Survival after liver transplantation: Is racial disparity inevitable?


  • Potential conflict of interest: Nothing to report.


Previous analyses have reported that minority patients undergoing orthotopic liver transplantation (OLT) have poorer survival than Caucasian recipients. The reason for this disparity is unclear. We examined whether racial differences in survival exist at select academic OLT centers. OLT recipients from 4 academic centers were prospectively enrolled in 2 multicenter databases. Data including demographics, liver disease diagnosis, and post-OLT follow-up were obtained for 2823 (135 African, 2448 Caucasian, and 240 other race) adult patients undergoing primary OLT between 1985 and 2000. The survival of patients and grafts after OLT was compared across race. The Kaplan-Meier estimates for 1-year recipient survival were 90.8% [95% confidence interval (CI): 86.0–95.9] for African Americans, 86.5% (95% CI: 85.1–87.9) for Caucasians, and 84.4% (95% CI: 79.8–89.2) for other races. The 5-year recipient survival probability was 69.2% (95% CI: 60.1–79.7) for African Americans, 72.2% (95% CI: 70.1–74.4) for Caucasians, and 67.5% (95% CI: 60.5–75.3) for other races. The 10-year recipient survival probability for African Americans was 54.4% (95% CI: 41.1–72.1), for Caucasians 50.7% (95% CI: 46.4–55.3), and for other races 55.7% (95% CI: 41.5–74.8). There was no difference in patient survival (P = 0.162) or graft survival (P = 0.582) among racial groups. A multivariable proportional hazards model confirmed the absence of an association between race and post-OLT survival after adjustments for age, gender, total bilirubin, creatinine, prothrombin time, and diagnosis. Conclusion: These data demonstrate that as a proof of principle, minority OLT recipients should not necessarily expect an OLT outcome inferior to that of Caucasians. (HEPATOLOGY 2007.)

There has been a controversy concerning whether the race of recipients affects patient and graft survival following orthotopic liver transplantation (OLT). After an initial report from the University of Pittsburgh in 1989,1 a number of studies have indicated poorer survival for African American recipients in comparison with Caucasian recipients.2–5 Other published studies have suggested no racial differences in outcome following OLT.6–10 The discrepancy between these studies may be explained by small sample sizes that lacked the power to detect meaningful differences between races or inadequate controls for confounding factors, such as the fact that recipient race may have been only a surrogate for other factors (for example, comorbidity or more advanced liver failure) that affect the outcome more directly.

There may be several mechanisms by which race may be associated with outcome after OLT. Some may have a biological basis, such as immunologic disparities between donors and recipients of different racial backgrounds and genetic differences in the pharmacokinetics or pharmacodynamics of immunosuppressant agents.2, 11–13 Alternatively, they may be socioeconomic, such as inadequate health insurance delaying presentation for OLT in minority patients, socioeconomic status affecting the affordability or choice of immunosuppressive agents, and cultural beliefs that may affect the acceptance of organ transplantation or posttransplant care.14–16 Because more than one of these factors may be in effect, it may be difficult to disentangle these complex relationships.

Recently, Nair et al.17 analyzed data from the United Network for Organ Sharing (UNOS) and reported that being African or Asian American was an independent risk factor for poorer outcome after OLT in comparison with being Caucasian. In a follow-up report evaluating the influence of socioeconomic status on OLT outcome, it was concluded that race was still a strong independent predictor of survival, even though socioeconomic status had some effect on survival.18 These data are strong because they captured all OLT recipients in the nation. However, one important confounding variable was not taken into account in these analyses: the effect of the hospital or region in which OLT was performed. For example, if minority patients are aggregated at OLT centers with inferior outcomes, it would lead to a spurious result that race is responsible for the poor outcome when, in fact, the problem may be that minority patients have limited access to centers with optimal OLT outcomes.

In this study, we consider racial differences in survival after OLT, using data from selected, high-volume, academic OLT centers after adjusting for other factors that affect post-OLT survival. Our data show no differences in OLT outcome among different races at these select centers, suggesting that disparities seen in the nationwide setting are not inevitable.


CI, confidence interval; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; HR, hazard ratio; MELD, model for end-stage liver disease; OLT, orthotopic liver transplantation; UNOS, United Network for Organ Sharing.

Materials and Methods

Data Source and Elements.

