit multivariable logistic regression model then were used to develop nomograms that predict the individual probability to receive more than 2 RBC units and to have a total intraoperative blood loss Z 900 mL.19,25 All variables of interest were tested for their linear effect. The linearity assumption was assessed by incorporating both the
3
variable of interest in its continuous and categorical form (ie, dichotomized at its quantiles) in the model (PROC GLIMMIX). For both outcomes (ie, transfusions and blood loss), the predictor variables in their categorical form were nonsignificant. That is, there is no additional effect of the variable beyond its linear effect. Subsequently, an internal validation of the nomogram to test its ability to correctly discriminate between those patients who received more or less than 2 units RBC was performed. The area under the receiver operating characteristic curve was used to quantify the nomogram’s discriminant properties and 800 bootstrap resamples were used to correct for over fit bias. Subsequently, the model’s calibration was tested. The LOESS smoother was used to plot the nomogram predicted probabilities of transfusing more or less than 2 RBC units or bleeding relative to the observed rate in the development cohort. Results Transfusion Exposure A total of 800 OLTs were performed ion 735 patients during the predefined period (679 patients had 1 OLT, 47 had 2 OLTs, and 9 had 3). Figure 1 shows the number of OLTs plotted against the number of RBC units transfused. The mean intraoperative transfusion of RBC units for all 800 cases was 0.6 7 1.4. Sixty-one transplantations (7.6%) required more than 2 RBC units, with a mean of 4.5 7 1.9 units in this subgroup. The remaining 739 transplantations (92.4%) who received 2 RBC units or less received a mean of 0.3 7 0.6 units. Six hundred and ten cases (76.3%) did not require any blood product. However, 14 OLTs (1.8%) required more than 5 RBC units. Population Characteristics and Outcomes The authors compared the demographic and health characteristics of the patients between their dichotomized outcomebased groups (Table 1), as well as their intraoperative characteristics (Table 2). As already mentioned, patients were divided between those requiring r 2 RBCs versus 42 RBCs
Fig 1. Orthotopic liver transplantation distribution by the number of RBC units transfused (total of 800 liver transplantations).
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Abbreviations: CTP, Child-Turcotte-Pugh; Hb, hemoglobin; INR, international normalized ratio; MELD, Model for End-Stage Liver Disease; OLT, orthotopic liver transplantation; RBC, red blood cell; REDO, reoperation (2nd or 3rd OLT). n Data available for only N ¼ 210 OLTs (N ¼ 187 and N ¼ 23 in the r2 RBC and 42 RBC groups, respectively; N ¼ 83 and N ¼ 127 in the r900 mL and 4900 mL groups, respectively).
33% 7 11 7 18 7 9 7 22 7 1.2 7 56 7 1.00 7 61 7 158 7 2.4 7 7 11% 52 77 170 101 2.0 85 1.81 105 137 10.4 24 33% 7 11 7 17 7 9 7 24 7 0.9 7 55 7 1.10 7 91 7 131 7 2.5 7 8 10% 51 75 169 111 1.7 98 2.32 103 106 9.3 21 0.0022 0.6213 0.4587 0.0077 o0.0001 0.0031 0.5223 0.6370 0.0062 0.0002 0.0007 0.0019 o0.0001 51% 7 12 7 18 7 10 7 12 7 1.5 7 55 7 1.17 7 68 7 229 7 2.4 7 8 26% 53 74 166 85 2.4 87 1.82 133 217 11.1 26 31% 711 7 17 7 9 7 23 7 1.0 7 56 7 1.05 7 77 7 135 7 2.5 7 9 9% 52 76 169 108 1.8 92 2.04 102 115 9.8 22 33% 7 11 7 17 7 9 7 23 7 1.0 7 56 7 1.06 7 77 7 146 7 2.5 7 9 10% 52 76 169 106 1.9 91 2.01 104 123 9.9 22 Sex (women) % Age (y) Weight (kg) Height (cm) Starting Hb value (g/L) Starting INR value Starting platelet count (109 pl/L) * Starting fibrinogen (g/L) Starting creatinine value (mmol/L) Starting bilirubin value (mmol/L) CTP score MELD score % REDO
p Value Z900 mL 420 OLTs o900 mL 379 OLTs p Value 42 RBC 61 OLTs r2 RBC 739 OLTs Total of 800 OLTs Variables
Table 1 Demographic and Health Characteristics for All 800 Liver Transplantations and for Patients Who (1) Received r2 RBC Units Versus 42 RBC Units and (2) Bled o900 mL Versus Z900 mL
0.9147 0.2056 0.0800 0.2936 o0.0001 0.0009 0.0045 0.0022 0.7906 0.0039 o0.0001 0.0001 0.5073
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
4
and between patients who bled o 900 mL versus Z900 mL. Patients who received more than 2 RBC units were smaller; were more often female; had more previous OLTs; had a lower starting hemoglobin value; had higher INR, creatinine, and bilirubin values; and had higher Child-Turcotte-Pugh (CTP) and Model of End-Stage Liver Disease (MELD) scores (the MELD score used was the available adjusted MELD score based on the United Network for Organ Sharing criteria). Both the starting CVP and the CVP before the vena cava clamping were higher in groups requiring more than 2 RBC units and with blood lost Z 900 mL. Transfusions of all labile blood products were higher in the same 2 groups (fresh frozen plasma, platelets, cryoprecipitate), as was the total volume of intravenous crystalloid, albumin 5%, and cell saver output infused. Even though the threshold for transfusion of RBCs was the same among the 4 groups, the final hemoglobin value was lower in the 42 RBCs and Z 900 mL groups. The percentage of patients having a phlebotomy performed before the anhepatic phase was lower in the groups with 42 RBC units transfused and Z900 mL blood lost. Patients who had a phlebotomy also had a lower mean blood lost (1,049 7 950 mL v 1,524 7 1,426 mL) and a lower mean of RBC transfusions (0.2 7 0.9 v 1.1 7 1.7) compared with those who did not have a phlebotomy (not shown in tables). Outcome Predictive Models Due to the small number of patients who received more than 2 RBCs, the number of independent variables to put into the final model was limited. Gender and height were highly correlated, so height was kept instead of gender. MELD and CTP scores were also correlated and the CTP score had a higher c-index (0.861 v 0.821). From theses 2 scores, the bilirubin value had a higher c-index (0.882) than either score alone (MELD: 0.821, CTP: 0.861). Components of the CTP and the MELD scores (bilirubin, INR, and creatinine) were included in the final model. Variables analyzed in a dichotomic way using the median were eliminated to keep only continuous variables (median of bilirubin, INR, creatinine, and hemoglobin). Table 3 summarizes the multivariate mixed logistic regression analysis for transfusion of “more than 2 RBCs”. Five variables were kept in the final model because of a significant association: patient’s height, baseline hemoglobin concentration, starting bilirubin value, CVP at the time of the vena cava clamping and the use of a phlebotomy (yes or no). Regarding the patients’ height, each increase of 1 centimeter (compared with the mean of 169 cm) is associated with a 5.3% decrease in the predicted odds of receiving more than 2 RBCs. Also, for each increase of 1 g/L of the starting hemoglobin value, the odds decrease by 5.0% and for each increase of 1 μmol/L of bilirubin concentration, the estimated odds of receiving more than 2 RBCs increases by 0.3%. The risk of multiple transfusion increases by 9.0% when the preclamping CVP increases by 1 mmHg above the mean (7.8 mmHg). Finally, the use of a phlebotomy decreases the odds of receiving more than 2 RBCs by 71.5% compared with “no phlebotomy.”
