Ke-Feng Jia ,Hao Wang,Chang-Lu Yu ,Wei-Li Yin ,Xiao-Dong Zhang ,Fang Wang ,Cheng Sun ,Wen Shen ,?

a Department of Radiology, First Central Clinical College, Tianjin Medical University, Tianjin 300192, China

b Department of Radiology, Tianjin Third Central Hospital, Tianjin 300170, China

c Department of Radiology, Tianjin First Central Hospital, School of Medicine, Nankai University, Tianjin 300192, China

d Department of Gastroenterology and Hepatology, Tianjin Third Central Hospital, Tianjin 300170, China

Keywords: Hepatocellular carcinoma Transarterial chemoembolization Scoring system Prognostic model Child-Pugh class Survival prediction

ABSTRACT Background: Due to the high heterogeneity among hepatocellular carcinoma (HCC) patients receiving transarterial chemoembolization (TACE),the prognosis of patients varies significantly. The decisionmaking on the initiation and/or repetition of TACE under different liver functions is a matter of concern in clinical practice. Thus,we aimed to develop a prediction model for TACE candidates using risk stratification based on varied liver function.Methods: A total of 222 unresectable HCC patients who underwent TACE as their only treatment were included in this study. Cox proportional hazards regression was performed to select the independent risk factors and establish a predictive model for the overall survival (OS). The model was validated in patients with different Child-Pugh class and compared to previous TACE scoring systems.Results: The five independent risk factors,including alpha-fetoprotein (AFP) level,maximal tumor size,the increase of albumin-bilirubin (ALBI) grade score,tumor response,and the increase of aspartate aminotransferase (AST),were used to build a prognostic model (ASARA). In the training and validation cohorts,the OS of patients with ASARA score ≤2 was significantly higher than that of patients with ASARA score > 2 ( P < 0.001,P = 0.006,respectively). The ASARA model and its modified version “AS(ARA)” can effectively distinguish the OS ( P < 0.001,P = 0.004) between patients with Child-Pugh class A and B,and the C-index was 0.687 and 0.706,respectively. For repeated TACE,the ASARA model was superior to Assessment for Retreatment with TACE (ART) and ALBI grade,maximal tumor size,AFP,and tumor response(ASAR) among Child-Pugh class A patients. For the first TACE,the performance of AS(ARA) was better than that of modified hepatoma arterial-embolization prognostic (mHAP),mHAP3,and ASA(R) models among Child-Pugh class B patients.Conclusions: The ASARA scoring system is valuable in the decision-making of TACE repetition for HCC patients,especially Child-Pugh class A patients. The modified AS(ARA) can be used to screen the ideal candidate for TACE initiation in Child-Pugh class B patients with poor liver function.

Introduction

Hepatocellular carcinoma (HCC) remains a global health challenge with a rising incidence [1] of approximately 1 million cases per year by 2025 [2] . According to Barcelona Clinic Liver Cancer (BCLC) staging system [3],guidelines of the American Association for the Study of Liver Diseases (AASLD) [4],and the European Association for the Study of the Liver Cancer (EASL) [5],transarterial chemoembolization (TACE) is the first-line treatment option for intermediate HCC patients with well-preserved liver function and asymptomatic and unresectable large/multifocal tumors in the absence of vascular invasion or extrahepatic metastasis. However,these patients who were recommended for TACE had a high degree of clinical heterogeneity with median survival ranging from 13-43 months [6] . As a result,recent guideline [5] have been proposed to address the risk stratification in this group of patients. Although several TACE initiation (hepatoma arterialembolization prognostic (HAP) score [7],modified HAP (mHAP)score [8],mHAP2 score [9],and mHAP3 score [10] ) and repetition scoring systems [Assessment for Retreatment with TACE (ART)score [11],alpha-fetoprotein (AFP),BCLC,Child-Pugh,and Response(ABCR) score [12],and albumin-bilirubin (ALBI) grade,maximal tumor size,AFP,and tumor response (ASAR) score [13] ] were suggested,these strategies may have a limited prognostic performance for TACE,which could be attributed to the lack of exclusion of other anti-tumor therapies,such as surgical resection,liver transplantation,radiofrequency ablation,and sorafenib.

In addition,maintaining liver function is equal important to tumor control in HCC patients. For Child-Pugh class A patients with satisfactory liver function,the evaluation of repeated TACE is crucial as ineffective TACE may result in the loss of optimal switch to other treatments [13] . Among Child-Pugh class B patients,“whether to start TACE session or not” is determined to prevent the deterioration of impaired liver function and survival worsening [14] . The present study aimed to develop a prognostic model for patients treated with TACE only and screen TACE candidates with different liver functions.

