Interim analyses in diagnostic versus treatment studies: differences and similarities.

The purpose of this paper was to contrast interim analyses in (randomized controlled) treatment studies with interim analyses in paired diagnostic studies of accuracy with respect to planning and conduct. The term 'treatment study' refers to a (randomized) clinical trial that aims to demonstrate the superiority or noninferiority of one treatment compared with another, and the term 'diagnostic study' to a clinical study that compares two diagnostic procedures, using a third diagnostic procedure as the gold standard. Though interim analyses in treatment studies and paired diagnostic studies show similarities in a priori planning of timing, decision rules, and the consequences of the analyses, they differ with respect to (1) the need for sample size adjustments, (2) the possibility of early decisions without early stopping, and (3) the impact of keeping results secret. These differences are due, respectively, to certain characteristics of paired diagnostic studies: the dependence of the sample size on the agreement rate between the modalities, multiple aims of diagnostic accuracy studies, and the advantages of early unblinding of results at the individual level. We exemplified our points by using a recent investigation at our institution on the detection of bone metastases from prostate cancer in patients with histologically confirmed prostate cancer in which (99m)Tc-MDP whole body bone scintigraphy was compared to positron emission tomography/computed tomography with (18)F-fluorocholine as tracer, using magnetic resonance imaging as a reference.