OA076] Logistic regression modelling of radiation induced mucositis of H&N cancer patients

Abstract

Purpose Radiation-induced mucositis is a serious side effect, which can jeopardize treatment compliance and influence patient weight during treatment. The aim of this study was to develop a mathematical model to predict the risk of severe mucositis. Methods 535 curative radiotherapy H&N cancer patients from one institution between 2011 and 2015 were included. Doses were 66, 68 or 76 Gy in 33, 34 or 56fx. Patients were treated with IMRT/VMAT and mucosal reactions were scored weekly during radiotherapy. The highest observed score was used as endpoint and dichotomised in stage two stages: milder than confluent mucositis vs. confluent mucositis and worse. DVH of the extended oral cavity (Brower et al.) was extracted from the TPS. Principal component analysis was used to uncouple the highly correlated dose metrics ( V 5 , … , V 75 ) . Predictors available for the logistic model were the first 5 principal dose components, gender, weekly low dose chemotherapy, radiosensitizer, treatment acceleration, age, smoking status, tumour site, and volume of extended oral cavity. Parameter selection was performed using Least Absolute Shrinkage and Selection Operator (LASSO) within the statistical package R. The LASSO tuning parameter was chosen using 10-fold cross validation and 95% confidence interval found with 2000 bootstraps. Results Gender, acceleration, current smoker, tumour in the vicinity of the oral cavity, radiosensitizer, and the two first principal dose components were selected as predictors using Lasso. Acceleration is a well-known risk factor while the tumour position indicates an increased risk beyond the prediction related to the oral cavity dose. The protective value of being male and current smoker is in line with previous findings of toxicity in oesophagus and lung. The risk related to dose is dominated by the PC1. The model calibration plot (predicted vs observed risk) show good agreement with the line of identity. The bootstrap adjusted area under the curve (AUC) was 0.77 (95% CI 0.73–0.81). Conclusions A robust logistic regression model for prediction of radiation induced mucositis of H&N has been developed, which can be used as risk assessment of mucosal toxicity during treatment plan optimisation. The AUC value of 0.77 is significantly larger than previous published models on mucositis.