Background: Elevated baseline HGF levels were associated with shorter OS in pts treated with BEV+INF. We evaluated on-treatment HGF levels to describe treatment-related changes and associations with outcome. Methods: We analyzed baseline EDTA plasma samples from 310 pts (148 INF; 162 BEV+INF) using an optimized multiplex ELISA platform for HGF at baseline and after 4-weeks (wks) on treatment. Primary endpoint of this analysis was OS. The Kaplan-Meier estimated the OS distribution and the proportional hazards model tested the prognostic importance of change at 4-wks from baseline in HGF levels in predicting OS, adjusting for treatment arm, bone metastases and stratification variables. Results: The median baseline HGF level in 310 pts was 161.4 pg/ml. Elevated HGF at 4-wks (>median) was associated with a worse OS (median OS = 14 vs 27 months; adjusted hazard ratio (HR)= 1.75, p< 0.0001). Only 9/155 pts (5.8%) with baseline HGF levels ≤ median developed elevated HGF (>median) at 4-wks; 66/155 pts (43%) with baseline HGF levels >median lowered HGF (<median) at 4-wks from baseline. Compared to pts with persistently elevated HGF levels, a decline in HGF levels at 4-wks (< median) was associated with improved OS (19 vs 13 months, adjusted HR=1.41, p=0.043). Conclusions: In RCC pts with low baseline HGF levels (< median), levels remain consistently low and are associated with improved OS. Conversely, in pts with high baseline HGF levels results are split; some patients continue to have high levels on treatment and are associated with a worse OS, suggesting that, HGF predicts for therapeutic benefit and represents a potential mechanism of resistance. Support: U10CA180821, U10CA180882. Clinical trial information: NCT00072046.