During the past 2 decades, an increase in the ageadjusted

Transcription

1 CLINICAL GASTROENTEROLOGY AND HEPATOLOGY 2006;4: Racial Differences in Survival of Hepatocellular Carcinoma in the United States: A Population-Based Study JESSICA A. DAVILA* and HASHEM B. EL SERAG*, *Section of Health Services Research and Section of Gastroenterology, The Houston Center for Quality of Care and Utilization Studies, Michael E. DeBakey Veterans Affairs Medical Center and Baylor College of Medicine, Houston, Texas Background & Aims: Survival after hepatocellular carcinoma (HCC) diagnosis is generally dismal, but there are patients with more favorable outcomes. Racial variation in survival of patients with HCC could be associated with observed differences in survival; however this has not been previously examined. Methods: During , HCC patients were identified from 9 Surveillance, Epidemiology, and End-Results registries. One- and 3-year survival rates were calculated and compared by race. Models were constructed to examine the effects of race on the mortality risk. Results: Asians had the highest 1- and 3-year observed and relative survival, followed by whites, Hispanics, and blacks. Compared with whites, Asians (odds ratio [OR], 1.37; 95% confidence interval [CI], ) were more likely to receive local or surgical therapy, whereas blacks (OR, 0.62; 95% CI, ) and Hispanics (OR, 0.81; 95% CI, ) were less likely to receive therapy. Adjusting for differences in receipt of therapy, stage of HCC, year of diagnosis, and other demographics, Asians (hazard rate [HR], 0.84; 95% CI, ) maintained a lower mortality risk compared with whites. In adjusted models, Hispanics (HR, 1.13; 95% CI, ) maintained a higher mortality risk, whereas the mortality risk for blacks became nonsignificant different from whites (HR, 1.06; 95% CI, ). Last, a 22% improvement in survival was observed between and , which was mostly explained by increased receipt of local or surgical therapy. Conclusions: We observed significant racial variation in survival. These variations in survival are partly explained by a lower likelihood of receipt of therapy and more advanced HCC at diagnosis among blacks and Hispanics. During the past 2 decades, an increase in the ageadjusted incidence of hepatocellular carcinoma (HCC) has been reported. The yearly age-adjusted incidence rates have increased from 1.4 per 100,000 in to 3.0 per 100,000 in However, the overall survival for patients diagnosed with HCC is generally dismal (1- and 5-year survival rates of 20% and 6%, respectively). 2 Nevertheless, for patients diagnosed with a small tumor and no distant spread, applying potentially curative therapy such as surgical resection, liver transplantation, or local ablation significantly prolongs survival (40% 50% at 5 years). 2,3 Although there is striking racial variation in the incidence of HCC, the association between race and survival is not known. One previous study found no significant differences in the crude survival rates between whites and blacks in the US. 2 However, the primary aim of that study was not to examine racial variation in survival, and no adjustments were made for potential confounding factors including age, stage of HCC at diagnosis, and receipt of therapy. Moreover, survival among other racial groups at high risk of developing HCC, including Asians and Hispanics, has not been examined. Therefore, additional studies that specifically examine racial differences in survival are warranted. With data from the population-based cancer registries of the Surveillance, Epidemiology, and End-Results (SEER) program, we conducted this study to determine whether racial/ethnic variation exists in survival after HCC diagnosis and to examine and adjust for other demographic and tumor features that might affect survival. Methods Data Source Beginning in 1973, the SEER program was established to identify all new cancer cases diagnosed within 7 geographic areas. By 1975, SEER included 9 geographic regions, 5 states (Connecticut, Hawaii, Iowa, New Mexico, and Utah) and 4 metropolitan areas (San Francisco-Oakland, Seattle-Puget Sound, Detroit, and Atlanta), representing approximately 10% of the US population. Within each SEER region, medical record abstractors routinely review inpatient and outpatient clinic records to identify incident cancer cases. For each case identified, the SEER program collects demographic features, date of cancer diagnosis, Abbreviations used in this paper: ARP, attributable risk percent; CI, confidence interval; HCC, hepatocellular carcinoma; HR, hazard rate; OR, odds ratio; SEER, Surveillance, Epidemiology, and End-Results Program by the American Gastroenterological Association /06/$32.00 PII: /S (05)

