Minimizing risk associated with elderly liver donors by matching to preferred recipients

Authors

  • Dorry L. Segev,

    Corresponding author
    1. Department of Surgery, The Johns Hopkins University School of Medicine, Baltimore, MD
    • Transplant Surgery, Johns Hopkins Medical Institutions, 600 N. Wolfe Street, Harvey 611, Baltimore, MD 21287
    Search for more papers by this author
    • Dorry Segev had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. The content of this article is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the United States Government.

    • fax: 410-614-2079

  • Warren R. Maley,

    1. Department of Surgery, The Johns Hopkins University School of Medicine, Baltimore, MD
    Search for more papers by this author
  • Christopher E. Simpkins,

    1. Department of Surgery, The Johns Hopkins University School of Medicine, Baltimore, MD
    Search for more papers by this author
  • Jayme E. Locke,

    1. Department of Surgery, The Johns Hopkins University School of Medicine, Baltimore, MD
    Search for more papers by this author
  • Geoffrey C. Nguyen,

    1. Department of Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD
    Search for more papers by this author
  • Robert A. Montgomery,

    1. Department of Surgery, The Johns Hopkins University School of Medicine, Baltimore, MD
    Search for more papers by this author
  • Paul J. Thuluvath

    1. Department of Medicine, The Johns Hopkins University School of Medicine, Baltimore, MD
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

Abstract

Elderly liver donors (ELDs) represent a possible expansion of the donor pool, although there is great reluctance to use ELDs because of reports that increasing donor age predicts graft loss and patient death. The goal of this study was to identify a subgroup of recipients who would be least affected by increased donor age and thus best suited to receive grafts from ELDs. A national registry of deceased donor liver transplants from 2002–2005 was analyzed. ELDs aged 70–92 (n = 1043) were compared with average liver donors (ALDs) aged 18–69 (n = 15,878) and ideal liver donors (ILDs) aged 18–39 (n = 6842). Recipient factors that modified the effect of donor age on outcomes were identified via interaction term analysis. Outcomes in recipient subgroups were compared using Kaplan-Meier survival analysis. Recipients preferred for ELD transplants were determined to be first-time recipients over the age of 45 with body mass index <35, non–status 1 registration, cold ischemic time <8 hours, and either hepatocellular carcinoma or an indication for transplantation other than hepatitis C. In preferred recipients, there were no differences in outcomes when ELD livers were used (3-year graft survival: ELD 75%, ALD 75%, ILD 77%, P > 0.1; 3-year patient survival: ELD 81%, ALD 80%, ILD 81%, P > 0.1). In contrast, there were significantly worse outcomes when ELD livers were used in nonpreferred recipients (3-year graft survival: ELD 50%, ALD 71%, ILD 75%, P < 0.001; 3-year patient survival: ELD 64%, ALD 77%, ILD 80%, P < 0.001). Conclusion: The risks of ELDs can be substantially minimized by appropriate recipient selection. (HEPATOLOGY 2007.)

Last year, over 10,000 new patients were registered for liver transplantation (LT), yet fewer than 6000 deceased donor LTs were performed. A similar discrepancy occurs yearly and has led to a list of over 17,000 patients who currently await LT.1 The response from the transplant community has been a continued expansion of the criteria for an acceptable liver donor, balancing the increased risks of expanded criteria donors against the risk of mortality while awaiting transplantation.2, 3

One of the most significantly expanded donor criteria has been donor age, likely reflecting continued clinical intuition that the liver may be less sensitive to senescence.2 Between 1994 and 2004, there was a 36% increase in the use of liver donors under the age of 50, while the use of those over 50 nearly tripled during the same period. Over half of the expansion of the donor pool during this period involved the use of donors over the age of 50.4 Further attempts to expand the donor pool have included using elderly liver donors (ELDs) over the age of 70, although these donors are much more likely to be discarded5, 6 and therefore comprise only 6% of liver transplants performed.4

Reluctance to use ELDs likely stems from reports that the risk of graft loss and death increases with donor age,2, 7–11 with as high as 65% increased risk associated with ELDs.12 These reports include studies from Europe as well as studies from the United States prior to the initiation of a new system of disease severity–based allocation using the model for end-stage liver disease (MELD) score.13–15

However, several transplant centers have reported excellent results with ELDs, in some cases equivalent to those from younger donors.16–22 We hypothesized that recipient selection practices might contribute to the improved outcomes experienced in these centers. Evidence that patients with hepatitis C are differentially affected by older donor age17, 23–26 supported this hypothesis and led us to perform a comprehensive evaluation of all potential effect modifiers in recipients of ELD allografts.

In this study, we analyzed a prospective, national cohort of liver transplant recipients to characterize ELDs and the recipients to whom they were allocated in the current MELD system, to evaluate graft and patient survival of ELD recipients, to elucidate effect modifiers associated with transplantation using ELDs, and to identify a subgroup of recipients who would be least affected by increased donor age and thus best suited to receive grafts from ELDs.

Abbreviations

ALD, average liver donor; BMI, body mass index; CIT, cold ischemic time; ELD, elderly liver donor; HCC, hepatocellular carcinoma; HR, hazard ratio; ILD, ideal liver donor; LT, liver transplantation; MELD, model for end-stage liver disease; UNOS, United Network for Organ Sharing.

Patients and Methods

Study Design and Population.

This was a secondary data analysis of a prospective cohort of adult deceased donor liver transplant recipients. This study evaluated the association between donor age and 2 recipient outcome measures: graft survival and patient survival. The study population consisted of 22,817 patients available for analysis in the United Network for Organ Sharing (UNOS) Standard Transplant Analysis and Research files who underwent LT between the onset of the MELD allocation system, February 27, 2002, and December 31, 2005. Follow-up was available through June 6, 2006.4 Excluded from analysis were pediatric recipients (age <18; n = 2169), live donor recipients (n = 1019), multiorgan recipients (n = 1131), donors missing information about age (n = 4), and pediatric donors (age <18; n = 1573). For comparisons of donor and recipient characteristics between ELDs and younger donors, unpaired 2-tailed t tests were used for continuous covariates, and χ2 tests of independence were used for categorical covariates.

