Albumin-bilirubin score is a prognostic factor in hepatocellular carcinoma treated with transarterial radioembolization: A retrospective cohort study

Abstract

BACKGROUND

While selective internal radiation therapy (SIRT), also known as transarterial radioembolization, is an established treatment for hepatocellular carcinoma (HCC), prognostic stratification remains challenging.

AIM

To identify prognostic factors for survival in patients with HCC treated with SIRT.

METHODS

A retrospective, single-center study was conducted, which included 50 patients with HCC treated with SIRT between February 2016 and May 2024. Demographic, clinical, and tumor-related variables were analyzed, including liver function, tumor burden, treatment response, and adverse events. Kaplan-Meier analysis determined overall survival (OS) and progression-free survival, while Cox regression models assessed prognostic factors.

RESULTS

Median OS was 22 months, with significant differences according to liver function assessed by the albumin-bilirubin (ALBI) score. Patients with ALBI grade 1 had longer survival than those with grade 2 [25 months vs 10 months; hazard ratio (HR) = 2.67, 95% confidence interval (CI): 1.2-6.0; P = 0.017]. A complete radiological response was achieved in 26% of patients and a partial response in 42%. The median progression-free survival was 9 months, but it was significantly shorter for patients with satellite nodules compared to those without (3 vs 11; HR = 2.44, 95%CI: 1.13-5.26; P = 0.013). Multivariate analysis confirmed the ALBI score as an independent predictor of OS (HR = 2.42, 95%CI: 1.02-5.74; P = 0.045).

CONCLUSION

The ALBI score is a useful and objective prognostic tool in patients with HCC undergoing SIRT. Its use may improve patient selection and clinical decision-making.

Key Words: Hepatocellular carcinoma; Selective internal radiation therapy; Albumin-bilirubin score; Yttrium-90; Satellite nodules

Core Tip: Selective internal radiation therapy is an established treatment for hepatocellular carcinoma, but prognostic stratification remains challenging. In this real-world cohort, the albumin-bilirubin score emerged as a key factor associated with overall survival, compared with conventional clinical parameters. These findings confirm that albumin-bilirubin is a simple, objective tool for patient selection and risk stratification, which can improve treatment decisions for patients undergoing selective internal radiation therapy.

INTRODUCTION

Hepatocellular carcinoma (HCC), the leading primary liver malignancy, typically develops within the context of cirrhosis[1]. Screening programs facilitate early detection, enabling curative options such as liver resection, transplantation, or ablation for Barcelona Clinic Liver Cancer (BCLC) stages 0-A, contingent on tumor burden (TB) and liver function. When these options are contraindicated or not feasible, locoregional therapies such as transarterial chemoembolization or selective internal radiation therapy (SIRT) may be used[2].

These established therapies for patients with intermediate-stage disease (BCLC-B) aim to manage disease progression and reduce TB. They serve to improve survival and potentially downstage patients, making them eligible for curative approaches. In this setting, meaningful complete response (CR) rates can be achieved. In advanced disease (BCLC-C), SIRT may represent an alternative for patients who are not candidates for systemic therapies[2,3].

Adequate liver function is essential for treatment eligibility. Within the BCLC staging system, the Child-Pugh (CP) score is the most frequently utilized tool, incorporating two clinical assessments (ascites and hepatic encephalopathy) alongside three laboratory parameters (bilirubin, albumin, and international normalized ratio).

Nonetheless, both CP and the model for end-stage liver disease score have limited prognostic accuracy for patients with preserved liver function. Clinical decompensation - marked by jaundice, ascites, or hepatic encephalopathy - signifies that liver function is already impaired. For patients with compensated liver disease, enhanced stratification is needed. In this context, the albumin-bilirubin (ALBI) score has gained increasing recognition, as it provides a more objective assessment of liver function and may improve the selection of candidates for locoregional therapies such as SIRT or transarterial chemoembolization.

MATERIALS AND METHODS

Study design

A retrospective, single-center study was conducted. All patients with HCC who underwent SIRT at our center from the introduction of the technique until May 2024 were included.

Patient selection

Patients were selected based on BCLC staging, which incorporates TB, liver function, and performance status. The analysis included all 50 patients who underwent the procedure during this period. In 2017, 3 patients were treated in the “STOP-HCC” clinical trial.

