Racial, ethnic, and socioeconomic disparities in the survival of patients with hepatocellular carcinoma (HCC) continue to exist. The authors of this report hypothesized that these differences result from inequities in access to care and in response to therapy.
Patients with HCC (n = 20,920) were identified from the Surveillance, Epidemiology, and End Results (SEER) database, and patients who underwent liver transplantation for HCC (n = 4735) were identified from the United Network for Organ Sharing (UNOS) database. Clinical and pathologic factors were compared after patients were stratified by race and ethnicity.
The survival of patients with HCC improved over time for all racial, ethnic, and income groups (P < .001). Black and low income individuals had the poorest long-term survival (P < .001). On multivariate analysis, black race was predictive of the poorest survival (hazard ratio [HR], 1.15; 95% confidence interval [CI], 1.09-1.22; P < .001), whereas Asian race was associated with the best survival (HR, 0.87; 95% CI, 0.83-0.91; P < .001). After liver transplantation, black patients had the worst graft survival and overall survival (median survival [MS], 30.5 months and 39.7 months, respectively; P < .001), whereas Hispanics had the best survival (MS, 83.4 months and 86.6 months, respectively; P < .001). In a multivariate analysis of transplantation patients, race and ethnicity were associated significantly with outcome.
The incidence of hepatocellular carcinoma (HCC) has more than doubled in the United States over the past 30 years, and there has been a proportionate increase in HCC-related hospitalizations and mortality during the same period.1 The etiology of this trend remains speculative, although several explanations have been proposed. The rising incidence of HCC in the United States may be related in large part to the late complications of chronic hepatitis C virus (HCV) infection, the prevalence of which increased considerably in the 1960s and reached a peak in the 1980s.1, 2 Given the long latency between HCV seroconversion and the development of cirrhosis, further increases in the number of new cases of HCC are expected and may continue for several more years.1-3 Other causes that potentially contribute to increasing HCC incidence include the improved medical management of cirrhosis,1, 4 which allows additional time for HCC to develop, and the increasing prevalence of obesity, diabetes, and associated nonalcoholic steatohepatitis.1
The prognosis for patients with HCC is generally dismal, and the reported 5-year overall survival rates range from 6% to 7%.5, 6 The observed poor outcomes are secondary to the advanced nature of the disease at presentation, which, in most patients, precludes curative surgical intervention.6 Despite the traditionally poor outcomes, there is evidence to suggest a progressive improvement in the survival of patients with HCC in recent years.6 Such improvements are likely related to advances in surgical technique as well as expanding indications for ablation, resection, and orthotopic liver transplantation (OLT), which have facilitated the successful treatment of larger and more advanced tumors.6 Other therapeutic measures, such as liver-directed/regional therapies and newly developed targeted agents, also may have contributed to improved survival,6, 7 although probably to a much lesser degree.
The overall increase in survival notwithstanding, there is good evidence to suggest that improved outcomes in patients with HCC have not been uniform with respect to race, with significantly worse survival reported in African-Americans and superior survival reported in Asian populations.7, 8 To some degree, the reasons for such disparities are related to the differences in etiology of HCC and underlying cirrhosis.9 However, experience with other cancers suggests that differences in healthcare use and access to surgical therapy may be involved, with minority racial and ethnic populations less likely to receive appropriate care.10 Racial and ethnic inequities in the delivery of appropriate treatment for HCC, especially surgery, may explain the observed survival disparities.7, 8
Given these observations, we performed a broad, population-based analysis using 2 major national databases to investigate trends in the survival of patients with HCC over the past 4 decades. In light of recent population-based evidence,6, 11 we hypothesized that this analysis would confirm the consistent improvement in survival over time. Our secondary objectives were to determine how these trends have varied among different racial, ethnic, and socioeconomic groups and whether HCC outcomes may have been affected by disparities in access to appropriate curative-intent surgical intervention.
MATERIALS AND METHODS
Data Sources and Patient Selection
Currently, the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute collects cancer incidence and survival data that encompass approximately 26% of the US population. The SEER limited-use database12 was queried for all adult patients (ie, aged ≥18 years) who were diagnosed with histologically confirmed HCC (according to the International Classification of Disease for Oncology, 3rd edition)13 with malignant behavior between 1973 and 2004.