Regression Modeling.

The Cox regression models for risk of graft loss and patient death were designed in the following manner. First, an unadjusted analysis was performed of the following biologically relevant donor, recipient, and transplant variables from the Standard Transplant Analysis and Research file: (1) donor age, body mass index (BMI), sex, ethnicity, cause of death, diabetes, hypertension, blood type, use of inotropes, blood urea nitrogen, creatinine, donor type; (2) recipient age, BMI, sex, ethnicity, diagnosis, MELD score (final calculated laboratory score at the time of transplantation), MELD priority exception, status 1 registration, hospitalized, on life support, albumin, ascites, encephalopathy, angina, diabetes, hypertension, history of bacterial peritonitis, history of portal vein thrombosis, prior abdominal surgery, prior liver transplant, prior transjugular intrahepatic portosystemic shunt, blood type, insurance; and (3) transplant cold ischemic time (CIT). The appropriate functional form of model covariates was determined via exploratory data analysis in unadjusted models. Absence of colinearity among the covariates was confirmed by testing variance inflation factors.

Patient diagnosis was categorized for the purposes of the regression model predicting outcomes after liver transplantation. Many diagnoses were infrequent (potentially leading to model instability), difficult to categorize (potentially leading to misclassification bias), unreported, or noncontributory to the outcome model on exploratory data analysis (potentially leading to overfitting) and as such were categorized as “other” (see Appendix). Hepatocellular carcinoma (HCC) was considered a separate category given the potential difference in outcomes specific to this malignancy. For consistency, this stratification was used for all analyses, including patient characteristics (Table 1), regression models (Table 2), and interaction term analyses (Table 3).

Table 1. Characteristics of 16,921 Deceased-Donor Liver Transplants Performed in the United States in the MELD Era, by Donor Age
 Donor Age <70Donor Age ≥70P ValueAge 70–74Age 75–79Age 80–92
  1. For each variable, the mean (standard error) are shown. P values represent the comparison between donor age ≥70 and donor age <70.

n15,878 (93.8%)1043 (6.2%) 569348126
Donor characteristics      
Age41.5 (0.1)74.8 (0.1)<0.00171.976.882.2
BMI26.6 (0.0)26.0 (0.2)<0.00126.325.924.5
Female (%)40.9 (0.4)54.3 (1.5)<0.00149.758.961.9
Ethnicity (%):      
 White70.4 (0.4)81.1 (1.2)<0.00179.284.580.2
 African American14.1 (0.3)9.4 (0.9) 9.78.011.9
 Hispanic12.2 (0.3)6.3 (0.8) 7.05.26.3
 Other3.3 (0.1)3.2 (0.5) 4.02.31.6
Cause of death (%)      
 Anoxia12.0 (0.3)5.5 (0.7)<0.0016.83.84.1
 CVA46.9 (0.4)82.4 (1.2) 81.184.881.3
 Head trauma40.2 (0.4)12.0 (1.0) 11.911.114.6
Diabetes (%)8.3 (0.2)18.9 (1.2)<0.00120.617.215.9
Hypertension (%)30.4 (0.4)64.8 (1.5)<0.00165.063.767.2
Share type (%)      
 Local71.6 (0.4)52.7 (1.5)<0.00156.150.344.4
 Regional22.3 (0.3)31.3 (1.4) 28.532.540.5
 National6.1 (0.2)16.0 (1.1) 15.517.215.1
Recipient characteristics      
Age51.8 (0.1)55.4 (0.3)<0.00155.354.658.5
BMI28.0 (0.0)28.1 (0.2)0.728.228.028.0
Female (%)31.0 (0.4)35.7 (1.5)0.00234.435.641.3
Ethnicity (%)      
 White74.2 (0.3)74.5 (1.4)0.00173.574.479.4
 African American8.8 (0.2)5.8 (0.7) 6.75.71.6
 Hispanic11.6 (0.3)13.9 (1.1) 13.714.712.7
 Other5.3 (0.2)5.8 (0.7) 6.25.26.3
Diagnosis (%)      
 Hepatitis C35.5 (0.4)26.0 (1.4)<0.00128.124.121.4
 Alcoholic13.0 (0.3)15.4 (1.1) 15.117.211.9
 Primary biliary cirrhosis3.4 (0.1)3.9 (0.6) 3.54.05.6
 Primary sclerosing cholangitis5.2 (0.2)6.1 (0.7) 6.56.63.2
 Hepatocellular carcinoma13.6 (0.3)11.9 (1.0) 11.611.813.5
 Other29.3 (0.4)36.6 (1.5) 35.136.244.4
MELD (calculated)20.4 (0.1)18.7 (0.3)<0.00119.318.317.2
MELD exception (%)26.2 (0.3)24.8 (1.3)0.325.722.128.6
Status 1 (%)6.6 (0.2)5.2 (0.7)0.15.35.73.2
Hospitalized (%)28.0 (0.4)25.7 (1.4)0.126.325.623.0
On life support (%)7.0 (0.2)6.4 (0.8)0.47.85.52.4
Diabetes (%)19.5 (0.3)25.4 (1.4)<0.00126.523.126.4
Hypertension (%)17.7 (0.3)21.7 (1.3)0.00221.422.021.8
Peritonitis (%)7.1 (0.2)5.3 (0.7)0.036.05.12.5
Prior transplant (%)7.9 (0.2)3.5 (0.6)<0.0014.03.22.4
Prior abdominal surgery (%)46.9 (0.4)47.8 (1.6)0.648.845.749.2
Table 2. Risk Factors for Graft Loss and Patient Death After Deceased Donor Liver Transplantation
 Graft LossPatient Death
Univariate HR (95% CI)P ValueMultivariate HR (95% CI)P ValueUnivariate HR (95% CI)P ValueMultivariate HR (95% CI)P Value
  • For each covariate, hazard ratio (95% confidence interval) are shown.

