Long-term results and modeling to predict outcomes in recipients with HCV infection: Results of the NIDDK liver transplantation database


  • Michael Charlton,

    Corresponding author
    • Mayo Clinic, Division of Gastroenterology & Hepatology, 200 First St. SW, Rochester, MN 55905
    Search for more papers by this author
    • telephone: 507-266-7054; FAX: 507-266-1856

  • Kris Ruppert,

  • Steven H. Belle,

  • Nathan Bass,

  • Daniel Schafer,

  • Russell H. Wiesner,

  • Katherine Detre,

  • Yuling Wei,

  • James Everhart

  • Prepared for the National Institute of Diabetes and Digestive and Kidney Diseases Liver Transplantation Database.


Hepatitis C virus (HCV)-associated liver disease is the most common indication for liver transplantation (LT). There are, however, no long-term (>5 year) studies of comparative outcomes for recipients with HCV infection, and no models capable of identifying recipients with HCV infection at greatest risk for adverse outcomes. We prospectively determined: 1) long-term outcomes, and 2) whether pretransplant patient or donor variables can be used to predict death and/or graft loss in recipients with HCV infection. A total of 165 HCV-infected recipients were eligible for this study. Pretransplant donor and recipient characteristics and patient and graft survival data were prospectively collected. Model building for outcomes was carried out using logistic regression. Receiver operating characteristic curves for different models were created and compared. Median follow-up was 8.5 years. Adjusted 10 year graft survival was 64% for recipients with HCV infection and 51% for uninfected recipients. A model incorporating pretransplant HCV ribonucleic acid (RNA), cytomegalovirus immunoglobulin (CMV IgG) serostatus, creatinine, bilirubin, prothrombin time international ratio (INR), recipient age, and donor age was developed to identify recipients at greatest risk of short-term mortality or graft loss (area under receiver operating characteristic curve = .83) In conclusion, long-term outcomes following LT for recipients with HCV infection are comparable to those for recipients undergoing LT for other indications. HCV-infected recipients at greatest risk for short-term mortality and graft loss can be identified using several readily identifiable pretransplant variables. Long-term death and graft loss specifically secondary to recurrence of HCV appears, however, to be determined primarily by factors other than those included in this analysis. (Liver Transpl 2004;10:1120–1130.)

End-stage liver disease associated with hepatitis C virus (HCV) infection is the most common indication for liver transplantation (LT) in the United States.1 Although short-term patient and graft survival for HCV-infected recipients have generally been reported to be similar to that of most other indications,2–7 liver transplant recipients with HCV infection are unique in that their original liver disease recurs almost universally (as measured by the detectability of HCV ribonucleic acid [RNA] in serum following LT). Recurrence of hepatitis C is a substantial source of morbidity, mortality, and graft loss, with 1 in 10 HCV-infected liver transplant recipients dying or requiring retransplantation due to HCV associated graft failure in the first 5 postoperative years.3 In the nontransplant setting, it is estimated that the mean time from infection to the development of cirrhosis is 20 years.8 Furthermore, a recent analysis of the United Network for Organ Sharing (UNOS) database found that patient and graft survival at 5 years posttransplantation may, in fact, be poorer for HCV-infected recipients than that of other indications.9 To date, meaningful data regarding posttransplant follow-up in HCV-infected recipients is available only for means of 3–5 years.2, 3, 6, 9–13 As the duration of post-LT follow-up of HCV-infected recipients lengthens, the impact of HCV recurrence on long-term patient and graft survival is likely to increase. There are, however, few prospective reports of long-term (more than 5 years) outcomes following LT in HCV-infected recipients. In addition, although individual host and viral characteristics have been associated with poor patient and graft survival,14, 15 no models exist that can reliably identify those HCV-infected recipients at greatest risk for patient or graft mortality following LT. Before transplantation, the model for end-stage liver disease (a numerical score derived from serum total bilirubin, serum creatinine, and international ratio for prothrombin time) has been shown to be predictive of mortality for patients with cirrhosis.16

The Liver Transplantation Database (LTD) was established by the National Institutes of Diabetes and Digestive and Kidney Diseases (NIDDK) in 1990 to prospectively collect serum, tissue, and clinical and demographic information from patients being evaluated for LT. We previously reported a detailed analysis of medium-term outcomes and predictors of patient and graft survival for HCV-infected liver transplant recipients.3 Using the NIDDK LTD, we have now prospectively determined: 1) long-term mortality and graft failure following LT among recipients with HCV-infection, and 2) whether pre-LT patient or donor variables can be used to identify HCV-infected LT recipients at risk of posttransplant mortality and/or graft loss.