After the National Transplant Act of 1984, the United Network for Organ Sharing (UNOS) was awarded a contract for the development and implementation of a national transplantation network. All US transplantation centers have functioned under the auspices of UNOS, which has maintained a registry responsible for accumulating scientific and demographic data for all organ recipients throughout the United States.14 Approval to use the UNOS transplantation database was requested and granted. A limited patient database was supplied by UNOS14 and was queried for all adult patients who underwent OLT for histologically confirmed HCC between October 1987 and February 2008.
Overall Survival Analysis
The primary outcome measure of interest for SEER data was overall survival. The prognostic factors of interest were year of diagnosis, race, and household income. Year of diagnosis was categorized by decade (ie, 1970s, 1980s, 1990s, and 2000s). Race and ethnicity were categorized as white, black, Hispanic, Asian/Pacific Islander, and other. By convention, the Hispanic category included only nonblack Hispanics. Those individuals coded as “black” by the race-ethnicity variable who were of Hispanic origin were categorized as black. Household income was categorized as low, middle, and high with cutoff points at the 33rd and 67th percentiles of income for all individuals. Income figures in the SEER database represent assigned values equivalent to the median household income of the individuals' county of residence based on US Census Bureau data. For consistency, the SEER Year 2000 median household income county-attribute variable was used in this study.12 Therefore, the figures used in this report represent approximations of income at the census-tract level rather than true measures of individual household income. Other data that were included in the analysis were age at diagnosis, sex, extent of disease, and type of therapy (including surgery). Extent of disease was grouped into metastatic and nonmetastatic disease; and type of therapy was grouped into no surgery, ablation/tissue destruction, and resection/transplantation. Resection was not differentiated from transplantation in the SEER database until 1998. Therefore, those 2 interventions were combined in this initial broad analysis.
Survival Analysis by Time, Race, Ethnicity, and Income
Differences in survival by decade, race, ethnicity, and income were assessed using the Kaplan-Meier method. A stratified subset analysis was performed separately for each variable of interest to determine whether similar trends in survival were observed across all subgroups. Three separate stratified analyses were performed. To determine whether the association between survival and time was similar for all subgroups, Kaplan-Meier curves comparing survival by decade were stratified by race, ethnicity, income, extent of disease, and type of therapy. To determine whether the associations between race/ethnicity and survival were similar in all subgroups, Kaplan-Meier curves comparing race and ethnicity were stratified by decade, extent of disease, and type of therapy. Finally, to determine whether the association between income and survival was similar in all subgroups, the income analysis was stratified similarly by decade, extent of disease, and type of therapy.
Next, a multivariate Cox regression analysis was performed to determine the significance of time period, race/ethnicity, and income adjusted for all other factors. The results were reported as hazard ratios (HRs) with 95% confidence intervals (CIs). P values ≤.05 were considered statistically significant.
Surgical Subgroup Analysis
SEER surgical coding differentiated OLT from other surgical procedures for HCC beginning in 1998. Therefore, the subset of patients diagnosed with HCC after 1998 was analyzed separately. This additional analysis was performed primarily to determine whether resection and transplantation rates varied by race and ethnicity in patients with nonmetastatic disease and to determine whether associations between race/ethnicity and survival were influenced by type of therapy, specifically resection and OLT. Type of intervention was categorized into 4 groups: no surgery, ablation/tissue destruction, resection, and transplantation. The rates of each intervention were determined for each race and ethnicity group and were compared using the chi-square test. To determine surgical and transplantation rates in a group for which surgery was indicated, the rates of intervention were recalculated for individuals who had solitary lesions that measured ≤5 cm in greatest dimension. This group represents a subset of patients that would have met Milan criteria.15 Differences in survival by intervention in all patients with nonmetastatic disease after 1998 were determined using the Kaplan-Meier method.
