Breast radiation treatment is burdened by acute and chronic toxicities, in most cases mild. However, considering the excellent life expectancy of patients with breast cancer, maintaining a low toxicity profile is of primary importance in order to guarantee a satisfactory quality of life. Currently there are numerous predictive models of toxicity (Normal Tissue Complication Probability, NTCP) which are based on dosimetric and sometimes also clinical data. To date, they do not include individual genetic variability. However, it is believed that inter-individual variability may be responsible for up to 40% of actinic toxicity. Multiparametric models that consider genetics, dose and clinical aspects probably better reflect the complexity of radiotoxicity than models that rely on a single parameter and it is possible to integrate such parameters using a machine learning approach. The definition of the molecular and genetic variables related to radiotoxicity and their integration into predictive molecular signatures would therefore allow the risk to be individualized. This would provide the clinician with a useful tool in order to personalize the radiation treatment, thus being able to choose the best technique or schedule for each patient.
