Radioembolization results in longer time-to-progression and reduced toxicity compared with chemoembolization in patients with hepatocellular carcinoma

Abstract

Background and Aims

Chemoembolization is a standard treatment for hepatocellular carcinoma (HCC). Radioembolization with ⁹⁰Y microspheres is a new, transarterial approach to radiation therapy. We performed a comparative effectiveness analysis of these therapies in patients with HCC.

Methods

We collected data from 463 patients who were treated with transarterial locoregional therapies (chemoembolization or radioembolization) over a 9-year period. We excluded patients who were not appropriate for comparison and analyzed data from 245 (122 who received chemoembolization and 123 who received radioembolization). Patients were followed for signs of toxicity; all underwent imaging analysis at baseline and follow-up timepoints. Overall survival was the primary outcome measure. Secondary outcomes included safety, response rate, and time-to-progression. Uni- and multi-variate analyses were performed.

Results

Abdominal pain and increased transaminase activity were more frequent following chemoembolization (P<.05). There was a trend that patients treated with radioembolization had a higher response rate than with chemoembolization (49% vs. 36%, P=0.104). Although time-to-progression was longer following radioembolization than chemoembolization (13.3 months vs 8.4 months, P=0.046), median survival times were not statistically different (17.4 months vs 20.5 months, P=0.232). Among patients with intermediate-stage disease, survival was similar between groups that received chemoembolization (17.5 months) and radioembolization (17.2 months, P=0.42).

Conclusion

Patients with HCC treated by chemoembolization or radioembolization with ⁹⁰Y microspheres had similar survival times. Radioembolization resulted in longer time-to-progression and less toxicity than chemoembolization. Post-hoc analyses of sample size indicated that a randomized study with >1000 patients would be required to establish equivalence of survival times between patients given the different therapies.

Introduction

Hepatocellular carcinoma (HCC) claims over half a million lives annually¹. While surgical treatments (transplantation/resection) provide the best curative outcomes, most patients present at an advanced stage². Systemic targeted therapies play a role in advanced disease^{3, 4}. Locoregional therapies (LRTs) have demonstrated therapeutic efficacy for HCC in selected patients, with chemoembolization conveying a survival benefit. Chemoembolization now represents one of the standards of care [Barcelona Clinic Liver Cancer (BCLC) stage B]. Yttrium-90 microspheres (⁹⁰Y radioembolization) has regulatory approval for liver cancer and is recognized as a treatment option for HCC by the National Comprehensive Cancer Network guidelines as well as the recent Consensus Recommendations of the National Cancer Institute Clinical Trials Planning Meeting^{5, 6}.

We recently demonstrated survival outcomes following ⁹⁰Y that are comparable to chemoembolization⁷. Given these findings, our center recently initiated a phase 2 randomized controlled trial (RCT) with response rate (RR)/time-to-progression (TTP) as endpoints comparing ⁹⁰Y to chemoembolization. However, it has become clear that in order to show survival equivalence between these two interventions, a prohibitively large number of patients would be required. Given this, we undertook a comprehensive comparative effectiveness analysis between ⁹⁰Y and chemoembolization using recently described methodology^7-12.

Methods

463 consecutive HCC patients were treated with either chemoembolization or ⁹⁰Y at our institution over a 9-year period. Patients were considered for transarterial therapies if they exhibited unresectable HCC (determined by transplant surgery) and bilirubin <3.0 mg/dL. Criteria used to create the comparison cohorts for this analysis were based on excluding patients who: a) were treated with both chemoembolization and ⁹⁰Y at any time, b) exhibited portal vein thrombosis (PVT); PVT is a well-recognized approved indication for ⁹⁰Y and a contraindication for chemoembolization (risk of ischemic hepatitis), c) exhibited extrahepatic metastases (confounding variable), or d) lacked imaging follow-up from expedited transplantation (see Figure 1: study flow chart). This resulted in the identification of 245 patients that could have been safely treated with either option (chemoembolization N=122, ⁹⁰Y N=123); they represent the study group. This project was compliant with the Health Insurance Portability and Accountability Act, approved by the Northwestern University Institutional Review Board, and was registered (NCT00530010). The data were closed on December 31, 2008.

Figure 1.

Study flow chart.

Evaluation and Staging

Diagnostic criteria for HCC included biopsy or radiographic findings as defined by guidelines¹³. Baseline staging was performed by Child-Pugh (liver function), United Network for Organ Sharing (UNOS TNM) (based on tumour size/number) and BCLC classification systems (composite of function, tumour size/number and symptoms) (eTable 1)¹³. Patients were classified as BCLC C if they exhibited HCC-related symptoms (e.g. tumour-related capsular or referred pain, cancer-related symptoms). The decision to treat with chemoembolization or ⁹⁰Y was consensus-based during weekly HCC conference represented by medical oncology, hepatology, transplant surgery and interventional radiology.