United Network for Organ Sharing: Transplantation Population Analysis
The outcome measures of interest were graft survival and overall survival as coded by UNOS. The primary prognostic factor of interest was race/ethnicity, which was categorized as white, black, Hispanic, and Asian. Additional clinicopathologic data included age, body mass index (BMI), transplantation Model for End-Stage Liver Disease (MELD) scores,16 HCV serostatus, hepatitis B virus (HBV) antigen positivity, the presence of diabetes mellitus, dialysis requirement before transplantation, education level, Medicaid status, number of tumors, combined tumor size, and total length of stay. Differences in these clinical factors were compared by race and ethnicity. A multivariate Cox regression analysis was performed to determine the association between survival and race/ethnicity controlling for potential confounding clinical factors. It is noteworthy that, because tumor size was not available for a significant number of patients and because MELD scores were used only during the last decade, the multivariate analysis initially did not include these variables, and a second multivariate analysis was performed to include tumor size and MELD score at transplantation.
The characteristics of the SEER HCC population are presented in Table 1. The mean age (±standard deviation) was 63.6 ± 13.0 years, and most patients were men (73%). Although only a minority of patients had metastatic disease (18%), most patients (72%) did not undergo surgical intervention. Most patients were white (54%), and the next largest group was Asian/Pacific Islanders (22%), followed by Hispanics (12%), and blacks (12%).
Table 1. Characteristics of the Surveillance, Epidemiology, and End Results Population of Patients With Hepatocellular Carcinoma From 1973 to 2004 (N=20,920)
There was a statistically significant improvement in survival over time in patients with HCC (Fig. 1) with a successive increase observed by decade (the median survival [MS] was 2 months in the 1970s, 2 months in the 1980s, 4 months in the 1990s, and 7 months in the 2000s; P < .001 for linear trend). Comparisons of survival by time after stratification by race/ethnicity and income also were performed (results not shown). A statistically significant linear increase in survival by time was noted in all racial/ethnic and income groups (all P values <.001). When stratified by the presence of metastatic disease and type of intervention (cancer-directed surgery vs no cancer-directed surgery), the same progressive increase was observed in all subgroups, but it was apparent that most of the increase in survival over time occurred in patients who had nonmetastatic disease and in those who underwent surgical treatment (results not shown).
Survival Analysis by Race/Ethnicity
There were statistically significant differences in outcome by race/ethnicity within the SEER HCC cohort (Fig. 2), and MS was lowest in black patients compared with all other groups (MS: 3 months for blacks vs 4 months for whites, 5 months for Hispanics, and 6 months for Asians; all pairwise P values <.001). The analysis was stratified by extent of disease and type of therapy (Fig. 3). Black patients demonstrated the worst survival compared with every other race/ethnicity in all subgroups—metastatic disease, nonmetastatic disease, and surgical and nonsurgical disease (all P values <.001). This disparity was most pronounced in patients who had nonmetastatic disease and in patients who underwent a cancer-directed surgical intervention. In patients who had surgical disease (ie, resection/transplantation), black race was associated with significantly poorer survival on pairwise comparison with all other races/ethnicities (MS, 23 months; P < .001 for all comparisons vs black race). However, no statistically significant differences in survival were observed between Asians, whites, and Hispanics.
Survival Analysis by Income
Similar to race and ethnicity, there were statistically significant differences in survival by income (Fig. 4), with the low-income group demonstrating the poorest MS compared with the middle-income and high-income groups (MS: 3 months, 4 months, and 5 months, respectively; all pairwise P values <.001). Stratified by extent of disease and type of treatment, survival disparities in income groups were observed predominantly in patients who had nonmetastatic disease (MS: 5 months vs 8 months for the low-income vs middle-income groups, respectively; P < .001) and patients who underwent surgical intervention (MS: 29 months vs 36 months for the low-income vs middle-income groups, respectively; P = .002), and little disparity was noted among patients who had metastatic or nonsurgical disease (results not shown).
