Optimization and Comparison of Normal Tissue Complication Probability Models in Radiotherapy

This work has been partially supported by French National Research Agency (ANR) through project TIGRE n° ANR-09-BLAN-0378-01 and by the NSFC under Grant 60911130370.International audienceIn the context of cancer radiotherapy, toxicity prediction is of the major importance to evaluate and compare dose plans. Normal tissue complication probability (NTCP) models are the major methods to predict and prevent the presentation of toxicities, but they have to be optimized and their predictive capacities have to be evaluated. In this investigation, the six main NTCP models were studied and their parameters were fitted on prostate cancer. The results argue that rectum toxicity within 2 years shows some characteristics of a serial organ (n=0.35). Poisson EUD and Logit EUD models have the better predictive abilities and their use in clinical routine should be studied in further works