BCLC is one of several staging systems used for HCC; it has been endorsed as the preferred system^{14, 15}. This system groups patients into six potential treatments: curative options for early disease (BCLC A: resection, transplantation, ablation), palliative options for intermediate (BCLC B: chemoembolization) or advanced disease (BCLC C: sorafenib), and supportive care for end-stage disease (BCLC D)¹³. However, HCC heterogeneity at presentation results in individualized patient decision-making that often does not permit the simple assignment of the suggested treatment within a BCLC stage; these patients usually receive other treatment (often chemoembolization) based on organ availability, patient-specific factors, contraindications and institutional practices. Despite this, we used the BCLC construct as our rationale for selecting chemoembolization as the standard of care and hence, the comparison group for radioembolization¹³.

Chemoembolization

Patients were admitted on the day of treatment and received prophylactic intravenous antibiotics. Chemoembolization was performed using 30 mg mitomycin, 30 mg adriamycin and 100 mg cisplatin mixed with lipiodol, followed by arterial embolization using permanent occlusive 300-500 micron particles. Patients were discharged 1-2 days following chemoembolization with a 5-day course of antibiotics and a 2-week course of analgesics and antiemetics as needed¹⁶.

Yttrium-90 Radioembolization

The microspheres used were glass-based (TheraSphere®, Ottawa, Canada) and are comprised of 20-30 micron beta-emitting particles. Pre-treatment mesenteric angiography and technetium-99m macroaggregated albumin scans were performed to assess gastrointestinal flow and lung shunting¹⁷. Patients were discharged 2-6 hours after treatment without the need for hospitalization on a prophylactic 5-day course of proton-pump inhibitors¹⁷.

Toxicity and Clinical Follow-up

Clinical and laboratory adverse events were recorded using National Cancer Institute Common Terminology Criteria v3.0 during routine clinic visits. Adverse events were assessed one month after treatment, and then every 2-3 months coinciding with imaging follow-up. All toxicities (laboratory, clinical) reported at anytime following treatment (no 30-day time cut-off) were recorded and are reported herein. Patients were followed until death; final date of death was confirmed by using the social security death index and/or direct confirmation by family members. Otherwise, patients were censored at the last known clinic follow-up.

AFP Analysis

Patients with baseline alpha-fetoprotein (AFP) >200 ng/ml were included in an exploratory tumour marker subanalysis. Baseline AFP was compared to the nadir per previous methodology¹⁸. Rates of >50% reduction, >90% reduction and complete normalization (AFP ≤13 ng/ml) were determined.

Imaging Analysis

Patients were followed using triphasic computerized tomography or contrast-enhanced magnetic resonance imaging one month following treatment and at scheduled 2-3 month intervals on protocol. Response rate (RR) was assessed using World Health Organization (WHO) size and European Association for Study of the Liver (EASL) necrosis criteria (eTable 2) ^{3, 14, 19-21}.

Statistical Analyses and Survival

Baseline patients characteristics between the two groups were compared using the t-test (means) or Mann-Whitney U test (median) for continuous variables, and Chi-square or Fisher exact test (for small n or highly imbalanced table cells) for categorical data. Data were summarized using descriptive statistics (mean and standard deviation for continuous variables; count and frequency for categorical variables). Time-to-response, TTP and survival analyses were performed from the date of first LRT and were censored to curative therapy^{7, 13}. A TTP endpoint was defined as any of the following: progression by WHO, EASL, UNOS stage, development of any portal vein thrombosis (tumour or bland), appearance of new lesions or extrahepatic metastases (eTable 1). Sub-stratification analyses by BCLC and Child-Pugh stage were also performed. Univariate (Kaplan-Meier) and multivariate analyses (Cox proportional hazards) were conducted to compare survival between groups^{22, 23}. In order to interpret the data most conservatively, P-values <0.1 on univariate analyses were corrected for multiple comparisons using Bonferroni methodology.

The following variables were entered into the multivariate model: age, gender, Eastern Cooperative Oncology Group (ECOG) performance status, AFP, Child-Pugh class, UNOS stage and treatment received (chemoembolization or ⁹⁰Y). Factors were included in the multivariate model if P<.25 in univariate analysis (unadjusted for multiple comparisons). Individual components of composite variables were excluded from multivariate analyses (see Table 4 legend, eTable 1, eTable 3 legend).

Table 4. Uni/Multivariate Analyses (Survival).

