The Institutional Review Board approved this retrospective study conducted across multiple institutions (IRB: B2022-694-01). The study adhered to the principles of the 1975 Helsinki Declaration, and the need for written informed consent was waived due to its retrospective design. Patient selection, data collection, and analysis followed the STROBE guidelines for observational studies [20]. The data from this study were uploaded to https://www.researchregistry.com/ (ID: researchregistry9425).
Between January 2009 and December 2022, 5,268 consecutive patients with intermediate-stage HCC, as defined by the BCLC guideline, received either initial cTACE or the FOLFOX regimen of HAIC. These patients were identified from an in-house medical database, with data source distribution detailed in Additional file 1: Table S1. Inclusion criteria were: (a) age over 18; (b) ECOG performance status of 0 or 1; (c) Child-Pugh class A or B liver function; (d) HCC diagnosis confirmed either pathologically or through AASLD clinical criteria [3]; (e) BCLC B stage verified by liver experts; (f) contrast-enhanced CT or MRI performed within 4 weeks before IAT. Patients were excluded if they: (a) had received previous antitumoral treatment before IAT; (b) had current or past other malignancies besides HCC; (c) had inadequate image quality for assessment; (d) underwent simultaneous TACE and HAIC treatment; (e) received drug-eluting beads TACE (dTACE); or (f) lacked follow-up information. Figure 1 outlines the inclusion and exclusion criteria for patients.
The TACE, HAIC procedures, and criteria for IAT discontinuation are outlined in Additional file 1: E1.1-1.2. The multidisciplinary tumor board selected a therapeutic protocol that includes immune checkpoint inhibitors (ICIs), tyrosine kinase inhibitors (TKIs), and sequential locoregional ablation, based on tumor progression or shrinkage. Details are available in Additional file 1: E1.3-1.4 and Additional file 1: Table S2.
Baseline and pre-IAT clinical data, including demographics, etiologies of chronic liver disease, liver function tests, and laboratory results such as alpha-fetoprotein (AFP), bilirubin, and albumin, were collected. The albumin-bilirubin (ALBI) score, calculated before IAT using the formula (log10 bilirubin [μmol/L] × 0.66) + (albumin [g/L] × -0.085), classifies ALBI grades as follows: grade 1 (≤ -2.60), grade 2 (> -2.60 to -1.39), and grade 3 (> -1.39) [21, 22].
The contrast-enhanced CT or MRI images before IAT were reviewed independently by two fellowship-trained abdominal radiologists at each participating center, each with 5-10 years of experience in liver imaging. Reviewers knew all patients had HCC but were blinded to other clinical, histopathologic, therapeutic, and follow-up details. Imaging features were assessed individually for each patient. For patients with multiple tumors, the imaging characteristics of the largest tumor were evaluated. When mixed radiologic patterns were present, the radiologic feature was determined based on the most dominant type. Radiologic features were classified into two categories: non-infiltrative and infiltrative types [23, 24]. Infiltrative HCC often blends with the cirrhotic parenchymal background and lacks demarcation of a discrete mass. Key CT/MRI features diagnostic of infiltrative HCC include: 1) A poorly marginated hepatic lesion; 2) Absence of fibrotic pseudocapsule; 3) Variable (heterogeneous or homogeneous) signal abnormalities; 4) Inhomogeneous contrast washout during portal venous or delayed phases. Additionally, the non-infiltrative type includes pseudo-capsulated, pseudocapsule breakthrough, and confluent multinodular subtypes. A senior radiologist with over 20 years of experience in liver imaging resolved any inter-rater disagreements at each participating center. Figure 2 summarizes the radiologic features of the four HCC types.
HCC patients after IAT underwent regular follow-ups at one month and then every three to six months using serum AFP levels and imaging techniques until death or the final follow-up on October 31, 2023. The primary outcome, based on imaging findings, was classified according to the modified Response Evaluation Criteria in Solid Tumors (mRECIST) [25] into complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). ORR is the percentage of CR and PR, while disease control rate (DCR) is the percentage of ORR and SD maintained for more than 4 weeks from initial radiological confirmation. Additional file 1: E1.5 outlines the response criteria for IAT. The secondary endpoints included progression-free survival (PFS) and overall survival (OS). PFS was measured from the start of IAT to the first confirmed tumor recurrence, identified radiologically or histologically as HCC, tumor-in-vein, or distant metastasis according to AASLD criteria, or death from any cause. OS was defined as the duration from the initiation of IAT to death from any cause. Additional file 1: Table S3 shows the post-study treatments administered during follow-up. Additional file 1: Figure S1-2 presents the medical records of patients treated with either cTACE or FOLFOX-HAIC by two radiologic partners. The third outcome involved IAT-related adverse events as defined by the Common Terminology Criteria for Adverse Events version 4.0 [26].
Statistical analysis was performed using SPSS version 23.0 (IBM Corp., NY, USA) and the RMS package in R software version 3.5.3. Inter-rater agreement was assessed with the intraclass correlation coefficient (ICC) for radiologic patterns and Cohen's kappa for binary imaging features. Propensity score matching (PSM) was performed using a 1:1 nearest-neighbor algorithm to balance the groups. Propensity scores from PSM were used to estimate case weights via inverse probability treatment weighting (IPTW) [27]. The weights for the HAIC group were calculated as the inverse of the propensity score, while the weights for the TACE group were calculated as the inverse of one minus the propensity score.
The Kaplan-Meier method, along with the log-rank test, was used to compare cumulative survival. Although a formal sample size calculation was not performed beforehand, the large number of events relative to the variables in the multivariable regression analysis ensured compliance with the 'ten events per variable' rule, supporting the accuracy of the regression estimates. Univariate and multivariable analyses of independent prognostic factors were conducted using a forward stepwise Cox regression model.