Potential conflict of interest: Nothing to report.
Address reprint requests to: Nathan J. Shores, Tulane Transplant Institute, 1415 Tulane Ave., New Orleans, LA 70112. E-mail: Nshores@tulane.edu; fax: 504-988-1105.
African American (AA) liver transplant (LT) recipients with hepatitis C virus (HCV) have higher rates of graft loss than other racial/ethnic groups. The Donor Risk Index (DRI) predicts graft loss but is neither race- nor disease-specific and may not be optimal for assessing donor risk for AA HCV-positive LT recipients. We developed a DRI for AA with HCV with the goal of enhancing graft loss predictions. All U.S. HCV-positive adult AA first deceased donor LTs surviving ≥30 days from March 2002 to December 2009 were included. A total of 1,766 AA LT recipients were followed for median 2.8 (interquartile range [IQR] 1.3-4.9) years. Independent predictors of graft loss were donor age (40-49 years: hazard ratio [HR] 1.54; 50-59 years: HR 1.80; 60+ years: HR 2.34, P < 0.001), non-AA donor (HR 1.66, P < 0.001), and cold ischemia time (CIT) (HR 1.03 per hour >8 hours, P = 0.03). Importantly, the negative effect of increasing donor age on graft and patient survival among AAs was attenuated by receipt of an AA donor. A new donor risk model for AA (AADRI-C) consisting of donor age, race, and CIT yielded 1-year, 3-year, and 5-year predicted graft survival rates of 91%, 77%, and 68% for AADRI <1.60; 86%, 67%, and 55% for AADRI 1.60-2.44; and 78%, 53%, and 39% for AADRI >2.44. In the validation dataset, AADRI-C correctly reclassified 27% of patients (net reclassification improvement P = 0.04) compared to the original DRI. Conclusion: AADRI-C identifies grafts at higher risk of failure and this information is useful for risk-benefit discussions with recipients. Use of AA donors allows consideration of older donors. (Hepatology 2013;58:1263–1269)
Hepatitis C virus (HCV) is the leading indication for liver transplantation (LT) in the United States. Compared to Caucasians, African-Americans (AA) have relatively superior outcomes with chronic HCV disease prior to transplantation,[2, 3] but experience more aggressive recurrence of HCV disease after liver replacement.[4, 5] The 2-year and 5-year graft survival for HCV-positive AA LT recipients has been reported to be as much as 10% lower than in non-AA recipients.[6, 7] The reason for this disparity in outcome is poorly understood. A lower likelihood of responding to antiviral therapy post-LT may be one factor.[8, 9] Donor factors are likely to be of importance also.
The Donor Risk Index (DRI)—derived from 20,023 predominantly pre-Model for Endstage Liver Disease (MELD) era U.S. liver transplants—was originally proposed in 2006 to predict LT recipient outcome based on available donor factors. Containing seven donor variables, DRI predicts post-LT graft failure using a continuous, numerical scoring system. The DRI was a milestone in highlighting the importance of donor quality on LT outcomes, and while the inclusion of a large, heterogeneous recipient pool maximized its generalizability, the DRI may have more limited prediction among specific subgroups, such as those transplanted for HCV. Prior retrospective studies have shown a strong and consistent association between donor age and severity of HCV recurrence.[11, 12] Interestingly, in the original DRI, allografts from AA donors, compared to Caucasian donors, were associated with an increased risk (hazard ratio [HR] 1.19, 95% confidence interval [CI] 1.10-1.29, P < 0.001) of posttransplant graft failure (death or re-LT); but several recent studies of HCV-infected transplant recipients have independently demonstrated a trend of improved graft outcomes when AA donor livers were paired with HCV-positive AA recipients.[4, 13, 14] With these observations in mind, we sought to define the donor factors of importance in AA recipients with HCV and to develop a donor risk model that accurately estimates risk of graft loss for this patient subgroup.
Materials and Methods
With Institutional Review Board (IRB) approval, we examined adult AA recipients of deceased donor liver transplants from March 1, 2002 to December 31, 2009 (MELD-era) with primary, secondary, or other diagnosis of HCV recorded in the UNOS Standard Transplant Analysis and Research (STAR) file created on June 30, 2011. We excluded liver retransplants and recipients with Status 1, human immunodeficiency virus (HIV-coinfection, or less than 30 days of follow-up. The primary outcome was post-LT graft loss (recipient death or retransplant).
