ORIGINAL ARTICLES

Preharvest Donor Hyperoxia Predicts Good Early GraftFunction and Longer Graft Survival After Liver

TransplantationStefano Ginanni Corradini,1 Walter Elisei,1 Rosanna De Marco,1 Maria Siciliano,1

Massimo Iappelli,2 Francesco Pugliese,2 Franco Ruberto,2 Francesco Nudo,2

Renzo Pretagostini,2 Alessandro Bussotti,2 Gianluca Mennini,2 Annarita Eramo,1

Francesca Liguori,1 Manuela Merli,1 Adolfo Francesco Attili,1

Andrea Onetti Muda,3 Stefania Natalizi,3 Pasquale Berloco,2 and Massimo Rossi2

A total of 44 donor/recipient perioperative and intraoper-ative variables were prospectively analyzed in 89 deceased-donor liver transplantations classified as initial good graftfunction (IGGF) or initial poor graft function (IPGF)according to a scoring system based on values obtainedduring the 1st 72 postoperative hours from the serumalanine aminotransferase (ALT) concentration, bile out-put, and prothrombin activity. The IGGF compared withthe IPGF group showed: 1) longer graft (P � .002) andpatient (P � .0004) survival; 2) at univariate analysis, ahigher (mean [95% confidence interval]) preharvestdonor arterial partial pressure of oxygen (PaO2) (152[136-168] and 104 [91-118] mmHg, respectively; P �.0008) and arterial hemoglobin oxygen saturation (97.9[97.2-98.7] and 96.7 [95.4-98.0]%, respectively; P �.0096), a lower percentage of donors older than 65 years(13 and 33%, respectively; P � .024), a lower percentageof donors treated with noradrenaline (16 and 41%,respectively; P � .012). At multivariate analysis, IGGFwas associated positively with donor PaO2 and negativelywith donor age greater than 65 years and with donortreatment with noradrenaline. Independently from thegrouping according to initial graft function, graft survivalwas longer when donor PaO2 was >150 mmHg thanwhen donor PaO2 was <150 mmHg (P � .045). In con-clusion, preharvest donor hyperoxia predicts IGGF andlonger graft survival. (Liver Transpl 2005;11:140–151.)

To expand the potential donor pool and meet thewaiting list demands for liver transplantation, the

use of marginal liver grafts, although at risk of initialpoor graft function (IPGF) or primary nonfunction(PNF), needs to be pursued.1 In the past, several studieshave shown that some donor characteristics define themarginal liver graft and other recipient perioperativeand intraoperative variables are also associated withIPGF and PNF.1–12 As the recipient and donor demo-graphics, the intensive care unit (ICU) management oforgan donors, the perioperative and intraoperativebehavior, and the allocation policy of the transplantteams change over time, the confirmation of the alreadydescribed predictors and a search for new variables asso-ciated with the transplant outcome are continuallyneeded.1 The present study was undertaken to obtainan updated investigation of variables associated with thequality of initial graft function and graft and patientsurvival after deceased-donor liver transplantation at asingle transplant center in Rome, Italy. We conducted aprospective investigation on an homogeneous popula-tion of primary, nonstatus 1, adult liver transplanta-tions performed using full-size livers.

Patients and potential organ donors in the ICU havePaO2 values varying widely, from hyperoxia, to nor-moxemia, and to hypoxemia.13 The degree of donorblood oxygenation, especially in the immediate prehar-vest period, could affect the liver graft’s early postoper-ative function by regulating hepatic metabolic path-ways, gene expression, and/or radical oxygenformation. In previously reported studies, the quality ofearly graft postoperative function after liver transplan-tation has been investigated in relation to donor PaO2

values and the occurrence of hypoxic episodes recordedat an undetermined time point during ICU stay or atthe time of donor ICU admittance.3,10,14,15 Among thevariables investigated in the present study, we includeddonor PaO2 measured no longer than 4 hours before

Abbreviations: ALT, alanine aminotransferase; ICU, intensivecare unit; IGGF, initial good graft function; IPGF, initial poor graftfunction; PNF, primary nonfunction; MELD, model for end-stageliver disease.

From the 1Division of Gastroenterology, Department of ClinicalMedicine, 2Department of General Surgery “Paride Stefanini”, and3Department of Experimental Medicine and Pathology, University “LaSapienza” Rome, Italy.

Address reprint requests to Stefano Ginanni Corradini, MD, PhD,Via Asmara 9-B, 00199 Rome, Italy. Telephone: 39-06-49972086;FAX: 39-06-4453319; E-mail: stefano.corradini@uniroma1.it

Copyright © 2005 by the American Association for the Study ofLiver Diseases

Published online in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/lt.20339

140 Liver Transplantation, Vol 11, No 2 (February), 2005: pp 140–151

organ harvest because we reasoned that it would haverepresented the degree of oxygenation, which influ-ences postoperative liver function, more efficiently thanin the aforementioned prior studies.3,10,14,15

Criteria used to define the quality of initial graft func-tion after liver transplantation have not been consistent inthe literature and grouping patients as IPGF or initial goodgraft function (IGGF), as well as the variables found to bepredictors of the quality of initial graft function, coulddiffer according to the criteria used.4–6,9,11,16–22 In addi-tion, controversial results on the prognostic value of thecriteria used to define the quality of initial graft functionhave been reported.4–6,9,16–19 It would be of low clinicalimpact to identify predictors of IPGF using criteria withpoor or no prognostic value. Thus, in the present study weused 2 different criteria to define initial graft function, toclassify the grafts as IPGF or IGGF, and to predict graftand patient survival.

Patients and Methods

Patients

Between February 2001 and April 2004, 107 consecutivetransplantations in 100 patients were performed at the LiverTransplantation Unit of University “La Sapienza” Hospital inRome. In this study we prospectively investigated only the 91patients submitted to nonstatus 1, full-size liver, primary,deceased-donor transplantation. A total of 2 patients wereexcluded from the analysis of the results: 1 because of intra-operative death and the other because, for technical reasons, itwas not possible to obtain donor biochemical data accordingto the study protocol. Therefore the study included theremaining 89 patients. Donor livers with macroscopic fattyinfiltration confirmed by a biopsy with more than 50% ofmacrovesicular steatosis were not utilized. The immunosup-pressive protocol was based on a triple therapy with cyclospor-ine microemulsion, methylprednisolone, and mycophenolatemofetil. Blood concentration of cyclosporine microemulsionwas monitored at time 0 and 2 hours after ingestion. Meth-ylprednisolone was rapidly tapered. Follow-up included theoccurrence of PNF or delayed nonfunction, graft and patientsurvival, and the incidence of complications. PNF wasdefined as a nonrecoverable hepatocellular function necessi-tating emergency retransplantation within 7 days. Delayednonfunction was defined as graft function necessitating emer-gency retransplantation between 7 and 30 days after trans-plantation.

