Fitting NTCP models to SBRT dose and carotid blowout syndrome data

Purpose: To estimate the radiobiological parameters of three popular NTCP models, which describe the dose–response relations of carotid blowout syndrome (CBOS) after stereotactic body radiotherapy (SBRT). To evaluate the goodness-of-fit and the correlation of those models with CBOS. Methods: The study included 61 patients with inoperable locally recurrent head and neck cancer treated with SBRT using CyberKnife (Accuray, Sunnyvale, CA) at the Department of Radiation Oncology, Hacettepe University, Ankara, Turkey between June 2007 and March 2011. The dose-volume histograms of the internal carotid were exported from the plans of all the patients. The follow-up results regarding the end point of carotid blowout syndrome were collected retrospectively. Initially, univariable analyses (Wilcoxon rank-sum or Chi-square tests) and a multivariate logistic regression analysis were performed between the outcome data and a list of clinical and treatment factors to identify significant correlations. Additionally, the Lyman–Kutcher–Burman (LKB), Relative Seriality (RS), and Logit NTCP models were used to fit the clinical data. The fitting of the different models was assessed through the area under the receiver operating characteristic curve (AUC), Akaike information criterion (AIC), and Odds Ratio methods. Results: The clinical/treatment factors that were found to have a significant or close to significant correlations with acute CBOS were Age at the time of CK (P-value = 0.03), Maximum carotid dose (P-value = 0.06), and CK prescription dose (P-value = 0.08). Using D_max, physical DVH, and EQD_2 Gy-DVH as the dosimetric metrics in the NTCP models, the derived LKB model parameters were: (a) D₅₀ = 45.8 Gy, m = 0.24, n = n/a; (b) D₅₀ = 44.8 Gy, m = 0.28, n = 0.01; and (c) D₅₀ = 115.8 Gy, m = 0.45, n = 0.01, respectively. The AUC values for the dosimetric metrics were 0.70, 0.68, and 0.61, respectively. The differences in AIC between the different models were less than 2 and ranged within ±0.9. Conclusion: The maximum dose to the internal carotid less than 34 Gy appears to significantly reduce the risk for CBOS. Age at the time of CK, Maximum carotid dose, and CK prescription dose were also found to correlate with CBOS. The values of the parameters of three NTCP models were determined for this endpoint. A threshold of gEUD <34.5 Gy appears to be significantly associated with lower risks of CBOS.