Home > Subjects > Transplant > Prediction of Intraoperative Transfusion Requirements During...
Anesthesia & Analgesia:
doi: 10.1213/ANE.0b013e3182a76f19
General Articles: Research Report

Prediction of Intraoperative Transfusion Requirements During Orthotopic Liver Transplantation and the Influence on Postoperative Patient Survival

Cywinski, Jacek B. MD*; Alster, Joan M. MS; Miller, Charles MD; Vogt, David P. MD; Parker, Brian M. MD*

Free Access
Article Outline
Collapse Box

Author Information

From the Departments of *General Anesthesiology and Transplant Center, Quantitative Health Sciences, and Hepato-pancreato-biliary and Transplant Surgery, Cleveland Clinic, Cleveland, Ohio.

Accepted for publication July 17, 2013.

Funding: Supported by internal funds.

The authors declare no conflicts of interest.

Reprints will not be available from the authors

Address correspondence to Brian M. Parker, MD, Cleveland Clinic, 9500 Euclid Ave., Cleveland, OH 44195. Address e-mail to parkerb1@ccf.org.

Collapse Box

Abstract

BACKGROUND: Predicting blood product transfusion requirements during orthotopic liver transplantation (OLT) remains difficult. Our primary aim in this study was to determine which patient variables best predict recipient risk for large blood transfusion requirements during OLT. The secondary aim was to determine whether the amount of blood products transfused during OLT impacted patient survival.

METHODS: Eight hundred four primary adult OLTs performed during a 9-year period were retrospectively analyzed, and predictive models were developed for blood product usage, usage >20 and usage >30 units of red blood cells (RBCs) plus cell salvage (CS). For survival analysis, potential predictors included all blood products administered during OLT.

RESULTS: For analyses of RBC + CS usage, we used several statistical techniques: regression analysis, logistic regression, and classification and regression tree analysis. Several preoperative factors were highly statistically significant predictors of intraoperative blood product usage in each of the analyses, namely lower platelet count and higher Model for End-Stage Liver Disease Score or one or more of its components (creatinine, total bilirubin, international normalized ratio). Despite these highly significant associations, the models were unable to predict reliably that patients might require the largest amount of blood products during OLT. For example, the classification and regression tree analyses were able to predict only 32% and 11% of patients requiring >20 and >30 units of RBC + CS, respectively. Survival analysis demonstrated poorer survival among patients receiving larger amounts of RBC + CS during OLT.

CONCLUSION: Prediction of intraoperative blood product requirements based on preoperatively available variables is unreliable; however, there is a strong measurable association between transfusion and postoperative mortality.

Predicting blood product transfusion requirements during orthotopic liver transplantation (OLT) has remained difficult. Because OLT is so resource intensive, understanding which variables predict transfusion requirements could allow improved blood product allocation and utilization. Data published both before, and after, the use of the Model for End-Stage Liver Disease (MELD) Score have shown select preoperative factors to be only minimally predictive of intraoperative transfusion requirements.1–4 Some of the preoperative variables reported in the literature affecting intraoperative transfusion requirements include: recipient age and weight, etiology of liver disease, starting hemoglobin level, prothrombin international normalized ratio (INR) and partial thromboplastin time, platelet count, serum creatinine (Cr), and presence of ascites. Unfortunately, many of these variables have not consistently been found to be associated with transfusion needs during OLT.

Higher transfusion requirements during OLT are associated with worse postoperative outcomes including prolonged length of intensive care unit stay, hospital stay, and increased mortality;5 however, these relationships have not been well characterized.6,7 Furthermore, an association rather than causality was established in these studies with varying strengths of correlation. Interestingly, when survival outcomes are compared among centers with different rates of blood product transfusion, centers with higher blood product administration rates do not necessarily have worse survival rates. This apparent anomaly highlights the fact that the relationship between transfusion and outcomes is complex and most likely dependent on factors unique to the particular transplant center.8

Recognizing that it may be very difficult to develop a single, reliable, and universally applicable model to predict transfusion requirements for patients undergoing OLT, the primary aim of this study was to determine which preoperatively available variables best predict recipient blood transfusion requirements during OLT in our program, with particular emphasis on predicting high transfusion requirements (>20 or >30 units of blood). Such models could identify the patients who consume the majority of the blood bank resources and possibly be modified to fit the practice of other institutions. The secondary aim of this study was to determine whether the amount of blood products transfused impacted survival after OLT.