Classification Criteria for Initial Graft Function

When the present study was planned, the quality of initialgraft function was defined according to a scoring system pre-viously utilized by a Spanish group, based on ALT, bile out-put, and prothrombin activity measured in the 1st 72 postop-erative hours (Table 1).5,10,20,21 In each patient, the score was

calculated from the sum of the assigned values for each param-eter. On the basis of this score, as Gonzalez et al.5 did in theirwork to investigate factors predictive of the quality of earlyliver graft function, in the present study we classified ourpatients into 2 groups. One group included patients with ascore ranging from 7 to 9, which we considered as IPGF,according to the Spanish criteria. The other group includedpatients with a score ranging from 3 to 6, which we consideredas IGGF, also according to the Spanish criteria.

Since several studies in the literature have used criteria todefine initial graft function based only on postoperative serumaminotransferase levels, in the present study we also groupedour transplantations using the Australian National LiverTransplantation Unit criteria for initial graft function.9,11,22

According to the Australian criteria, the group with IPGFincluded patients with an increased level of serum aspartateaminotransferase and / or ALT of greater than 1500 IU/Lobtained in at least 2 consecutive measurements within the 1st72 hours after liver transplantation. All the other patients wereclassified in the IGGF group according to the Australian cri-teria.

Clinical Variables Analyzed as Predictors of theQuality of Initial Graft Function and ofSurvival

Pretransplant recipient-associated factors analyzed in relationto the quality of initial graft function were age, gender, causeof liver disease, biochemical data, and the model for end-stageliver disease (MELD) score.23 Recipient biochemical datawere obtained a few hours before surgery. Donor-associatedfactors analyzed in relation to the quality of initial graft func-tion were age, gender, body mass index, duration of ICU stay,

Table 1. Parameters Used for the Calculation of ScoringSystem Assessing Early Postoperative Graft Function in

Liver Transplant Patients According to the Spanish Criteria

Parameter Assigned value

Serum ALT (U/L)*�1,000 11,000–2,500 2�2,500 3

Bile output (mL/day)†�100 140–100 2�40 3

Prothrombin activity (%)‡�60 (spontaneous) 1�60 (with fresh-frozen plasma) 2�60 (with fresh-frozen plasma) 3

*Highest value in the 72 hours after liver transplantation.†Mean value during the first 72 hours after liver transplanta-tion.‡Lowest value in the first 72 hours after liver transplantation.IPGF � score 7–9; IGGF � score 3–6.

141Donor Hyperoxia and Liver Transplantation

cause of brain death, occurrence of cardiac arrest and hypo-tension episodes during ICU stay, use of vasoactive drugsduring at least the last 12 hours before organ harvesting,venous biochemical data, arterial blood gas analysis, and livergraft steatosis. Donor venous blood samples were obtainedduring the last 6 hours before organ harvesting. Donor arterialblood gas analysis was obtained between 4 and 1 hour beforeorgan harvesting with a standard technique, taking particularcare to avoid trapping air bubbles in the syringe. Blood sam-ples were immediately analyzed at each local laboratory andthe gas analyzers were calibrated before each measurement.Aliquots of preexplant allograft biopsy obtained from donorswhose hearts were beating before the aorta was clamped wereimmediately fixed in formalin for steatosis assessment. Thesteatosis level was blindly scored independently by 2 of us(A.O.M. and S.N.). The percentage of hepatocytes withmicrovesicular and macrovesicular steatosis was measured.Data were then analyzed according to the Australian NationalLiver Transplantation Unit scale of macrovesicular steatosiswhich is as follows: S0, steatosis absent; S1, less than 30% ofhepatocytes with steatosis; S2, 30 to 60%; and S3, more than60%.9,11,24 The organ allocation committee was not aware ofdonor gas analytical variables and graft steatosis score beforeeach allocation decision.

Cold and warm ischemia times, the use of venovenousbypass, and of the piggyback surgical techniques, and thehemoglobin concentration in the recipient at the end of sur-gery, were analyzed as intraoperative and perioperative vari-ables in relation to the quality of initial graft function.

Irrespective of any early postoperative graft function clas-sification, graft and patient survival were analyzed as a func-tion of donor age, sodium serum concentration, length ofICU stay, use of noradrenaline and PaO2, cold ischemia time,and graft macrovesicular steatosis.

Statistical Analysis

Agreement beyond chance between the Spanish and the Aus-tralian criteria for initial graft function classification wasassessed by Kappa test statitistic.25 The following quantitativevalues were applied to the Kappa test statistic: 0– .2 � “slight”agreement; .2– .4 � “fair”; .4– .6 � “moderate”; .6– .8 �“substantial”; .8–1.0 � “almost perfect.”25,26 Univariateanalyses (�2 test for categorical factors and analysis of variancewith Kruskal-Wallis test for continuous factors) were used forthe comparison of the IGGF with the IPGF group. Multivar-iate analysis (logistic-regression analysis) was used to identifyindependent predictive factors of the quality of initial graftfunction. Among the quantitative variables, cold ischemiatime, donor age, length of ICU stay, and serum sodium con-centration were analyzed as both continuous and dichoto-mous variables. Liver graft steatosis was analyzed as bothsemiquantitative and dichotomous variables. The cutoff val-ues for the dichotomous variables were chosen according to arecent review that describes the marginal donor features.1

Probability curves of patient and graft survival were calculatedaccording to the Kaplan-Meier method, and the curves were

statistically compared by means of the log-rank test. Correla-tions were calculated with the Spearman’s correlation test.The significance level was set at P � .05. Contingency tables(2 � 2) were used to determine positive and negative predic-tive values. The NCSS 97 Statistical System for Windows(NCSS, Kaysville, UT) was used for all statistical analyses.

