This paper proposes to summarize and analyze the daily patient setup shifts based on megavoltage computed tomography (MVCT) image registration results for Helical TomoTherapy® (HT) treatment. One hundred and fifty-five consecutive treatment plans for a total of 137 patients delivered by the HT unit through one year were collected in this study. The patient data included pelvis (26%), abdomen (23%), lung (21%), head and neck (10%), prostate (8%), and others (12%). All the translational and roll rotational shifts made via auto MVCT and kilovoltage computed tomography (kVCT) image registration were recorded at each fraction. Manual fine-tuning was followed if automatic registration result was not satisfactory. The mean shift ± one standard deviation (1 SD) was calculated for each patient based on the entire treatment course. For each treatment site, the average shift was analyzed as well as displacement in 3D vector. Statistical tests were performed to analyze the relationship of patient-specific, tumor site-specific, and fraction number association with the patient setup shifts. For all the treatment sites, the largest average shift was found in the anterior-posterior direction. The population standard deviations were between 1.2 and 5.6 mm for the X, Y, and Z directions and ranged from 0.2 to 0.6 degrees for the roll rotational correction. The largest standard deviations of the setup reproducibility in X, Y, and Z directions were found in lung patients (4.2 mm), abdomen, lung and spine patients (4.4 mm), and prostate patients (5.6 mm), respectively. The maximum 3D displacement was 10.9 mm for prostate patients' setup. ANOVA tests demonstrated the setup shifts were statistically different between patients even for those that were treated at the same tumor site in the translational directions. No strong correlation between the setup and the fraction number was found. In conclusion, the MVCT guided function in the HT treatment enables us to generate relatively accurate daily setup through registration with KVCT data sets. Our results indicate that lung, prostate, and abdominal patients are more prone to setup uncertainty and should be carefully evaluated.