HCV, hepatitis C virus; RNA, ribonucleic acid; LT, liver transplantation; UNOS, United Network for Organ Sharing; LTD, Liver Transplantation Database; NIDDK, National Institute of Diabetes and Digestive and Kidney Diseases.


Study Patients

The LTD consisted of 805 adult patients who underwent LT at 3 medical centers:17 Mayo Clinic and Foundation, Rochester, MN; the University of Nebraska, Omaha, NE; and the University of California, San Francisco, CA. The study, coordinated at the University of Pittsburgh's Graduate School of Public Health, was approved by the Institutional Review Boards of each of the centers. Dates of initial transplants occurred between April 15, 1990 to June 30, 1994. Clinical follow-up data and vital statistics were collected through November, 2002.

To be eligible for this study, recipients had to be adults (at least 16 years old), receive a single organ transplant, receive a liver from a donor who was core hepatitis B negative, survive for at least 1 day postoperatively, and have a diagnosis of HCV (n = 165). Patients were considered to have hepatitis C if they tested positive for anti-HCV by second generation enzyme-linked immunosorbent assay and were also reactive for anti-HCV by recombinant immunoblot assay, or had HCV RNA present in serum as detected by reverse-transcriptase polymerase chain reaction.


Induction immunosuppression at Mayo Clinic consisted of cyclosporine, prednisone, and azathioprine. The University of Nebraska used cyclosporine and prednisone. The University of California at San Francisco utilized antilymphocyte globulin followed by cyclosporine, prednisone, and azathioprine. All centers participated in the FK506 Primary Immunosuppression Trial, resulting in a subgroup of 92 recipients receiving a tacrolimus based regimen.


Frozen serum and tissue samples were stored as part of a comprehensive HCV study for the NIDDK LTD. To minimize specimen degradation, tissue and serum were stored at −20°C and transported on dry ice to the Mayo Clinic and Foundation Tissue and Serum Bank, where samples were thawed, aliquoted, and stored at −70°C before testing. Of the samples, 98% were drawn within 3 days before transplantation; 2% were drawn between 1 day and 3 months before transplantation.

Viral RNA Extraction and Amplification

Aliquots of 100 μl of sera were extracted by a guanidine thiocyanate lysis protocol using reagents supplied in the Amplicor Hepatitis C Test kit (Roche Diagnostics, Branchburg, NJ) for amplification of the 5′ untranslated region and genotyping reactions. Modifications included the addition of 2 μl of Pellet Paint (Novagen, Madison, WI) to facilitate recovery of nucleic acid pellets after isopropanol precipitation and the addition of 40 units of recombinant RNasin in a final 25 μl volume with RNase free water to stabilize the recovered RNA before reverse-transcriptase polymerase chain reaction.

Deoxyribonucleic Acid Sequencing and Genotyping

Specimens were subjected to deoxyribonucleic acid sequencing and gel electrophoresis on an ABI Prism 377 Genetic Analyzer (Applied Biosystems, Foster City, CA) to generate deoxyribonucleic acid sequence information. Sequences were compared to published HCV type reference sequences using the FASTA algorithm (Wisconsin Genetics Computer Group, Madison, WI). Genotypes were assigned based upon percentage match scores to the reference strains.18

Nucleotide primers specific for a 401 bp target sequence within the NS5 region of HCV were utilized as described by Simmonds et. al.18 Reverse transcription was carried out as previously described.19

HCV-RNA Quantitation (Branched Deoxyribonucleic Acid Assay)

Pretransplantation HCV RNA, expressed in HCV RNA viral equivalents per milliliter (vEq/mL), was quantitated in serum by signal amplification employing branched deoxyribonucleic acid in a sandwich hybridization assay (Quantiplex Version 2.0; Chiron, Emeryville, CA). The lower limit of sensitivity of this assay is .2 × 106 vEq/mL. A test of trend demonstrated that the percentage of recipients with HCV RNA titers greater than the overall median of 340,000 vEq/mL did not change significantly over time (P = .66, mean storage time 4.5 years).3