Recipient and donor factors were described with frequency distributions and medians (interquartile ranges [IQRs]). Covariates evaluated included recipient age, gender, height, body mass index (BMI), blood type, diabetes, life support at transplant, region of transplant center, previous abdominal surgery, dialysis week prior to transplant, HBV surface antigen, hepatocellular carcinoma (HCC), simultaneous kidney transplant, laboratory values at transplant (MELD, creatinine, bilirubin, albumin), and donor age, gender, gender match, ethnicity, height, weight, BMI, blood type, blood type match, cytomegalovirus status match with recipient, antihypertensives pre-cross clamp, HCV antibody, hepatitis B core antibody, vasodilators, diabetes, history of hypertension, blood urea nitrogen, creatinine, SGOT, SGPT, bilirubin, donation after cardiac death, cause of death, share type, partial/split liver, cold ischemic time (CIT), and transplant year. Missing CIT (7%) and CIT less than 2 hours or greater than 20 hours (1.5%) were imputed with the median CIT for the region by share type.
The Kaplan-Meier method was used to estimate observed posttransplant graft survival. The log-rank test compared survival estimates across strata and Bonferroni corrected P values adjusted for multiple comparisons.
We used the Cox proportional hazards model to evaluate recipient and donor factors associated with graft loss. Time to graft loss was defined as days from liver transplant to the first of retransplant or death. Patients alive or lost to follow-up were censored at the date of last follow-up. When valid Social Security death dates were available for patients coded as alive or lost to follow-up, posttransplant follow-up status and date were updated with data from the Social Security death certificate master file. Donor factors with a prespecified statistical significance of P < 0.1 were analyzed by multivariate Cox regression models. Backwards elimination with P < 0.05 was used to select the multivariate donor model. The final model was adjusted for recipient age, gender, HCC, blood type match, laboratory MELD and albumin at transplant, and region. A novel donor risk model specific for AA recipients with HCV (AADRI-C) was developed. We investigated the interaction between donor age and donor race. The adjusted donor model was stratified by donor race (AA versus non-AA) to quantify and demonstrate differences in the risk of graft failure for the donor age by donor race interaction. Predicted survival estimates for tertiles of AADRI-C (tertile 1, AADRI-C <1.6; tertile 2, AADRI-C 1.6-2.44; and tertile 3, AADRI-C >2.44) and DRI (tertile 1, DRI <1.18; tertile 2, DRI 1.18-1.55; and tertile 3, DRI >1.55) were derived from the Cox proportional hazards model.
To compare the AADRI-C to the DRI, we identified a separate cohort of 294 HCV-positive AA patients receiving liver transplants between January 1, 2010 and January, 31, 2011 in the UNOS STAR file (created April 30, 2012) meeting our study selection criteria. These patients were not included in the original development dataset. In this validation dataset, we measured model discrimination (the ability of a model to correctly classify subjects into events and nonevents) with the overall C-index. We assessed improvement in model performance by quantifying the proportion of correct risk reclassification by AADRI-C at 1 year post-LT using the net reclassification improvement (NRI). NRI utilized a priori 1-year graft loss risk groups stratified as <7.5%, 7.5% to <10%, 10% to <12.5% and 12.5% to <15% and ≥15% to compare the AADRI-C model to DRI. Statistical analyses were conducted using SAS v. 9.2 (Cary, NC) and figures were created using Stata v. 11.1 (College Station, TX).
Recipient and Donor Characteristics
A total of 1,766 MELD-era AA LT recipients followed for a median of 2.8 (IQR 1.3-4.9) years were included (Table 1). Recipients were 70% male, had median age of 54 years, and 38% were transplanted with HCC. The corresponding donors (Table 2) were 60% male with a median age of 42 years (IQR: 26-53), 22% were AA and 7.3% were anti-HCV positive. The median CIT was 7 (IQR: 5.3-8.3) hours.
Table 1. Recipient Characteristics and Univariate Association With Graft Loss
Table 2. Donor and Transplant Characteristics and Univariate Association With Graft Loss
HR (95% CI)
Cold ischemic time (per hr >8 hrs)
BUN mg/dL (ln)
Creatinine mg/dL (ln)
SGOT U/L: (ln, centered at 10)
SGPT U/L (ln)
Total bilirubin g/dL (ln)
Gender match (recip/donor)
Partial Split Liver
Cause of death
Anti-hypertensives pre-cross clamp
Hep C antibody positive
HBV Core antibody positive
History of hypertension
Donor Factors Associated With Graft Survival
Overall, 1-year, 3-year, and 5-year graft survival rates for HCV-positive AA LT recipients were 85%, 65%, and 54%, respectively. Donor characteristics associated with graft loss in univariate analysis (Table 2), including age, female donor/female recipient match, non-AA/AA mismatch, cause of death, HBV core antibody, diabetes, history of hypertension, cold ischemia time, BMI, and blood urea nitrogen met the criteria for evaluation in multivariate analysis. After adjusting for recipient age, gender, HCC, blood type match, region, and laboratory values at transplant (MELD and albumin), the only donor characteristics independently predicting graft loss were older donor age (40-49 years: HR 1.54; 50-59 years: HR 1.80; 60-69 years: HR 2.03; ≥70 years: HR 2.83; P < 0.001), donor non-AA (HR 1.66, P < 0.001), and CIT per hour increase over 8 hours (HR 1.03 per hour increment, P = 0.03) (Table 3).
