Liver International ISSN 1478-3223
ORIGINAL ARTICLE
Donor information based prediction of early allograft dysfunction and outcome in liver transplantation Dieter P. Hoyer1, Andreas Paul1, Anja Gallinat1, Ernesto P. Molmenti2, Renate Reinhardt1, Thomas Minor3, 1 € rgen W. Treckmann1, Georgios C. Sotiropoulos1 and Zoltan Mathe Fuat H. Saner1, Ali Canbay4, Ju 1 2 3 4
Department of General, Visceral and Transplantation Surgery, University Hospital, Essen, Germany Department of Surgery, North Shore University Hospital, Manhasset, NY, USA Department of Experimental Surgery, University Hospital, Bonn, Germany Department of Hepatology and Gastroenterology, University Hospital, Essen, Germany
Keywords donor risk factors – early allograft dysfunction – liver transplantation – multivariable analysis – scoring system
Correspondence Dieter P. Hoyer, MD, Department of General, Visceral and Transplantation Surgery, University Hospital Essen, Hufelandstraße 55, 45122 Essen, Germany Tel: +0049-201-7231100 Fax: +0049-201-7231142 e-mail: [email protected] Received 25 September 2013 Accepted 12 December 2013 DOI:10.1111/liv.12443
Abstract Background & Aims: Poor initial graft function was recently newly defined as early allograft dysfunction (EAD) [Olthoff KM, Kulik L, Samstein B, et al. Validation of a current definition of early allograft dysfunction in liver transplant recipients and analysis of risk factors. Liver Transpl 2010; 16: 943]. Aim of this analysis was to evaluate predictive donor information for development of EAD. Methods: Six hundred and seventy-eight consecutive adult patients (mean age 51.6 years; 60.3% men) who received a primary liver transplantation (LT) (09/2003–12/2011) were included. Standard donor data were correlated with EAD and outcome by univariable/multivariable logistic regression and Cox proportional hazards to identify prognostic donor factors after adjustment for recipient confounders. Estimates of relevant factors were utilized for construction of a new continuous risk index to develop EAD. Results: 38.7% patients developed EAD. 30-day survival of grafts with and without EAD was 59.8% and 89.7% (P < 0.0001). 30-day survival of patients with and without EAD was 68.5% and 93.1% (P < 0.0001) respectively. Donor body mass index (P = 0.0112), gGT (P = 0.0471), macrosteatosis (P = 0.0006) and cold ischaemia time (CIT) (P = 0.0031) were predictors of EAD. Internal cross validation showed a high predictive value (c-index = 0.622). Conclusions: Early allograft dysfunction correlates with early results of LT and can be predicted by donor data only. The newly introduced risk index potentially optimizes individual decisions to accept/decline high risk organs. Outcome of these organs might be improved by shortening CIT.
As efforts to increase the number of donor organs evolve world-wide beyond hitherto existing boundaries (1), so does the incidence of suboptimal graft function, a complication characterized by its high morbidity and mortality. Early Allograft Dysfunction (EAD) was recently newly defined (2). This definition aimed at identifying a group of liver grafts with clinical expression of suboptimal function and showed a strong correlation with patient and graft survival. The concept of this new definition, based exclusively on laboratory data, represents an objective tool that can help in the assessment of early allograft function/dysfunction. In fact, this definition is outstanding since the pre-operative status of the recipient does not have any influence on this post-operative classification. Recipient, surgical and donor variables, as well as scoring/grading systems have been proposed as predictors of outcomes in orthotopic liver transplantation (LT) (3–9). The establishment of the donor risk index (DRI) by Feng et al. (10) and its validation within the Liver International (2014) © 2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Eurotransplant area (11) (ET-DRI) stressed the importance of donor factors for the outcome of LT. While these definitions analyzed risk factors for the overall graft survival, the early graft dysfunction and the accompanying complicated clinical course was not specifically taken into account. Using early graft function/ dysfunction as defined by EAD as outcome parameter has the advantage to address exactly these post-operative courses of high risk with the opportunity to characterize risk factors that predict the immediate post-operative function of the allograft. The post-operative function of the allograft in turn is a major aspect of donor-recipient matching. Therefore, adequate recipients could be selected based on the expected graft function. The exclusive use of donor variables as a way to assess risk is both reasonable and advantageous. Firstly, post-operative function is greatly dependent on the extent of ischaemic injury sustained by the graft during procurement and cold storage. Secondly, donor data are
1
Hoyer et al.
