We aimed to determine the impact of air inflow into vacuum-type immobilization devices (VIDs) on setup errors. We assigned 70 patients undergoing radiotherapy for head and neck cancer to groups V (n = 34) or N (n = 36) according to whether the VIDs were deflated weekly or not deflated during treatment, respectively. We calculated systematic errors (Σ) as the standard deviations (SDs) of mean errors, and random errors (σ) as the root mean square of SDs in each patient. We compared overall means (μ), SDs (SDoverall), random errors and systematic errors. We also measured temporary pressure changes in VIDs to determine the influence of pressure changes in VIDs on setup errors. The μ was within 0.20mm and 0.2° in both groups, whereas SDoverall significantly differed between them. The SDoverall differed the most in the Roll axes of groups N (0. 87°) and V (0.58°). The Σ and σ values were lower in all axes of group V than in group N. Despite the initial deflation target of -70kPa, the pressure in VIDs reached -5kPa at the end of treatment. However, weekly deflation apparently maintained pressure at -20kPa. Effective pressure control in VIDs can reduce patient-by-patient variation and improve setup reproducibility for individual patients, consequently resulting in small variations among overall setup errors.