Assessing correlation between toxicity and efficacy outcomes in oncology
In Radiation Oncology, there are often relatively large datasets containing detailed information on radiation dose as well as outcomes such as toxicity. Studies with biomarkers are becoming common but the number of patients in such studies tends to be smaller. One goal is often to develop predictive models for outcome using both dose and biomarkers. This raises the statistical question of how to utilize the historical or larger dataset with dose and toxicity but not biomarkers. We propose to study statistical approaches to this problem. As part of the Michigan Radiation Oncology Quality Consortium, we are collected dose and toxicity outcomes on many (over 1000) lung cancer patients. We also have detailed biomarker data (along with dose and toxicity outcomes) on approximately 120 patients enrolled on protocols here at UM. We will utilize these datasets to illustrate the methods and develop predictive models.