The incidence of hepatocellular carcinoma (HCC) is increasing in the United States, and the care of these patients remains highly specialized and complex. Multiple treatment options are available for HCC but their use and effectiveness remain unknown.
Using Surveillance, Epidemiology, and End Results (SEER)-Medicare linked data, 8730 patients who were diagnosed with HCC between 1991 and 2005 were identified. Therapy included surgical resection (8.7%), liver transplantation (1.4%), ablation (3.6%), or transarterial chemoembolization (16%). Patients who received no or palliative-only treatment were grouped together (NoTx; 70.3%). Patient, disease, and tumor factors were examined as determinants of therapy.
HCC is increasing in the Medicare population. The median age at diagnosis was 75.1 years and 73.6% of patients were coded as white, 17.2% as Asian, 8.3% as black, and 0.9% as other race. The rate of therapy increased over time, but only 29.7% of patients overall underwent therapy. In patients with early stage HCC, only 43.1% underwent therapy. In the NoTx group, 49.4% did not have cirrhosis, 36.0% had tumors that measured <5 cm, and 39.8% were diagnosed with stage I or II disease when variables were complete. The use of therapy for all HCC patients increased over time, correlating with a commensurate increase in median survival. In multivariate regression analysis, patients who received any modality of treatment achieved significant benefit compared with the NoTx group (odds ratio, 0.41; 95% confidence interval, 0.39-0.43).
Hepatocellular carcinoma (HCC) is one of the most common malignancies in the world, with over 500,000 new cases diagnosed worldwide per year.1 The number of cases of HCC is projected to rise dramatically over the next 10 to 15 years in North America2 because of the high prevalence of chronic hepatitis infections. Patients with hepatitis B or C virus are estimated to develop HCC at yearly rates of 5% and 0.5%, respectively, in the United States.3-5 Four procedure-based treatments have been shown to be potentially beneficial for HCC: surgical resection, ablation, transarterial chemoembolization (TACE), and liver transplantation.6 For selected patients, liver transplantation offers superior results, but the prolonged time on the waiting list may negate th.is apparent survival advantage. Surgical resection and ablation are potentially curative options in patients with small tumors and adequate liver reserve.
Regional, socioeconomic, and racial differences in access to definitive treatment and care for patients with HCC have been suggested.7, 8 Using population-based data, we examined whether these differences exist on a national level and whether this variation and underuse of treatment may contribute to the poor outcomes noted after a diagnosis of HCC.
MATERIALS AND METHODS
Patients were identified from the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database between 1991 and 2005. The SEER program is a National Cancer Institute-sponsored tumor registry; the diagnoses for 2000 through 2005 in the SEER-Medicare linked data were obtained from 16 SEER registries. These registries represent approximately 26% of the US population. The 1991 through 1999 diagnoses are from 12 SEER registries, which comprised approximately 14% of the US population. The SEER population is comparable to the US elderly populations with regard to age, sex, and socioeconomic factors. However, cancer mortality rates are slightly lower in the SEER-Medicare cohort compared with the US population.9, 10 Each registry reports information on incident cancer cases, including patient demographics, date of diagnosis and stage, primary tumor site, histological type, initial cancer-directed surgery and/or radiotherapy, and date and cause of death. All newly diagnosed HCC cases are obtained from multiple reporting sources such as hospitals, outpatient clinics, and laboratories. Through SEER's linkage with Medicare beginning in 1991, additional information regarding patients aged ≥65 years is captured about any service for which a claim may be filed. This includes inpatient and outpatient procedures and physician-generated and laboratory-generated claims, as well as home health and hospice billings.11, 12
We selected all Medicare-eligible patients aged ≥65 years at time of diagnosis and registered in 1 of the SEER registries between January 1, 1991 and December 31, 2005 who were diagnosed with HCC (using International Classification of Diseases for Oncology [ICD-O] code 817- or SITER [Site Recode Definition] 29). Two groups were constructed: those treated with therapy and those receiving no therapy. Therapy was defined as undergoing at least 1 of the following: surgical resection, liver transplantation, ablation, or TACE. Liver resection was identified in the Medicare databases using the ICD ninth revision (ICD-9) procedure codes (50.3 or 50.22) and Healthcare Common Procedure Coding System (HCPCS) codes (47120, 47122, 47125, 47130, 47140, 47141, and 47142). Liver transplantation was identified in the Medicare databases using ICD-9 procedure codes (50.4, 50.5, 50.51, and 50.59) and HCPCS codes (47131, 47133, 47135, 47136, 47140, 47141, 47142, 47143, 47144, 47145, 47146, and 47147). Ablation was identified in the Medicare databases using ICD-9 procedure codes (50.23, 50.24, 50.25, and 50.26) and HCPCS codes (47380, 47381, 47382, 47370, and 47371). TACE was identified in the Medicare databases using ICD-9 procedure code 99.25 and HCPCS codes (37204, 36249, 75726, 75894, and 75898). A hierarchical model was constructed to account for patients who may undergo more than 1 therapy. This model, in descending order, was liver transplantation, surgical resection, ablation, and TACE. Each patient was assigned to only 1 treatment group; patients who underwent multiple treatments were assigned to the therapy group with the highest order. This order was established to account for the magnitude and potential for cure with the specific procedure.
