Decreasing Disparity in Liver Transplantation among White and Asian Patients with Hepatocellular Carcinoma, California, 1998–2005

Abstract

Background

A preliminary study using national cancer surveillance data from 1998–2002 suggested there were significant differences between non-Hispanic whites (“whites”) and Asian/Pacific Islanders (APIs) in the use of liver transplantation as a treatment for hepatocellular carcinoma (HCC).

Methods

We conducted a study to examine whether differences in liver transplantation between white and API HCC patients were changing over time. Using a population-based statewide cancer registry, we obtained data on all HCC cases diagnosed in California between 1998 and 2005, but limited the study to white and API patients with non-metastatic HCC, sized ≤ 5.0 cm (n = 1728).

Results

During 1998–2003 (n = 1051), the odds of receiving a liver transplant were 2.56 times higher for white patients than for API patients (95% confidence interval [CI], 1.72 to 3.80 times higher), even after adjusting for age, sex, marital status, year of diagnosis, TNM stage, and tumor grade. In contrast, during 2004–2005 (n = 677), there were no significant differences in the odds of receiving a transplant. Between 2002 and 2004, changes in liver transplantation policy assigned priority points to HCC patients (initially to stage I and II, then to stage II only). After the policy changes, API HCC patients experienced a significant increase in stage II diagnoses, while white patients did not.

Conclusions

In California, there was a large and significant disparity in liver transplantation among white and API patients with HCC during 1998–2003, but not during 2004–2005. Changes in liver transplantation policy during 2002–2004 may have played a role in decreasing this difference.

Between 1995 and 2004, the incidence of liver cancer in the United States increased by 22%, with only thyroid cancer showing a greater increase. More worrisome, the mortality rate for liver cancer increased by 17% during this period, which was larger than all other cancer sites, and was in stark contrast to the decreasing mortality rates for the majority of cancer sites.¹ The largest contribution to these incidence and mortality trends was the rise in hepatocellular carcincoma (HCC). Unfortunately, HCC has one of the poorest survival rates of all cancers²; for patients diagnosed during 1995–1999 and followed through 2004, the 5-year relative survival rate was only 8.8%. Thus far, the only therapies thought to be curative for HCC are resection and liver transplantation (LT) when resection is not possible.^3,4 The use of LT to treat HCC was revitalized by Mazzaferro in 1996, when he identified cases with specific tumor criteria that clearly demonstrated a survival benefit after transplantation (no evidence of extrahepatic tumor and unifocal tumor mass < 5 cm in diameter, or ≤ 3 multifocal tumors each < 3 cm in diameter), with an overall actuarial 4-year survival rate of 75% and a recurrence-free 4-year survival of 83%.⁵ Based on the success of this therapy and the small window of opportunity to gain this survival benefit, the United Network for Organ Sharing (UNOS) implemented changes in LT policy between 2002 and 2004 that assigned priority points to HCC patients (initially to stage I and II, then to stage II only). Following the initial policy change in 2002, the number of transplants for liver cancer has risen rapidly.⁶

Groups who may derive a large benefit from these policy changes include Asian and Pacific Islanders (APIs) and other ethnic and racial groups who are disproportionately affected by liver disease.⁷ Of these groups, Vietnamese men and Korean women have the highest incidence of liver cancer, and Chinese men and women have the highest liver cancer mortality rate of any demographic group in the US.^7–9 In contrast to other groups, the etiology of liver cancer in APIs is predominantly due to chronic infection with hepatitis B virus (HBV). In patients with chronic HBV infection, the majority of patients who develop HCC do not experience hepatic decompensation before or by the time of diagnosis, indicating that HCC usually arises in the clinical setting of compensated cirrhosis which may be clinically silent.^10,11 However, prior to the adoption of the UNOS policy changes, in the context of a relatively large pool of patients awaiting transplantation of a limited organ supply, patients with liver cancer were only competitive for transplantation after they had developed decompensated cirrhosis. These factors may in part explain why, in a preliminary study using national cancer registry from 1998 to 2002 on adult patients with non-metastatic HCC sized 5.0 cm or less, non-Hispanic whites (“whites”) were 1.5 times more likely than APIs to receive a liver transplant, even after adjusting for year of diagnosis, age, marital status, tumor size, and histologic grade (P = 0.005).¹²