Variables and definitions

Baseline demographic, clinical, and tumor-related variables were analyzed, including liver disease etiology, liver function (assessed by CP and ALBI scores), TB, and alpha-fetoprotein levels. The ALBI score was calculated using the formula: (log₁₀bilirubin × 0.66) + (albumin × -0.085), where bilirubin is expressed in μmol/L and albumin in g/L. Patients were classified as ALBI grade 1 (≤ -2.60), grade 2 (> -2.60 to ≤ -1.39), and grade 3 (> -1.39). TB was evaluated using the TB score, calculated as: TB score = [(maximum tumor diameter)² + (number of tumors)²]^(1/2)[4]. The presence of satellite nodules (defined as one or more small tumor nodules located near the main tumor, usually within the same hepatic segment) and portal vein thrombosis (tumoral and non-tumoral) was also recorded. Treatment pathways before and after SIRT were documented.

Treatment procedure

Following our institutional protocol, SIRT was performed with yttrium-90 (Y-90) microspheres, preceded by a simulation scan using technetium-99m-labeled albumin macroaggregates approximately 14 days before treatment. This simulation was used to assess lung shunt fraction and evaluate extrahepatic tracer distribution, ensuring treatment safety. All procedures were performed with Y-90 glass microspheres (TheraSphere^®; Boston Scientific, Marlborough, MA, United States), with Y-90 glass microsphere dosimetry calculated based on TB and liver volume according to standard manufacturer guidelines. Dosimetric data for 31 patients were available; however, due to missing data in the remaining cases, dosimetric variables were not analyzed in this study.

Follow-up and outcomes

Imaging with dynamic computed tomography or contrast-enhanced magnetic resonance imaging was used to assess treatment response 1 month after SIRT and then every 3 months during follow-up, with results classified as CR, partial response (PR), and stable disease or progression according to modified Response Evaluation Criteria in Solid Tumours criteria[5]. Early progression was defined as imaging evidence of disease advancement at the first check-up, 1 month after SIRT. Because SIRT often has a delayed effect, all patients received follow-up imaging within 3 months. This follow-up assessment confirmed and further characterized tumor responses, particularly for initial stable disease or PR, while also identifying delayed CRs. Accordingly, imaging within this 3-month period was used to define the initial treatment response (stable disease, PR, or CR), whereas early progression status was defined exclusively based on the first imaging assessment.

Follow-up analysis included overall survival (OS), progression-free survival (PFS), and type of progression (treated lesion, new intrahepatic, or extrahepatic). Adverse events (AEs) were recorded for 1-month post-treatment. To evaluate the learning curve and patient selection effects, the cohort was stratified by treatment date (before vs after January 2022). Twenty procedures were performed before January 2022, with thirty thereafter.

Statistical analyses

IBM SPSS Statistics version 21 (IBM Corp., Armonk, NY, United States) was used for data analysis. A descriptive analysis of the study variables was performed. Because the quantitative data did not follow a normal distribution (Kolmogorov-Smirnov test), variables are expressed as the medians and interquartile ranges (IQRs). Qualitative variables are presented as absolute frequencies and percentages. Factors associated with initial treatment response were analyzed using the χ² test or Fisher’s exact test for qualitative variables and the Mann-Whitney U test for quantitative variables. Relative risks (RRs) were calculated using contingency tables and represent unadjusted estimates of association between baseline variables and early progression. OS was analyzed using the Kaplan-Meier method, and survival curves were plotted with 95% confidence intervals (CIs). The log-rank (Mantel-Cox) test was used to compare survival curves and determine statistically significant differences between groups. A Cox proportional hazards regression model was used to identify potential prognostic factors for survival, with results expressed as hazard ratios (HRs) with 95%CIs. Due to the limited number of events, variables were selected based on clinical relevance instead of statistical significance. To prevent overfitting, the model was adjusted only for BCLC stage, which is well-established as a key prognostic factor in HCC. P < 0.05 was considered statistically significant.

Ethical considerations

This study was conducted in accordance with the Declaration of Helsinki. Data were obtained from the RACH-001 regional registry, which was approved by the Research Ethics Committee of the Virgen Macarena-Virgen del Rocío University Hospitals (Approval No. SICEIA-2025-002463). Given the retrospective, anonymized nature of part of the study, the requirement for informed consent was waived for these patients; conversely, written informed consent was obtained from all, concurrently enrolled patients.

RESULTS

Demographics and clinical characteristics

Fifty patients received Y-90 microsphere SIRT during the study period; their characteristics are detailed in Table 1. The main etiology of liver disease was alcohol-related in 56% of cases, with 84% of patients presenting with advanced chronic liver disease. However, most patients maintained good function, with 86% classified as CP A and 68% as ALBI grade 1.

Table 1 Patient cohort characteristics, median (interquartile range)/n (%).