Patients who were alive but discontinuously enrolled in Medicare Parts A and B after diagnosis were excluded; for patients in the therapy group, this was defined as 1 year after surgery. Patients were also excluded if the diagnosis was obtained only from a death certificate or autopsy; if the date of death differed by >2 months in the SEER and Medicare databases; if the diagnosis date was missing or before 1991; or, for those in Medicare, for end-stage renal disease or disability.
Definition of Patient, Tumor, and Health System Characteristics
Age and sex was recorded and captured in all patients. Race was categorized as white, black, Asian/Pacific Islander, or other. Specific tumor variables included tumor size >5 cm or <5 cm in diameter and number of tumor nodules. Tumor nodules were listed as the median or > or <1. Other demographic variables included region (Northeast, South, Midwest, and West), hospital location, and hospital size (number of beds). Stage was determined using the American Joint Committee on Cancer (AJCC) HCC staging system and the histological staging system. Patients with stage I or II disease were grouped together and patients with higher than stage II disease were grouped together. Presence of and previously recorded diagnosis of cirrhosis identified from Medicare claims files were used (ICD-9 codes 571.2, 571.5, or 571.6).
The SEER database (Patient Entitlement and Diagnosis Summary File [PEDSF]) was used to determine patient demographic information, including socioeconomic status (SES) as previously described.12 Census data from 2000 were used in conjunction with median income and per capita income in both the patient's census tract and ZIP code of residence to extrapolate SES; this was then subdivided into fractions at approximately the average poverty line for the time period of the study. Because much of these data are incompletely entered in SEER, the following hierarchy was used. Census tract median income was the first choice value. If this information was missing, ZIP code median income was substituted. If ZIP code median income was missing, census tract per capita income was used, and if that was missing, ZIP code per capita income was used. Finally, patients for whom all of the above data were missing were classified in the lowest SES bracket.11 SES was categorized into quartiles with 0 indicating the lowest quartile and 3 indicating the highest quartile.
Patient comorbidity was assessed using both inpatient and outpatient Medicare claims for the 13 months to 1 month before cancer diagnosis. The Li et al, Klabunde et al, and Deyo et al modifications of the Charlson comorbidity index were used.13-15 The comorbidity index was categorized into quartiles of 0, 1, 2 or 3 and higher. A missing comorbidity index was incorporated in the lowest quartile.
Vascular invasion was defined as a single lesion with intrahepatic vascular invasion or multiple (satellite) tumors with intrahepatic vascular invasion or >1 lobe involved by contiguous growth or extension to extrahepatic blood vessel(s) such as the hepatic artery, inferior vena cava, or portal vein.
Survival was calculated as the time from diagnosis to death from any cause or date of last follow-up. The date of therapy was determined using Medicare Provider Analysis & Review (MEDPAR) and non-MEDPAR claims; the date of death was captured in PEDSF through its linkage with the National Death Index, which is maintained by the National Centers for Health Statistics and updated on an annual basis. For those who did not die, the date of last follow-up was December 31, 2007.