Methods

Study cohort

From January 2008 to December 2013,a total of 524 HCC patients underwent TACE treatment for the first time at Tianjin Third Central Hospital,China. The diagnostic criteria for HCC were based on the AASLD [4] and the EASL [5] guidelines and were confirmed by histological or radiographic evaluation. The diagnosis of liver cirrhosis was based on the "2019 Chinese guidelines on the management of liver cirrhosis" [15],which include the etiology,medical history,clinical manifestations,complications,treatment process,laboratory test,imaging and histology. The inclusion criteria were as follows: 1) age>18 years; 2) received ≥2 TACE sessions;3) liver function of Child-Pugh class A or B; 4) BCLC stage A,B,or C HCC,and the portal vein tumor thrombus (PVTT) involved only branches; and 5) Eastern Cooperative Oncology Group (ECOG) activity score of 0 or 1 point. The exclusion criteria were as follows:1) any previous anti-tumor therapies or received TACE as a bridge for any other therapy before the end of follow-up (including surgical resection,liver transplantation,radiofrequency ablation,sorafenib,or other systemic therapies); 2) imaging evaluation after first TACE showed complete response (CR); and 3) data were incomplete. Thus,302 patients were excluded,and 222 patients were enrolled in this study ( Fig. 1 ). The study was conducted in accordance with the ethical guidelines oftheDeclarationofHelsinkiand approved by the Ethics Committee of Tianjin Third Central Hospital.

Fig. 1. Flow diagram of the screening of patients. CR: complete respons; HCC: hepatocellular carcinoma; TACE: transarterial chemoembolization.

TACE session, efficacy evaluation, and follow-up

TACE was performed by radiologists with more than 10 years of experience,and superselective method was used to embolize the tumor. A suspension of doxorubicin (10-60 mg) and iodinated oil (2-20 mL) was infused,followed by gelatin sponge embolization. The dose of chemotherapeutic drugs and embolic materials was determined according to the tumor burden and the liver function of each patient [16] . The laboratory indexes were tested within 3 days before each TACE session. Computed tomography (CT) or magnetic resonance imaging (MRI) examination was performed 6-8 weeks following each TACE session. If residual or newly found lesions were detected,repeated TACE was considered. The tumor response was evaluated by modified Response Evaluation Criteria in Solid Tumors (mRESIST) [17] . The follow-up was ended on November 30,2020.

Characteristics and risk stratification of previous TACE scores

According to previous studies,the baseline and follow-up characteristics and patient risk stratification included in 5 first and 3 repeated TACE scores are shown in Table S1. ASA(R) [13] is a modified version of the ASAR score,which assesses the prognosis of first-time TACE therapy after removing the radiological response from the original model.

Statistical analysis

Quantitative data with normal distribution were presented as mean ± standard deviation (SD),and Student’st-test was used to compare between groups. The quantitative data with skewed distribution were described with median and interquartile range(IQR),qualitative data were described as frequencies and percentages,and Mann-WhitneyUtest or Chi-square test was used for comparison between groups. The study cohort was randomly divided into training and validation datasets with a ratio of 7:3. The overall survival (OS) was defined as TACE initiation until death from any cause. Cox proportional hazard regression model was used to investigate the prognostic factors for OS in HCC patients treated with TACE. Factors showing significant differences (P<0.05) in univariate analyses were included in a multivariate model.The prediction model of OS was established by the forward and backward stepwise selection,in which a minimum Akaike information criterion value was selected. Schoenfeld’s residuals method was applied to evaluate the proportional-hazards hypothesis of the model. The final regression coefficients were weighed and assigned to each factor. The determination of optimal cutoff value was based on pairwise log-rank test results. Harrell’s concordance (C)-index was used to evaluate the predictive ability of the model. Based on the cutoff value,the training and validation cohorts were divided into different risk subgroups. The Kaplan-Meier survival curve was plotted,and the log-rank test was used to further verify the model.Then,the model was validated in patients with different Child-Pugh grades and compared to the previous TACE scores with respect to the predictive and calibration performance. The integrated Brier score (IBS) within 60 months was calculated as the summary measure of the predictive error. A low Brier score indicated a lower predictive error and an optimal calibration performance. SPSS software version 20.0 (SPSS Inc.,Chicago,IL,USA) and R version 3.3.1(R Foundation for Statistical Computing,Vienna,Austria) were used for all statistical analyses. Two-sidedP<0.05 indicated statistical significance.

