Introduction

Hepatocellular carcinoma (HCC) ranks as the sixth most common cancer globally and the third leading cause of cancer-related mortality.1 The majority of HCC cases are diagnosed at intermediate or advanced stages. Patients categorized with intermediate-stage HCC, defined as Barcelona Clinic Liver Cancer (BCLC) stage B or China Liver Cancer (CNLC) stage IIa/IIb, represent a substantial subgroup, comprising 20–30% of cases.2 3

Intermediate-stage HCC poses significant clinical challenges and is a major health concern worldwide. This stage often falls into a treatment gap, being beyond curative therapies like resection or ablation yet not advanced enough for systemic treatments to be the primary option. Intermediate-stage HCC exhibits high heterogeneity in tumor burden. While transarterial chemoembolization (TACE) is the preferred treatment, it fails to yield desired outcomes for patients exceeding the up-to-seven criteria (seven defined by the sum of the largest tumor size (in cm) and the count of tumors).4 5 For these patients treated with TACE alone, the median progression-free survival (PFS) was 3.0 months (95% CI: 2.1 to 4.3), and the objective response rate (ORR) was 38.7% per modified Response Evaluation Criteria in Solid Tumors (mRECIST).6 The latest BCLC version, released in 2022, categorizes BCLC-B into three groups, with systemic treatment recommended, particularly for patients with diffuse, infiltrative, or extensive bi-lobar liver involvement.7 Thus, there is an urgent need for the development of more effective therapeutic approaches for intermediate-stage HCC beyond up-to-seven criteria.

The emergence of systemic treatment with immunotherapy, particularly immune checkpoint inhibitors (ICIs), has introduced new possibilities for HCC treatment. Numerous studies using ICIs have demonstrated ORRs in advanced HCC ranging from 10% to 20%, with many showing durable responses.8–11 Various ongoing clinical trials exploring combination settings with ICIs or ICI-based therapies are anticipated to yield higher response rates and longer durability in patients with HCC. Therefore, combination settings involving ICIs or ICI-based therapies hold the potential to revolutionize the landscape of HCC treatment.

Combination therapies leverage the patient’s immune system to better recognize and combat cancer cells. By combining TACE with immunotherapy, the goal is to enhance overall treatment efficacy, potentially overcoming the limitations of each approach when used alone. This combination is theorized to not only target the tumor directly but also modulate the tumor microenvironment, augmenting the immune response against HCC. However, the effectiveness of combining TACE and ICIs for intermediate-stage HCC beyond up-to-seven criteria remains incompletely elucidated.

Despite its theoretical advantages, the application of this combined approach in intermediate-stage HCC beyond up-to-seven criteria is still in its early stages, with ongoing research focused on identifying patient populations that would derive the greatest benefit. Imaging omics represents a key area in clinical prediction. This study has conducted research in this domain to inform personalized and more effective treatment strategies.

Thus, this prospective study enrolled patients with intermediate-stage HCC beyond up-to-seven criteria who underwent TACE plus ICIs, aiming to investigate the efficacy, safety, and exploratory model of this combined therapy.

MethodsStudy design and participants

This single-arm, phase II prospective study was conducted at Zhongshan Hospital of Fudan University and registered at ClinicalTrials.gov. Eligible patients were aged ≥20 and ≤75 years and clinically diagnosed or pathologically confirmed with intermediate-stage HCC (BCLC stage B or CNLC IIa/IIb) beyond the up-to-seven criteria. Patients were either newly diagnosed or had recurrent disease more than 6 months after radical surgery. Those with chronic hepatitis B virus (HBV) infection had to have an HBV DNA viral load <500 IU/mL at screening and be on antiviral therapy per regional standard of care guidelines before initiating study therapy. Patients were required to have at least one measurable site of disease as defined by mRECIST and Response Evaluation Criteria in Solid Tumors (RECIST) V.1.1 criteria on spiral CT scan or MRI. Adequate blood count, liver enzymes, and renal function were necessary, with specific thresholds including hemoglobin ≥85 g/L, absolute neutrophil count ≥1,5 × 109/L, platelets ≥70 × 109/L, total bilirubin ≤3 × upper normal limit (ULN), aspartate aminotransferase (AST) (serum glutamic-oxaloacetic transaminase (SGOT)), alanine aminotransferase (ALT) (serum glutamic-pyruvic transaminase (SGPT))≤5 × ULN, international normalized ratio ≤1.25, albumin ≥31 g/L, serum creatinine ≤1.5 × institutional ULN, or creatinine clearance ≥30 mL/min (if using the Cockcroft-Gault formula). Patients were required to be willing and able to comply with the study protocol, including undergoing treatment, adhering to contraceptive measures, attending scheduled visits and examinations, and participating in follow-up. Female patients of reproductive potential had to have a negative urine or serum pregnancy test within 7 days prior to the start of the trial. Patients with Child-Pugh A class liver function and an Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–1 were eligible for inclusion. Additionally, normal T3 and T4 levels or T3 and T4 controlled within the normal range through treatment were required for eligibility. The study protocol is available as part of the supplementary information (online supplemental file 2).

