Prediction Models of Donor Arrest and Graft Utilization in Liver Transplantation From Maastricht-3 Donors After Circulatory Death


  • D. Davila,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
    • Diego Davila and Ruben Ciria have equally contributed to the development of this manuscript.

  • R. Ciria,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
    • Diego Davila and Ruben Ciria have equally contributed to the development of this manuscript.

    • Ruben Ciria is supported by a scholarship from the Spanish Society of Liver Transplantation (Sociedad Española de Trasplante Hepático SETH 2009–2010).

  • W. Jassem,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • J. Briceño,

    1. Unit of Liver Transplantation, University Hospital Reina Sofia, Córdoba, Spain
    Search for more papers by this author
  • W. Littlejohn,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • H. Vilca-Meléndez,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • P. Srinivasan,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • A. Prachalias,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • J. O’Grady,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • M. Rela,

    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
    Search for more papers by this author
  • N. Heaton

    Corresponding author
    1. Institute of Liver Studies, King's Health Partners at King's College Hospital, London, United Kingdom
      Nigel Heaton,
    Search for more papers by this author

Nigel Heaton,


Shortage of organs for transplantation has led to the renewed interest in donation after circulatory–determination of death (DCDD). We conducted a retrospective analysis (2001–2009) and a subsequent prospective validation (2010) of liver Maastricht-Category-3-DCDD and donation-after-brain-death (DBD) offers to our program. Accepted and declined offers were compared. Accepted DCDD offers were divided into donors who went on to cardiac arrest and those who did not. Donors who arrested were divided into those producing grafts that were transplanted or remained unused. Descriptive comparisons and regression analyses were performed to assess predictor models of donor cardiac arrest and graft utilization. Variables from the multivariate analysis were prospectively validated. Of 1579 DCDD offers, 621 were accepted, and of these, 400 experienced cardiac arrest after withdrawal of support. Of these, 173 livers were transplanted. In the DCDD group, donor age < 40 years, use of inotropes and absence of gag/cough reflexes were predictors of cardiac arrest. Donor age >50 years, BMI >30, warm ischemia time >25 minutes, ITU stay >7 days and ALT ≥ 4× normal rates were risk factors for not using the graft. These variables had excellent sensitivity and specificity for the prediction of cardiac arrest (AUROC = 0.835) and graft use (AUROC = 0.748) in the 2010 prospective validation. These models can feasibly predict cardiac arrest in potential DCDDs and graft usability, helping to avoid unnecessary recoveries and healthcare expenditure.


accepted offers


alanine transaminotransferase


analysis of variance




body mass index


donation after brain death


donation after cardiac death


donation after circulatory determination of death


declined offers


gamma –glutamyl transferase


Intensive Therapy Unit


inferior vena cava


no arrest


not used


receiver-operating characteristic


superior mesenteric vein


total offers


used grafts


warm ischaemia time


The continuing shortage of donors has led to the use of increasingly marginal grafts which, 20 years ago, would not had been considered suitable for liver transplantation (1,2). Various types of donation including split (3), domino and living donor (4,5) liver transplantation have also emerged but have not made inroads into organ shortage (6). Prior to the introduction of brain stem death criteria in 1968 (7), donation after cardiac death (DCD) was the only source of grafts. Subsequently, donation after brain death (DBD) became the dominant source of grafts, displacing DCD, because of the improved graft quality and the potential for multiorgan donation. Over the last 10 years, donation after circulatory determination of death (DCDD) has been reintroduced with the hope that it would increase the number of grafts available for transplantation (8,9) by 10–20% (10).

An early experience of DCDD liver transplantation was reported by the Pittsburgh group (11), with primary nonfunction rates of 50% and graft and patient survival rates of 16% and 67% for uncontrolled and 50% and 50%, respectively for controlled DCDD (11). Further clinical studies quickly identified risk factors for graft failure, including prolonged donor warm ischemia, cold ischemia and donor age above 50 years. In 2001, our unit established a liver recovery service for DCDD Maastrich III and IV (12), which has grown over 9 years and accounts for 20–25% of our liver grafts. It is clear that DCDD recovery is different from DBD. The demands on the recovering team include the short notice given to attend donation often on a background of patient deterioration, of waiting for death to occur after withdrawal of medical support, the need for rapid but careful surgery and methods of perfusion in situ and on the back table and the need to make an early decision about organ suitability to allow for recipient preparation with a short cold ischemic time. The number of potential offers has grown significantly and the majority are turned down as unsuitable. Even when accepted, the potential donor may not arrest and thus, the recovery of a liver occurs in fewer than 50% of cases. This is in marked contrast to DBD where recovery to transplantation rates exceeds 90%. DCDD recovery is therefore labor intensive and carries significant expense for organ recovery and transplant services (13). Identification of appropriate donors/mode of death would potentially improve organ utilization without losing usable grafts. The aims of our study were to analyze the profile of DCDD and DBD donation and the rates of donor arrest and graft usability; second, to develop prediction models to improve DCDD recover; and, to validate the previous models in a prospective setting.