Prediction of EAD in LT
universally available at the time of organ allocation. Thirdly, since recipients are determined according to the Model of End Stage Liver Disease (MELD) allocation system, the only variables that can be altered are those associated with the donor. The aim of this study was to further evaluate prognostic donor factors and to develop a scoring system predictive of EAD. Patients and methods Study population
We evaluated data corresponding to all LTs performed from September 2003 to December 2011 at the
University Hospital of Essen, Germany. Recipients 2000 IU/L within the first 7 days (2). Each case was classified as ‘EAD’ or ‘no-EAD’. For recipients who died within 7 days after transplantation, laboratory and clinical parameters up to the time of death were considered for the classification.
Prediction of EAD in LT
laboratory values, the DRI and the Eurotransplant Donor Risk Index. Statistical analysis
Data were expressed as mean and standard deviation and as median and range values where appropriate. Multiple imputation (12) was performed for incomplete donor data (e.g. macrovesicular steatosis in donor biopsies). Graft and patient survival were calculated using the Kaplan–Meier method and compared with the logrank test. Univariable and multivariable regression analyses were performed with binary logistic regression and Cox proportional hazard models. Variables statistically significant by univariable analyses were subsequently evaluated by multivariable analysis. To isolate the impact of donor characteristics on the development of EAD, the multivariable logistic regression model was adjusted for recipient and procedural factors that may impact the outcome. Risk ratios or odd ratios were obtained from regression models. Using results of the multivariable logistic regression analysis, a formula was constructed based on significant donor factors and their regression coefficients. P < 0.05 was considered significant. Statistical analyses were performed using JMP (version 10.0.0 SAS, SAS Institute Inc., Cary, NC, USA).
Definition of primary non function
Primary non function was defined as post-transplant liver dysfunction requiring retransplantation or leading to death within 7 days. Definition of rescue allocation
Livers refused by more than three different centres for allocated candidates with the highest MELD scores on the national waiting list were characterized as ‘organ rescue offers’. These grafts were then either offered to the nearest centre with a suitable recipient or allocated to the first centre to accept them (multiple-refusal/competitive rescue offer procedure). ‘Organ rescue offers’ were also occasionally encountered in instances of donor instability, prolonged cold ischaemic times (CITs), or unfavourable logistic reasons. Risk factors for development of EAD
The following donor variables were evaluated: age, gender, height, weight, body mass index (BMI), cause of death, CIT, length of stay in the ICU, vasopressor requirement (no vasopressor support, low = 0.5 lg/ kg/min), biopsy-determined steatosis (total, macro-, and microvesicular), rescue offer allocation, procurement team assessment of organ quality, split liver transplantation, organ preservation solution, most recent Liver International (2014) © 2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Results
This study population consisted of 678 patients who underwent LT between September 2003 and December 2011. Cohort A for primary analysis consisted of 475 subjects while cohort B for validation consisted of the remaining 203 subjects. No statistically significant differences were found between these cohorts in terms of basic recipient and donor data (data not shown). Donor characteristics
Mean donor age was 53 years (±17.6 years). 247 (52%) were men. Mean donor height was 172 cm (±13.2 cm) with a mean body weight of 79.2 kg (±18.2 kg), resulting in a mean BMI of 26.4 kg/m² (±4.9 kg/m²). Cause of death was cerebrovascular accident in 307 (65.3%), hypoxia in 40 (8.5%), trauma in 65 (13.8%), and other causes in 58 (12.3%). Median donor ICU length of stay was 3.8 days (0.5–47 days). Vasopressor requirements were documented as follows: high dosage in 46 (10.4%), moderate dosage in 163 (36.9%), low dosage in 169 (38.3%), and none in 63 (14.3%). Mean DRI was 1.7 (±0.39). Mean Eurotransplant DRI was 1.9 (±0.40). Perfusion solution was Histidine-Tryptophane-Ketoglutarate in 72.9% and University of Wisconsin in 27.1%. Mean cold and warm ischaemic times were 451 min (±143 min) and 34 min (±10 min) respectively. The frequency of liver biopsies correlated with the macroscopic liver quality as assessed by the procurement team.
