Introduction IMRT has been in clinical use for nearly 20 years, yet the issue of optimal beam directions remains open. Although there exist many beam angle optimization algorithms, none has established itself yet, nor is it clear how to assess the optimality of results. Purpose To investigate how the choice of beam angles in IMRT of the prostate affects the dosimetry of the rectum, bladder and the large bowel. Methods and materials Five patients, previously treated with a 5-beam IMRT plan to 86.4 Gy, were selected for this study. Candidate beams were defined every 20° in a 360° arc around the isocenter. Treatment plans were generated for each set of 5 beams taken out from the set of the candidate beams (total 8568 plans per patient). For each patient, all other constraints and parameters, except the beam angles, were kept fixed as in the clinical plan. All plans were normalized so that rectum Dmax was 99%. The dosimetric effect was assessed using D54, D87.5 and V47 for the rectum; Dmax, D54 and V47 for the bladder, and Dmax for the large bowel. Results All dosimetric indices varied between 2- and 3-fold. Subsequently, the best 5% of plans with respect to each index were analyzed (428 plans). Lateral beams were the most favored for every index. Secondly, anterior beams were favored for the rectum but not for the bladder or the large bowel. Conclusion The choice of beam directions may significantly affect the dosimetry of the rectum, bladder and large bowel.