Patients and Methods

Patient selection

A combined retrospective (2001–2009) and prospective (2010) data collection and analysis were performed of all donation activity from DCDD in our Liver Transplant program at King's College Hospital London. According to the Maastricht classification of DCDD (12), four categories are recognized: (1) dead on arrival, (2) unsuccessful resuscitation, (3) awaiting cardiac arrest, and (4) cardiac arrest while brain death. To date, only category 3 and 4 DCDD (controlled) have been recovered within our program. To avoid a potential selection bias from category-4 donors, they were excluded from the analysis because of the predictability of cardiac arrest and the short warm ischemia time.

DCDD recovery procedure and definitions

A donor offer is made by the donor co-ordinator and depending on the clinical summary is accepted or declined. Following acceptance, the recovery team is sent to the donor hospital. Withdrawal of care in the Intensive Therapy Unit (ITU) or in the anaesthetic theatre room depends on local hospital policy. At treatment withdrawal, extubation is performed and supportive drugs are stopped by the medical team. Painkillers are administered, but no other drugs or medication (including heparin) are administered prior to cardiac arrest certification. The patient is monitored by electrocardiography (ECG), pulse oximetry and blood pressure recordings. Warm ischemia time (WIT) was defined as from the time systolic blood pressure fell to 50 mmHg or oxygen saturation fell to less than 70% and ended with donor ‘cold’ aortic perfusion. After withdrawal of treatment, the liver team remains for up to 1 hour waiting for asystole. If this has not occurred, the liver team withdraws. Death is declared when asystole is confirmed on ECG, with cessation of breathing and pulse oximetry showing 0%. After death has been declared by the medical team, a 5-minute stand-off time is observed. The donor is then transferred to theatre and recovery performed using the “superrapid technique” (10,11). Following aortic cannulation, University of Wisconsin (UW) solution is infused. Superior mesenteric vein (SMV) or portal vein cannulation is performed and the liver is perfused. Special attention is given to peritoneal cooling with ice to aid cooling of the liver, and to early venting of the inferior vena cava (IVC) to prevent congestion. Organ recovery is performed in a standardized fashion (Figure 1) (10).

Figure 1.

King's College Hospital donor recovery procedure for DCDD. After withdrawal, criteria for not proceeding with the recover are long-standing low saturation (prolonged hypoperfusion) and no arrest (>1 hour to reach asystole). Warm ischemia time (WIT) is considered the time between low blood pressure (<50 mmHg) and aortic perfusion.

Data collection and stratification

We analyzed DCDD and DBD donor offers from January 2001 to December 2010. Data from 2001 until 2009 were used for the retrospective analysis. Data from 2010 were prospectively obtained to proceed with the validation of results previously obtained. The yearly rate of DBD in the United Kingdom has been provided for baseline comparisons (UKTransplant). All donor offers were collected on a proforma and retrospectively reviewed. From 2007, details of the presence or absence of gag and cough reflexes were added in the light of clinical experience. DCDD and DBD offers were either accepted or declined. In the DCDD group, accepted offers were further classified into those patients who underwent cardiac arrest and liver recovery and those who did not, either because the mode of death was prolonged or because it did not occur. Graft viability was assessed by clinical and histological criteria and a decision regarding usage was made by the responsible surgeon. Thus, a recovered graft could be declared as not usable for transplant, leading to a further classification of grafts recovered from donors who died within the defined time frame which were used or not used.

Statistical analysis

All the analysis was performed using the Statistical Package for the Social Sciences (SPSS) release 11.0 for Windows (SPSS, Inc., Chicago, IL, USA). Data from 2001–2009 were used for the baseline comparisons, retrospective comparative analysis and multivariate regression models. Data from 2010 were prospectively collected according to the findings in the previous models, so they are not included in the previous calculations.

