Background: Neoadjuvant use of immune checkpoint inhibitors (ICI) is feasible and achieves pathological responses in a subset of patients with hepatocellular carcinoma (HCC). However, it is not clear whether pathological response to ICI translates into long-term survival benefit. Methods: We analysed patient-level data from 86 subjects recruited to 4 prospective phase I/II clinical trials of ICI prior to liver resection (LR) in 9 centres in the United States, Europe, and Asia and included a cohort of 23 patients (pts) receiving neoadjuvant ICI off trial. Radiological response was assessed with RECISTv1.1. Major (MPR) and complete pathological response (pCR) were considered as ≥70% and 100% non-viable tumour in the resected specimen, respectively. Pathological responses were correlated with radiologic overall response rates (ORR) and relapse-free survival (RFS). Results: Out of 109 pts treated between 10/2017 and 09/2023, 55 pts (50.5%) received a double ICI combination, 35 (32.1%) ICI monotherapy, and 19 (17.4%) an antiangiogenic/ICI combination. In 19 pts (17.4%) an adjuvant ICI course was administered for a median time of 5.5 months (m) (interquartile range [IQR] 3.7-6.0). Most of pts were male (77.6%) with underlying viral chronic liver disease (68.2%), performance status 0 (83.5%), BCLC 0-A (55%), a median tumour diameter of 6.0 cm (IQR 3.9-9.9) without portal vein thrombosis (75.7%). After a median follow-up of 27.3 m (95% CI 21.5-33.0), 35 (32.1%) relapses and 14 (12.8%) deaths occurred. Radiologic ORR was 28.5% (n=31), with 3.7% complete responses (CR, n=4) and 24.8% partial responses (PR, n=27). Out of 100 pts evaluable for a pathological response, 33% achieved MPR and 19% pCR. MPR was more likely in patients who had achieved a radiologic PR or CR (82%) as compared to those who did not (14% p<0.001), however linear correlation suggested an imperfect correlation between ORR and MPR (R² 0.43, p<0.0001). Median RFS was 43.5m (95%CI, 27.5-59.6), while median overall survival was not reached (NR). mRFS was significantly improved in pts with MPR (NR vs 28.3m [95%CI, 13.1-43.5], hazard ratio [HR] 0.27 [95%CI, 0.11-0.69], p = 0.003) and with pCR (NR vs 32.8m [95%CI, 15.2-50.4], HR 0.20 [95%CI, 0.05-0.82], p=0.013). Conclusions: This study is the first to qualify the role of MPR as a putative surrogate endpoint for recurrence-free survival in pts treated with neoadjuvant immunotherapy for HCC. Validation in larger phase III clinical trials is warranted.