The state of California has the largest population of APIs in the US (4.6 million as of July 2003).¹³ Using the unique resources of the California Cancer Registry, a population-based cancer registry which has covered the entire state of California since 1988, we obtained data on all HCC cases diagnosed among whites and APIs in California between 1998 and 2005. Using three additional years of data (2003–2005) beyond the earlier preliminary study, we examined differences in the use of LT among white and API patients diagnosed in California with non-metastatic HCC, sized ≤ 5.0 cm. We also examined whether these differences were changing over time. On the basis of the considerations above, we hypothesized that the largest benefit of the UNOS policy changes would be seen among APIs with HCC.

METHODS

Study Population

HCC cases were identified through the California Cancer Registry (CCR), a population-based registry that has collected cancer incidence data for the entire population of California since 1988 through a system of regional registries. With the state’s population at 33.9 million in the 2000 US Census, nearly 140,000 cases of invasive cancer are added to the Registry each year, and data on nearly 2.3 million invasive cases have been collected since 1988. CCR is a participant in the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program, which requires the highest standards of data quality, as judged by completeness, accuracy, and timeliness.

We restricted eligibility to patients diagnosed between 1998 and 2005, because collection of detailed liver surgery codes began in 1998, and because 2005 was the latest year with complete case ascertainment. Cases were followed through the end of 2005, the latest year with complete cause-of-death ascertainment. Because of racial differences in the etiology of liver cancer and thus, in the potential impact of changes in transplantation policy, we limited to study to white and API patients with HCC. Between January 1, 1998 and December 31, 2005, 6421 white and API adults (age ≥ 18 years) were newly diagnosed in California with primary invasive cancer of the liver (SEER site recode 21071). We restricted study eligibility to patients with hepatocellular carcinoma (International Classification of Diseases for Oncology, Third Edition [ICD-O-3] histology code 8170), without nodal or distant metastasis (i.e., N0/M0 in the TNM staging nomenclature), and tumor size ≤ 5.0 cm (n = 1767). In the American Joint Committee on Cancer (AJCC) TNM staging system, these patients ranged from Stage I (T1/N0/M0) to Stage IIIA (T3/N0/M0) or IIIB (T4/N0/M0). Because of insufficient numbers for analysis, we excluded patients with unknown marital status or TNM stage. We also excluded cases identified by autopsy or death certificate only. Thus, the final sample included 1728 cases. Each case included detailed information on the most definitive site-specific surgery; a total of 31 surgery codes were available, with two codes representing transplantation (code 61, total hepatectomy and transplant; and code 75, bile duct and hepatectomy with transplant).

During the study period, CCR employed two distinct but complementary schemes for collecting data on tumor characteristics. In the beginning of 2004, the new Collaborative Staging (CS) scheme was put into place, and required the collection of a large set of tumor, node, and metastasis fields, as well as supplementary clinical and laboratory data. For cases diagnosed during 2004–2005, the registry data contained values for AJCC TNM stage groups, derived from the CS fields. Prior to 2004, a simpler Extent of Disease (EOD) system had been used for many years, which required the collection of a smaller set of tumor, node, and metastasis fields. For cases diagnosed during 1998–2003, these EOD fields were used to derive AJCC TNM stage groups, using the same staging rules that were in effect in 2004–2005.

Statistical Methods

We used t tests and chi-squared tests to compare baseline characteristics related to the probability of LT. To simultaneously assess the effect of race/ethnicity, age, sex, marital status, year of diagnosis, TNM stage, and histologic grade on the probability of receiving a liver transplant, we used logistic regression models that included terms for all of these variables. For each term, we estimated the odds ratio (OR) and associated 95% confidence interval (CI). In order to assess whether the effect of race was changing over time, we performed these analyses for three time intervals: 1) the entire study period, 1998–2005; 2) the earlier part of the time period, 1998–2003; and 3) the later part of the time period, 2004–2005. Survival estimates were computed by the Kaplan-Meyer method, using deaths from all causes. All statistical analyses were conducted using SAS for Windows, version 9.1 (SAS Institute, Cary, NC).