Baseline comparisons are as follows:

  • – Descriptive analyses of qualitative and quantitative variables were expressed as number and percentage, and as mean, maximum, minimum and standard deviation, respectively. Basal intergroup comparisons were performed by chi-square, analysis of variance (ANOVA), and two-tailed nonpaired T-tests for qualitative, and multiple and dichotomic mean comparisons, respectively.
  • – Retrospective analysis: To predict the probability of a potential donor to arrest and of graft usability for liver transplant, logistic regression models were performed. All the variables with a p<0.1 in the univariate analysis or those biologically important of potential confounding factors were included step by step until a final model was obtained with variables achieving a p value <0.05 and assuming risk and protective factors as those over or below 1, respectively (including their highest and lowest confidence intervals).
  • – Prospective validation: In order to confirm the accuracy of the previous donor arrest and graft use models, we used a prospectively collected database form 2010. We constructed a new variable for each accepted donor in 2010 according to the following formulas from the previous regression models:
    • ○ Probability of donor arrest = e2.4 + 1.6*(0 = no inotropes; 1 = yesinotropes) – 1.27*(0 = <40 years; 1 = >40 years) – 1.92*(0 = no gag/cough; 1 =gag/cough present)
    • ○ Probability of graft refuse = e −5.8 + 1,03*(0 = age<50; 1 = age>50) +1,8*(0 = BMI<30; 1 = BMI<30) + 1,26*(0 = ITU<7d; 1 = ITU>7d) + 1,159*(0 = ALT<4x;1 = ALT≥4x) +1.72*(0 =WIT<25; 1 = WIT>25)
    • ○ After the creation of the previous two variables, receiver-operating characteristic (ROC) curves were performed with donor arrest and graft use as primary endpoints. The area under the ROC curve (AUROC) was considered a useful predictor above 0.7.
    • ○ To obtain the best sensitivity and specificity cut-off values, their own “B” statistic value from the regression model was applied to the 3 and 5 variables in both cardiac arrest and graft use models, considering negative or “zero” values when the variables were present but negative or absent, respectively. The overall score was introduced in the ROC curve, choosing the cut-off values that showed the highest sensitivity with the lowest “1–Specificity” values. A p value<0.05 was assumed for statistical significance.


Baseline results

In the last decade, 4632 donation offers have been received in our Unit. Of these, 3037 were from DBD and 1595 were from DCDD, 1579 of which were from category-III DCDD. From 2001 to 2009, the Unit received 966 DCDD offers. Of these, 420 (44%) were accepted for further evaluation and 546 (56%) declined, respectively. Of the accepted offers, 300 (72.9%) patients had cardiac arrest and their livers were recovered and assessed for graft viability. Of these, 132 grafts (49.2%) were used for adult and pediatric liver transplantation (Figure 2).

Figure 2.

Schematic overview of the population included in the study, stratified in the retrospective (2001–2009) and prospective (2010) periods. TO = total offers; DBD = donor after brain death; DCDD = donor after circulatory determination of death; AO = accepted offer; DO = declined offer; ARR = cardiac arrest; NoARR = no cardiac arrest.

Trends in DCDD donation (Figure 3)

Figure 3.

King's College Hospital annual evolution since the start of the DCDD programme (2001–2010) of total donation offers and DCDD accepted, declined offers and used grafts.

The number of total offers (TO) has progressively increased since the introduction of the DCDD program. In the last 4 years, the number of donor offers has increased by more than 100% (951 in 2010 vs. 429 in 2007). The overall number of accepted total offers has increased by 80% (472 in 2010 vs. 262 in 2007). The number of DCDD offered is 55 and 4 times higher in 2010 than in 2001 and 2007, respectively (11 in 2001; 147 in 2007 and 613 in 2010).

Declined and accepted offers

Global comparisons for the whole series and for accepted and declined offers are summarized in Table 1. Baseline statistically significant differences were observed in clinical (age, ITU stay and cause of death) and biochemical parameters (GGT, ALT and Bilirubin) in the initial acceptance and declining setting. Of note, body mass index (BMI) was not an initial parameter for donor acceptance.

Table 1.  Baseline distribution of the variables studied in the global series (2001–2009) and comparisons between declined and accepted offers
 Global dataDeclined offerAccepted offerp
  1. Results are expressed as % for qualitative variables and mean (standard deviation) for quantitative variables. Two-tailed t-test and chi-square tests were used for quantitative or qualitative comparisons. BMI = body mass index; ITU = intensive therapy unit; GGT = gamma-glutamyl-transferase; ALT = alanine-transaminotransferase.