2 January 2006 RACIAL DIFFERENCES IN SURVIVAL OF HCC IN THE US 105 Table 1. Demographic and Tumor Features Compared by Race/Ethnicity Among Patients With HCC between in 9 SEER Registries (N 9919) White (n 5426) Black (n 1278) Asian a (n 1573) Hispanic (n 746) Other b (n 896) P value Time period of diagnosis (n 2418) 1352 (55.9) 318 (13.2) 426 (17.6) 149 (6.2) 173 (7.2) (n 3214) 1765 (54.9) 389 (12.1) 521 (16.2) 235 (7.3) 304 (9.5) (n 4287) 2309 (53.9) 571 (13.3) 626 (14.6) 362 (8.4) 419 (9.8) Age (y) Mean (standard deviation) 66.9 (12.5) 60.7 (13.0) 64.3 (13.2) 62.4 (13.1) 59.0 (12.9) (2.01) 56 (4.4) 75 (4.8) 21 (2.8) 60 (6.7) (23.1) 533 (41.7) 418 (26.5) 293 (39.3) 388 (43.3) (74.9) 689 (53.9) 1080 (68.7) 432 (57.9) 448 (50.0) Gender Men 3961 (73.0) 950 (74.3) 1172 (74.5) 571 (76.5) 651 (72.7).9072 SEER registry San Francisco-Oakland 872 (16.1) 300 (23.5) 830 (52.8) 260 (34.9) 230 (25.7).0001 Connecticut 1015 (18.7) 123 (9.6) 17 (1.1) 68 (9.1) 15 (1.7) Detroit-Metropolitan 967 (17.8) 540 (42.3) 30 (1.9) 25 (3.4) 30 (3.4) Hawaii 130 (2.4) 3 (0.2) 527 (33.5) 1 (0.1) 235 (26.2) Iowa 679 (12.5) 23 (1.8) 6 (0.4) 15 (2.0) 25 (2.8) New Mexico 244 (4.5) 5 (0.4) 9 (0.6) 292 (39.1) 74 (8.3) Seattle-Puget Sound 924 (17.0) 84 (6.6) 128 (8.1) 43 (5.8) 203 (22.7) Utah 249 (4.6) 2 (0.2) 9 (0.6) 26 (3.5) 25 (2.8) Atlanta-Metropolitan 346 (6.4) 198 (15.5) 17 (1.1) 16 (2.1) 59 (6.6) Stage at diagnosis Localized 1694 (31.2) 376 (29.4) 462 (29.4) 266 (35.6) 298 (33.3).0001 Regional spread 1317 (24.3) 357 (27.9) 518 (32.9) 173 (23.1) 278 (31.0) Distant spread 1142 (21.0) 315 (24.7) 336 (21.4) 151 (20.2) 179 (20.0) Unstaged 1273 (23.5) 230 (18.0) 257 (16.3) 518 (21.1) 141 (15.7) Receipt of therapy None 4558 (84.0) 1115 (87.3) 1251 (79.4) 654 (87.6) 734 (81.8).0001 Local c 96 (1.8) 18 (1.4) 36 (2.4) 16 (2.1) 22 (2.4) Surgical d 569 (10.5) 90 (7.0) 214 (13.6) 55 (7.4) 116 (13.1) Unknown 203 (3.7) 55 (4.3) 72 (4.6) 22 (2.9) 24 (2.7) a Asian comprises Chinese, Japanese, and Filipino. b Other comprises American Indian, Hawaiian, and unspecified. c Photodynamic therapy, electocautery, cryosurgery, laser, alcohol, heat. d Resection or transplantation. cancer site, method of diagnosis (histology, cytology, microscopic confirmation with an unspecified method, laboratory test/marker study, direct visualization, positive radiology test results, clinical diagnosis only, or unknown method of confirmation), stage of disease at diagnosis, therapy, and follow-up status from the medical record. Data for this study were obtained from SEER-STAT public-use data files, available on CD-ROM from the National Cancer Institute. 4 Demographic and cancer-related information included in the SEER database are obtained by medical record review. For race, abstractors categorize patients into one of a total of 28 racial groups on the basis of information contained in the medical record. Ethnicity is collected as a separate variable indicating Spanish ethnicity or origin. The exact source of race/ethnicity information in the medical record is not specified in the registry database. Studies are conducted annually at each SEER registry site to verify that data are being collected accurately, and that case ascertainment is 98% or greater. Cancers are coded according to the second International Classification of Disease-Oncology. 5 Study Population We used second International Classification of Disease- Oncology code 8170 to identify patients with HCC among all patients with primary liver cancer diagnosed during Subsequently, only cases with some diagnostic confirmation of HCC were included in our study cohort (94.9% of total). Diagnostic confirmation was defined as having positive histology (63.1%), cytology (20.3%), laboratory test/marker study (3.0%), direct visualization (0.3%), or positive radiology test results (13.3%). Patients with clinical diagnosis only or unknown method of confirmation were excluded (5.1% of total). Patients with unknown race/ethnicity ( 1%) were also excluded. Statistical Analysis We compared patients in 5 categories of race/ethnicity (white, black, Asian, Hispanic, other) for time period of HCC diagnosis ( , , ), demographic characteristics (age, gender), SEER registry, stage of HCC at diagnosis, and receipt of therapy (Table 1). Asian race comprised Chinese, Japanese, and Filipino, whereas other race