Variable	Patients (n = 50)
Age, years	67 (60-77)
Sex (men)	41 (82)
Smoking habit
Smoker	12 (24)
Ex-smoker	27 (54)
Never smoker	11 (22)
Diabetes mellitus	23 (46)
Obesity	18 (36)
High blood pressure	27 (54)
Etiology
Viral	12 (24)
MASLD	8 (16)
Alcohol	7 (14)
MetALD	7 (14)
Alcohol + viral	8 (16)
MetALD + viral	6 (12)
Advanced chronic liver disease	42 (84)
Esophageal varices	22 (44)
Child-Pugh A	44 (86)
ALBI score - grade 1	34 (68)

MASLD: Metabolic dysfunction-associated steatotic liver disease; MetALD: Metabolic dysfunction-associated and alcohol-related liver disease; ALBI: Albumin-bilirubin score.

Neoplastic disease

The median time from HCC diagnosis to SIRT was 7 months (IQR 4-14). Eighteen patients (36%) had received at least one prior treatment for HCC, including seven who received more than one, as detailed in the baseline characteristics in Table 2.

Table 2 Tumor disease characteristics, median (interquartile range)/n (%).

Variable	Patients (n = 50)
Previous treatments	18 (36)
TACE	9 (18)
Systemic therapy	6 (12)
Thermal ablation	3 (6)
Hepatectomy	1 (2)
Unilobar hepatocellular carcinoma	38 (76)
Satellite nodules	16 (32)
Portal infiltration	4 (8)
Maximum diameter of the main lesion, cm	4.6 (3.5-6.6)
Number of HCC
1	26 (52)
2	7 (14)
3	8 (16)
More than 3	9 (18)
Tumor burden score1	5.5 (4-7.1)
BCLC stage
A	15 (30)
B	32 (64)
C	3 (6)
AFP < 400 ng/mL, n (%)	42 (84)

¹Tumor burden score = [(maximum tumor diameter)² + (number of tumors)²]^(1/2).

TACE: Transarterial chemoembolization; HCC: Hepatocellular carcinoma; BCLC: Barcelona Clinic Liver Cancer; AFP: Alpha-fetoprotein.

Initial response to SIRT

CR was achieved in 13 patients (26%), PR in 21 (42%), and stable disease in 3 (6%). Thirteen patients (26%) developed tumor progression. Early progression was more frequent in procedures performed before January 2022 (9/20 vs 4/30), with a RR of 1.58 (95%CI: 1.04-2.40; P = 0.015). No statistically significant associations were observed between baseline variables and early progression; however, ALBI score (RR = 2.48; 95%CI: 0.99-6.10; P = 0.055) and arterial hypertension (P = 0.053) showed similar non-significant results. TB was not significantly associated with early progression (Table 3). Initial radiological response was associated with both PFS and OS.

Table 3 Univariate analysis of predictors of early progression, median (interquartile range)/n (%).

Variable		Early progression		P value
		Yes (n = 13)	No (n = 37)
Age, years		66 (58-71)	69 (60-77)	0.382
Sex	Men	12 (24)	29 (58)	0.25
	Women	1 (2)	8 (16)
Diabetes mellitus	No	9 (18)	18 (36)	0.17
	Yes	4 (8)	19 (38)
Obesity	No	9 (18)	23 (46)	0.458
	Yes	4 (8)	14 (28)
Arterial hypertension	No	3 (6)	20 (40)	0.053
	Yes	10 (20)	17 (34)
Viral etiology	No	8 (16)	16 (32)	0.208
	Yes	5 (10)	21 (42)
MASLD etiology	No	8 (16)	20 (40)	0.446
	Yes	5 (10)	17 (34)
Alcohol etiology	No	8 (16)	14 (28)	0.124
	Yes	5 (10)	23 (46)
Advanced chronic liver disease	No	2 (4)	6 (12)	0.659
	Yes	11 (22)	31 (62)
ALBI score	1	6 (12)	28 (56)	0.055
	2	7 (14)	9 (18)
Child-Pugh	A5	7 (14)	26 (52)	0.322
	A6	3 (6)	7 (14)
	B7	3 (6)	4 (8)
Unilobar HCC	No	4 (8)	29 (58)	0.376
	Yes	9 (18)	8 (16)
Satellite nodules	No	7 (14)	27 (54)	0.176
	Yes	6 (12)	10 (20)
Tumor burden score1		4.92 (3.93-6.10)	6.26 (4.08-7.52)	0.241
BCLC stage	A	4 (8)	11 (22)	0.949
	B	8 (16)	24 (48)
	C	1 (2)	2 (4)
AFP < 400 ng/mL	No	11 (22)	31 (62)	0.659
	Yes	2 (4)	6 (12)
Sample division	Before January 2022	9 (18)	11 (22)	0.015
	After January 2022	4 (8)	26 (52)

¹Tumor burden score = [(maximum tumor diameter)² + (number of tumors)²]^(1/2).

