Predictive factors for the benefit of triple‐drug transarterial chemoembolization for patients with unresectable hepatocellular carcinoma

Abstract Background Compared with single‐drug TACE, our previous phase III study demonstrated that triple‐drug transarterial chemoembolization (TACE) prolonged overall survival (OS) in patients with unresectable hepatocellular carcinoma (HCC). The aim of this study was to find which patients can benefit from the triple drugs TACE compared with single‐drug TACE. Methods Patients in the triple‐drug TACE arm received sponge embolization and emulsions composed of 50 mg epirubicin, 50 mg lobaplatin, 6 mg mitomycin C, and lipiodol, while patients in the single‐drug TACE arm received sponge embolization and emulsions composed of 50 mg epirubicin and lipiodol. From July 2007 to November 2009, 244 patients (224 men and 20 women; age ranged from 21 to 75 years) from our phase III study formed the initial cohort. From January 2010 to June 2015, external validation cohort was composed of 449 patients (411 men and 38 women; age ranged from 18 to 75 years) from another institution. The validation cohort after propensity score matching (PSM) (n = 374) was analyzed. Cox proportional hazard model was used to evaluate the interaction term between treatments for each subgroup. This retrospective study was approved by the institutional review board at each center. Results No difference was observed in the baseline characteristic of three cohorts. This exploratory analysis showed that triple‐drug TACE brought a survival benefit in the initial cohort, validation cohort (before PSM), and validation cohort (after PSM) compared with single‐drug TACE. The outcomes of three cohorts all showed that a significantly greater OS triple‐drug chemotherapy benefit versus single‐drug chemotherapy was seen in patients with large tumors (larger than 10 cm) while no survival difference was seen in patients with small tumors (10 cm or smaller). Conclusions Triple‐drug TACE seems to benefit patients with HCC larger than 10 cm in particular compared with single‐drug TACE.


| INTRODUCTION
Transarterial chemoembolization (TACE) is the first-line treatment for patients with unresectable hepatocellular carcinoma (HCC) at intermediate stage, and nexavar (sorafenib) is the first-line treatment for patients with HCC at advanced stage. 1,2 According to Barcelona Clinic Liver Cancer (BCLC) staging system, intermediate stage (BCLC B) is formed by those patients with single large HCCs and those with multifocal disease who are asymptomatic and do not present vascular invasion or extrahepatic spread, and advanced stage (BCLC C) is formed by those patients with symptoms and/or present vascular invasion or extrahepatic spread. 3 Different from Western guidelines, Asian guidelines recommend TACE for advanced HCC. 4,5 Additionally, a randomized, phase III of advanced HCC showed that TACE plus sorafenib had noninferiority in overall survival (OS) (12.8 vs 10.8 months, P = 0.29) but superiority in time to progression (5.3 vs 3.5 months, P = 0.003), progression-free survival (5.2 vs 3.6 months, P = 0.01) and response rate (60.6% vs 47.3%, P = 0.005) compared with sorafenib monotherapy, and TACE plus sorafenib had longer survival rate in patients who received ≥2 TACE sessions (18.6 vs 10.8 months, P = 0.0055) in the subgroup analysis. 6 Conventional TACE is composed of chemotherapy and embolization. However, which one played the leading role was controversial. Some studies showed that chemotherapy made a positive effect, and adding embolization did not improve survival. [7][8][9][10] In contrast, others believed that embolization was important, and chemotherapy was useless except adding adverse events. [11][12][13] Additionally, a meta-analysis including six randomized controlled trials demonstrated that there was no superiority between TACE and bland embolization in HCC. 14 However, our multicenter, randomized, phase III trial including HCC at intermediate or advanced stage showed that triple-drug chemolipiodolization with gelatin-sponge embolization improved OS compared single-drug chemolipiodolization with gelatin-sponge embolization ( 15 In addition, no statistically significant difference in the incidence of adverse event or serious adverse event was observed between tripledrug TACE and single-drug TACE.
Although there are a number of clinical scoring systems that help assess patient prognosis after TACE, [16][17][18] predictive factors for the additional benefit of triple-drug chemotherapy compared with single chemotherapy are lacking. A predictive factor, which is a parameter used to distinguish subgroup of patients who can benefit most from a specific treatment, is different from a prognostic factor, which is a baseline characteristics related to the natural processes of the disease in spite of treatment. 19 The data of prospective randomized trials are ideal to research predictive factors. The purpose of this study was to find out baseline parameters that predicted the survival benefit of triple-drug chemotherapy on OS compared with single-drug TACE in patients with unresectable HCC at intermediate or advanced stage.