Statistical Analysis

The cohort was described using estimates of central tendency (means, medians) and spread (standard deviation, range) for continuous data and frequencies and percentages for categorical data. Groups were compared using the chi-square test for differences in proportions (categorical data) and the Wilcoxon Rank Sum test (continuous data). The groups that were compared are those who had all the available data for the short-term analysis (n = 118) to those missing at least 1 data item (n = 47), and those above (n = 46) and below (n = 72) the LTD HCV Risk Score cutoff value of 2.05. Probable ratio estimates were used to establish a more meaningful interpretation of the LTD HCV Risk Score.

Short-Term Outcomes Modeling

The purpose of this analysis was to develop a model, using pretransplant characteristics, to predict short-term mortality or graft failure. Short-term graft survival (<90 days) was modeled using logistic regression. The coefficients from the model were used to calculate an overall score (the LTD HCV Risk Score). The score was used to make a prediction concerning outcome and the area under the receiver operating characteristic curve was calculated. A single cut-off value was determined that maximized the sum of the sensitivity and specificity of the model. Survival curves were estimated for scores above and below that cut-off point. Classification and regression tree analysis is an empirical method for discriminant analysis.20 This methodology creates recursive binary splits of the sample until a stopping criterion is met. For these analyses, each binary split occurred at a node defined by age that maximally discriminated short-term retransplantation-free survivors from those that either died or underwent retransplantation within 90 days. It is from these trees that the age categories for both the donor and recipient were obtained.

Long-Term Outcomes Modeling

To identify potential predictors of patient or graft failure, Cox proportional hazards models were constructed. Only pretransplant characteristics were used in the models. These included basic recipient age, race, gender, INR, total bilirubin, creatinine, CMV IgG serostatus, and donor age, race, and gender. Adjusted survival is reported for males, 40–50 years of age at transplant, who received a liver from a donor between the ages of 30–40 years, and had a Pugh score of 9 pretransplantation.


Cohort Characteristics

The cohort consisted of 165 HCV patients of whom 118 had complete data to calculate the short-term LTD HCV model (14 had missing viral loads, 3 missing bilirubin, 1 missing creatinine, 30 missing CMV serostatus, 1 missing INR). There was no significant difference between those who had complete data vs. those that did not with respect to the pretransplant characteristics. Demographic data for both the long-term and the short-term study groups can be found in Table 1. For the entire 165, follow-up ranged from 0–12.0 years. Median follow-up was 8.5 years.

Table 1. Recipient Demographics
VariableAll HCV Recipients (n = 165) N (%)LTD HCV Model (n = 118) N (%)
  1. Abbreviations: HCV, hepatitis C virus; LTD, liver transplantation database; UNMC, University of Nebraska Medical Center; UCSF, University of California, San Francisco; CMV, cytomegalovirus; IgG, immunoglobulin G.

 Mayo22 (13.3)22 (18.6)
 UNMC46 (27.9)42 (35.6)
 UCSF97 (58.8)54 (45.7)
 Male120 (72.7)85 (72.0)
 Female45 (27.3)33 (28.0)
 White124 (75.2)88 (74.6)
 Black6 (3.6)6 (5.1)
 Other35 (21.2)24 (20.3)
 < 56124 (75.2)86 (72.9)
 ≥5641 (24.9)32 (27.1)
 Mean (SD)48.1 (10.1)48.6 (10.2)
Donor Age  
 <3070 (42.7)53 (44.9)
 ≥3094 (57.3)65 (55.1)
 Mean (SD)35.3 (16.1)35.2 (16.2)
 Positive100 (74.6)90 (76.3)
 Negative34 (25.4)28 (23.7)
Viral load  
 <1 Meq/mL104 (68.9)78 (66.1)
 ≥1 Meq/mL47 (31.1)40 (33.9)
 1a52 (34.4)44 (37.3)
 1b54 (35.8)40 (33.9)
 2b11 (7.3)10 (8.5)
 3a18 (11.9)11 (9.3)
 Other13 (8.6)11 (9.3)
 Unsequenced3 (2.0)2 (1.7)