Surveillance, Epidemiology, and End Results: Multivariate Analysis
The results of multivariate Cox regression analyses are presented in Table 2. Time period, race/ethnicity, income, age, sex, type of intervention, extent of disease, and grade all were associated independently with survival. Specifically, there was a progressive increase in survival (ie, a decreased risk of death) with each successive decade (Table 2). Black patients (HR, 1.15; 95% CI, 1.09-1.22; P < .001) had significantly worse survival compared with white patients, whereas Asian race was an independent predictor of improved survival (HR, 0.87; 95% CI, 0.83-0.91; P < .001). With respect to income, the middle-income group (HR, 0.89; 95% CI, 0.85-0.93; P < .001) and the high-income group (HR, 0.95; 95% CI, 0.91-0.99; P = .029) had significantly better survival compared with patients in the low-income group, and there was no difference in survival between the middle-income and high-income groups (P value not significant; data not shown). Patients who underwent ablation (HR, 0.40; 95% CI, 0.36-0.44; P < .001) fared better than patients who had no cancer-directed intervention. As expected, patients who underwent resection/transplantation (HR, 0.27; 95% CI, 0.25-0.28; P < .001) had the best overall outcome.
Table 2. Multivariate Analysis of Patients Diagnosed With Hepatocellular Carcinoma From 1973 to 2004 (N=14,906)
To determine whether differences in outcome by race and ethnicity were related to the type of hepatic intervention, we re-examined SEER data for the period from 1998 to 2004, when OLT was coded separately from resection. Patients with metastatic disease were excluded from this analysis. Rates of intervention by race and ethnicity within this cohort were assessed (Table 3). There was a statistically significant difference in surgery rates by race and ethnicity, with African Americans demonstrating the lowest rate of transplantation and of combined resection/transplantation (7% and 22%, respectively; P < .001) (Table 3). In addition, both blacks and Hispanics had significantly higher rates receiving no hepatic intervention (67% and 69%, respectively) compared with Asians/Pacific Islanders and whites (55% and 62%, respectively).
Table 3. The Rate of Intervention by Race/Ethnicity for Patients With Nonmetastatic Disease From 1998 to 2004 (valid n=8277) and for Patients Who Had Solitary Lesions ≤5 Centimeters With Nonmetastatic Disease From 1998 to 2004 (N=1618)
No. of Patients (%)
PI indicates Pacific Islander.
Patients with nonmetastatic disease
Patients with solitary lesions
We compared types of intervention by race/ethnicity in patients who had solitary lesions that measured ≤5 cm in greatest dimension, which represented a group of patients who potentially were eligible for OLT (Table 3). Black patients had the lowest rate of resection and transplantation combined and had a significantly lower rate of transplantation than Hispanic patients and white patients. Asian patients had the lowest rate of transplantation but the highest rate of surgery (resection and/or transplantation). We compared survival by type of intervention in all patients with nonmetastatic disease (Fig. 5). There were statistically significant differences in survival by intervention; patients who underwent transplantation had the best survival (MS not reached) followed by those who underwent resection (MS, 34 months), ablation (MS, 23 months), and no invasive therapy (MS, 5 months; all pairwise P values < .001).
United Network for Organ Sharing
Of the 88,973 liver transplantations, 4735 were performed for HCC. The mean age (±standard deviation) of patients who underwent transplantation for HCC was 55.4 ± 8.8 years. Most patients were whites (67%), followed by Hispanics (14%), Asians (10%), and blacks (7%). Approximately half of the cohort (48%) were seropositive for HCV, and 10% were positive for HBV surface antigen. Ten percent of patients had diabetes mellitus, but only 3% of the entire cohort required dialysis before transplantation. Thirty-one percent of patients had a college level or higher education, and 35% had a high school education or less. A minority of patients (11%) required Medicaid to pay for transplantation costs.
In a comparison of descriptive characteristics by race and ethnicity (Table 4), black patients were significantly younger (P < .001). Black patients also had higher rates of HCV positivity, dialysis requirement, and Medicaid use and a lower rate of college education than all other groups (P < .001 for all). Hispanics had a higher rate of diabetes mellitus than all other groups (P < .05 for all), and Asians the highest rate of positivity for HBV surface antigen (50%).
Table 4. Comparison of Clinicopathologic Factors by Race/Ethnicity in The United Network for Organ Sharing Hepatocellular Carcinoma Transplantation Cohort (N=4735)
No. of Patients (%)
SE indicates standard error of the mean; BMI, body mass index; HBV, hepatitis B virus; HCV, hepatitis C virus; MELD, model for end-stage liver disease.