Age 48 (18) 51 (18) 45 (16) 0.001
Blood GR   
BMI 26.6 (5.78) 26.9 (10.74) 26.33 (5.22)
ITU 5 (9) 6 (12) 4 (4) 0.001
 0–2 d44.740.249.60.017
 2–5 d28.429.826.9
 >5 d26.93023.50.03
Cause of death    
Sodium 144 (9) 142 (21) 144 (9)
Potasium 4.1 (0.7) 4.7 (0.6) 4.1 (0.7)
Creatinine 91 (64) 142 (4) 91 (64)
GGT 138 (207) 193 (248) 87 (143) 0.001
ALT 134 (427) 182 (566) 77 (116) 0.002
Bilirubin 16 (38) 17 (21) 15 (50)
Reasons for decline    
 Age 13.3  
 Donor characteristics 17.1  
 Donor may not demise 5.9  
 Prolonged arrest/ITU 6.5  
 Tumor history 8.9  
 History drug user 2.1  
 History alcohol intake 6.6  
 Medical history 15  
 Logistics 4.6  
 No suitable recipient 5.1  
 Liver function tests 3.8  
 Others 11.2  

Arrest versus no arrest

Almost three out of four accepted donors experienced cardiac arrest (73%) after withdrawal of support. Those donors who arrested were significantly younger and had higher serum sodium and creatinine levels. The presence of gag or cough reflexes was more frequent in the patients who did not have cardiac arrest. In contrast, patients who had cardiac arrest were more likely to be supported with inotropes (Table 2).

Table 2.  Baseline (2001–2009) comparisons between donors who had cardiac arrest after withdrawal and those who did not
 No arrestArrestp
  1. Statistically significant (p < 0.05) differences (chi-square or t-test) are displayed as for the favorable comparing group (NAR-no arrest; ARR-arrest). Results are expressed as% for qualitative variables and mean (standard deviation) for quantitative variables. BMI = body mass index; ITU = intensive therapy unit; GGT = gamma-glutamyl-transferase; ALT = alanine-transaminotransferase; DONOR ACC = total number of donors accepted from a hospital; DONOR EXP = donor offer order for every hospital. FiO2 = fraction of inspired oxygen; PEEP (cmH2O) = positive end-expiratory pressure.

Age 51 (13) 43 (17) 0.001
Blood GR   
BMI 26.2 (5.06) 26.38 (5.3)
ITU 4 (4) 4 (4)
 0–7 d86.786
 >7 d13.314
Cause of death   
Sodium 141 (7) 146 (9) 0.001
Potasium 4.1 (0.5) 4.1 (0.7)
Creatinine 76 (36) 97 (71) 0.001
GGT 76 (104) 91 (155)
ALT 64 (93) 79 (120)
Bilirubin 11 (6) 16 (58)
FiO2 (%)44 (19)49 (22)
PEEP (cmH2O)5.5 (1.6)6.2 (2.7) 
pH7.5 (0.2)7.3 (0.5)
pO2 (kPa)18 (12)18 (14)
pCO2 (kPa)5.5 (3.6)6.1 (5.3)
HCO3 (mEq/L)24.8 (3.6)23.7 (4.1)

Used and not-used grafts

Graft usability (as expected) had a different distribution of clinical (donor age, BMI) and biochemical (potassium, creatinine, GGT and ALT) parameters. There was a higher proportion of grafts used when the donor offer was more than the 10th from that particular donor hospital (donor offering experience seems to be an important graft usability factor). For our surgical team, WIT was a major concern for graft selection, and thus, significant differences can be observed between groups (Table 3).

Table 3.  Baseline (2001–2009) comparisons used and not used grafts
 Not usedUsedp
  1. Statistically significant (p<0·05) differences (chi-square or T-test) are displayed as for the favorable comparing group (NOU—not used graft; USE—used graft). Results are expressed as % for qualitative variables and mean (standard deviation) for quantitative variables. BMI = body mass index; ITU = intensive therapy unit; GGT = gamma-glutamyl-transferase; ALT = alanine-transaminotransferase; DONOR ACC = total number of donors accepted from a hospital; DONOR EXP = donor offer order for every hospital; WIT = warm ischemia time.