  • Abbreviations: CVA, cerebrovascular accident; HCC, hepatocellular carcinoma; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis.

  • *

    Donor characteristics in the multivariate model also included blood type, inotropes, blood urea nitrogen, creatinine, and donor type.

  • Recipient characteristics in the multivariate model also included blood type, ascites, encephalopathy, albumin, angina, portal vein thrombosis, prior transjugular intrahepatic portosystemic shunt, and insurance.

Donor*        
Age        
 <40Reference Reference Reference Reference 
 40–491.17 (1.07–1.28)<0.0011.14 (1.01–1.28)0.0331.14 (1.04–1.25)0.0051.13 (1.00–1.28)0.044
 50–591.38 (1.26–1.51)<0.0011.39 (1.23–1.57)<0.0011.23 (1.12–1.36)<0.0011.28 (1.13–1.46)<0.001
 60–691.60 (1.45–1.77)<0.0011.56 (1.35–1.80)<0.0011.45 (1.30–1.61)<0.0011.44 (1.24–1.67)<0.001
 ≥701.78 (1.57–2.01)<0.0011.83 (1.55–2.17)<0.0011.60 (1.40–1.82)<0.0011.64 (1.37–1.97)<0.001
BMI ≥351.04 (0.92–1.17)0.51.03 (0.89–1.19)0.71.07 (0.95–1.22)0.31.10 (0.94–1.28)0.2
Female1.11 (1.04–1.19)0.0011.08 (0.99–1.18)0.11.06 (0.99–1.14)0.11.03 (0.94–1.13)0.5
Ethnicity        
 WhiteReference Reference Reference Reference 
 African American1.14 (1.04–1.25)0.0061.19 (1.06–1.33)0.0031.05 (0.95–1.16)0.41.08 (0.95–1.22)0.2
 Hispanic1.19 (1.08–1.32)<0.0011.22 (1.08–1.38)0.0011.22 (1.10–1.35)<0.0011.15 (1.01–1.31)0.035
 Other1.26 (1.07–1.49)0.0071.15 (0.93–1.42)0.21.32 (1.11–1.57)0.0021.26 (1.02–1.57)0.036
Cause of death        
 AnoxiaReference Reference Reference Reference 
 CVA1.27 (1.14–1.42)<0.0011.27 (1.11–1.47)<0.0011.18 (1.05–1.32)0.0051.21 (1.04–1.40)0.013
 Head trauma0.91 (0.81–1.02)0.11.10 (0.95–1.28)0.20.89 (0.79–1.00)0.0481.04 (0.89–1.21)0.6
Diabetes1.31 (1.18–1.46)<0.0011.16 (1.01–1.32)0.0311.28 (1.15–1.43)<0.0011.18 (1.02–1.36)0.027
Hypertension1.28 (1.20–1.37)<0.0010.95 (0.86–1.05)0.31.19 (1.11–1.28)<0.0010.93 (0.84–1.04)0.2
Recipient        
Age1.00 (1.00–1.01)0.21.01 (1.00–1.01)0.0121.01 (1.01–1.02)<0.0011.02 (1.01–1.02)<0.001
BMI ≥351.11 (1.01–1.23)0.0311.13 (1.00–1.27)0.0451.15 (1.03–1.27)0.0111.17 (1.03–1.33)0.014
Female1.06 (0.99–1.14)0.11.01 (0.93–1.11)0.81.07 (1.00–1.15)0.11.07 (0.97–1.17)0.2
Ethnicity        
 WhiteReference Reference Reference Reference 
 African American1.39 (1.25–1.54)<0.0011.38 (1.21–1.58)<0.0011.29 (1.15–1.44)<0.0011.28 (1.11–1.48)<0.001
 Hispanic0.91 (0.82–1.01)0.10.83 (0.72–0.94)0.0050.93 (0.84–1.04)0.20.83 (0.72–0.96)0.01
 Other0.92 (0.80–1.07)0.30.84 (0.69–1.02)0.10.98 (0.84–1.14)0.80.93 (0.76–1.14)0.5
Diagnosis        
 Hepatitis CReference Reference Reference Reference 
 Alcoholic0.84 (0.75–0.93)0.0010.84 (0.74–0.96)0.0110.82 (0.73–0.92)<0.0010.79 (0.68–0.91)<0.001
 PBC0.71 (0.58–0.87)<0.0010.73 (0.58–0.94)0.0140.68 (0.55–0.85)<0.0010.67 (0.52–0.87)0.003
 PSC0.80 (0.68–0.94)0.0070.78 (0.65–0.95)0.0140.61 (0.50–0.73)<0.0010.63 (0.50–0.79)<0.001
 HCC0.94 (0.84–1.04)0.21.02 (0.88–1.17)0.80.94 (0.85–1.05)0.30.94 (0.82–1.09)0.4
 Other1.03 (0.95–1.11)0.50.84 (0.76–0.93)<0.0010.93 (0.85–1.01)0.10.73 (0.65–0.81)<0.001
MELD (calculated)1.02 (1.02–1.03)<0.0011.01 (1.00–1.02)<0.0011.02 (1.02–1.03)<0.0011.02 (1.01–1.02)<0.001
MELD exception0.86 (0.80–0.93)<0.0011.09 (0.97–1.22)0.10.92 (0.85–1.00)0.0391.21 (1.07–1.37)0.002
Status 11.72 (1.54–1.92)<0.0010.94 (0.78–1.14)0.51.64 (1.46–1.85)<0.0010.87 (0.71–1.07)0.2
Hospitalized1.58 (1.48–1.69)<0.0011.14 (1.02–1.28)0.0171.64 (1.53–1.76)<0.0011.16 (1.03–1.30)0.016
On life support2.27 (2.06–2.50)<0.0011.50 (1.27–1.76)<0.0012.34 (2.11–2.59)<0.0011.59 (1.35–1.88)<0.001
Diabetes1.22 (1.13–1.32)<0.0011.21 (1.10–1.34)<0.0011.33 (1.22–1.44)<0.0011.29 (1.16–1.42)<0.001
Hypertension1.01 (0.93–1.10)0.80.92 (0.82–1.02)0.11.10 (1.01–1.21)0.0320.98 (0.88–1.10)0.7
Peritonitis1.15 (1.02–1.30)0.0280.99 (0.85–1.16)0.91.19 (1.04–1.36)0.0090.98 (0.84–1.16)0.8
Prior transplant2.23 (2.03–2.46)<0.0011.86 (1.62–2.14)<0.0012.18 (1.97–2.41)<0.0011.86 (1.60–2.17)<0.001
Prior abdominal surgery1.29 (1.21–1.38)<0.0011.18 (1.09–1.29)<0.0011.33 (1.24–1.42)<0.0011.21 (1.10–1.32)<0.001
Cold ischemic time        
 <8 hoursReference Reference Reference Reference 
 8–12 hours1.13 (1.05–1.22)0.0021.14 (1.05–1.24)0.0021.08 (0.99–1.17)0.11.10 (1.01–1.21)0.037
 12–16 hours1.37 (1.20–1.55)<0.0011.41 (1.23–1.63)<0.0011.32 (1.15–1.51)<0.0011.36 (1.16–1.58)<0.001
 ≥16 hours1.64 (1.31–2.05)<0.0011.61 (1.26–2.05)<0.0011.39 (1.08–1.78)0.0111.37 (1.04–1.82)<0.026
Table 3. Univariate Interaction Term Analysis to Identify Recipient or Transplant Factors that Modify the Effects of Older Donor Age (≥70 Years) on Risk of Graft Loss and Patient Death
Interaction TermsGraft LossPatient DeathTotal Number of RecipientsNumber of ELD Liver Recipients
  1. Abbreviations: HCC, hepatocellular carcinoma; PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis.

