[Cost effective planning of controlled trials with diagnostic instruments of limited reproducibility]

This paper proposes a simple approach for sample size prediction in controlled diagnostic or clinical trials for group comparison when the diagnostic (imaging) instrument under consideration shows limited reliability and thus calls for taking replicate measurements on each study subject. Using a specified cost function for the overall trial, a cost optimum replication number per subject is predicted, resulting in a cost-optimum predicted group size. This approach results in remarkable cost-effectiveness, as single assessment costs based on the imaging method under consideration only present a small fraction of the overall costs of preparing study subjects for a series of measurement replications by, for example, injection of nuclear or contrast agents.