Early progression was more frequent in procedures performed before January 2022 (9/20 vs 4/30), with a relative risk of 1.58 (95% confidence interval: 1.04-2.40; P = 0.015). No statistically significant associations were observed between baseline variables and early progression. MASLD: Metabolic dysfunction-associated steatotic liver disease; ALBI: Albumin-bilirubin; HCC: Hepatocellular carcinoma; BCLC: Barcelona Clinic Liver Cancer; AFP: Alpha-fetoprotein.

AEs attributable to the procedure

Eighteen patients (36%) experienced at least one AE in the month following treatment. The most common AE was asthenia, which occurred in 12 patients (24%), followed by abdominal pain (14%) and new-onset ascites (14%). Three cases of hepatic encephalopathy and two cases of radioembolization-induced liver disease (REILD) were also recorded. All cases were managed on an outpatient basis, without the need for hospitalization.

Subsequent treatments

While all patients received SIRT as a standalone locoregional treatment without concomitant systemic therapy, 48% (n = 24) required further treatment during follow-up. Among these, 22% (n = 11) received additional locoregional treatments - most commonly a second SIRT session (18%, n = 9), while others underwent thermal ablation (6%, n = 3), liver transplantation (6%, n = 3), or hepatectomy (n = 1). An additional 20% (n = 10) of the cohort later received systemic therapy following tumor progression, provided they maintained good liver function. However, subsequent treatment was not associated with OS, with a median of 25 months (95%CI: 19.8-30.2) for treated patients vs 22 months (95%CI: 12.6-31.4) for untreated patients (P = 0.25).

Follow-up

With a median follow-up of 14 months (IQR 7-22), 60% of patients (n = 30) experienced tumor progression, including 20 with progression of the treated lesion. Median PFS was 9 months (95%CI: 7-11). Kaplan-Meier analysis showed that the presence of satellite nodules was significantly associated with a shorter PFS (Figure 1). Regarding PFS based on initial response, patients who achieved a CR had a median PFS that was not reached (estimated at 33 months), significantly longer than the 11-month median PFS (95%CI: 8.7-13.3) for those patients with a PR (P < 0.001).

Figure 1 Kaplan-Meier curve of progression-free survival by presence of satellite nodules. A: Kaplan-Meier survival curve; B: Number of patients at risk at different time points. Median progression-free survival was longer in patients without satellite nodules [11 months; 95% confidence interval (CI): 7.3-14.7] than in those with satellite nodules (3 months; 95%CI: 0-7); hazard ratio 2.44 (95%CI: 1.13-5.26), P = 0.013.

Follow-up revealed a 50% mortality (n = 25), with most deaths resulting from liver dysfunction or disease progression. The median OS was 22 months (95%CI: 16.7-27.3; Figure 2). Among the liver function variables analyzed, only the ALBI score showed a significant association with OS (Figure 3), while CP and BCLC stages did not. This lack of significance may be due to 3 patients with preserved liver function (CP A) and low TB (BCLC A) who experienced shorter survival due to concomitant malignancies. In univariate Cox regression analysis, clinically relevant variables were evaluated. CP score (HR = 1.56, 95%CI: 0.93-2.62; P = 0.095), TB score (HR = 0.85, 95%CI: 0.70-1.03; P = 0.089), Alpha-fetoprotein levels (HR = 1.77, 95%CI: 0.70-4.53; P = 0.230), and the presence of satellite nodules (HR = 1.45, 95%CI: 0.65-3.23; P = 0.365) were not significantly associated with OS. Multivariate Cox regression analysis adjusted for BCLC stage (BCLC A vs BCLC B/C) was performed, as this variable integrates TB and liver function and is widely used for prognostic stratification in HCC. However, given the limited number of events, this analysis was restricted and should be interpreted with caution. In this model, ALBI grade remained associated with OS (HR = 2.42, 95%CI: 1.02-5.74; P = 0.045).

Figure 2 Kaplan-Meier curve of overall survival. Estimated median overall survival was 22 months (95% confidence interval: 16.7-27.3).

Figure 3 Kaplan-Meier curve of overall survival stratified by albumin-bilirubin score (grade 1 vs grade 2). A: Kaplan-Meier survival curve; B: Number of patients at risk at different time points. Patients with albumin-bilirubin grade 1 had a median overall survival of 25 months [95% confidence interval (CI): 20.2-29.8] compared with 10 months (95%CI: 2.8-17.2) in those with albumin-bilirubin grade 2; hazard ratio = 2.67 (95%CI: 1.2-6), P = 0.017. ALBI: Albumin-bilirubin.