| Study design
This prior article (ClinicalTrials.gov number NCT00493402) dealt with the development of the important role of triple-drug chemotherapy in the TACE 15 whereas in this manuscript, we reported on which patients can benefit from triple-drug TACE in particular. In this retrospective analysis, from July 2007 to November 2009, 244 patients who have been previously reported in our phase III clinical study 15 formed the Results: No difference was observed in the baseline characteristic of three cohorts.
This exploratory analysis showed that triple-drug TACE brought a survival benefit in the initial cohort, validation cohort (before PSM), and validation cohort (after PSM) compared with single-drug TACE. The outcomes of three cohorts all showed that a significantly greater OS triple-drug chemotherapy benefit versus single-drug chemotherapy was seen in patients with large tumors (larger than 10 cm) while no survival difference was seen in patients with small tumors (10 cm or smaller). Conclusions: Triple-drug TACE seems to benefit patients with HCC larger than 10 cm in particular compared with single-drug TACE.

K E Y W O R D S
overall survival, predictive factor, propensity score matching, single-drug chemotherapy, transarterial chemoembolization, triple-drug chemotherapy initial cohort. Besides, between January 2010 and June 2015, 449 patients who met the following criteria and received the same treatment as our phase III study therapy at the First Affiliated Hospital of Sun Yat-Sen University formed the external validation cohort. Criteria of patients enrolled in the external validation cohort were the same as the phase III clinical study: (a) age ranging from 18 to 75 years; (b) the size of main tumor larger than 7 cm; (c) Child-Pugh A liver cirrhosis and adequate organ function (platelet count greater than 60000/μL; hemoglobin greater than 8.5 g/dL; and prothrombin time less than 3 seconds above control; albumin greater than 3.5 g/dL; total bilirubin less than 1.5 mg/dL; and alanine aminotransferase and aspartate aminotransferase less than 5× upper limit of normal; serum creatinine less than 1.5 × upper limit of normal); (d) Eastern Cooperative Oncology Group performance status (ECOG PS) of 0 to 1; (e) unresectable HCC at intermediate or advanced stage; and (f) with no previous treatment. Exclusion criteria were as follows: (a) evidence of extrahepatic metastases; (b) cardiac ventricular arrhythmias requiring antiarrhythmic therapy; (c) evidence of hepatic decompensation including ascites, gastrointestinal bleeding or hepatic encephalopathy.
Even though we have demonstrated that triple-drug TACE was superior to single-drug TACE, no guideline recommended the triple-drug regimen as the standard regimen. Therefore, some doctors used this triple-drug regimen while others still used single-drug regimen at the First Affiliated Hospital of Sun Yat-sen University. This study procedure was approved by the ethics committee of the Sun Yat-sen University Cancer Center and the First Affiliated Hospital of Sun Yat-sen University, and was performed in accordance with the Declaration of Helsinki. Each patient included in the study signed informed consent.