HCV-infected recipients were younger than HCV-negative controls, included a greater proportion of non-Caucasian recipients, and were more likely to be male. HCV-infected recipients had higher mean Child-Pugh score (mean 8.9 vs. 8.6, P = .04) and longer mean cold ischemia time (mean 10.8 vs. 10.2 hours, P = .02) than HCV-negative controls. The cohorts were similar with respect to final UNOS status, proportion of recipients with renal insufficiency (creatinine >2 mg/dL), donor liver quality, and donor age. A full description of demographics in the HCV-negative cohort has been published previously.3

Overall Posttransplant Mortality and Graft Loss

Adjusted patient survival for recipients with HCV infection (80% at 5 and 67% at 10 years posttransplantation) was comparable to that of recipients without HCV infection (75% at 5 and 59% at 10 years posttransplantation). Adjusted graft survival for recipients with HCV infection (77% at 5 and 64% at 10 years posttransplantation) was also comparable to that of recipients without HCV infection (68% at 5 and 51% at 10 years posttransplantation (Fig. 1).

Figure 1.

Cumulative graft survival following LT for recipients according to primary diagnosis before transplantation is shown. Chol, cholestatic liver diseases; non-B-C, chronic non-B-C hepatitis; metab, metabolic liver disease; ALD, alcoholic liver disease; HBV, end-stage liver disease secondary to chronic hepatitis B infection.

Recipients undergoing transplantation for hepatitis B infection or malignancies had significantly poorer cumulative survival (Relative Risk [RR] = 2.4 and 2.2, P = .003 and P = .02, respectively) than the HCV-infected recipients, while patients transplanted because of cholestatic liver disease (primary biliary cirrhosis and primary sclerosing cholangitis) had significantly better cumulative survival (RR = .4, P = .001) than the HCV-infected recipients. Patient survival for recipients with alcoholic liver disease, metabolic liver disease, and chronic non-B-C hepatitis was not significantly different from the HCV-infected cohort. Similar results were observed for graft survival.

Early Causes of Mortality or Graft Loss Among Recipients With HCV Infection (Less Than 90 Days)

Causes and timing of mortality and graft loss for the entire cohort are summarized in Table 2. The most common cause of early death was non-HCV infections (n = 4), and the most common cause of early graft failure was primary nonfunction (n = 3). Other causes of early death were stroke or myocardial infarction (MI) (n = 2), perforated peptic ulcer (n = 1), and central pontine myelinolysis (n = 1). Other causes of early graft failure were hepatic artery thrombosis/occlusion (n = 2) and rejection (n = 1).

Table 2. Causes and Timing of Death/Graft Loss (< 90 Days)*
Reason for ReTxCause of DeathTime to Graft Failure (Days)Time to Death (Days)
  • Abbreviation: CPM, central pontine myelinolysis.

  • *

    Each line indicates one person.

Primary non-functionPerforated peptic ulcer51,072
 Stroke 20
Primary non-function 163,528
Severe cholestasis 452,901
 Pneumonia 2
 Pneumonia 17
 Opportunistic infection 30
Hepatic artery thrombosis 873,431
Primary non-function 33,255
 Opportunistic infection 38
Myocardial Infarction  58

Short-Term LTD HCV Model

For the 118 recipients who had complete data, the model incorporating the following readily identifiable pretransplantation characteristics had the largest area under the curve: donor age (stratified into less than 30 years vs. at least 30 years), recipient age (less than 56 years vs. at least 56 years), pre-LT viral load (less than 1 miliequivalents [Meq]/mL vs. at least 1 Meq/mL), pre-LT creatinine, total bilirubin, INR, and recipient CMV IgG serostatus. Scores for each patient were calculated using the following formula:

equation image

where donor age = [0 = less than 30 years; 1 = at least 30 years], recipient age = [0 = less than 56 years; 1 = at least 56 years], viral load = [0 = less than 1 Meq/mL; 1 = at least 1 Meq/mL], CMV IgG = [0 = negative; 1 = positive], total bilirubin, creatinine, INR = [continuous].

