Dose distributions in HN-IMRT are complex and may be sensitive to the treatment uncertainties. The goals of this study were to evaluate: 1) dose differences between plan and actual delivery and implications on margin requirement for HN-IMRT with rigid setup errors; 2) dose distribution complexity on setup error sensitivity; and 3) agreement between average dose and cumulative dose in fractionated radiotherapy. Rigid setup errors for HN-IMRT patients were measured using cone-beam CT (CBCT) for 30 patients and 896 fractions. These were applied to plans for 12HN patients who underwent simultaneous integrated boost (SIB) IMRT treatment. Dose distributions were recalculated at each fraction and summed into cumulative dose. Measured setup errors were scaled by factors of 2-4 to investigate margin adequacy. Two plans, direct machine parameter optimization (DMPO) and fluence only (FO), were available for each patient to represent plans of different complexity. Normalized dosimetric indices, conformity index (CI) and conformation number (CN) were used in the evaluation. It was found that current 5 mm margins are more than adequate to compensate for rigid setup errors, and that standard margin recipes overestimate margins for rigid setup error in SIB HN-IMRT because of differences in acceptance criteria used in margin evaluation. The CTV-to-PTV margins can be effectively reduced to 1.9 mm and 1.5 mm for CTV1 and CTV2. Plans of higher complexity and sharper dose gradients are more sensitive to setup error and require larger margins. The CI and CN are not recommended for cumulative dose evaluation because of inconsistent definition of target volumes used. For fractionated radiotherapy in HN-IMRT, the average fractional dose does not represent the true cumulative dose received by the patient through voxel-by-voxel summation, primarily due to the setup error characteristics, where the random component is larger than systematic and different target regions get underdosed at each fraction.