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Variables *
RBC transfused (units)
FFP transfused (units) Platelet transfused (units) Cryoprecipitate transfused (units) Albumin 5% (mL) Pentastarch or 6% hydroxyethyl starch (mL) † Hb trigger of RBC transfusion (g/L) Final Hb value (g/L) Blood loss (mL) Duration of surgery (min) % of use of phlebotomy % of use of the cell saver Diuresis (mL) Starting CVP (mmHg) CVP before clamping (mmHg) Crystalloid transfused (mL)
Total of 800 OLTs
r2 RBC 739 OLTs
42 RBC 61 OLTs
p Value
o900 mL 379 OLTs
Z900 mL 420 OLTs
p Value
0 ¼ 610 (76.3%) 1 ¼ 62 (7.7%) 2 ¼ 67 (8.4%) 3 ¼ 22 (2.7%) 4 ¼ 19 (2.4%) 5-12 ¼ 20 (2.5%) 0 ¼ 730 (91.4%) 40 ¼ 69(8.6%) 0 ¼ 774 (96.8%) 40 ¼ 26(3.2%) 0 ¼ 774 (96.9%) 40 ¼ 25 (3.1%) 0 ¼ 746 (93.3%) 40 ¼ 54 (6.7%) 0 ¼ 242 (31.2%) 40 ¼ 534 (68.8%) 59 7 10 93 7 20 1,256 7 1,203 244 7 63 56% 71% 419 7 314 13.1 7 4.8 7.8 7 3.8 4,084 7 1,607
0 ¼ 610 (82.5%) 1 ¼ 62 (8.4%) 2 ¼ 67 (9.1%)
3 ¼ 22 (36.1%) 4 ¼ 19 (31.1%) 5-12 ¼ 20 (32.8%)
N/App
r2 ¼ 373 (98.4%) 42 ¼ 6 (1.6%)
r2 ¼ 365 (86.9%) 42 ¼ 55 (13.1%)
o0.0001
0 ¼ 707 (95.8%) 40 ¼ 31(4.2%) 0 ¼ 730 (98.8%) 40 ¼ 9(1.2%) 0 ¼ 727 (98.5%) 40 ¼ 11 (1.5%) 0 ¼ 699 (94.6%) 40 ¼ 40 (5.4%) 0 ¼ 222 (31.1%) 40 ¼ 493 (69.0%) 60 7 9 93 7 20 1,080 7 801 239 7 57 60% 70% 427 7 315 12.9 7 4.6 7.7 7 3.7 3,909 7 1,322
0 ¼ 23(37.7%) 40 ¼ 38(62.3%) 0 ¼ 44 (72.1%) 40 ¼ 17(27.9%) 0 ¼ 47 (77.1%) 40 ¼ 14 (22.9%) 0 ¼ 47 (77.1%) 40 ¼ 14 (22.9%) 0 ¼ 20 (32.8%) 40 ¼ 41 (67.2%) 59 7 10 85 7 15 3,381 7 2,531 310 7 91 13% 87% 322 7 283 15.4 7 5.8 9.6 7 4.9 6,192 7 2,824
o0.0001
0 ¼ 375 (98.9%) 40 ¼ 4(1.1%) 0 ¼ 377 (99.5%) 40 ¼ 2(0.5%) 0 ¼ 378 (99.7%) 40 ¼ 1 (0.3%) 0 ¼ 371 (97.9%) 40 ¼ 8 (2.1%) 0 ¼ 121 (33.4%) 40 ¼ 241 (66.6%) 62 7 8 97 7 20 554 7 190 229 7 54 65% 56% 425 7 296 12.6 7 4.3 7.3 7 3.5 3,637 7 1,153
0 ¼ 355 (84.7%) 1 ¼ 64(15.3%) 0 ¼ 396 (94.3%) 40 ¼ 24(5.7%) 0 ¼ 395 (94.3%) 40 ¼ 24 (5.7%) 0 ¼ 374 (89.1%) 40 ¼ 46 (10.9%) 0 ¼ 121 (29.3%) 40 ¼ 292 (70.7%) 59 7 10 88 7 18 1,889 7 1370 258 7 70 49% 85% 414 7 329 13.6 7 5.1 8.3 7 4.0 4,491 7 1,839
o0.0001
o0.0001 o0.0001 o0.0001 0.7764 0.4036 0.0033 o0.0001 o0.0001 o0.0001 0.0100 0.0379 0.0014 0.0017 o0.0001
Abbreviations: CVP, central venous pressure; FFP, fresh frozen plasma; Hb, hemoglobin; N/App, non applicable; OLT, orthotopic liver transplantation; RBC, red blood cells. n No RBC transfused in N ¼ 338/373 and N ¼ 271/365 for the r900 mL and 4900 mL groups, respectively. † Data available for only N ¼ 188 OLTs (N ¼ 127 and N ¼ 61 in the r2 RBC and 42 RBC; N ¼ 41 and N ¼ 147 in the r900 mL and 4900 mL groups, respectively).
0.0010 0.0022 o0.0001 0.2198 0.3035 o0.0001 N/App o0.0001 o0.0001 o0.0001 0.6224 0.0038 0.0004 o0.0001
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Table 2 Surgical Characteristics for All 800 Liver Transplantations and for Patients Who (1) Received r2 RBC Units Versus 42 RBC Units and (2) Bled o900 mL Versus Z900 mL
5
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
6
Table 3 Summary of the Multivariate Logistic Regression Model and Odds Ratio for More than 2 RBC Transfusions Variables
Estimate
Height –0.05427 Starting Hb –0.05085 Starting bilirubin 0.002787 Phlebotomy –1.2545 Pre-clamping CVP 0.08656
SE
0.01733 0.01434 0.001038 0.4705 0.03952
Odds Ratio
0.947 0.950 1.003 0.285 1.090
95% Wald CI Lower
Upper
0.916 0.924 1.001 0.113 1.009
0.980 0.978 1.005 0.717 1.178
Abbreviations: CI, confidence interval; CVP, central venous pressure; Hb, starting hemoglobin value; SE, standard error.
To analyze bleeding above the median (Z 900 mL), all available variables were included in the model due to the greater number of patients (420 patients). Table 4 summarizes the multivariate analysis for bleeding Z900 mL. Three variables were associated with outcome: starting hemoglobin value, CTP score, and CVP at the time of the vena cava clamping. For each 1 g/L increase in starting hemoglobin value above the mean, the risk of bleeding Z900 mL decreased by 1.6%. For each increase of 1 point on the CTP score, the risk of bleeding increased by 12%. Finally, the risk increased by 5.7% for each 1 mmHg increase in CVP at the time of clamping. Nomogram-Based Individual Outcome Prediction Figure 2 shows the nomogram itself that predicts individual probabilities of being transfused more than 2 RBC units. A low starting hemoglobin value represents the most powerful predictor of the transfusion risk. Two examples summarize this nomogram: a patient with height of 178 cm, with a starting hemoglobin value at 140 g/L, a bilirubin value at 50 μmol/L, with a phlebotomy, and a CVP value of 5 mmHg at the time of the vena cava clamping predicts a near zero probability of receiving more than 2 RBC units. At the opposite end of the spectrum, a patient with height of 150 cm, a starting hemoglobin of 70 g/L, a serum bilirubin concentration of 300 μmol/L, a CVP of 15 mmHg, and no phlebotomy has a predicted 80% risk of receiving more than 2 RBC units. Figure 3 shows the nomogram that predicts the risk of Table 4 Summary of the Multivariate Logistic Regression Model and Odds Ratio for Blood Loss, More or Less than the Median (900 mL) Variables
Starting Hb CTP score Preclamping CVP
Estimate
–0.01599 0.1136 0.05553
SE
0.00369 0.03264 0.02084
Odds Ratio
0.984 1.120 1.057
95% Wald CI Lower
Upper
0.977 1.051 1.015
0.991 1.194 1.101
Abbreviations: CI, confidence interval; CTP, Child-Turcotte-Pugh; CVP, central venous pressure; Hb, starting hemoglobin value; SE, standard error.
bleeding 900 mL or more. Starting hemoglobin value, CTP score, and CVP at the time of clamping were the significant variables. For example, a patient with a starting hemoglobin value of 100 g/L, a CTP score of 12, and a CVP value at 12 mmHg at the time of the vena cava clamping has a 65% risk of bleeding 900 mL or more. Discussion Transfusions of blood products are associated with increased morbidity and mortality. As previously suggested, the prophylactic correction of coagulation laboratory abnormalities reduces neither bleeding nor the need for transfusion and might contribute to morbidity, as intraoperative RBC transfusions were associated with a decrease in 1-year survival among liver transplant recipients.2,7,26 Indeed, prevention of excessive blood loss and multiple transfusions has become a major improvement goal in the perioperative management of liver transplantation patients.2,8–10 The main objective of this work was to describe the authors’ liver transplantation population further and to better identify, based on baseline and intraoperative characteristics, which patients will bleed more and require multiple blood transfusions. Data on 800 consecutive OLTs were available in the database for the authors’ predefined observational period. In this paper, the authors showed that their population still has a low transfusion exposure compared with many other liver transplantation populations described in the literature.3–5 Based on available variables, the authors built multivariate predictive models for multiple transfusions and major bleeding. The authors showed that preoperative characteristics, such as height, anemia, and severity of liver disease, are associated with a higher risk of major bleeding and a higher risk of being transfused more than 2 RBC units. The use of a phlebotomy prior to the anhepatic phase is associated with a reduced risk of receiving multiple RBC transfusions. The authors’ strongest predictor for major bleeding and multiple transfusions is a high CVP prior to vena cava clamping, based on the strength of association in the logistic regression model. Finally, the authors created nomograms that quantify the individual risk of multiple RBC transfusions and major bleeding. These nomograms represent a graphical display of the effect of several risk factors on the outcomes and provide the individual probability of receiving more than 2 RBC units or bleeding more than 900 mL in a 0% to 100% scale. Such a format is substantially more intuitive than odd ratios. These nomograms are tools that combine the most significant and informative predictors of multiple transfusions and major bleeding within 1 display. They will help target higher risk patients for interventions and further studies.21,25 This study has many limitations. First, only baseline and intraoperative characteristics gathered in the authors’ database were included in the analyses. However, these characteristics are considered important by many liver transplantation programs.3–5 Other variables could have been included, such as variables from medical history; patient hospitalization status; donor characteristics, including the donor risk index; and other
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
7
Points
Height
Starting Hemoglobin
Pre-op Bilirubin No
Phlebotomy Yes
CVP
Total Points
Predicted Value
Fig 2. Nomogram predicting transfusion of more than 2 units of red blood cells.