Results

Baselines and follow-up patient characteristics

The mean age of 222 patients in the baseline study cohort was 59.7 ± 8.8 years old,and 80.2% (178/222) were males. The majority were at BCLC stage B (65.8%,146/222),and 214 patients(96.4%) had liver cirrhosis. Most of them had chronic hepatitis B virus (HBV) infection (79.0%,169/214). The patients were randomly divided into a training cohort (n= 155) and a validation cohort(n= 67) at the ratio of 7:3. No significant difference was observed in the baseline and follow-up characteristics between the two groups (allP>0.05,Table 1 ). The median follow-up time was 34.5 (24.0-43.6) months. At the end of the follow-up,9 (5.8%) patients survived,146 (94.2%) patients died,and the median OS was 25.3 (14.5-45.0) months in the training cohort. In the validation cohort,4 (6.0%) patients survived,and 63 (94.0%) patients died; the median OS was 20.0 (13.4-40.0) months.

Table 1Baseline and follow-up characteristics of the training and validation cohort.

The analysis of prognostic factors for HCC patients after TACE

Univariate Cox analysis showed that the following five characteristics were statistically significant prognostic factor for HCC:maximal tumor diameter>7 cm,AFP ≥400 ng/mL,AST increase>25%,ALBI score increase and tumor response (allP<0.05). Forward and backward stepwise selection was applied for multivariate Cox analysis to identify the independent prognostic factors as follows: AFP ≥400 ng/mL [hazard ratio (HR) = 1.575; 95% CI: 1.082-2.291;P= 0.018],maximal tumor diameter>7 cm (HR = 1.493;95% CI: 1.029-2.166;P= 0.035),ALBI score increase (HR = 1.469;95% CI: 1.049-2.057;P= 0.025),tumor response (HR = 2.024; 95%CI: 1.448-2.829;P<0.001),and AST increase>25% (HR = 1.826;95% CI: 1.292-2.581;P<0.001) ( Table 2 ).

Table 2Uni- and multivariate analysis for OS in the training cohort ( n = 155).

Prediction model establishment

The prediction model “ASARA” (stands for AFP,size,ALBI,tumor response,and AST) ( Table 3 ) was established by screening.The regression coefficient (β) of each factor was 0.454 (AFP ≥400 ng/mL),0.401 (maximal tumor diameter>7 cm),0.385 (ALBI score increase),0.705 (no tumor response),and 0.602 (AST increase>25%). Each of the five factors was assigned a weighted score (0,1) based on the regression coefficient,and the total scores ranged from 0-5 ( Table 3 ); of these,2 were determined as the cutoff values based on the pairwise log-rank result. The C-index of the training and the validation cohort was 0.687 (95% CI: 0.640-0.734) and 0.675 (95% CI: 0.602-0.748),respectively.

Table 3ASARA (AFP-size-ALBI-response-AST) scoring system.

Validation of ASARA and sensitivity analysis

The training cohort was divided into two subgroups based on the prognosis according to cutoff value. Log-rank test showed that the median OS of patients with ASARA ≤ 2,ASARA>2 was 30.6 (IQR: 21.6-51.0),14.0 (IQR: 6.0-24.0) months,respectively (P<0.001,Fig. 2 A). The 1-,3-,and 5-year survival rates in ASARA ≤2 group were 89.0%,42.2%,19.7% and in ASARA>2 group,53.3%,15.6%,2.2%,respectively.

Fig. 2. Kaplan-Meier survival curve of ASARA ≤2 group and ASARA > 2 group. A : Training cohort; B : validation cohort.

In the validation cohort,the median OS of patients with ASARA≤2 and ASARA>2 was 26.0 (IQR: 15.8-47.3) and 13.6 (IQR: 9.0-24.0) months,respectively (P= 0.006,Fig. 2 B). The 1-,3-,and 5-year survival rates in the ASARA ≤2 and ASARA>2 groups were 87.5%,35.4%,16.7%,and 63.2%,19.7%,0.0%,respectively.

