This study was designed to evaluate prospectively magnetic resonance imaging for the prediction of the circumferential resection margin in rectal cancer to identify in which patient magnetic resonance imaging could accurately assess the circumferential resection margin before surgery and in which patients it could not. During a 17-month period, a preoperative magnetic resonance imaging for the assessment of circumferential resection margin was obtained prospectively in 38 patients with mid or low rectal cancer. The agreement of magnetic resonance imaging and pathologic examination for assessment of circumferential resection margin was analyzed. Overall, magnetic resonance imaging agreed with histologic examination of the circumferential resection margin assessment in 28 patients (73 percent; kappa = 0.47). In all cases of disagreement between magnetic resonance imaging and pathology, magnetic resonance imaging overestimated the circumferential resection margin involvement. For the 11 patients with mid rectal cancer, circumferential resection margin was well predicted by magnetic resonance imaging in all cases (kappa = 1). For 27 patients with low rectal tumor, overall agreement between magnetic resonance imaging and histologic assessment was 63 percent (kappa = 0.35). Agreement was 22 percent (kappa = 0.03) for the 9 patients with low anterior and 83 percent (kappa = 0.67) for the 18 patients with low posterior rectal tumor. Univariate analysis revealed that only low and anterior rectal tumor was risk factor of overestimation of the circumferential resection margin by magnetic resonance imaging. Although magnetic resonance imaging remains the best imaging tool for the preoperative assessment of the circumferential resection margin in patients with rectal cancer, it can overestimate the circumferential resection margin involvement in low and anterior tumor with the risk of overtreating the patients.