		UNIVARIATE ANALYSIS (Kaplan Meier and Logrank Test)			MULTIVARIATE ANALYSIS (Cox Proportional Hazards Model)**
Predictor	Category	Hazard ratio (CI)	P	Adj. P*	Hazard ratio (CI)	P
Age	<65	1.70 (1.16-2.50)	0.007	0.091	1.29 (0.83-1.99)	0.260
	≥65	1.00			1.00
Gender	Female	1.40 (0.89-2.17)	0.138		1.69 (1.06-2.68)	0.027
	Male	1.00			1.00
Ethnicity	African-American	0.98 (0.58-1.67)	0.941		-	-
	Hispanic	1.06 (0.58-1.92)	0.849		-	-
	Asian	0.72 (0.38-1.37)	0.319		-	-
	Caucasian	1.00			-	-
Child-Pugh Class***	A	0.79 (0.54-1.15)	0.215		0.58 (0.39-0.88)	0.010
	B/C	1.00			1.00
Baseline Bilirubin	<2 mg/dL	1.22 (0.69-2.15)	0.490		-	-
	≥2 mg/dL	1.00			-	-
Baseline Albumin	>3.5 mg/dL	0.55 (0.29-1.07)	0.077		-	-
	≤3.5 mg/dL	1.00			-	-
Ascites	Absent	0.56 (0.35-0.88)	0.011	0.143	-	-
	Present	1.00			-	-
ECOG Performance Status	0	0.50 (0.34-0.74)	0.005	0.065	0.57 (0.37-0.87)	0.009
	>0	1.00			1.00
AFP (ng/mL)	≤200	0.59 (0.40-0.87)	0.008	0.104	0.77 (0.51-1.16)	0.209
	>200	1.00			1.00
UNOS Stage***	T1/T2	0.37 (0.23-0.60)	<.001	<.001	0.33 (0.19-0.56)	<0.001
	T3	0.45 (0.29-0.70)	<.001	0.005	0.48 (0.30-0.77)	0.003
	T4a	1.00			1.00
Maximum Baseline Dimension	≤5 cm	0.78 (0.46-1.33)	0.366		-	-
	5-10 cm	1.07 (0.61-1.88)	0.812		-	-
	>10 cm	1.00			-	-
Solitary/Multifocal	Solitary	0.56 (0.37-0.83)	0.004	0.052	-	-
	Multifocal	1.00			-	-
Treatment	Radioembolization	0.80 (0.55-1.16)	0.232		1.06 (0.70-1.62)	0.780
	Chemoembolization	1.00			1.00

Abbreviations: CI, confidence interval; EASL, European Association for Study of the Liver; ECOG, Eastern Cooperative Oncology Group; HR, hazard ratio; UNOS, United Network for Organ Sharing; WHO, World Health Organization.

^*P-value adjusted for multiple comparisons (correction factor n=13).

^**Factors were included in multivariate analysis if P <.25 in univariate analysis (unadjusted for multiple comparisons).

^***Only 1 variable was chosen if there was overlapping of factors of the categorical variables; ie, (1) ascites, baseline bilirubin and baseline albumin were excluded because these variables are captured in Child-Pugh class; (2) distribution (solitary/multifocal) and maximum baseline dimension were excluded because these variables are captured in UNOS stage.

Hazard ratio estimates were based on simultaneous analysis of all predicated variables. Assumption of proportionality was tested using log-minus-log plot and was met. Uni/multivariate analyses were also performed in the BCLC B subgroup. This study was powered (80%) to detect a hazard ratio of >1.45 or <0.69 between ⁹⁰Y and chemoembolization using two-sided alpha of 0.05 and median survival time of 17.5 months with chemoembolization. All analyses were conducted using SAS 9.2 (SAS, Cary, NC); a P-value <.05 was considered statistically significant.

Results

Patient Sample and Baseline Characteristics

Table 1 summarizes the baseline patient demographics. ⁹⁰Y patients were older (66 vs. 61 yrs, P<.001). Both groups were treatment naïve (>90%) and had comparable rates of portal hypertension, ascites, cirrhosis, tumour distribution, bilirubin and cancer stage (UNOS, BCLC and Child-Pugh).

Table 1. Baseline Patient Characteristics.

