Dose–volume modeling of salivary function in patients with head-and-neck cancer receiving radiotherapy

Abstract

We investigated the factors that affect salivary function after head-and-neck radiotherapy (RT), including parotid gland dose-volume effects, potential compensation by less-irradiated gland tissue, and functional recovery over time. Sixty-five patients with head-and-neck tumors were enrolled in a prospective salivary function study. RT was delivered using intensity-modulated RT (n = 45), forward-planning three-dimensional conformal RT (n = 14), or three-dimensional conformal RT with an intensity-modulated RT boost (n = 6). Whole salivary flow was measured before therapy and at 6 months (n = 61) and 12 months (n = 31) after RT. A wide variety of dose-volume models to predict post-RT salivary function were tested. Xerostomia was defined according to the subjective, objective, management, analytic (SOMA) criteria as occurring when posttreatment salivary function was < 25% of the pretreatment function. Multivariate logistic regression analysis was used to assess the combined effect of dose-volume, patient-, and treatment-related factors. A significant correlation was observed between the relative quality-of-life scores and relative stimulated saliva values at 6 months after RT (Spearman's correlation coefficient [R(s)] = 0.46, p < 0.001). The dose-volume factors were by far the strongest correlates with stimulated saliva flow, although other factors showed modest significance in multimetric models (chemotherapy, gender, and Karnofsky performance status). Several fitted dose-volume models provided a good mathematical description of the data. Significant noise in the salivary measurements (repeated measurement coefficient of variation was 27% in normal subjects) precluded selection of any one of the models presented solely on the basis of the objective fit criteria. Nevertheless, the mean dose-exponential model, in which each parotid gland's relative salivary gland function equaled exp(-A x mean gland dose), with A equal to 0.054/Gy (68% confidence interval 0.052-0.059), provided a good representation of the data and was incorporated into our multimetric analysis. Using that model, we estimated that a mean parotid dose of 25.8 Gy, on average, was likely to reduce a single parotid gland's flow to 25% of its pretreatment value, regardless of the treatment delivery method. Significant correlations were observed between a logistic multivariate model (incorporating the mean dose-exponential equation, gender, and Karnofsky performance status) and stimulated saliva flow at 6 months (R(s) = 0.73), stimulated saliva flow at 12 months (R(s) = 0.54), and quality-of-life score at 6 months (R(s) = 0.35) after RT. Stimulated parotid salivary gland dose-volume models strongly correlated with both stimulated salivary function and quality-of-life scores at 6 months after RT. The mean stimulated saliva flow rates improved from 6 to 12 months after RT. Salivary function, in each gland, appeared to be lost exponentially at a rate of approximately 5%/1 Gy of mean dose. Additional research is necessary to distinguish among the models for use in treatment planning. The incidence of xerostomia was significantly decreased when the mean dose of at least one parotid gland was kept to < 25.8 Gy with conventional fractionation. However, even lower mean doses imply increased late salivary function.