intraoperative characteristics and interventions. Second, only bleeding and transfusion outcomes were analyzed in this study. Many members of the liver transplantation community consider bleeding and transfusions being important surrogate outcomes that already have been associated with more longterm significant outcomes, such as mortality.2,9,16,20 However, other outcomes could have been considered, such as acute renal failure, graft complications, pulmonary and infectious complications, and burden of care outcomes, such as need for
postoperative organ support, length of stay, and readmissions. Predictive models associated for such intermediate and longterm outcomes probably should be done in future studies. Third, the specific role of phlebotomy is not well defined because it is performed only in lower risk patients, introducing many potential confounders. These confounders might be hard to identify and correct in any model. Furthermore, the authors used logistic regression models instead of linear ones to classify their cohort in more clinically significant categorized
Points
Starting Hb
CTP score
CVP clamping
Total Points
Probability
Fig 3. Nomogram predicting the risk of bleeding more than the median (900 mL). Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
8
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
outcomes. The size of their cohort allowed sufficient statistical power to do this. However, the authors looked at 2 outcomes and did multiple statistical tests; some of their findings could be chance alone. Finally, the predictive models and the nomograms presented here were developed within a single cohort and should be validated prospectively in an external cohort with similar bleeding and transfusion rates. However, the authors previously published a similar model to predict transfusion of 1 or more RBC units in a cohort of 400 patients that was prospectively validated in a second cohort. Because the authors obtained similar results in this study, these results emphasize the significance of some of those predictors in their population.21 Of the 5 variables linked to transfusion of more than 2 RBC units, 2 are not modifiable: the patient’s height and the starting serum bilirubin concentration. Smaller patients probably have a greater risk of receiving 42 RBC units because a similar blood loss is higher compared with their own blood volume. The MELD score is used widely to prioritize recipients for OLT, but some of the authors’ previous work suggested that this score predicts neither bleeding nor transfusions.27 Nevertheless, this study suggests that 1 variable part of the MELD score, the baseline bilirubin value, is associated with multiple transfusions. Therefore, hyperbilirubinemia itself seems to be a predictive factor for transfusions. The 3 other significant variables are either modifiable characteristics or an intervention itself. The authors’ previous work, based on smaller cohorts, identified 2 predictors of RBC transfusions: the nonuse of a phlebotomy and the starting hemoglobin value.21 Then, transfusion of RBC units was analyzed as a dichotomic outcome divided between no RBC transfusion and Z 1 RBC unit transfused and the baseline hemoglobin value was a stronger predictor for transfusions than any traditional disease severity score (CTP and MELD score).21 The present study focused on OLTs requiring transfusions of more than 2 RBC units, based on the fact that an intraoperative transfusion greater than 2 RBC units is a more significant exposure and represents a subgroup of patients that might benefit from blood sparing interventions.2 Here again, baseline hemoglobin value is a major determinant of multiple transfusions and severe blood loss. Such a significant prognostic role of the preoperative hemoglobin value could suggest that any intraoperative intervention to reduce transfusions might be limited. However, the use of a phlebotomy still is independently associated with a reduced risk of transfusion, suggesting that therapeutic interventions may be possible, even though this association might be inaccurate due to multiple confounders.2 Finally, preclamping CVP appears to be a powerful and significant predictor for both outcomes in this study. This finding suggests the role of volume overload in the risk of bleeding and transfusion in this population. It also emphasizes the possible mechanistic pathway of phlebotomy that may reduce portal pressure and sometimes CVP through a decrease in volume overload, as already suggested by the authors’ group.28,29 Some of the variables independently associated with the outcomes in the authors’ multivariate model are clinically correlated when it comes to interventions. Improvement of the
baseline hemoglobin value could be part of a preoperative optimization strategy and artificially increased during the waiting period prior to the OLT. However, a RBC transfusion close to the surgery might contribute to an increase in CVP, a harmful factor in this study. The value of such RBC prophylactic transfusions earlier in the waiting period should be explored further in clinical trials. As already mentioned, benefits of phlebotomy might be influenced by confounding factors reflecting patients’ better overall condition because it is performed only when hemoglobin value is higher than 85 g/L and when patients have normal renal function. However, an independent benefit from phlebotomy has been suggested in this study and in previous work, where patients who underwent a phlebotomy bled less, even if they had identical preclamping CVP.2 As already suggested, this effect might come from an isolated decrease in portal venous pressure, even though a high CVP at the time of the vena cava clamping might be associated with a higher portal pressure.30 This high CVP might be secondary to preoperative volume overload, a preanhepatic liberal fluid resuscitation strategy (high amounts of crystalloids, colloids, or blood products), heart failure, or any other cause of overall volume overload, including not having received a phlebotomy. The independent clinical role of the aforementioned variables and interventions must be explored further. Indeed, the optimal combination of various interventions, including early preoperative RBC transfusion and a restrictive fluid resuscitation strategy that encompasses phlebotomy and will minimize bleeding, reduce transfusions, and improve outcomes of liver transplant recipients remains to be determined. However, the external validity of these findings might be limited by the fact that many liver transplantation programs observe much higher transfusion rate. The authors’ findings probably may be extrapolated only to other low bleeding populations. Nonetheless, this study provides preliminary data to support further studies and clinical trials on the impact of aggressive preoperative correction of anemia and different perioperative fluid strategy on postoperative outcomes.
Conclusion This observational study of 800 consecutive OLTs was performed in a cohort with few patients having multiple RBC transfusions (7.6% received 42 RBC units). In this cohort, 5 variables were associated with transfusion of 2 RBC units: patient’s height, starting hemoglobin value, starting bilirubin value, the nonuse of a phlebotomy, and CVP at the time of the vena cava clamping. From these 5 variables, 2 were also associated with bleeding of Z900 mL: hemoglobin and CVP. Nomograms were developed, based on these variables, to predict each individual patient’s risk of receiving more than 2 RBCs and having major bleeding. This study will help predict risk of multiple transfusions and major bleeding and provides preliminary data to support future investigations on possible interventions that could improve postoperative outcomes in a low bleeding liver transplantation population.
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Acknowledgments The authors thank Mrs Nadine Gaudreau for her secretarial work. Appendix I. – Local Transfusion Guidelines The thresholds for transfusing blood products in the authors’ center were Red blood cells: hemoglobin between 60 and 70 g/L Fresh frozen plasma (10-15 mL/kg): INR 4 1.5 Platelets (5-10 units): platelet count r30 109/L Cryoprecipitate (5-10 units): fibrinogen r 2 g/L All procoagulant blood products were used only to treat significant bleeding and not to “correct” laboratory values. References 1 Butler P, Israel L, Nusbacher J, et al. Blood transfusion in liver transplantation. Transfusion 1985;25:120–3. 2 Massicotte L, Thibeault L, Roy A. Classical notions of coagulation revisited in relation with blood losses, transfusion rate for 700 consecutive liver transplantations. Semin Thromb Hemost 2015;41:538–46. 3 Roullet S, Biais M, Millas E, et al. Risk factors for bleeding and transfusion during orthotopic liver transplantation. Ann Fr Anesth Reanim 2011;30:349–52. 4 McCluskey SA, Karkouti K, Wijeysundera DN, et al. Derivation of risk index for the prediction of massive blood transfusion in liver transplantation. Liver Transpl 2006;12:1184–93. 5 Cywinski JB, Alster JM, Miller C, et al. Prediction of intraoperative transfusion requirements during orthotopic liver transplantation and the influence on postoperative patient survival. Anesth Analg 2014;118: 428–37. 6 Tripodi A, Mannucci PM. The coagulopathy of chronic liver disease. N Engl J Med 2011;365:147–56. 7 Lisman T, Porte RJ. Value of preoperative hemostasis testing in patients with liver disease for perioperative hemostatic management. Anesthesiology 2017;126:338–44. 8 Shan WL, Barkun J, Metrakos P, et al. Blood product use during orthotopic liver transplantation. Can J Anaesth 2004;51:1045–6. 9 Massicotte L, Beaulieu D, Thibeault L, et al. Coagulation defects do not predict blood product requirements during liver transplantation. Transplantation 2008;85:956–62. 10 Ozier Y, Klinck JR. Anesthetic management of hepatic transplantation. Curr Opin Anaesthesiol 2008;21:391–400. 11 Massicotte L, Denault AY, Beaulieu D, et al. Transfusion rate for 500 consecutive liver transplantations: Experience of one liver transplantation center. Transplantation 2012;93:1276–81. 12 Pereboom IT, de Boer MT, Haagsma EB, et al. Platelet transfusion during liver transplantation is associated with increased postoperative mortality due to acute lung injury. Anesth Analg 2009;108:1083–91.