Characteristic N (%)		TACE N=122	90Y N=123	P
Demographics
Age (years)	Median	61	66	<.001
	Range	33-88	30-88
Ethnic Group	Caucasian	79 (65)	91 (74)	0.40
	Asian	13 (11)	12 (10)
	Hispanic	15 (12)	9 (7)
	African-American	15 (12)	11 (9)
Gender	Male	102 (84)	87 (71)	0.14
	Female	20 (16)	36 (29)
Etiology	Alcohol	21 (17)	20 (16)	0.24
	HCV	56 (46)	42 (35)
	HCV + Alcohol	2 (2)	6 (4)
	HBV	12 (10)	13 (11)
	NASH	1 (1)	5 (4)
	Cryptogenic	25 (20)	28 (23)
	Unknown	5 (4)	9 (7)
Method of Diagnosis	Imaging	71 (58)	68 (55)	0.49
	Biopsy	51 (42)	55 (45)
Prior Failed Therapies	None	112 (92)	112 (91)	0.81
	RFA	6 (5)	8 (7)
	Resection	4 (3)	3 (2)
Imaging Characteristics
Portal Hypertension	Present	98 (80)	90 (73)	0.22
	Absent	24 (20)	33 (27)
Ascites	Present	21 (17)	16 (13)	0.38
	Absent	101 (83)	107 (87)
Cirrhosis	Present	112 (92)	106 (86)	0.22
	Absent	10 (8)	17 (14)
Lobar Distribution	Unilobar	80 (66)	79 (64)	0.9
	Bilobar	42 (34)	44 (36)
Distribution	Solitary	57 (47)	55 (45)	0.85
	Multifocal	65 (53)	68 (55)
Largest Tumour Size (cm)	Median (Inter-quartile range)	3.6 (2.6-5.7)	4.5 (3.1-6.6)	0.10
AFP (ng/mL)	>200	38 (30) *	41 (33) *	0.82
	≤200	84 (70)	82 (67)
Staging
UNOS	T1	4 (3)	1 (1)	0.44
	T2	46 (38)	44 (36)
	T3	35 (29)	43 (35)
	T4a (≥4 tumours)	37 (30)	35 (28)
BCLC	A	47 (39)	43 (35)	0.95
	B	61 (50)	65 (53)
	C	12 (9)	13 (10)
	D	2 (2)	2 (2)
Child-Pugh	A	67 (55)	67 (54)	0.99
	B	53 (43)	54 (44)
	C	2 (2)	2 (2)

^*Follow-up AFP was not available in 1 chemoembolization and 1 radioembolization patient, TACE=chemoembolization

Treatment Characteristics

The median number of chemoembolization treatments was 2 [inter-quartile range (IQR):1-3]; it was 1 for ⁹⁰Y (IQR:1-2) (P=0.09). Mean days hospitalized per treatment session was 1.8 (range:1-11) for chemoembolization and 0 for ⁹⁰Y (P<.001). Mean cumulative days hospitalized was 3.4 (range:1-22) for chemoembolization and 0 for ⁹⁰Y (P<.001). Median administered liver and lung radiation doses of ⁹⁰Y were 110 Gy [95% confidence interval (CI):102-115] and 3.7 Gy (CI:2.9-5), respectively. For ⁹⁰Y, prophylactic coil embolization of non-target vessels was performed in 40/123 patients (33%); this rate was consistent throughout the study period.

Clinical/Laboratory Toxicities

Table 2 lists the clinical and laboratory toxicities. Although fatigue and fever were more common following ⁹⁰Y and diarrhea was more common with chemoembolization, these observations were not maintained upon P-value correction. Both abdominal pain (P<.001) and elevated transaminases (29% vs. 11%, P=0.004) were more common with chemoembolization. There were no gastrointestinal ulcers.

Table 2. Toxicity Analyses.

Characteristic		TACE N=122	90Y N=123	P	Adj. P*
Clinical Toxicities
Fatigue	Total	47 (38)	68 (55)	0.012	0.074
	Child-Pugh A	28/67 (42)	41/67 (62)	0.038	0.226
	Child-Pugh B/C	19/55 (35)	27/56 (48)	0.179
Abdominal Pain	Total	46 (38)	18 (15)	<.001	<.001
	Child-Pugh A	25/67 (37)	5/67 (7)	<.001	<.001
	Child-Pugh B/C	21/55 (38)	13/56 (23)	0.818
Nausea/Vomiting	Total	25 (20)	18 (15)	0.299
	Child-Pugh A	14/67 (21)	11/67 (16)	0.658
	Child-Pugh B/C	11/55 (20)	7/56 (13)	0.314
Anorexia	Total	16 (13)	13 (11)	0.675
	Child-Pugh A	8/67 (12)	7/67 (10)	1.000
	Child-Pugh B/C	8/55 (15)	6/56 (11)	0.580
Fever/Chills	Total	2 (2)	10 (8)	0.034	0.205
	Child-Pugh A	2/67 (3)	9/67 (13)	0.055	0.328
	Child-Pugh B/C	0/55 (0)	1/56 (2)	1.000
Diarrhea	Total	10 (8)	2 (2)	0.019	0.115
	Child-Pugh A	9/67 (13)	2/67 (3)	0.055	0.328
	Child-Pugh B/C	1/55 (2)	0/56 (0)	1.000
Grade 3-4 Laboratory Toxicities
Bilirubin	Total	25 (20)	20 (16)	0.490
	Child-Pugh A	11/67 (16)	8/67 (12)	0.621
	Child-Pugh B/C	14/55 (25)	12/56 (21)	0.659
Alkaline Phosphatase	Total	3 (2)	3 (2)	1.000
	Child-Pugh A	1/67 (2)	1/67 (2)	1.000
	Child-Pugh B/C	2/55 (4)	2/56 (4)	1.000
ALT/AST	Total	36 (29)	14 (11)	0.0008	0.004
	Child-Pugh A	18/67 (27)	10/67 (15)	0.136
	Child-Pugh B/C	18/55 (33)	4/56 (7)	0.0008	0.004
Albumin	Total	26 (21)	15 (12)	0.082	0.41
	Child-Pugh A	10/67 (15)	4/67 (6)	0.156
	Child-Pugh B/C	16/55 (29)	11/56 (19)	0.275
Creatinine	Total	4 (3)	4 (3)	1.000
	Child-Pugh A	1/67 (2)	1/67 (2)	1.000
	Child-Pugh B/C	3/55 (4)	3/56 (6)	1.000

