We are optimistic about the promising results reported by Michael S Lee and colleagues1 regarding the use of atezolizumab with or without bevacizumab (GO30140 study) for patients with unresectable hepatocellular carcinoma. 1 The investigators found that atezolizumab plus bevacizumab showed promising efficacy (objective response rate 36%[95% CI 26-46]) in group A (single arm); the combination also resulted in longer progression-free survival (5· 6 months [95% CI 3· 6–7· 4] vs 3· 4 months [1· 9–5· 2]; hazard ratio 0· 55 [80% CI 0· 40–0· 74], p= 0· 011) in group F (randomised group) than atezolizumab alone in patients with unresectable hepatocellular carcinoma. This phase 1b study was followed by a successful phase 3 study (IMbrave150 study), 2 which showed longer overall survival and progression-free survival with atezolizumab plus bevacizumab than with sorafenib, indicating an updated first-line systemic therapy for patients with unresectable hepatocellular carcinoma, which was approved by the US Food and Drug Administration. 2 However, we still have concerns and suggestions about this combination therapy. First, the duration of treatment for long survival might be important. For group A, with long follow-up, the median progression-free survival was 7· 3 months (95% CI 5· 4–9· 9), and the median treatment duration was 8· 3 months (IQR 3· 4–12· 5) with atezolizumab and 8· 2 months (2· 8–12· 2) with bevacizumab. 1 These findings indicate that with continued treatment even progression might lead to clinical benefits, consistent with the IMbrave 150 study. 2 However, adverse events, especially serious immune-related adverse events (such as pneumonitis) and anti-VEGF-related adverse events (such as haemorrhages or hypoproteinemia) should be monitored during followup, and cost-effectiveness should be considered.

Second, this study found that atezolizumab plus bevacizumab conferred a clinically meaningful benefit, irrespective of tumour burden, and patients with high-risk disease (defined as patients with bile duct invasion, main portal vein invasion, or liver occupancy of 50% or more) had similar outcomes to those with low-risk disease. 1 However, the proportion of patients with highrisk disease in group A was only 25%, which suggests that this conclusion might not be stable. 1 Many studies have found that patients with hepatocellular carcinoma with lung or lymph metastasis have an even higher objective response rate for immune checkpoint inhibitor (ICI) monotherapy. 3, 4 However, in these patients, metastases are usually small and more likely to respond to ICIs. 3, 4 Therefore, other tumour burden indexes (such as tumour burden score) might need to be further evaluated in ICI-based treatment. Finally, although the effects of atezolizumab plus bevacizumab are promising, many patients still do not respond. Considering benefit heterogeneity, drug accessibility, and cost in real-world practice, biomarkers (such as PD-L1 expression and hepatitis C virus infection) might still be useful for predicting which patients with hepatocellular carcinoma can benefit from ICI-based therapy. 1, 5