Age 46 (16) 38 (16) 0.001
Blood GR   
BMI 27.89 (5.86) 24.75 (4.05) 0.001
ITU 4 (5) 4 (3)
 0–7 days82.589.9
 >7 days17.510.1
Cause of death   
Sodium 145 (9) 147 (9)
Potasium 4.3 (0.8) 4 (0.7) 0.003
Creatinine 110 (91) 83 (37) 0.002
GGT 120 (202) 64 (82) 0.023
ALT 93 (125) 70 (121)
Bilirubin 21 (83) 12 (8)
WIT 20 (8) 16 (5) 0.001
Reasons for not use   
 Prolonged WIT14  
 Fatty graft40.1  
 Other graft impairments23.6  
 Incidental tumors4.5  
 Other donor characteristics17.9  

Prediction of donor cardiac arrest and graft usability

All variables initially noted by our co-ordinators were included in multiple conditional forward logistic binary regression models in order to get predictive models of donor arrest and graft usability (Table 4). Donor arrest was predicted by age, donor sodium levels and inotrope status. By 2007, it was apparent that the presence of cough and gag reflexes was an important factor in determining likelihood of death and liver donation and were included in the donor data form. A regression model for the last two years has been performed. Donor age and inotropes maintained their influence, but the presence of gag or cough reflexes emerged as important protective factors for cardiac arrest. Regarding graft usability, a clear and stable model involving five variables (donor age and BMI, ITU stay, WIT and ALT levels) was obtained and appeared to be a good predictor model. Figure 4 graphically depicts the donor arrest and graft usability probabilities according to the adding of risk and protective factors and their specific relative risk weight according to confidence interval levels.

Table 4.  Regression models for prediction of graft usability and donor arrest. A subanalysis of donor arrest prediction model is performed for the years 2008–2009 including gag/cough reflexes
Prediction of donor arrest (global model) Accepted offers: arrest (300) vs. no arrest (117)
 BpRRConfidence intervals
Age > 45 years−0.6320.0270.5310.3030.932
Use of inotropes1.7670.0015.8513.20510.679
BMI > 290.6680.0441.9501.0183.734
Serum Na > 1450.9510.0032.5871.3934.807
Prediction of donor arrest (partial model: 2008–2009) Accepted offers: arrest (115) vs. no arrest (63)
 BpRRConfidence intervals
Age > 40 years1.2880.00283.6261.15111.424
Use of inotropes−1.6430.0010.1930.0720.520
Gag/cough reflex1.8660.0016.4632.16819.267
Prediction of graft usability Accepted offers: used (142) vs. not used (158)
 BpRRConfidence intervals
  1. In brackets () is the inverse arresting prediction model to find out the relative power of inotropes use.

Age > 50 years1.2290.0013.4191.6367.148
BMI > 291.8270.0016.2182.58614.954
WIT > 24 minutes1.8580.0036.4131.91521.477
ITU stay > 7 days1.4140.0234.1121.21213.954
ALT ≥ 4×1.2110.0453.3581.02610.989
Figure 4.

Global adding risk increments according to the accumulation of risk and protective factors from the regression models of graft usability and donor arrest. The relative risk ratios and highest and lowest values of confidence intervals were added and calculated according to the accumulation in the same donor of protective or risk factors. Thus, in the cardiac arrest model, a potential donor could have a combination of risk factors for not arresting (gag/cough + and/or age>40 years) and a protective factor for not arresting (to be on inotropes). In the graft usability model, all five variables (Age > 50 years; BMI > 29; WIT > 24 minutes; ITU stay > 7 days; ALT ≥ 4×) are risk factors for not using the graft, and their addition in the same donor increases this risk. Because all the factors have different risk strengths, three main areas are displayed: blue (highest risk area with the highest upper limits of the RR confidence intervals); red (intermediate risk area with the higher and lower possible values of the RR combinations); and green (lowest risk area with the lowest low limits of the RR confidence intervals). In the arrest prediction model, an inverted calculation for estimating the power of inotropes use (risk factor for cardiac arrest) was performed (coming from Table 4). Thus, its specific weight is considered as negative risk increments for no arrest prediction (“protective factor for no arrest”).