^*P-value adjusted for multiple comparisons (correction factor n=6 and 5 for clinical and laboratory toxicities respectively); TACE=chemoembolization

Alpha-fetoprotein Subanalysis

Both chemoembolization (N=37) and ⁹⁰Y (N=40) resulted in significant reduction of AFP [(chemoembolization: baseline 985 ng/ml (IQR:445-4148); nadir 390 ng/ml (IQR:66-1606), (P=0.008)]; [⁹⁰Y: baseline 1246 ng/ml (IQR:593-9205); nadir 73 ng/ml (IQR:27-2321), (P<.001)]. 22 (59%) and 32 (80%) of the chemoembolization and ⁹⁰Y patients exhibited >50% AFP reduction (P=0.086), with a median time to >50% AFP reduction of 2.6 months (CI:2.1-3.9) for ⁹⁰Y and 3.9 months (CI:2.8-7.5) for chemoembolization. 14 (38%) and 19 (48%) of the chemoembolization and ⁹⁰Y patients exhibited >90% AFP reduction, respectively (P=0.532). Complete normalization was seen in 4 (11%) and 9 (22%) of the chemoembolization and ⁹⁰Y patients, respectively (P=0.288).

Imaging Outcomes

Response rate

Table 3 lists the imaging outcomes stratified by stage. 1065 scans were reviewed (mean: 4.3 scans/patient). RR by WHO trended towards ⁹⁰Y over chemoembolization (49% vs. 36%, P=0.104). Median time to WHO partial response was shorter with ⁹⁰Y than chemoembolization (6.6 vs. 10.3 months, P=0.050). By EASL, RRs were similar (chemoembolization: 69%, ⁹⁰Y: 72%, P=0.748), with faster time to EASL response with ⁹⁰Y (1.2 vs. 2.2 months, P=0.016).

Table 3. Imaging and Survival Outcomes.

Characteristic		TACE N=122	90Y N=123	P	Adj. P*	TACE N=122	90Y N=123	P	Adj. P*
		WHO Response				EASL Response
Time to Response (Months) (CI)		10.3 (7.7-16)	6.6 (4.2-8.6)	0.025	0.050	2.2 (1.5-3.0)	1.2 (1.1-2.1)	0.008	0.016
OVERALL RESPONSE RATE		44/122 (36)	60/123 (49)	0.052	0.104	84/122 (69)	88/123 (72)	0.748
Child-Pugh A	Overall	25/67 (37)	36/67 (54)	0.082	0.164	43/67 (64)	52/67 (78)	0.128
	T1/T2	7/24 (29)	12/22 (55)	0.134		20/24 (83)	19/22 (86)	1.00
	T3	9/18 (50)	15/24 (63)	0.533		12/18 (67)	21/24 (88)	0.139
	T4a (≥ 4 tumours)	9/25 (36)	9/21 (43)	0.764		11/25 (44)	12/21 (57)	0.554
Child-Pugh B	Overall	19/53 (36)	24/54 (44)	0.478		40/53 (75)	36/54 (67)	0.429
	T1/T2	9/25 (36)	8/22 (36)	1.00		19/25 (76)	15/22 (68)	0.745
	T3	4/16 (25)	11/19 (58)	0.086	0.172	12/16 (75)	16/19 (84)	0.677
	T4a (≥ 4 tumours)	6/12 (50)	5/13 (38)	1.00		9/12 (75)	5/13 (38)	0.111
Child-Pugh C	Overall	0/2 (0)	0/2 (0)	-	-	1/2 (50)	0/2 (0)	-
BCLC	A	15/47 (32)	20/43 (47)	0.229		38/47 (81)	33/43 (77)	0.827
	B	27/61 (44)	33/65 (51)	0.581		40/61 (66)	46/65 (71)	0.664
	C	2/12 (17)	7/13 (54)	0.097	0.194	5/12 (42)	9/13 (69)	0.238
	D	0/2 (0)	0/2 (0)	-		1/2 (50)	0/2 (0)	-
		Time-to-Progression				Survival
OVERALL (Months) (CI)		8.4 (7.3-10.6)	13.3 (9.3-25)	0.023	0.046	17.4 (13.9-18.7)	20.5 (15.7-29.1)	0.232
Child-Pugh A	Overall	9.6 (7.6-11.7)	21.9 (10.8-27.1)	0.19		17.5 (14.8-20.7)	22.1 (17.2-32.6)	0.13
	T1/T2	11.3 (7.2-32.9)	27.1 (8, -)	0.8		- (11.4, -)	26.9 (17.1-27.4)	0.7
	T3	12.8 (6.1, -)	21.9 (10.8-25.9)	0.42		19.2 (13.9-31.5)	35.7 (18.2, -)	0.049	0.098
	T4a (≥ 4 tumours)	9.4 (7.3-11.5)	8.6 (5, -)	0.6		15.1 (12.6-18.2)	14.9 (7.7-22.1)	0.68
Child-Pugh B	Overall	7.7 (5.8-10.7)	13 (8.4-25.1)	0.06	0.12	17.4 (10.7-24.4)	17.0 (11.2-30.2)	0.7
	T1/T2	6.8 (4.3-20.2)	13 (6.2-25.1)	0.19		26 (12.8-46.1)	29.1 (17, -)	0.68
	T3	7 (3.1-19.6)	17.7 (8.4-33.8)	0.057	0.114	17.4 (10.6-26.4)	38.3 (11.4-41.6)	0.8
	T4a (≥ 4 tumours)	7.7 (5.8, -)	5.4 (3, -)	0.56		10.0 (4.7-17.9)	6.8 (5.3-13.6)	0.97
Child-Pugh C	Overall	4.4 (-, -)	1.6 (-, -)	0.7		11.3 (-, -)	2.4 (-, -)	0.09	0.18
BCLC	A	8.8 (6.8-20.2)	25.1 (8-27.2)	0.4		45.4 (15.1-46.1)	27.3 (17.1-30.2)	0.74
	B	9.4 (7.2-11.5)	13.3 (8.4-25.9)	0.047	0.094	17.5 (14.8-18.7)	17.2 (11.4-29.6)	0.42
	C	7.9 (3.6, -)	13.8 (10.8, -)	0.38		9.3 (6.2-11.5)	22.1 (11.2, -)	0.04	0.08
	D	4.4 (-, -)	1.6 (-, -)	0.7		11.3 (-, -)	2.4 (-, -)	0.09	0.18