This study was based on 2 National Institutes of Health–sponsored multicenter OLT databases. First, the National Institute of Diabetes and Digestive and Kidney Disease Liver Transplantation Database prospectively enrolled patients who underwent OLT between 1985 and 1994 at the Mayo Clinic (Rochester, MN), the University of California at San Francisco (San Francisco, CA), and the University of Nebraska (Omaha, NE). Second, the Model for End-Stage Liver Disease (MELD) Study Group Database included OLT recipients between 1994 and 2000 from the Mayo Clinic and Baylor University (Dallas, TX). From these databases, all adult (age > 18 years) primary liver transplant recipients were included in this analysis.

In both databases, the recipient race was designated by the healthcare provider (physician or transplant coordinator) and categorized as Caucasian, African, or other race. In addition to race, pre-OLT variables that may have a potential impact on post-OLT survival were extracted as covariates in the analysis. These included demographic (age and gender), clinical (liver disease diagnosis), and laboratory data (MELD components at the time of the initial evaluation and OLT). Liver disease diagnoses were coded as fulminant hepatic failure, cancer (primary liver malignancy), alcoholic liver disease, cholestatic liver disease, hepatitis B, hepatitis C, and other. The causes of fulminant hepatic failure were extracted if this information was available.

Statistical Methods.

For a comparison of the baseline characteristics across the racial groups, the Wilcoxon-Rank sum test was used for continuous variables, and a chi-square test was used for categorical variables. Laboratory data at the time of the initial evaluation were used for this comparison. Survival was calculated with the last known follow-up date in comparison with the date of OLT. Patients who were alive on the last day of follow-up were censored at that time. The Kaplan-Meier method was used to estimate recipient and graft survival after transplant by race. The proportional hazards regression analysis was used to examine the impact of race on a recipient's survival, while stratifying on the center. Covariates to adjust for differences in baseline characteristics included age, gender, liver disease diagnosis, and laboratory data to represent the severity of liver disease at the time of OLT. A P value less than 0.05 was used for statistical significance in all analyses.


Recipient Characteristics.

A total of 2823 OLT recipients were included in this study after the exclusion of 28 patients (1.0%) whose race was not reported. As shown in Table 1, 135 (4.8%) recipients were African, 2448 (86.7%) were Caucasian, and 240 (8.5%) were of other races, including Asian and Hispanic. African American recipients were significantly younger than Caucasians (P < 0.01). There were more female recipients among African Americans than other groups (P = 0.01). Overall, hepatitis C was the commonest single disease entity. In comparison with Caucasians, the proportion of African Americans with fulminant (P < 0.01) and other liver diseases (P = 0.02) was higher, whereas the proportion with alcoholic (P = 0.047) and cholestatic liver diseases (P = 0.02) was lower. With respect to other races, the frequency of hepatic malignancy (P < 0.01) and hepatitis B virus (HBV; P < 0.01) was higher and the frequency of alcoholic (P < 0.01) and cholestatic liver diseases (P < 0.01) was lower in comparison with Caucasians.

Table 1. Recipients' Characteristics by Race
  • *

    P < 0.05(reference: Caucasian).

  • Mean ± the standard deviation.

Total number2448 (86.7%)135 (4.8%)240 (8.5%)2823
 A1407 (89.9%)101 (6.5%)57 (3.6%)1565 (55.4%)
 B630 (89.9%)7 (1.0%)64 (9.1%)701 (24.8%)
 C189 (87.5%)7 (3.2%)20 (9.3%)216 (7.7%)
 D222 (65.1%)20 (5.9%)99 (29.0%)341 (12.1%)
Age (years)49.8 ± 10.943.1 ± 11.7*48.2 ± 11.049.3 ± 11.0
Male gender1351 (55.2%)59 (43.7%)*126 (52.5%)1536 (54.4%)
Diagnosis group    
 Fulminant hepatic failure100 (4.1%)20 (14.8%)*14 (5.8%)134 (4.7%)
 Cancer91 (3.7%)2 (1.5%)18 (7.5%)*111 (3.9%)
 Alcoholic liver disease326 (13.3%)10 (7.4%)*17 (7.1%)*353 (12.5%)
 Cholestatic disease605 (24.7%)21 (15.6%)*30 (12.5%)*656 (23.2%)
 Hepatitis B virus109 (4.5%)5 (3.7%)42 (17.5%)*156 (5.5%)
 Hepatitis C virus573 (23.4%)29 (21.5%)67 (27.9%)669 (23.7%)
 Other644 (26.3%)48 (35.6%)*52 (21.7%)744 (26.4%)

Table 2 compares laboratory data at the time of the initial evaluation among the 3 racial groups. Of the patients with nonfulminant liver disease, African Americans had significantly higher serum total bilirubin (P = 0.03), creatinine (P = 0.045), and prothrombin time (P < 0.01) in comparison with Caucasians. This trend for African American patients presenting with more advanced liver disease was not seen with fulminant disease. African American patients with fulminant liver failure had significantly lower total bilirubin than Caucasians (P = 0.01). The etiology of fulminant hepatic failure was most commonly unknown (n = 49), followed by viral hepatitis (n = 42), drug toxicity (n = 7), and autoimmune and other miscellaneous causes. No apparent pattern was found among the racial groups (data not shown).