Patients with diffuse HCC were excluded from the study. Additionally, individuals with a history of other malignant tumors within the past 3 years or concurrently (except for cured skin basal cell carcinoma and cervical carcinoma in situ) were ineligible. Other exclusion criteria included a known history of hepatic encephalopathy within the last 6 months, cardiac disease within the preceding 12 months before the first dose of the study drug, clinically significant hemoptysis or tumor bleeding of any cause within 2 weeks before the first dose of the study drug, severe unhealed wounds, ulcers, or fractures, prior treatment with TACE, uncontrolled hypertension (systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg), any active autoimmune disease or history of autoimmune disease, major surgery within 4 weeks of starting the study treatment, or subjects who had not recovered from the effects of major surgery, a history of allogeneic tissue/solid organ transplant, a urine routine test showing urine protein ≥ ++ and confirmed 24-hour urine protein content >1.0 g, prior therapy with an anti-programmed cell death protein 1 (anti-PD-1), anti-programmed cell death-ligand 1 (anti-PD-L1), anti-programmed cell death-ligand 2, anti-CD137 (4-1BB ligand, a member of the tumor necrosis factor receptor family), or anti-cytotoxic T-lymphocyte-associated antigen-4 antibody (including ipilimumab or any other antibody or drug specifically targeting T-cell co-stimulation or checkpoint pathways), and female patients who were pregnant or breastfeeding.

Procedures

Eligible patients with intermediate-stage HCC beyond up-to-seven criteria received conventional TACE alongside sintilimab (200 mg, intravenous, every 3 weeks). The initial dose of sintilimab was administered following the first TACE session once liver function recovery was confirmed, starting no earlier than 3 days after the initial TACE treatment. Subsequent doses followed a 3-week schedule. TACE procedures involved ultra-fluid lipiodol and gelatin sponge embolization, with chemotherapeutic agents (such as fluorouracil (0.5–1.0 g), oxaliplatin (50–150 mg), or epirubicin (10–30 mg)) tailored to individual patient needs. The TACE strategy was determined by the investigator based on patient-specific requirements. Enhanced MRI was conducted approximately 6 weeks (±14 days) post-TACE to evaluate treatment efficacy.12–14 The timing of subsequent TACE sessions was at the discretion of the investigator, with a minimum interval of 4 weeks between sessions.

Follow-up and outcomes

Patients underwent follow-up every 6–8 weeks, which included routine physical examination, laboratory blood tests, and enhanced MRI scans.

The primary objectives of the study were to assess PFS and safety. Secondary objectives included measuring the ORR and disease control rate (DCR) according to both RECIST V.1.1 and mRECIST criteria, as well as overall survival (OS). Safety assessments were conducted based on the National Cancer Institute’s Common Terminology Criteria for Adverse Events (V.4.0). Additionally, tumor characteristics were analyzed through radiomics to identify predictors influencing the efficacy of tumor treatment.

Statistical analysis

The results are presented as mean (SD), number (%), or median (95% CI). Survival outcomes were calculated using the Kaplan-Meier method and compared using the log-rank test. Statistical analyses were conducted using SPSS software (V.16.0; IBM).

ResultsCharacteristics of enrolled patients

Between April 2021 and February 2023, 20 patients were enrolled in the study. Baseline characteristics are summarized in table 1. The mean age was 62±11.3 years, with the majority being men (17/20, 85.0%). All patients had an ECOG status of 0. Most patients were classified as CNLC stage IIb (11/20, 55.0%). The etiology of HCC was predominantly HBV infection (11/20, 55.0%) and hepatitis C virus infection (1/20, 5.0%). Among 11 patients with confirmed hepatitis B, regular nucleoside analog therapy was initiated (tenofovir disoproxil fumarate for eight, and entecavir for the remaining three). HBV-DNA was detectable in two patients, with the rest exhibiting levels below the detection threshold. Patients underwent a median of 4 cycles of sintilimab (IQR, 2–7), and 2 sessions of TACE procedures (IQR, 1–3). The median time interval between TACE treatment and subsequent sintilimab therapy in patients was 10 days (IQR, 8–13).