Back to Top | Article Outline

METHODS

After receiving IRB approval which waived the need for informed consent, 835 consecutive primary adult (age ≥18 years) OLTs performed from January 1, 2001 to June 30, 2010 at the Cleveland Clinic were retrospectively analyzed. During this time frame, 47 retransplants were performed; in these cases, we analyzed data from the first transplant only. Patients undergoing combined liver and other organ transplants (liver-kidney n = 36, liver-heart n = 3, liver-lung n = 3, liver-pancreas n = 2, liver-pancreas-small intestine n = 1) as well as split liver graft (n = 13) were included in the analysis. In addition, 24 living donor liver transplants were performed during this period. None of the patients undergoing retransplantation or living donor liver transplants was included in the analyses nor were 7 transplants for which we did not have blood product usage information. This resulted in a final analysis of 804 OLTs.

All blood products administered intraoperatively to OLT recipients were analyzed, including allogeneic red blood cells (RBCs), fresh frozen plasma (FFP), cryoprecipitate, platelets, and intraoperative red blood cell salvage (CS) units returned to the patient. A predictive model was developed for RBC + CS (using RBC equivalents) usage during liver transplantation. Stepwise multivariable regression analysis was used to identify recipient, donor, and transplant procedure variables associated with blood product usage. The natural logarithm of (RBC + CS) was used to produce a nearly normally distributed analysis variable with stable error variance. Bootstrap aggregation (bagging) was used for variable selection,9 with P < 0.07 for entry into the models and P < 0.05 for variable retention. Variables appearing in at least 50% of bootstrap analyses were considered reliably statistically significant at P < 0.05 (median rule). Several transformations of continuous predictors (x2, 1/x, 1/x2, sqrt[x], ln[x]) were plotted against ln(RBC + CS) to check whether transformation could result in more nearly linear relationships between the predictors and ln(RBC + CS). Since the transformations did not result in significantly improved linearity, the predictors were analyzed on their original scales. In addition, goodness of fit of the regression model was confirmed by plotting residuals versus predicted values. To predict large transfusion requirements, use of >20 and >30 units of RBC + CS (approximately top 20% and 10% of blood product needs in our study) were modeled using logistic regression and bootstrap aggregation, as above. Model fit was confirmed using Hosmer-Lemeshow goodness of fit tests.

The large transfusion requirement models were further investigated using bootstrap forest partitioning methods with 100 trees in the forest, 16 terms sampled per split, and minimum split size of 5. Twenty percent of the population sample was withheld for validation purposes. The 15 most important classification variables found by the bootstrap forest technique were compared with results from the logistic regression analyses, and the top 10 of these were also input to a classification and regression tree (CART) analysis. Resulting trees for predicting >20 and >30 units of RBC + CS were depicted.10 The CART analyses used 8 steps with likelihood ratio χ2 tests used to determine the best splitting variables at each step. No minimum number of cases were required for final branches in the trees.

Risk unadjusted survival was estimated nonparametrically by the method of Kaplan and Meier and parametrically by a multiphase hazard decomposition method.11 Nonproportional multiphase, multivariable hazard methodology11 was used to identify recipient, donor, and transplant procedure variables associated with each hazard phase simultaneously. Bootstrap aggregation (bagging) was used for variable selection as described above.