3 106 DAVILA AND EL-SERAG CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 1 comprised American Indian, Hawaiian, and other nonspecified races. Stage of disease at diagnosis was categorized by SEER as localized, regional spread, distant spread, or unknown. Receipt of therapy included both local and surgical therapy. Local therapy was defined by SEER treatment categories of photodynamic therapy, electrocautery, cryosurgery, laser, alcohol, and heat, whereas surgical therapy was defined by resection or transplant. Chi-square tests were conducted for dichotomous variables and Mann-Whitney U tests for continuous variables. One- and 3-year observed and relative survival and their 95% confidence intervals (CIs) were calculated. The observed survival rate was equal to the number of patients remaining alive for a specific length of time divided by the total number of patients at risk of death during that specified time period. The relative survival, which is an alternative to calculating cancer-specific mortality, was calculated as the ratio of the observed survival to the expected survival for a group of people in a general population that is similar to that of the patient group with respect to race, sex, age, and calendar period of observation. Expected survival rates are available from the US census bureau. Countyspecific population estimates were used to determine expected survival rates for each SEER region. The time to death after date of HCC diagnosis was also modeled as the outcome variable in a Cox proportional hazards survival analysis that examined the effect of race/ethnicity (white, black, Asian, Hispanic, other) on the mortality risk while adjusting for year of diagnosis, age ( 40, 40 59, 60 years), gender (women, men), SEER registry (San Francisco- Oakland, Connecticut, Detroit-Metropolitan, Hawaii, Iowa, New Mexico, Seattle-Puget Sound, Utah, Atlanta-Metropolitan), stage of HCC (localized, regional spread, distant spread, unknown), and receipt of therapy (none, local, surgical). The follow-up period for these analyses began at the date of HCC diagnosis and ended in December 2001 or at time of death. Hazard rates (HRs) and 95% CIs were calculated for each variable. The log-log survival plots were used to examine the proportional hazards assumption, which was met in all models. A multivariable logistic regression model was developed to examine the association between stage of HCC and race/ethnicity, adjusting for age, gender, and geographic region. In addition, another multivariable logistic regression model was constructed to examine the association between race/ethnicity and receipt of local or surgical therapy (vs no therapy), adjusting for age, gender, geographic region, and stage of HCC. Wald 2 tests were used to determine the significance of each variable. Adjusted odds ratios (ORs) and 95% CIs were calculated for each variable. In addition, the attributable risk percent (ARP) was calculated individually for blacks, Asians, and Hispanics compared with whites. The ARP is the percent difference in risk between the exposed and the unexposed groups. For these analyses, data from the SEER public-use CD-ROM were converted into SAS data sets for further analyses (SAS version 8.2; SAS Institute, Cary, NC). Results During , we identified a total of 9919 patients with HCC. Among the 5 racial/ethnic groups, there were significant differences in the time period of diagnosis, age at diagnosis, SEER registry, stage of HCC, and receipt of therapy (all P values.0001). The proportions of Asians diagnosed with HCC declined over time, whereas the proportions of Hispanic and other race groups increased during more recent years. No significant change in the proportions of whites and blacks