Table 1

Patient baseline demographic and clinical characteristics

Antitumor activity

At a median follow-up duration of 22.0 months, the median PFS by both RECIST V.1.1 and mRECIST was 8.4 months (95% CI: 4.7 to 19.7) (figure 1A), while the median OS was not yet reached (figure 1B). 95% of patients (19/20) exhibited tumor shrinkage, with the best per cent change in viable (enhancing) tumor diameter from baseline depicted in figure 2A. Swimmer plots for each patient are illustrated in figure 2B. The ORR was 30.0% (95% CI: 14.6% to 51.9%) per RECIST V.1.1 and 60% (95% CI: 38.7% to 78.1%) per mRECIST. The DCR was 95.0% (95% CI: 76.4% to 99.1%) per both RECIST V.1.1 and mRECIST (table 2). Four patients underwent successful conversion to curative surgical resection, with one experiencing tumor recurrence 5 months post-surgery.

Figure 1

Kaplan-Meier plot of survival for progression-free survival and overall survival. (A) Kaplan-Meier curves of progression-free survival for the full analysis set (n=20); (B) Kaplan-Meier curves of overall survival for the full analysis set (n=20).

Figure 2

Characteristics of tumor response. (A) Waterfall plot of the best per cent change in viable (enhancing) tumor diameter from baseline (full analysis set, n=20); (B) swimmer plots of patients (per modified Response Evaluation Criteria in Solid Tumors). CR, complete response; PD, progressive disease; PR, partial response; SD, stable disease.

Safety profiles

Treatment-related adverse events (TRAEs) included elevated AST levels (19/20, 95.0%), followed by elevated ALT levels (18/20, 90.0%), hypothyroidism (18/20, 90.0%), decreased appetite (10/20, 50.0%), fever (7/20, 35.0%), hyperbilirubinemia (3/20, 15.0%), and rash (1/20, 5.0%). Among all 20 patients, only one experienced a grade 3 TRAE (myocarditis). The incidence of AEs observed with the combination therapy is comparable to that reported for monotherapy,15 and no new safety signals were identified in this study. Additionally, no grade 4–5 AEs or treatment-related deaths were observed (table 3). In this study, one patient discontinued sintilimab due to immunotherapy-related myocarditis, and there were no instances of temporary drug discontinuation due to adverse reactions. One patient passed away at the local hospital due to gastrointestinal bleeding after a stable tumor treatment assessment (Patient 10 in figure 2B).

Table 3

Treatment-related adverse events

RadiomicsData acquisition and radiomics feature extraction

The tumor volume of interest (VOI) was manually delineated in each sequence (T1-weighted phase, T2-weighted phase, arterial phase, venous phase, delayed phase, and diffusion-weighted phase) of MR images using the open-source tool 3D-Slicer (https://www.slicer.org). A binary dilation operation was applied to expand the spatial extent of the VOI by 1 cm around the tumor region. Radiomics features, including texture features, first-order features, and shape features, were extracted for both the tumor and peritumoral VOIs using PyRadiomics (V.3.1; https://pyradiomics.readthedocs.io) (figure 3A).

Figure 3

Comprehensive set of radiomics features, including texture features, first-order features, and shape features, were extracted for both the tumor and peritumoral volume of interest using PyRadiomics (A). ElasticNetCV was employed to perform a grid search over different values of the L1 ratio and alpha hyperparameters, using 10-fold cross-validation. The optimal alpha and L1 ratio values were determined based on achieving the lowest mean squared error (B, C). A subset of the top 10 features, determined by their importance according to the Elastic Net model with the optimal hyperparameters (D, E). The ROC curve was calculated (F). Online supplemental file 1 of radiomics codes. AUC, area under the curve; CR, complete response; DWI, diffusion-weighted imaging; LHH/HHH/LHL/LLH/HLH, L represents low-frequency, H represents high-frequency; MCC, maximal correlation coefficient; PD, progressive disease; PR, partial response; SD, stable disease; T1WI, T1-weighted imaging; T2WI, T2-weighted imaging.

Feature selection and reduction

Z-score standardization was employed to ensure that all features were on the same scale. Independent-sample t-tests and Mann-Whitney U tests were used to assess the significance of their relationship with the label, based on normality assumptions. A correlation matrix was calculated to identify highly correlated features (correlation coefficient >0.75) to reduce multicollinearity.