Validation of ASARA in different liver function groups and comparison with previous TACE scoring system

In the study cohort,214 cirrhotic patients were divided into Child-Pugh A (n= 116) and Child-Pugh B (n= 98). Among the Child-Pugh A patients,the median OS of patients with ASARA ≤2 and ASARA>2 was 30.6 (IQR: 19.0-69.4) months,15.2 (IQR:9.0-32.0) months,respectively (P<0.001,Fig. 3 A). The 1-,3-,and 5-year survival rates in the ASARA ≤2 group were 88.7%,43.0%,and 29.9% respectively,while those in the ASARA>2 group were 66.7%,21.2%,and 3.0%,respectively. The predictive performance of ASARA score (C-index = 0.687,95% CI: 0.624-0.749)was significantly better than that of ART and ASAR (P= 0.001 and 0.035,respectively) in Child-Pugh A patients ( Table 4 ). With reference to Kaplan-Meier curve,ASARA score (5-year IBS,0.159)was also better than ART and ASAR (0.178 and 0.172,respectively) in evaluating the calibration performance of each prediction model within 60 months. The prediction error curve is illustrated in Fig. 4 A.

Table 4Comparison of TACE score in Child-Pugh A patients.

Fig. 3. Survival difference between ASARA ≤2 and ASARA > 2 in Child-Pugh A patients ( A ) and between AS(ARA) ≤1 and AS(ARA) > 1 in Child-Pugh B patients ( B ).

Fig. 4. Prediction error curve. A: Comparison of ASARA with other repeated TACE score; B: comparison of AS(ARA) with other initial TACE scores.

Since the TACE initiation is a major issue for Child-Pugh B patients,we modified the ASARA model by excluding the parameters obtained after the first TACE and evaluated the first TACE treatment. The modified model contains only two factors: AFP and maximal tumor size,namely “AS(ARA)”. We used 1 as the cutoff value ( ≤1 vs.>1) based on pairwise log-rank method and used to divide the Child-Pugh B patients into two subgroups with different prognosis. The median OS of patients with AS(ARA) ≤1 and AS(ARA)>1 was 25.0 (IQR: 15.7-42.7) months,12.5 (IQR: 4.5-24.0) months,respectively (P= 0.004,Fig. 3 B). The 1-,3-,and 5-year survival rates in AS(ARA) ≤ 1 group were 81.0%,33.3%,and 7.1%,respectively,while those in AS(ARA)>1 group were 50.0%,14.3%,and 0.0%,respectively. The predictive performance of AS(ARA) score (C-index = 0.706,95% CI: 0.617-0.797) in Child-Pugh B patients was significantly better than that of mHAP,mHAP3,and ASA(R) (P= 0.001,0.003,0.007,respectively) ( Table 5 ). The calibration performance of AS(ARA) (5-year IBS,0.150) was superior to that of other scores ( Table 5 ). The prediction error curve is demonstrated in Fig. 4 B.

Table 5Comparison of AS(ARA) with other initial TACE scores in Child-Pugh B patients.

Discussion

The ASARA prognostic model established in this study would be valuable in predicting whether patients with different liver functions are suitable for initial or repeated TACE therapy. Compared to the previous studies,this model has the following advantages:(1) for the first time,patients who underwent TACE as the only treatment were selected. The predictive performance could be improved by excluding the effects of other anti-tumor treatments on survival. (2) These phenomena prove the applicability of the model in patients with different liver functions,and the prediction and calibration performance was compared to five initial and three repeated TACE scoring systems established previously.

The median follow-up duration in this study was 34.5 (24.0-43.6) months. The ASARA score comprises five clinically accessible factors,including three tumor burden-related factors,such asAFP,maximal tumor diameter,and tumor response,and two liver function-related factors,such as the increase of ALBI and AST. AFP remains a reliable marker for tumor burden and invasive tumor behavior [ 18,19 ]. Baseline AFP>400 ng/mL in HCC patients undergoing TACE therapy is associated with a three-fold increase in the risk of death [7] . Typically,tumor size affects survival. The larger the tumor,the greater the risk of vascular invasion and extrahepatic metastasis [19–22] . Currently,there is no consensus yet on the cutoff diameter (between 3 and 10 cm),which should be used as the threshold for worsening survival [19–21] . Similar tothe current study,tumor diameter>7 cm was also included in other models such as HAP,mHAP,and mHAP2. Tumor response is often used in expert consensus or guidelines to define refractory TACE [23–25] and has also been included in the previous repeated TACE score [11–13] . In this study,tumor response after the first TACE was associated with a two-fold risk of death (HR = 2.024,95% CI: 1.448-2.829;P<0.001).