3
Hoyer et al.
Prediction of EAD in LT
‘Good’ macroscopic liver quality was associated with a significantly lower (59.8%) number of biopsies than ‘acceptable’ (76.8%) or ‘poor’ (80%) ratings (P = 0.0008) (data not shown).
respectively (P = 0.0008). The corresponding median lengths of total hospital stay were 22 days (0–168) and 23 days (0–155) (P = 0.2477). Risk factors for developing EAD
Recipient characteristics
Mean recipient age at transplantation was 51.6 years (±10.5 years). 289 (60.8%) were male participants. Mean height was 173 cm (±10 cm) with a mean body weight of 79.7 kg (±16.8 kg), resulting in a mean BMI of 26.7 kg/m² (±4.9 kg/m²). Mean laboratory MELD score (labMELD) prior to transplantation was 19.2 (±9.9). Main indications for liver transplantation included viral hepatitis cirrhosis, hepatocellular carcinoma and alcoholic cirrhosis in 30.8%, 24.8% and 21.7% of recipients respectively. Patient and graft outcomes
One hundred and eighty-four (38.7%) of the 475 recipients developed EAD. The most frequent criterion leading to the diagnosis of EAD was elevated transaminases within the first seven post-operative days. Graft survival after 30 days was 59.8% in EAD patients compared to 89.7% in no-EAD patients (P < 0.0001). 12 month graft survival rates in EAD and no-EAD patients were 49.1% and 79.4% respectively (P < 0.0001). Patients with EAD showed a 30-day survival of 68.5% compared to 93.1% in those without EAD (P < 0.0001). 12-month survival for EAD and no-EAD patients was 59.4% and 81.4% respectively (P < 0.0001) (Fig. 2). The overall incidence of PNF was 8.4%. Forty (21.7%) of the 184 recipients with EAD developed PNF. Twenty-eight (70%) of the 40 were retransplanted after ‘high-urgency’ listing. Median length of ICU stay in instances of EAD and no-EAD was 6 days (1–161) and 4 days (0–127)
(A)
The following donor variables were found by univariable analysis to be associated with the development of EAD: Donor height (P = 0.0121), donor weight (P < 0.0001), donor BMI (P = 0.0009), last donor gGT (P = 0.0057), last donor serum sodium (P = 0.0407), macrovesicular steatosis by liver biopsy (P = 0.0002) and CIT (P < 0.0001). Donor male gender (P = 0.0514) showed a tendency to predict EAD (P = 0.0586). None of the other variables, including donor vasopressor therapy and total steatosis (data not shown), reached significance. Statistical cut-off values for significant factors of the univariable analysis were defined by ROC-analysis and showed the following thresholds: BMI = 25.3 kg/m²; last donor gGT = 55 U/L; last donor serum sodium = 155 mmol/L, macrovesicular steatosis = 15%; CIT = 7.4 h (444 min). Univariable analysis of recipient and surgical (procedure) factors showed that only length of surgery (P < 0.0002) and warm ischaemic time (P = 0.0053) were significantly associated with EAD (Table 1). Since recipient attributes are static and procedural variables are yet to be determined at the time of organ acceptance, a multivariable binary logistic regression analysis was built on the donor parameters that reached significance by univariable analyses (BMI, last gGT, last serum sodium, macrovesicular steatosis, CIT, donor male gender) and adjusted for significant procedural variables (length of surgery, warm ischaemic time) as well as basic recipient variables (age, BMI, labMELD). Donor male gender was included as well with its marginal significance. Donor height and donor weight were not included because of close correlation with donor BMI. Donor BMI and donor macrovesicular steatosis
(B)
No EAD
No EAD
EAD EAD P < 0.0001 log rank
No-EAD At risk: 291 Events: 0 EAD At risk: 184 Events: 0
P < 0.0001 log rank
236 51
200 59
93 88
79 93
No-EAD At risk: 291 Events: 0 EAD At risk: 184 Events: 0
243 33
206 42
112 69
96 74
Fig. 2. (A) 12-months-graft-survival in EAD and no-EAD patients. (B) 12-months-patient-survival in EAD and no-EAD patients.