Machine learning model

The Elastic Net regression model was chosen for its ability to handle both L1 (Lasso) and L2 (Ridge) regularization. ElasticNetCV was used to conduct a grid search over various values of the L1 ratio and alpha hyperparameters, employing 10-fold cross-validation. Optimal alpha and L1 ratio values were determined based on achieving the lowest mean squared error (figure 3B,C). Additionally, recursive feature elimination was applied to select a subset of the top 10 features, determined by their importance according to the Elastic Net model with the optimal hyperparameters (figure 3D,E). The final Elastic Net model was trained using the selected features and the optimal hyperparameters obtained from the cross-validation step. Class probabilities were generated by the Elastic Net model for each sample in the data set, and the receiver operating characteristic (ROC) curve was calculated (figure 3F). The ROC curve demonstrates that radiomic features of the tumor and peritumoral regions can predict treatment efficacy, aiding in the selection of appropriate patients for this therapy. The radiomics analysis codes are provided in the online supplemental file 1.

Discussion

To our knowledge, this is the first prospective study to report the therapeutic efficacy and safety of TACE combined with PD-1 inhibitor for patients with intermediate-stage HCC beyond up-to-seven criteria. In our study, the ORR was 60.0% per mRECIST and 30.0% per RECIST V.1.1, with a median PFS of 8.4 months and median OS not reached, indicating acceptable safety.

HCC often reaches an intermediate stage where treatment options typically involve TACE. However, TACE alone may not suffice for disease control in patients exceeding the up-to-seven criteria, and the number of interventional treatments becomes limited by the patient’s liver function. According to the report by Kudo et al,6 the median PFS and ORR were 3.0 months (95% CI: 2.1 to 4.3) and 38.7% per mRECIST, respectively, for patients with intermediate-stage HCC beyond the up-to-seven criteria who received TACE alone. PD-1 inhibitors, such as sintilimab, are immunotherapies that enhance the immune response against cancer cells by blocking the PD-1 pathway. The phase III IMbrave150 trial demonstrated that atezolizumab plus bevacizumab achieved better OS and PFS compared with sorafenib in BCLC-B patients with unresectable HCC, highlighting the substantial potential of immunotherapy-based combination regimens in the treatment of unresectable HCC.16 Combining PD-1 inhibitors with TACE could potentially enhance the tumor’s response to treatment. TACE cuts off the tumor’s blood supply and delivers chemotherapy agents directly to the tumor, leading to tumor cell death. This process may release tumor-associated antigens, activating the immune system’s recognition and attack on the tumor. PD-1 inhibitors act on immune checkpoints, blocking the mechanism by which tumor cells evade immune system attacks through the PD-1 pathway. By alleviating immune suppression, PD-1 inhibitors enhance the immune system’s ability to attack the tumor. The synergistic effect could lead to higher tumor response rates and longer survival periods. It has been shown that, TACE is associated with lower intratumoral density of immune-exhausted effector cytotoxic and regulatory T cells (Tregs), with significant upregulation of pro-inflammatory pathways. This highlights the pleiotropic effects of TACE in modulating the tumor microenvironment and strengthens the rationale for developing immunotherapy alongside TACE.17 The combination of local control by TACE and systemic immune activation by PD-1 inhibitors may improve treatment outcomes, especially in patients who do not respond well to monotherapies. According to our study, the combination therapy has been shown to improve ORR, PFS, and DCR, with 95% of patients (19/20) experiencing tumor shrinkage. Meanwhile, 20% of patients initially deemed unresectable (4/20) were successfully converted to curative surgical resection.

