Effectiveness of Initial Transarterial Chemoembolization for Hepatocellular Carcinoma Among Medicare Beneficiaries

Abstract

Background

Optimal administration of transarterial chemoembolization (TACE), the standard approach for intermediate stage hepatocellular carcinoma (HCC), requires clinical and technical expertise. We sought to evaluate if TACE retains its effectiveness when administered across a broad range of healthcare settings. With increasing use of yttrium⁹⁰ radioembolization (Y⁹⁰), we explored the effectiveness of Y⁹⁰ compared with TACE.

Methods

HCC patients diagnosed from 2004–2009 treated initially with TACE or Y⁹⁰ were identified from the Surveillance Epidemiology and End-Results (SEER)-Medicare linkage. Key covariates included prediagnosis AFP screening, complications of cirrhosis, and tumor extent. Effect of treatment, patient, and healthcare system factors on overall survival (OS) was evaluated using multivariable Cox proportional hazards. Stratified OS estimates are provided. Propensity score (PS) weighting was used to compare effectiveness of Y⁹⁰ to TACE.

Results

Of 1,528 with intra-arterial embolization, 577 had documentation of concurrent chemotherapy (e.g. TACE). Median OS was 21 months (95% CI 18–23) following TACE, 9 months (95% CI 1–41) following Y⁹⁰. Refined survival estimates stratified by stage, AFP screening, and liver comorbidity are presented. Ninety day mortality after TACE was 21–25% in patients with extrahepatic spread or vascular invasion. In the PS-weighted analysis, Y⁹⁰ was associated with inferior survival, aHR 1.39 (95% CI 1.02–1.90).

Conclusions

The effectiveness of TACE is generalizable to Medicare patients receiving care in a variety of treatment settings. However, early post-treatment mortality is high in patients with advanced disease. We found no evidence for improved outcomes with Y⁹⁰ compared with TACE. Survival estimates from this large cohort can be used to provide prognostic information to patients considering palliative TACE.

INTRODUCTION

Hepatocellular carcinoma (HCC) is one of a handful of malignancies with rising incidence and mortality rates in the United States.[1, 2] HCC carries a very poor prognosis, in large part because two thirds of patients have underlying cirrhosis and most have multifocal cancer, both of which limit treatment options.[2, 3] Two intra-arterially delivered locoregional therapy (LRT) options are considered to be reasonable initial therapy for patients with unresectable HCC and compensated cirrhosis: transarterial chemoembolization (TACE) and yttrium⁹⁰ radioembolization (Y⁹⁰). TACE delivers chemotherapy directly to the cancer through the hepatic arterial system with subsequent arterial embolization, or by the contemporary approach of concurrent chemotherapy delivery and embolization in the form of doxorubicin-eluting beads. Among patients with nonmetastatic HCC and compensated cirrhosis, TACE improves survival over supportive care[4–6] and is widely considered the first-line treatment option.[7] . Bland embolization performed without chemotherapy is used for patients with less compensated cirrhosis at some centers to minimize the risk of the procedure. Bland embolization appears to offer less robust survival benefit.[4]

The optimal delivery of TACE requires considerable clinical expertise to appropriately select patients for therapy and technical expertise for safe and effective administration.[8] Therefore TACE effectiveness might be diminished in centers without ready access to multidisciplinary clinical teams including expert diagnostic and interventional radiology, hepatology, transplant surgery, and oncology. In a prior evaluation of HCC treatment and outcomes using the Surveillance Epidemiology and End Results (SEER) Medicare linkage of patients treated in the 1990s, outcomes following TACE were quite poor with a median survival of less than a year.[9] A more recent SEER-Medicare analysis of patients treated from 2000–2005 also showed a short median survival, only 14 months.[10] Both of these estimates were well below the quality metric of 20 months proposed by the Society of Interventional Radiology Standards of Practice Committee.[8] However, as the more contemporary analysis excluded patients who underwent ablation, resection, or transplant at any time, only patients with the most advanced disease were evaluated. Therefore, we sought to evaluate the effectiveness of first-line TACE across the spectrum of cancer stages in a contemporary cohort.

