The prognostic value of positive circumferential resection margins (CRM) in resected esophageal squamous cell carcinoma (ESCC) is unclear. The Royal College of Pathologists criteria and the College of American Pathologists criteria are the two commonly used definitions of CRM involvement. The aim of this report was to compare the prognostic performance of the two criteria and to propose a modified stratification in patients who underwent radical esophagectomy for ESCC. We retrospectively reviewed 112 patients with pathologically confirmed T3N0M0 ESCC and without neoadjuvant therapy from June 2009 and July 2011. The optimal cutoff point was obtained by the X-tile. The prognostic performance of different classifications of CRM was assessed in terms of homogeneity, discriminatory ability, and monotonicity. According to the Royal College of Pathologists criteria, a positive CRM was detected in 87 patients (77.7%); and 24 patients (21.4%) were found with positive CRM according to the College of American Pathologists criteria. Non-significant associations between overall survival and CRM were observed according to either of the two criteria. The analysis of reclassifying the CRM criteria demonstrated that the optimal cutoff CRM value for best prognostic power was 600 µm. Patients with CRM more than 600 µm showed better overall survival (P<0.05) than the cases with CRM less than 600 µm. Furthermore, the improved homogeneity, discriminatory ability, and monotonicity gradients were also found in this modified criteria, as compared with the two existing criteria. Our study highlighted that CRM was an independent prognostic factor for survival in esophageal cancer patients, and the modified CRM criteria had better prognostic power than the traditional criteria in patients with ESCC.