Due to its insidious onset, approximately 80% of patients with hepatocellular carcinoma (HCC) are first diagnosed with intermediate or advanced stage, which means that they are ineligible for curative therapies such as resection, transplantation, or ablation (1, 2). For these patients, transarterial chemoembolization (TACE) is the major recommendation according to several international guidelines if they have preserved liver function and good performance status (3–5). In real-world clinical practice, especially in Asia, TACE is also widely applied in intermediate and advanced stages (6). By embolizing targeted arteries of the lesions to achieve tumor control, TACE induces ischemic and hypoxic changes that lead to an increase in vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF) expression locally in the residual lesion(s), which theoretically promotes recurrence or progression (7–9). In addition, repeated TACE may lead to the deterioration of liver function, resulting in poor prognosis (10).

Sorafenib, a potent multikinase inhibitor that mainly targets VEGF receptors, RAF, and PDGF receptor, is the major treatment approach for advanced HCC (2, 11, 12). It has antiangiogenic and antiproliferative effects that inhibit tumor growth and disrupt tumor microvasculature (11, 12). Therefore, the combination of TACE and sorafenib should have synergistic action for unresectable HCC. A preclinical study identified that the combination of TACE and antiangiogenic therapy reduces tumor volume and vessel density and prolongs survival when compared to TACE monotherapy (13). Nevertheless, several randomized controlled trials (RCTs) demonstrated a negative benefit for such combination therapy (14–16). Apart from the study design, including the treatment duration of sorafenib and sequence of the combination treatment, the high heterogeneity of the included patients is also the main reason contributing to the negative results of these RCTs (14–17).

Different from other solid tumors, the prognosis and treatment algorithms of HCC are based not only on tumoral characteristics but also on liver function, which leads to high heterogeneity in HCC (18). Currently, the most widely applied model to assess liver functional status is the Child-Turcotte-Pugh (CTP) grade. Nevertheless, the CTP grade was initially established for patients with cirrhosis and portal hypertension undergoing surgery for variceal bleeding rather than for HCC (19, 20). In addition, the CTP grade includes two objective variables: grading of ascites and encephalopathy. Such subjectivity influences the accuracy of the CTP grade for HCC in clinical practice (21).

In 2015, Johnson and colleagues reported a new model, the albumin-bilirubin (ALBI) grade, based solely on albumin and bilirubin, as an alternative method for liver function assessment for HCC (22). Afterward, several studies compared the prognostic performance between the ALBI and CTP grades in HCC cohorts, and the majority of them demonstrated that the ALBI grade performed at least no worse than the CTP grade (23–27). Considering its easy application and objectivity, the ALBI grade has the potential to act as an alternative method to the CTP grade. Notably, the patients included in the negative RCTs about TACE combined with sorafenib for HCC mentioned above were CTP grade A or B7-B8 (14–16). Previous analyses suggested that the ALBI grade performed better than the CTP grade in discriminatory ability for HCC patients with CTP grade A treated with sorafenib monotherapy (25). Therefore, the ALBI grade may have a wider application in clinical research about sorafenib for HCC. Nevertheless, no studies have yet been reported regarding the discriminatory ability comparison between the ALBI and CTP grades for HCC patients treated with TACE combined with sorafenib.

Using artificial intelligence, this study was carried out to compare the prognostic performance associated with overall survival between the ALBI and CTP grades for HCC patients treated with TACE combined with sorafenib.

In this multicenter retrospective series of Chinese HCC patients treated with TACE combined with sorafenib, we demonstrated that the ALBI grade and CTP grade provide comparable prognostic performance using artificial intelligence. The majority of ALBI grade 1 patients were also classified as CTP A5, and a statistically significant difference in median OS was observed between CTP A5, A6, and B for patients with ALBI grade 1.

By combining sorafenib with TACE, a decrease in angiogenesis after the administration of TACE and improved efficacy of TACE should be expected (13). In addition, the treatment effect on liver function may be suppressed by reducing the frequency of TACE through this combination treatment (10). Therefore, the combination treatment should achieve synergetic efficacy for HCC. Nevertheless, the majority of the RCTs (Post-TACE, SPACE, and TACE-2) comparing TACE combined with sorafenib vs. TACE monotherapy demonstrated negative results so far (14–16). Apart from the study design (timing of sorafenib administration in Post-TACE trial, unsuitable scheduled TACE in the SPACE trial, and unsuitable progression assessment in the TACE-2 trial), the heterogeneity of HCC may be another major reason contributing to the negative results of these trials (17). Unlike other solid tumors in which the prognosis is mainly based on the tumor burden, the prognosis of HCC is also influenced by liver function and general status (18).

Currently, the most widely accepted liver function assessment tool for HCC is the CTP grade, which was initially devised to assess the prognosis of patients with cirrhotic-related portal hypertension undergoing surgery for variceal bleeding (19, 20). Nevertheless, several issues should be noted regarding the CTP grade (34). First, the grade includes two subjective variables: grade of ascites and encephalopathy. Second, albumin levels and grade of ascites are inadequate to put together for one grading system, since they are interrelated variables. Third, for the majority of patients, a score of 1 is obtained if converting the prothrombin time into the international normalized ratio. More importantly, for those with CTP A, which is the major inclusion criterion for liver function in many RCTs, the prognosis also varies (25). Therefore, a new assessment tool that overcomes the limitations of the CTP grade mentioned above is warranted for HCC, especially for the study design of HCC-related RCTs.