9
13 Boyd SD, Stenard F, Lee DK, et al. Alloimmunization to red blood cell antigens affects clinical outcomes in liver transplant patients. Liver Transpl 2007;13:1654–61. 14 Hendriks HG, van der Meer J, de Wolf JT, et al. Intraoperative blood transfusion requirement is the main determinant of early surgical reintervention after orthotopic liver transplantation. Transpl Int 2005;17: 673–9. 15 Boin IF, Leonardi MI, Luzo AC, et al. Intraoperative massive transfusion decreases survival after liver transplantation. Tranpl Proc 2008;40:789–91. 16 Benson AB, Burton Jr JR, Austin GL, et al. Differential effects of plasma and red blood cell transfusions on acute lung injury and infection risk following liver transplantation. Liver Transpl 2011;17:149–58. 17 Massicotte L, Sassine MP, Lenis S, et al. Survival rate changes with transfusion of blood products during liver transplantation. Can J Anaesth 2005;52:148–55. 18 Porte RJ, Molenaar IQ, Begliomini B, et al. Aprotinin and transfusion requirements in orthotopic liver transplantation: A multicentre randomised double-blind study. EMSALT study group. Lancet 2000;355:1303–9. 19 Massicotte L, Denault AY, Beaulieu D, et al. Aprotinin versus tranexamic acid during liver transplantation: Impact on blood product requirements and survival. Transplantation 2011;91:1273–8. 20 Molenaar IQ, Wanaar N, Groen H, et al. Efficacy and safety of antifibrinolytic drugs in liver transplantation: A systematic review and meta-analysis. Am J Transplant 2007;7:185–94. 21 Gurusamy KS, Pissanou T, Pikhart H, et al. Methods to decrease blood loss and transfusion requirements for liver transplantation. Cochrane Database Syst Rev CD009052, 2011. 22 Massicotte L, Umberto C, Beaulieu D, et al. Independent validation of a model predicting the need for packed red blood cell transfusion at liver transplantation. Transplantation 2009;88:386–91. 23 Von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. J Clin Epidemiol 2008;6: 344–9. 24 Royston D, Bidstrup BP, Taylor KM, et al. Effect of aprotinin on need for blood transfusion after repeat open-heart surgery. Lancet 1987;2: 1289–91. 25 Ferguson DA, Hébert PC, Mazer CD, et al. A comparison of aprotinin and lysine analogues in high-risk cardiac surgery. N Engl J Med 2008;358: 2319–31. 26 Capitano U, Jeldres C, Shariat SF, et al. Clinicians are more familiar with nomograms and rate their clinical usefulness highest, look-up tables are second best. Eur Urol 2008;54:958. 27 Reyle-Hahn M, Rossaint R. Coagulation techniques are not important in directing blood product transfusion during liver transplantation. Liver Transpl 1997;3:663–5. 28 Massicotte L, Beaulieu D, Roy JD, et al. MELD score and blood product requirements during liver transplantation: No link. Transplantation 2009;87:1689–94. 29 Massicotte L, Lenis S, Thibeault L, et al. Effect of low central venous pressure and phlebotomy on blood product transfusion requirements during liver transplantations. Liver Transpl 2006;12:117–23. 30 Massicotte L, Perrault MA, Denault AY, et al. Effects of phlebotomy and phenylephrine infusion on portal venous pressure and systemic hemodynamics during liver transplantation. Transplantation 2010;89: 920–7.
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Contents lists available at ScienceDirect
journal homepage: www.jcvaonline.com
Original Article
Development of a Predictive Model for Blood Transfusions and Bleeding During Liver Transplantation: An Observational Cohort Study Luc Massicotte, MDn,1, François Martin Carrier, MD†, André Y. Denault, MD, PhD†, Pierre Karakiewicz, MD, MSc‡, Zoltan Hevesi, MD§, Mickael McCormack, MD‡, Lynda Thibeault, MD, MSc||, Anna Nozza¶, Zhe Tiannn, Michel Dagenais, MD, MSc††, André Roy, MD†† n
Anesthesiology Department, Centre Hospitalier de l’Université de Montréal (CHUM), Hôpital St-Luc, Montreal, QC, Canada † Anesthesiology Department and Critical Care Division, Centre hospitalier de l’Université de Montréal (CHUM), Hôpital St-Luc, Montreal, QC, Canada ‡ Urology Division, Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada § Anesthesiology Department, University of Wisconsin || Epidemiology Department ¶ Montreal Health Innovation Coordinating Center (MHICC), Montreal, QC, Canada nn Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada †† Hepato-biliary Division, Centre Hospitalier de l’Université de Montréal (CHUM), Montreal, QC, Canada
Objective: Orthotopic liver transplantation (OLT) frequently is associated with major blood loss and considerable transfusion requirements. The goal of this study was to define the risk factors for multiple transfusions and major bleeding during OLT and to help identify higher risk patients that could benefit from targeted interventions. Design: OLTs were studied for this observational cohort study. Setting: Community hospital. Participants: A total of 800 consecutive OLTs were studied. Intervention: No intervention. Measurements and Main Results: Baseline and intraoperative data were gathered. Multivariate logistic regression analyses were performed to find variables associated with 2 outcomes: transfusion of more than 2 units of red blood cells (RBC) and bleeding Z900 mL. Two nomograms were developed to predict individual risks. The overall intraoperative RBC transfusion was 0.6 7 1.4 units on average, and 61 surgeries (7.6%) received more than 2 units of RBC (4.5 7 1.9). Some variables were associated with the outcomes: 5 were associated with transfusion of more than 2 units of RBC (patient’s height, starting hemoglobin concentration, starting bilirubin value, the use of a phlebotomy, and central venous pressure [CVP] at the time of vena cava clamping) and 3 with blood loss of Z900 mL (starting hemoglobin value, Child-Turcotte-Pugh score, and CVP at the time of vena cava clamping). Preclamping CVP showed the strongest association with both outcomes. Nomograms were developed to predict the individual OLT recipients’ risk of requiring more than 2 units RBC and suffering from major bleeding. Among the variables associated with multiple RBC transfusions and major bleeding, 3 can lead to interventions: baseline hemoglobin value, the use of a phlebotomy, and the preclamping CVP.
1 Address reprint requests to Dr Luc Massicotte, CHUM- Hôpital St-Luc, 1058, St-Denis, Montreal, QC H2X 3J4, Canada. E-mail address: [email protected] (L. Massicotte).
http://dx.doi.org/10.1053/j.jvca.2017.10.011 1053-0770/& 2017 Elsevier Inc. All rights reserved.
