Use of pathological response versus RECIST assessment in predicting relapse-free survival following neoadjuvant immunotherapy for hepatocellular carcinoma.

633 Background: Immune checkpoint inhibitors (ICI) administered prior to liver resection (LR) lead to pathological responses in patients with hepatocellular carcinoma (HCC). However, the relative value of pathological versus radiological response in predicting relapse-free survival (RFS) remains unclear. Methods: We pooled patient-level data from 111 patients (pts) with HCC receiving ICI prior to LR as part of 5 phase I/II trials and observational clinical studies conducted in 12 centres in the United States, United Kingdom and Asia, as part of an academic consortium (NeoHCC). Pathological response was measured as the percentage of non-viable tumour in the resected specimen, with major (pMR) and complete pathological response (pCR) corresponding to ≥70% and 100% tumour regression. Radiological overall response rate (ORR) was assessed by RECIST v1.1 and modified RECIST (mRECIST) criteria. We correlated pathological response and ORR with RFS using Cox regression. Results: Pts received preoperative ICI between Oct 5, 2017, and Nov 15, 2023, mostly ICI combinations (69%, n=76). Most pts were male (78%, n=87) with viral chronic liver disease (66%, n=73), BCLC stage A HCC (55%, n=61). ORR was 28% per RECIST v1.1 and 32% per mRECIST criteria (available for n=81). Out of the 104 pathologically evaluable pts, pMR and pCR rates were 32% (n=33) and 18% (n=19). Radiological response by RECIST v1.1 showed a significant correlation with pathological response (R 2 = 0.43, p<0.001). When using RECIST v1.1, 74% of pts with ORR achieved pMR vs 14% of those without ORR (n=23/31 vs 10/73, p<0.001). However, 30% of pMR were not predicted by ORR (n=10/33). The discrepancy between pMR and ORR decreased using mRECIST. ORR per mRECIST was 83% in pMR pts (n=19/23), whereas it was 10% in non-pMR (n=5/51). After a median follow-up of 27.2 mos (95%CI 22.3-32.1), median RFS for the whole cohort was 43.6 mos (95%CI 28.3-NE). Achievement of pMR, pCR and ORR was associated with improved RFS (Table). However, reduction in the risk of relapse and/or death was higher in pts achieving pMR or pCR than ORR, regardless of the use of RECIST v1.1 or mRECIST. Conversely, pts without pMR had a lower mRFS than pts without ORR (28.3 mos [95%CI 12.8-43.8] and 32.8 mos [95%CI 13.7-51.9], respectively). Conclusions: Whilst radiological responses to neoadjuvant ICI are associated with improved RFS in pts with HCC, achievement of pMR is more accurate than RECIST and mRECIST-based responses in predicting for RFS, with lack of pMR identifying pts at a higher risk of relapse or mortality. Cox regression for RFS HR (95% CI) p value pCR 0.19 (0.05-0.78) 0.02 pMR 0.25 (0.10-0.66) 0.005 RECIST 1.1 0.34 (0.13-0.86) 0.02 mRECIST 0.38 (0.13-1.11) 0.08 RFS, relapse-free survival; HR, hazard ratio; CI, confidence interval; pCR, complete pathological response; pMR, major pathological response; mRECIST, modified RECIST.