BackgroundImmobilization devices are essential for maintaining accurate and repeatable patient positioning in radiotherapy. This study aimed to evaluate the setup errors and dosimetric deviation induced by the deformation of immobilization devices in thoracic cancer radiotherapy using CT‐linac.Materials and methodsA retrospective analysis was conducted on 40 thoracic cancer patients who underwent radiotherapy, using vacuum cushion (VC) and thermoplastic mask (TM) for immobilization. A total of 206 weekly Fan‐beam CT (FBCT) images (4–7 per patient) were analyzed to manually delineate immobilization devices and assess their geometric deformations against setup errors. Dosimetric deviations between the clinical plan (CT‐plan) and the delivered plan (FBCT‐plan) were compared for planning target volume (PTV) and organs at risk (OARs). Correlations between dose variations and setup errors were analyzed in lateral (LAT), longitudinal (LNG), and vertical (VRT) axes.ResultsThe conformity of the VC (Nup) and TM (Ndown) with the patient in simulation CT exhibited moderate to strong correlations with VRT setup errors (Ndown: r = −0.484, p < 0.01;Nup: r = −0.697, p < 0.01). However, intra‐fraction deformation of immobilization devices (in FBCT) showed no significant correlation with setup errors. In the dosimetric analysis of OARs, lung dose parameters (Dmean, V5, V20) and heart Dmean exhibited a consistent absolute difference with increasing setup errors. Dose variation decreased significantly when errors exceeded 5 mm, particularly in the VRT direction for most PTV indices, with the exception of CI and HI. Spinal cord Dmax variation correlated significantly with setup accuracy along the LNG axis, but not along other axes.ConclusionThe conformity of immobilization devices in simulation CT exhibits a stronger correlation with setup accuracy than the deformation of these devices in intra‐fraction FBCT. FBCT is recommended for improving treatment precision through dosimetric assessment and planning adjustments.