Descriptive analyses included means and frequency distributions. Cochrane-Armitage trend tests were used to evaluate for linear trends in the percentage of patients receiving each type of therapy previously described. Univariate analyses of associations between patient demographics, patterns of care, and the receipt of therapy were performed using chi-square tests. Variables chosen a priori to be tested included patient age, sex, race, SES, and comorbidity score; tumor size and stage; and hospital location (region and urban vs rural) and teaching status. Univariate analyses were stratified according to year group and adjuvant treatment. Survival was assessed according to the method of Kaplan and Meier and compared using the log-rank test. Cox models were constructed by including patient age, sex, race, SES, and comorbidity score; tumor size and stage; hospital types; and therapy. The final models were determined by stepwise covariate selection procedures, and enforcing age and gender in the model. Overall therapy and specific procedures were run separately. All data manipulation and statistical analysis were performed using SAS statistical software (version 9.2; SAS Institute, Inc, Cary, NC).
This study was evaluated and determined to be exempt from review according to the University of Massachusetts Medical School Institutional Review Board.
Incidence and Overall Survival
The incidence of HCC increased over time (P < .001) (Fig. 1A. This rate of diagnosis in the SEER-Medicare database was controlled for the number of states included in the cohort so that the enrollment of cases by state was kept constant. Cases of HCC from 1991 accounted for only 3.0% compared with 11.4% in 2005. The use of therapy for HCC also increased over time (P<.001) (Fig.1B); this trend was most apparent after 1999. Over the time period of this study (1991-2005), the overall survival of patients diagnosed with HCC improved as well (Fig. 1C). Specifically, the median survival increased over time (2002-2005: 5.3 months; 1999-2001: 3.5 months; and 1991-1998: 3.1 months [P < .001]). This improvement in survival was greatest in the therapy group (2002-2005: 18.8 months; 1999-2001: 14.7 months; and 1991-1998: 10.9 months [P < .0001]).
Table 1 displays the general demographics of the entire cohort of 8730 HCC patients. The median age of the patients was 75.1 years and the majority of patients were white (73.6%) or male (66.7%). Asian patients were the second most common race diagnosed (17.2%). Only 32.2% of patients harbored nodules measuring >5 cm. A large percentage of patients in the SEER-Medicare cohort resided in the West region (56.0%). Vascular invasion was present in 11.2% of patients. Of all patients diagnosed with HCC, only 29.7% underwent therapy. This included liver resection (8.7%), liver transplantation (1.4%), ablation (3.6%), or TACE (16.0%).
Table 1. General Demographics of 8730 HCC Patients in the SEER Medicare Population
Comparison of the demographics and characteristics of the patients who underwent therapy (n = 2595; 29.7%) and those who received no therapy (n = 6135; 70.3%) is shown on Table 2. The therapy group was younger (mean age, 73.3 years vs 75.8 years; P < .0001) and demonstrated associations with race (P < .0001). When patients underwent therapy, their tumors were more likely to measure <5 cm (49.8% vs 36.0%; P < .0001) and 58.8% of patients were diagnosed with stage I or II HCC. In the no-therapy group, 49.6% of patients were not cirrhotic, 36.0% had tumors measuring <5 cm, and 39.8% were diagnosed with stage I or II disease when these variables were available.
Table 2. Patient Demographics and Hospital Characteristics of 8730 Patients With HCC Stratified by Therapy Versus No Therapy
No Therapy (n=6135)
HCC indicates hepatocellular carcinoma; AJCC, American Joint Committee on Cancer; SES, socioeconomic status.
Median age (range), y
Tumors >5 cm
Tumors ≤5 cm
AJCC stage I or II HCC
Presence of cirrhosis
Charlson comorbidity score
Hospital bed size
Predictors of Survival: Univariate Analysis
Overall survival was found to be improved with therapy compared with no therapy (P < .001) (Fig. 2A). Tumor size was found to be a significant predictor of survival on univariate analysis (P < .001) (Fig. 2B). Time from diagnosis to therapy was calculated to determine whether this interval change could be a contributor to improved survival over time. The median time intervals from diagnosis to specific therapy were worse over time for surgical resection (1991-1998: 19 months; 1999-2001: 27 months; and 2002-2002: 35 months) and for TACE (1991-1998: 28 months; 1999-2001: 34 months; and 2002-2002: 46 months) (P < .001). Time intervals for liver transplantation were not analyzed given waiting list time and geographic variability.
