Background: We aimed to identify tissue microRNAs (miRs) that could predict outcome for patients with metastatic colorectal cancer (mCRC) treated with first line bevacizumab (BEV) and chemotherapy (CT) but not for patients treated with CT alone. Methods: Patients with mCRC treated with first line capecitabine and oxaliplatin (CT) with or without BEV at ten hospitals were identified and data was extracted retrospectively. Formalin-fixed paraffin-embedded tissue samples from primary tumors were collected and RNA was purified from 3 x 10 µm sections, without micro-dissection. miR expression was measured using Applied Biosystems TaqMan Custom LDA cards profiling 22 selected miRs in duplicate. The 22 miRs were selected from a previous discovery study profiling 754 miRs. miR expression was related to time to disease progression (TTP) and overall survival (OS) in multivariate analyses using Cox proportional hazards models with adjustment for age, prior adjuvant treatment, and no. of metastatic sites. We have previously found that patients with primary tumors originating in the sigmoid colon and rectum (S+R) experienced a better outcome than patients with other primary tumor locations (caecum to descending colon) when treated with BEV, so our analyses were stratified by primary tumor location. Results: miR expression was measured in samples from 399 patients: 155 samples from the original CT+BEV discovery study, 119 samples from a new CT+BEV cohort, and 125 samples from a CT alone cohort. Expression of miR-664 showed a significant positive association with increasing TTP, OS, and response rate (RR), but only in the cohort of patients with sigmoid colon- and rectal primary tumors treated with CT+BEV (n=183). Conclusions: We have identified a subgroup of patients with mCRC that are likely to benefit from BEV addition to first line CT using the combined information of location of the primary tumor and expression level of miR-664.