Reasons for those failures have been discussed previously: poor understanding of drivers of progression, flaws in trial design, underestimation of liver toxicity, marginal antitumour activity, and an absence of trial enrichment with adequate biomarkers of benefit.3 Many histology-based, molecular-based or immune-based classifications have now been published to assist that enrichment.4 In The Lancet Oncology, Shukui Qin and colleagues report their evaluation of the safety and activity of camrelizumab, a new immune checkpoint inhibitor developed by Jiangsu Hengrui Medicine. Unfortunately, only 30 (14%) of 217 patients had PD-L1 expression data from liver biopsy, and no relationship between PD-L1 expression and activity has been shown. ...camrelizumab, reproducing the results of other checkpoint inhibitors, might be a good option, but only in a small and unidentified fraction of the hepatocellular carcinoma population. On the basis of their study results, the authors have launched two phase 3 trials: one in June, 2019 comparing sorafenib with camrelizumab plus apatinib (from the same company), and a second involving FOLFOX (leucovorin, fluorouracil, and oxaliplatin; approved in China for the treatment of locally advanced and metastatic liver cancers when surgery and local treatments are unsuitable, and frequently given owing to the lower cost compared with sorafenib, but not considered the standard of care in the rest of the world) versus camrelizumab plus FOLFOX.