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
2
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Conclusion: Some variables were able to predict the risk of multiple transfusions and major bleeding in this low bleeding liver transplantation population. Further studies based on these variables should be done to better define the role of targeted interventions in higher risk liver transplant recipients. & 2017 Elsevier Inc. All rights reserved. Key Words: liver transplantation; bleeding; transfusion; coagulation
ORTHOTOPIC LIVER TRANSPLANTATION (OLT) often is associated with major blood loss and a need for massive blood product transfusions.1 During the past 2 decades, a significant decrease in perioperative blood loss and blood product requirements has been observed during OLT.2 Although some centers used very limited amount of transfusions, major intraoperative bleeding was still common in many other programs.3–5 Hemostasis in cirrhotic patients is a complex state characterized by a reduced level of both pro- and anticoagulant proteins. A low platelet count counterbalanced by an elevated von Willebrand factor and a low plasma levels of ADAMST13 (a von Willebrand factor cleaving protease) are other characteristics of this state.6,7 Hyperfibrinolysis also is very common.6,7 With such an understanding of these hemostatic characteristics, some clinicians adopted a wait-and-see approach instead of a prophylactic one, using transfusion of blood products only when active, nonsurgical bleeding occurs rather than to “correct” abnormal hemostasis laboratory values.8,9 Such an approach likely has contributed to the observed decrease in the volume of blood transfusions associated with OLT today.2,8 However, some studies still report populations with high amounts of RBC transfusions (mean Z 6 RBC units, Z8 RBC units, or 20-30 RBC units) and have tried to predict transfusions with different models based on cohorts between 150 and 800 OLTs.3–5 They obtained conflicting results for many potential predictive variables (age, starting hemoglobin value, international normalized ratio [INR], platelets count, creatinine, albumin, and second OLT). A substantial and growing body of evidence suggests that the excessive use of blood products is associated with an increased morbidity and mortality during OLT.9–16 Because it is likely that decreased blood loss could prevent the deleterious effects of transfusions and affect outcomes, interventions to reduce bleeding in liver transplant recipients are needed. However, few high-quality evidence supports interventions that decrease blood product exposure and improve outcomes in this population, except the use of antifibrinolytic drugs.17–20 To help target interventions that reduce OLT patients’ exposure to blood products, predictive models to identify higher risk patients for bleeding and transfusions should be developed. Previous studies suggested that it is possible to develop such predictive models, even in a liver transplantation center with a low transfusion rate.2,21 Therefore, the authors completed a longitudinal observational cohort study to further define the risk factors for major bleeding and multiple transfusions in their OLT population and to help identify higher risk patients. The primary objectives of this study were
to describe the authors’ OLT population further with a large cohort study and to develop a model that could predict the risk of using 42 units RBC and major intraoperative bleeding based on baseline and intraoperative characteristics. The secondary objective was to develop a logistic regression-based nomogram that predicts the individual risk of these outcomes. Patients and Methods This is a nonexperimental observational cohort study. The authors report it according to the STROBE statement guidelines for observational studies.22 This study was undertaken at the Centre Hospitalier de l’Université de Montréal (CHUM) (registered on Clinical Trial.gov, NCT: 02548130). After approval by the review ethics board (REB approval #15.113), all consecutive OLTs in adults that took place in the authors’ center between October 2002 and February 2016 were included. No exclusion criteria were applied. Data Collection and Management Baseline population characteristics, including baseline laboratory values, as well as intraoperative data were collected prospectively with a standardized data report form during and after each OLT. Collected intraoperative data included duration of surgery, baseline and preanhepatic CVP, volume of fluid resuscitation, type of fluid used, volume of phlebotomy performed, blood products transfused, volume of cell saver reinfused, total bleeding, and total intraoperative diuresis. Postoperative hemoglobin, creatinine, and INR values, as well as mortality up to 1 year, also were collected. Data were kept in a secured database. Because all OLTs were included, no selection bias was expected. Outcomes The authors’ outcomes of interest were the number of RBC units transfused and the total intraoperative blood lost. Intraoperative blood lost was measured from the blood suctioned through the cell saver minus the irrigating fluid put in the surgical field. Intraoperative Management Protocol A standard monitoring and anesthesia technique were used in all patients, as previously described.2,8,18 All grafts were harvested from cadaveric donors and were transplanted using a total cross-clamping technique with vena cava replacement. The transfusion of procoagulant blood products was based on a “wait and see,” approach and all blood products were
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
transfused based on local guidelines (see Appendix I).2,8 Bedside hemoglobin measurements were used, but no other point-of-care testing was available. Aprotinin was administered for the first 300 OLTs according to the Hammersmith protocol,23 and tranexamic acid was used for the last 500 OLTs according to the BART protocol (Blood conservation using Antifibrinolytics in a Randomized Trial).24 A cell saver device was part of the authors’ routine care for every case, except for the initial 75 OLTs. From the beginning of this cohort and based on a local protocol, anesthesiologists used a restrictive fluid resuscitation strategy before the anhepatic phase. This strategy included a 33% lowering of the central venous pressure (CVP) from baseline, the use of a phlebotomy (withdrawal of 7-10 mL/kg of blood without any infusion of fluid to replace the withdrawn volume), or a combination of both.2 Phlebotomies were performed if patients had both a hemoglobin concentration above 85 g/L and normal renal function. They were interrupted if the mean arterial blood pressure dropped by more than 20% of the baseline value despite vasopressor administration. Phlebotomies were infused back at the end of surgery or earlier if needed clinically. Statistical Analyses The number of RBC units transfused were dichotomized in 2 categories because the authors considered, based on previous data, that transfusing more than 2 units is a more clinically significant outcome than being exposed to 2 units or less.2 Total intraoperative bleeding also was dichotomized at the median ( Z900 mL) for similar statistical analyses. Data are expressed as mean 7 standard deviation for continuous data or as a percentage for proportions. Statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC). Baseline and intraoperative characteristics were compared between the authors’ dichotomized outcome-based groups using Student t tests and Fisher exact tests. Univariate mixed logistic regression models (PROC GLIMMIX) then were used to assess the association of 16 different variables with transfusion of more than 2 RBCs and with bleeding above the median. To take into account the correlation within subjects (ie, because a subject can have multiple OLTs), a random intercept was included in the model. A mixed logit multivariate model then was used by incorporating the significant factors identified in the univariate analyses. Due to the small number of OLTs in the 4 2 unit RBC group (N ¼ 61), the number of independent variables to put into the final model was limited. The selection for both final models was based on a number of factors: eliminating all variables that were correlated and calculating the discriminant accuracy quantified in terms of the c-index for different combinations of potential independent variables. The regression coefficients from the mixed logit multivariable logistic regression model then were used to develop nomograms that predict the individual probability to receive more than 2 RBC units and to have a total intraoperative blood loss Z 900 mL.19,25 All variables of interest were tested for their linear effect. The linearity assumption was assessed by incorporating both the
3
variable of interest in its continuous and categorical form (ie, dichotomized at its quantiles) in the model (PROC GLIMMIX). For both outcomes (ie, transfusions and blood loss), the predictor variables in their categorical form were nonsignificant. That is, there is no additional effect of the variable beyond its linear effect. Subsequently, an internal validation of the nomogram to test its ability to correctly discriminate between those patients who received more or less than 2 units RBC was performed. The area under the receiver operating characteristic curve was used to quantify the nomogram’s discriminant properties and 800 bootstrap resamples were used to correct for over fit bias. Subsequently, the model’s calibration was tested. The LOESS smoother was used to plot the nomogram predicted probabilities of transfusing more or less than 2 RBC units or bleeding relative to the observed rate in the development cohort. Results Transfusion Exposure A total of 800 OLTs were performed ion 735 patients during the predefined period (679 patients had 1 OLT, 47 had 2 OLTs, and 9 had 3). Figure 1 shows the number of OLTs plotted against the number of RBC units transfused. The mean intraoperative transfusion of RBC units for all 800 cases was 0.6 7 1.4. Sixty-one transplantations (7.6%) required more than 2 RBC units, with a mean of 4.5 7 1.9 units in this subgroup. The remaining 739 transplantations (92.4%) who received 2 RBC units or less received a mean of 0.3 7 0.6 units. Six hundred and ten cases (76.3%) did not require any blood product. However, 14 OLTs (1.8%) required more than 5 RBC units. Population Characteristics and Outcomes The authors compared the demographic and health characteristics of the patients between their dichotomized outcomebased groups (Table 1), as well as their intraoperative characteristics (Table 2). As already mentioned, patients were divided between those requiring r 2 RBCs versus 42 RBCs
Fig 1. Orthotopic liver transplantation distribution by the number of RBC units transfused (total of 800 liver transplantations).
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Abbreviations: CTP, Child-Turcotte-Pugh; Hb, hemoglobin; INR, international normalized ratio; MELD, Model for End-Stage Liver Disease; OLT, orthotopic liver transplantation; RBC, red blood cell; REDO, reoperation (2nd or 3rd OLT). n Data available for only N ¼ 210 OLTs (N ¼ 187 and N ¼ 23 in the r2 RBC and 42 RBC groups, respectively; N ¼ 83 and N ¼ 127 in the r900 mL and 4900 mL groups, respectively).
