Prognostic and predictive markers are needed to predict the clinical outcomes of patients with advanced colorectal cancer (CRC) who receive standard first-line treatments. We performed a prospective cohort study in advanced CRC patients to identify a miRNA signature that could predict the benefit of receiving first-line chemotherapy for these patients. Twenty-one paired tumours and adjacent normal tissues were collected from advanced CRC patients and analysed by miRNA microarrays. Between tumour and normal tissues, 33 miRNAs were differentially expressed and was confirmed by qRT-PCR from another group of 67 patients from a prospective cohort study. A two-miRNA-based signature was obtained using the LASSO Cox regression model based on the association between the expression of each miRNA and the PFS of individual patients. Internal and external validation cohorts, including 40 and 44 patients with advanced CRC, respectively, were performed to prove the prognostic and predictive value of this signature. A signature was built based on two miRNAs, miR-125b-2-3p and miR-933. CRC patients were classified into low- and high-risk groups for disease progression based on this tool. The patients with low risk scores generally had better PFS than those with high risk scores. In the training set, the median PFS in the low- and high-risk groups were 12.00 and 7.40 months, respectively. In the internal validation set, the median PFS in the low- and high-risk groups were 9.90 and 5.10 months, respectively. In the external validation set, the median PFS in the low- and high-risk groups were 9.90 and 6.40 months, respectively. Furthermore, we detected miR-125b-2-3p associated with CRC cell sensitivity to first-line chemotherapy. Our two-miRNA-based signature was a reliable prognostic and predictive tool for tumour progression in patients with advanced CRC, and might be able to predict the benefit of receiving standard first-line chemotherapy in CRC.