Table 3. Adjusted Independent Variables Included in AADRI-C Model Predicting Risk of Graft Loss
*Adjusted for recipient age, gender, HCC, blood type match, laboratory MELD, and albumin at transplant and region.
Donor age 10-39
CIT (per hr >8 hrs)
We detected a significant interaction between donor age and donor race (P = 0.047). Stratifying the model by donor race (AA n = 395, non-AA n = 1371) revealed an attenuation of the increased risk of graft loss with increasing age among AA donors (Table 4; Supporting Fig. 1). Risk of graft loss increased with increasing donor age among recipients of non-AA donor grafts across all donor age categories (P < 0.001) compared to donors age 10-39. In contrast, risk of graft loss was not significantly increased in recipients of AA donors ages 40-49 (HR 1.09, P = NS) or 50-59 (HR 1.17, P = NS) compared to donors age 10-39. Risk of graft loss did not increase until AA donors were ≥60 years of age (HR 1.93, P = 0.02). Overall, the 5-year post-LT graft survival in AAs receiving an AA donor 40 years of age or older was significantly higher compared to AA receiving a non-AA donor of similar age (P = 0.02 to P < 0.001) (Supporting Fig. 1).
Table 4. Risk of Graft Loss Stratified by Donor Age and Race
AA Donor (n = 395)
Non-AA Donors (n = 1371)
*Adjusted for recipient age, gender, HCC, blood type match, laboratory MELD and albumin at transplant and region.
AADRI-C Development and Evaluation
Donor age, AA donor status, and CIT were included in a new risk model for HCV-positive AA liver transplant recipients (AADRI-C). Observed 5-year graft survival estimates by tertiles of AADRI-C (tertile 1, AADRI-C <1.6; tertile 2, AADRI-C 1.6-2.44; and tertile 3, AADRI-C >2.44) were 69%, 54%, and 39%, respectively (P < = 0.001) (Fig. 1). The 1-year, 3-year, and 5-year predicted graft survival for AADRI-C tertile 1 were 91%, 77%, and 68%; for AADRI-C tertile 2 were 86%, 67%, and 55%; and for AADRI-C tertile 3 were 79%, 53%, and 39%, respectively. Predicted graft survival for tertiles of AADRI-C and DRI are shown (Supporting Fig. 2).
Examples of combinations of donor age, donor race, and CIT and the corresponding predicted AADRI-C survival rates are shown in Table 5. These examples reflect the strong favorable influence of AA donor race on HCV-positive AA recipient graft outcomes. For example, an HCV-positive AA receiving a 59-year-old graft from an AA with 8 hours CIT would be predicted to have ∼15% higher graft survival than receiving a similar graft donated by a non-AA or comparable graft survival to receiving a <40-year-old graft from a non-AA donor with 8 hours of CIT.
Table 5. Example Combinations of Donor Risk Factors and the Corresponding AADRI-C Predicted Survival
Calculation: AADRI-C = exp[(0.433 if 40≤age<50)+(0.588 if 50≤age<60)+(0.850 if age≥60)+(0.504 if non-AA race)+((CIT-8)*0.033)].
Compared to the original DRI, AADRI-C better predicted risk of graft failure in AA HCV-positive recipients in both the development (C-index 0.56 and 0.60, respectively) and validation (C-index 0.51 and 0.55, respectively) datasets. Furthermore, estimated 1-year risk of graft loss calculated by AADRI-C correctly reclassified 19% of patients (NRI P < 0.001) in the development dataset and 27% of patients (NRI P = 0.04) in the validation dataset.
In our disease-specific and race-specific assessment of donor quality and its association with graft failure in HCV-infected AA transplant recipients, the only donor factors of importance were age, race, and CIT. The AADRI-C classifies risk of graft loss among AA recipients more accurately than the original DRI. Donor age—as in the original DRI—remains the dominant predictor of graft outcome in HCV-positive recipients in the AADRI-C model. However, for the first time we identify a potential age effect modifier, namely, donor AA race. We found that receipt of an AA donor liver attenuated the negative effect of increasing donor age on graft survival. Specifically, compared to AA recipients with donors under the age of 40 years, AA recipients of livers from AA donors had no statistically significant decline in graft survival until the donor age was 60 years or greater. This is a particularly important finding given that the original DRI found that, among all transplant recipients, graft outcomes were inferior with use of livers from AA donors.
