Abstract
Hepatocellular carcinoma (HCC) is a major cause of cancer-related mortality worldwide. Around half of patients with HCC will receive systemic therapies during their life span. The pivotal positive sorafenib trial and regulatory approval in 2007 was followed by a decade of negative studies with drugs leading to marginal antitumoral efficacy, toxicity, or trials with a lack of enrichment strategies. This trend has changed over the last 2 years with several compounds, such as lenvatinib (in first-line) and regorafenib, cabozantinib, ramucirumab and nivolumab (in second-line), showing clinical benefit. These successes came at a cost of increasing the complexity of decision-making, and ultimately, impacting the design of future clinical trials. Nowadays, life expectancy with single active agents has surpassed the threshold of 1 year and sequential strategies have provided encouraging outcomes. Overall survival (OS) remains the main endpoint in phase III investigations, but as in other solid tumours, there is a clear need to define surrogate endpoints that both reliably recapitulate survival benefits and can be assessed before additional efficacious drugs are administered. A thorough analysis of 21 phase III trials published in advanced HCC demonstrated a moderate correlation between progression-free survival (PFS) or time to progression (TTP) and OS (R = 0.84 and R = 0.83, respectively). Nonetheless, the significant differences in PFS identified in 7 phase III studies only correlated with differences in OS in 3 cases. In these cases, the hazard ratio (HR) for PFS was ≤0.6. Thus, this threshold is herein proposed as a potential surrogate endpoint of OS in advanced HCC. Conversely, PFS with an HR between 0.6-0.7, despite significance, was not associated with better survival, and thus these magnitudes are considered uncertain surrogates. In the current review, we discuss the reasons for positive or negative phase III trials in advanced HCC, and the strengths and limitations of surrogate endpoints (PFS, TTP and objective response rate [ORR]) to predict survival.