Analysis of intraoperative RBC + CS usage included recipient factors (age, gender, race, height, weight, body mass index, body surface area, blood type, Rh factor, primary diagnosis [primary cause of liver failure], previous abdominal surgery [yes/no], cytomegalovirus status, transjugular intrahepatic portosystemic shunt, total bilirubin, serum creatinine, assigned MELD Scorea, pretransplant INR, and platelets), donor factors (age, gender, race, expanded criteria donor, cause of death, circumstance of death, and mechanism of death), and procedure variables (surgeon, date of transplant, liver-only transplant, whole liver transplant). Survival analysis incorporated the above variables plus donor blood type and Rh factor, donor/recipient matching variables (gender match [male to male, female to female, female to male, male to female], Rh compatibility, and length of surgery [time from the surgical incision to closure]), graft cold ischemic time, and blood products given during transplantation.

Back to Top | Article Outline

RESULTS

Recipient, donor, and transfusion data are presented in Tables 1–4. Cryoprecipitate was not analyzed due the majority of this product being given to 3 patients who received massive transfusions and were later dropped from further survival analysis. Ten surgeons performed the OLTs with minimum 29, and maximum 122 transplants performed by each.

Table 1
Table 1
Image Tools
Table 2
Table 2
Image Tools
Table 3
Table 3
Image Tools
Table 4
Table 4
Image Tools
Back to Top | Article Outline
RBC + CS

The following factors were associated with increased RBC and CS administration during OLT (Table 5): older recipient age (P = 0.004); lower platelet count (P < 0.0001); greater recipient pretransplant INR (P < 0.0001), total bilirubin (P = 0.007), and Cr (P = 0.0005); and expanded criteria donor (P = 0.03). Two factors were associated with decreased combined RBC and CS administration: recipient diagnosis of hepatocellular carcinoma (P < 0.0001) and Surgeon X (P < 0.0001). The r2 for the model was small, with these predictors accounting for only 22% of the variance in amount of blood product needed.

Table 5
Table 5
Image Tools

Analyses of blood product requirements >20 units and >30 units of RBC + CS during OLT (approximate 80th and 90th percentiles of usage) are shown in Tables 6–11. Since the 75 patients with RBC + CS >30 were also among the 156 patients with RBC + CS >20, we expected some compatibility in the models developed.

Table 6
Table 6
Image Tools
Table 7
Table 7
Image Tools
Table 8
Table 8
Image Tools
Table 9
Table 9
Image Tools
Table 10
Table 10
Image Tools
Table 11
Table 11
Image Tools
Back to Top | Article Outline
RBC + CS >20

Logistic regression analysis found the following factors to be associated with increased odds of requiring >20 units of RBC + CS (Table 6): higher levels of pretransplant Cr (P < 0.0001) and INR (P = 0.003), lower pretransplant platelet count (P = 0.002), previous abdominal surgery (P = 0.003), blood type A (P = 0.007), and surgeon Y (P = 0.009). Model c (concordance) statistic = 0.70 (95% CI, 0.66–0.75).

Table 7 shows univariate descriptive summaries of the logistic regression predictors by RBC + CS >20 units.

Table 8 shows the 15 most important variables found by bootstrap forest analysis of RBC + CS >20 units. There is good overlap in the 2 analyses: the logistic regression variables Cr, platelet count, and INR are the top 3 variables used for partitioning in the bootstrap forest analysis, and previous abdominal surgery is the 15th ranked variable. Blood type and surgeon were not important in the bootstrap forest analysis so may be of lesser importance in predicting RBC + CS >20. As the 15th ranked variable, previous abdominal surgery may also be of only marginal importance. The CART analysis yielded the representative tree shown in Figure 1 with the following variables used in the 8 partitions: Cr, total bilirubin, INR, MELD Score, platelet count, and body weight. After the 8 partitions, 620 of 648 (96%) patients with RBC + CS ≤20 were correctly classified, and 50 of 156 (32%) patients with RBC + CS >20 were correctly classified. Classification improves with more partitions; however, trees with many partitions can be difficult to interpret and likely would not replicate well on new data.

