Cumulative dose-volume histograms (DVH) are crucial in evaluating the quality of radioactive seed prostate implants. When calculating DVHs, the choice of voxel size is a compromise between computational speed (larger voxels) and accuracy (smaller voxels). We quantified the effect of voxel size on the accuracy of DVHs using an in-house computer program. The program was validated by comparison with a hand-calculated DVH for a single 0.4-U iodine-125 model 6711 seed. We used the program to find the voxel size required to obtain accurate DVHs of five iodine-125 prostate implant patients at our institution. One-millimeter cubes were sufficient to obtain DVHs that are accurate within 5% up to 200% of the prescription dose. For the five patient plans, we obtained good agreement with the VariSeed (version 6.7, Varian, USA) treatment planning software's DVH algorithm by using voxels with a sup-inf dimension equal to the spacing between successive transverse seed implant planes (5 mm). The volume that receives at least 200% of the target dose, V200, calculated by VariSeed was 30% to 43% larger than that calculated by our program with small voxels. The single-seed DVH calculated by VariSeed fell below the hand calculation by up to 50% at low doses (30 Gy), and above it by over 50% at high doses (>250 Gy).