Y⁹⁰-radioembolization, in which either glass or resin microspheres laden with yttrium⁹⁰ are delivered to cancers through the hepatic arteries, is emerging as an excellent TACE alternative as data emerge reporting survival following Y⁹⁰ to be comparable to TACE.[11–21] With overlapping clinical indications and similar outcomes, both TACE and Y⁹⁰ are reasonable initial therapy choices for HCC in a patient with compensated cirrhosis. However, the comparative effectiveness of these procedures as they are being administered throughout the United States is unknown. Therefore using this large observational dataset, we also sought to explore the comparative effectiveness of Y⁹⁰ versus TACE.

PATIENTS and METHODS

Study Population

The cohort was derived from the Surveillance Epidemiology and End Results (SEER) Medicare linkage (described in detail elsewhere[22, 23]). We included patients diagnosed with HCC from 2004 to 2009, excluding autopsy diagnoses. To avoid misclassification of liver metastases, patients with prior invasive cancer within 5 years were excluded.[9] To ensure availability of claims, only patients with continuous enrollment in Medicare A and B and patients not enrolled in Medicare Managed Care in the 12 months before and after diagnosis (or death) were included. This analysis was deemed exempt from review by the Office of Human Research Ethics at the University of North Carolina (#12-1828).

Covariates

Sociodemographic patient data, tumor size and number and presence of macrovascular invasion were obtained from SEER (see Supplemental Methods). Laboratory parameters were not available, therefore diagnosis codes for complications of cirrhosis were used to control for confounding liver disease.[24] HCC screening behavior was ascertained by pre-diagnosis AFP[25, 26] as a surrogate measure of performance status as use of cancer screening is associated with decreased probability of frailty in cancer patients.[27] Cause of liver disease was ascertained from claims using ICD-9 diagnosis codes for hepatitis B virus, hepatitis C virus, alcoholic cirrhosis, and other cirrhosis. Non-cirrhotic comorbidity was determined using the Klabunde modification of the Charlson Comorbidity Index (CCI).[28] All claims-based covariates were ascertained in the 12 months before diagnosis.

Treating hospital was defined for the majority of patients from inpatient claims on the date of initial procedure. Hospital characteristics were derived from the SEER hospital file. Hospital volume was defined as LRT volume at the treating hospital in the 12 months preceding each individual patient’s treatment.

Treatment

Treatment group was determined by initial therapy delivered. Codes for arterial occlusion without intra-arterial radiotherapy were categorized as embolization (ETable 1); those with codes for chemotherapy or intra-arterial chemotherapy delivery at time of embolization were classified as chemoembolization (TACE) and those without as bland embolization (TAE). Because chemotherapy claims are reliable when present but are less well identified in records of hospitalized patients,[29] some patients receiving chemoembolization were likely misclassified as bland embolization. Because the intent of the procedure cannot be ascertained from claims, some patients classified as TAE may have received embolization for reasons other than anticancer therapy such as to control bleeding from a ruptured tumor or in preparation for Y90-radioembolization that was subsequently aborted. There is not a specific code for drug-eluting beads thus we could not compare specific TACE strategies. Patients with codes for intra-arterial radiotherapy delivery were categorized as receiving Y⁹⁰regardless of embolization. Because pre-procedure arterial embolization is required to ensure dose and safety of Y⁹⁰embolization codes in the 60 days before Y⁹⁰ administration were considered part of Y⁹⁰.

Statistical Analysis

Overall survival (OS) was measured from date of initial procedure until death. Kaplan-Meier methods estimated OS for the entire cohort and within strata of tumor extent, liver comorbidity, and use of pre-diagnosis AFP screening. Cox proportional hazards (PH) models were used to evaluate factors associated with OS. Robust variances were used in all analyses.