RESULTS

Our study population consisted of 991 white and 737 API patients. During the entire study period (1998–2005), there were large differences in the probability of LT among white and API patients with HCC (Figure 1), with unadjusted transplantation rates higher for whites (19.8%) than for APIs (14.1%) (P = 0.002). In analyses not shown, we found that in our study population, patients who received a transplant had a 5-year survival of 68.2%, versus 23.4% for patients who did not receive a transplant.

Figure 1.

Unadjusted Differences in Liver Transplantation among White and API* Patients with Hepatocellular Carcinoma,† California, 1998–2005 (n = 1728)

* White denotes non-Hispanic white and API denotes Asian/Pacific Islander.

† Only patients without nodal or distant metastasis and tumor size ≤ 5.0 cm were included in the analysis.

Table 1 shows racial differences in baseline characteristics related to the probability of LT. Compared to whites, API patients were older, more were female, more were currently married, and more were diagnosed during 1988–2002. There were no significant racial differences in TNM stage or histologic grade. To assess the role of race while adjusting for these potential confounding variables, we performed a multivariate logistic regression analysis, simultaneously measuring the effect of race, age, sex, marital status, year of diagnosis, TNM stage, and histologic grade (Table 2). Even after adjustment for these variables, the odds of receiving a transplant were 1.88 times higher for whites than for APIs (95% CI, 1.40 to 2.53 times higher). Other factors associated with higher odds of transplantation were younger age, being currently married, diagnosis during 2002, TNM stage II, and histologic grade other than unknown.

Table 1.

Baseline Characteristics of White and API^* Patients with Hepatocellular Carcinoma,^† California, 1998–2005 (n = 1728)

Characteristic	White (n = 991)	API (n = 737)	P Value
Mean age (years)	62.1	63.3	0.03
Sex (%)
Male	74.9	63.6	< 0.0001
Female	25.1	36.4
Marital status (%)
Never married	19.2	7.5	< 0.0001
Formerly married	25.0	16.8
Currently married	55.8	75.7
Year of diagnosis (%)
1998	6.1	6.8	0.04
1999	7.6	7.6
2000	7.8	6.5
2001	10.7	10.7
2002	11.6	16.2
2003	17.3	12.9
2004	18.8	17.1
2005	20.3	22.2
TNM stage (%)
I	57.7	54.1	0.32
II	38.6	42.1
III	3.7	3.8
Histologic grade (%)
Well differentiated	22.3	18.1	0.16
Moderately differentiated	18.7	19.4
Poorly differentiated/undifferentiated	6.0	7.1
Unknown	53.1	55.5

^*White denotes non-Hispanic white and API denotes Asian/Pacific Islander.

^†Only patients without nodal or distant metastasis and tumor size ≤ 5.0 cm were included in the analysis.

Table 2.

Multivariate Model of Liver Transplantation, White and API^* Patients with Hepatocellular Carcinoma,^† California, 1998–2005 (n = 1728)

Variable	Multivariate OR for Liver Transplant (95% CI)
Race
API	1.00 (referent)
White	1.88 (1.40–2.53)
Age (1-year increase)	0.93 (0.92–0.95)
Sex
Male	1.00 (referent)
Female	1.12 (0.81–1.57)
Marital status
Never married	1.00 (referent)
Formerly married	1.33 (0.79–2.25)
Currently married	2.58 (1.68–3.95)
Year of diagnosis
1998	1.00 (referent)
1999	1.19 (0.56–2.53)
2000	0.82 (0.37–1.81)
2001	1.15 (0.57–2.33)
2002	2.13 (1.10–4.13)
2003	0.83 (0.42–1.66)
2004	1.36 (0.71–2.58)
2005	1.23 (0.65–2.34)
TNM stage
I	1.00 (referent)
II	1.71 (1.29–2.26)
III	0.52 (0.20–1.36)
Histologic grade
Well differentiated	1.00 (referent)
Moderately differentiated	0.99 (0.69–1.44)
Poorly differentiated/undifferentiated	0.84 (0.49–1.45)
Unknown	0.26 (0.19–0.37)

Abbreviations: OR, odds ratio; CI, confidence interval.

