Diagnostic test validation studies when there is a perfect reference standard

Before tools became available to consider diagnostic test validation studies where a ‘gold-standard’ is not available, new diagnostic tests were compared to a reference standard assumed to be highly accurate if not perfect. This paper reviews such ‘traditional’ situations with examples and methods of study design and analysis. Three situations are described, two where a perfect reference is available for either positive or negative animals, and one where the reference is perfect for both. Thus, here the authors review circumstances to be considered when validating a diagnostic test with a credible reference standard. An appropriate study design requires an unbiased selection of animals from the population to which a new test will be applied. Examples for calculating sample size and data analysis are provided. Finally, the authors discuss situations where it may be appropriate to include influential variables (‘covariates’) in a diagnostic test validation study..