Normal tissue complication probability modeling of radiation-induced sensorineural hearing loss after head-and-neck radiation therapy

Purpose: The aim of this study was to generate the dose-response curves by six normal tissue complication probability (NTCP) models and ranking the models for prediction of radiation induced sensorineural hearing loss (SNHL) caused by head and neck radiation therapy (RT).

Materials and methods: Pure tone audiometry (PTA) was performed on 70 ears of patients for 12 months after the completion of RT. The SNHL was defined as a threshold shift ≤15 dB at two contiguous frequencies according to the common toxicity criteria for adverse events scoring system. The models evaluated were: Lyman and Logit; Mean Dose; relative seriality (RS); Individual critical volume (CV); and population CV models. Maximum likelihood analysis was used to fit the models to experimental data. The appropriateness of the fit was determined by the two-sample Kolmogorov-Smirnov test. Ranking of the models was made according to Akaike's information criterion.

Results: The dose of 50% complication rate (D₅₀) was 51-60 Gy. Three of the examined models fitted well with clinical data in a 95% confidence interval. The RS model was ranked as the best model of prediction for radiation induced SNHL.

Conclusions: Cochlea shows a different behavior against different NTCP models; it may be due to its small size.