Figure 1
Figure 1
Image Tools
Back to Top | Article Outline
RBC + CS >30

Logistic regression analysis found the following factors to be associated with increased odds of requiring >30 units of RBC + CS (Table 9): higher assigned MELD Score (P = 0.004), lower pretransplant platelet count (P = 0.006), surgeon Y (P = 0.004), and surgeon Z (P = 0.004). Model c (concordance) statistic = 0.67 (95% CI, 0.60–0.73).

Table 10 shows univariate descriptive summaries of the logistic regression predictors by RBC + CS >30 units.

Table 11 shows the 15 most important variables found by bootstrap forest analysis of RBC + CS >30 units. There is moderate consistency in the 2 analyses: the logistic regression variables platelet count and MELD Score were the third and fourth most important variables used for partitioning in the bootstrap forest analysis; however, the bootstrap forest analysis found Cr and age to be of greater importance than platelet count or MELD Score. The bootstrap forest method identified Surgeon Y as only the 15th most important variable and did not identify Surgeon Z at all. Note that the bootstrap forest method identified the interval from January 1, 2001 to transplant as important. Surgeon Y’s transplants were mostly performed much later than average in the sequence of transplants, so there is confounding between surgeon and date of transplant. It is not surprising that 2 analytic methods might choose differently among the confounded factors. The CART analysis yielded the representative tree shown in Figure 2 with the following variables used in the 8 partitions: Cr, assigned MELD Score, height, platelet count, years since January 1, 2001, and age. After the 8 partitions, 727 of 729 (99.7%) patients with RBC + CS ≤30 were correctly classified; however, only 8 of 75 (11%) patients with RBC + CS >30 were correctly classified. Creating a tree with more splits did not improve prediction appreciably.

Figure 2
Figure 2
Image Tools
Back to Top | Article Outline
Survival Analysis

Mean follow-up time for survivors in the study was 3.6 ± 2.3 years. Because the goal of the analysis was to assess whether the quantity of blood products transfused affected survival, it was important to consider whether high numbers of units may have been transfused at the time of transplant in order to rescue dying patients. If so, this could bias the conclusions toward finding association of transfusions with death. Indeed, we found that among the 10 patients who died within a day of transplantation, 8 had 40 or more units of RBC + CS transfused, placing them in the top 5% of RBC + CS usage among study patients. Cases that resulted in mortality within the first postoperative day (10 patients) were excluded from the primary survival analysis. However, the analysis was repeated with the 10 patients included to assure that the results from the 2 analyses were compatible.

Instantaneous risk of death after liver transplantation consisted of an early hazard phase of greatest risk, followed, after approximately 9 months, by a late hazard phase of nearly constant risk (Figs. 3, A and B). The hazard of death peaked at about 2 weeks posttransplant. The results of the multivariable hazard modeling are shown in Table 12. More units of blood transfused, greater pretransplant INR, older donor age and donor race other than African American or Caucasian were all associated with a greater hazard of death in the early posttransplant period. A primary diagnosis of viral hepatitis and disease etiology of liver cancer were associated with a greater hazard of death in the ensuing late hazard phase. The effects of donor race and units of RBC + CS transfused on predicted 1-year survival are shown graphically in Figure 4. Kaplan-Meier unadjusted survival curves by units of RBC + CS transfused are shown in Figure 5. As demonstrated in Figure 5, patients receiving >15 units of RBC + CS have markedly worse survival after OLT.

Table 12
Table 12
Image Tools
Figure 3
Figure 3
Image Tools
Figure 4
Figure 4
Image Tools
Figure 5
Figure 5
Image Tools
Back to Top | Article Outline

DISCUSSION

During the evolution and establishment of OLT as the most effective treatment option for end-stage liver disease, there has been significant interest in the literature in elucidating the predictors of blood transfusion during this surgery.2,3 Intraoperative blood loss and subsequent blood transfusion can influence multiple postoperative outcomes (i.e., need for re-exploration, incidence of sepsis, poorer patient and graft survival); however, cause and effect relationships have not been clearly established.

The difficulties in predicting transfusion requirements during OLT are many. Quantification of intraoperative surgical conditions (intra-abdominal adhesion, venous collaterals, etc.) is considered by some to be very important but may be difficult to anticipate in advance.2 These factors, although intuitively easy to appreciate, do not reliably translate into intraoperative blood loss and transfusion requirements. Previous investigations have attempted to identify preoperative predictors of blood transfusion; however, the predictive value of preoperative variables and patient characteristics were inconsistent and weak at best.2,3 Transfusion requirements depend not only on the intraoperative blood loss but also on the threshold for when transfusions of different products are initiated. Hevesi et al.12 demonstrated that requirements for RBC and FFP can be reduced almost 2- and 3-fold, respectively, if the anesthesia team universally followed protocols including goal-directed transfusion practices. Therefore, comparison of intraoperative transfusion requirements from different transplant centers may be inherently biased by an inability to account for differences in transfusion triggers and clinical practices. Consequently, the predictive models developed in one institution may hardly, if ever, be applicable in the others. For example, Massicotte et al developed a predictive model to estimate the probability of RBC transfusion based on 3 individually weighted risk factors derived from their cohort of OLT patients (transfusion of FFP, inability to perform phlebotomy, and starting hemoglobin). Despite the high c-statistic of that model, its applicability in other centers is questionable because the rate of transfusion in the program where the model was developed was extremely low with only 19.5% of the patients transfused at all.4 In contrast, many other centers report significantly higher transfusion rates.

Our results demonstrate that a model created to predict transfusion requirements for patients undergoing OLT is unreliable, and hence, the practical utility remains minimal. Although there was substantial overlap in the sets of variables that predicted >20 and >30 units of RBCs among our models, no set of variables provided good prediction for large transfusion requirements. The CART analysis model was best in identifying patients with intraoperative RBC + CS requirements <20 units, which clinically may not be very useful to facilitate allocation of blood bank resources in advance of OLT. Thus, it seems that construction of a reliable algorithm for prediction of intraoperative blood product requirements, even in the setting of fairly uniform practice (single institution) is difficult, if not impossible. Of multiple analyzed variables, indicators of liver disease severity and serum Cr were most consistently predictive of intraoperative blood product utilization; however, we were not able to construct a model to preoperatively identify those patients needing a large amount of blood products. In contrast, McClusky et al.13 reported a high c statistic (0.79) for a model to predict massive transfusion during the first 24 hours using similar preoperative variables.

Surgical blood loss and subsequent blood replacement can impact outcomes after OLT.5,7 In this analysis, increasing transfusion of RBC and CS was consistently associated with increased risk of mortality which peaked 2 weeks posttransplant. There is still no agreement whether patients who require more intraoperative blood products are a “sicker” population and therefore have a higher postoperative morbidity and mortality, or whether these outcomes are directly associated with the blood products transfused.5 Multiple studies have demonstrated an association between intraoperative blood product administration and worse outcomes; however, causality has not been proven.7,14,15 In the literature, patients who received more blood products were sicker at the time of transplantation which is also the case in our analysis.5,7 Furthermore, according to United Network for Organ Sharing data, overall survival of recipients in our program is better than the national average and certainly no worse than survival in programs where the transfusion exposure is much less. It may be that when there is a substantial difference in patient populations between programs (acuity of disease, more patients with hepatocellular carcinoma, etc.), the benefit of transfusion overshadows the risk associated with exposure to blood products. Although it is plausible to assume that increased transfusion requirements are associated with immediate outcomes based on the temporal proximity of these events (transfusion and death), explanation of the long-term effects of blood transfusion is more difficult. It is possible that the effect of intraoperative transfusion on long-term survival is dwarfed by the overall postoperative management of the recipients: choice and management of immunosuppression and its side effects and other posttransplant medical comorbidities.

Our investigation included a large cohort of patients transplanted in a high volume transplant center; however, even after analysis with sophisticated statistical methods, we were unable to construct a clinically useful model to predict intraoperative requirements for blood products solely based on preoperative variables. Even the prediction of a large volume transfusion (>20 units of RBC and CS) proved difficult.

We did confirm, however, a strong association between intraoperative transfusion and early mortality, as well as quantified the risk of transfusion on posttransplant survival.

The main limitation of this study is its retrospective nature and the fact that we could not account for all the factors influencing transfusion requirements, including transfusion triggers. Also, the results of this study reflect our institutional practice and should not be interpreted outside of this context. It is possible, although unlikely, that a similar analysis conducted on patients from other institutions could produce different results, simply based on differences in practice and the population presenting for OLT.

In conclusion, prediction of intraoperative blood product requirements based on preoperatively available variables is unreliable in all transfusion range categories analyzed. We were able to quantify a strong measurable association between intraoperative blood transfusion and recipient mortality after OLT.

Back to Top | Article Outline

APPENDIX

INTERPRETATION OF REGRESSION, LOGISTIC REGRESSION, AND SURVIVAL MODEL EFFECTS
Multivariable Regression:

Predictors of red blood cell (RBC) + cell salvage (CS) usage (Table 5).

Back to Top | Article Outline
Recipient Age at Transplant

For each year older a patient is, there is a 0.9% increase in the predicted number of units of RBC + CS required. For every increase of 10 years in recipient age, there is a 9% increase in the estimated number of units of RBC + CS required.

Back to Top | Article Outline
Recipient Diagnosis: HCC + Malignant Liver Tumor

For patients with this diagnosis, the predicted need for blood products is 30% lower than for patients with other diagnoses.

Back to Top | Article Outline
Recipient Pretransplant Platelet Count

Each increase of 1k/mm3 in pretransplant platelet count results in a 0.3% decrease in the predicted number of units of RBC + CS required. Each increase of 10k/mm3 in pretransplant platelet count results in a 3% decrease in the predicted number of units of RBC + CS required.

Back to Top | Article Outline
Recipient Pretransplant INR

Each increase of 1 unit of INR results in a 36% increase in predicted number of units of RBC + CS required.

Back to Top | Article Outline
Recipient Pretransplant Total Bilirubin

Each increase of 1 unit of total bilirubin results in a 0.9% increase in predicted number of units of RBC + CS required.

Back to Top | Article Outline
Recipient Pretransplant Creatinine

Each increase of 1 unit of creatinine results in a 9% increase in predicted number of units of RBC + CS required.

Back to Top | Article Outline
Expanded Criteria Donor

Use of expanded criteria donor results in 26% increase in predicted number of units of RBC + CS required.

Back to Top | Article Outline
Surgeon X

When Surgeon X performs the transplant, the predicted number of units of RBC + CS required is 30% lower than when another surgeon performs the transplant.

Back to Top | Article Outline
Logistic Regression: Predictors of RBC + CS usage > 20 Units (Table 6)

(Effect size percentages are followed by 95% confidence intervals on the effect sizes).

Back to Top | Article Outline
Recipient Pretransplant Creatinine

Each 1 unit increase in creatinine increases odds of needing >20 units of RBC + CS by 34% (17%–53%).

Back to Top | Article Outline
Recipient Pretransplant INR

Each 1 unit increase in INR increases odds of needing >20 units of RBC + CS by 81% (23%–167%).

Back to Top | Article Outline
Recipient Pretransplant Platelet Count

Each decrease of 1k/mm3 in pretransplant platelet count results in a 0.6% (0.2%–0.9%) increase in the odds of needing >20 units of RBC + CS. Each decrease of 10k/mm3 in pretransplant platelet count results in a 5.8% (2.0%–8.6%) increase in the odds of needing >20 units of RBC + CS.

Back to Top | Article Outline
Recipient Previous Abdominal Surgery

History of previous abdominal surgery increases odds of needing >20 units of RBC + CS by 79% (23%–162%).

Back to Top | Article Outline
Recipient Blood Type A

Blood type A increases odds of needing >20 units of RBC + CS by 68% (15%–143%).

Back to Top | Article Outline
Surgeon Y

Having transplant Surgeon Y increases odds of needing >20 units of RBC + CS by 112% (21%–272%).

Back to Top | Article Outline
Logistic Regression: Predictors of RBC + CS usage > 30 Units (Table 9)

(Effect size percentages are followed by 95% confidence intervals on the effect sizes).

Back to Top | Article Outline
Recipient MELD Score

Each 1 unit increase in the assigned MELD Score increases odds of needing >30 units of RBC + CS by 7% (2%–11%).

Back to Top | Article Outline
Recipient Pretransplant Platelet Count

Each decrease of 1k/mm3 in pretransplant platelet count results in a 0.7% (0.2%–1.3%) increase in the odds of needing >30 units of RBC + CS. Each decrease of 10k/mm3 in pretransplant platelet count results in a 6.8% (1.9%–12.3%) increase in the odds of needing >30 units of RBC + CS.

Back to Top | Article Outline
Surgeon Y

Having transplant Surgeon Y increases odds of needing >30 units of RBC + CS by 164% (53%–871%).

Back to Top | Article Outline
Surgeon Z

Having transplant Surgeon Z increases odds of needing > 30 units of RBC + CS by 286% (66%-986%).

Back to Top | Article Outline
Survival Analysis Hazard Model (Table 12)

(Effect size percentages are followed by 95% confidence intervals on the effect sizes).

Back to Top | Article Outline
RBC + CS Units Transfused

Each additional unit transfused increases the hazard of death in the early hazard phase by 4.3% (3.0%–5.6%).

Back to Top | Article Outline
INR

Each increase of 0.1 in INR is associated with a 3.3% (1.4%–5.3%) increase in hazard of death in the early hazard phase.

Back to Top | Article Outline
Donor Age

Each increase of 1 year in donor age is associated with a 3.0% (1.2%–4.9%) increase in hazard of death in the early hazard phase. Each increase of 10 years in donor age is associated with a 34% (12%–61%) increase in hazard of death in the early hazard phase.

Back to Top | Article Outline
Donor Race Other than African American or Caucasian

Patients whose donor is of race other than African American or Caucasian have 4.4 (1.6–11.7) times the hazard of death in the early hazard phase as patients with African American or Caucasian donors.

Back to Top | Article Outline
Diagnosis Viral Hepatitis

Patients with a primary diagnosis of viral hepatitis have 3.0 (1.4–6.1) times the hazard of death in the late hazard phase as patients with other diagnoses.

Back to Top | Article Outline
Disease Etiology Liver Cancer

Patients with disease etiology of liver cancer have 4.2 (2.0–8.8) times the hazard of death in the late hazard phase as patients with other disease etiologies.

Back to Top | Article Outline

DISCLOSURES

Name: Jacek B. Cywinski, MD.

Contribution: This author helped design and conduct the study, analyze the data, and write the manuscript.

Attestation: Jacek B. Cywinski has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.

Name: Joan M. Alster, MS.

Contribution: This author helped analyze the data and write the manuscript.

Attestation: Joan M. Alster has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Charles Miller, MD.

Contribution: This author helped write the manuscript.

Attestation: Charles Miller has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: David P. Vogt, MD.

Contribution: This author helped design and conduct the study.

Attestation: David P. Vogt has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

Name: Brian M. Parker, MD.

Contribution: This author helped design and conduct the study, analyze the data, and write the manuscript.

Attestation: Brian M. Parker has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.

This manuscript was handled by: Edward C. Nemergut, MD.

Back to Top | Article Outline

ACKNOWLEDGMENTS

The authors wish to thank Lucy Thuita, MS, for data management and programming and Mrs. Tanya Smith for her editorial assistance.

Back to Top | Article Outline

FOOTNOTE

a The assigned MELD Score is a composite of the biologic MELD Score (which is determined using the patient’s most recent bilirubin, prothrombin INR time and creatinine values) and special case exception points which are assigned for urgent situations including the presence of hepatocellular carcinoma. Cited Here...

Back to Top | Article Outline

REFERENCES

1. Deakin M, Gunson BK, Dunn JA, McMaster P, Tisone G, Warwick J, Buckels JA. Factors influencing blood transfusion during adult liver transplantation. Ann R Coll Surg Engl. 1993;75:339–44

2. Findlay JY, Rettke SR. Poor prediction of blood transfusion requirements in adult liver transplantations from preoperative variables. J Clin Anesth. 2000;12:319–23

3. Steib A, Freys G, Lehmann C, Meyer C, Mahoudeau G. Intraoperative blood losses and transfusion requirements during adult liver transplantation remain difficult to predict. Can J Anaesth. 2001;48:1075–9

4. Massicotte L, Sassine MP, Lenis S, Roy A. Transfusion predictors in liver transplant. Anesth Analg. 2004;98:1245–51, table of contents

5. Hendriks HG, van der Meer J, de Wolf JT, Peeters PM, Porte RJ, de Jong K, Lip H, Post WJ, Slooff MJ. Intraoperative blood transfusion requirement is the main determinant of early surgical re-intervention after orthotopic liver transplantation. Transpl Int. 2005;17:673–9

6. Schroeder RA, Johnson LB, Plotkin JS, Kuo PC, Klein AS. Total blood transfusion and mortality after orthotopic liver transplantation. Anesthesiology. 1999;91:329–30

7. Massicotte L, Sassine MP, Lenis S, Seal RF, Roy A. Survival rate changes with transfusion of blood products during liver transplantation. Can J Anaesth. 2005;52:148–55

8. de Boer MT, Molenaar IQ, Hendriks HG, Slooff MJ, Porte RJ. Minimizing blood loss in liver transplantation: progress through research and evolution of techniques. Dig Surg. 2005;22:265–75

9. Breiman L. Bagging predictors. Mach Learn. 1996;24:123–40

10. Breiman L. Random forests. Mach Learn. 2001;45:5–32

11. Blackstone EH, Naftel DC, Turner ME Jr. The decomposition of time-varying hazard into phases, each incorporating a separate stream of concomitant information. J Am Stat Assoc. 1986;81:615–24

12. Hevesi ZG, Lopukhin SY, Mezrich JD, Andrei AC, Lee M. Designated liver transplant anesthesia team reduces blood transfusion, need for mechanical ventilation, and duration of intensive care. Liver Transpl. 2009;15:460–5

13. McCluskey SA, Karkouti K, Wijeysundera DN, Kakizawa K, Ghannam M, Hamdy A, Grant D, Levy G. Derivation of a risk index for the prediction of massive blood transfusion in liver transplantation. Liver Transpl. 2006;12:1584–93

14. Pereboom IT, de Boer MT, Haagsma EB, Hendriks HG, Lisman T, Porte RJ. Platelet transfusion during liver transplantation is associated with increased postoperative mortality due to acute lung injury. Anesth Analg. 2009;108:1083–91

15. Ramos E, Dalmau A, Sabate A, Lama C, Llado L, Figueras J, Jaurrieta E. Intraoperative red blood cell transfusion in liver transplantation: influence on patient outcome, prediction of requirements, and measures to reduce them. Liver Transpl. 2003;9:1320–7

© 2014 International Anesthesia Research Society

Login

Become a Society Member

Article Level Metrics