diagnosed with HCC was observed over time. The greatest proportion of patients age 60 years and older was among whites (74.9%), followed by Asians (68.7%), Hispanics (57.9%), blacks (53.9%), and patients of other race (50.0%). The greatest proportion of whites was observed in Connecticut, blacks in Detroit-Metropolitan, Asians in San Francisco-Oakland, Hispanics in New Mexico, and patients of other race in Hawaii. Patients of other race (64.3%) and Asians (62.3%) had the largest proportion of patients diagnosed with localized or regional spread disease, followed by Hispanics (58.7%), blacks (57.3%), and whites (55.5%). Asians were most likely to receive local ablation or surgical therapy (13.6%), followed by patients of other race (13.1%), whites (10.5%), Hispanics (7.4%), and blacks (7.0%). No significant differences in the proportions of men were observed by race; approximately 75% of all patients were men. We also examined the proportion of patients who received therapy among those considered most favorable (single lesion, 3 cm) to receiving therapy. Among the 356 patients who met these criteria, 53.4% (n 190) received local ablation or surgical therapy. Of these, 84% (n 160) received surgical therapy. Among racial/ethnic groups, Asians had the greatest proportion receiving local ablation or surgical therapy (60.4%), followed by Hispanics (59.3%), whites (53.3%), patients of other race (51.2%), and blacks had the lowest proportion (44.2%). Among those patients who received any therapy, Asians were also most likely to receive surgical therapy (50.0%), followed by Hispanics (48.2%), whites (46.2%), patients of other race (41.9%), and blacks were the least likely (34.9%). The cumulative observed survival for patients categorized by race is presented in Figure 1. Complete follow-up information was available for greater than 98% of patients included in this study (n 9735). Asians and patients of other race had significantly longer overall observed survival compared with whites. Blacks and Hispanics had the shortest survival of all race groups. These differences were most prominent during the first year after HCC diagnosis.

4 January 2006 RACIAL DIFFERENCES IN SURVIVAL OF HCC IN THE US 107 Table 2. Results From the Cox Proportional Hazards Model Examining Mortality Risk and Race/Ethnicity Through the End of 2001 Predictor variable HR 95% CI P value Figure 1. The cumulative observed survival within the first 3 years after HCC diagnosis categorized by race/ethnicity. Data are shown for patients with complete follow-up information diagnosed in 9 SEER registries during (n 9735). Both the unadjusted observed and relative survivals were similar for each race group, thus indicating that the majority of deaths among these patients were due to HCC rather than other causes. The 1-year unadjusted relative survival among Asians was significantly 10% (31.5% vs 21.4%) higher as compared with blacks and 7% (31.5% vs 24.4%) higher compared with whites. However, there were no significant differences between Asians and Hispanics or patients of other race (Figure 2). Similarly, the 3-year unadjusted relative survival was approximately 5% (12.9% vs 7.9%) higher among Asians compared with blacks; however, this difference was not statistically significant. The unadjusted Cox proportional hazards models showed several significant differences in mortality risk on Race White 1.00 Reference Black Asian Hispanic Other Time period of cancer diagnosis Reference Age at diagnosis (y) Reference Gender Male 1.00 Reference Female SEER registry Connecticut 1.00 Reference San Francisco Oakland Seattle Detroit Iowa New Mexico Hawaii Utah Atlanta Stage at diagnosis Distant spread 1.00 Reference Regional spread Localized Unknown Receipt of therapy None 1.00 Reference Local Surgical The model adjusts for time period of diagnosis, age, gender, SEER registry, stage of disease, and receipt of therapy among patients with HCC during This analysis includes only patients with complete follow-up information (N 9735). Figure 2. The 1- and 3-year relative survival in 5 race/ethnic groups during Relative survival is calculated as the ratio of the observed survival to the expected survival for a group of people in a general population that is similar to that of the