Statistical problems in assessing the true value of new in vitro diagnostic tests, with special reference to the chemiluminescent assay (CLA, MAST)

Abstract

By virtue of familiarity and usage over many years, diagnostic tests may become established and be regarded as infallible "Gold Standards". In the case of allergic disease no test except perhaps allergen challenge can ever precisely reflect the clinical disease at the actual site of the allergic reaction. Thus when a new test is introduced there are no fixed values available for statistical comparison when directly comparing established tests with more recent additions. This is illustrated by many articles comparing the new multiple allergosorbent test system (CLA, MAST) which is a chemiluminescent assay, to other previously established tests. The major reports on CLA in the diagnosis of allergy are here discussed and summarized. Most trials have compared the accuracy of CLA with clinical history, with the skin prick test (SPT), the universally used in vivo method, or with RAST, the most common in vitro diagnostic method. CLA performed similarly to SPT and most CLA and RAST results matched; nevertheless in statistical analyses the problems of comparing diagnostic methods without a Gold Standard against which to judge become obvious-if two fallible methods disagree, which is correct? These statistical problems are discussed as well as other difficulties, such as the impact of the use of different allergen sources and the differing prevalence of allergies in the various test populations analysed.