As TACE is the standard treatment, we also sought to ask “what would the effect of Y⁹⁰ be among patients typically treated with TACE” by comparing the effectiveness of Y⁹⁰ to TACE in a population of patients resembling those selected for initial TAE/TACE. To do so, a weighted pseudo-population was created from the propensity score (PS) for TAE/TACE (see Supplemental Methods).[30] This created treatment groups balanced on key patient characteristics in which multivariable Cox PH models were used to adjust for residual covariate imbalances and OS was compared by treatment. In sensitivity analyses, we compared the effectiveness of Y⁹⁰ to TAE/TACE after restricting to only those TACE patients with codes for chemotherapy. We also conducted a PS-trimming sensitivity analysis to address the potential for unmeasured confounding by frailty.[31] As patients treated contrary to prediction are most likely to have unmeasured confounding that influences treatment selection (e.g. frailty), omitting them may improve the validity of the treatment effect estimate. If the observed treatment effect were due to unmeasured confounding, it would be expected to approach the null with trimming. Because a larger proportion of TACE patients underwent subsequent curative surgery, sensitivity analysis explored the magnitude of this effect by censoring patients at the time of curative surgery.

RESULTS

Initial LRT was administered to 1,651 patients, including 1,528 treated with TAE/TACE and 121 with first-line Y⁹⁰ (Figure 1). Of TAE/TACE patients, 577 (38%) received TACE. Median age of the TAE/TACE group was 72 (range 27–94), 66% of whom were white and 14% Asian (Table 1). The majority of TACE patients had liver-confined unifocal (207, 42%) or multifocal (207, 36%) disease without macrovascular invasion, though 126 (22%) had either macrovascular invasion or extrahepatic disease. Second-line therapy was administered to 1,010 (66%) TAE/TACE patients, including repeat TAE/TACE in 635 (42%), curative surgery in 99 (6%), ablation in 129 (8%), and drug therapy in 104 (7%) (ETable 2). Many second procedures likely represent planned contralateral treatment as 295 of 635 repeat TAE/TACEs occurred within 60 days of first treatment.

Figure 1.

Cohort Assembly

Table 1.

Patient Characteristics and Association with Survival

Characteristic	TACE + TAE n=1525	TACE n=577 (38%)	Multivariate HR, 95% CI	p-value
Treatment
TACE, no chemo code	-		1
TACE, with chemo code			0.78, 0.69–0.89	<0.001
Median Age (range)	72 (27–94)	72 (27–94)	-	-
Age
< 65 years	296 (19%)	108 (19%)	1
65–74 years	635 (42%)	259 (45%)	1.02, 0.85–1.22	0.82
75+ years	597 (39%)	210 (36%)	1.21, 1.00–1.46	0.06
Sex
Male	1,045 (68%)	390 (68%)	1
Female	483 (32%)	187 (32%)	1.03, 0.90–1.17	0.71
Race/Ethnicity
White	1,008 (66%)	374 (65%)	1
African American	136 (9%)	46 (8%)	0.94, 0.75–1.18	0.60
Asian	213 (14%)	92 (16%)	0.88, 0.68–1.15	0.35
Hispanic/Other	171 (11%)	65 (11%)	0.93, 0.75–1.15	0.48
Place of Birth
USA	820 (54%)	278 (48%)	1
Asia	203 (13%)	85 (15%)	0.67, 0.51–0.88	0.004
Other/Missing^	505 (33%)	214 (37%)	0.56, 0.49–0.65	<0.001
Hepatitis B*	147 (10%)	72 (12%)	0.89, 0.70–1.13	0.34
Hepatitis C*	529 (35%)	211 (37%)	1.08, 0.94–1.26	0.28
Alcohol*	182 (12%)	73 (13%)	0.99, 0.82–1.21	0.96
Other*	127 (8%)	51 (9%)	97, 0.77–1.22	0.78
Modified Charlson Score
0	330 (22%)	111 (19%)	1
1	440 (29%)	181 (31%)	1.01, 0.85–1.19	0.95
2+	758 (50%)	285 (49%)	1.06, 0.90–1.24	0.50
# of Cirrhosis Complications
0	1161 (76%)	464 (80%)	1
1	240 (16%)	78 (14%)	1.37, 1.16–1.62	<0.001
2+	127 (8%)	35 (6%)	1.51, 1.18–1.94	0.002
# of Pre-Diagnosis AFP Claims
0	639 (42%)	204 (35%)	1
1	456 (30%)	176 (31%)	0.79, 0.69–0.92	0.002
2+	433 (28%)	197 (34%)	0.72, 0.61–0.85	<0.001
Extent of Tumor
Single, no vascular invasion	639 (42%)	244 (42%)	1
Multiple, no vascular invasion	503 (33%)	207 (36%)	1.34, 1.17–1.54	<0.001
Single, vascular invasion	108 (7%)	38 (7%)	1.18, 0.93–1.52	0.17
Multiple, vascular invasion	130 (8%)	48 (8%)	1.62, 1.30–2.01	<0.001
Extrahepatic Extension or NOS	148 (10%)	40 (7%)	1.75, 1.40–2.19	<0.001
Maximum Tumor Size
≤3 cm	344 (23%)	150 (26%)	1
3–5 cm	375 (25%)	151 (26%)	1.26, 1.5–1.51	0.02
> 5 cm	568 (37%)	202 (35%)	1.81, 1.51–2.16	<0.001
Unknown	241 (16%)	74 (13%)	1.70, 1.36–2.13	<0.001
Teaching Hospital+	1,339 (88%)	508 (89%)	1.06, 0.87–1.30	0.55
Hospital NCI Designation in 2005
None	1,202 (80%)	471 (82%)	1
Clinical	36 (2%)	18 (3%)	1.18, 0.78–1.79	0.42
Comprehensive	272 (18%)	84 (15%)	0.75, 0.63–0.89	<0.001
Hospital with Solid Organ Transplant	1,076 (71%)	422 (74%)	0.72, 0.62–0.85	<0.001
Hospital Volume in Prior 12 months
<5	802 (52%)	258 (45%)	1
5–20	420 (27%)	165 (29%)	0.92, 0.79–1.07	0.26
>20	306(20%)	154 (27%)	0.88, 0.73–1.07	0.19