patient group with respect to race, sex, age, and calendar period of observation. the basis of race. Compared with whites, the mortality risk was 16% lower among Asians (unadjusted HR, 0.84; 95% CI, ) and persons of other race (unadjusted HR, 0.84; 95% CI, ), whereas blacks and Hispanics had an 11% (unadjusted HR, 1.11; 95% CI, ) and 9% (unadjusted HR, 1.09; 95% CI, ), respectively, higher mortality risk compared with whites. When adjusting for year of cancer diagnosis, age, gender, SEER registry, stage of HCC, and receipt of therapy (Table 2), Asians continued to have a 16% lower mortality risk, whereas patients of other race showed a 9% reduction in risk when receipt of therapy and age at diagnosis were added to the model. Compared with whites, the ARPs among Asians and patients of other race were 19.1% and 9.9%, respectively. On

5 108 DAVILA AND EL-SERAG CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 1 Figure 3. Changes in the proportion of patients with HCC who received any therapy (local or surgical) during 3 time periods ( , , ) by race ). No significant differences by gender were observed. An increase in the receipt of local or surgical therapy over time was observed among all race/ethnic groups (Figure 3). Between and , the proportion of patients receiving therapy doubled among Asians (10.1% to 21.1%) and approximately tripled among blacks (4.1% to 12.3%) and patients of other race (8.7% to 22.9%). A logistic regression model was developed to evaluate the association between receipt of local or surgical therapy and race/ethnicity. Compared with whites, Asians and patients of other race were 36% and 32%, respectively, more likely to receive therapy, whereas blacks and Hispanics were 34% and 25% less likely to receive therapy, respectively. Adjusting for time of diagnosis, age, gender, SEER registry, and stage of HCC, the likelihood of receiving therapy continued to be significantly higher among Asians and significantly the other hand, differences between blacks and whites became nonsignificant when receipt of therapy was adjusted for in the model (Table 2). The ARP among blacks compared with whites was 5.6%. Last, the difference between Hispanics and whites increased to 13% when stage of disease was adjusted for in the model. The ARP among Hispanics compared with whites was 11.5%. Other variables that were significantly associated with reduced mortality risk included younger age at diagnosis, female gender, localized or regional spread, and receipt of local or surgical therapy (Table 2). Women had a 10% lower risk than men. HCC patients ages 40 years and years had a 23% and 11% lower mortality risk, respectively, compared with patients 60 years. Patients with localized disease had a 49% lower mortality risk compared with patients with distant disease, whereas those with regional spread had a 33% lower mortality risk. HCC patients who received local or surgical therapy were at a 67% and 62% lower mortality risk, respectively, compared with those patients who did not receive either therapy. A multivariable logistic regression model was developed to examine the association between stage of HCC and race/ethnicity. No significant associations between stage of HCC and race/ethnicity were observed, adjusting for time of diagnosis, age, gender, and SEER registry. In this model, there was an increased likelihood of having local or regional stage of disease during the more recent time period of diagnosis ( : OR, 1.65; 95% CI, ; : OR, 1.32; 95% CI, ), in patients with older age ( 40 years: OR, 0.75; 95% CI, , years: OR, 0.83; 95% CI, Table 3. Results From the Multivariable Logistic Regression Model Examining the Association Between Receipt of HCC Therapy and Race/Ethnicity Predictor variable Adjusted OR 95% CI P value Race White 1.00 Reference Black Asian Hispanic Other Time period of HCC diagnosis Reference Age at diagnosis (y) Reference Gender Male 1.00 Reference Female SEER registry Connecticut 1.00 Reference San Francisco Oakland Seattle Detroit Iowa New Mexico Hawaii Utah Atlanta Stage at HCC diagnosis Distant spread 1.00 Reference Regional spread Localized Unknown The model adjusts for time period of diagnosis, age, gender, SEER registry, and stage of disease among patient with HCC during (N 9919).