| TACE procedure
Both the initial cohort and validation cohort were conducted on the basis of our previously reported protocol. 15 Patients in the triple-drug TACE arm (triple arm) received emulsions composed of 50 mg epirubicin (H20000497; Pharmorubicin, Pfizer, Wuxi, Jiangsu, China), 50 mg lobaplatin (H20080359; Hainan Changan International Pharmaceutical Co. Ltd., Haikou, Hainan, China), 6 mg mitomycin C (H33020786; Zhejiang Hisun Pharmaceutical Co. Ltd., Taizhou, Zhejiang, China), and lipiodol (H20150099; Lipiodol UltraFluide; Guerbet Laboratories, Aulnay Sous Bois, Paris, France). The injection was stopped when the point of near stasis within the feeding artery occured. Then gelatin-sponge particles (H32024096; Gelfoam; Hanzhou alc Ltd, China) were used to achieve embolization of the tumor-feeding artery. Patients in the single-drug TACE arm (single arm) were treated with 50 mg epirubicin, lipiodol and gelatin-sponge particles as above, but without mitomycin C and lobaplatin.
Patients who received TACE with single-drug regimen or TACE with triple-drug regimen at the First Affiliated Hospital of Sun Yat-Sen University were divided into the single-drug TACE arm or the triple-drug TACE arm, respectively. Patients who received only one kind of TACE were included in analysis, and patients who received treatment crossover from TACE with single-drug regimen to TACE with triple-drug regimen or from TACE with triple-drug regimen to TACE with single-drug regimen were excluded.
All patients with HBV received antiviral therapy. Subsequent TACE was performed as the same as initial TACE and on an "on-demand" as follows: the presence of active lesions, and adequate liver function [Child-Pugh score 5-6]). The initial cohort used the original OS date. The validation cohort was censored on 31 December 2016.

| Statistical analysis
The two tailed chi-squared test for categorical variables were used to compare baseline characteristics. The Kaplan-Meier method was used to calculate survival curves, and the log-rank test was used for univariate analysis. A multivariate Cox analysis included all factors irrespective of P value in the univariate analysis. Enter method was used in cox regression (Enter all variables in the model in one single step, without checking). The significance level of 0.05 was used to confirm statistical significance. The Statistical Package used to perform analysis was SAS software (version 9.0; SAS, Cary, NC).
Retrospective variables between groups in the validation cohort might have adverse impact on the outcomes. Therefore, propensity score matching (PSM) analysis was conducted to reduce the influence of selection bias and potential confounding factors between arms, and the data after PSM formed the validation cohort (after PSM). 20,21 All parameters were included in PSM (age, sex, neutrophil:lymphocyte ratio [NLR], prothrombin time, hepatitis B surface antigen, alanine aminotransferase, aspartate aminotransferase, alkaline phosphatase, glutamyl transpeptidase [GGT], albumin (ALB), total bilirubin (TBil), alpha-fetoprotein (AFP), tumor size, tumor number, and portal vein tumor thrombus [PVTT]). Matched pairs were then formed using a one-to-one nearest-neighbor caliper of width 0.2. The purpose of this exploratory analysis was to find out possible predictors for OS benefit from triple-drug chemotherapy when compared with single chemotherapy. OS was measured from the date of initial TACE to death from any cause (or was censored at the time of data cutoff). The cutoff of continuous variables was consistent with our previous phase III study. 15 To separately evaluate the predictive value of each parameter for predicting the survival benefit, each parameter's interaction term with treatment was tested in a model containing only the baseline parameter, the treatment and their interaction. The above statistical method was applied in the validation cohort.

