Novel approaches to improve the therapeutic index of head and neck radiotherapy: an analysis of data from the PARSPORT randomised phase III trial

Abstract

Purpose: Subjective xerostomia is a common side-effect following radiotherapy for the treatment of head-and-neck cancer. Standard mean dose models previously used to model xerostomia only that partially predict the occurrence of xerostomia. Studies in animal models have suggested that there are regional variations in the radiosensitivity of the parotid glands. In this work we tested the hypothesis that this is also true for the human parotid gland.

Methods: We present novel dose-response models explicitly taking the spatial distribution of the radiation dose into account. We considered dose to the submandibular gland and other clinical factors and used a variable-selection algorithm to select the best dose-response model. This methodology was applied to 63 head and neck cancer patients and validated using two independent patient cohorts of 19 and 29 patients, respectively.

Results: The predictive accuracy of dose-response models improved significantly when including regional variations of radiosensitivity of the parotid glands compared to standard mean-dose models (p = 0.001, t-test). Beneficial dose-pattern analysis demonstrated the importance of minimising dose to the lateral and cranial component of the human parotid gland in order to avoid xerostomia. Furthermore we found an evidence that surgical removal of the sub-mandibular gland significantly increases the risk of radiation-induced xerostomia.

Conclusion: Dose-response models which take the shape of the dose-distribution into account predicted xerostomia significantly better than standard mean-dose models. Our novel model could be used to rank potential treatment plans more reliably according to their therapeutic index and may be useful to generate better treatment plans.