Table 2. Recipients' Characteristics at the Time of the Initial Evaluation by Race (Median and Interquartile Range)
 Nonfulminant Hepatic Failure
Caucasian (n = 2348)African (n = 115)Other (n = 226)
  • *

    P < 0.05 (reference: Caucasian).

Total bilirubin (mg/dL)2.9 (1.8–5.9)4.1 (1.7–10.1)*2.9 (1.6–6.1)
Creatinine (mg/dL)0.9 (0.8–1.2)1.0 (0.8–1.4)*0.9 (0.8–1.2)
Prothrombin time (seconds)13.7 (12.4–15.2)14.2 (12.8–17.3)*13.8 (12.6–15.5)
 Fulminant Hepatic Failure
Caucasian (n = 100)African (n = 20)Other (n = 14)
Total bilirubin (mg/dL)25.6 (17.7–31.0)18.8 (13.9–20.7)*17.4 (3.6–22.7)
Creatinine (mg/dL)1.3 (0.8–2.6)1.0 (0.8–1.2)0.9 (0.7–1.0)
Prothrombin time (seconds)25.0 (18.5–34.4)26.0 (17.5–47.8)26.5 (13.3–27.9)

Survival of OLT Recipients by Race.

Figure 1 shows patient and graft survival for the 3 racial groups. The Kaplan-Meier estimates for 1-year, 5-year, and 10-year patient survival was 86.5%, 71.6%, and 50.8%, respectively, and the graft survival was 81.3%, 66.3%, and 46.9%, respectively. There was no difference among the racial groups for patient survival (P = 0.16) or graft survival (P = 0.58). In particular, patient survival between Caucasian and African American patients overlapped extensively.

Figure 1.

Patient (A) and graft (B) survival by race.

Figure 2 compares the survival of patients by race with fulminant and nonfulminant liver diseases. The 1-year, 5-year, and 10-year survival of patients with nonfulminant disease was 87.1%, 71.7%, and 49.9%, respectively. Among patients with fulminant liver disease, the 1-year, 5-year, and 10-year survival was 77.2%, 68.9%, and 63.6%, respectively. Overall, the patients with fulminant liver disease had survival results comparable to those of the patients with nonfulminant disease (P = 0.78). More importantly, in neither group was there any difference among the racial groups in patient survival (P = 0.74 and 0.13 for fulminant and nonfulminant diagnoses, respectively).

Figure 2.

Patient survival in nonfulminant (A) and fulminant (B) hepatic failure by race.

The univariate proportional hazards regression analysis showed that age [hazard ratio (HR) = 1.25], female gender (HR = 0.81), serum creatinine (HR = 1.05), liver cancer (HR = 1.54), alcoholic liver disease (HR = 1.27), viral hepatitis (HR = 1.57), and cholestatic liver disease (HR = 0.47) were significantly associated with patient survival (Table 3). In agreement with Fig. 2, fulminant liver disease was not associated with an increased risk of death. Other race was associated with an increased risk of death (HR = 1.36, P = 0.03) in comparison with Caucasian race, whereas African race was not (HR = 0.90, P = 0.53).

Table 3. Univariate Proportional Hazards Regression Analysis for Patient Survival**
FactorHazard Ratio (95% Confidence Interval)P
  • *

    Laboratory data at the time of transplantation.

  • **

    Analysis stratified on center.