^*P-value adjusted for multiple comparisons (correction factor n=2); TACE=chemoembolization

Time-to-Progression

96 patients progressed (⁹⁰Y: 42, chemoembolization: 54). By median TTP, ⁹⁰Y outperformed chemoembolization (13.3 months vs. 8.4 months, P=0.046).

Survival

Table 3 also reports survival. The number of chemoembolization and ⁹⁰Y patients censored was 44 and 31, respectively (transplantation N=73, resection N=2). 113 patients died (chemoembolization: 59, ⁹⁰Y: 54) during the median follow-up time of 32.6 months for chemoembolization and 22.7 months for ⁹⁰Y. Overall, both treatments resulted in similar overall survival (Figure 2). Exceptions to this were noted in two subgroups; there was improved survival with ⁹⁰Y in Child-Pugh A T3 and BCLC C patients (Table 3) on univariate analysis; this was not maintained when correcting for multiple comparisons. BCLC B patients exhibited nearly identical survival times (17.5 months vs. 17.2 months, P=0.42). Survival was also not different between the groups after excluding patients that had been censored to curative therapies, or by intention-to-treat analysis when comparing the groups prior to excluding those that had no follow-up imaging (chemoembolization N=150, ⁹⁰Y N=126; see study flow chart, survival data not shown).

Figure 2.

Survival distribution function by treatment group adjusted for covariates (P = .7803).

Uni/Multivariate Analyses

Uni/multivariate analyses are presented in Table 4. Female gender [Hazard ratio (HR): 1.69, CI:1.06-2.68], ECOG 0 (HR:0.57, CI:0.37-0.87), Child-Pugh class A (HR: 0.58, CI:0.39-0.88), UNOS T3 (HR:0.48, CI:0.30-0.77) and UNOS T1/T2 (HR:0.33, CI:0.19-0.56) were significant prognosticators of survival. Treatment received (chemoembolization or ⁹⁰Y) was not (HR:1.06, CI:0.70-1.62) (Figure 2). In the BCLC B subanalysis (eTable 3), Child-Pugh class A, ECOG 0 and UNOS T3 were significant prognosticators of survival. Again, treatment received was not (HR:0.86, CI:0.50-1.47).

Testing for Survival Equivalence: Post Hoc Projected Sample Size Analysis

Based on these results and the lack of clear survival difference between treatment groups, we conducted a post-hoc sample size estimate needed for a RCT testing for therapeutic (survival) equivalence. Assuming 17.5 months median survival for chemoembolization (standard of care) and a clinical equivalence limit set at 2 months (lower limit: 15.5 months, upper limit: 19.5 months; 60 months follow-up), approximately 870 would be needed to demonstrate survival equivalence with an 80% power and a type I error of 0.05. This increases to >1000 patients if a screening failure rate of 15% is applied.

