This retrospective study reviewed data of 854 unresectable HCC patients receiving TACE-based therapy from April 2016 to December 2021 (Fig. 1).

The flowchart of patients’ selection

The inclusion criteria were as follows: (1) histologically, cytologically or clinically confirmed diagnosis of HCC according to the European Association for the Study of the Liver (EASL) criteria[5]; (2) at least 1 measurable lesion according to modified Response Evaluation Criteria in Solid Tumors (mRECIST, version 1.1); (3) with BCLC stage B or C; (4)received TACE or TACE + TKIs + ICIs as the first-line treatment, (5) received at least one cycle of TACE based therapy.

The exclusion criteria were as follows: (1) with Child–Pugh C liver function; (2) Eastern Cooperative Oncology Group (ECOG) score of ≥ 3 points; (3) liver transplantation, liver resection, or local–regional therapies before the TACE procedure; (4) currently had or had a history of malignant tumors in addition to HCC; (5) incomplete data;(6) not meeting the definition of combination therapy.

In the end, a total of 456 patients were selected for the study. Patients were assigned into 2 groups according to different treatment strategies: TACE monotherapy group and TACE in combination with TKIs and ICIs therapy group (combination group) (Fig. 1). Combination therapy was defined as treatment with TKIs and ICIs within 1 month before, or within 3 months after the first TACE procedure, while the interval between TKIs and ICIs must be within 1 week in the TACE + TKIs + ICIs group. To reduce the impact of potential confounders, we employed propensity score matching, performed 1:2 optimal pair matching to balance the covariates between the treatment groups (Table 1).

This study was designed and performed according to the Declaration of Helsinki and was approved by the Medical Ethics Committee of Nanfang Hospital, Southern Medical University, and written, informed consent was obtained from each patient to retrospectively review and report on their medical records.

The TACE protocol was performed as the previous study stated. Briefly, an emulsion mixed with lipiodol (5–20 mL) and adriamycin (20–50 mg) and subsequent 500–700 μm absorbable gelatin sponge particles were injected into the tumor feeding artery for chemoembolization.

TKIs was given at a metronomic oral dosage (400 mg, twice daily, for sorafenib; 8 mg, daily, for lenvatinib; 80 mg, daily, for Regorafenib; 250 mg, daily, for Apatinib).

Toripalimab was given at a fixed dose of 240 mg; Camrelizumab, Sintilimab and Tislelizumab was given at a fixed dose of 200 mg; Atezolizumab was given at a fixed dose of 1200 mg; Pembrolizumab was given at 2 mg/kg body weight; Nivolumab was given at 3 mg/kg body weight; All ICIs were given every 3 weeks intravenously.

The grading of adverse events related to TKIs, and ICIs were conducted according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE 5.0). If adverse events ≥ 3 grade (defined as serious adverse events, SAEs) were observed, dosages of TKIs and ICIs were reduced, suspended, or discontinued.

The primary endpoints of this study were OS (defined as the interval from treatment initiation (TACE or the first dose of TKIs/ICIs administration) to patient death or the last follow-up) and progression free survival (PFS) (defined as the time from treatment initiation to the first reported disease progression or death from any cause). The secondary endpoints were objective response rate (ORR) and disease control rate (DCR), according to the tumor response evaluated by two experienced radiologists using mRECIST. The ORR was defined as the percent of the total number of patients who had a complete response (CR) and partial response (PR), while DCR was defined as the percent of the total number of patients who had a CR, PR and stable disease (SD). Adverse events (AEs) were graded according to NCI-CTCAE 5.0. In addition, we had analyzed the predictors of affecting OS in the study. Whether the timing of TACE in combination with TKIs and ICIs affects OS was further explored in the TACE + TKIs + ICIs group.

To address the imbalance of potential confounders between two groups, we performed 1:2 optimal pair matching, propensity scores were calculated using logistic regression, taking into account the aforementioned demographic and clinical characteristic, including Age, Gender, BCLC, ECOG, Tumor Number, Tumor Size, Extrahepatic metastasis, PVTT, AFP, Ascites, Child–Pugh, WBC, NEU, ALT, AST, TBIL, ALB, PT, PLT, HBsAg, HBsAb, HBeAg, HBeAb, HBcAg, HBVDNA, and HCV, TACE time for both groups.

The patient characteristics were summarized using median with IQR for continuous variables and frequencies with proportions for categorical variables. Student’s t-test or Mann–Whitney U test was used to analyze continuous variables. Chi-squared test or Fisher exact test was applied to analyze categorical variables. The difference in PFS and OS between the two groups were compared with the use of a log-rank test. The survival curves were plotted by the method of Kaplan–Meier.

Cox proportion hazards model was used for multivariable analyses on the propensity-matched sample. Forest plots were used to display these data.

A two-tailed P-value of < 0.05 was considered statistically significant. All the above statistical analyses were performed using R (version 4.2.2; R Project for Statistical Computing, http://

www.r-project.org), MSTATA software and SPSS (version 25.0; IBM, SPSS).