Results

General Donor and Graft Characteristics

Overall, the mean age of the donors was 49.1 years(median 51.0 years; range 11-83 years). A total of 51%(45 / 89), 29% (26 / 89), and 19% (17 / 89) of thedonors were �50, �60, and �65 years of age, respec-tively. Only 26% (23 / 89) of the donors died due to atraumatic event. A total of 42% (37 / 89) of the donorshad a body mass index between 25 and 30; and 8% (7 /89) of the donors had a body mass index over 30. A totalof 25% (22 / 89) of the donors stayed in the ICU morethan 5 days. A total of 11% (10 / 89) of the grafts had acold ischemia time longer than 8 hours; and 10% (7 /71) of the grafts had a macrovesicular steatosis �30%.

Concordance of the Spanish and AustralianClassification Criteria for Initial Graft Function

A total of 9 grafts classified as IPGF according to theSpanish criteria were included in the IGGF group whengrouping was based on the Australian criteria. On theother hand, 2 grafts classified as IPGF according to theAustralian criteria were included in the IGGF groupwhen grouping was based on the Spanish criteria. The 2classification criteria were concordant in 18 and 60grafts that were classified as IPGF and IGGF, respec-tively. Thus, the 2 criteria used to define initial poorgraft function were concordant in the classification of atotal of 78 grafts (88%). The Kappa test statistic was .68(95% confidence interval .52-.85), showing a “substan-tial” agreement in the classification of grafts betweenthe 2 criteria used to define early postoperative graftfunction. The 2 classification criteria were also concor-dant with regard to the differences of serum amino-transferase concentrations and bile output during the1st 3 days after transplantation between the IPGF andthe respective IGGF group. In fact, the mean serumaspartate aminotransferase and ALT concentrationswere higher and the mean bile output was lower in asimilar fashion in each IPGF, as compared with therespective IGGF group (Table 2).

During follow-up (median 16.5 months; range.1-41 months), 12 and 6 grafts were lost because of liverrelated and liver unrelated causes, respectively. The 2criteria used to define postoperative graft function were

142 Corradini et al.

completely concordant on the distribution between theIPGF and the respective IGGF group, with regard tothe short-term (within 4 months after transplantation)liver related and liver unrelated causes of graft loss. Inparticular, there were 5 PNF, 1 delayed nonfunction,and 1 acute rejection in the IPGF group according toboth of the classification criteria. There were 2 hepaticarterial thrombosis in the IGGF group according toboth of the classification criteria. With regard to theliver unrelated causes of graft loss, 1 patient classified inthe IPGF group according to both of the classificationcriteria died of pulmonary embolism, and the others (1pulmonary embolism, 2 sepsis, and 2 abdominal hem-orrhages) had been classified in the IPGF group accord-ing to both of the classification criteria.

As shown in Figure 1, graft and patient survivalswere significantly better in the IGGF as compared with

the IPGF group according to the Spanish criteria. Thesame results were found when we compared graft andpatient survival in the IGGF and the IPGF groupdefined according to the Australian criteria (data notshown). In addition, the initial graft function classifica-tion according to the Spanish and the Australian criteriashowed similar positive (41 and 50%, respectively) andnegative (89 and 88%, respectively) predictive valuesfor graft loss and similar positive (41 and 50%, respec-tively) and negative (92 and 91%, respectively) predic-tive values for patient death.

A total of 27 (30%) and 20 (22%) grafts were clas-sified as IPGF according to the Spanish and the Austra-lian criteria, respectively. Since the 2 methods used toclassify initial graft function were substantially concor-dant in the distribution of grafts between the IPGF andthe IGGF group and had the same prognostic power,

Table 2. Serum Aminotransferase Concentrations and Bile Output During the First 72 Hours After Transplantation According toEarly Postoperative Graft Function Classifications*

Parameter IPGF-S† n � 27 IGGF-S† n � 62 IPGF-A† n � 20 IGGF-A† n � 69

Serum AST (U/L) atpostoperative day 1 1808‡ 775 2137§ 784

(1,288–2,328) (608–943) (1,589–2,756) (626–944)Serum ALT (U/L) at

postoperative day 1 1332‡ 579 1530§ 598(1,029–1,635) (466–691) (1,179–1,881) (490–705)

Serum AST (U/L) atpostoperative day 2 2170*‡ 549 2764§ 542

(1,449–2,891) (446–653) (1,933–3,595) (452–631)Serum ALT (U/L) at

postoperative day 2 2145*‡ 601 2664§ 607(1,595–2,695) (488–713) (2,047–3,280) (521–693)

Serum AST (U/L) atpostoperative day 3 1423‡ 379 1753§ 389

(946–1,899) (282–476) (1,191–2,316) (293–485)Serum ALT (U/L) at

postoperative day 3 2021‡ 580 2469§ 597(1,522–2,520) (458–703) (1,901–3,036) (495–699)

Bile output (mL/day) atpostoperative day 1 55‡ 132 62� 122

(17–94) (106–158) (10–114) (98–146)Bile output (mL/day) at

postoperative day 2 70‡ 155 82§ 142(9–130) (132–177) (1–164) (120–165)

Bile output (mL/day) atpostoperative day 3 70† 165 87� 150

(41–99) (145–184) (48–126) (130–170)

*Data expressed as mean (95% CI).†IPGF-S � IPGF according to the Spanish criteria; IGGFS � IGGF according to the Spanish criteria; IPGF-A � IPGF according to theAustralian criteria; IGGF-A � IGGF according to the Australian criteria.‡P � 0.0001 IPGF vs. IGGF according to the Spanish criteria by Kruskal-Wallis test.§P � 0.0001 IPGF vs. IGGF according to the Australian criteria by Kruskal-Wallis test.�P � 0.01 IPGF vs. IGGF according to the Australian criteria by Kruskal-Wallis test.

143Donor Hyperoxia and Liver Transplantation

we report in the following analyses only the data relativeto the Spanish criteria for initial graft function classifi-cation.

Univariate Analysis of Clinical VariablesPredictive of the Quality of Initial GraftFunction

A total of 44 variables were investigated at univariateanalysis as factors predictive of the quality of initial graftfunction according to the Spanish criteria (Tables3–7). All the transplant recipient qualitative and quan-titative variables were not different in the IGGF ascompared with the IPGF group (Table 3). Among thequalitative variables from organ donors, the percentageof donors treated with noradrenaline was significantlylower in the IGGF than in the IPGF group (Table 4).No intergroup difference was found for donor age,body mass index, length of ICU stay, and venous bloodquantitative variables evaluated as continuous variables(Table 5). When donor age (�60 vs. �60 years or �65vs. �65 years), sodium serum concentration (�160 vs.�160 mEq/L) and length of ICU stay (�5 vs. �5days), were dichotomized, only donor age �65 vs. �65years was significantly (P � .024) different between theIGGF and the IPGF group. In fact the proportion ofdonors over 65 years was lower in the IGGF (8 / 62;13%) than in the IPGF (9 / 27; 33%) group. The meanand median donor arterial PaO2 were in the hyperoxicrange and significantly higher in the IGGF as compared

Figure 1. Probability of graft (upper section) and patient(lower section) survival in liver transplantation with initialpoor (IPGF; n � 27) and good (IGGF; n � 62) postoperativegraft function according to the Spanish criteria.

