The purpose of this study was to evaluate the effect of an iterative reconstruction method (IR) on shape reproducibility in three-dimensional (3D) computed tomography images. We used an IR (sinogram-affirmed iterative reconstruction: SAFIRE) implemented in a 64-channel multi slice computed tomography system (Siemens, SOMATOM Definition AS). We scanned a simulated vessel phantom with 180-HU vessel contrast at various radiation doses and reconstructed axial images of filtered back projection (FBP) and SAFIRE. Then we reconstructed multi-planar reconstruction (MPR) and volume rendering (VR) images from the axial images. Roundness was evaluated from MPR images and vessel surface roughness was evaluated from pixel value profiles of a vessel in VR images and visual evaluation by radiological technologists. In comparison on equivalent dose, the roundness and roughness were improved with increase of SAFIRE strength level. However, in comparison on equivalent noise (standard deviation: SD) of axial images, SAFIRE strength levels of 2 to 5 were significantly inferior to FBP (p<0.05). SAFIRE strength of 1 with a dose reduction of 19% maintained the shape reproducibility in all evaluations. Therefore, it would be not appropriate to use the SD of axial image as index for the dose reduction rate determination of 3D-CT images.