33% 7 11 7 18 7 9 7 22 7 1.2 7 56 7 1.00 7 61 7 158 7 2.4 7 7 11% 52 77 170 101 2.0 85 1.81 105 137 10.4 24 33% 7 11 7 17 7 9 7 24 7 0.9 7 55 7 1.10 7 91 7 131 7 2.5 7 8 10% 51 75 169 111 1.7 98 2.32 103 106 9.3 21 0.0022 0.6213 0.4587 0.0077 o0.0001 0.0031 0.5223 0.6370 0.0062 0.0002 0.0007 0.0019 o0.0001 51% 7 12 7 18 7 10 7 12 7 1.5 7 55 7 1.17 7 68 7 229 7 2.4 7 8 26% 53 74 166 85 2.4 87 1.82 133 217 11.1 26 31% 711 7 17 7 9 7 23 7 1.0 7 56 7 1.05 7 77 7 135 7 2.5 7 9 9% 52 76 169 108 1.8 92 2.04 102 115 9.8 22 33% 7 11 7 17 7 9 7 23 7 1.0 7 56 7 1.06 7 77 7 146 7 2.5 7 9 10% 52 76 169 106 1.9 91 2.01 104 123 9.9 22 Sex (women) % Age (y) Weight (kg) Height (cm) Starting Hb value (g/L) Starting INR value Starting platelet count (109 pl/L) * Starting fibrinogen (g/L) Starting creatinine value (mmol/L) Starting bilirubin value (mmol/L) CTP score MELD score % REDO
p Value Z900 mL 420 OLTs o900 mL 379 OLTs p Value 42 RBC 61 OLTs r2 RBC 739 OLTs Total of 800 OLTs Variables
Table 1 Demographic and Health Characteristics for All 800 Liver Transplantations and for Patients Who (1) Received r2 RBC Units Versus 42 RBC Units and (2) Bled o900 mL Versus Z900 mL
0.9147 0.2056 0.0800 0.2936 o0.0001 0.0009 0.0045 0.0022 0.7906 0.0039 o0.0001 0.0001 0.5073
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
4
and between patients who bled o 900 mL versus Z900 mL. Patients who received more than 2 RBC units were smaller; were more often female; had more previous OLTs; had a lower starting hemoglobin value; had higher INR, creatinine, and bilirubin values; and had higher Child-Turcotte-Pugh (CTP) and Model of End-Stage Liver Disease (MELD) scores (the MELD score used was the available adjusted MELD score based on the United Network for Organ Sharing criteria). Both the starting CVP and the CVP before the vena cava clamping were higher in groups requiring more than 2 RBC units and with blood lost Z 900 mL. Transfusions of all labile blood products were higher in the same 2 groups (fresh frozen plasma, platelets, cryoprecipitate), as was the total volume of intravenous crystalloid, albumin 5%, and cell saver output infused. Even though the threshold for transfusion of RBCs was the same among the 4 groups, the final hemoglobin value was lower in the 42 RBCs and Z 900 mL groups. The percentage of patients having a phlebotomy performed before the anhepatic phase was lower in the groups with 42 RBC units transfused and Z900 mL blood lost. Patients who had a phlebotomy also had a lower mean blood lost (1,049 7 950 mL v 1,524 7 1,426 mL) and a lower mean of RBC transfusions (0.2 7 0.9 v 1.1 7 1.7) compared with those who did not have a phlebotomy (not shown in tables). Outcome Predictive Models Due to the small number of patients who received more than 2 RBCs, the number of independent variables to put into the final model was limited. Gender and height were highly correlated, so height was kept instead of gender. MELD and CTP scores were also correlated and the CTP score had a higher c-index (0.861 v 0.821). From theses 2 scores, the bilirubin value had a higher c-index (0.882) than either score alone (MELD: 0.821, CTP: 0.861). Components of the CTP and the MELD scores (bilirubin, INR, and creatinine) were included in the final model. Variables analyzed in a dichotomic way using the median were eliminated to keep only continuous variables (median of bilirubin, INR, creatinine, and hemoglobin). Table 3 summarizes the multivariate mixed logistic regression analysis for transfusion of “more than 2 RBCs”. Five variables were kept in the final model because of a significant association: patient’s height, baseline hemoglobin concentration, starting bilirubin value, CVP at the time of the vena cava clamping and the use of a phlebotomy (yes or no). Regarding the patients’ height, each increase of 1 centimeter (compared with the mean of 169 cm) is associated with a 5.3% decrease in the predicted odds of receiving more than 2 RBCs. Also, for each increase of 1 g/L of the starting hemoglobin value, the odds decrease by 5.0% and for each increase of 1 μmol/L of bilirubin concentration, the estimated odds of receiving more than 2 RBCs increases by 0.3%. The risk of multiple transfusion increases by 9.0% when the preclamping CVP increases by 1 mmHg above the mean (7.8 mmHg). Finally, the use of a phlebotomy decreases the odds of receiving more than 2 RBCs by 71.5% compared with “no phlebotomy.”
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Variables *
RBC transfused (units)
FFP transfused (units) Platelet transfused (units) Cryoprecipitate transfused (units) Albumin 5% (mL) Pentastarch or 6% hydroxyethyl starch (mL) † Hb trigger of RBC transfusion (g/L) Final Hb value (g/L) Blood loss (mL) Duration of surgery (min) % of use of phlebotomy % of use of the cell saver Diuresis (mL) Starting CVP (mmHg) CVP before clamping (mmHg) Crystalloid transfused (mL)
Total of 800 OLTs
r2 RBC 739 OLTs
42 RBC 61 OLTs
p Value
o900 mL 379 OLTs
Z900 mL 420 OLTs
p Value
0 ¼ 610 (76.3%) 1 ¼ 62 (7.7%) 2 ¼ 67 (8.4%) 3 ¼ 22 (2.7%) 4 ¼ 19 (2.4%) 5-12 ¼ 20 (2.5%) 0 ¼ 730 (91.4%) 40 ¼ 69(8.6%) 0 ¼ 774 (96.8%) 40 ¼ 26(3.2%) 0 ¼ 774 (96.9%) 40 ¼ 25 (3.1%) 0 ¼ 746 (93.3%) 40 ¼ 54 (6.7%) 0 ¼ 242 (31.2%) 40 ¼ 534 (68.8%) 59 7 10 93 7 20 1,256 7 1,203 244 7 63 56% 71% 419 7 314 13.1 7 4.8 7.8 7 3.8 4,084 7 1,607
0 ¼ 610 (82.5%) 1 ¼ 62 (8.4%) 2 ¼ 67 (9.1%)
3 ¼ 22 (36.1%) 4 ¼ 19 (31.1%) 5-12 ¼ 20 (32.8%)
N/App
r2 ¼ 373 (98.4%) 42 ¼ 6 (1.6%)
r2 ¼ 365 (86.9%) 42 ¼ 55 (13.1%)
o0.0001
0 ¼ 707 (95.8%) 40 ¼ 31(4.2%) 0 ¼ 730 (98.8%) 40 ¼ 9(1.2%) 0 ¼ 727 (98.5%) 40 ¼ 11 (1.5%) 0 ¼ 699 (94.6%) 40 ¼ 40 (5.4%) 0 ¼ 222 (31.1%) 40 ¼ 493 (69.0%) 60 7 9 93 7 20 1,080 7 801 239 7 57 60% 70% 427 7 315 12.9 7 4.6 7.7 7 3.7 3,909 7 1,322
0 ¼ 23(37.7%) 40 ¼ 38(62.3%) 0 ¼ 44 (72.1%) 40 ¼ 17(27.9%) 0 ¼ 47 (77.1%) 40 ¼ 14 (22.9%) 0 ¼ 47 (77.1%) 40 ¼ 14 (22.9%) 0 ¼ 20 (32.8%) 40 ¼ 41 (67.2%) 59 7 10 85 7 15 3,381 7 2,531 310 7 91 13% 87% 322 7 283 15.4 7 5.8 9.6 7 4.9 6,192 7 2,824
o0.0001
0 ¼ 375 (98.9%) 40 ¼ 4(1.1%) 0 ¼ 377 (99.5%) 40 ¼ 2(0.5%) 0 ¼ 378 (99.7%) 40 ¼ 1 (0.3%) 0 ¼ 371 (97.9%) 40 ¼ 8 (2.1%) 0 ¼ 121 (33.4%) 40 ¼ 241 (66.6%) 62 7 8 97 7 20 554 7 190 229 7 54 65% 56% 425 7 296 12.6 7 4.3 7.3 7 3.5 3,637 7 1,153
0 ¼ 355 (84.7%) 1 ¼ 64(15.3%) 0 ¼ 396 (94.3%) 40 ¼ 24(5.7%) 0 ¼ 395 (94.3%) 40 ¼ 24 (5.7%) 0 ¼ 374 (89.1%) 40 ¼ 46 (10.9%) 0 ¼ 121 (29.3%) 40 ¼ 292 (70.7%) 59 7 10 88 7 18 1,889 7 1370 258 7 70 49% 85% 414 7 329 13.6 7 5.1 8.3 7 4.0 4,491 7 1,839
o0.0001
o0.0001 o0.0001 o0.0001 0.7764 0.4036 0.0033 o0.0001 o0.0001 o0.0001 0.0100 0.0379 0.0014 0.0017 o0.0001
Abbreviations: CVP, central venous pressure; FFP, fresh frozen plasma; Hb, hemoglobin; N/App, non applicable; OLT, orthotopic liver transplantation; RBC, red blood cells. n No RBC transfused in N ¼ 338/373 and N ¼ 271/365 for the r900 mL and 4900 mL groups, respectively. † Data available for only N ¼ 188 OLTs (N ¼ 127 and N ¼ 61 in the r2 RBC and 42 RBC; N ¼ 41 and N ¼ 147 in the r900 mL and 4900 mL groups, respectively).
