Evaluation of Statistical Methods for Sample and Diagnostics Test Stability

• Activity Number: 306
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Zhonggai Li*+ and Holger Hoefling and Anthony Rossini
• Companies: Novartis Molecular Diagnostics (MDx) and Novartis Molecular Diagnostics (MDx) and Novartis Molecular Diagnostics (MDx)
• Keywords: sample stability ; diagnostic test comparisons ; constant bias ; archived samples ; personalized medicine ; diagnostics equivalence

Abstract

A common statistical quantity that must be assessed in the development of clinical diagnostics is the stability of results over time. The concerns over stability of diagnostic results can be due to clinical sample storage, change in reagent batches, change of assay machines, introduction of new technologies into the protocol for scientific or economic reasons, or similar changes. Our research addresses the selection of statistical approaches to resolve the specific question ``how can we evaluate whether the storage of a clinical sample biases the stability of the diagnostic readout?'. We focus on the evaluation of constant bias/difference which could be due to sample storage or changes between a new and original test. These two situations often occur during development of companion diagnostics in supporting of personalized medicine as a result of the limited availability of precious clinical samples and desires to leverage newer technology for existing diagnostics for scientific or economic reasons. This type of study is a special case of the more general category of study called ``diagnostics equivalence', which is different from the well studied ``bio-equivalence' evaluation.