Mean age ± SE, y
55.7 ± 0.15
52.7 ± 0.53
55.5 ± 0.32
55.4 ± 0.13
Mean ± SEM BMI, kg/m2
28.2 ± 0.09
27.6 ± 0.31
28.7 ± 0.19
24.4 ± 0.18
Mean ± SEM MELD score at transplantation
23.8 ± 0.11
24.5 ± 0.32
25.3 ± 0.24
25.0 ± 0.30
No. of tumors
Mean ± SE combined greatest tumor dimension, cm
3.39 ± 0.04
3.38 ± 0.11
3.42 ± 0.09
3.38 ± 0.10
Mean ± SEM length of stay, d
14.8 ± 0.32
14.5 ± 0.79
14.4 ± 0.60
15.2 ± 0.84
Kaplan-Meier analysis revealed that black patients who underwent transplantation for HCC had significantly worse graft survival compared with the other groups (median graft survival: 30.5 months for blacks vs 69.7 months for Asians, 63.8 months for whites, and 82.4 months for Hispanics; P < .001). It is noteworthy that graft survival in black patients who underwent OLT was less than half that of all other groups. Hispanics demonstrated the best graft survival, and Asians and whites demonstrated intermediate and near-equivalent graft survival. With respect to overall survival, again, black patients fared the worst with an MS that was less than half that of Hispanics, which was the group that had the best outcome (MS: 39.7 months for blacks vs 72.4 months for Asians, 74.7 months for whites, and 86.6 months for Hispanics; P < .001). Asians/Pacific Islanders and whites again demonstrated intermediate and near-equivalent overall survival (Fig. 6).
In multivariate analysis, black race was an independent predictor of both decreased graft survival and decreased overall survival when controlling for age, sex, BMI, diabetes status, dialysis requirement, and HBV and HCV serostatus (HR, 1.63; 95% CI, 1.29-2.04 [P < .001] and HR, 1.66; 95% CI, 1.29-2.12 [P < .001], respectively) (Table 5). After adding combined tumor size and MELD score, black race remained a significant predictor of poorer graft survival and overall survival (HR, 1.67; 95% CI, 1.17-2.36 and HR, 1.60; 95% CI, 1.10-2.34, respectively; P < .05) (table not shown). This second multivariate analysis was representative of the post-MELD era of OLT.
Table 5. Multivariate Cox Regression Analysis of Graft and Overall Survival (N=3017)
HR indicates hazard ratio; CI, confidence interval; SE, standard error; BMI, body mass index; HBV, hepatitis B virus; HBV, hepatitis C virus.
Mean ± SE age, y
HBV surface antigen
The increasing use of major hepatic resection and OLT for HCC, with concomitant decreases in procedural morbidity and mortality, have afforded effective and often curative therapy for individuals with previously untreatable disease.15, 17-21 In the current study, we demonstrated that the survival of patients with HCC has improved over the past 31 years. This improvement in survival was most notable for the subset of patients who had localized disease and in patients who underwent surgery with curative intent. A similar temporal improvement in survival was noted by El-Serag et al,6 who suggested that a significant portion of the improvement over time may have been caused by lead-time bias related to earlier diagnosis of disease. The improved outcomes observed in surgical cohorts, however, likely reflect true increases in survival related to advances in surgical technique and more widespread implementation of OLT.6 We also demonstrated that the same pattern of improved survival over time has occurred in all racial/ethnic and income groups, an indication that all groups have benefited to some degree from advances in screening, diagnosis, and treatment.
Despite these overall trends, our study demonstrates that black patients and lower income patients continue to have the worst survival. Racial inequities in outcome have been observed previously among patients with HCC and other cancers.7, 8 The reasons for this disparity are unclear but may be related both to differences in underlying disease and to inconsistencies in access to appropriate care.7 Both Davila and El-Serag7 and Sloane et al8 demonstrated differences in the receipt of therapy by race and reported that black populations were much less likely to receive appropriate surgical treatment for localized disease. In multivariate analysis, those authors demonstrated that race was less predictive of survival when the receipt of therapy was taken into account and that black patients no longer had a survival deficit. Our results demonstrate, however, that black patients continue to fare worse even after adjusting for type of therapy on multivariate analysis.
