Statistical Power in Quantitative Diffusion MRI of Tumor Response

Abstract

Rationale and Objectives Diffusion magnetic resonance imaging may be useful in tracking tumor growth and response to treatment. However, studies employing these measures may lack statistical power to draw definitive conclusions regarding changes in tumor cellularity. Using apparent diffusion coefficient (ADC) values taken from the literature, we estimate sample size for a range of changes to the mean. Materials and Methods We performed a literature search of studies measuring the average diffusion coefficient for various bodily tissues and recorded the mean and standard deviation from each article. We then performed analyses of statistical power using these values and comparing them to a population of healthy controls. Results Tumor cellularity as measured by ADC may have high sensitivity, but our analyses indicate that investigations in this field may potentially suffer from low statistical power. For example, our findings indicate that samples of less than 20 patients may require a mean change of approximately 25% between study conditions. Conclusions We offer suggestions for improvements in methodological approach and in data reporting to assist in overcoming limitations of small sample sizes. Based on our literature review, reference values are provided to help investigators estimate study sample size to achieve adequate statistical power.