The purpose of this study was to quantify the systematic and random errors for various disease sites when daily MVCT scans are acquired, and to analyze alterna- tive off-line verification protocols (OVP) with respect to the patient setup accuracy achieved. Alignment data from 389 patients (9,418 fractions) treated at ten differ- ent anatomic sites with daily image-guidance (IG) on a helical tomotherapy unit were analyzed. Moreover, six OVP were retrospectively evaluated. For each OVP, the frequency of the residual setup errors and additional margins required were calculated for the treatment sessions without image guidance. The magnitude of the three-dimensional vector displacement and its frequency were evaluated for all OVP. From daily IG, the main global systematic error was in the vertical direction (4.4-9.4 mm), and all rotations were negligible (less than 0.5°) for all anatomic sites. The lowest systematic and random errors were found for H&N and brain patients. All OVP were effective in reducing the mean systematic error to less than 1 mm and 0.2° in all directions and roll corrections for almost all treatment sites. The treatment margins needed to adapt the residual errors should be increased by 2-5 mm for brain and H&N, around 8 mm in the vertical direction for the other anatomic sites, and up to 19 mm in the longitudinal direction for abdomen patients. Almost 70% of the sessions presented a setup error of 3 mm for OVPs with an imaging frequency above 50%. Only for brain patients it would be feasible to apply an OVP because the residual setup error could be compensated for with a slight margin increase. However, daily imaging should be used for anatomic sites of difficult immobilization and/or large interfraction movement.