^*White denotes non-Hispanic white and API denotes Asian/Pacific Islander.

^†Only patients without nodal or distant metastasis and tumor size ≤ 5.0 cm were included in the analysis.

In order to examine whether these racial differences were changing over time, we performed multivariate logistic regression analyses after stratifying by year of diagnosis (two periods, 1998–2003 and 2004–2005). Table 3 shows that during 1998–2003, the odds of receiving a transplant were 2.56 times higher for whites than for APIs (95% CI, 1.72 to 3.80 times higher), but that during 2004–2005, there were no significant racial differences in the odds of transplantation (OR for white vs. API, 1.23; 95% CI, 0.77–1.96). The importance of other factors also changed over time. During 1998–2003, being currently married was associated with a very large increase in the odds of receiving a transplant (OR, 4.24; 95% CI, 2.22–8.08), but it did not play a significant role during 2004–2005 (OR, 1.64; 95% CI, 0.89–2.99). Patients who were diagnosed in 2002 had more than a 2-fold increase in odds of receiving a transplant compared to those diagnosed in 1998 (OR, 2.09; 95% CI, 1.08–4.03), reflecting the initial decision in 2002 to award (what were retrospectively deemed excessive) priority points to HCC patients. The impact of UNOS policy decisions is also seen in the fact that a stage II diagnosis did not significantly increase the odds of receiving a transplant during 1998–2003 (OR, 1.32; 95% CI, 0.92–1.91), but during 2004–2005, the odds of receiving a transplant were 2.61 times higher among patients with a stage II diagnosis (95% CI, 1.65 to 4.14 times higher).

Table 3.

Decreasing Disparity in Liver Transplantation among White and API^* Patients with Hepatocellular Carcinoma,^† California, 1998–2005 (n = 1728)

Variable	Multivariate OR for Liver Transplant (95% Cl)
	1998–2003 (n = 1051)	2004–2005 (n = 677)
Race
API	1.00 (referent)	1.00 (referent)
White	2.56 (1.72–3.80)	1.23 (0.77–1.96)
Age (1-year increase)	0.93 (0.91–0.95)	0.93 (0.91–0.95)
Sex
Male	1.00 (referent)	1.00 (referent)
Female	1.17 (0.77–1.79)	1.06 (0.61–1.84)
Marital status
Never married	1.00 (referent)	1.00 (referent)
Formerly married	2.13 (1.01–4.49)	0.78 (0.35–1.74)
Currently married	4.24 (2.22–8.08)	1.64 (0.89–2.99)
Year of diagnosis
1998	1.00 (referent)
1999	1.12 (0.53–2.37)
2000	0.78 (0.35–1.69)
2001	1.10 (0.55–2.23)
2002	2.09 (1.08–4.03)
2003	0.78 (0.39–1.56)
2004		1.00 (referent)
2005		0.88 (0.57–1.37)
TNM stage
I	1.00 (referent)	1.00 (referent)
II	1.32 (0.92–1.91)	2.61 (1.65–4.14)
III	0.64 (0.21–1.95)	0.33 (0.04–2.68)
Histologic grade
Well differentiated	1.00 (referent)	1.00 (referent)
Moderately differentiated	1.43 (0.87–2.35)	0.56 (0.31–1.00)
Poorly differentiated/undifferentiated	1.05 (0.52–2.12)	0.63 (0.26–1.53)
Unknown	0.38 (0.24–0.59)	0.14 (0.08–0.25)

Abbreviations: OR, odds ratio; CI, confidence interval.