4
Liver International (2014) © 2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Hoyer et al.
Prediction of EAD in LT
Table 1. Univariable analysis of donor and recipient details for risk of development of EAD
Donor data Gender Male Female Age (years) BMI (kg/m²) Cause of death Cerebrovascular accident Hypoxia Trauma Other ICU stay (days) Steatosis micro (%) Steatosis macro (%) LTX quality as assessed by surgeon Good Moderate Poor AST (U/L) ALT (U/L) gGT (U/L) Total bilirubin (lmol/L) INR Creatinine (lmol/L) Serum Sodium (mmol/L) Rescue offer Split liver Perfusion solution HTK UW Cold ischaemic time (min) DRI ET-DRI Recipient data Gender Male Female Age (years) BMI (kg/m²) Pre-LT mechanical ventilation (days) Pre-LT ICU stay (days) labMELD before LT ‘high-urgency’ listing Procedural data Time for surgery (min) Warm ischaemic time (min)
No-EAD n = 291
EAD n = 184
141 (48.5%) 150 (51.5%) 53.1 ± 18.9 25.9 ± 4.8
106 (57.6%) 78 (42.4%) 52.8 ± 15.4 27.4 ± 4.7
0.0514
185 (64.0%) 24 (8.3%) 43 (14.9%) 37 (12.8%) 3.6 (0.5–47) 14 (0–95) 5 (0–50)
122 (67.4%) 16 (8.8%) 22 (12.2%) 21 (11.6%) 4 (0.7–30.3) 13.5 (0–90) 10 (0–45)
0.8083
214 (76.7%) 61 (21.9%) 4 (1.4%) 50 (9–731) 32 (4–1200) 36.5 (5–766) 9.9 (0.4–246.5) 1.15 (0.87–5.76) 79.2 (0.9–1142.8) 148 (115–171) 159 (61.9%) 7 (2.4%)
127 (70.6%) 49 (27.2%) 4 (2.2%) 48 (12–1001) 35 (7–733) 50 (4–1346) 8.8 (0.3–180.2) 1.11 (0.82–8.95) 79.2 (26.4–3256) 148 (128–167) 106 (67.9%) 9 (4.9%)
202 (72.9%) 75 (27.1%) 429 ± 126 1.7 ± 0.4 1.9 ± 0.4
127 (73.0%) 47 (27%) 488 ± 160 1.7 ± 0.4 1.9 ± 0.4
174 (59.8%) 117 (40.2%) 51.3 ± 10.6 26.5 ± 4.7 0 (0–17)
115 (62.5%) 69 (37.5%) 52.2 ± 10.2 26.9 ± 5.3 0 (0–17)
0.5557
0 (0–60) 19.5 ± 9.9 27 (9.5%)
0 (0–17) 18.7 ± 9.9 19 (10.4%)
0.7492 0.4238 0.7479
284 ± 83 32.9 ± 9.8
319 ± 97 35.8 ± 10.5
0.0002 0.0053
P-value
0.8798 0.0009
0.1919 0.7293 0.0002 0.3273
0.4999 0.9502 0.0057 0.6645 0.9143 0.3739 0.0407 0.2099 0.1468 0.9881
ORIGINAL ARTICLE
Donor information based prediction of early allograft dysfunction and outcome in liver transplantation Dieter P. Hoyer1, Andreas Paul1, Anja Gallinat1, Ernesto P. Molmenti2, Renate Reinhardt1, Thomas Minor3, 1 € rgen W. Treckmann1, Georgios C. Sotiropoulos1 and Zoltan Mathe Fuat H. Saner1, Ali Canbay4, Ju 1 2 3 4
Department of General, Visceral and Transplantation Surgery, University Hospital, Essen, Germany Department of Surgery, North Shore University Hospital, Manhasset, NY, USA Department of Experimental Surgery, University Hospital, Bonn, Germany Department of Hepatology and Gastroenterology, University Hospital, Essen, Germany
Keywords donor risk factors – early allograft dysfunction – liver transplantation – multivariable analysis – scoring system
Correspondence Dieter P. Hoyer, MD, Department of General, Visceral and Transplantation Surgery, University Hospital Essen, Hufelandstraße 55, 45122 Essen, Germany Tel: +0049-201-7231100 Fax: +0049-201-7231142 e-mail: [email protected] Received 25 September 2013 Accepted 12 December 2013 DOI:10.1111/liv.12443
