Statistics Seminar - Amy Liu - Statistical Models for Radiotherapy Induced Normal Tissue Complications
Speaker: Amy Liu (Princess Margaret Cancer Centre)
Title: Statistical Models for Radiotherapy Induced Normal Tissue Complications
Abstract: The dose-volume histogram (DVH), which relates the radiation dose to tissue volume receiving it, is a commonly used measure for radiation treatment planning, as well as for associating radiation dose to toxicity outcomes - also known as normal tissue complication probability (NTCP) modeling. Such results may be used clinically to calculate probability of toxicity under alternative treatment plans; however to guide modification of the plans, interpretation of the models is also important. In principle DVH is a functional exposure and its effect on toxicity can be modeled through functional generalized linear models. On the other hand, the complement of the DVH is the probability distribution of the dose. In this talk I review the limitations of conventional functional principal component regression in this context, and discuss the alternative of monotonic functional regression which is more biologically plausible. We propose a new monotonic regression model for the DVH-toxicity relationship with a view of extending it to interactions between dose and patient characteristics and between doses received by multiple organs. This research is motivated by NTCP modeling for anal canal cancer at Princess Margaret Cancer Centre.