The ALBI grade was introduced in 2015 by Johnson PJ and colleagues with the aim of overcoming the limitations of the CTP grade by calculating the score based on only two objective variables (serum albumin and bilirubin levels) (22). The prognostic performance of the ALBI grade was then validated and compared in several studies with different kinds of treatments for HCC, with the conclusion that the ALBI grade performed as least no worse than the CTP grade (23–27). Previous studies found that the ALBI grade has the ability to stratify prognosis for patients treated with sorafenib monotherapy in the CTP A group (25). Nevertheless, no effort has been made to explore the performance of the ALBI grade for patients treated with TACE combined with sorafenib. To the best of our knowledge, this study is the first multicenter study comparing the ALBI and CTP grades in a large number of patients treated with TACE combined with sorafenib.

This study demonstrated that the prognostic performance and discrimination of the ALBI grade and the CTP grade were comparable for HCC patients treated with TACE combined with sorafenib. In addition, the ALBI grade has the ability to stratify prognosis for the CTP A group. Notably, the ANN model showed that the importance of the ALBI grade associated with OS was higher than that of the CTP grade, which indicated that the ALBI grade might have a slightly better performance than the CTP grade. Although no significant difference was observed between the ALBI grade and the CTP grade, the easy application and objectiveness of the ALBI grade make it more applicable in large-scale multicenter or international studies. Furthermore, with the ability to select subgroups with better prognosis within the CTP A group, the ALBI grade has the potential to be considered an alternative approach in RCTs, especially for TACE combined with sorafenib for the treatment of HCC.

The interaction among independent risk factors that associated with prognosis in clinical is nonlinear, which makes it relatively inaccurately to distinguish the importance and association with prognosis of each independent risk factor when using conventional linear discriminant analysis (35, 36). ANN has been identified outperform conventional discriminant analysis based on its advances with computer technology application to model a biological neural system structurally and functionally (35, 36). Nomograms are an effective way to predict survival outcomes when the data used to build the nomograms reflect the evolution of the question the nomogram is tasked to answer (37). In this study, the data is retrospectively collected from three academic hospitals in China. Patient and data management is performed in a standard way.

This study has several limitations. First, the retrospective nature of the study may lead to selection bias of the included patients. Nevertheless, the majority of the baseline characteristics of the included patients were not significantly different among the three cohorts. Second, the median OS in validation cohort 1 was longer than that in the other two cohorts. The longer median OS in validation cohort 1 may be due to the better general status and lower tumor burden of these patients compared to those in the other two cohorts. The validation results showed that the prognostic performance of the nomograms was high in validation cohort 1 and comparable with validation cohort 2. Third, considering treatment safety, patients with ALBI grade 3 were excluded. Fourth, we did not analyze other outcomes, such as progression-free survival or the newly introduced time to untreatable progression, that were used in the recently published positive trial about TACE combined with sorafenib for HCC, the TACTICS trial. Finally, and importantly, post-TACE and sorafenib treatment was hard to follow up and analyze; it may be a risk factor associated with survival. Nevertheless, even well-designed RCTs are also hard to manage post-treatment after targeted therapy. Further studies with well-designed protocols and detailed data on this topic are warranted.

In conclusion, the ALBI grade and CTP grade have comparable prognostic performance and discrimination for HCC patients treated with TACE combined with sorafenib. Considering its objectiveness and easy application, the ALBI grade has the potential to be an alternative for liver function assessment tool. ALBI grades 1 and 2 have the potential to serve as stratification factors for clinical trials of TACE combined with systemic therapy. Additional well-designed studies with more detailed information are warranted.

The raw data supporting the conclusions of this manuscript will be made available by the authors, without undue reservation, to any qualified researcher.

The study was approved by the institutional review boards at the three participating hospitals, and the requirement for informed consent was waived due to its retrospective nature. The study was performed in accordance with the Declaration of Helsinki. The studies involving human participants were reviewed and approved by The First Affiliated Hospital of Soochow University. The ethics committee waived the requirement of written informed consent for participation.

C-FN, X-LZ, B-YZ, Z-PY, J-HS, LZ, W-SW, and ZL contributed to the study concept and design. LZ, Z-HH, M-JY, G-HZ and SZ contributed to the acquisition of clinical data. B-YZ contributed to the statistical analysis. B-YZ wrote the first draft of the manuscript. C-FN and X-LZ supervised and oversaw the study. All authors contributed to the article and approved the submitted version.

This study was supported by the National Natural Science Foundation of China (81901847, 81771945, and 81971713), the Jiangsu Medical Innovation Team (CXTDB2017006), the Natural Science Foundation of Jiangsu Province (BK20190177), the Natural Science Foundation of Zhejiang Province (LZ18H180001), the Suzhou Science and Technology Youth Plan (KJXW2018003) and “Six One Projects” for High-Level Health Personnel in Jiangsu Province (LGY2018077). Funding source had no involvement in the financial support for the conduct of the research and preparation of the article.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The reviewer JJ declared a past co-authorship with several of the authors B-YZ, J-HS, G-HZ and C-FN to the handling editor.