Although the observation period in our study was not long enough, the combination of TACE and PD-1 inhibitors showed a favorable therapeutic response. A literature-based meta-analysis indicated that the objective response measured by mRECIST could predict OS for patients receiving locoregional therapies, and ORR could be considered as primary endpoints in phase II trials, as recommended by the European Association for the Study of the Liver guideline.18 Our study demonstrated a promising and favorable result with a 60.0% ORR per mRECIST. Hiraoka et al reported on the efficacy of 95 BCLC-B patients with HCC beyond the up-to-seven criteria who received Atezo/Bev treatment.19 The ORR of this study was only 17.7% and 42.5% per RECIST 1.1 and mRECIST, respectively. Similar results were reported for patients with unresectable HCC who received Atezo/Bev combined with TACE or lenvatinib combined with TACE, with ORRs of 61.9% and 68.3% per mRECIST,20 21 respectively. However, these were retrospective studies. TACE combined with other molecular targeted treatments and ICIs also holds promise.22–24 Results from a multicenter retrospective study (CHANCE001) showed that 78 BCLC-B and 150 BCLC-C patients who received TACE combined with PD-(L)1 inhibitors and molecular targeted treatments in real-world situations achieved an ORR of 60.1% per mRECIST.24 Meanwhile, several combinations of three treatment regimens for intermediate HCC are ongoing in phase III studies.13 The ORR per RECIST V.1.1 of the EMERALD-1 study (TACE combined with PD-L1 inhibitor (durvalumab)±bevacizumab) for patients with locoregional HCC was 41.0–43.6%. Our results suggest that TACE combined with PD-1 inhibitors may be a promising treatment regimen for intermediate-stage HCC, especially for HCC beyond the up-to-seven criteria. According to the PETAL phase Ib study, preliminary activity was observed for TACE plus pembrolizumab, and the mechanisms of efficacy were explored. 15 patients were included in the safety and efficacy population. The PFS rate at 12 weeks was 93% and the median PFS was 8.95 months (95% CI: 7.30 to NE (not estimable).25 Though our study focused on patients with intermediate-stage HCC exceeding the up-to-seven criteria, we achieved similar PFS outcomes with a manageable safety profile.

Combining two treatments with different mechanisms may reduce the resistance associated with monotherapy.26–28 For patients with unresectable HCC, effective systemic therapy (PD-1 inhibitor) combined with local therapy (TACE) can achieve a high rate of tumor regression and objective response. The treatment-related adverse reactions are controllable, which is expected to provide new options for extending the survival of patients with unresectable HCC.29 30 The safety profile in our study was also investigated. The most common AEs were AST elevation, ALT elevation, hypothyroidism, decreased appetite, fever, hyperbilirubinemia, and rash. The combination therapy might increase the rate of certain toxicities inevitably, but most cases were mild in severity. Only one patient experienced grade 3 TRAE (myocarditis). No treatment-related deaths were reported in our study.

Radiomics involves extracting a large number of quantitative features from medical images, providing detailed information about tumor characteristics not visible to the naked eye.31 In this study, analyzing radiomic features from both tumor and peritumoral regions can offer significant insights into the tumor’s biological behavior and response to treatment. The utilization of radiomic features in predicting the efficacy of combined PD-1 inhibitor and TACE treatment marks a significant advancement. It indicates that certain image-based features may indicate how well a tumor could respond to treatment, thereby assisting in personalized medicine. ROC curve analysis can aid in selecting the most predictive radiomic features, crucial for developing a robust model that accurately predicts treatment outcomes.32 By predicting treatment efficacy based on pretreatment imaging, clinicians can tailor treatment plans to individual patients, potentially leading to more effective and targeted treatments and improved patient outcomes. Identifying patients more likely to benefit from combined PD-1 inhibitor and TACE treatment can optimize patient care and resource allocation, minimizing exposure to ineffective treatments and associated side effects. Standardization in the extraction and analysis of radiomic features is needed, along with validation of predictive models in larger, independent cohorts to ensure reliability and generalizability. Combining radiomic features with other clinical and molecular biomarkers could enhance predictive power and provide a more comprehensive understanding of the tumor microenvironment and its response to therapy. In conclusion, applying radiomic features to predict the efficacy of combined PD-1 inhibitor and TACE treatment represents a promising area in cancer treatment, demonstrating the potential of integrating advanced imaging analysis with personalized treatment strategies to improve patient outcomes. However, further research and validation are necessary to fully realize its potential in clinical practice.

Although promising therapy responses were identified in the present cohort, the study has several limitations. The follow-up period was not long enough for the evaluation of OS, and this is a prospective study with a small sample size. The compelling outcomes derived from this study necessitate further substantiation via a phase 3 trial encompassing a broader cohort. Additionally, our inability to identify valid biomarkers in blood/tissue highlights the necessity for larger-scale studies to further investigate this area. Moreover, the fact that over half of the patients included in this study were HBV patients limits the broad applicability of these study outcomes, and further evidence is required to support their application to a more general population. However, patients enrolled in this study are highly homogeneous; all patients are classified as intermediate-stage HCC beyond the up-to-seven criteria, allowing for the evaluation of the efficacy and safety of this combination treatment regimen.

In conclusion, the synergistic effect of combined PD-1 inhibitors and TACE treatment in patients with intermediate-stage HCC beyond up-to-seven criteria offers a potential new treatment strategy. This strategy may present an acceptable safety profile, improve local tumor control and systemic immune response, but more research is needed to determine its optimal application.