Dose response of the parotid and submandibular salivary glands was determined for 25 head and neck cancer patients treated by IMRT. Individual salivary gland functions were assessed by scintigraphy before and 6 months after radiotherapy. Accuracy requirements were estimated by using the maximal slope of the fitted dose response model and average value of the dose gradients within the glands. In addition, systematic and random set-up errors were estimated for each patient by at least weekly portal imaging. We investigated the changes in the salivary gland mean doses (Dmean) that would have occurred without correction of patient positioning. This was done by shifting the planned isocenter according to the obtained systematic set-up error and by recalculating the dose distribution in treatment planning system (TPS).
The maximal slope and D50 values of the dose response model were −0.041 1/Gy and 30.4 Gy, respectively. The results suggested that spared fraction of individual salivary gland function can be estimated with an accuracy of ±10 % , if actual Dmean of the gland is within ±2.4 Gy with the planned value. On the average, this was achieved with maximal systematic positional 3D shift of 3.0 mm for the parotid glands and 2.7 mm for the submandibular glands. The magnitude of systematic 1D set-up errors was 1.7 ± 1.3 mm (mean ± SD) while that of systematic 3D errors was 3.4 ± 1.6 mm. The SD of random set-up errors was 1.5 mm. The magnitude of Dmean shifts due to set-up errors was 1.5 ± 1.4 Gy. The steepness of dose gradients within the glands was 0.8 ± 0.5 Gy/mm in the most critical direction (toward the glands).
When substantial part of salivary gland function is intended to be spared in head and neck IMRT, narrow dosimetric and positional tolerances should be adopted for the major salivary glands due to steep dose response curve obtained for the glands.