Discussion

HCC is a condition where many patients present beyond potentially curative options². Sorafenib has been shown to extend survival in advanced disease^{3, 4}. Radiofrequency ablation for small tumours is potentially curative, although large RCTs comparing to resection are lacking²⁴. In non-ablatable disease, palliation with chemoembolization in selected patients improves survival when compared to best supportive care²⁵. Chemoembolization has therefore become the transarterial standard of care for unresectable HCC patients^{12, 13}.

Although chemoembolization and ⁹⁰Y are both delivered through the hepatic artery, they are mechanistically quite different. Chemoembolization involves the injection of chemotherapy into liver tumours with a macroembolic effect and arterial occlusion²⁶. This results in a cascade of events including overexpression of hypoxia-inducible factor 1α and vascular endothelial growth factor²⁷. Clinically, this is manifest as pain and post-embolization syndrome, often requiring anti-inflammatories, narcotics and hospitalization for management²⁷. Permanent occlusion of hepatic vessels is also a possibility. In contradistinction, ⁹⁰Y involves the hepatic injection of radioactive particles without arterial occlusion (microembolic effect)²⁸. Hence, there is no hypoxia-initiated cascade; clinical sequelae are derived from radiation and include post-radioembolization syndrome, dominated by fatigue managed without hospitalization. Vessel patency also permits subsequent arterial treatment if necessary.

⁹⁰Y represents a novel transarterial brachytherapy device with regulatory approval for the treatment of HCC with or without PVT, or as a bridge to surgery or transplantation. Our center implemented this therapy after observing enhanced tolerability in elderly patients (no post-embolization syndrome), the convenience of outpatient treatment, and its unique role in PVT. This represents a paradigm shift from chemoembolization, where hospitalization is necessary and PVT is a contraindication¹⁵. In the presence of PVT, the hepatic artery may be the sole source of blood supply to the liver; a microembolic therapy such as ⁹⁰Y provides a new option for patients with compromised portal flow²⁸. Further studies in PVT are required, particularly since systemic agents have shown survival benefit in advanced HCC^{3, 4, 29}.

We and others have demonstrated that ⁹⁰Y can be used safely and effectively with survival outcomes similar to chemoembolization^{7, 30, 31}. Although a RCT comparing ⁹⁰Y to chemoembolization would be ideal, enrollment requirements for such a trial may compromise its feasibility. Therefore, while a randomized phase 2 study is underway at our institution, we chose to foreshadow potential challenges of a phase 3 RCT (survival) by performing a comparative effectiveness analysis. Comparative effectiveness research is being proposed as a methodology to gain insight into treatment options that reflect “real-world” clinical practices and patient treatment algorithms^{9, 10, 32}.

Safety

Although both groups experienced fatigue, nausea and anorexia, differences in toxicities were noted. Chemoembolization patients were more likely to experience abdominal pain (P<.001). Furthermore, similar to previous reports, chemoembolization patients exhibited significantly higher hepatic transaminase elevation (P=0.004)^{33, 34}. Although ⁹⁰Y patients received fewer treatments, this did not reach statistical significance.

⁹⁰Y patients were significantly older than those undergoing chemoembolization. Despite this, our data showed that ⁹⁰Y was a well-tolerated outpatient treatment with no hospitalizations, underscoring a lower need for in-patient resource utilization.

AFP Response Subanalysis

Both treatments resulted in a significant reduction of AFP. This finding has been shown to prognosticate therapeutic benefit following LRT and chemotherapy^{18, 35}.

Imaging Outcomes

Time-to-response was notably shorter with ⁹⁰Y. Comparing with the standard of care (chemoembolization), the WHO RR favored ⁹⁰Y but was not statistically significant. The most striking difference was in overall TTP, where ⁹⁰Y significantly outperformed chemoembolization. However, prolonged TTP following ⁹⁰Y did not translate directly into improved survival. In advanced disease where survival is limited (10-11 months), TTP has been correlated with survival³. However, in this patient cohort where overall survival was slightly longer (16-20 months), prolonging TTP may not have had the same effect. One potential explanation might rest in survival time; survival in these patients may have been sufficiently long for any potential survival benefit from delaying tumour growth (prolonged TTP) to be offset by the deleterious effects of chronic background cirrhosis and liver failure. Despite this, TTP is still clinically relevant; delaying TTP as a bridge to transplantation might decrease drop-out rates and improve the chance of a life-saving transplant before death from cirrhosis. This finding favoring ⁹⁰Y is consistent with two recent radiologic-pathologic analyses^{36, 37}

Survival

Survival data demonstrated similar outcomes between the groups on both univariate and multivariate analyses for the entire cohort as well as BCLC B patients. As previously reported, Child-Pugh A T3 and BCLC C patients demonstrated a longer survival with ⁹⁰Y over chemoembolization³¹. This subgroup observation of survival difference should be interpreted with caution given the small sample size and loss of statistical significance following P-value correction.

