Background: Standardized conditions to distinguish subpopulations of colorectal cancer (CRC) patients more and less sensitive to cetuximab therapy remain undefined. Materials and methods: We retrospectively analyzed epidermal growth factor receptor (EGFR) copy number by fluorescence in situ hybridization (FISH) in paraffin-embedded tumor blocks from 85 chemorefractory CRC patients treated with cetuximab. Results were analyzed according to different score systems previously reported in colorectal and lung cancers. The primary end point of the study was identification of the EGFR FISH score that best associates with response rate (RR). Results: Using receiver operating characteristic (ROC) analysis, the cut-off that best discriminated responders versus nonresponders to cetuximab was a mean of 2.92 EGFR gene copies per cell. This model showed sensitivity of 58.6% 95% confidence interval (CI) = 47.1–70.1) and specificity of 93.3% (95% CI  = 80.6–100). EGFR FISH-positive patients (N = 43, 50.6%) had significantly higher RR (P = 0.0001) and significantly longer time to disease progression (P = 0.02) than EGFR FISH negative (N = 42, 49.4%). Other scoring systems resulted less accurate in discriminating patients with the highest likelihood of response to cetuximab therapy. Conclusions: CRC patients with high EGFR gene copy number have an increased likelihood to respond to cetuximab therapy. Prospective clinical trials with a careful standardization of assay conditions and pattern interpretation are urgently needed.