| Patient characteristics
In the initial cohort, 244 patients (122 patients in each arm) from our phase III study were analyzed, ranging from 21 to 75 years and including 224 men and 20 women. In the external validation cohort, 449 patients (226 patients in single arm and 223 patients in triple arm) were included in the study, ranging from 18 to 75 years and including 411 men and 38 women. After performing propensity score matching, we derived oneto-one paired cohorts (187 patients in each arm), ranging from 21 to 75 years and including 346 men and 28 women. No significantly difference was observed in the baseline characteristics of the initial cohort, validation cohort (before PSM), and validation cohort (after PSM) (

| Prognostic factors for survival
Because BCLC stage was directly related to the presence or absence of PVTT, the multivariate analysis used Model 1 ( Table 2, excluding presence or absence of PVTT) and Model 2 (Table S1, excluding BCLC stage). Table 2 showed the results of univariate analysis, and Table S2 showed the survival rate (%) depending on the age, tumor size, and TACE type separately. In the initial cohort, survival rates were statistically significantly better in the triple arm (median OS = 10.567 months; 95% CI = 8.325-12.809) than in the single arm (median OS = 5.967 months; 95% CI = 4.266-7.668) (P = 0.0003) ( Figure 1A). In the validation cohort (before PSM) and validation cohort (after PSM), survival rates were also statistically significantly better in the triple arm than in the single arm (11. Figure 1C). Multivariate analysis using Model 1 ( Table 2, excluding presence or absence of PVTT) showed that prognostic indicators for OS of the initial cohort, validation cohort (before PSM) and validation cohort (after PSM) were treatment allocation, GGT and BCLC stage; treatment allocation, NLR, ALP, GGT, ALB, AFP, and BCLC stage; treatment allocation, sex, ALB, AFP, and BCLC stage, respectively. The multivariate analysis using Model 2 (Table S1, excluding BCLC stage) showed that prognostic indicators for OS of the initial cohort, validation cohort (before PSM) and validation cohort (after PSM) were treatment allocation, GGT, and the presence or absence of PVTT; treatment allocation, NLR, ALP, GGT, ALB, AFP, and the presence

| Predictive factors of treatment benefit
Predictive factors of survival benefit from triple-drug chemotherapy in the initial cohort, validation cohort (before PSM), and validation cohort (after PSM) were summarized in Tables 3-5, respectively. Tumor size was predictive factor for the benefit of TACE with triple-drug chemotherapy in all cohorts (interaction P = 0.0452, 0.0159, and 0.0153, respectively). After stratification by tumor size, there was no survival difference between the triple-drug TACE and the single-drug TACE in the three cohorts for tumor 10 cm or smaller (Figure 2A-C). However, significantly greater triple-drug chemotherapy benefit was observed in patients with tumors larger than 10 cm in the largest diameter in the initial cohort (P < 0.0001, Figure 2D). The survival rates were also particularly better in the triple arm than in the single arm for patients with tumors larger than 10 cm in the validation cohort (before PSM) (P < 0.0001, Figure 2E) and validation cohort (after PSM) (P = 0.0003, Figure 2F).

| Safety
Our previous showed that no difference in the incidence of adverse event or serious adverse event was observed between the triple arm and single arm. 15 For the validation cohort, there was no difference in any grade or grade 3-4 adverse event between the two arms (Table S3).