To further examine the issue of appropriate treatment use, we looked more closely at the subset of patients with nonmetastatic disease in the SEER database who were diagnosed after 1998. This group of patients represents individuals who, based on the absence of distant disease, may have been eligible for surgery. Black patients again had the lowest rate of surgery and the lowest rate of transplantation, which is the treatment modality associated with the best chance for long-term survival. Examining the patients who had solitary lesions that measured ≤5 cm, a subset of patients that met Milan criteria, black patients had the lowest rate of surgery compared with all other patient groups and had the lowest rate of transplantation excluding Asian patients.9, 22 These finding underscore the presence of racial/ethnic inequity in the administration of curative surgical options, specifically, liver transplantation.
The suggestion from our initial findings was that racial and ethnic disparities in survival may be related primarily to inequities in the administration of appropriate treatment. The assumption that follows is that equivalent treatment should result in equivalent survival. To test this assertion, we compared outcomes by race and ethnicity in patients who underwent transplantation for HCC using the UNOS database. Although black patients generally are under represented on transplantation waiting lists,23, 24 the subset of patients in the UNOS database who undergo transplantation represents a particularly homogenous population given the ranking methods used for patient listing, particularly in the post-MELD era. This point was reported by Moylan et al,25 who noted that there was no correlation between race and the receipt of liver transplantation in the era of MELD scores.
Our study demonstrates that, after OLT for HCC, significant racial and ethnic disparities in both graft survival and overall survival persist, with black patients again demonstrating a significant survival deficit compared with all other groups. Because this finding also was noted when the MELD score was incorporated into the multivariate model—which limited the sample only to those patients who had a MELD score recorded—significant racial and ethnic disparities were present even in the MELD era, a period during which differences in access to OLT based on race and ethnicity are less prevalent.25 Therefore, our study demonstrates that survival disparities by race and ethnicity cannot be explained by access issues alone, and other factors need to be considered.
The clinical basis of our findings is unclear. Previous studies that have examined racial and ethnic differences in survival for all patients who underwent transplantation, including those with nonmalignant disease, also demonstrated worse post-transplantation outcomes in black patients26, 27 but reported no clear underlying reason for such disparity. Potential causes include the increased prevalence of HCV-related cirrhosis in black patients, which has been associated with worse post-OLT outcomes,28 as well as increased post-OLT recurrence of HCV, resulting in a poorer response to subsequent medical therapy.29 Other potential explanations include differences in response to immunosuppression,30 severity of end-stage liver disease not accounted for by the MELD score, and differences in comorbidity.
Our results should be interpreted in light of several limitations. The SEER database lacks specific information related to comorbidity, and this information is incomplete in the UNOS database. In addition, we could not directly address the issues of access to care. Although it is possible to determine the specific type of treatment rendered adjusted for extent of disease, the reasons for differences in intervention are difficult to ascertain. Furthermore, we could not assess the compliance with postoperative care or response to immunosuppression. Finally, we were unable to assess biologic factors that may have affected survival. Despite these limitations, our study is unique to the literature in that, in addition to demonstrating racial disparity, we used a homogenous population with equivalent treatment to demonstrate that racial outcome differences may be caused in part by factors other than disparities in treatment and healthcare use.
In conclusion, we have demonstrated that the survival of patients with HCC has improved over time, with the greatest improvement noted in patients who have localized and surgical disease, probably secondary to advancements in surgical techniques and the more widespread use of transplantation. We also have demonstrated that all racial/ethnic and income groups have experienced a proportional increase in survival, although significant socioeconomic and racial/ethnic disparities in survival remain. These may be related to issues like the use of surgery and access to appropriate care. By using a transplantation database for all patients with OLT in the United States we have demonstrated that, even with appropriate and equivalent treatment, racial and ethnic inequities in survival persist that are unexplained by access to care. The cause of these inequities remains uncertain, but the implications are significant. On the basis of our results, interventions aimed at eliminating such disparity cannot be focused solely on equalizing the use of transplantation but also may need to address post-treatment follow-up and care. Prospective, population-based studies are needed to resolve these questions.