Image-based data mining (IBDM) and voxel-based analysis (VBA) have shown great promise in the retrospective analysis of routine clinical data, offering a way to analyze a patient population without selection. In radiation therapy, IBDM analyzes whole dose distributions for their effect on a given outcome with no prior assumptions. Recent extensions in IBDM methodologies include per voxel survival analyses, allowing the generation of more robust and testable hypotheses. An essential aspect to produce solid conclusions includes applying the correct statistical techniques to account for multiple testing. There is a growing interest in applying IBDM/VBA techniques in radiation therapy, which is a relatively new area with limited examples in the literature. We present this editorial to discuss guidelines for best practice. In doing so, we highlight 3 recently published papers applying IBDM techniques to radiation therapy data 1 , 2 , 3 and suggest alternative, more robust analysis methods.