Age <45 years1.50 (0.99–2.26)1.22 (0.75–1.99)3030110
BMI >351.29 (0.88–1.89)1.44 (0.96–2.15)1913142
New diagnosis    
 Hepatitis C (no HCC)1.29 (0.96–1.73)1.40 (1.02–1.91)5907271
 Hepatitis B (no HCC)0.81 (0.35–1.88)0.61 (0.22–1.69)56535
 Acute hepatic necrosis0.91 (0.36–2.28)0.94 (0.34–2.62)39223
 Nonalcoholic fatty liver disease1.03 (0.42–2.50)1.17 (0.48–2.87)27428
 Alcoholic0.84 (0.55–1.29)0.76 (0.47–1.22)2231161
 PBC0.84 (0.36–1.96)1.01 (0.43–2.38)58441
 PSC0.80 (0.40–1.61)0.55 (0.22–1.38)89264
 Autoimmune0.41 (0.10–1.71)0.84 (0.26–2.73)49124
 Cryptogenic0.83 (0.52–1.31)0.60 (0.35–1.05)1355133
 Hepatitis C (with HCC)1.05 (0.57–1.91)1.25 (0.68–2.29)132758
 Hepatitis B (with HCC)0.88 (0.12–6.68)1.01 (0.13–7.60)2186
 HCC (no hepatitis)0.90 (0.46–1.76)0.95 (0.48–1.86)73560
MELD (calculated)1.00 (0.98–1.02)1.00 (0.98–1.02)  
MELD exception1.15 (0.83–1.59)1.14 (0.80–1.61)4420259
Status 11.71 (1.09–2.67)1.62 (1.00–2.61)110854
Hospitalized1.12 (0.83–1.49)1.17 (0.86–1.59)4715267
On life support1.14 (0.74–1.75)1.16 (0.74–1.81)117566
Diabetes1.05 (0.76–1.44)0.93 (0.66–1.30)3229253
Hypertension0.92 (0.65–1.31)1.05 (0.73–1.50)2852212
Peritonitis0.93 (0.48–1.78)0.90 (0.45–1.80)110852
Prior transplant1.78 (1.12–2.83)2.35 (1.48–3.73)128737
Prior abdominal surgery0.85 (0.64–1.13)0.98 (0.73–1.33)7770485
Cold ischemic time >8 hours1.21 (0.89–1.64)1.26 (0.91–1.75)6327389

Because all variables examined were believed to be biologically relevant and thus potential confounders, all variables were used in the multivariate regression models. To confirm that inclusion in a forced model did not affect our conclusions, the models were repeated with serial omissions of covariates that were not statistically significant on univariate analysis. The hazard ratio for graft loss or patient death for donor age ≥70 years did not vary by more than 5% when any covariates were omitted.

Schoenfeld residuals were minimized with stratified Cox models, where proportional hazards assumptions were confirmed. The hazard ratios (HRs) for donor age ≥70 years in stratified models were compared with nonstratified models and did not vary by more than 5%. For ease of interpretation of the covariates, results from nonstratified models are presented in this manuscript.

Missing Data.

