Background: Sorafenib (S) is the current standard therapy for advanced HCC but validated biomarkers predicting its clinical outcome are lacking. The present study aimed to identify biomarkers predicting prognosis and/or response to S ± erlotinib (E) in HCC patients from the randomized phase III SEARCH trial. Methods: A total of 720 patients were randomized to receive oral S 400 mg bid plus either E 150 mg qd or placebo (P). VEGF-A, VEGF-C, PDGF-BB, KIT (extracellular domain), HGF, bFGF, IGF-2, amphiregulin, betacellulin, EGF, epigen, epiregulin, heregulin, hbEGF, and TGF-a were measured in baseline plasma samples. Mutations in 19 oncogenes were analyzed in archival biopsies (Sequenom OncoCarta). Results: Baseline plasma biomarker data were available for 494 (69%) patients; n=243 S/E and 251 S/P. Treatment-independent analyses (combining both treatment arms) indicated that elevated HGF was associated with poor overall survival (OS; HR=0.598 [low vs high expression], multiplicity adjusted (adj) p=0.0007). Elevated plasma VEGF-A levels and low levels of KIT showed a similar trend towards poor survival (VEGF-A: HR=0.722 [low vs high expression], p=0.03, adj-p=0.39; KIT: HR=0.713 [high vs low expression], p=0.05, adj-p=0.60). High levels of plasma VEGF-C correlated with longer time to tumor progression (HR=0.626 [high vs low expression], adj-p=0.0042). Finally, in 67% of evaluable patients (339/494), a multi-marker composite signature consisting of HGF, VEGF-A, KIT, epigen, and VEGF-C correlated with improved OS: median OS 349 vs 184 days, HR=0.505, p=0.00002. None of the 15 baseline plasma biomarkers predicted efficacy from E in biomarker-treatment interaction analyses. Oncogenic mutations were detected in only 2 patient samples. Conclusions: HGF, VEGF-A, KIT, and VEGF-C baseline plasma levels were associated with clinical outcomes in HCC patients treated with S ± E, and these biomarkers plus epigen constituted a multi-marker composite signature for improved OS. Because S was used in both arms, whether these biomarkers are prognostic or also predictive of benefit from S alone in this population remains to be determined. Clinical trial information: NCT0901901.