^*White denotes non-Hispanic white and API denotes Asian/Pacific Islander.

^†Only patients without nodal or distant metastasis and tumor size ≤ 5.0 cm were included in the analysis.

Since the UNOS policies to award priority points were targeted at patients with stage II HCC, we investigated whether there were changes in the proportion of HCC patients with stage II over time, and whether these time trends differed by race. Table 4 shows that for white patients, there was no significant change in the TNM stage of patients over time, while for API patients there was a significant shift in TNM stage. Between the earlier and later years of the study period, the proportion of API patients diagnosed with stage II HCC increased from 39.4% to 46.2%, a 17% relative increase.

Table 4.

Changes in TNM Stage among White and API^* Patients with Hepatocellular Carcinoma,^† California, 1998–2005 (n = 1728)

TNM Stage	White (n = 991)		API (n = 737)
	1998–2003	2004–2005	1998–2003	2004–2005
	Stage I	60.3	53.8	55.7	51.7
Stage II	36.8	41.3	39.4	46.2
Stage III	3.0	4.9	4.9	2.1
P value for change over time	0.07		0.04

^*White denotes non-Hispanic white and API denotes Asian/Pacific Islander.

^†Only patients without nodal or distant metastasis and tumor size ≤ 5.0 cm were included in the analysis.

Because small tumors (< 2.0 cm) are more appropriately treated with surgical resection rather than transplantation, we investigated whether there were racial differences in the prevalence of these tumors that might explain the racial differences in transplantation. In our study population, there were no significant racial differences in the proportion of HCC patients with tumors sized < 2.0 cm in either the earlier (white, 16.7% versus API, 13.4%; P = 0.14) or later (white, 14.7% versus API, 14.8%; P = 0.97) years of the study period.

DISCUSSION

LT for HCC has led to dramatic improvements in survival for many patients if the tumor is detected at an early stage. Identifying these patients and assigning their priority relative to others awaiting LT has been a difficult balance. Realizing that the opportunity for LT in patients with liver cancer was small, beginning in 2002, UNOS assigned additional MELD (Model of End-Stage Liver Disease) points to these patients, to increase their chances of undergoing LT. Between 2002 and 2004, the methodology for assigning these points underwent a number of modifications to make it equitable. The goal is that appropriate numbers of patients with liver cancer are transplanted, but not to the extent that others are unfairly excluded. Although imperfect, these policy changes have led to a remarkable improvement in the number of liver transplants being performed for liver cancer in the US. While these new rules should benefit all patients with early stage HCC, we hypothesized that APIs would benefit the most, because of the frequency of HCC developing in hepatitis B patients in the absence of decompensated cirrhosis. Our analyses may be among the first to document the particularly favorable outcomes seen among APIs after the implementation of these policy changes.

We did find that 5-year survival was markedly improved for patients who received a liver transplant (68%), compared to those who did not (23%). The number of transplants for liver cancer increased dramatically over time: a patient with HCC diagnosed in 2002 was more than twice as likely to receive a liver transplant than one diagnosed in 1998. The UNOS policy changes were targeted toward those with stage II disease, and we did find that patients with stage II cancer were 2.61 times more likely to receive a transplant, compared to patients with stage I disease. Thus, the policy changes appear to have been broadly successful for patients with HCC. However, it did appear that the largest benefits were seen for API patients, because the policy changes were temporally associated with the elimination of a large disparity in LT between whites and APIs seen during 1998–2003. In this earlier time period, the odds of transplantation were 2.56 times higher for white patients, but during 2004–2005, after the policy changes, there was no significant racial difference in the odds of transplantation for HCC.

