Objective To compare the errors of final isocenter marking method and reference pointmarking method for CT simulation positioning in intensity-modulated radiotherapy (IMRT). MethodsFrom 2009 to 2012, 327 patients with head and neck cancer for IMRT underwent CT simulation positioning using the Philips Brilliance CT Big Bore scanner and Philips Tumor LOC workstation and were divided into final isocenter marking group (n=208) and reference point marking group (n=119) according to positioning methods. Target volume delineation and treatment plan design were performed on the Varian Eclipse treatment planning system (TPS). Before treatment, kilovoltage cone-beam CT scans and registration were performed with the Varian EX on-board imager system to obtain beam position errors in the right-left (RL), superior-inferior (SI), and anterior-posterior (AP) directions, and then comparisons of errors between the two groups were made by independent-samples t test. Finally, the TPS was used to measure the changes in the doses to the organs at risk after moving isocenters in the RL, SI, and AP directions among 5 patients with nasopharyngeal carcinoma. Results The mean beam position errors in the three directions were less in the final isocenter marking group than in the reference point marking group (P=0.02,0.01,0.03). After moving isocenters in the three directions, the target dose was reduced and the dose to the normal tissue around the target tumor was increased significantly. The error in the AP direction had the maximum influence on the spinal cord and brainstem. Conclusions Final isocenter marking method leads to less beam position error than reference point marking method in CT simulation positioning. Small isocenter motion can cause large changes in the doses to the organs at risk.