For most covariates, data were missing for fewer than 5% of observations; the exceptions were angina (5.7%), CIT (13.5%), hypertension (6.3%), and peritonitis (6.2%). Observations with missing covariates were analyzed via casewise deletion. To confirm that this method of handling missing data did not affect our conclusions, the models were repeated with missing values of each covariate recoded as “missing” categories. Additionally, for each missing covariate, the most extreme cases were also tested—that is, all missing data were recoded as each of the potential values for that covariate and tested in the models. The HR for graft loss or patient death for donor age ≥70 years did not vary by more than 5% regardless of the way in which missing values were handled.

Survival Analyses.

Unadjusted overall graft survival and patient survival rates were estimated using the Kaplan-Meier method and compared using log-rank tests. Comparisons of HRs for graft loss and patient death between recipients of livers from ELDs and younger donors were performed using univariate and multivariate Cox regression models.

Interaction Analysis.

To allow for internal validation, the interaction analysis and subsequent identification of a preferred recipient subgroup was performed on a derivation cohort, a random two-thirds sample of transplants that occurred throughout the study period. Cox regression analysis of first-degree interaction terms with donor age ≥70 years was performed in order to identify recipient factors that modified the effect of donor age on graft loss and patient death. A subgroup of “preferred” recipients was identified that predicted the lowest risk based on examination of the model interaction coefficients, but with the goal of including as many patients as possible in this subgroup. Each factor with HRs for graft loss and patient death ≥1.20 was considered nonpreferred.

Graft survival and patient survival were compared by donor age in both the preferred (patients without the identified risk factors) and nonpreferred (patients with any of the identified risk factors) subgroups using Kaplan-Meier survival estimates and log-rank tests. Internal validation was performed using a validation cohort that was separate from the derivation cohort. In other words, results were validated using the remaining data (the one-third not used to generate the model). Additionally, to test applicability over time, results were further validated using datasets limited to transplants performed in each of years 2002, 2003, 2004, and 2005. The results from each of these validations were similar to the results derived from the entire dataset and reported in this article.

In a separate analysis, to determine the proportion of patients awaiting liver transplantation that would be considered preferred, we tested the applicability of these preferred criteria to the 17,396 active deceased donor liver transplant registrations as reported by UNOS on June 2, 2006.

Statistical Significance.

Unless otherwise specified, all tests were 2-sided with statistical significance set at α = 0.05. All analyses were performed using Stata 9.1 for Linux (StataCorp, College Station, TX).

Results

Patient, Donor, and Transplant Characteristics.

All adult patients (n = 16,921) undergoing LT in the MELD era who met the criteria for analysis were included in this study (Table 1). Of these patients, 1043 (6.2%) received livers from ELDs. The oldest donor was 92 years of age (n = 1), and 126 donors were 80 years of age or older. ELDs were more likely to be female, white, diabetic, hypertensive, and die of cerebrovascular accident (P < 0.001). In general, recipients of ELD livers were somewhat older and more likely to be female, diabetic, and hypertensive (P < 0.005). Furthermore, recipients had marginally lower MELD scores and were less likely to have prior LTs, hepatitis C, or HCC (P < 0.001). ELD livers were much more likely to be shared across regions (31.3% versus 22.3%; P < 0.001) and nationally (16.0% versus 6.1%; P < 0.001). Almost half (48.1%) of ELD transplants were performed at high-volume centers (upper tertile of center volume, performing >300 LT in the MELD era), compared to 31.8% of transplants from younger donors (P < 0.001).

Risk Factors for All Recipients.

When all transplants reflecting current judgment and practice were considered, after adjusting for all available potential confounders, patients receiving ELD livers were at significantly higher risk of graft loss (multivariate HR 1.83, CI 1.55–2.17; P < 0.001) and death (multivariate HR 1.64, CI 1.37–1.97; P < 0.001) compared with patients receiving livers from donors under 40. Other factors associated with poor outcomes after LT were: (1) donor ethnicity, cause of death, and diabetes; (2) recipient age, BMI, ethnicity, diagnosis, MELD score, MELD exception, hospitalization, life support, diabetes, prior transplant, prior abdominal surgery; and (3) transplant CIT (Table 2).

Effect Modifiers and Determining Preferred Recipients.

The modifying effects of various recipient and transplant characteristics on the relative risk of graft loss and patient death associated with donor age ≥70, as well as the sizes of each subgroup, are shown in Table 3. The risk of graft loss from an ELD was 1.29 times worse in patients with hepatitis C (but without HCC) when compared with the risk of graft loss from an ELD in patients with other diagnoses (i.e., the adverse effects of an ELD were amplified in patients with hepatitis C). Similarly, prolonged CIT was not only a risk factor for worse graft outcome from any donor (Table 2), but was also an effect modifier and amplified the adverse effects of an ELD (Table 3). Conversely, older recipient age was a risk factor for worse outcomes from any donor; however, as an effect modifier, it actually attenuated the adverse effects of ELD. Different mechanisms are responsible for the effect modification seen in different covariates. Figure 1 illustrates examples with BMI [where average liver donor (ALD) outcomes are similar for both recipient BMI subgroups, but ELDs are associated with profoundly worse outcomes when BMI >35 is compared with BMI <35] and recipient age (where ELD outcomes are similar for both recipient age subgroups, but ALD outcomes are much better for recipients <45 years of age). It is noteworthy that disease severity, as defined by MELD score, was a risk factor for worse outcomes but did not modify the effect of donor age, indicating that some patients with high MELD scores might do similarly well with ELD or younger livers.

Figure 1.

Effect modification by (A) BMI and (B) age of recipients of ELD livers (age ≥70).