6 January 2006 RACIAL DIFFERENCES IN SURVIVAL OF HCC IN THE US 109 lower among blacks (Table 3). In this model, there was an increased likelihood of receiving therapy during the more recent time period of diagnosis, as well as for younger age, female gender, and in patients with localized or regional spread (Table 3). We also observed significant improvement over time in the observed and relative survival for patients diagnosed with HCC between 1987 and The greatest improvement has occurred within the first year after diagnosis, with a smaller improvement seen by the third year after diagnosis. The 1-year observed survival increased from 19.1% (95% CI, ) during to 28.5% during (95% CI, ). The 3-year observed survival increased from 7.3% (95% CI, ) during to 12.3% (95% CI, ) during In the unadjusted Cox proportional hazards model, mortality risk was 9% and 22% higher during and , respectively, as compared with However, when receipt of therapy was adjusted for in the model, no significant change in the mortality risk was observed during , and only a 10% reduction in mortality risk was observed during compared with (Table 2). Discussion In this population-based study, we examined racial/ethnic variation in survival of patients with HCC in the United States. Compared with whites, Asians and patients of other race (American Indian, Hawaiian, and other nonspecified races) had a 16% lower mortality risk, whereas blacks and Hispanics had an 11% and 9% higher risk, respectively. For Asians, Hispanics, and patients of other race, these differences could not be fully explained by differences in time period of HCC diagnosis, age, gender, registry, the available information on stage at diagnosis, or receipt of specific HCC therapy. However, the mortality risk in blacks became nonsignificantly different from whites when receipt of therapy was adjusted for in the model, suggesting that differences in receipt of therapy could explain the increased mortality risk among blacks. Blacks were less likely to receive local or surgical therapy compared with whites, independent of time period of diagnosis, age, gender, registry, or stage of HCC. Lastly, irrespective of race, improvements in survival also occurred during more recent time periods. Compared with , the mortality risk during and was 9% and 22% lower, respectively. This temporal improvement in survival was largely explained by the increased use of local ablation and surgical therapy for HCC. Previous studies have reported racial/ethnic differences in the incidence of HCC in the US. One study found that the age-adjusted incidence of HCC was highest among Asians (8.4 per 100,000), followed by blacks (4.2 per 100,000) and whites (2.2 per 100,000). 1 Another study reported a greater age-adjusted incidence of HCC among Hispanic men (3.29 per 100,000) and women (1.23 per 100,000) compared with white men (1.82 per 100,000) and women (0.60 per 100,000). 6 Racial/ethnic differences in the prevalence of major risk factors among patients with HCC have also been observed. One study conducted in the US reported that the proportion of HBV-related HCC was highest among Asians (49.5%), followed by patients of other race (19.0%), blacks (15.8%), and whites (5.9%). 7 In contrast, HCV-related HCC was highest among blacks (53.7%), followed by whites (51.7%), patients of other race (34.2%), and Asians (29.9%). 7 Only one study has addressed differences in HCC survival by race and found no significant differences in survival between whites and blacks. 2 However, in that study, no adjustments were made for other important predictors of survival, including age, stage of HCC at diagnosis, and receipt of therapy. Furthermore, survival among Asians and Hispanics was not examined. In the current study, differences in the receipt of therapy explained some of the observed racial/ethnic variation in survival. Compared with other race/ethnic groups, Asians in this study had the greatest 3-year relative survival (12.9%) and the largest proportion of patients with localized or regional spread HCC. Asians also had the greatest proportion of patients receiving local and surgical therapy (13.6%), which is partly explained by the relatively large proportion of early HCC. The higher proportion of Asians with relatively early HCC could be associated with the high prevalence of HBV-related HCC among Asians. 8,9 Recent immigrants from HBV-endemic areas (such as southeast Asia) and their descendants are at high risk of developing chronic HBV infection and HBV-related HCC. 10,11 Compared with HCV-related HCC, HBV-related HCC is more likely to occur in the absence of advanced cirrhosis especially in patients with perinatal acquisition of HBV. 11,12 The presence of cirrhosis accompanied by liver decompensation prohibits the majority of HCC patients from receiving surgical or local therapy or even chemoembolization. However, the likelihood of receiving local or surgical therapy was greatest among Asians, independent of stage of HCC, time period of diagnosis, age, gender, and SEER registry. Other patient, provider, or system factors related to receipt of therapy need to be examined. Differences in access to care among race/ethnicity groups might also contribute to variation in the receipt of therapy. However, information re-