Table 3. Transplant Recipient Qualitative and Quantitative Variables Analyzed as Factors Predictive of Early Postoperative GraftFunction According to the Spanish Criteria*

Parameter IPGF (n � 27) IGGF (n � 62) P value†

Gender [number (%)] .884Male 20 (74) 45 (73)Female 7 (26) 17 (27)

Cause of liver disease [number (%)] .309Viral 16 (59) 41 (66)Alcoholic 2 (7) 10 (16)Alcoholic � viral 3 (11) 5 (8)Other 6 (22) 6 (10)

Mean age (years) 54 (50–57) 54 (51–56) .796Serum bilirubin (mg/dL) 5.1 (1.7–8.6) 3.6 (2.6–4.6) .950Serum albumin (g/dL) 3.2 (3.1–3.4) 3.1 (3.0–3.2) .164Prothrombin activity (International Normalized Ratio) 1.5 (1.3–1.8) 1.6 (1.4–1.7) .171Serum creatinine (mg/dL) 1.1 (1.0–1.2) 1.1 (1.0–1.1) .349MELD (score;‡) 14 (12–17) 14 (12–15) .789

*Data expressed as number of liver transplantation recipients (% of the total for each group) for qualitative variables and as mean (95%CI) for quantitative variables.†P values by �2 test and by Kruskal-Wallis test for qualitative and quantitative variables, respectively.‡See Kamath et al.23

144 Corradini et al.

with the IPGF group in which they were in the nor-moxic range (Table 6; Fig. 2). The mean donor arterialhemoglobin oxygen saturation was significantly higherin the IGGF as compared with the IPGF group, whileall the other donor gas analytical parameters and thefraction of inspired oxygen were not different in the twogroups (Table 6). A significant (P � .01) negative cor-relation was found between donor preharvest PaO2 val-ues and recipient serum ALT concentrations on the 1stpostoperative day (Fig. 3). Similar correlations werefound between donor PaO2 and recipient serum ALTconcentration on the 2nd and 3rd postoperative daysand between donor PaO2 and recipient serum aspartateaminotransferase concentration on the 1st, 2nd, and3rd postoperative days (data not shown). When donorPaO2 values (�150 vs. �150 mmHg) were dichoto-mized, the proportion of donors with a PaO2 value

�150 mmHg was significantly (P � .001) higher in theIGGF (26 / 60; 43%) than in the IPGF (2 / 26; 8%)group. Intraoperative and perioperative qualitative andquantitative variables were not different in the IGGF ascompared with the IPGF group (Table 7). When coldischemia time (�8 vs. �8 hours) was dichotomized,the proportion of grafts with an ischemia time longerthan 8 hours did not differ (P � .151) between theIGGF (5 / 62; 8%) and the IPGF (5 / 27; 19%)groups.

A total of 71 liver donor biopsies were available forsteatosis scoring, 50 in the IGGF and 21 in the IPGFgroup. At univariate analysis, the proportion of grafts inthe IGGF and in the IPGF group with each steatosisscore did not differ (P � .886). In fact, the number ofgrafts classified as IGGF and as IPGF with a steatosisscore S0 (27 [54%] vs. 11 [52%], respectively), S1 (18[36%] vs. 8 [38%], respectively), S2 (4 [8%] vs. 1 [5%],respectively), and S3 (1 [2%] vs. 1 [5%], respectively),was similar. When the percentage of hepatocytes withmacrovesicular steatosis (�30 vs. �30%) was dichoto-mized, the proportion of grafts with �30% of steatosisdid not differ (P � .951) between the IGGF (5 / 50;10%) and the IPGF (2 / 21; 10%) groups.

Independently from the initial graft function classi-fication, in the 7 transplants performed using graftswith �30% macrovesicular steatosis, total ischemiatime never exceeded 9 hours, in only 1 transplant wasthe recipient positive for hepatitis C virus and the donorolder than 65 years.

Multivariate Analysis of Variables Predictive ofthe Quality of Initial Graft Function

Multivariate analysis was performed including initialgraft function as group variable, use of venovenousbypass and treatment with noradrenaline as indepen-dent qualitative variables and the following as inde-pendent quantitative continuous variables: cold andwarm ischemia times, recipient age and MELD score,donor age, body mass index, length of ICU stay,serum sodium, ALT, aspartate aminotransferase andcreatinine concentrations, PaO2, arterial hemoglo-bin oxygen saturation and pH. IGGF was positivelyassociated with donor PaO2 (P � .005), while it wasnegatively associated with the percentage of donorstreated with noradrenaline (P � .02). No other pre-dictor of initial graft function was found. When thesame analysis was repeated substituting donor age(�65 vs. �65 years), sodium serum concentration(�160 vs. �160 mEq/L), length of ICU stay (�5 vs.�5 days), and cold ischemia time (�8 vs. �8 hours)

Table 4. Qualitative Variables From Organ DonorsAnalyzed as Factors Predictive of Early Postoperative Graft

Function According to the Spanish Criteria*

ParameterIPGF

n � 27IGGFn � 62

Pvalue†

Gender .446Male 12 (44) 33 (53)Female 15 (56) 29 (47)

Cause of brain death .991Traumatic 7 (26) 16 (26)Vascular/anoxic 20 (74) 46 (74)

Prolonged arterial hypotension .955Yes 8 (30) 18 (29)No 19 (70) 44 (71)

Cardiac arrest .451Yes 1 (4) 5 (8)No 26 (96) 57 (92)

Use of any vasoactive drug .430Yes 24 (89) 51 (82)No 3 (11) 11 (18)

Use of �1 vasoactive drug .372Yes 9 (33) 15 (24)No 18 (67) 47 (76)

Use of dopamine � 10 �g/kg/minute .359Yes 5 (19) 7 (11)No 22 (81) 55 (89)

Use of dobutamine .667Yes 5 (19) 14 (23)No 22 (81) 48 (77)

Use of noradrenaline .012Yes 11 (41) 10 (16)No 16 (59) 52 (84)

*Data expressed as number of liver transplantation donors (%of the total for each group).†P values by �2 test.