| DISCUSSION
This study showed that triple-drug TACE significantly prolonged the survival of patients with unresectable HCC in all cohorts, which was consistent with our previous phase III study. 15 It may be explained that chemotherapy played a positive and major effect in TACE, and our another study showed that hepatic artery infusion chemotherapy (abandoning embolization completely and strengthening chemotherapy) had better tumor responses and less serious adverse events than TACE. 9 Then, the triple-drug TACE were common in clinical used in our hospital, 9,22 and the overall efficacy of triple-drug TACE in patients with intermediate HCC seen in another study 22 was comparable with that reported in our previous phase III study 15 (17.67 vs 13 months). Treatment allocation and BCLC stage (or PVTT) were independent negative prognostic factors in all cohorts. There were discrepancies of identified prognostic factors between this study and original study. 15 That might be explained: First, sample size varied greatly; Second, initial and validation cohort were from different hospital; Finally, the multivariate analysis of original study included three arms (triple-drug TACE with embolization arm, single-drug TACE arm with embolization, and triple-drug TACE without embolization arm) while this study only included two arms (triple-drug TACE with embolization arm and single-drug TACE arm with embolization). The purpose of this study was to find out particular patients who gained the greatest benefit from triple-drug chemotherapy. In the initial cohorts, validation cohort (before PSM), and validation cohort (after PSM), tumor size was the strongest baseline predictive factor for the benefit of triple-drug chemotherapy. A significant survival difference was seen in large tumors (>10 cm) but no survival difference was seen in small tumors (≤10 cm) between tripledrug chemotherapy and single-drug chemotherapy. Several explanations could account for this result. First, small HCCs had fewer feeding arteries and were less invasive; thus, it is feasible to completely block all tumor-feeding arteries. However, it is especially difficult for large HCCs to completely block all tumor-feeding arteries because these HCCs usually have multiple intrahepatic or extrahepatic collateral circulation. 23,24 Second, embolization could only block the hepatic artery. However, as tumors grow, more HCCs accompanied hepatic arteriovenous shunts, 25 and the hepatic artery supply decreased while the portal vein supply increased. Third, large tumors needed more embolic agents while the amount of embolic agents was limited because patients with large tumors frequently had poor liver reserve. 23,26 Finally, chemotherapy drugs can flow to the whole liver by means of a hepatic artery portal vein traffic branch that was not affected by the tumor blood supply vessels. Tumor size was also one of the strongest prognostic factors for survival in HCC, 27 and previous studies showed that TACE for large HCC had a poor prognosis. 23,28 Guidelines from the American Association for the Study of Liver Diseases showed that tumor burden (size >10 cm) is relative to contraindications for TACE. 29 Our previous studies also showed that TACE without embolization did not reduce survival, 15 and hepatic artery infusion chemotherapy had better tumor responses and less serious adverse events than TACE. 9 Therefore, embolization might be ineffective, and chemotherapy might play the major role in large HCCs. Whether embolization plays a major role in small HCCs requires further research. Our findings could explain why different studies investigating TACE with combination chemotherapy vs TACE with mono-chemotherapy had contrasting results. One study with large tumors (mean size = 8 cm) showed that TACE with combination chemotherapy had better survival than did TACE with mono-chemotherapy, 30 while another study with small tumors (mean size ≤3 cm) showed no survival difference. 31 In addition, the reason why studies of TACE vs transarterial embolization without chemotherapy [32][33][34][35][36][37] showed no survival difference could be explained. First, most studies had small tumors (mean size ≤5 cm), except for the study of Malagari K (mean size >8 cm). 37 Second, all studies used a single chemotherapeutic drug. Doxorubicin or cisplatin alone might be ineffective. Whether a single chemotherapeutic drug is sufficient to cause a survival difference requires further study. Third, different chemotherapeutic drugs might cause a survival difference. Compared with cisplatin, lobaplatin was reported to have less toxicity, better therapeutic index, and higher solubility. 38,39 There were many limitations in this study. The first limitation of this trial was the retrospective nature of the external validation cohort; however, there were no significant differences in the baseline characteristics between the two arms, and the same results were attained after PSM was used. Because this was not a prespecified hypotheses but a post hoc exploratory analysis, this result needs confirmation in other prospective trials. The second limitation was that TACE was not the current standard treatment for HCC with PVTT. However, previous studies showed that TACE was the safe and effective treatment for HCC patients with PVTT. 40,41 Finally, this study was done only in China. The predominant etiology of HCC in China was hepatitis B virus. Therefore, whether the results could be adapted to western countries where the etiology of HCC is mainly hepatitis C virus remains to be proved.
In summary, this retrospective explorative trial showed that TACE with triple-drug chemotherapy had better survival rates than TACE with single-drug chemotherapy, particularly in patients with large tumors (>10 cm). However, patients with small tumors (≤10 cm) could not benefit from the chemotherapy of TACE. Tumor size was a predictive factor for the benefit of TACE with triple-drug chemotherapy.