Predictors of Survival: Multivariate Analysis
In multivariate Cox regression analysis, patients who received any modality of therapy achieved benefit compared with those receiving no therapy (hazard ratio [HR], 0.41; 95% confidence interval [95% CI], 0.39-0.43) (Table 3). Specifically, adjusted overall survival was greatest in the patients who underwent liver transplantation (HR, 0.19; 95% CI, 0.15-0.24), surgical resection (HR, 0.28; 95% CI, 0.26-0.31), ablation (HR, 0.36; 95% CI, 0.31-0.41), and TACE (HR, 0.52; 95% CI, 0.49-0.56). Survival was found to be independently lower with a tumor size >5 cm (HR, 1.33; 95% CI, 1.25-1.42), increased comorbidity (HR, 1.25; 95% CI, 1.18-1.33), white race (HR, 1.14; 95% CI, 1.08-1.21), and black race (HR, 1.20; 95% CI, 1.09-1.33). Improved survival was independently observed with early stage HCC (HR, 0.62; 95% CI, 0.59-0.66) and care received at a teaching hospital (HR, 0.91; 95% CI, 0.87-0.95).
Table 3. Cox Multivariate Analysis for Survival of 8730 Patients From Time of Diagnosis of HCC
A separate subgroup analysis was performed for patients with early stage HCC (stage I or II). Overall, 43.1% of patients with early stage tumors received therapy. Patients who received therapy were more likely to have received care at large (47.7% vs 22.5%) or teaching (64.3% vs 58.8%) centers. There was a survival benefit for receiving therapy compared with no therapy for patients with early stage tumors (HR, 0.11; 95% CI, 0.08-0.14). The use of therapy for early stage HCC increased over time, suggesting improved use of therapy over time for patients with treatable or early stage tumors (Fig. 3).
The results of the current study demonstrated that in the Medicare population, HCC patients who received any therapy experienced a substantial survival advantage over their peers who did not receive therapy. Despite evidence that many patients had favorable biological characteristics, <30% of patients diagnosed with HCC received therapy. These findings suggest that many patients who did not undergo therapy had potentially treatable tumors in noncirrhotic livers. This was further supported by the finding that the rate of therapy in patients with early stage HCC was only 43.1%. The frequency of HCC in the Medicare population is increasing and despite improved outcomes with therapy over time, the median survival from diagnosis remains short. Barriers to treatment and its underuse must be identified to improve survival in patients diagnosed with HCC in the United States.
Given the lack of population-based studies in the US, the true rising rate of HCC incidence and its peak remains unknown. Previous reports from population-based databases and transplant registries have shown an increase in HCC in the US.2, 16-18 Given the rising incidence of hepatitis C and population-based migration of patients infected with hepatitis B, the rate of HCC will likely only increase.19-22 Lang et al estimated the current total annual burden of HCC to be $454.9 million based on 13,824 cases of HCC in the US.23 The results of the current study demonstrate that therapy increased over time whereas the number of patients who did not receive therapy decreased. When examining patients with early stage tumors, there was an increased use of therapy over no therapy compared with patients with later-stage tumors, as one would expect; this effect increased over time as well. However, much improvement remains needed, because less than half of the patients with early stage tumors received therapy. There was an increase in the rate of surgical resection and TACE over the course of the study. One potential explanation for this was, because of the older Medicare population, patients with small HCC underwent surgical resection and not liver transplantation. The increase in the use of TACE may reflect the increasing expertise in interventional techniques over time. Further methods to improve diagnostic and therapeutic measures for these patients will remain a health policy issue.
From 1991 through 2005, the care and survival of patients diagnosed with HCC in the Medicare population improved over time. This benefit was observed in all groups examined: patients receiving therapy, patients receiving no therapy, and all patients. The greatest benefit was observed in the therapy group. The reasons behind this are not directly addressable in a population-based database, but may include improved diagnosis, multidisciplinary care, and therapy techniques such as surgical resection, liver transplantation, ablation, and TACE. Over the time period of this study, these therapies and their use have evolved tremendously.6, 18, 24-28 Changes in practice and safety may have resulted in more patients being treated for their HCC over time. The substantial survival benefit that a patient receives from any therapy suggests that improving access to care may be the best strategy to rapidly improve HCC outcomes overall.
