The purpose of this study is to assess the efficacy and safety of tiragolumab, an anti-TIGIT monoclonal antibody, when administered in combination with atezolizumab and bevacizumab as first-line treatment, in participants with unresectable, locally advanced or metastatic HCC.

Per amendment version 5, following a memo issued by the Sponsor, participants receiving treatment in the atezolizumab plus bevacizumab plus tiragolumab arm are recommended to discontinue tiragolumab treatment unless the investigator decides the benefit outweighs the risk. Participants receiving treatment in atezolizumab plus bevacizumab plus placebo arm must discontinue placebo treatment. Participants may continue receiving active treatment(s) per protocol until loss of clinical benefit or unacceptable toxicity, whichever occurs first.

Inclusion Criteria:

• Locally advanced or metastatic and/or unresectable HCC with diagnosis confirmed by histology/cytology or clinically by American Association for the Study of Liver Diseases (AASLD) criteria in cirrhotic participants
• Disease that is not amenable to curative surgical and/or locoregional therapies
• No prior systemic treatment for locally advanced or metastatic and/or unresectable HCC-measurable disease according to RECIST v1.1
• Eastern cooperative oncology group (ECOG) performance status of 0 or 1 within 7 days prior to randomization
• Child-pugh Class A within 7 days prior to randomization
• Adequate hematologic and end-organ function
• Female participants of childbearing potential must be willing to avoid pregnancy within 5 months after the final dose of atezolizumab, within 6 months after the final dose of bevacizumab, and within 90 days after the final dose of tiragolumab/placebo
• Male participants with a female partner of childbearing potential or pregnant female partner must remain abstinent or use a condom during the treatment period and for 6 months after the final dose of bevacizumab and for 90 days after the final dose of tiragolumab/placebo to avoid exposing the embryo.

Exclusion Criteria:

• Pregnancy or breastfeeding within 5 months after the final dose of atezolizumab, within 6 months after the final dose of bevacizumab, and within 90 days after the final dose of tiragolumab/placebo
• Prior treatment with cluster of differentiation 137 (CD137) agonists or immune checkpoint blockade therapies
• Treatment with investigational therapy within 28 days prior to initiation of study treatment
• Treatment with locoregional therapy to liver within 28 days prior to initiation of study treatment, or non-recovery from side effects of any such procedure
• Treatment with systemic immunostimulatory agents
• Treatment with systemic immunosuppressive medication
• Untreated or incompletely treated esophageal and/or gastric varices with bleeding or that are at high risk for bleeding
• A prior bleeding event due to esophageal and/or gastric varices within 6 months prior to initiation of study treatment
• Active or history of autoimmune disease or immune deficiency
• History of idiopathic pulmonary fibrosis, organizing pneumonia, drug-induced pneumonitis, or idiopathic pneumonitis, or evidence of active pneumonitis on screening chest computed tomography (CT) scan
• History of malignancy other than HCC within 5 years prior to screening, with the exception of malignancies with a negligible risk of metastasis or death
• Mixed histology or other subtypes/variants of HCC, including, but not limited to, known liver adenocarcinoma, fibrolamellar HCC, sarcomatoid HCC, other rare HCC variant, or mixed cholangiocarcinoma and HCC
• Co-infection with hepatitis B virus (HBV) and hepatitis C virus (HCV)
• Acute epstein-barr virus (EBV) infection or known or suspected chronic active EBV infection
• Symptomatic, untreated, or actively progressing central nervous system (CNS) metastases.

For eligible studies, qualified researchers may request access to individual patient level clinical data. See Roche's commitment to transparency of clinical study information here: https://go.roche.com/data\_sharing

The study has been unblinded as of amendment version 5.