One possible explanation for worse outcomes with ELDs in young patients is that providers only choose young patients for ELDs when these patients are in extremis. To better understand the recipient selection process for young patients receiving ELDs, we compared the registration and disease severity of young patients who received ELD livers with young patients who received ALD livers. Minimal difference was seen in status 1 registration, with 18% of ALD recipients and 19% of ELD recipients listed as status 1 (P = 0.8). However, young recipients of ELD livers had a much lower mean MELD score than young recipients of ALD livers [21.6 versus 24.5; P = 0.001 (2-sided Student t test)]. If ELDs were reserved for only desperate and heroic attempts at performing LT in young patients, we would have expected the opposite effect, with higher MELD scores in ELD recipients. We do acknowledge that more information about practice patterns might be obtained by evaluating the candidates for whom providers turn down ELDs, and as such our inferences are limited to patients who received liver transplants.

To best delineate the effect modification contributions of hepatitis C, hepatitis B, and HCC, we performed separate interaction term analyses for hepatitis C without HCC, hepatitis C with HCC, hepatitis B without HCC, hepatitis B with HCC, and HCC without hepatitis. Of these, the only diagnosis that met our a priori criteria for significant effect modification (HR ≥1.2 for both death and graft loss) was hepatitis C without HCC.

We then selected preferred characteristics that were predicted to minimize the risks associated with ELD while maximizing the size of the subgroup selected. Our goal was to identify a reasonably large subgroup of recipients whose outcomes from ELD livers might be similar to their outcomes from younger livers. We selected “preferred recipients” to be first-time recipients over the age of 45 with BMI <35, non–status 1 registration, CIT <8 hours, and either HCC or an indication for transplantation other than hepatitis C. Of patients awaiting LT as of June 2006, approximately 50% fit this profile (without the CIT restriction); of patients receiving LT during our study period, approximately 25% fit this profile (with the CIT restriction). Furthermore, recipients determined to be preferred for ELDs spanned the entire spectrum of disease severity, with a similar distribution of MELD scores as those determined to be nonpreferred for ELDs (Fig. 2).

Figure 2.

Distribution of MELD score at transplantation, comparing recipients determined to be preferred and nonpreferred for elderly allografts. The mean MELD score was 23.7 for both subgroups (P = 0.7).

The question of whether HCC or hepatitis C takes diagnostic “priority” when evaluating effect modifiers merits further discussion. Hepatitis C without HCC amplified the adverse effects of ELDs and met our a priori criteria for significant interaction (HR 1.29 for graft loss and 1.40 for death). HCC without hepatitis C did not modify the effects of ELDs, with near-null HRs for both graft loss (0.90) and death (0.95). In patients with combined HCV/HCC, there was no appreciable effect modification in terms of graft loss (HR 1.05) but a moderate effect modification in terms of death (HR 1.25). The decision of whether to prefer ELD allocation to patients with hepatitis C who have HCC thus depends on the goals of defining this subgroup. Keeping with the goal of creating an empirically derived and more inclusive preferred subgroup without compromising the results from this subgroup, we only included patients with hepatitis C and no HCC in the nonpreferred group and allowed all other categories to be preferred. All of the analyses that follow were calculated using this definition of preferred recipients. We also considered an alternate classification system in which HCV diagnosis takes precedence over all other diagnoses—that is, considering all patients with HCV to be nonpreferred, including those with HCC. We repeated all of the analyses that follow this alternate classification system and found no differences regarding the performance of the preferred group—that is, ELD livers still performed as well as ALD livers or livers from ideal liver donors (ILDs) in the preferred group. However, this alternate classification system was more selective, allowing fewer patients to be preferred for ELD livers. Under the classification system that we developed empirically, patients with HCV/HCC were preferred for ELD livers. Under the alternate HCV precedence classification system, patients with HCV/HCC would not be preferred for ELD livers. Given that both classification systems allowed for similar results with ELD, ALD, and ILD livers in preferred recipients, we picked the classification system that most closely followed the statistical analysis and was the most inclusive—in other words, the classification that considered HCV/HCC to be preferred. However, equally convincing arguments could be made to pick the classification system that excluded patients with HCV/HCC from the preferred subgroup. Future prospective data will likely clarify this issue.

Subgroup Analysis.

When preferred recipients of ELD livers were compared with preferred recipients of ALD livers under 70 years of age or preferred recipients of ILD livers under 40 years of age, graft and patient survival were similar (Fig. 3, left panels). For preferred recipients, 3-year graft survival with ELD allografts was 74.9% (CI 68.5–80.2) versus 75.0% (CI 73.1–76.8) with ALD allografts (P = 0.6) and 77.3% (CI 74.4–79.9) with ILD allografts (P = 0.2). Similarly, for preferred recipients, 3-year patient survival was 81.2% (CI 76.0–85.4) after transplantation with an ELD allograft versus 80.2% (CI 78.7–81.6) with an ALD allograft (P = 0.9) and 81.2% (CI 78.9–83.3) with an ILD allograft (P = 0.8). Furthermore, patient survival was not significantly affected over various strata of donor age (Fig. 4A, upper panel) or donor risk index, a combination of risk factors defined by Feng and colleagues12 (Fig. 4B, upper panel).

Figure 3.

Kaplan-Meier estimates of (A) graft survival and (B) patient survival after deceased donor LT. Elderly donors (age ≥70) are compared with ALDs (age <70) (upper panels) and ILDs (age <40) (lower panels). Effect of donor age is compared between recipients determined to be preferred (left panels) or nonpreferred (right panels) for elderly allografts.

Figure 4.

Kaplan-Meier estimates of patient survival (left panels) and Cox regression hazard ratios or risk of death (right panels) after deceased donor liver transplantation, stratified by (A) donor age and (B) donor risk index. Effects are compared between recipients determined to be preferred (upper panels) or nonpreferred (lower panels) for elderly allografts.