As a standard tool for liver function assessment,the Child-Pugh score includes subjective factors (ascites and hepatic encephalopathy grades) and interrelated factors (albumin and ascites) [26] . Recently,the ALBI score developed by Johnson et al [27] . provided an objective and simple method for assessing liver function and has a significant predictive value on the prognosis of HCC patients undergoing TACE treatment [ 26,28-29 ]. Unlike the ASAR score [13],the present study did not include ALBI grade but demonstrated that the score increase was an independent prognostic factor for patients undergoing TACE (P= 0.025). The elevation of bilirubin and the decrease in albumin after TACE could deteriorate liver function. We speculated that an increase in ALBI scores is a robust predictor of liver function impairment,and only when this subtle impairment has accumulated to a certain degree,ALBI grade is altered. As a reflection of liver function,AST increase>25% has been included in the ART score and our model [11] . Unlike 4 points in the ART score,the weighted score for AST increase>25% in this model was 1,which increased the risk of death by 1.8-fold(HR = 1.826; 95% CI: 1.292-2.581;P<0.001). Thus,it could be deduced that AST changes should not account for four points in the prognostic model,which might be a reason why many follow-up studies have been unable to validate ART scores [30–33] .

In Child-Pugh A patients,ASARA was superior to ART,ASAR scores and not second to ABCR score,suggesting that ASARA could be used as a screening tool to determine the benefit of TACE retreatment and prevent these patients from losing the opportunity to switch to other therapies due to TACE failure. In Child-Pugh class B patients with poor liver function,AS(ARA) is a crucial parameter to decide whether TACE should be initiated and to avoid otherwise further deterioration of the liver function caused by TACE and worsened survival. Herein,we established a modified model AS(ARA) by excluding three follow-up factors of ASARA and retaining only AFP and maximal tumor diameter. Although it only contains two baseline factors reflecting the tumor burden,AS(ARA)is a satisfactory predictor of prognosis,and the C-index was 0.706 with respect to the condition of Child-Pugh B liver function. In addition,the performance of AS(ARA) is better than that of the other TACE scores,such as mHAP,mHAP3,and ASA(R),and no less than HAP and mHAP2.

Although ASAR score is also based on liver function,it differed from ASARA that included TACE-only patients. Moreover,the model is suitable for clinical practice to include patients with Child-Pugh class B9 and BCLC stage C. Nevertheless,the present study has some limitations. Firstly,the patients included in this study constitute Chinese HCC population with HBV,thereby necessitating that the current results be extrapolated cautiously and be externally validated by multicenter inclusion of different populations;Secondly,since TACE was the only treatment for the patients included,whether it is applicable to the current combination therapy needs to be explored further; Thirdly,the retrospective nature of this study could lead to several biases; (4) Only patients who received conventional TACE were included in this study,while those who received drug-eluting bead TACE (DEB-TACE) were not evaluated.

In summary,the objective and simple ASARA score might shed light on the decision-making of repeated TACE for HCC patients,especially in Child-Pugh class A patients. When ASARA>2,the model can predict the TACE failure and help the patients to switch to systemic therapy before deterioration of the liver function. For Child-Pugh class B patients with impaired liver function,the modified AS(ARA) model could be applied to screen the ideal candidates for initial TACE treatment.

Acknowledgments

We would like to express our gratitude to all the colleagues that cooperate and comment on this study.

CRediT authorship contribution statement

Ke-FengJia:Conceptualization,Funding acquisition,Investigation,Methodology,Visualization,Writing – original draft,Writing– review & editing.HaoWang:Conceptualization,Data curation,Investigation,Methodology,Validation,Writing – review & editing.Chang-LuYu:Formal analysis,Funding acquisition,Methodology,Resources,Validation,Writing – original draft.Wei-LiYin:Conceptualization,Data curation,Investigation,Visualization,Writing– original draft.Xiao-DongZhang:Conceptualization,Investigation,Methodology,Resources,Software.FangWang:Data curation,Investigation,Methodology,Software,Validation,Writing – original draft.ChengSun:Data curation,Investigation,Methodology,Software,Visualization.WenShen:Formal analysis,Methodology,Project administration,Resources,Supervision,Validation,Writing– review & editing.

Funding

This study was supported by a grant from Tianjin Key Medical Discipline (Specialty) Construction Project.

Ethical approval

The study was conducted in accordance with the ethical guidelines oftheDeclarationofHelsinkiand approved by the Ethics Committee of Tianjin Third Central Hospital.

Competing interest

No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.

Supplementary materials

Supplementary material associated with this article can be found,in the online version,at doi: 10.1016/j.hbpd.2022.02.007 .