The area under the curve for this model (.83) was basically greater than all other models tested, including the model for end-stage liver disease (area under the curve = .52) (Table 3 and Fig. 2). The 1 model whose area under the curve = .84 was not used due to the rarity of the 3a genotype. We believe that this score could have been obtained by chance. The mean LTD HCV score for those surviving without retransplantation for at least 90 days post-LT (n = 107) was 1.29 (median 1.40) vs. 3.15 (median 3.32) for those who died or underwent retransplantation in the first 90 days (n = 11). An LTD HCV score of 2.05 had a sensitivity of 82% and specificity of 65% for predicting early post-LT mortality or graft failure. Kaplan-Meier survival curves and demographics of recipients according to LTD HCV scores are shown in Fig. 3 and Table 4, respectively. Table 5 demonstrates the likelihood ratio of mortality by LTD HCV score. Higher scores lead to a higher probability of early mortality or retransplantation compared to a lower score. After testing various models, we were unable to find variables significantly associated with long-term patient or graft survival.

Table 3. LTD-HCV Model Compared to Other Models Tested
ModelFormulaArea Under Curve
  1. Abbreviations: dagecat4, donor age categories <30, > = 30; vload, pre-operative viral load < 1 Meq/ml, > = 1 Meq/ml; agecat2, recipient age <56, > = 56; cmvgr, CMV IgG 0 = negative 1 = positive; creat, creatinine, tbili, total bilirubin; INR, international normalized ratio; genotype 1, 3a; 0, others.

Original Meld10 × (.957 × log(creat) + .378 × log(tbili) + 1.120 × log(INR) + .643)0.52
LTD-HCV model2.6288 × dagecat4 + 1.4394 × agecat2 + −0.6978 × cmvgr + 0.9862 × vload + 0.0194 × log(tbili) + −0.1681 × log(INR) + −1.9881 × log(creat)0.83
A2.3794 × dagecat4 + 1.7925 × agecat2 + .7493 × vload + −1.0686 × cmvgr + 2.2064 × genotype)0.84
B1.4772 × agecat2 + 2.3556 × dagecat4 + −0.8074 × cmvgr + 0.9779 × vload0.79
Figure 2.

Receiver operator characteristic curve for LTD HCV model is shown. The mean LTD HCV model risk score for recipients surviving at least 90 days posttransplantation (n = 107) was 1.29 (median 1.40) vs. 3.15 (median 3.32) for those who died or underwent retransplantation in the first 90 days (n = 11). An LTD HCV score of 2.05 had a sensitivity of 82% and specificity of 65% in predicting post-LT mortality or graft failure.

Figure 3.

Variation in cumulative mortality according to pretransplant LTD HCV model risk score is shown. A single cut-off value (2.05) was determined that maximized the sum of the sensitivity and specificity of the model.

Table 4. Variation in Demographics According to LTD HCV Score
VariableScore < 2.05 (n = 72) n (%)Score ≥ 2.05 (n = 46) n (%)P-Value
  • Note: 2.05 is the value with the highest sensitivity and specificity.

  • Chi-square test.

  • Wilcoxon Rank Sum test.

Gender  0.10
 M48 (66.7)37 (80.4) 
 F24 (33.3)9 (19.6) 
Race  0.19
 White53 (73.6)35 (76.1) 
 Black6 (8.3)0 (0.0) 
 Hispanic7 (9.7)8 (17.4) 
 Other6 (8.3)3 (6.5) 
Viral Load  0.001
 ≤1 Meq/ml56 (77.9)22 (47.8) 
 >1 Meq/ml16 (22.2)24 (52.2) 
Donor Age (yr)  0.0001
 <3051 (70.8)2 (4.4) 
 ≥3021 (29.2)44 (95.7) 
Recipient age (yr)  0.06
 <5657 (79.2)29 (63.0) 
 ≥5615 (20.8)17 (37.0) 
CMV IgG  0.02
 Neg12 (16.7)16 (34.8) 
 Pos60 (83.3)30 (65.2) 
Total Bilirubin (mg/dL)  0.8
 Mean (SD)4.6 (9.1)3.5 (3.1) 
Creatinine (mg/dL)  0.01
 Mean (SD)1.3 (0.6)1.0 (0.2) 
INR  0.9
 Mean (SD)1.5 (0.2)1.5 (0.4) 
Table 5. Distribution of LTD-HCV Scores by Short-Term Mortality
LTD HCV ScoreDeath and/or Retransplanted (%)Alive and Not Retransplanted (%)Likelihood Ratio*
  • *

    The likelihood ratio offers a measure of the impact of a variable (e.g., LTD HCV score) on the probability of an outcome (e.g., early mortality). The likelihood ratio of short-term death and/or retransplantation increases with LTD HCV score.

