Background: This study evaluated the feasibility of combining XELOX with erlotinib in a 2 different schedules, and the prognostic significance of serum AMR and TGFa in previously untreated mCRC. Methods: Eligible patients (Pt) were randomized to the ‘continuous’ arm (Arm CON) [ERL 100mg daily D1-21, oxaliplatin (Ox) 130mg/m² D1, capecitabine (Xe) 825mg/m² bd D1-14, q3w], or the ‘intermittent’ arm (Arm INT) [ERL 150mg daily at alternate day on D 1-14, then 150mg daily on D15-21, Ox 130mg/m² on D1, Xe 750mg/m²bd on D1-14, in a q3w]. Serum levels of AMR, TGFa and CEA were determined serially. KRAS mutation was determined from archived tumors. Results: 60 pts were randomized and there was no difference in the baseline characteristics between the 2 arms. Of the 58 pts evaluated for response, the overall response rates (ORR) were 56.3% (Arm CON) and 66.3% (Arm INT). At a median follow-up of 2.8 yrs, the median overall survival (OS) of pts in Arm CON and Arm INT were 9 m (95% CI: 7.2-18m) and 10.3m (95% CI: 7.2-15.6 m) respectively. KRAS mutation (n = 44 tumors) status did not predict ORR or survival. Gr 3-4 toxicities occurred in 53.3% (Arm CON) and 46.7% (Arm INT) of pts. Multivariate analysis showed that elevated baseline serum TGFa and AMR were independent predictors of inferior PFS and OS, respectively. Post-treatment drop in serum TGFa was associated with shorter OS and PFS, while a drop in serum TGFa and a rise in serum AMR were associated with inferior PFS and OS, respectively. Conclusions: Although there was no statistical difference in the toxicity and efficacy of the 2 arms, the ORR from XELOX and intermittent ERL compared favorably with historic control. KRAS mutation status was not predictive. Serum TGFa and AMR have prognostic significance in this cohort and further prospective studies are warranted. Clinical trial information: NCT01243047.