TACE, transarterial chemoembolization; AJCC, American Joint Commission on Cancer; NCI, National Cancer Institute; HR, Hazard Ratio. Notes: HR adjusted for variables shown and census track median income, SEER region.

^{^}n=360 missing place of birth and n=145 non-US non-Asian born.

^*Causes are not mutually exclusive. HR referent= no.

⁺Because of missing data, hospital variables do not sum to 1528.

Survival After TACE

Median survival in TACE patients was 21 months (95% CI 18–23), ranging from 24 months (95% CI 21–28) in patients with solitary tumors without vascular invasion to 11 months (95% CI 5–20) in patients with multiple tumors and vascular invasion. Ninety day mortality was high in patients with extrahepatic spread (25%) or multiple tumors with vascular invasion (21%). Stratification by pre-diagnosis AFP screening and codes for liver comorbidity further refined these estimates (Table 2, Figure 2).

Table 2.

Survival Estimates by Patient Subgroup Following Transarterial Chemoembolization

Patient Subgroup			TACE with CHEMOTHERAPY GROUP Overall Survival Estimates
Tumor Extent	Liver Comorbidity	AFP Screening	N	Median Survival (months), 95% CI	90 day n (%)	24 month† n (%)
Single, no vascular invasion	All	All	244	24 (21–28)	231 (95%)	112 (49%)
	No	No	79	20 (17–24)	71 (90%)	28 (38%)
	No	Yes	112	29 (25–41)	109 (97%)	63 (60%)
	Yes	No	<11	11 (1–33)	<11	<11
	Yes	Yes	43	23 (12-NA)	42 (98%)	17 (46%)
Multiple, no vascular invasion	All	All	207	21 (16–25)	191 (92%)	90 (46%)
	No	No	57	16 (8–24)	51 (89%)	21 (37%)
	No	Yes	109	25 (20–31)	100 (92%)	55 (54%)
	Yes	No	12	11.5 (7–23)	12 (100%)	<11
	Yes	Yes	29	23 (15–39)	28 (97%)	13 (48%)
Singe, vascular invasion	All	All	38	19.5 (12– 38)	33 (87%)	14 (42%)
Multiple, vascular invasion	All	All	48	11 (5– 20)	38 (79%)	14 (33%)
Any^, vascular invasion	No	No	30	14.5 (12–29)	25 (83%)	<11
	No	Yes	44	19 (8–46)	39 (89%)	18 (45%)
	Yes	No	<11	3.5 (2–5)	<11	0
	Yes	Yes	<11	5.5 (0–10)	<11	<11
Extrahepatic Extension or NOS	All	All	40	13.5 (5–18)	30 (75%)	<11

^†22 patients were censored at 24 months. Cells with <11 patients suppressed to preserve confidentiality.

