Voxel-based Analysis Identifies New Functional Subregions Predictive of Xerostomia After Head and Neck Cancer Radiation Therapy.

PURPOSE: Xerostomia is a common and debilitating toxicity following head and neck cancer radiation therapy (RT). Although adaptive RT (ART) strategies can reduce mean parotid gland (PG) dose, clinical trials have shown limited benefit in reducing xerostomia, suggesting global dose metrics may be insufficient. Voxel-based analysis (VBA) enables assessing local dose-toxicity relationships and may identify symptom-related subregions (SRS) more predictive of toxicity. This study aimed to identify xerostomia-associated SRS using VBA and assess their relevance using a functional salivary atlas.

METHODS AND MATERIALS: Sixty patients with locally advanced oropharyngeal cancer treated with chemoradiation therapy were randomized to receive standard RT or weekly ART. Xerostomia was assessed at 12 months posttreatment using stimulated salivary flow, with values ≤500 mg/min indicating xerostomia. A customized VBA was applied to identify subregions from voxelwise dose differences between patients with and without xerostomia using a permutation test (planned dose in the standard arm and accumulated dose in the ART arm). To assess clinical relevance, a functional salivary atlas was generated from prostate-specific membrane antigen-positron emission tomography/computed tomography (PSMA-PET/CT) scans of 13 patients with prostate cancer and compared with the subregions. A dosimetric analysis evaluated the predictive value of dose in overlapping regions.

RESULTS

VBA identified a subregion (300.2 cc) with significantly higher doses in patients with xerostomia, located contralaterally near the nasopharynx. Overlap with the functional PSMA-PET atlas revealed 2 clinically relevant subregions in the contralateral PG (SRSPG, 6.8 cc) and tubarial gland (SRStubarial, 12.7 cc), with doses >11.5 Gy higher in patients with xerostomia. Logistic regression models yielded high predictive performance, with maximum areas under the curve of 0.88 for SRSPG and 0.88 for SRStubarial.

CONCLUSIONS

VBA revealed 2 predictive subregions-within the contralateral PG and tubarial glands-strongly associated with xerostomia at 12 months post-RT. Our findings offer a plausible explanation for the persistence of xerostomia despite PG sparing and support incorporating subregional dose metrics and tubarial gland sparing into RT planning.