145Donor Hyperoxia and Liver Transplantation

as dichotomous variables for the respective continu-ous variables, IGGF was positively associated withdonor PaO2 (P � .004), while it was negatively asso-ciated with the percentage of donors older than 65years (P � .018) and of those treated with noradren-aline (P � .014). No other predictor of initial graftfunction was found.

In the subgroup of 71 transplants for which the donorliver steatosis evaluation was available, the same 2 modelsof multivariate analyses were performed as for the entirestudy population, also including graft steatosis as a furtherindependent variable. When graft steatosis was included asthe Australian National Liver Transplantation Unit semi-quantitative score and all the quantitative variables wereincluded as continuous independent variables, IGGF waspositively associated with donor PaO2 (P � .006), while itwas negatively associated with the percentage of donors

treated with noradrenaline (P � .028). No other predictorof initial graft function was found. When graft steatosis(�30 vs. �30% of hepatocytes with macrovesicular ste-atosis), donor age, serum sodium concentration, length ofICU stay and cold ischemia time were included as dichot-omous independent variables, IGGF was positively asso-ciated with donor PaO2 (P � .009), while it was negativelyassociated with the percentage of donors treated with nor-adrenaline (P � .028). No other predictor of initial graftfunction was found.

Clinical Variables Predictive of SurvivalIndependent of the Initial Graft FunctionClassification

Irrespective of any early postoperative graft functionclassification, graft and patient survival were analyzed as

Table 5. Quantitative Variables From Organ Donors Analyzed as Factors Predictive of Early Postoperative Graft FunctionAccording to the Spanish Criteria*

Parameter IPGF (n � 27) IGGF (n � 62) P value†

Mean age (years) 52 (45–60) 48 (44–52) .190Body mass index (kg/m2) 26.0 (24.1–27.8) 25.2 (24.5–26.0) .704ICU stay (days) 5.6 (3.2–7.9) 4.5 (3.7–5.3) .698Serum AST (U/L) 52 (39–64) 55 (40–71) .500Serum ALT (U/L) 45 (29–62) 47 (30–65) .824Prothrombin activity (International Normalized Ratio) 1.3 (1.1–1.5) 1.2 (1.1–1.3) .241Serum creatinine (mg/dL) 1.0 (.8–1.2) 1.0 (0.9–1.1) .947Serum glucose (mg/dL) 178 (146–210) 163 (139–186) .362Serum sodium (mEq/L) 150 (145–155) 153 (150–156) .196Serum potassium (mEq/L) 3.7 (3.5–3.9) 3.7 (3.5–3.9) .870Red blood cell count (106/mm3) 3.79 (3.45–4.12) 3.49 (3.26–3.71) .296Hemoglobin concentration (g/dL) 10.9 (10.1–11.8) 10.6 (10.1–11.2) .636White blood cell count (103/mm3) 13.0 (10.7–15.2) 14.4 (10.4–18.4) .887

*Data expressed as mean (95% CI).†P values by Kruskal-Wallis test.

Table 6. Arterial Gas Analytical Variables and Fraction of Inspired Oxygen From Organ Donors Analyzed as Factors Predictive ofEarly Postoperative Graft Function According to the Spanish Criteria*

Parameter IPGF n � 26 IGGF n � 60 P value†

PaO2 (mmHg) 104 (91–118) 152 (136–168) .0008Hemoglobin O2 saturation (%) 96.7 (95.4–98.0) 97.9 (97.2–98.7) .0096pH 7.43 (7.40–7.45) 7.43 (7.41–7.44) .724HCO3

� (mEq/L) 25.2 (23.7–26.8) 24.9 (23.9–25.8) .443PCO2 (mmHg) 36.7 (34.3–39.2) 37.4 (35.4–39.3) .858FiO2

‡ (%) 45 (40–50) 44 (41–47) .497

*Data expressed as mean (95% CI). Arterial gas analysis was not available in 3 out of the 89 donors included in the study.†P values by Kruskal–Wallis.‡Fraction of inspired oxygen.

146 Corradini et al.

a function of donor age (�65 vs. �65 years), donorsodium serum concentration (�160 vs. �160 mEq/L),donor length of ICU stay (�5 vs. �5 days), donorinfusion with noradrenaline, donor PaO2 (�150 vs.�150 mmHg), cold ischemia time (�8 vs. �8 hours),graft macrovesicular steatosis (�30 vs. �30% and �15vs. �15%). Graft and patient survival were longer inrecipients who received the graft from a donor with aPaO2 value greater than 150 mmHg (n � 28) than thatof recipients who received the graft from a donor with aPaO2 value lower than 150 mmHg (n � 58), althoughstatistical significance was reached only for graft survival(Fig. 4). A trend (P � .070) for longer patient survivalwas found in recipients who received the graft after acold ischemia time shorter than 8 hours (n � 79) than

that of recipients who received the graft after a coldischemia time longer than 8 hours (n � 10). No othervariable was able to predict graft and / or patient sur-vival.

Discussion

The main and novel finding of this study is that prehar-vesting donor PaO2 is an independent predictor of thequality of early graft function and survival after adult,nonstatus 1, full-size, primary, deceased-donor livertransplantation. Both the univariate and the multivari-ate analyses showed that a relatively high (i.e., in thehyperoxic range) donor PaO2 predicts good early post-operative liver function which, in turn, was associated

Table 7. Intraoperative and Perioperative Qualitative and Quantitative Variables Analyzed as Factors Predictive of EarlyPostoperative Graft Function According to the Spanish Criteria*

Parameter IPGF n � 27 IGGF n � 62 P value†

Venovenous bypass .440Yes 16 (59) 42 (68)No 11 (41) 20 (32)

Piggyback .535Yes 11 (41) 21 (34)No 16 (59) 41 (66)

Graft ischemia times (minutes)Cold 394 (359–428) 390 (372–408) .817Warm 65 (54–75) 62 (56–68) .954Total 458 (420–495) 452 (432–472) .620

Hemoglobin concentration‡ (g/dL) 9.9 (9.2–10.5) 10.2 (9.7–10.6) .571

*Data expressed as number of liver transplantation recipients (% of the total for each group) for qualitative variables and as mean (95%CI) for quantitative variables.†P values by �2 test and by Kruskal-Wallis test for qualitative and quantitative variables, respectively.‡At the end of operation.