−3 to 11/11 (9%)48/107 (45%)0.20
1–2.041/11 (9%)22/107 (21%)0.43
2.05–3.96/11 (55%)35/107 (33%)1.7
4+3/11 (27%)2/107 (2%)13.5

Later Causes of Death or Graft Loss (at Least 90 Days)—Entire Cohort

Recurrent HCV with ensuing graft failure was the most frequent primary or secondary cause of death, occurring in 22 / 56 (39%) of the HCV-infected recipients who died. Recurrent hepatitis C was also the most frequent indication for retransplantation (5 / 8, 63%). Other causes of death were non-HCV infections (n = 8), malignancy (n = 7), primary cardiac or cerebrovascular events (n = 3), other (n = 11), and unknown (n = 2). The proportion of graft failure (resulting in death or retransplantation) due to recurrence of HCV increased with duration of follow-up, accounting for 43% (10 / 23) of graft failure in years 1–5 posttransplant vs. 48% (10 / 21) in years 6–10. Median time to graft failure or death due to recurrence of HCV was 1,581 days (range 266–3,858). Using the previously mentioned pretransplant donor and patient characteristics, proportional hazards regression models21 were created to identify parameters that predicted patient death or graft failure due to recurrent disease. None of the characteristics were significantly associated with these outcomes.


In this study, using the combined results of 3 large centers in the United States, long-term outcomes following LT for a highly characterized, prospectively followed cohort of recipients with HCV infection have been described. One of the most important observations of this analysis is that liver transplant recipients with HCV infection have 10 year patient and graft survival rates that are similar to those of recipients undergoing LT for other indications. Although overall graft survival is comparable to that of other indications, HCV-associated allograft injury is the most common cause of both death and graft failure among HCV-infected recipients. Furthermore, the proportion of graft failure secondary to HCV increased with duration of follow-up, accounting for 43% of graft failure in years 1–5 posttransplant vs. 48% in years 6–10. Thus, recurrence of HCV is a major cause of graft failure in HCV-infected recipients. These results seem to be superficially in contrast to a recent analysis of the UNOS9 that reported relatively poorer 5 year graft survival among recipients with HCV infection when compared to non-HCV-infected recipients. Differences between the UNOS and the NIDDK have been extensively discussed previously.22 Importantly, patient and graft survival rates for HCV-infected recipients in the NIDDK LTD were greater than or equal to that of all groups in the UNOS analysis at 1, 3, and 5 years posttransplantation, suggesting the patient populations or the databases are substantially different and making direct comparisons difficult. The period of study (year of transplantation) was also different between the UNOS and NIDDK cohorts. This may be significant as more recent years of transplantation have been reported to be associated with more rapid progression of recurrence of HCV.23 Finally, patient and graft survival for recipients with HCV infection in the UNOS database was only significantly worse than that of recipients with primary biliary cirrhosis. The overall findings of the 2 databases are thus generally similar. The applicability of our results to patients currently undergoing LT for HCV infection is unknown. Outcomes may be affected by the use of newer immunosuppressive agents, such as mycophenolate mofetil. This is an inherent limitation to long-term follow up studies, which, by definition, span considerable lengths of time. Nonetheless, a prospective study of long-term outcomes in a relatively large cohort of liver transplant recipients with HCV infection provides new and important insights into the impact of HCV recurrence on patient and graft survival.

A second specific aim of this study was to develop a model that predicted posttransplant patient death or graft failure following LT for hepatitis C using variables that can be identified before transplantation. Such a model was constructed using several readily determined pretransplant variables. After the first 90 postoperative days, rates of mortality and graft loss did not vary significantly with LTD HCV score. The LTD HCV score thus identifies recipients most likely to experience early graft loss and/or mortality. The rates of mortality and graft loss at 1 and 5 years posttransplantation were 22 and 39%, respectively, among recipients with a high (2.05) pretransplant LTD HCV score vs. 10 and 22% for recipients with a lower (<2.05) LTD HCV score. This model incorporates several parameters that have been shown to be independently predictive of poor patient and/or graft survival in HCV-infected recipients (pretransplant viral load, greater recipient age, bilirubin, and INR) and 2 variables that have not previously been associated with adverse outcomes specifically in HCV-infected recipients (pretransplant CMV serostatus and older donor age).

