Background: GAN and conatumumab (CON) are investigational, fully human, monoclonal antibodies against type 1 insulin-like growth factor receptor (IGF1R) and death receptor 5, respectively. A 3-arm, P-controlled, phase 2 study randomized 125 mPC pts 1:1:1 to GAN + G, CON + G, or P + G. OS was improved (HR 0.67; 95%CI 0.41 to 1.12) in the GAN +G arm (N=42) vs P+ G arm (N=42) (Kindler et al, JCO 2010:28 abstr 4035). We hypothesized that high BL levels of the IGF1R ligands, IGF-1 and IGF-2; high BL levels of IGFBP-3 (stabilizes IGF-1 and -2); or low BL levels of IGFBP-1, -2, and -4 (sequester IGF-1 and -2) may predict improved OS with GAN treatment. Methods: BL serum or plasma was available from 30 to 33 pts in the GAN + G arm and 26 to 33 pts in the P + G arm. IGF-1 (free [f] and total [t]), tIGF-2 and IGFBP-1, -2, -3, and -4 were analyzed by immunoassay. Treatment effects on OS in high and low BL biomarker subgroups (defined by median BL levels) were evaluated by Cox proportional hazard models, and the interactions between treatment and BL biomarker subgroups were tested. Similar analyses were also conducted treating biomarkers as a continuous variable. Results: GAN +G was associated with significantly improved OS in the following BL biomarker groups: low IGFBP-2 (HR=0.19, 95% CI: 0.06 to 0.55; p=0.002); high tIGF-1 (HR=0.25, 95% CI: 0.09 to 0.67; p=0.006); high IGFBP-3 (HR = 0.28, 95% CI: 0.11 to 0.73; p=0.009); and high IGF-2 (HR=0.24, 95% CI: 0.09 to 0.68, p=0.007). p-values for testing differences in treatment effects among BL biomarker subgroups were 0.379 for IGFBP-2, 0.272 for tIGF-1, 0.11 for IGFBP-3 and 0.051 for IGF-2 (study not powered to detect these differences). Results were consistent when similar analyses for testing interactions were done treating BL biomarker levels as a continuous variable. Conclusions: In support of our hypothesis, low BL IGFBP-2 and high BL tIGF-1, IGF-2, or IGFBP-3 appeared to be associated with improved OS among patients treated with GAN + G. These results are informative for attempting to predict a biomarker-defined population enriched for GAN treatment effects and will be further tested in the ongoing GAN P3 trial.