Abstract Background & Aims: Poor initial graft function was recently newly defined as early allograft dysfunction (EAD) [Olthoff KM, Kulik L, Samstein B, et al. Validation of a current definition of early allograft dysfunction in liver transplant recipients and analysis of risk factors. Liver Transpl 2010; 16: 943]. Aim of this analysis was to evaluate predictive donor information for development of EAD. Methods: Six hundred and seventy-eight consecutive adult patients (mean age 51.6 years; 60.3% men) who received a primary liver transplantation (LT) (09/2003–12/2011) were included. Standard donor data were correlated with EAD and outcome by univariable/multivariable logistic regression and Cox proportional hazards to identify prognostic donor factors after adjustment for recipient confounders. Estimates of relevant factors were utilized for construction of a new continuous risk index to develop EAD. Results: 38.7% patients developed EAD. 30-day survival of grafts with and without EAD was 59.8% and 89.7% (P < 0.0001). 30-day survival of patients with and without EAD was 68.5% and 93.1% (P < 0.0001) respectively. Donor body mass index (P = 0.0112), gGT (P = 0.0471), macrosteatosis (P = 0.0006) and cold ischaemia time (CIT) (P = 0.0031) were predictors of EAD. Internal cross validation showed a high predictive value (c-index = 0.622). Conclusions: Early allograft dysfunction correlates with early results of LT and can be predicted by donor data only. The newly introduced risk index potentially optimizes individual decisions to accept/decline high risk organs. Outcome of these organs might be improved by shortening CIT.
As efforts to increase the number of donor organs evolve world-wide beyond hitherto existing boundaries (1), so does the incidence of suboptimal graft function, a complication characterized by its high morbidity and mortality. Early Allograft Dysfunction (EAD) was recently newly defined (2). This definition aimed at identifying a group of liver grafts with clinical expression of suboptimal function and showed a strong correlation with patient and graft survival. The concept of this new definition, based exclusively on laboratory data, represents an objective tool that can help in the assessment of early allograft function/dysfunction. In fact, this definition is outstanding since the pre-operative status of the recipient does not have any influence on this post-operative classification. Recipient, surgical and donor variables, as well as scoring/grading systems have been proposed as predictors of outcomes in orthotopic liver transplantation (LT) (3–9). The establishment of the donor risk index (DRI) by Feng et al. (10) and its validation within the Liver International (2014) © 2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Eurotransplant area (11) (ET-DRI) stressed the importance of donor factors for the outcome of LT. While these definitions analyzed risk factors for the overall graft survival, the early graft dysfunction and the accompanying complicated clinical course was not specifically taken into account. Using early graft function/ dysfunction as defined by EAD as outcome parameter has the advantage to address exactly these post-operative courses of high risk with the opportunity to characterize risk factors that predict the immediate post-operative function of the allograft. The post-operative function of the allograft in turn is a major aspect of donor-recipient matching. Therefore, adequate recipients could be selected based on the expected graft function. The exclusive use of donor variables as a way to assess risk is both reasonable and advantageous. Firstly, post-operative function is greatly dependent on the extent of ischaemic injury sustained by the graft during procurement and cold storage. Secondly, donor data are
1
Hoyer et al.