^{^}Because of small numbers, patients with single and multiple tumors with any vascular invasion were combined for analysis by AFP and liver comorbidity. Also because of small sample size, further subgroup analysis was not performed on patients with extrahepatic spread.

Figure 2.

Kaplan-Meier Estimates of Survival Following Transarterial Chemoembolization. Survival estimates for patients with documented chemotherapy administration at time of TACE procedure for the entire cohort n=577 by tumor extent (A); and for single tumors without vascular invasion n=244 (B) and multiple tumors without vascular invasions n=207 (C) by AFP screening and liver comorbidity.

Patient sex, race/ethnicity, census tract median income, non-liver comorbidity, and cause of liver disease were not associated with survival (Table 1). Patients born in Asia or other non-US sites had significantly longer OS compared with US born patients—a possible surrogate for hepatitis B infection. Treatment at an NCI-designated Comprehensive Cancer Center and at a hospital with a solid organ transplant program was associated with significantly better OS.

Comparative Effectiveness

Median survival in Y⁹⁰ patients was 9 months (95% CI 1–41). In the PS-weighted population balanced across key clinical covariates (ETable 2), Y⁹⁰ was associated with an increased risk of death compared with TAE/TACE, adjusted HR 1.39 (95% CI 1.02–1.90). The effect was more pronounced when restricting just to TACE patients, adjusted HR 1.88 (95% CI 1.40–2.53, ETable 3). We found little evidence that the greater risk of death with Y⁹⁰ resulted from residual unmeasured confounding as trimming of patients treated contrary to prediction did not lead to an attenuation of the hazard ratio towards the null.

DISCUSSION

In this large cohort of patients with newly diagnosed HCC, the survival estimate of 21 months for the true chemoembolization patients suggest TACE effectiveness is retained across the broad range of health system settings where Medicare beneficiaries are treated in the United States; though treatment at a center with a solid organ transplant program or NCI Comprehensive status was associated with the best outcomes. However, we found that in the subgroup of TACE-treated patients whose disease was more extensive than is typically considered TACE amenable (extrahepatic disease or multifocal disease with macrocascular invasion), immediate post-treatment mortality was very high with 21 to 25% of patients dying within 90 days of treatment.

The expected duration of survival following TACE varies widely as it is dependent upon both the extent of cancer and the degree of underlying cirrhosis.[32–34] In well done[35] randomized clinical trials using conventional (non-bead) TACE for unresectable HCC, median OS has been reported between 13.8–28.7 months.[4, 5, 36] Better outcomes have been reported recently in patients with well compensated cirrhosis using the contemporary approach of doxorubicin-eluting bead TACE.[37–39] Based on the existing data, the Standards of Practice Committee of the Society of Interventional Radiology has set the quality threshold of median survival following TACE at 20 months.[8] While we found the median survival among Medicare beneficiaries treated with initial TACE to meet this quality threshold, two prior population-based observational studies also using SEER-Medicare reported post-TACE median survivals of only <12 and 14 months.[9, 10] The first of these investigations was conducted in the 1990s when only 4% of patients were treated with TACE. As use and techniques of TACE have evolved since the 1990s, it is likely that survival has also improved over this time. The more recent investigation of HCC patients treated between 2000 and 2005 was designed to compare palliative approaches in the elderly; it intentionally excluded patients younger than 65 and those receiving subsequent curative therapies. In contrast we chose to retain younger patients eligible for Medicare on the basis of disability to more broadly investigate the effectiveness of TACE in the United States. We also retained patients who underwent subsequent curative surgery, transplant and ablative procedures as excluding such patients underestimates TACE effect by selectively excluding those with the best response. Because of these differences we do not believe these data to be conflicting, rather, our report expands on this prior work by providing survival estimates across a broader range of treatment scenarios.