In contrast, nonpreferred recipients had drastically lower graft and patient survival with ELD allografts when compared with nonpreferred recipients who received allografts from younger donors (Fig. 3, right panels). For these recipients, 3-year graft survival with ELD allografts was 50.4% (CI 44.8–55.8) versus 70.7% (CI 69.6–71.8) with ALD allografts (P < 0.001) and 74.7% (73.1–76.3) with ILD allografts (P < 0.001). Similarly, 3-year patient survival was 64.4% (CI 60.0–68.5) for recipients of ELD allografts versus 77.4% (CI 76.5–78.3) for recipients of ALD allografts (P < 0.001) and 80.0% (CI 78.7–81.2) for recipients of ILD allografts (P < 0.001). When analyzed by strata of donor age and donor risk index, incremental risk patterns were seen (Fig. 4, lower panels).

Center Effects.

Because almost half of ELD transplants were performed at high-volume centers, we analyzed the association between center volume and outcomes. Consistent with the recent report by Northup et al.,27 center volume was not independently associated with differences in outcomes when all donors were included. In adjusted multivariate Cox proportional hazards models, the risk of patient death at high volume centers was 0.99 (95% CI 0.91–1.09, P = 0.9) and the risk of graft loss was 0.98 (95% CI 0.90–1.06, P = 0.6). When outcomes specific to ELDs were analyzed, a trend toward better outcomes at high-volume centers was seen, but these were not statistically significant (patient death HR 0.90, 95% CI 0.65–1.26, P = 0.6; graft loss HR 0.91, 95% CI 0.67–1.24, P = 0.6).

To evaluate if the preferred recipient criteria identified in this study could be generalized to centers of various volumes, outcomes were compared between ELD and ALD/ILD recipients in preferred and nonpreferred recipients, stratified by higher-volume and lower-volume centers. As seen with pooled estimates, nonpreferred patients receiving ELD livers had a 50%–100% increased risk of death and graft loss at higher-volume and lower-volume centers (P < 0.001 for all comparisons). Similarly, as seen with pooled estimates, there was no statistically significant difference between outcomes of preferred patients receiving ELD livers and those patients receiving ALD or ILD livers, at higher-volume or lower-volume centers (P > 0.1 for all comparisons). It is possible that high-volume centers may have established better criteria for selecting ELD livers (by examining confounders not captured by our data, such as graft steatosis), or may have more effective clinical protocols for managing recipients of ELD livers. Regardless, the adverse effects of ELD livers were significantly attenuated in all preferred patients, regardless of the transplant center's volume.

Furthermore, there are likely some practice pattern differences between transplant centers that the UNOS data cannot capture. To make sure that the concept of preferred recipients for ELDs is generalizable to all transplant centers, regardless of differences in practices, the following analysis was performed. Graft and patient survival of preferred recipients of ELD transplants were compared with preferred recipients of ILD and ALD transplants on a center-by-center basis (for all centers where the number of ELD transplants was considered large enough for log-rank test comparison, which we defined as n ≥ 10). For every center that we studied, there was no statistical difference (P > 0.1 for every transplant center) comparing ELD with ILD and ALD for preferred recipients, while there was a significantly worse graft and patient survival when nonpreferred recipients received ELDs compared with ILDs and ALDs (P < 0.001 for every transplant center). These findings were similar to those obtained from the pooled analysis that included all centers.

Discussion

We reviewed the characteristics and outcomes of 1043 LTs performed using livers from donors over 70 years of age in the MELD era in the United States. Twelve percent of these allografts were procured from donors over 80 years of age, and the oldest donor reported in this time period was 92 years of age. Almost 50% of these LTs were performed at upper-tertile volume centers, 31% were shared regionally, and 16% were shared nationally. Despite the donor scrutiny and recipient selection process that currently occurs, patients receiving livers from ELDs incurred an 83% increased risk of graft loss (CI 1.55–2.17; P < 0.001) and a 64% increased risk of death (CI 1.37–1.97; P < 0.001) when compared with similar patients receiving livers from ILDs. However, we identified a subgroup of recipients whose outcomes were not affected by donor age. These were first-time recipients over the age of 45 with BMI <35, non–status 1 registration, CIT <8 hours, and either HCC or an indication for transplantation other than hepatitis C. In these patients, outcomes with ELD livers were comparable to outcomes with livers from younger, including ideal, donors (3-year graft survival: ELD 75%, ALD 75%, ILD 77%, P > 0.1; 3-year patient survival: ELD 81%, ALD 80%, ILD 81%, P > 0.1).

Our findings demonstrate clinically important effect modification, the interaction between donor and recipient attributes, that occurs in liver transplantation with expanded criteria allografts. Donor risk factors identified by a population-based approach, where all recipients are studied as a single cohort, are limited by an assumption that these factors affect all recipients in the same manner or in a manner accountable for by other measured confounders. However, in this subgroup analysis, we have shown that some patients are less affected by the mechanism through which older allografts cause poor outcomes.

Inferences made from our findings must be based on the assumption that there was no systematic bias in reporting. This bias is less likely given the prospective nature of the UNOS cohort, although observational data collected in a nonrandomized, nonblinded fashion remain at risk of systematic reporting errors, thereby limiting our study. One advantage of the study is that in the MELD era, during which allocation has been based on disease severity, the key recipient risk factors that comprise the MELD score have been well documented and audited, because they affect organ allocation. In addition, reporting of graft survival was likely accurate, because most patients with failed allografts either die or return to the waiting list, and therefore must register with UNOS. Finally, the reporting of patient death was supplemented with record linkage to the Social Security Master Death File. We therefore believe that any minor reporting errors were unlikely to have influenced our observations. Furthermore, our sensitivity analysis further suggests that our conclusions were not biased by missing data.

