Background: REACH was a global, multicenter, randomized, double-blind, phase 3 study evaluating the efficacy and safety of RAM as a single agent for the treatment of pts with advanced HCC after prior sorafenib therapy. The primary outcome for REACH was presented at ESMO 2014. The overall survival (OS) HR for the ITT population (RAM 283; placebo [PBO] 282) was 0.866 (95% CI 0.717, 1.046; p=0.1391); median OS was 9.2m for RAM vs 7.6m for PBO. The pre-specified subgroup analysis of baseline AFP (cutoff 400 ng/mL) suggested AFP is a predictive marker for RAM survival benefit. Methods: Pre-specified subgroup analysis was performed based on baseline AFP with a cutoff of 400 ng/mL. Additional analyses were conducted using stratified/unstratified cox regression models and corresponding log rank test to evaluate the relationship between baseline AFP and RAM treatment effect. Results: In 250 pts with baseline AFP ≥400 ng/mL (RAM 119; PBO 131), OS HR was 0.67 (95% CI 0.51–0.90; p=0.0059). Median OS was 7.8m for RAM vs 4.2m for PBO. In 417 pts with a baseline AFP ≥1.5 × upper limit of normal (ULN; RAM 205; PBO 212), mOS was 8.6m for RAM vs 5.7m for PBO and the HR was 0.749 (95% CI: 0.603, 0.930) (p=0.0088). The interaction testing of baseline AFP and RAM treatment effect on OS using both cutoffs (400 ng/mL and 1.5 xULN) are significant (p-value = 0.0272 and 0.0372, respectively). The safety profile in these pt populations was similar to that observed in the overall safety population. Additional REACH analyses demonstrating the predictive value of baseline AFP for RAM treatment effect on OS will be presented. Conclusions: A clinically meaningful improvement in OS was observed in populations with a baseline AFP ≥400 ng/mL or ≥1.5 × ULN. Additional analyses demonstrated a consistent RAM OS benefit for the pt population with baseline AFP over a wide range of values above the normal range. Baseline AFP is a likely predictive marker for RAM OS benefit. Clinical trial information: NCT01140347