A model that stratifies recipients according to the likelihood of adverse outcome has a principle goal of assigning an estimate of risk. The suitability of an individual as a potential liver transplant recipient is, primarily, determined by the anticipated likelihood of procedural success, as measured by patient and graft survival. Patients for whom the anticipated risk is perceived to be too great, e.g., recipients with poor cardiac function or metastatic adenocarcinomatosis, are precluded from being listed for LT unless the risk can be modified. One of the more intriguing aspects of the LTD HCV model is the potential to modify the risk score for individual patients, e.g., by matching higher risk recipients with younger donors and/or by administering antiviral therapy to lower pretransplant viral load. In a multicenter pilot study, we found administering antiviral therapy to patients with Child's B cirrhosis to be associated with poor tolerability.24 The routine administration of antiviral therapy to patients with decompensated liver disease to reduce HCV levels before transplantation is thus, currently, a potentially hazardous intervention of unproven benefit. The utility of antiviral therapy in more compensated cirrhosis is well established.25 We found donor and recipient age to be the most powerfully predictive of early posttransplant mortality/graft loss. Several studies have associated more advanced donor age with poor patient and graft survival in adults26–28 and also in children.29 More recently, advanced donor age has been specifically associated with increased risk for graft loss among HCV-infected recipients.30, 31 Although the mechanism remains unclear, livers from older donors have been shown to have a higher prevalence of ischemic injury following LT.32 Donor age is a potentially safe modifiable variable. In the LTD HCV model donor, ages were categorized as <30 years and 30 years. To lower the LTD HCV model risk score (keep below 2.05), recipients with higher scores might be limited to receiving organs from donors <30 years of age. Whether this would translate into a survival benefit is unknown. Although advanced age at onset of HCV infection has also been associated with more rapidly progressive fibrosis,33 this is unlikely to account for the excess in early deaths/graft loss seen in our study as none were attributable to recurrence of HCV.

The value of pretransplant CMV serological status (presence or absence of anti-CMV IgG) in modeling posttransplant patient and graft survival is, perhaps, not surprising. Cytomegalovirus infection is common among liver transplant recipients, particularly among those who are immunologically naive for this β-herpes virus.34 CMV has been shown to be directly immunosuppressive and to increase the risk of other opportunistic infections.35, 36 Two studies have reported that posttransplant CMV infection mortality in HCV-infected recipients is associated with increased risk for allograft cirrhosis and non-HCV infection related mortality.30, 37, 38 Negative CMV IgG serology before LT is a known risk for posttransplant CMV disease, particularly when the donor is positive for CMV IgG.39 It is also possible that pretransplant CMV seronegativity may be indicative of relatively poor host immunity. Alternatively, virus–virus interactions have also been postulated to modify the pathogenesis of specific human viral infections. The observed association between CMV disease and HCV in this and previous studies30, 37, 40 may be analogous to the described interaction between human immunodeficiency virus and HCV.41, 42 Whether CMV prophylaxis in higher risk recipients, e.g., recipients who are CMV IgG negative who receive organs from CMV IgG positive donors, would be associated with improved outcomes is not known.

The finding of a significant association between HCV genotype 3a early mortality/graft loss was unexpected. HCV genotype 3a increased the area under the receiver operator curve by a relatively small amount in our model (area under the curve .81 without vs. .84 with adjusting for genotype 3a). The relatively small numbers of recipients with genotype 3a infection (n = 11), precludes definitive comparisons of demographic characteristics of genotype 3a infected recipients with those of recipients infected with other HCV genotypes. Previous reports of the impact of HCV genotype on outcomes in HCV-infected recipients following orthotopic liver transplantation have been mixed, with several groups finding that HCV genotype 1b does2 or, conversely, does not,5, 7, 43 impart an increased risk for more severe chronic allograft injury.2, 14, 19, 44, 45 Although genotype 3a added to the predictivity of our model for early posttransplant mortality/graft loss in our current study, it has no independently predictive value in identifying recipients at risk for medium- or long-term mortality or graft loss.2, 14, 19, 44, 45 A larger study would be needed to determine whether infection with HCV genotype 3a confers an increased relative risk of mortality/graft loss or is indicative of some other host or viral characteristic.

