IMbrave251: Final analysis of atezolizumab (atezo) + lenvatinib (lenva) or sorafenib (sora) vs lenva or sora alone in locally advanced or metastatic hepatocellular carcinoma (LA/mHCC) previously treated with atezo and bevacizumab (bev).

4002 Background: Several immuno-oncology therapies are available for 1L HCC, but data are limited after progression. IMbrave251 (NCT04770896) is the first global randomized Phase 3 trial to assess atezo + lenva or sora (TKIs) vs lenva or sora alone in patients (pts) with LA/mHCC that progressed after atezo + bev treatment (tx). This is the final analysis. Methods: Eligible pts had LA/mHCC that was responsive to atezo + bev tx but then progressed (atezo + bev was given for ≥4 cycles and ≥2 tumor assessments, of which ≥1 showed stable disease or complete/partial response). In the atezo + TKI arm, pts received atezo (1200 mg Q3W) + lenva (8 mg for pts < 60 kg or 12 mg for pts ≥60 kg QW) or sora (800 mg QW). Pts in the TKI-only arm received lenva or sora alone. The primary endpoint was OS. Secondary endpoints included PFS, ORR by RECIST 1.1, DOR, time to tumor progression (TTP), quality of life, and safety. Stratification factors were disease etiology (hepatitis B/C infection vs non-viral), ALBI score (1 vs 2/3), AFP ( < 400 vs ≥400 ng/mL), and TKI therapy (lenva vs sora). Results: IMbrave251 enrolled 557 pts (full analysis population). Median duration of follow-up was 12 mo. Pts were assigned to receive atezo + TKI (n = 279) or TKI-only (n = 278). Most pts in both arms received lenva (92.1%) vs sora (7.9%). At clinical cutoff (Oct 10, 2025), median (m)OS with atezo + TKI vs TKI-only was 14.6 vs 12.5 mo (stratified HR 0.88; P = 0.2115) and mPFS was 4.3 vs 4.8 mo (stratified HR 1.05), respectively. Confirmed ORR was 7.5% vs 5.8% with atezo + TKI vs TKI-only. DOR was 10.2 vs 6.9 mo, respectively (stratified HR 0.47) (Table). Additional subgroup analyses and complete duration of systemic therapy will be shown. In the atezo + TKI vs TKI-only arm, the median tx duration was 5.6 vs 4.7 mo for lenva and 2.9 vs 2.3 mo for sora. No new or unexpected safety signals were seen (safety analysis n = 554; Table). Conclusions: IMbrave251 did not meet its primary OS endpoint but showed that the addition of atezo to TKIs had a manageable safety profile. This study provides clinically relevant Phase 3 evidence addressing an important, unanswered clinical question regarding 2L tx sequencing after immunotherapy and establishes a new OS benchmark for 2L tx in LA/mHCC. Clinical trial information: NCT04770896 . Efficacy (95% CI) Atezo + lenva or sora(n=279) Lenva or sora(n=278) Stratified HR (95% CI) mOS, mo 14.6 (12.6, 16.8) 12.5 (10.7, 15.5) 0.88 (0.72, 1.08) P =0.2115 mPFS, mo 4.3 (4.1, 5.4) 4.8 (4.2, 5.6) 1.05 (0.88, 1.25) ORR, % 7.5 (4.7, 11.3) 5.8 (3.3, 9.2) Δ1.77 (−2.72, 6.26) a mDOR, mo 10.2 (5.6, 13.0) 6.9 (5.5, 11.7) 0.47 (0.18, 1.23) mTTP, mo 4.4 (4.2, 5.5) 5.4 (4.2, 5.6) 1.06 (0.87. 1.28) Safety, n (%) n=279 n=275 Grade 3/4 | 5 AE 207 (74.2) | 15 (5.4) 182 (66.2) | 16 (5.8) - Serious AE | AE leading to withdrawal from any study drugs 112 (40.1) |33 (11.8 101 (36.7) |19 (6.9) - a Difference in confirmed ORR (95% CI).