Normal tissue complication probability (NTCP) models can be used to individualize radiation therapy treatment planning, potentially by guiding the design of the dose distribution. Our goal is to produce improved NTCP models, based on dose-volume, patient, and disease characteristics, for head and neck and lung treatment complications. Under the previous grant, we developed a software system, which enables the construction and convenient analysis of databases of 3-D treatment plans, including datasets from multiple institutions. Using data thus obtained, we will construct predictive models using multi-metric logistic regression methods, which include dose-volume terms as well as other patient and disease-related factors. The robustness of variable selection will be tested with bootstrap methods. For lung treatment plans, we will recompute lung and esophagus dose-volume histograms using a novel Monte Carlo-based technique, to improve the consistency and accuracy of the database dose distributions.
Under Specific Aim (SA) #1, Improvements in post-RT late pneumonitis/fibrosis NTCP models, we will: (a) expand the currently available Wash. Univ. dataset (from 166 pts. to an estimated 450 in 4 years), (b) study the inclusion of new factors such as spatially-varying sensitivity and pretreatment pulmonary function tests, (c) test and refine our model using the RTOG 93-11 dataset (113 pts.), and (d) test and refine our model against data contributed by Duke University and the Netherlands Cancer Institute (an estimated 550 pts.). Under SA #2, Improvements in acute esophagitis NTCP models, we will: (a) accrue more patients (from 166 to an estimated 450 in 4 years), (b) incorporate new factors such as partial-circumferential irradiation and other metrics based on the shape of the high dose region, and (c) test and refine our model using new data contributed by the Netherlands Cancer Institute (an estimated 300 pts.). Under SA #3, Improvements in post-RT parotid salivary function/xerostomia models, we will: (a) test the effect of spatial placement of high-dose regions, (b) use the model to analyze the radio-protective effect of Amifostine on salivary function in an ongoing intensity modulated radiation therapy trial, and (c) test/refine our model against the University of Michigan xerostomia dataset. In addition, we will establish publicly archived databases with convenient and freely available software tools. We hypothesize that this research will result in a significantly improved ability to predict, on an individualized basis, the risk of xerostomia, pneumonitis, or esophagitis, and could thereby lead to improved radiation therapy treatments.
Showing the most recent 10 out of 38 publications