0.0010 0.0022 o0.0001 0.2198 0.3035 o0.0001 N/App o0.0001 o0.0001 o0.0001 0.6224 0.0038 0.0004 o0.0001
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
Table 2 Surgical Characteristics for All 800 Liver Transplantations and for Patients Who (1) Received r2 RBC Units Versus 42 RBC Units and (2) Bled o900 mL Versus Z900 mL
5
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
6
Table 3 Summary of the Multivariate Logistic Regression Model and Odds Ratio for More than 2 RBC Transfusions Variables
Estimate
Height –0.05427 Starting Hb –0.05085 Starting bilirubin 0.002787 Phlebotomy –1.2545 Pre-clamping CVP 0.08656
SE
0.01733 0.01434 0.001038 0.4705 0.03952
Odds Ratio
0.947 0.950 1.003 0.285 1.090
95% Wald CI Lower
Upper
0.916 0.924 1.001 0.113 1.009
0.980 0.978 1.005 0.717 1.178
Abbreviations: CI, confidence interval; CVP, central venous pressure; Hb, starting hemoglobin value; SE, standard error.
To analyze bleeding above the median (Z 900 mL), all available variables were included in the model due to the greater number of patients (420 patients). Table 4 summarizes the multivariate analysis for bleeding Z900 mL. Three variables were associated with outcome: starting hemoglobin value, CTP score, and CVP at the time of the vena cava clamping. For each 1 g/L increase in starting hemoglobin value above the mean, the risk of bleeding Z900 mL decreased by 1.6%. For each increase of 1 point on the CTP score, the risk of bleeding increased by 12%. Finally, the risk increased by 5.7% for each 1 mmHg increase in CVP at the time of clamping. Nomogram-Based Individual Outcome Prediction Figure 2 shows the nomogram itself that predicts individual probabilities of being transfused more than 2 RBC units. A low starting hemoglobin value represents the most powerful predictor of the transfusion risk. Two examples summarize this nomogram: a patient with height of 178 cm, with a starting hemoglobin value at 140 g/L, a bilirubin value at 50 μmol/L, with a phlebotomy, and a CVP value of 5 mmHg at the time of the vena cava clamping predicts a near zero probability of receiving more than 2 RBC units. At the opposite end of the spectrum, a patient with height of 150 cm, a starting hemoglobin of 70 g/L, a serum bilirubin concentration of 300 μmol/L, a CVP of 15 mmHg, and no phlebotomy has a predicted 80% risk of receiving more than 2 RBC units. Figure 3 shows the nomogram that predicts the risk of Table 4 Summary of the Multivariate Logistic Regression Model and Odds Ratio for Blood Loss, More or Less than the Median (900 mL) Variables
Starting Hb CTP score Preclamping CVP
Estimate
–0.01599 0.1136 0.05553
SE
0.00369 0.03264 0.02084
Odds Ratio
0.984 1.120 1.057
95% Wald CI Lower
Upper
0.977 1.051 1.015
0.991 1.194 1.101
Abbreviations: CI, confidence interval; CTP, Child-Turcotte-Pugh; CVP, central venous pressure; Hb, starting hemoglobin value; SE, standard error.
bleeding 900 mL or more. Starting hemoglobin value, CTP score, and CVP at the time of clamping were the significant variables. For example, a patient with a starting hemoglobin value of 100 g/L, a CTP score of 12, and a CVP value at 12 mmHg at the time of the vena cava clamping has a 65% risk of bleeding 900 mL or more. Discussion Transfusions of blood products are associated with increased morbidity and mortality. As previously suggested, the prophylactic correction of coagulation laboratory abnormalities reduces neither bleeding nor the need for transfusion and might contribute to morbidity, as intraoperative RBC transfusions were associated with a decrease in 1-year survival among liver transplant recipients.2,7,26 Indeed, prevention of excessive blood loss and multiple transfusions has become a major improvement goal in the perioperative management of liver transplantation patients.2,8–10 The main objective of this work was to describe the authors’ liver transplantation population further and to better identify, based on baseline and intraoperative characteristics, which patients will bleed more and require multiple blood transfusions. Data on 800 consecutive OLTs were available in the database for the authors’ predefined observational period. In this paper, the authors showed that their population still has a low transfusion exposure compared with many other liver transplantation populations described in the literature.3–5 Based on available variables, the authors built multivariate predictive models for multiple transfusions and major bleeding. The authors showed that preoperative characteristics, such as height, anemia, and severity of liver disease, are associated with a higher risk of major bleeding and a higher risk of being transfused more than 2 RBC units. The use of a phlebotomy prior to the anhepatic phase is associated with a reduced risk of receiving multiple RBC transfusions. The authors’ strongest predictor for major bleeding and multiple transfusions is a high CVP prior to vena cava clamping, based on the strength of association in the logistic regression model. Finally, the authors created nomograms that quantify the individual risk of multiple RBC transfusions and major bleeding. These nomograms represent a graphical display of the effect of several risk factors on the outcomes and provide the individual probability of receiving more than 2 RBC units or bleeding more than 900 mL in a 0% to 100% scale. Such a format is substantially more intuitive than odd ratios. These nomograms are tools that combine the most significant and informative predictors of multiple transfusions and major bleeding within 1 display. They will help target higher risk patients for interventions and further studies.21,25 This study has many limitations. First, only baseline and intraoperative characteristics gathered in the authors’ database were included in the analyses. However, these characteristics are considered important by many liver transplantation programs.3–5 Other variables could have been included, such as variables from medical history; patient hospitalization status; donor characteristics, including the donor risk index; and other
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
7
Points
Height
Starting Hemoglobin
Pre-op Bilirubin No
Phlebotomy Yes
CVP
Total Points
Predicted Value
Fig 2. Nomogram predicting transfusion of more than 2 units of red blood cells.
intraoperative characteristics and interventions. Second, only bleeding and transfusion outcomes were analyzed in this study. Many members of the liver transplantation community consider bleeding and transfusions being important surrogate outcomes that already have been associated with more longterm significant outcomes, such as mortality.2,9,16,20 However, other outcomes could have been considered, such as acute renal failure, graft complications, pulmonary and infectious complications, and burden of care outcomes, such as need for
postoperative organ support, length of stay, and readmissions. Predictive models associated for such intermediate and longterm outcomes probably should be done in future studies. Third, the specific role of phlebotomy is not well defined because it is performed only in lower risk patients, introducing many potential confounders. These confounders might be hard to identify and correct in any model. Furthermore, the authors used logistic regression models instead of linear ones to classify their cohort in more clinically significant categorized
Points
Starting Hb
CTP score
CVP clamping
Total Points
Probability
Fig 3. Nomogram predicting the risk of bleeding more than the median (900 mL). Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
8
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
outcomes. The size of their cohort allowed sufficient statistical power to do this. However, the authors looked at 2 outcomes and did multiple statistical tests; some of their findings could be chance alone. Finally, the predictive models and the nomograms presented here were developed within a single cohort and should be validated prospectively in an external cohort with similar bleeding and transfusion rates. However, the authors previously published a similar model to predict transfusion of 1 or more RBC units in a cohort of 400 patients that was prospectively validated in a second cohort. Because the authors obtained similar results in this study, these results emphasize the significance of some of those predictors in their population.21 Of the 5 variables linked to transfusion of more than 2 RBC units, 2 are not modifiable: the patient’s height and the starting serum bilirubin concentration. Smaller patients probably have a greater risk of receiving 42 RBC units because a similar blood loss is higher compared with their own blood volume. The MELD score is used widely to prioritize recipients for OLT, but some of the authors’ previous work suggested that this score predicts neither bleeding nor transfusions.27 Nevertheless, this study suggests that 1 variable part of the MELD score, the baseline bilirubin value, is associated with multiple transfusions. Therefore, hyperbilirubinemia itself seems to be a predictive factor for transfusions. The 3 other significant variables are either modifiable characteristics or an intervention itself. The authors’ previous work, based on smaller cohorts, identified 2 predictors of RBC transfusions: the nonuse of a phlebotomy and the starting hemoglobin value.21 Then, transfusion of RBC units was analyzed as a dichotomic outcome divided between no RBC transfusion and Z 1 RBC unit transfused and the baseline hemoglobin value was a stronger predictor for transfusions than any traditional disease severity score (CTP and MELD score).21 The present study focused on OLTs requiring transfusions of more than 2 RBC units, based on the fact that an intraoperative transfusion greater than 2 RBC units is a more significant exposure and represents a subgroup of patients that might benefit from blood sparing interventions.2 Here again, baseline hemoglobin value is a major determinant of multiple transfusions and severe blood loss. Such a significant prognostic role of the preoperative hemoglobin value could suggest that any intraoperative intervention to reduce transfusions might be limited. However, the use of a phlebotomy still is independently associated with a reduced risk of transfusion, suggesting that therapeutic interventions may be possible, even though this association might be inaccurate due to multiple confounders.2 Finally, preclamping CVP appears to be a powerful and significant predictor for both outcomes in this study. This finding suggests the role of volume overload in the risk of bleeding and transfusion in this population. It also emphasizes the possible mechanistic pathway of phlebotomy that may reduce portal pressure and sometimes CVP through a decrease in volume overload, as already suggested by the authors’ group.28,29 Some of the variables independently associated with the outcomes in the authors’ multivariate model are clinically correlated when it comes to interventions. Improvement of the
baseline hemoglobin value could be part of a preoperative optimization strategy and artificially increased during the waiting period prior to the OLT. However, a RBC transfusion close to the surgery might contribute to an increase in CVP, a harmful factor in this study. The value of such RBC prophylactic transfusions earlier in the waiting period should be explored further in clinical trials. As already mentioned, benefits of phlebotomy might be influenced by confounding factors reflecting patients’ better overall condition because it is performed only when hemoglobin value is higher than 85 g/L and when patients have normal renal function. However, an independent benefit from phlebotomy has been suggested in this study and in previous work, where patients who underwent a phlebotomy bled less, even if they had identical preclamping CVP.2 As already suggested, this effect might come from an isolated decrease in portal venous pressure, even though a high CVP at the time of the vena cava clamping might be associated with a higher portal pressure.30 This high CVP might be secondary to preoperative volume overload, a preanhepatic liberal fluid resuscitation strategy (high amounts of crystalloids, colloids, or blood products), heart failure, or any other cause of overall volume overload, including not having received a phlebotomy. The independent clinical role of the aforementioned variables and interventions must be explored further. Indeed, the optimal combination of various interventions, including early preoperative RBC transfusion and a restrictive fluid resuscitation strategy that encompasses phlebotomy and will minimize bleeding, reduce transfusions, and improve outcomes of liver transplant recipients remains to be determined. However, the external validity of these findings might be limited by the fact that many liver transplantation programs observe much higher transfusion rate. The authors’ findings probably may be extrapolated only to other low bleeding populations. Nonetheless, this study provides preliminary data to support further studies and clinical trials on the impact of aggressive preoperative correction of anemia and different perioperative fluid strategy on postoperative outcomes.
