Artificial intelligence based on serum biomarkers predicts the efficacy of lenvatinib for unresectable hepatocellular carcinoma

Lenvatinib has been effective not only as a first-line but also as a later-line systemic therapy for unresectable hepatocellular carcinoma (uHCC) in real-world clinical practice. How to predict the efficacy of lenvatinib and guide appropriate therapy selection in patients with uHCC have become important issues. This study aimed to investigate the impact of serum biomarkers on the treatment outcomes of patients with uHCC treated with lenvatinib in a real-world setting using an artificial intelligence algorithm. We measured serum biomarkers, including alpha-fetoprotein (AFP), albumin-bilirubin (ALBI) grade, and circulating angiogenic factors (CAFs [i.e., vascular endothelial growth factor, angiopoietin-2, fibroblast growth factor-19 [FGF19], and FGF21]) and analyzed treatment outcomes, including objective response rate (ORR), progression-free survival (PFS), and overall survival (OS) in patients with uHCC treated with lenvatinib. The results of this study demonstrated that an AFP reduction ≥ 40% from baseline within 8 weeks after lenvatinib induction was associated with a higher ORR. With baseline biomarkers using a decision tree-based model, we identified patients with high, intermediate, and low ORRs (84.6%, 21.7% and 0%, respectively; odds ratio, 53.04, P < 0.001, high versus intermediate/low groups). Based on the decision tree-based survival predictive model, baseline AFP was the most important factor for OS, followed by ALBI grade and FGF21.