[en] A radiotherapy treatment plan optimisation algorithm has been applied to 48 prostate plans and the results compared with those of an experienced human planner. Twelve patients were used in the study, and a 3, 4, 6 and 8 field plan (with standard coplanar beam angles for each plan type) were optimised by both the human planner and the optimisation algorithm. The human planner 'optimised' the plan by conventional forward planning techniques. The optimisation algorithm was based on fast-simulated-annealing. 'Importance factors' assigned to different regions of the patient provide a method for controlling the algorithm, and it was found that the same values gave good results for almost all plans. The plans were compared on the basis of dose statistics and normal-tissue-complication-probability (NTCP) and tumour-control-probability (TCP). The results show that the optimisation algorithm yielded results that were at least as good as the human planner for all plan types, and on the whole slightly better. A study of the beam-weights chosen by the optimisation algorithm and the planner will be presented. The optimisation algorithm showed greater variation, in response to individual patient geometry. For simple (e.g. 3 field) plans it was found to consistently achieve slightly higher TCP and lower NTCP values. For more complicated (e.g. 8 fields) plans the optimisation also achieved slightly better results with generally less numbers of beams. The optimisation time was always ≤5 minutes; a factor of up to 20 times faster than the human planner