Future Studies

Radioembolization presents an evolving therapeutic tool in the management of HCC. However, econometric cost analyses comparing in-patient chemoembolization (hospitalization, chemotherapy) and outpatient ⁹⁰Y (device) will need to be performed, particularly since they appear to exhibit similar patient survival. Although our data did not find statistically significant differences in survival between ⁹⁰Y and chemoembolization, we cannot conclude survival equivalence.

One of the questions we set out to answer was whether a RCT comparing these two treatments was feasible. Given our experience, we believe several factors limit the feasibility of such a study. HCC is unique in that survival as an outcome is dependent on two co-existing conditions (malignancy and cirrhosis); death may be caused by either factor. Therefore, the choice of a staging system incorporating these variables is essential, especially in the context of treatment decisions. As demonstrated in this study, survival was similar between the two treatments. However, survival equivalence would be difficult to confirm since >1000 patients would be required. Such a study would require a multi-center approach, itself presenting a problem given the current lack of technical and methodological standardization of what is considered the current standard of care-chemoembolization. Studies in HCC necessitating this magnitude of enrollment are rare. As an example, the seminal study of sorafenib in advanced HCC required 121 centers worldwide and reported on 602 randomized patients³. It is also difficult to envision the source of funding for such a study given the current economic environment. Moreover, similar to surgical expertise required for resection/transplantation, the use of chemoembolization or radioembolization also require specific skills; the number of centers that possess this expertise may be limited. Therefore, with this comparative effectiveness analysis, we have attempted to provide an alternative to a RCT while our institutional randomized phase 2 study is conducted.

Strengths and Limitations

There are strengths and limitations to this study. Strengths include the relevance of the study question, as well as the comprehensive comparison of toxicity, imaging and survival from a real-world clinical scenario at a large comprehensive cancer and transplant center; this is the intent of comparative effectiveness¹⁰. All toxicities were reported without the usual 30-day post treatment time cut-off, thereby minimizing attribution bias. Although controversial, the statistical methods included adjusted P-values and conclusions were drawn following adjustment, permitting cautious interpretation^{38, 39}. It is noteworthy that most HCC studies exclude Child-Pugh B patients; these patients were well-represented in both cohorts and provided insight into this understudied group. Nearly identical patient baseline cancer stage(s) permitted a thorough comparison between treatments; any heterogeneity would theoretically affect both arms equally. Our center has significant experience with chemoembolization and ⁹⁰Y, thereby permitting a comprehensive comparison and internal methodological standard^{7, 12}. The importance of in-patient stay and total days hospitalized was recognized as a relevant clinical parameter; this was measured and described. This type of analysis also addresses the reality that clinicians commonly discuss and offer patients several potential treatments for any given HCC status, despite the lack of direct or randomized comparative data between the proposed options (e.g. resection versus ablation for HCC). The weaknesses of the study were mitigated by utilizing conservative reporting methodology, especially for variables assessing response to therapy. Outcome variables (response, TTP, survival) were calculated from time of first treatment, with censoring to curative therapy. TTP was very conservatively measured and, although not clearly stated in the guidelines, included the development of any portal vein thrombosis as an endpoint (clinically relevant finding). Given the improved tolerability and outpatient nature, there was a tendency to treat the elderly with ⁹⁰Y, representing an inherent bias in favor of chemoembolization. Younger patients were also more likely to receive chemoembolization as a bridge to transplantation. Although both groups had mature follow-up, it was longer with chemoembolization, highlighting the slight difference in time when the treatments were performed. Cost factors were not addressed; this is a complex matter that plagues both treatments as well as all cancer therapies (surgical, systemic, locoregional) in general. These will be the subject of future analyses. Finally, in the absence of randomization, comparative effectiveness analyses may be limited by unmeasured confounders (clinical, social factors) and other selection biases (e.g. physician, local expertise) that may affect treatment decisions, outcomes, as well as the generalizability of these findings^{10, 40, 41}.

Conclusion

Our data show similar survival outcomes between ⁹⁰Y and chemoembolization for HCC^{21, 42-44}. Despite the older age in the ⁹⁰Y group, it was better tolerated than chemoembolization with fewer toxicities and treatments, without the need for in-patient hospitalization as a resource. The longer TTP associated with ⁹⁰Y has drop-out implications in potential recipients being bridged or downstaged to transplantation. While the importance of RCTs is fully acknowledged, we caution that a study aimed at showing survival equivalence may not be logistically feasible and unlikely to yield further clinically relevant information. Although currently there are ongoing randomized studies with transarterial LRTs, none have selected overall survival as the primary endpoint, supporting our findings reported here, as well as the notion of a statistical barrier to demonstrating survival equivalence. Given the results of this comparative effectiveness analysis, we believe a strong case can be made to consider ⁹⁰Y at least equivalent to chemoembolization as one of the standards of care for HCC. Rather than complete a >1000 patient RCT comparing these two treatments, we believe a better utilization of resources would be to move the science forward, combine this therapy with other approaches, and integrate with the rapidly evolving field of targeted biologics.