Kaplan-Meier analysis showed that OS was significantly associated with initial radiological response, with a median survival of 17 months (95%CI: 6.3-27.7), 23 months (95%CI: 16-30), and 53 months (median not reached) in patients with progression, PR, and CR, respectively (P = 0.015). Subgroup analyses according to radiological response were not performed due to the limited number of patients in each subgroup, which would have reduced statistical reliability.

DISCUSSION

SIRT has demonstrated promising results in patients with HCC, both in intermediate and early stages. Our study included mainly patients with intermediate-stage disease, although 30% were in the early stage of BCLC. Previous studies have documented variable response rates. Muglia et al[6] reported a CR rate of 31% and PR rate of 38% in a cohort with moderate TB (42.3% single nodule, with a mean diameter of 65 mm).

In patients with early-stage disease, concomitant malignancies negatively influenced survival despite a low TB. By contrast, previous studies have shown excellent long-term outcomes with SIRT in carefully selected patients with early-stage HCC. For example, Lewandowski et al[7] reported high response rates and prolonged survival in patients with single tumors smaller than 5 cm.

Survival in HCC is influenced not only by TB but also by liver function. Because 80%-90% of HCC occurs in cirrhotic livers, a patient’s hepatic reserve often dictates OS more than tumor characteristics. Multiple studies, including those focused on patients receiving locoregional and systemic therapies, have demonstrated that the ALBI score is a more accurate prognostic factor for liver function in patients with HCC than the classic CP classification[8,9]. Regarding SIRT, Hickey et al[10] validated its prognostic utility in a North American cohort, whereas Demirtas et al[11] highlighted its utility for functional stratification. Several studies have consistently shown improved survival in patients with better ALBI grades after SIRT compared to those classified by CP[12,13]. Gui et al[14] demonstrated that patients with ALBI grade 1 experienced significantly better survival, highlighting the ALBI grade as a superior objective tool for assessing liver function compared to CP classification, which did not effectively differentiate outcomes.

Studies indicate that post-treatment toxicity, specifically the development of REILD, is more common in patients with pre-existing liver dysfunction, particularly those with CP ≥ B7[11]. A retrospective study highlighted the ALBI score’s ability to predict liver toxicity[15], supporting its role in selecting candidates for aggressive locoregional therapies such as SIRT. Nevertheless, the incidence of REILD (4%) in this study was low, precluding meaningful analysis of its association with the ALBI score.

Another noteworthy finding is that patients treated after January 2022 experienced lower initial progression rates than those treated before this date. These findings highlight the impact of the learning curve alongside improved patient selection, evidenced by the rise in BCLC A stage cases from 20% before January 2022 to 36.7% thereafter. Mosconi et al[16] reported a marked improvement in median OS between two treatment periods, which they attributed to improved technical experience, more rational dosimetry, and better patient selection for SIRT.

This study had several limitations. First, its retrospective single-center design and the absence of a control group limit the generalizability of the findings. In addition, the relatively small sample size and number of events may have reduced the statistical power and limited the ability to perform more comprehensive multivariate adjustments. Dosimetric data were available in only 31 patients (approximately 60% of the cohort) and were therefore not included in the analysis due to missing data. This represents a relevant limitation, as radiation dose is a key determinant of both treatment efficacy and liver toxicity. The absence of dosimetric information in a substantial proportion of patients may have influenced the observed outcomes and limit the interpretation of treatment response and survival results, potentially introducing selection bias. In addition, the relatively small sample size may have limited the statistical power of subgroup analyses, particularly those exploring the relationship between ALBI score and treatment-related toxicity such as REILD. With a median follow-up exceeding 1 year, these findings are preliminary and hypothesis-generating, rather than confirmatory. Therefore, larger, multicenter studies with systematic dosimetric data collection are needed to validate these findings and further clarify the prognostic role of treatment-related factors.

CONCLUSION

In this real-world cohort of patients with HCC treated with SIRT, the ALBI score effectively predicted OS, confirming its value as a simple, objective tool for prognostic stratification. The presence of satellite nodules was associated with PFS, highlighting the relevance of tumor-related features in disease progression. These findings highlight that careful patient selection is key to optimizing treatment outcomes. Further prospective studies with larger cohorts are warranted to validate these results.

ACKNOWLEDGEMENTS

The authors would like to thank the Hepatology Unit and the Interventional Radiology Unit at Virgen del Rocío Hospital for their collaboration in this study.