The published data on the predictive and prognostic value of KRAS mutations in metastatic colorectal cancer (mCRC) treated with cetuximab seemed inconclusive. To derive a more precise estimation of the relationship, a meta-analysis was performed. Systematic computerised searches of the PubMed, EMBase, BIOSIS, and SCOPUS were performed. A total of 22 studies were identified. Random-effects model or fix-effects model was used according to between-study heterogeneity. A total of 2188 mCRC patients were included in the final meta-analysis. The rate of KRAS mutations was 38% (829/2188). The overall response rate (ORR) of mutant KRAS patients was 14% (119/829), whereas the ORR of wild-type KRAS patients was 39% (529/1359). The overall pooled relative ratio (RR) for ORR was 0.24 (95% confidence intervals (CI): 0.16–0.38; P<0.01) when mutant KRAS patients were compared with wild-type KRAS patients. Median PFS was significantly shorter in mutant KRAS patients compared with that in wild-type KRAS patients (3.0 versus 5.8 months; HR=1.94; 95% CI: 1.62–2.33; P<0.01). Similarly, median OS was significantly shorter in mutant KRAS patients compared with that in wild-type KRAS patients (6.9 versus 13.5 months; HR=2.17; 95% CI: 1.72–2.74; P<0.01). The meta-analysis strongly suggests that KRAS mutations represent adverse predictive and prognostic biomarkers for tumour response and survival in mCRC patients treated with cetuximab. Patients with tumours that harbour mutant-type KRAS are more likely to have a worse response, PFS, and OS when treated with cetuximab.