We divided our cohort into a therapy group, which was comprised of surgical and nonsurgical treatments, and a no-therapy group for HCC. This was done to determine the use of any therapy for HCC in this population. Because of the limitations of the large database, clinical criteria for certain therapies would be difficult to determine retrospectively and without all clinical characteristics of a patient and their tumor. Nonetheless, all 4 therapies examined in this study, both individually and collectively as a group, provided a survival benefit from diagnosis in the Cox multivariate model. Liver transplantation and surgical resection provided the largest survival benefit. This was not surprising given the selection bias that is inherently involved in those patients undergoing those procedures. This would include favorable tumor biology and grade of cirrhosis and potentially better access to these procedures. Further research is currently underway to attempt to address the role that patient characteristics may play in the receipt of therapy for HCC.
On the basis of the results of the current study, it would appear that <30% of Medicare patients with HCC undergo therapy. Considering that TACE comprised nearly half of the treatment administered to this group and is not curative, this low number is alarming and represents underuse of therapy in this country. El Serag et al have previously shown that 19.2% of patients with unfavorable tumor features received potentially curative therapy.8 In this study, approximately y39.8% of patients who did not receive therapy were diagnosed with stage I or II HCC. Reasons for not receiving therapy could not be clearly elucidated in this population-based study. We did find that in the no-therapy group, there was a higher incidence of black race and lower SES, and patients were more likely to be seen at a small or medium-sized hospitals. In the multivariate model, advanced age, white race, black race, and lowest SES quartile were associated with decreased overall survival after a diagnosis of HCC. In an analysis of the SEER registry, Asians were found to have the highest survival and were more likely to receive local and surgical therapy compared with whites or blacks.29 Racial and SES disparities have been found previously in cancer models, such as analyses of prostate cancer in the Medicare population.30 Davila et al postulated that geographic variation may, in part, play a role in the differences in age, gender, and race of patients with HCC.17 Using SEER, Artinyan et al recently demonstrated that racial and ethnic disparities exist in HCC despite comparable receipt of therapy.31 Reasons for potential disparities in the receipt of therapy for patients with HCC will need further elucidation. The results of the current study indicate that the improvement of HCC patient survival over time could be largely because of the increase in the rate of therapy. Conversely, there is room for improvement in the time interval between diagnosis and therapy.
The current study is subject to many limitations. We relied on administrative Medicare claims data for patients aged ≥65 years. This limits our ability to truly assess access to liver transplantation because many patients aged >70 years are not considered suitable candidates for transplantation surgery. It also limits the evaluation of patients aged <65 years, who may represent missed cases because of infection from hepatitis B or C, which may occur at a younger age than other solid organ cancers.32, 33 Important clinical variables such as performance and psychosocial status, patient preference, or the presence of portal hypertension could be included in the analysis. Physician and hospital preferences for therapy were not known and may influence treatment options as well. In addition, as a claims-based database, Medicare data will fail to reflect services for which no bill was submitted, which may happen for a variety of reasons. This claims data also do not allow for access to individual laboratory tests such as α-fetoprotein, which may have important prognostic value in therapy for HCC. Furthermore, claims data do not provide specific pertinent details regarding therapies used. For example, types of ablation (eg, alcohol, radiofrequency), medications used during TACE, and extent of surgical resection remain unknown but may have significant therapeutic implications.
This study is important because it is a population-based study that comprises a large proportion of the SEER registries, which is one of the most comprehensive and authoritative data sources on cancer. Using the Medicare claims data allows one to identify information regarding patient comorbidity and use of and access to therapy. All of these could be confounders for survival. Most importantly, however, the SEER-Medicare linked databases allow us to report long-term survival across patients who are treated surgically, medically, and palliatively; this unique feature allows one to make broad statements regarding the overall care with respect to the denominator or total number of patients with HCC.
In the Medicare population, HCC patients who received therapy experienced a substantial survival advantage over patients who received no therapy. Despite evidence that many patients had favorable biological characteristics, <28% of patients diagnosed with HCC received any treatment. Barriers to treatment and its underuse must be identified to improve survival in patients diagnosed with HCC in the United States.
We thank Elizabeth Dobie and Nonna Akkerman for their assistance with data acquisition.
CONFLICT OF INTEREST DISCLOSURES
Supported by the American Society of Transplant Surgeons Faculty Development Award (to S.A.S.) and a Howard Hughes Medical Institute Early Career Grant (to J.F.T.).