Prospective validation of donor arrest and graft usability models

Both previous models (cardiac arrest [2008–2009]; and graft usability [2001–2009]) had excellent AUROC when applied to the prospective data collection in 2010 (Figure 5). Considering the “B” value from the regression model, a value of 1.42 had a sensitivity of 86% and specificity of 80% to predict that the donor will not experience cardiac arrest. In the graft usability model, a score of 1.66 had a sensitivity of 69% and a specificity of 72% to predict that the graft may not be used for liver transplantation.

Figure 5.

Prospective 2010 area under the receiver-operating curves (AUROC) from the retrospective models developed for the prediction of a potential donor cardiac arrest (above)and graft usability (below) models, showing excellent 0.835 and 0.748 values, respectively. CI = confidence intervals.


The development of DCDD liver transplantation has led to significant changes in clinical practice. Organ shortage has driven this resurgence in the countries with appropriate legislation (14). According to UK Transplant data, DCDD (primarily for kidney transplantation) increased the overall donor pool in the United Kingdom by 5.6% in 2001 (42 DCDD and 744 DBDs). Currently, DCDD account for 15% of transplanted livers (564 DBD vs. 99 DCDD) and 36% of kidneys (952 DBD vs. 530 DCDD). However, the donor rate was 13.1 per million of population (pmp) in 2001 and 15.5 pmp in 2010 [ September 2010]. Similar trends have been reported in countries with different cultures of donation such as Korea (DCDD increased from 0.7% in 2006 to 4.2% in 2009 with organ donation rates of 2.9 and 5.3 pmp, respectively) (15) and the United States (DCDD account for 10.6% of donation). However, graft utilization has declined progressively with the use of DCDD (currently at 68% in comparison to 90.3% of DBD). The number of livers recovered at DCDD but not used has increased due to concerns over organ quality and subsequent graft dysfunction, cholangiopathy and DCDD long-term outcomes (16).

Analysis of grafts used and accepted DCDD offers shows clear trends over the past 10 years: The utilized graft/total DCDD offer ratio has dropped from 0.45 in 2001 to 0.07 in 2010; in contrast to the DBD population which has remained stable at 0.5–0.6 over the last decade (Figure 6A). The utilized graft/accepted DCDD offer ratio reflects a similar trend, from 0.45 in 2001 to 0.20 in 2010, with little change in the DBD group of 0.8 (Figure 6B). An overview of our data shows that 12.2% of all DCDD offers have turned into usable grafts over the study period. Almost 35% of accepted donors did not experience cardiac arrest and death after treatment withdrawal and more than 50% of those who did have a cardiac arrest did not have a usable graft. Thus, the effectiveness and efficiency of DCDD seems to fall short of what was expected. Prediction models of donor arrest and graft usability would help to maximize the yield of DCDD and reduce the costs of recovery.

Figure 6.

Trends in the last decade in the used graft/total offers (6A) and used graft/accepted offers (6B) ratios, both in DCDD and DBD groups.

Initial offers, which can be declined at the outset, include patients with cirrhosis, severe sepsis, liver ischemia and those with active cancer. It was not possible to report differences in trends for the initial offer acceptance because the policy has not been restrictive. Some countries (United Kingdom is an example) have tried to achieve a full-potential-donor offer policy; however, this requires resource and increased demands on surgical teams and coordinators.

Few publications have reported tools that can be used to predict if a patient is a suitable candidate for DCDD (17). The most recent one on risk predictors for death 60 minutes after treatment withdrawal was reported by DeVita et al. (18,19). They identified risk factors, including Glasgow coma scale = 3, SaO2/FiO2 ratio <230, peak inspiratory pressure (PIP) ≥35 mmHg, respiratory rate off the ventilator <8/min, PaO2 <72 mmHg, inotropes ≥0.2 μg/kg/min, all treatments withdrawn within 10 minutes and extubation withdrawal. Our model has the advantage that it was performed on 982 donor offers, of which 432 were accepted and we were able to assess the likelihood of donor arrest and liver graft utilization. In contrast, DeVita et al. included 533 patients who were not organ donation candidates.

