In this prospective study we investigated the absolute accuracy of the conventional simulation in head and skull base tumors. 41 isocenters in 40 consecutive patients with tumors of the head and skull base were included. In all cases a rigid stereotactic mask system was used for non-invasive fixation. The stereotactic (?calculated?) coordinates of the isocenter were defined by the treatment planning computer. Each patient underwent a physical simulation using exclusively anatomical reference points to define the ?preliminary? isocenter. The displacement between its coordinates and those of the stereotactic target point was recorded in X-, Y- and Z-direction with help of the targeting device, and the spatial error was calculated. Additionally, the patients were stratified by basal or calvarial tumor site to estimate the importance of the basal bone structures in the simulation accuracy. The influence of the learning effect on simulation accuracy was also determined.

The results showed an accuracy of set-up at the linac within 1 mm in all three directions as calculated from orthogonal portal films. Mean shift of the isocenter coordinates obtained from physical simulation compared to the calculated stereotactic coordinates was 2.15 mm, 2.54 mm, and 2.69 mm for X-, Y-, and Z-direction, respectively. Mean spatial displacement amounted 5.06 mm, and the median was 4.50 mm. No significant difference could be noted between basal and calvarial location of the isocenter. A significant ?learning effect? was observed with a decrease in spatial shift with increasing patient numbers. This effect was stronger in basal lesions, whereas calvarial lesions showed only a minor, insignificant effect. In conclusion, a physical simulation requires a safety margin of 5 mm in PTV definition in addition to other factors, e.g. organ movement.

Key words: Stereotactic irradiation, Simulation, Accuracy, Quality assurance.