Figure 2. Box plots of donor PaO2 values in liver trans-plantation with initial poor (IPGF; n � 26) and good(IGGF; n � 60) postoperative graft function according tothe Spanish criteria.

Figure 3. Correlation plot (r � –.397; n � 86; P < .001)between donor preharvest PaO2 values and recipientserum ALT concentration on the 1st postoperative dayafter liver transplantation.

147Donor Hyperoxia and Liver Transplantation

with longer graft and patient survival compared withgrafts with poor early function. In addition, donorhyperoxia (PaO2 value greater than 150 mmHg) wasassociated with better graft survival even independentlyfrom the classification of the quality of early postoper-ative graft function. From a practical point of view, thepresent study suggests that a preharvest donor PaO2

value greater than 150 mmHg helps to define the opti-mal liver donor for deceased-donor adult liver trans-plantation. Consequently, donor PaO2 values might beincluded among the variables to estimate the risk ofIPGF or PNF, which is increased these days with thefrequent utilization of marginal (non-optimal) livergrafts.1,27 The marginal liver graft has been defined in arecent review by Busuttil and Tanaka1 according to thepresence of 1 or more of the following features: donorage over 60 years, donor ICU stay longer than 5 days,donor hypernatremia exceeding 160 mEq/L, donor

requirement for more than 1 vasopressor, graft coldischemia time longer than 8 hours, and graft macrove-sicular steatosis between 30 and 60%. In agreementwith the definition of the marginal liver graft, we foundthat donor age over 65 years was positively associatedwith IPGF.1,10 However, we found only a weak nonsig-nificant association between a graft cold ischemia timelonger than 8 hours and the quality of initial graftfunction or patient survival and no association betweengraft steatosis and the transplant outcome. It is likelythat the negative effect of prolonged graft cold ischemiatime and graft macrovesicular steatosis on the trans-plant outcome has been blunted in the present study bythe small number of grafts with a cold ischemia timelonger than 8 hours or with severe steatosis which, inturn, could have caused a �-type error. The highlyrestricted range of cold ischemia time and graft steatosisin the present study is the consequence of a changedbehavior and more restricted criteria for discardingseverely steatotic grafts by our transplant team, as wellas of the majority of transplantation surgeons in the lastyears, based upon previous studies which have identi-fied prolonged ischemia time and moderate to severesteatosis as features of the marginal liver graft. A furtherpossible reason why graft steatosis was not associatedwith poor graft function in the present study is thenegligible occurrence in the 7 transplants performedusing grafts with �30% macrovesicular steatosis ofcofactors known to negatively affect early postoperativefunction of steatotic livers, such as a total ischemia timelonger than 10 hours, donor age greater than 65 years,and recipient hepatitis C virus positivity.12 In accor-dance to our findings, efforts of transplant teams tominimize the occurrence of negative cofactors, togetherwith improved preoperative and postoperative patientcare, and refinement in surgical techniques can explainthe fact that the negative effect of graft liver steatosis onearly postoperative function has been less marked intransplants performed in recent years9,10 as compared totransplants performed earlier,8,11,12,28–30 and evenabsent in 1 recent large series.31 Similarly, an improveddonor ICU management in the last years can explainour present finding of a lack of association with trans-plant outcome of the donor length of ICU stay, thenumber of vasoactive drug infused and the occurrenceof hypernatriemia.10

Our present data show that if variables previouslyshown to be predictors of initial graft function with theexception of donor age are modified, preharvest donorhyperoxia is a marker of good early liver posttransplan-tation function. However, our finding is discordantfrom the previously reported lack of association of

Figure 4. Probability of graft (upper section) and patient(lower section) survival in liver transplantation performedusing grafts obtained from donors with PaO2 > 150mmHg (n � 28) or from donors with PaO2 < 150 mmHg(n � 58), irrespective of early postoperative graft func-tion.

148 Corradini et al.

donor PaO2 values with the quality of initial graft func-tion after human liver transplantation and from theeven beneficial effect of transient donor hypoxia previ-ously reported.14,15 One possibility is that donor hyper-oxia is a good predictor of IGGF only when the otherabove mentioned predictors of initial graft function areeliminated. On the other hand, an alternate explanationfor our discrepant results on donor PaO2 from previousstudies is the different timing of gas analysis measure-ment with respect to organ harvesting. In fact, in thepreviously published papers donor PaO2 measurementsobtained at an undetermined time point during ICUstay or at a mean of 2.9 days before organ procurementdid not reflect the degree of preharvest oxygenation andare difficult to relate with postoperative liver func-tion.14,15 At variance, although we cannot exclude thatin some cases donor PaO2 could have deteriorated aftermeasurement, our gas analytical data obtained nolonger than 4 hours before organ harvest should reflectthe degree of preharvest oxygenation. Our present datashow that donor hyperoxia represent a predictor ofIGGF, while no causative relation can be demonstrated.Thus, donor hyperoxia could simply be a nonmodifi-able marker of an optimal liver donor. In theory, abeneficial effect of donor hyperoxia might be possible,since it has been experimentally shown that preharvest-ing systemic hyperoxia attenuates ischemic-reperfusioninjury of the heart.32–34 In addition, donor hyperoxiamight have a beneficial effect similar to that of venous-systemic oxygen persufflation during cold preservationof the rat liver.35 Donor hyperoxia could induce goodearly postoperative liver function by increasing ATPhepatic content, by increasing protein synthesis andreducing the proteolytic processes in the liver and / orby inducing a sublethal burst of oxygen radicals such asthat involved in the protection of ischemic precondi-tioning against liver cold ischemic injury.35–40 How-ever, only further studies will clarify whether the induc-tion of systemic donor hyperoxia improves the livertransplantation outcome.