Combining the arrest and usability prediction models, we were able to identify two groups. The first group were donors under 40 years old, with no gag or cough reflexes, on inotropes, with a BMI under 30, less than 7 days on ICU and with ALT of less than 3 times normal values which will almost certainly lead to a usable graft. The second is a donor over 50 years old, with gag or cough reflexes and hemodynamically stable with BMI higher than 30 and prolonged ICU stay, with or without high ALT levels which may be declined at outset because of low probability of graft use. In the middle, there is a category with intermediate cardiac arrest and usability probabilities where utilization may vary depending on the risk of the potential recipient of dying on the waiting list. The results from the prospective validation, suggesting that in the cardiac arrest model, the presence of age >40 years and/or gag/cough reflexes without inotropes had excellent sensitivity and specificity. The same happened in the graft usability model, where BMI > 30 or WIT > 25 minutes, or the combination of two of the other three variables, reached sensitivity and specificity values around 70%. This prospective analysis has given our study a significant strength, for three reasons: the avoidance of an arbitrary assignation of a risk value to each variable, considering the previous “B” score of the regression, making the statistical analysis stronger; secondly, results are consistent with the previous findings and their AUROC, sensitivity and specificity are promising; and third, using a validation set from a short period of time (year 2010) with such a marked increase in organ offers and utilization in a short period of time minimizes biases or confounding factors. Obviously, statistical models cannot predict 100% of cases. Moreover, progression to cardiac arrest is a hardly process dictated by natural history and by management of withdrawal care. However, we could get high sensitivity and specificity values with easy-to-remember and easily reproducible variables, what can make our models clinically useful. Although liver transplant teams around the world may have their own ‘center-criteria’, after our analysis, a bit more of evidence highlights the DCDD environment, and could turn their criteria into “science-guided criteria”.

The cost of DCDD liver transplantation is approximately 120–150% of DBD depending on the postoperative complications (13). This excludes the donor recovery cost, which is labor intensive. The recovery team is composed of two surgeons, one scrub nurse and a perfusionist/technician and all the kit to perform an organ recovery. In addition, the recovery team attends to an increased number of non-productive attempts, which may have implications for staff motivation and “burn-out”. The emphasis of DCDD is one of turning down donors or grafts rather than accepting, which is in contrast to DBD where it is one of looking to accept. This change of emphasis may be damaging to transplantation in the longer term.

DCDD practice has changed dramatically over the last decade with a significant increase in the number of DCDD offers. The number of accepted DCDD offers has increased 10-fold over the past 10 years and reflects increased work and expense. Is the number of grafts worth the effort? Should more emphasis be placed on ensuring that DBD donation takes place when feasible? It appears that in the United Kingdom the pool of DCDD has increased and that DBD offers have decreased by approximately the same proportion. It could be argued that DCDD has prevented a greater shortfall in organ supply, however, it does appear to be seen as more convenient and does appear to be replacing DBD. The reasons for favoring DCDD over DBD include the faster time to donation, shorter ITU stay, and avoid waiting for drug effects to wean off, and more “efficient” donor management with no requirement for brain stem death testing. In addition, the approach to families with a more easily understood donation process may be a factor.

DCDD livers should currently be considered as extended criteria grafts, and continue to be an important factor in donor risk indexes (20). Although many centers are reporting improving outcomes with DCDD grafts (10), overall they appear to have lower graft survival (21,22) and increased risk of posttransplant cholangiopathy (23,24) than DBD liver transplantation. According to our reported (Valente R., et al. ESOT Congress; September 2011) experience, 3-months, 1- and 5-years patient and graft actuarial survival are 95.4%, 88.3%, 78.3% and 90.8%, 82.4%, 73.3%, respectively. Main complications have been 7 (3.6%) primary non-functions, 6 (3.4%) arterial thromboses, 17 (9.7%) biliary complications, of which 5 (2.7%) were ischemic cholangiopathies (25). Informed consent of the recipient for use of a DCDD graft (26,27) and the potential increased risk of graft loss and cholangiopathy should be centre specific. The use of other marginal livers such as steatotic (28) or old (29) grafts is well established and contribute significantly to the size of the donor pool. DCDD were expected to do the same (30); however, it is still not clear whether this is true or whether DBD multiorgan donor are continuing to be lost.

Liver transplantation is an effective therapeutic option and graft shortage continues to be the major limitation to its application. DCDD may have helped to cover the shortfall in DBD liver transplantation. Much of the increased cost of DCDD liver transplantation from graft complications currently remains hidden. Our study is the first analysis of DCDD activity and donor offers with the development of donor arrest and graft usability models from “controlled donation”. Concerns regarding the increasing number of DCDD offers, the proportion of donors who do not have cardiac arrest, the number of unused grafts and the trend change from DBD to DCDD require further study.


The authors of this manuscript have no conflicts of interest to disclose as described by the American Journal of Transplantation.