Intratumoral Proton Density Fat Fraction Predicts the Outcome of HAIC Combined With PD-1 Inhibitors in Advanced...

This study aimed to explore the predictive value of intratumoral proton density fat fraction (PDFF) and the clinical efficacy of hepatic arterial infusion chemotherapy (HAIC) combined with anti-programmed cell death protein 1 (anti-PD-1) therapy in advanced hepatocellular carcinoma (HCC).

In this retrospective cohort study, patients with advanced HCC received FOLFOX-HAIC or HAIC combined with anti-PD-1 (camrelizumab). Progression-free survival (PFS) was evaluated as the time-to-event outcome, while therapeutic efficacy was assessed using tumor response rates. The Kaplan-Meier method and log-rank test were used to compare PFS. In the MRI-PDFF subset, receiver operating characteristic (ROC) analysis was used to determine the optimal PDFF cutoff for predicting nonresponse (SD or PD).

Between September 2020 and August 2025, 103 patients were included, of whom 47 received HAIC monotherapy and 56 received HAIC combined with anti-PD-1 therapy. The HAIC-PD1 group demonstrated significantly longer PFS compared with the HAIC group (HR 0.423; 95% CI 0.218-0.818; p = 0.0085), and a higher objective response rate (ORR: 46.4% vs. 21.3%; p = 0.012). In the MRI-PDFF subset, baseline intratumoral PDFF was associated with treatment response. ROC analysis identified an optimal PDFF cutoff of 2.64% for predicting nonresponse (AUC 0.784; 95% CI 0.664-0.903). Patients with PDFF < 2.64% achieved a higher ORR and longer PFS compared with those with PDFF ≥ 2.64%. Longitudinal analyses showed treatment-dependent changes in PDFF after HAIC-PD1 therapy; however, ΔPDFF did not differ significantly between responders and nonresponders.

HAIC combined with anti-PD-1 therapy demonstrated superior efficacy compared with HAIC monotherapy in advanced HCC. Baseline intratumoral PDFF may serve as a promising imaging biomarker associated with treatment response in patients receiving HAIC-PD1 therapy. Its potential prognostic relevance for time-to-event outcomes requires further validation in prospective cohorts.