Race (reference: Caucasian)  
 African0.90 (0.64, 1.26)0.53
 Other1.36 (1.03, 1.79)0.03
Age (decades)1.25 (1.17, 1.34)<0.001
Female gender0.81 (0.70, 0.93)0.004
Creatinine*1.05 (1.01, 1.11)0.03
Total bilirubin*1.01 (1.00, 1.01)0.15
Prothrombin time*1.01 (1.00, 1.02)0.15
Diagnosis group  
 Fulminant hepatic failure (versus not)1.04 (0.75, 1.44)0.81
 Cancer (versus not)1.54 (1.10, 2.16)0.02
 Alcoholic liver disease (versus not)1.27 (1.04, 1.57)0.02
 Viral hepatitis (versus not)1.57 (1.34, 1.83)<0.001
 Cholestatic disease (versus not)0.47 (0.38, 0.58)<0.001
 Other (versus not)0.91 (0.78, 1.07)0.26

Table 4 summarizes a multivariable proportional hazards model correlating pre-OLT characteristics with post-OLT survival among recipients with nonfulminant liver disease. After adjustments for age, gender, diagnosis, and baseline laboratory data, there was no association between African race (HR = 0.85, P = 0.42) or other race (HR = 1.25, P = 0.13) and post-OLT survival. In the same model, compared with recipients with viral hepatitis, recipients with cholestatic disease (HR = 0.39, P < 0.01) and other liver disease categories (HR = 0.69, P < 0.01) had significantly better survival. Finally, we examined whether the lack of an effect of race on post-OLT survival was dependent on time periods. Multivariable models stratifying on transplant centers for patients before and after 1994 were created. In the first era (before 1994), with adjustments for age, gender, and pretransplant laboratory data, a significantly increased risk of mortality was found for patients of other races (HR = 1.60, P < 0.01) in comparison with Caucasians, whereas African race was not associated with mortality (HR = 0.99, P = 0.95). The effect of other race became attenuated once the liver disease diagnosis [HBV and hepatocellular carcinoma (HCC)] was included in the model (HR = 1.39, P = 0.04). In the second era (after 1994), neither African race (HR = 0.47, P = 0.14) nor other race (HR = 1.06, P = 0.89) was associated with recipient survival.

Table 4. Multivariable Proportional Hazards Regression for Survival Among Patients with Nonfulminant Liver Disease**
FactorHazard Ratio (95% Confidence Interval)P
  • *

    Reference: Caucasian.

  • **

    Analysis stratified on center.

  • Laboratory data at the time of transplantation.

  • Reference: viral hepatitis.

 African0.85 (0.57, 1.26)0.42
 Other1.25 (0.94, 1.66)0.13
Age (decades)1.29 (1.20, 1.39)<0.001
Female gender0.97 (0.83, 1.14)0.70
Total bilirubin1.01 (1.003, 1.02)0.01
Creatinine1.02 (0.95, 1.09)0.58
Prothrombin time1.02 (1.00, 1.03)0.08
Diagnosis group  
 Alcoholic liver disease0.85 (0.67, 1.07)0.16
 Cholestatic disease0.39 (0.30, 0.50)<0.001
 Cancer1.23 (0.85, 1.76)0.27
 Other0.68 (0.56, 0.82)<0.001


In this work, we compared the survival of liver transplant recipients among Caucasian, African American, and other races, using data from 4 academic transplant centers. Although there were noticeable differences in baseline characteristics, such as age, gender, and diagnosis, among the 3 groups, we found no relationship between recipient race and survival. This observation held up in our multivariable analyses, which took into account differences in pre-OLT characteristics. These results are in direct contradiction to a report (Nair et al.17) based on the UNOS data, which indicated that African and Asian races were associated with significantly shorter patient and graft survival after OLT in comparison with Caucasian recipients. An examination of our Kaplan-Meier curves and multivariable HRs indicates that the lack of a difference in survival between Caucasian and African races that we observed was not a result of insufficient power, although our sample size of African American patients was indeed smaller (n = 135) than that of Nair et al. (n = 1042). Furthermore, in a post hoc sample size calculation, our sample size had enough power to detect a difference of 10% in survival, which was about the size of the difference in outcome in the UNOS report.

The most immediate difference between the 2 analyses is that the current work is based on patients from select, high-volume academic medical centers, which may be expected to provide better outcomes than the national average.19 For example, data from UNOS have shown that center volume has a significant impact on recipient survival.20 In fact, there is a substantial difference in survival between UNOS patients and ours. The 5-year patient survival in our database was 72% for Caucasian recipients and 69% for African American counterparts, whereas Nair et al.17 reported 5-year survival of 58% for the former and 48% for the latter. Thus, our patients experienced substantially better survival than the national average.

