Statistical issues in study design

This chapter discusses the building blocks of an imaging research study, common biases in these studies, strategies for efficient designs, and sample size considerations. Imaging studies typically have five common building blocks: a research question, study sample, imaging test, reference standard (if applicable), and study outcomes. Design biases are common in research studies. Their effects vary depending on the type and magnitude of the bias. Investigators often design studies, recognizing that bias exists in the design they have chosen but realizing there is no practical alternative to avoid the bias. Planning the size of a research study is a critical component of the study design phase. The sample size needed for a study can tell us if a study is feasible or not, how long the study will take, and what it will cost. Most sample size calculations for imaging studies require an experienced statistician to perform.