Prediction of EAD in LT
universally available at the time of organ allocation. Thirdly, since recipients are determined according to the Model of End Stage Liver Disease (MELD) allocation system, the only variables that can be altered are those associated with the donor. The aim of this study was to further evaluate prognostic donor factors and to develop a scoring system predictive of EAD. Patients and methods Study population
We evaluated data corresponding to all LTs performed from September 2003 to December 2011 at the
University Hospital of Essen, Germany. Recipients 2000 IU/L within the first 7 days (2). Each case was classified as ‘EAD’ or ‘no-EAD’. For recipients who died within 7 days after transplantation, laboratory and clinical parameters up to the time of death were considered for the classification.
Prediction of EAD in LT
laboratory values, the DRI and the Eurotransplant Donor Risk Index. Statistical analysis
Data were expressed as mean and standard deviation and as median and range values where appropriate. Multiple imputation (12) was performed for incomplete donor data (e.g. macrovesicular steatosis in donor biopsies). Graft and patient survival were calculated using the Kaplan–Meier method and compared with the logrank test. Univariable and multivariable regression analyses were performed with binary logistic regression and Cox proportional hazard models. Variables statistically significant by univariable analyses were subsequently evaluated by multivariable analysis. To isolate the impact of donor characteristics on the development of EAD, the multivariable logistic regression model was adjusted for recipient and procedural factors that may impact the outcome. Risk ratios or odd ratios were obtained from regression models. Using results of the multivariable logistic regression analysis, a formula was constructed based on significant donor factors and their regression coefficients. P < 0.05 was considered significant. Statistical analyses were performed using JMP (version 10.0.0 SAS, SAS Institute Inc., Cary, NC, USA).
Definition of primary non function
Primary non function was defined as post-transplant liver dysfunction requiring retransplantation or leading to death within 7 days. Definition of rescue allocation
Livers refused by more than three different centres for allocated candidates with the highest MELD scores on the national waiting list were characterized as ‘organ rescue offers’. These grafts were then either offered to the nearest centre with a suitable recipient or allocated to the first centre to accept them (multiple-refusal/competitive rescue offer procedure). ‘Organ rescue offers’ were also occasionally encountered in instances of donor instability, prolonged cold ischaemic times (CITs), or unfavourable logistic reasons. Risk factors for development of EAD
The following donor variables were evaluated: age, gender, height, weight, body mass index (BMI), cause of death, CIT, length of stay in the ICU, vasopressor requirement (no vasopressor support, low = 0.5 lg/ kg/min), biopsy-determined steatosis (total, macro-, and microvesicular), rescue offer allocation, procurement team assessment of organ quality, split liver transplantation, organ preservation solution, most recent Liver International (2014) © 2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Results
This study population consisted of 678 patients who underwent LT between September 2003 and December 2011. Cohort A for primary analysis consisted of 475 subjects while cohort B for validation consisted of the remaining 203 subjects. No statistically significant differences were found between these cohorts in terms of basic recipient and donor data (data not shown). Donor characteristics
Mean donor age was 53 years (±17.6 years). 247 (52%) were men. Mean donor height was 172 cm (±13.2 cm) with a mean body weight of 79.2 kg (±18.2 kg), resulting in a mean BMI of 26.4 kg/m² (±4.9 kg/m²). Cause of death was cerebrovascular accident in 307 (65.3%), hypoxia in 40 (8.5%), trauma in 65 (13.8%), and other causes in 58 (12.3%). Median donor ICU length of stay was 3.8 days (0.5–47 days). Vasopressor requirements were documented as follows: high dosage in 46 (10.4%), moderate dosage in 163 (36.9%), low dosage in 169 (38.3%), and none in 63 (14.3%). Mean DRI was 1.7 (±0.39). Mean Eurotransplant DRI was 1.9 (±0.40). Perfusion solution was Histidine-Tryptophane-Ketoglutarate in 72.9% and University of Wisconsin in 27.1%. Mean cold and warm ischaemic times were 451 min (±143 min) and 34 min (±10 min) respectively. The frequency of liver biopsies correlated with the macroscopic liver quality as assessed by the procurement team.
