Statistical Challenges in the Development and Validation of Diagnostic Tests: A Regulatory Perspective
View Presentation View Presentation
*Gregory Campbell, U.S. Food and Drug Administration 

Keywords: microarrays, diagnostic test validation

The explosion of genetic and genomic information has posed challenges concerning the development and promulgation of good statistical practice for the construction and subsequent validation of complex diagnostic tests based on genetic, genomic or proteomic data. Various types of diagnostic tests or biomarkers are introduced and illustrated. The FDA Draft Guidance on In Vitro Diagnostic Multivariate Index Assays (IVDMIAs) is discussed. One instructive project that can illustrate the myriad of statistical challenges involving classifier construction and validation is the FDA-led Microarray Quality Control Project (MAQC). Its second phase (MAQC-II) was designed to build and evaluate the performance of predictive models based on microarray information. Statistical experiences and lessons learned associated with MAQC-II are reported.