Conclusion This observational study of 800 consecutive OLTs was performed in a cohort with few patients having multiple RBC transfusions (7.6% received 42 RBC units). In this cohort, 5 variables were associated with transfusion of 2 RBC units: patient’s height, starting hemoglobin value, starting bilirubin value, the nonuse of a phlebotomy, and CVP at the time of the vena cava clamping. From these 5 variables, 2 were also associated with bleeding of Z900 mL: hemoglobin and CVP. Nomograms were developed, based on these variables, to predict each individual patient’s risk of receiving more than 2 RBCs and having major bleeding. This study will help predict risk of multiple transfusions and major bleeding and provides preliminary data to support future investigations on possible interventions that could improve postoperative outcomes in a low bleeding liver transplantation population.
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011
L. Massicotte et al. / Journal of Cardiothoracic and Vascular Anesthesia ] (]]]]) ]]]–]]]
Acknowledgments The authors thank Mrs Nadine Gaudreau for her secretarial work. Appendix I. – Local Transfusion Guidelines The thresholds for transfusing blood products in the authors’ center were Red blood cells: hemoglobin between 60 and 70 g/L Fresh frozen plasma (10-15 mL/kg): INR 4 1.5 Platelets (5-10 units): platelet count r30 109/L Cryoprecipitate (5-10 units): fibrinogen r 2 g/L All procoagulant blood products were used only to treat significant bleeding and not to “correct” laboratory values. References 1 Butler P, Israel L, Nusbacher J, et al. Blood transfusion in liver transplantation. Transfusion 1985;25:120–3. 2 Massicotte L, Thibeault L, Roy A. Classical notions of coagulation revisited in relation with blood losses, transfusion rate for 700 consecutive liver transplantations. Semin Thromb Hemost 2015;41:538–46. 3 Roullet S, Biais M, Millas E, et al. Risk factors for bleeding and transfusion during orthotopic liver transplantation. Ann Fr Anesth Reanim 2011;30:349–52. 4 McCluskey SA, Karkouti K, Wijeysundera DN, et al. Derivation of risk index for the prediction of massive blood transfusion in liver transplantation. Liver Transpl 2006;12:1184–93. 5 Cywinski JB, Alster JM, Miller C, et al. Prediction of intraoperative transfusion requirements during orthotopic liver transplantation and the influence on postoperative patient survival. Anesth Analg 2014;118: 428–37. 6 Tripodi A, Mannucci PM. The coagulopathy of chronic liver disease. N Engl J Med 2011;365:147–56. 7 Lisman T, Porte RJ. Value of preoperative hemostasis testing in patients with liver disease for perioperative hemostatic management. Anesthesiology 2017;126:338–44. 8 Shan WL, Barkun J, Metrakos P, et al. Blood product use during orthotopic liver transplantation. Can J Anaesth 2004;51:1045–6. 9 Massicotte L, Beaulieu D, Thibeault L, et al. Coagulation defects do not predict blood product requirements during liver transplantation. Transplantation 2008;85:956–62. 10 Ozier Y, Klinck JR. Anesthetic management of hepatic transplantation. Curr Opin Anaesthesiol 2008;21:391–400. 11 Massicotte L, Denault AY, Beaulieu D, et al. Transfusion rate for 500 consecutive liver transplantations: Experience of one liver transplantation center. Transplantation 2012;93:1276–81. 12 Pereboom IT, de Boer MT, Haagsma EB, et al. Platelet transfusion during liver transplantation is associated with increased postoperative mortality due to acute lung injury. Anesth Analg 2009;108:1083–91.
9
13 Boyd SD, Stenard F, Lee DK, et al. Alloimmunization to red blood cell antigens affects clinical outcomes in liver transplant patients. Liver Transpl 2007;13:1654–61. 14 Hendriks HG, van der Meer J, de Wolf JT, et al. Intraoperative blood transfusion requirement is the main determinant of early surgical reintervention after orthotopic liver transplantation. Transpl Int 2005;17: 673–9. 15 Boin IF, Leonardi MI, Luzo AC, et al. Intraoperative massive transfusion decreases survival after liver transplantation. Tranpl Proc 2008;40:789–91. 16 Benson AB, Burton Jr JR, Austin GL, et al. Differential effects of plasma and red blood cell transfusions on acute lung injury and infection risk following liver transplantation. Liver Transpl 2011;17:149–58. 17 Massicotte L, Sassine MP, Lenis S, et al. Survival rate changes with transfusion of blood products during liver transplantation. Can J Anaesth 2005;52:148–55. 18 Porte RJ, Molenaar IQ, Begliomini B, et al. Aprotinin and transfusion requirements in orthotopic liver transplantation: A multicentre randomised double-blind study. EMSALT study group. Lancet 2000;355:1303–9. 19 Massicotte L, Denault AY, Beaulieu D, et al. Aprotinin versus tranexamic acid during liver transplantation: Impact on blood product requirements and survival. Transplantation 2011;91:1273–8. 20 Molenaar IQ, Wanaar N, Groen H, et al. Efficacy and safety of antifibrinolytic drugs in liver transplantation: A systematic review and meta-analysis. Am J Transplant 2007;7:185–94. 21 Gurusamy KS, Pissanou T, Pikhart H, et al. Methods to decrease blood loss and transfusion requirements for liver transplantation. Cochrane Database Syst Rev CD009052, 2011. 22 Massicotte L, Umberto C, Beaulieu D, et al. Independent validation of a model predicting the need for packed red blood cell transfusion at liver transplantation. Transplantation 2009;88:386–91. 23 Von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies. J Clin Epidemiol 2008;6: 344–9. 24 Royston D, Bidstrup BP, Taylor KM, et al. Effect of aprotinin on need for blood transfusion after repeat open-heart surgery. Lancet 1987;2: 1289–91. 25 Ferguson DA, Hébert PC, Mazer CD, et al. A comparison of aprotinin and lysine analogues in high-risk cardiac surgery. N Engl J Med 2008;358: 2319–31. 26 Capitano U, Jeldres C, Shariat SF, et al. Clinicians are more familiar with nomograms and rate their clinical usefulness highest, look-up tables are second best. Eur Urol 2008;54:958. 27 Reyle-Hahn M, Rossaint R. Coagulation techniques are not important in directing blood product transfusion during liver transplantation. Liver Transpl 1997;3:663–5. 28 Massicotte L, Beaulieu D, Roy JD, et al. MELD score and blood product requirements during liver transplantation: No link. Transplantation 2009;87:1689–94. 29 Massicotte L, Lenis S, Thibeault L, et al. Effect of low central venous pressure and phlebotomy on blood product transfusion requirements during liver transplantations. Liver Transpl 2006;12:117–23. 30 Massicotte L, Perrault MA, Denault AY, et al. Effects of phlebotomy and phenylephrine infusion on portal venous pressure and systemic hemodynamics during liver transplantation. Transplantation 2010;89: 920–7.
Please cite this article as: Massicotte L, et al. (2017), http://dx.doi.org/10.1053/j.jvca.2017.10.011