7 110 DAVILA AND EL-SERAG CLINICAL GASTROENTEROLOGY AND HEPATOLOGY Vol. 4, No. 1 lated to access to care, such as insurance status, is not available from SEER. In contrast, blacks had the lowest 3-year survival (7.9%) and were the least likely to receive surgical therapy (7.0%). Blacks were also slightly more likely to have distant spread or unstaged cancers compared with other race groups. On the basis of results obtained from the multivariable analysis, receipt of local or surgical therapy appears to explain the differences in survival observed between blacks and whites. Blacks were approximately 40% less likely than whites to receive local or surgical therapy, adjusting for stage of HCC and other demographic features. However, these findings suggest that differences in stage of HCC at diagnosis cannot fully explain the differences in receipt of therapy. Further studies are needed to explain these findings. Hispanics were also more likely to have lower survival compared with whites. This increased mortality risk could not be explained by receipt of therapy, time period of diagnosis, age, gender, or SEER registry. A greater proportion of Hispanics were diagnosed with localized disease, and by controlling for this confounding factor, the mortality risk among Hispanics was further increased. Limitations of our study include the absence of information in the SEER database about other factors that affect survival, including disease comorbidity and severity of liver disease. For example, we could not adjust for the presence of cirrhosis or underlying risk factors, which could affect both the receipt of therapy and survival. Unrecognized factors associated with differences in survival remain, particularly among Asians and Hispanics. Because of different practices in evaluating and managing HCC, the inclusion of new immigrants into the SEER database could impose several confounding factors if such patients were diagnosed in their primary countries and managed, treated, and subsequently followed in the US. Differences in screening methods might also impact their results. However, we were able to adjust for several important variables that explained at least some of the observed racial variations in survival. Last, we were not able to examine the effect of type or intensity of screening or testing before HCC diagnosis. Improvements in HCC detection could potentially result in lead-time bias. The use of detection methods for HCC could vary by race. For example, alpha-fetoprotein greater than 200 ng/ml has traditionally been used to confirm HCC in patients with HCV-related cirrhosis and a hepatic mass. However, alpha-fetoprotein is insensitive of the diagnosis of HCC in African Americans. 13 There are several strengths to our study. This is the first population-based study to examine racial/ethnic differences in survival, and thus our results should be generalizable to the US population. All cases of HCC included in this study were diagnostically confirmed. Last, complete follow-up information was available for greater than 98% of patients. In conclusion, survival among patients with HCC varies significantly by race/ethnicity. These variations in survival are partly explained by differences in receipt of therapy and, to a lesser extent, stage of disease at presentation. Blacks have the lowest survival and are least likely to receive surgical or local therapy, irrespective of stage of disease at presentation. Additional studies are also needed to clarify the role of underlying risk factors for HCC, including HCV and HBV, on receipt of therapy and survival among non-white patients with HCC. References 1. El-Serag HB, Davila JA, Petersen NJ, et al. The continuing increase in the incidence of hepatocellular carcinoma in the United States: an update. Ann Intern Med 2003;139: Erratum: Ann Intern Med 2004;140: El-Serag HB, Mason AC, Key C. Trends in survival of patients with hepatocellular carcinoma between in the United States. Hepatology 2001;33: Liu JH, Chen PW, Asch SM, et al. Surgery for hepatocellular carcinoma: does it improve survival? Ann Surg Oncol 2004;11: National Cancer Institute. Surveillance, Epidemiology, and End- Results (SEER) Program Public-Use CD-ROM ( ). Bethesda, MD: National Cancer Institute, DCPC Surveillance Program, Cancer Statistics Branch, World Health Organization. International classification of disease for oncology. Geneva, Switzerland: World Health Organization, Shea KA, Fleming LE, Wilkinson JD, et al. Hepatocellular carcinoma incidence in Florida: ethnic and racial distribution. Cancer 2001;91: Di Bisceglie AM, Lyra AC, Schwartz M, et al. Hepatitis C-related hepatocellular carcinoma in the United States: influence of ethnic status. Am J Gastroenterol 2003;98: Kim WR, Benson JT, Therneau TM, et al. Changing epidemiology of hepatitis B in a US community. Hepatology 2004;39: McQuillan GM, Alter MJ, Everhart JE. Viral hepatitis. In: Everhart JE, ed. Digestive disease in the United States: epidemiology and impact. Bethesda, MD: National Institutes of Health NIH publication no , 1994: McQuillan GM, Townsend TR, Fields HA, et al. Seroepidemiology of hepatitis B virus infection in the United States to Am J Med 1989;87:5S 10S. 11. McGlynn KA, Tsao L, Hsing AW, et al. International trends and patterns of primary liver cancer. Int J Cancer 2001;94: Di Bischeglie AM. Hepatitis C and hepatocellular carcinoma. In: Liang JT, Hoofnagle JH, eds. Hepatitis. San Diego, CA: Academic Press, 2000: Nguyen MH, Garcia RT, Simpson PW, et al. Racial differences in effectiveness of alpha-fetoprotein for diagnosis of hepatocellular carcinoma in hepatitis C virus cirrhosis. Hepatology 2002;36: Address requests for reprints to: Hashem B. El-Serag, MD, MPH, The Houston Veterans Affairs Medical Center, 2002 Holcombe Blvd (152), Houston, TX hasheme@bcm.tmc.edu; fax: