We have studied and analysed the magnitude of interfraction set-up errors and gold seed marker and prostate displacement in 118 patients using three gold seeds implanted within the prostate. Set-up errors and gold seed marker displacements were determined from bony anatomy and gold seed marker mismatch between the electronic portal image and the simulation digitally reconstructed radiograph (DRR), respectively. Prostate displacement relative to bony anatomy was determined from the difference between gold seed marker and bony anatomy displacement. Daily online repositioning of patients was accomplished through image matching using Varian Portal-Vision software. A total of 4878 electronic portal images and 236 DRRs from 118 patients were acquired over the course of the study. The means and standard deviations of the systematic error of gold seed marker displacement of 118 patients were 2.1+/-2.7 mm for anteroposterior (AP), -0.5+/-1.7 mm for left-right (L-R), and 1.0+/-1.9 mm for superoinferior (SI) directions; the random errors were 3.2 mm (0.9-4.9 mm) for AP, 1.9 mm (0.7-5.3 mm) for L-R, and 2.1 mm (0.7-4.5 mm) for SI directions. The mean and standard deviation of the isocentre set-up systematic error of 20 patients was 1.2+/-2.2 mm for AP, -0.1+/-1.4 mm for L-R, and -0.8+/-2.6 mm for SI directions. The isocentre set-up random errors were 1.6 mm (1.2-4.8 mm) for AP, 1.3 mm (0.6-2.5 mm) for L-R and 1.3 mm (1.0-2.6 mm) for SI directions. The mean and standard deviation of the prostate displacement systematic error relative to bony anatomy was 0.0+/-1.4 mm for AP, 0.0+/-1.1 mm for L-R and -0.2+/-2.4 mm for SI directions. Prostate displacement random errors were 1.5 mm (1.2-3.3 mm) for AP, 0.9 mm (0.4-1.5 mm) for L-R and 1.4 mm (1.2-2.4 mm) for SI directions.