3
Hoyer et al.
Prediction of EAD in LT
‘Good’ macroscopic liver quality was associated with a significantly lower (59.8%) number of biopsies than ‘acceptable’ (76.8%) or ‘poor’ (80%) ratings (P = 0.0008) (data not shown).
respectively (P = 0.0008). The corresponding median lengths of total hospital stay were 22 days (0–168) and 23 days (0–155) (P = 0.2477). Risk factors for developing EAD
Recipient characteristics
Mean recipient age at transplantation was 51.6 years (±10.5 years). 289 (60.8%) were male participants. Mean height was 173 cm (±10 cm) with a mean body weight of 79.7 kg (±16.8 kg), resulting in a mean BMI of 26.7 kg/m² (±4.9 kg/m²). Mean laboratory MELD score (labMELD) prior to transplantation was 19.2 (±9.9). Main indications for liver transplantation included viral hepatitis cirrhosis, hepatocellular carcinoma and alcoholic cirrhosis in 30.8%, 24.8% and 21.7% of recipients respectively. Patient and graft outcomes
One hundred and eighty-four (38.7%) of the 475 recipients developed EAD. The most frequent criterion leading to the diagnosis of EAD was elevated transaminases within the first seven post-operative days. Graft survival after 30 days was 59.8% in EAD patients compared to 89.7% in no-EAD patients (P < 0.0001). 12 month graft survival rates in EAD and no-EAD patients were 49.1% and 79.4% respectively (P < 0.0001). Patients with EAD showed a 30-day survival of 68.5% compared to 93.1% in those without EAD (P < 0.0001). 12-month survival for EAD and no-EAD patients was 59.4% and 81.4% respectively (P < 0.0001) (Fig. 2). The overall incidence of PNF was 8.4%. Forty (21.7%) of the 184 recipients with EAD developed PNF. Twenty-eight (70%) of the 40 were retransplanted after ‘high-urgency’ listing. Median length of ICU stay in instances of EAD and no-EAD was 6 days (1–161) and 4 days (0–127)
(A)
The following donor variables were found by univariable analysis to be associated with the development of EAD: Donor height (P = 0.0121), donor weight (P < 0.0001), donor BMI (P = 0.0009), last donor gGT (P = 0.0057), last donor serum sodium (P = 0.0407), macrovesicular steatosis by liver biopsy (P = 0.0002) and CIT (P < 0.0001). Donor male gender (P = 0.0514) showed a tendency to predict EAD (P = 0.0586). None of the other variables, including donor vasopressor therapy and total steatosis (data not shown), reached significance. Statistical cut-off values for significant factors of the univariable analysis were defined by ROC-analysis and showed the following thresholds: BMI = 25.3 kg/m²; last donor gGT = 55 U/L; last donor serum sodium = 155 mmol/L, macrovesicular steatosis = 15%; CIT = 7.4 h (444 min). Univariable analysis of recipient and surgical (procedure) factors showed that only length of surgery (P < 0.0002) and warm ischaemic time (P = 0.0053) were significantly associated with EAD (Table 1). Since recipient attributes are static and procedural variables are yet to be determined at the time of organ acceptance, a multivariable binary logistic regression analysis was built on the donor parameters that reached significance by univariable analyses (BMI, last gGT, last serum sodium, macrovesicular steatosis, CIT, donor male gender) and adjusted for significant procedural variables (length of surgery, warm ischaemic time) as well as basic recipient variables (age, BMI, labMELD). Donor male gender was included as well with its marginal significance. Donor height and donor weight were not included because of close correlation with donor BMI. Donor BMI and donor macrovesicular steatosis
(B)
No EAD
No EAD
EAD EAD P < 0.0001 log rank
No-EAD At risk: 291 Events: 0 EAD At risk: 184 Events: 0
P < 0.0001 log rank
236 51
200 59
93 88
79 93
No-EAD At risk: 291 Events: 0 EAD At risk: 184 Events: 0
243 33
206 42
112 69
96 74
Fig. 2. (A) 12-months-graft-survival in EAD and no-EAD patients. (B) 12-months-patient-survival in EAD and no-EAD patients.