Because registry data do not contain laboratory data or performance status, we could not evaluate outcomes by HCC-relevant clinical categories of Child Pugh or BCLC stage. To address this limitation we used surrogates for the components of BCLC stage (extent of cancer, performance status, and Child Pugh score)[40] to refine our survival estimates. We used the cancer-specific tumor extent from SEER that includes macrovascular invasion. However, the predominantly clinical staging paradigm of HCC differs greatly from other cancers. Concordance between registrar-reported vascular invasion has not been evaluated to the best of our knowledge, but may be poor. Codes for complications of cirrhosis prior to diagnosis were used as a marker of the extent of cirrhosis. As this approach relies on the thoroughness of physician coding, these codes likely underestimate the extent of cirrhosis and limit the direct clinical applicability of the estimates provided for patients with and without liver comorbidity in this analysis. Pre-diagnosis AFP screening was chosen given prior evidence of cancer screening as a marker of frailty,[41] though AFP is likely a marker for multiple factors associated with improved outcomes including possible earlier detection with a lower burden of disease and engagement with the healthcare system. While these means of stratifying patients is less granular than validated HCC staging systems, each component was strongly associated with survival.

Fifteen percent of TACE-patients in this cohort had disease beyond that for which TACE is currently recommended,[7] including multifocal disease with macrovascular invasion and extrahepatic disease. In this subgroup of patients with advanced cancer, 90 day mortality was very high, 21–25%. Because untreated patients are inherently different, we did not compare outcomes between treated and untreated patients to explore the extent to which this early mortality was the result of hepatic decompensation following treatment or ineffective treatment in terminally ill individuals. In this sicker group of patients it would be particularly relevant to know more detail about the technical aspects of TACE, specifically whether the procedure was performed using super selective techniques that minimize the damage to non-neoplastic liver, or whether a more extensive embolization was performed. Unfortunately this quality metric cannot be ascertained from embolization codes in any reliable fashion. However, if indeed patients with poorly compensated cirrhosis were treated with less selective embolization this might account for the poor outcomes of TACE in this advanced disease group. But, without this information we cannot definitively conclude whether it is all TACE that does not benefit patients with advanced disease, or merely poor quality TACE.

Patients with a poor performance status, diffuse infiltrative cancer, Child-Pugh C cirrhosis, and Child Pugh B cirrhosis with portal vein thrombosis all have an exceptionally poor prognosis even with TACE or Y⁹⁰.[11, 32, 33] Unfortunately, there are few data available to help determine whether LRT improves outcomes in this high risk group, or if they are better served by sorafenib or supportive care. Studies designed to evaluate if TACE and/or Y⁹⁰ improve outcomes over supportive care or sorafenib such as NCT01887717 (Y⁹⁰ versus sorafenib in patients with portal vein occlusion) are difficult to complete, but of paramount importance.

Our exploration of the comparative effectiveness of Y⁹⁰-radioembolization to TACE was limited by a small number of Y⁹⁰ treated patients; however we found no suggestion that Y⁹⁰ offered a survival advantage over TACE. The median survival of 9 months was well below what has been reported for patients with compensated cirrhosis following Y⁹⁰-radioembolization,[11, 12, 15, 21, 32] and may simply reflect use of Y⁹⁰ in patients with more advanced liver disease and a larger burden of cancer.

Our study also shows that while registry-linked claims data allow for a broad overview of outcomes of HCC treatment, the lack of HCC-relevant staging—specifically laboratory parameters and components of Child Pugh score—markedly limits the feasibility of a nuanced study. As HCC is one of a few cancers with a rising incidence and mortality in the US,[42] efforts should be made to incorporate laboratory parameters relevant to HCC staging into mandatory cancer registry reporting. Such efforts would improve the ability to study the comparative safety and effectiveness of emerging therapies for this deadly disease.

Figure 3.

Kaplan-Meier Estimates of Survival by Treatment. Survival estimates for TACE and yttrium^90-radioembolization are shown for the propensity score weighted population.