Our data also show that age, marital status, and histologic grade other than unknown improve the odds of undergoing LT for HCC. This is likely a reflection of some of the requirements for patients prior to undergoing LT. Although there is no specific age limitation to LT,^14,15 older liver transplant recipients have been shown to have poorer long-term outcomes than younger patients,¹⁶ in part because of their increased risk of cancer¹⁷ and coronary artery disease.¹⁸ Medical comorbidities and functional status are important factors in deciding who will undergo LT, and thus, younger patients are more likely to receive transplants because of fewer comorbidities and better functional status. Another important factor is the social support available to a patient who will undergo a major operation with a prolonged recovery time, and the need for close adherence to medications, lab testing, and clinic visits after surgery. The improved odds of those who are married may be a reflection of these requirements, although it is not clear why the effect of marital status would change over time. Histology is often not a factor in deciding LT for HCC and now is infrequently done, since imaging in conjunction with a serum alpha fetoprotein level can often make the diagnosis without the risk of tumor seeding. However, if a biopsy is done and is inconclusive, one may be less likely to be offered a transplant.

In addition to changes in UNOS policy, other factors were changing during the study period. Screening for hepatitis B, and surveillance for liver cancer among hepatitis B carriers (using abdominal imaging and alpha fetoprotein) have also increased over the past decade, especially in the API community. The American Association for the Study of Liver Diseases recommends surveillance for liver cancer in API hepatitis B carriers, beginning at age 40 for men and age 50 for women.¹⁹ The practice of screening has been controversial in terms of its cost-effectiveness²⁰ and true benefit to survival. There have been many retrospective studies suggesting benefit, which have suffered from potential lead and length time biases.²¹ To date, there has been only one randomized trial showing a benefit from screening, a study of hepatitis B patients in China which found a 37% reduction in mortality.²²

Our data suggest that in California and particularly among APIs, screening for liver cancer is leading to an increase in diagnosis at early stages, when LT can be offered. In many parts of California, there have been major hepatitis B education and screening campaigns directed at Asians,²³ and specialized centers have emerged which focus on liver disease in Asian patients. Despite these developments, screening rates for HBV infection and liver cancer among APIs remain low.²⁴ These screening efforts must be encouraged, and new methods must be identified to reach the numerous and diverse ethnicities that comprise the API population. Reducing barriers to LT will also require different approaches across the many API ethnicities.²⁵

Our study is subject to several limitations. We aggregated all API groups together to achieve a greater sample size, although results may have differed for some API subgroups.¹³ In fact, the “API” category consisted of more than 16 ethnicities, including Chinese, Japanese, Vietnamese, Korean, and Hawaiian, with varying incidence of liver cancer, and differing rank order of liver cancer incidence and liver cancer mortality.²⁶ Barriers to screening for these different API groups may include linguistic, cultural, and religious factors, and we must design interventions that acknowledge and incorporate these factors. Our study is also limited by the lack of data on risk factors for HCC, as well as by the lack of clinical details on patients’ liver disease, both of which may play a role in determining therapy for HCC and LT. Long-term follow-up will be necessary to see whether the trends observed in this study continue.

It is possible that the increase in use of LT among APIs may represent overtreatment with LT and undertreatment with surgical resection. It is also possible that factors we did not measure may have made whites less competitive candidates for LT. For example, if a substantially lower proportion of API HCC patients had cirrhosis, we would expect fewer to undergo LT, since surgical resection is the treatment of choice in non-cirrhotic patients with HCC. On the other hand, if the prevalence of alcohol abuse or intravenous drug use were substantially greater among whites, this could lead to fewer receiving LT. In a recent study of patients with HCC due to HBV or hepatitis C virus (HCV),²⁷ Barazani et al. found that HBV patients were predominantly Asian (84%) and HCV patients were predominantly Caucasian (72%). Cirrhosis was present in 98% of HCV patients versus 79% of HBV patients. However, 78% of HCV patients had alcohol abuse or intravenous drug use, versus 13% of HBV patients.

Despite these limitations, our data suggest that the assignment of MELD priority points for patients with HCC has led to an increase in LT, and these policy changes were temporally associated with the elimination of a large disparity between whites and APIs seen during 1998–2003. The UNOS policy changes have provided a rational means to increase LT among groups with historically lower use of this life-saving treatment.