In the present study we found that donor treatmentwith noradrenaline was associated with initial poorgraft function. Data from the literature on liver donorvasopressor use and graft function and survival are con-flicting.1,41 Our finding that donor treatment withnoradrenaline was associated with initial poor graftfunction independently from donor serum aminotrans-ferase concentration and other indirect markers of per-fusion as serum creatinine and arterial pH could beexplained by several reasons. First, noradrenaline infu-sion could not reduce organ perfusion in the donor. Infact, although experimental data have shown that nor-

adrenaline infusion in vivo in healthy animals reducesliver perfusion, the use of vasopressors in the critically illand unstable organ donor could be beneficial and notdetrimental for liver perfusion by increasing blood pres-sure.42 Second, under uncontrolled (not experimental)clinical conditions donor aminotransferases, creatinineand arterial pH could be affected by factors differentfrom organ perfusion. Third, even if noradrenalineinfusion would reduce liver perfusion, the negativeeffect on hepatocytes could occur only after ischemiaand reperfusion, but not be clinically measurable in thedonor. In keeping with this hypothesis, in a previousstudy hepatic sympathetic nerve stimulation reducedliver perfusion flow in both galactosamine damaged andcontrol rat livers but increased LDH and aspartate ami-notransferase leakage in galactosamine damaged liversand not in controls.43 Finally, noradrenaline infusioncould exert a direct negative effect on the ischemia-reperfusion injury independently from the reduction ofpreharvest organ perfusion, by upregulating tumornecrosis factor-� production in hepatic Kupffer cells,which are known to be involved in the ischemia-reper-fusion injury process.44,45

Criteria used to define the quality of initial graftfunction after liver transplantation have not been con-sistent in the literature.4–6,9,11,16–22 In the present studywe found that two of these criteria which have beenwidely used were substantially concordant in the classi-fication of grafts as IPGF or IGGF. In addition,although we showed only the analysis performed usingthe Spanish criteria, both criteria permitted recognitionof the same predictors of the quality of initial graftfunction. Furthermore, the predictors of the quality ofinitial graft function we found have a clinical impactbecause the 2 criteria we used to define the quality ofinitial graft function had a good prognostic power.Heise et al.19 have recently criticized the criteria used toclassify initial graft function, which are based on arbi-trarily chosen cutoff levels of biochemical variables andbile output recorded on a single postoperative day,because they are unable to predict survival after livertransplantation. However, as shown in the presentstudy, the criteria for initial graft function definition,even if based on arbitrarily chosen cutoff levels, corre-late with graft and patient survival, provided that con-sideration is given, as originally described, to the aver-age bile output and worst biochemical variablesobtained during the entire time span of 72 hours aftertransplantation.5,22

In conclusion, this study is the first demonstrationthat preharvest donor hyperoxia helps to define theoptimal donor for deceased donor adult liver transplan-

149Donor Hyperoxia and Liver Transplantation

tation and is associated with longer graft survival. Inaddition, early postoperative graft function assessedduring the first 72 postoperative hours predicts graftand patient survival.

Acknowledgment

We thank Mary Waters for revising the English text.

References1. Busuttil RW, Tanaka K. The utility of marginal donors in liver

transplantation. Liver Transpl 2003;9:651–663.2. Yanaga K, Tzakis AG, Starzl TE. Personal experience with the

procurement of 132 liver allografts. Transpl Int 1989;2:137–142.

3. Mor E, Klintmalm GB, Gonwa TA, Solomon H, Holman MJ,Gibbs JF, et al. The use of marginal donors for liver transplanta-tion. Transplantation 1992;53:383–386.

4. Ploeg RJ, D’Alessandro AM, Knechtle SJ, Stegall MD, PirschJD, Hoffmann RM, et al. Risk factors for primary dysfunctionafter liver transplantation—a multivariate analysis. Transplanta-tion 1993;55:807–813.

5. Gonzalez FX, Rimola A, Grande L, Antolin M, Garcia-Valdeca-sas JC, Fuster J, et al. Predictive factors of early postoperativegraft function in human liver transplantation. Hepatology 1994;20:565–573.

6. Strasberg SM, Howard TK, Molmenti EP, Hertl M. Selectingthe donor liver: risk factors for poor function after orthotopicliver transplantation. Hepatology 1994;20:829–838.

7. De Carlis L, Sansalone CV, Rondinara GF, Colella G, Slim AO,Rossi O, et al. Is the use of marginal donors justified in livertransplantation? Analysis of results and proposal of modern cri-teria. Transpl Int 1996;9:S414–S417.

8. Chui AK, Shi LW, Rao ARN, Verran DJ, Painter D, Koorey D,et al. Donor fatty (steatotic) liver allografts in orthotopic livertransplantation: a revisit. Transplant Proc 2000;32:2101–2102.

9. Nanashima A, Pillay P, Verran DJ, Painter D, Nakasuji M,Crawford M, et al. Analysis of initial poor graft function afterorthotopic liver transplantation: experience of an Australian sin-gle liver transplantation center. Transplant Proc 2002;34:1231—1235.

10. Rull R, Vidal O, Momblan D, Gonzalez FX, Lopez-Boado MA,Fuster J, et al. Evaluation of potential liver donors: limitsimposed by donor variables in liver transplantation. LiverTranspl 2003;9:389–393.

11. Verran D, Kusyk T, Painter D, Fisher J, Korrey D, Strasser S, etcl. Clinical experience gained from the use of 120 steatotic donorlivers for orthotopic liver transplantation. Liver Transpl 2003;9:500–505.

12. Salizzoni M, Franchello A, Zamboni F, Ricchiuti A, Cocchis D,Fop F, et al. Marginal grafts: finding the correct treatment forfatty livers. Transpl Int 2003;16:486–493.

13. McKinley BA, Parmley CL. Clinical trial of an ex vivo arterialblood gas monitor. J Crit Care 1998;13:190–197.

14. Figueras J, Busquets J, Grande L, Jaurrieta E, Perez-Ferreiroa J,Mir J, et al. The deleterious effect of donor high plasma sodiumand extended preservation in liver transplantation. Transplanta-tion 1996;61:410–413.

15. Totsuka E, Fung JJ, Urakami A, Moras N, Ishii T, Takahashi K,

et al. Influence of donor cardiopulmonary arrest in human livertransplantation: possible role of ischemic preconditioning.Hepatology 2000;31:577–580.

16. Maring JK, Klompmaker IJ, Zwaveling JH, Kranenburg K, TenVergert EM, Slooff MJ. Poor initial graft function after ortho-topic liver transplantation: can it be predicted and does it affectoutcome? An analysis of 125 adult primary transplantations.Clin Transplant 1997;11:373–379.

17. Ardite E, Ramos C, Rimola A, Grande L, Fernandez-Checa JC.Hepatocellular oxidative stress and initial graft injury in humanliver transplantation. J Hepatol 1999;31:921–927.

18. Fernandez-Merino FJ, Nuno-Garza J, Lopez-Hervas P, Sanjuan-benito A, Lopez-Buenadicha A, Quijano-Collazo Y, et al. Sur-vival and complication rates in liver transplant patients showingprimary dysfunction versus normofunction. Transplant Proc2003;35:1815–1817.

19. Heise M, Settmacher U, Pfitzmann R, Wunscher U, Muller AR,Jonas S, Neuhaus P. A survival-based scoring-system for initialgraft function following orthotopic liver transplantation.Transpl Int 2003;16:794–800.

20. Garcia-Valdecasas JC, Gonzalez FJ, Grande L, Rimola A, NavasaM, Fuster J, et al. The use of the University of Wisconsin andEuro-Collins solutions either alone or in a combined method.Transpl Int 1992;5:77–80.

21. Grande L, Rimola A, Garcia-Valdecasas JC, Mas A, Fuster J,Navasa M, et al. Recovery of liver graft after initial poor function.Transplantation 1992;53:228–230.

22. Verran DJ, Gurkan A, Dilworth P, Painter D, Chui A, Dorney S,et al. Inferior liver allograft survival from cadaveric donors � 50years of age? Clinical Transplant 2001;15:106–110.

23. Kamath PS, Wiesner RH, Malinchoc M, Kremers W, TherneauTM, Kosberg CL, et al. A model to predict survival in patientswith end-stage liver disease. Hepatology 2001;33:464–470.

24. Chui AK, Haghighi K, Painter D, Jayasundera M, Hall G, RaoARN,, et al. Donor fatty (steatotic) liver allograft in orthotopicliver transplantation. Transplant Proc 1998;30:3286–3287.

25. Sackett DL, Haynes RB, Guyatt GH, Tugwell P. Clinical Epi-demiology: a basic science for clinical medicine. Little Brown andCo., Boston/Toronto, 2nd ed., 1991:24–34.

26. Osborn DA, Lockley C, Jeffery HE, Lloyd J, Lazarus R. Inter-observer reliability of the click test: a rapid bedside test to deter-mine surfactant function. J Paediatr Child Health 1998;34:544–547.

27. Briceno J, Solorzano G, Pera C. A proposal for scoring marginalliver grafts. Transpl Int 2000;13:S249–S252.

28. Todo S, Demetris AJ, Makowka L, Teperman L, Podesta L, ShaverT, et al. Primary nonfunction of hepatic allografts with preexistingfatty infiltration. Transplantation 1989;47:903–905.

29. D’Alessandro AM, Kalayoglu M, Sollinger HW, Hoffmann RM,Reed A, Knechtle SJ, et al. The predictive value of donor liverbiopsies on the development of primary nonfunction after ortho-topic liver transplantation. Transplant Proc 1991;23:1536–1537.

30. Adam R, Reynes M, Johann M, Morino M, Astarcioglu I, Kaf-etzis I, et al. The outcome of steatotic grafts in liver transplanta-tion. Transplant Proc 1991;23:1538–1540.

31. Yoo HY, Molmenti E, Thuluvath PJ. The effect of donor bodymass index on primary graft nonfunction, retransplantation rate,and early graft and patient survival after liver transplantation.Liver Transpl 2003;9:72–78.

32. Tahepold P, Valen G, Starkopf J, Karaine C, Zilmer M, Vaage J.Pretreating rats with hyperoxia attenuates ischemia-reperfusioninjury of the heart. Life Sci 2001;68:1629–1640.

150 Corradini et al.

33. Tahepold P, Ruusalepp A, Li G, Vaage J, Starkopf J, Valen G.Cardioprotection by breathing hyperoxic gas-relation to oxygenconcentration and exposure time in rats and mice. Eur J Cardio-thorac Surg 2002;21:987–994.

34. Tahepold P, Vaage J, Starkopf J, Valen G. Hyperoxia elicitsmyocardial protection through a nuclear factor �B-dependentmechanism in the rat heart. J Thorac Cardiovasc Surg 2003;125:650–660.

35. Minor T, Klauke H, Nagelschmidt M, Isselhard W. Reductionof proteolysis by venous-systemic oxygen persufflation during ratliver preservation and improved functional outcome after trans-plantation. Transplantation 1997;63:365–368.

36. Kamiike W, Burdelski M, Steinhoff G, Ringe B, Lauchart W,Pichlmayr R. Adenine nucleotide metabolism and its relation toorgan viability in human liver transplantation. Transplantation1988;45:138–143.

37. Lanir A, Jenkins RL, Caldwell C, Lee RGL, Khettry U, ClouseME. Hepatic transplantation survival: correlation with adeninenucleotide level in donor liver. Hepatology 1988;8:471–475.

38. Yanagi K, Oshima N. Improvement of metabolic performance ofcultured hepatocytes by high oxygen tension in the atmosphere.Artif Organs 2001;25:1–6.

39. Calmus Y, Cynober L, Dousset B, Lim SK, Soubrane O, Conti F, et

al. Evidence for the detrimental role of proteolysis during liver pres-ervation in humans. Gastroenterology 1995;108:1510–1516.

40. Sindram D, Rudiger HA, Upadhya AG, Strasberg SM, ClavienPA. Ischemic preconditioning protects against cold ischemicinjury through an oxidative stress dependent mechanism.J Hepatol 2002;36:78–84.

41. Broughan TA, Douzdjian V. Donor liver selection. The South-Eastern Organ Procurement Foundation Liver Committee. AmSurg 1998;64:785–790.

42. Wang P, Tait SM, Chaudry IH. Sustained elevation of norepi-nephrine depresses hepatocellular function. Biochim BiophysActa 2000;1535:36–44.

43. Iwai M, Shimazu T. Exaggeration of acute liver damage byhepatic sympathetic nerves and circulating catecholamines inperfused liver of rats treated with d-galactosamine. Hepatology1996;23:524–529.

44. Yang S, Koo DJ, Zhou M, Chaudry IH, Wang P. Gut-derivednorepinephrine plays a critical role in producing hepatocellulardysfunction during early sepsis. Am J Physiol Gastrointest LiverPhysiol 2000;279:G1274–G1281.

45. Zhou M, Yang S, Koo DJ, Ornan DA, Chaudry IH, Wang P.The role of Kupffer cell alpha(2)-adrenoceptors in norepineph-rine induced TNF-alpha production. Biochim Biophys Acta2001;1537:49–57.

151Donor Hyperoxia and Liver Transplantation