4
Liver International (2014) © 2013 John Wiley & Sons A/S. Published by John Wiley & Sons Ltd
Hoyer et al.
Prediction of EAD in LT
Table 1. Univariable analysis of donor and recipient details for risk of development of EAD
Donor data Gender Male Female Age (years) BMI (kg/m²) Cause of death Cerebrovascular accident Hypoxia Trauma Other ICU stay (days) Steatosis micro (%) Steatosis macro (%) LTX quality as assessed by surgeon Good Moderate Poor AST (U/L) ALT (U/L) gGT (U/L) Total bilirubin (lmol/L) INR Creatinine (lmol/L) Serum Sodium (mmol/L) Rescue offer Split liver Perfusion solution HTK UW Cold ischaemic time (min) DRI ET-DRI Recipient data Gender Male Female Age (years) BMI (kg/m²) Pre-LT mechanical ventilation (days) Pre-LT ICU stay (days) labMELD before LT ‘high-urgency’ listing Procedural data Time for surgery (min) Warm ischaemic time (min)
No-EAD n = 291
EAD n = 184
141 (48.5%) 150 (51.5%) 53.1 ± 18.9 25.9 ± 4.8
106 (57.6%) 78 (42.4%) 52.8 ± 15.4 27.4 ± 4.7
0.0514
185 (64.0%) 24 (8.3%) 43 (14.9%) 37 (12.8%) 3.6 (0.5–47) 14 (0–95) 5 (0–50)
122 (67.4%) 16 (8.8%) 22 (12.2%) 21 (11.6%) 4 (0.7–30.3) 13.5 (0–90) 10 (0–45)
0.8083
214 (76.7%) 61 (21.9%) 4 (1.4%) 50 (9–731) 32 (4–1200) 36.5 (5–766) 9.9 (0.4–246.5) 1.15 (0.87–5.76) 79.2 (0.9–1142.8) 148 (115–171) 159 (61.9%) 7 (2.4%)
127 (70.6%) 49 (27.2%) 4 (2.2%) 48 (12–1001) 35 (7–733) 50 (4–1346) 8.8 (0.3–180.2) 1.11 (0.82–8.95) 79.2 (26.4–3256) 148 (128–167) 106 (67.9%) 9 (4.9%)
202 (72.9%) 75 (27.1%) 429 ± 126 1.7 ± 0.4 1.9 ± 0.4
127 (73.0%) 47 (27%) 488 ± 160 1.7 ± 0.4 1.9 ± 0.4
174 (59.8%) 117 (40.2%) 51.3 ± 10.6 26.5 ± 4.7 0 (0–17)
115 (62.5%) 69 (37.5%) 52.2 ± 10.2 26.9 ± 5.3 0 (0–17)
0.5557
0 (0–60) 19.5 ± 9.9 27 (9.5%)
0 (0–17) 18.7 ± 9.9 19 (10.4%)
0.7492 0.4238 0.7479
284 ± 83 32.9 ± 9.8
319 ± 97 35.8 ± 10.5
0.0002 0.0053
P-value
0.8798 0.0009
0.1919 0.7293 0.0002 0.3273
0.4999 0.9502 0.0057 0.6645 0.9143 0.3739 0.0407 0.2099 0.1468 0.9881
