BACKGROUND: Tivantinib (ARQ 197), a selective, oral MET inhibitor, improved overall survival and progression-free survival compared with placebo in a randomised phase 2 study in patients with high MET expression (MET-high) hepatocellular carcinoma previously treated with sorafenib. The aim of this phase 3 study was to confirm the results of the phase 2 trial.METHODS: We did a phase 3, randomised, double-blind, placebo-controlled study in 90 centres in Australia, the Americas, Europe, and New Zealand. Eligible patients were 18 years or older and had unresectable, histologically confirmed, hepatocellular carcinoma, an Eastern Cooperative Oncology Group performance status of 0-1, high MET expression (MET-high; staining intensity score ≥2 in ≥50% of tumour cells), Child-Pugh A cirrhosis, and radiographically-confirmed disease progression after receiving sorafenib-containing systemic therapy. We randomly assigned patients (2:1) in block sizes of three using a computer-generated randomisation sequence to receive oral tivantinib (120 mg twice daily) or placebo (twice daily); patients were stratified by vascular invasion, extrahepatic spread, and α-fetoprotein concentrations (≤200 ng/mL or >200 ng/mL). The primary endpoint was overall survival in the intention-to-treat population. Efficacy analyses were by intention to treat and safety analyses were done in all patients who received any amount of study drug. This study is registered with ClinicalTrials.gov, number NCT01755767.FINDINGS: Between Dec 27, 2012, and Dec 10, 2015, 340 patients were randomly assigned to receive tivantinib (n=226) or placebo (n=114). At a median follow-up of 18·1 months (IQR 14·1-23·1), median overall survival was 8·4 months (95% CI 6·8-10·0) in the tivantinib group and 9·1 months (7·3-10·4) in the placebo group (hazard ratio 0·97; 95% CI 0·75-1·25; p=0·81). Grade 3 or worse treatment-emergent adverse events occurred in 125 (56%) of 225 patients in the tivantinib group and in 63 (55%) of 114 patients in the placebo group, with the most common being ascites (16 [7%] patients]), anaemia (11 [5%] patients), abdominal pain (nine [4%] patients), and neutropenia (nine [4%] patients) in the tivantinib group. 50 (22%) of 226 patients in the tivantinib group and 18 (16%) of 114 patients in the placebo group died within 30 days of the last dose of study medication, and general deterioration (eight [4%] patients) and hepatic failure (four [2%] patients) were the most common causes of death in the tivantinib group. Three (1%) of 225 patients in the tivantinib group died from a treatment-related adverse event (one sepsis, one anaemia and acute renal failure, and one acute coronary syndrome).INTERPRETATION: Tivantinib did not improve overall survival compared with placebo in patients with MET-high advanced hepatocellular carcinoma previously treated with sorafenib. Although this METIV-HCC trial was negative, the study shows the feasibility of doing integral tissue biomarker studies in patients with advanced hepatocellular carcinoma. Additional randomised studies are needed to establish whether MET inhibition could be a potential therapy for some subsets of patients with advanced hepatocellular carcinoma.FUNDING: ArQule Inc and Daiichi Sankyo (Daiichi Sankyo Group).