Another limitation of this study is the possibility of unmeasured confounding. The most critical missing potential confounder was graft steatosis, which unfortunately was not well captured by the current UNOS dataset. It is likely that the selection practices in place for evaluating ELD have already limited the use of ELD livers to those with little evidence of harmful steatosis.28–30

Finally, our study is limited by the possibility of unmeasured outcomes. Even with equivalent graft and patient survival, transplantation with ELD livers could be complicated by other morbidities, including blood transfusions, initial poor function, hepatic artery thrombosis, prolonged intensive care unit stay, biliary tract ischemia, and cholestasis.2, 21, 31, 32 However, if such complications occurred, they did not effect a difference in ultimate survival of the graft or patient. We do acknowledge that no more than 4 years of follow-up were available since the initiation of MELD-based organ allocation, so inferences must be limited to outcomes within this period.

Despite the positive changes introduced with the MELD liver allocation system, a severe discrepancy still exists between the number of patients who would benefit from liver transplantation and the number for whom this treatment is available. Our analysis of national experience with over 1000 ELD liver transplants suggests that, when appropriately allocated, donors over the age of 70 represent a viable source for significantly expanding the liver donor pool and addressing the growing organ shortage. We have identified a subgroup of recipients who seem to fare equally well with livers from older or younger donors. Because the definition of this subgroup does not depend on MELD score, it is possible to allocate older livers to preferred recipients without violating the spirit of disease severity–based organ allocation. For nonpreferred recipients with high MELD scores, the balance between the increased risk of graft loss and death with an ELD liver versus the risk of waiting for a better organ offer depends on the attributes of the donor, the recipient, and regional organ availability, and as such remains in the hands of the experienced clinician.

Appendix

Table  . UNOS Diagnosis Codes for Patients Classified as “Other”
Recipient Diagnosis (UNOS diagnosis code)% of Recipients (Donor Age <70)% of Recipients (Donor Age ≥70)
AHN: DRUG OTHER (4100)2.82.9
AHN: TYPE A (4101)0.20.0
AHN: TYPE B- HBSAG+ (4102)2.51.0
AHN: TYPE D (4105)0.00.0
AHN: TYPE B AND D (4107)0.10.0
AHN: ETIOLOGY UNKNOWN (4108)4.52.1
AHN: OTHER4.92.6
CIRRHOSIS: DRUG/INDUST EXPOSURE OTHER SPECIFY (4200)0.60.8
CIRRHOSIS: TYPE A (4201)0.20.0
CIRRHOSIS: TYPE B- HBSAG+ (4202)8.17.9
CIRRHOSIS: TYPE D (4205)0.10.3
CIRRHOSIS: TYPE B AND D (4207)0.20.0
CIRRHOSIS: CRYPTOGENIC- IDIOPATHIC (4208)17.124.3
CIRRHOSIS: CHRONIC ACTIVE: ETIOLOGY UNKNOWN (4209)0.20.5
CIRRHOSIS: OTHER2.82.4
CIRRHOSIS: AUTOIMMUNE (4212)9.46.3
CIRRHOSIS: CRYPTOGENIC (IDIOPATHIC) (4213)8.38.9
CIRRHOSIS: FATTY LIVER (NASH) (4214)4.86.3
SEC BILIARY CIRRHOSIS: CAROLI'S DISEASE (4230)0.20.0
SEC BILIARY CIRRHOSIS: CHOLEDOCHOL CYST (4231)0.00.0
SEC BILIARY CIRRHOSIS: OTHER SPECIFY (4235)0.60.5
FAMILIAL CHOLESTASIS: BYLER'S DISEASE (4250)0.00.0
FAMILIAL CHOLESTASIS: OTHER SPECIFY (4255)0.20.3
CHOLES LIVER DISEASE: OTHER SPECIFY (4260)0.90.3
NEONATAL HEPATITIS OTHER SPECIFY (4265)0.20.0
BILIARY ATRESIA: EXTRAHEPATIC (4270)0.30.0
BILIARY HYPOPLASIA: NONSYNDROMIC IHBD PAUCITY (4271)0.00.3
BILIARY HYPOPLASIA: ALAGILLEÆS SYNDROME (4272)0.10.0
BILIARY ATRESIA OR HYPOPLASIA: OTHER0.30.0
CONGENITAL HEPATIC FIBROSIS (4280)0.10.0
CYSTIC FIBROSIS (4285)0.10.0
BUDD-CHIARI SYNDROME (4290)1.30.3
METDIS: ALPHA-1-ANTITRYPSIN DEFIC A-1-A (4300)3.22.6
METDIS: WILSON'S DISEASE (4301)1.20.8
METDIS: HEMOCHROMATOSIS- HEMOSIDEROSIS (4302)1.32.1
METDIS: GLYC STOR DIS TYPE I (GSD-I) (4303)0.20.0
METDIS: PRIMARY OXALOSIS/OXALURIA- HYPER (4307)0.10.0
METDIS: OTHER SPECIFY (4315)0.81.0
PLM: CHOLANGIOCARCINOMA (CH-CA) (4403)0.51.0
PLM: HEMANGIOENDOTHELIOMA-HEMANGIOSARCOM (4405)0.30.0
PLM: OTHER SPECIFY (4410)0.50.5
BILE DUCT CANCER (CHOLANGIOMA-BILIARY TR (4420)0.00.3
SECONDARY HEPATIC MALIGNANCY OTHER SPECIFY (4430)0.20.8
BENIGN TUMOR: HEPATIC ADENOMA (4450)0.10.3
BENIGN TUMOR: POLYCYSTIC LIVER DISEASE (4451)0.60.8
BENIGN TUMOR: OTHER SPECIFY (4455)0.10.0
GRAFT VS. HOST DIS SEC TO NON-LI TX (4510)0.30.3
TRAUMA OTHER SPECIFY (4520)0.20.0
OTHER SPECIFY (999)19.721.7

Ancillary