The DRI remains a landmark innovation for discussing donor risk in LT. However, limitations in the DRI noted since its original presentation may hamper its current utility. For instance, DRI includes a great deal of pre-MELD era data that may not reflect post-MELD trends in donor quality. Also, over time, donors have become older and more obese, while recipients have become more ill on average.[17, 18] Most important in terms of our analysis, the DRI evaluated donor factors in a heterogeneous cohort of adult deceased-donor liver recipients, including all varieties of transplant indications and recipient races/ethnicities. Therefore, it may perform differently in patient and disease subsets. For example, Maluf et al. found that the same DRI score predicts significantly worse outcomes for HCV-positive patients than in HCV-negative recipients. For these reasons, we developed a donor risk model specific to HCV-positive recipients in the MELD era of LT, and focused on AA recipients because of their previously described poor long-term graft survival.
There is currently a donor shortage in Western countries. In 2009, the United States alone had 26% of patients listed for liver transplant die or become too ill to transplant. Most patients removed from the list without transplant receive at least one offer before they dropped off the list and most of those offers are refused for perceived issues of donor quality. The ability to utilize older donors in specific patient subsets without compromising outcomes provides a modest means of expanding the donor pool and potentially reducing wait-list mortality.
The matching of AA donors with HCV-positive AAs has previously been criticized as too impractical to apply to day-to-day donor selection. However, given the significant risk of graft loss within 5 years for AA with non-AA donors, especially older non-AA donors, plausible clinical scenarios that may allow matching of AA donors to AA recipients should be considered. The AADRI-C may also be useful in identifying AA recipients at highest risk for graft loss who may benefit from more intensive monitoring and/or early HCV treatment post-LT. An HCV-positive AA recipient transplanted with a high AADRI-C graft (>2.44) has a predicted 3-year graft survival of only 53% compared to 3-year survival with a low AADRI-C (<1.6) donor of 77%. A clinician might target this high AADRI-C recipient for timely antiviral therapy.
The underlying pathogenesis linking AA-derived allografts with improved postliver transplant outcomes in AAs is unclear. In a pretransplant setting, AAs carry a disproportionate burden of HCV infection in the U.S. population and there is epidemiological evidence suggesting AAs spontaneously clear acute HCV infection less often than non-AAs.[22-24] However, chronically infected AAs may actually progress to cirrhosis more slowly than Caucasians. Investigators have looked for racial differences in immune response to HCV that explain the apparent dichotomy in AA outcomes with HCV infection acutely and chronically. It has been theorized that ethnic trends in HLA typing and KIR type predicts spontaneous viral clearance and sustained virological response to interferon-based therapy. For example, HLA-A*02 and HLA-DRB1*12 genotypes were associated with treatment-induced viral clearance in non-Caucasians but not in Caucasians, and natural killer cell immunoglobulin receptor KIR2DL3 was associated with both treatment and spontaneous clearance in HLA-C patients.[26, 27] Also, differences in CD4 T-cell responses and programmed cell death differ significantly among Caucasians and AAs and independently associate with odds of viral response to treatment prior to liver transplant.[28, 29] It is likely that complex, ethnically based differences in immune response to HCV underlie the benefit of matching grafts from AA donors to AA liver recipients.
Most famously, IL28-B CC (versus non-CC) genotype has a well-described linkage to viral clearance pretransplant; and the disparity of CC prevalence in AAs versus non-AAs partially explains poorer response to interferon-based treatments.[23, 30] Charlton and colleagues have recently confirmed that IL28B CC recipient status and CC donor status are positively associated with postliver transplant sustained viral response. Interestingly, however, genotype CC donors were associated with greater posttransplant fibrosis, graft failure, and liver-related death. Biggins et al. recently confirmed these latter findings with more severe HCV disease seen with IL28B CC grafts, especially when transplanted into non-CC recipients. It may be that the lower likelihood of IL28B-CC genotype among AA donors underlies the superior outcomes in HCV-positive AA recipients receiving AA donor grafts.
Our study has limitations inherent to the retrospective collection of donor characteristics and recipient outcomes in a large database. However, the size of the database and the relatively standardized, prospective collection of pretransplant recipient and donor data add statistical power and generalizability to our results. It represents the largest possible cohort of HCV-positive AAs recipients and is consistent with prior results from multicenter and center-specific studies of HCV disease outcomes in AA recipients.[5, 14]
In summary, we identified the key donor factors associated with graft survival among AA LT recipients with HCV: donor age, donor race, and CIT. The AADRI-C will be helpful to clinicians making decisions about specific donor offers for HCV-positive AAs, in guiding the intensity of post-LT monitoring and timing of post-LT antiviral therapy, and in framing discussions with AA recipients regarding graft selection. Ultimately, with the use of AADRI-C, as well as improved therapeutic interventions, it is anticipated that AA LT recipients with HCV will enjoy the same post-LT outcomes as other non-AA liver recipients.