RESUMO
BACKGROUND AND PURPOSE: Sampling theory and operator characteristic curves are methods that can determine an optimal schedule for quality control tests. We apply this method to positional data for whole breast radiotherapy since several surveys report inconsistent image guidance practice for this technique. MATERIALS AND METHODS: Positional errors were defined, for 55 consecutive breast cancer patients, by comparing the central lung distance measured on portal images with that obtained from the corresponding digitally reconstructed radiograph. From the distribution of positional errors, the probability of a setup error >5 mm in the direction of the mediastinum was established. Using operator characteristic curves, we compared the effectiveness of various image-guidance schedules in dealing with such errors. We also calculated the dosimetric impact of undetected errors. RESULTS: Setup errors >5 mm towards the mediastinum for this cohort were unlikely, at 2.7%. Imaging half of the fractions protects most patients against three or more undetected errors. Undetected, such an error increases, on average, the maximum dose to 10 cm3 of the heart by 50 cGy, the mean heart dose by 4 cGy, and the left lung V20Gy by 0.2%; therefore, the clinical impact is minute. Given that detected positional errors outside of tolerance are corrected, their residual likelihood decreases with the ratio of fractions being imaged. CONCLUSIONS: For most tangential breast radiotherapy patients, setup errors >5 mm towards the mediastinum are unlikely, and their dosimetric impact is remote. Imaging half of the fractions of a course of whole breast radiotherapy prevents these errors to occur more than twice.