There were intriguing differences in the baseline characteristics between Caucasian and African recipients. African American patients were more likely to have fulminant liver disease, and this was also the case in the UNOS data. Among nonfulminant patients, African American recipients presented with higher MELD labs at the time of the initial evaluation, although they were younger than the patients of other racial groups. This may suggest that African American patients were referred to transplant centers late in the course of their disease progression. The relative abundance of fulminant disease among African Americans may in fact be due to fewer nonfulminant patients being referred to transplant centers. We are not aware of any evidence that African race is associated with higher susceptibility to acute, fulminant liver disease. In our data, fulminant disease was not associated with poorer survival in comparison with nonfulminant disease.

Of the potential mechanisms by which a recipient's race may affect the outcome after OLT, some may have a biological basis. These may include increased tissue incompatibility between donors and recipients of different racial backgrounds12 and genetic differences in the metabolism or effectiveness of immunosuppressant agents.13 Our data suggest that these biological factors are not likely to account for outcome differences. In the best circumstances (that is, selected academic centers), African American patients did just as well as recipients of other races.

Alternatively, race may be a surrogate for socioeconomic factors, such as barriers in access to healthcare or insurance coverage, socioeconomic status affecting the affordability of post-OLT care, and cultural beliefs that may affect the acceptance of organ transplantation or posttransplant care.14 An earlier report from the United Kingdom suggested that the higher incidence of chronic rejection observed among non-Caucasian recipients could not be attributed to lower socioeconomic status or noncompliance.7 A more recent report from the United States observed that although socioeconomic status had some effects on the survival of OLT recipients, race was still an important predictor of survival.18 These studies were limited because the indicators of socioeconomic status included only education level, health insurance, or imputed income based on the postal code of the recipient's residence.21 For our study, one may postulate that African American patients referred to the centers represented here may be systematically different (for example, higher socioeconomic status) than patients represented in the UNOS data. Thus, studies conducted so far have not been adequate to dissect the complex economic, social, and cultural dynamics that aggregate along racial and ethnic boundaries.

Our data seem to point to possible barriers for African American patients to liver transplantation: African American patients with liver disease may be less likely to be referred for OLT, unless a young patient is stricken with a dramatic illness (that is, fulminant liver disease), or their referral may be more likely to be delayed until complications of advanced liver disease develop (that is, high MELD). This interpretation does not necessarily imply a bias or discrimination on the part of the provider, as the delay may be attributed to the payer (or lack thereof) or other socioeconomic circumstances.

It is well established that the pre-OLT recipient status affects the post-OLT outcome. On the one hand, African American patients who present with higher levels of sickness may be expected to experience poorer outcome after OLT. On the other hand, the younger age of African American patients may give them an advantage in having a better outcome. Our multivariable analysis, which included these pre-OLT predictors, showed that neither recipient race nor pre-OLT disease severity had much impact on post-OLT survival. Perhaps the impact of the pre-OLT condition on the post-OLT outcome may be mitigated, at least in part, by the experience and resources of the transplant center. In contrast, if minority patients have barriers in access to these centers and are served by centers less capable of dealing with higher levels of sickness prior to OLT, minority race will have an apparent association with poor outcome. Thus, we postulate that the difference between our data and those of others stems from unequal access to high-quality OLT centers between races. This, however, remains a hypothesis that may be tested in the future: the standard UNOS data files made available to researchers do not include center information. Ultimately, to disentangle this and other complex issues involving race and transplant outcome, a focused investigation is necessary to address issues such as patient preference, referral patterns for OLT in different race groups, and biases, if any, in the healthcare system with respect to referring and evaluating minority patients for OLT.

One brief comment may be made about previous observations that Asian race may also be associated with poor patient survival. A likely explanation for this observation is that there have been more Asian patients with HBV infection and HCC than Caucasian or African patients. For example, the analysis by Nair et al.17 was based on data from 1988-1996, when modern methods for prophylaxis against HBV recurrence were not available and a high incidence of reinfection with HBV and recurrent hepatitis led to graft and patient loss.22–24 In addition, the data preceded the establishment of criteria for OLT in patients with HCC, and this may have led to a higher incidence of recurrent HCC and reduced survival in Asian patients in comparison with others.25 The comparison of pre-1994 and post-1994 data in our analysis further corroborates accumulating literature that post-OLT survival in recent Asian recipients is comparable to that in recipients of other racial categories.26–28

As a proof of principle, our data, although put together from different databases from different time periods, indicate that OLT recipients of a minority race, especially the African race, should not necessarily be expected to have poorer survival than Caucasians. Although recipient race in and of itself may not be a direct predictor of poor outcome after liver transplantation, socioeconomic factors that correlate with minority race may explain previous observations that have shown such an association. Focused investigations to understand and eliminate barriers along socioeconomic and racial boundaries are warranted.