Assigning statistical significance to tumor changes in patient monitoring using FDG pet

Abstract

In PET-based patient monitoring, tumor changes can be assessed using standardized uptake values (SUV), tumor volume (V), or total lesion glycolysis (TLG). We studied the impact of the SUV, V and TLG estimation methods on the interpretation of tumor changes between 2 PET scans. We also propose a bootstrap approach to assign statistical significance to the observed tumor changes. In 17 tumor changes, the SUV variations were the least dependent on the estimation method compared to the V or TLG changes. In 16/17 cases, SUV changes were significant. In 2 out of these 16 significant cases, at least one SUV index suggested non significant change. Testing the significance of tumor feature changes might reduce errors in interpreting tumor changes.