The basis of the increased risk of patient mortality and graft loss associated with higher pretransplant HCV RNA levels is not clear. The most common cause of early death in this study was non-HCV infections (n = 4), and the most common cause of early graft failure was primary nonfunction (n = 3). Other causes of death were stroke (n = 1), perforated peptic ulcer (n = 1), and central pontine myelinolysis (n = 1). Other causes of early graft failure were hepatic artery thrombosis/occlusion (n = 2) and rejection (n = 1). Thus, HCV recurrence did not account for any of the early mortality or graft loss. We have previously reported, in this same cohort, that higher pretransplant HCV RNA titers are associated with a higher frequency of severe posttransplant HCV recurrence and also death and graft loss due to non-HCV infections were also more common (19.8 vs. 3.7% at 5 years posttransplantation, P = .003).3 An increased incidence of serious infections following LT for end-stage liver disease secondary to HCV has also been reported by others.46 It is possible that HCV recurrence either mimics or causes allograft rejection, resulting in greater exposure to immunosuppressive agents and, thereby, increased susceptibility to serious infection.

It is possible that the factors found to be predictive of early posttransplant mortality and graft loss are markers of poor overall recipient immunity (CMV IgG seronegativity, higher HCV titer, and donor and recipient age). The association of higher bilirubin and INR with early posttransplant mortality suggests that these variables reflect general debility of the recipient.

In conclusion, liver transplant recipients with HCV infection experience overall patient and graft survival rates that are similar to those of recipients without HCV infection. HCV-associated allograft injury is, however, the most common cause of both death and graft failure among HCV-infected recipients. We have developed a sensitive and specific model that identifies patients with HCV infection who are at greatest risk for early post-LT mortality and graft loss. This model includes variables that are readily identifiable before transplantation. Because at least 1 of the parameters, donor age, can be used for selecting donors to improve the risk score generated using this model, it is possible that the LTD HCV model may have utility in optimizing patient and graft survival following LT in recipients with HCV infection. An immediate goal is to prospectively validate this model in a separate, more contemporary, cohort.


Members of the NIDDK LTD who contributed to this study included the following—from the Department of Epidemiology, University of Pittsburgh Graduate School of Public Health, Pittsburgh, PA: Katherine M. Detre, M.D., Ph.D. (Principal Investigator), A. Jake Demetris, M.D. (Co-Investigator), Steven H. Belle, Ph.D. (Co-Investigator), Yuling L. Wei, M.S. (Co-Investigator), Kris Ruppert, M.S.N. (Statistician), Tatiana Ledneva, M.S. (Data Manager), Carol Penovich (Data Manager), Heather Eng, B.A. (Data Analyst); from the Departments of Medicine and Surgery, Mayo Clinic Foundation, Rochester, MN: Russell H. Wiesner, M.D. (Principal Investigator), Michael R. Charlton, M.D. (Co-Investigator), Ruud A.F. Krom, M.D. (Co-Investigator), Lori Schwerman, R.N. (Research Nurse Coordinator), Nancy Evans (Manuscript Preparation); from the Departments of Medicine and Surgery, the University of Nebraska Medical Center, Omaha, NE: Daniel Schafer, M.D. (Principal Investigator), Byers W. Shaw, Jr., M.D. (Co-Investigator), Rowen K. Zetterman, M.D. (Co-Investigator), Elizabeth Strudthoff (Research Nurse Coordinator); from the Departments of Medicine and Surgery, University of California, San Francisco, CA: John Roberts, M.D. (Principal Investigator), Nathan Bass, M.D., Ph.D. (Co-Investigator), Bev Nikolai, CCRC (Research Coordinator); and from the Division of Digestive Diseases and Nutrition, NIDDK, National Institutes of Health, Bethesda, MD: James Everhart, M.D., M.P.H. (Program Director), Jay H. Hoofnagle, M.D. (Division Director).