Statistical issues on the diagnostic multivariate index assay for targeted clinical trials.

In the past decade, pharmacogenomics and microarrays are considered two of the most important scientific breakthroughs for detection and treatment of diseases with many other applications. After completion of the Human Genome Project (HGP), the importance of diagnostic tests for identification of molecular targets increases as more targeted clinical trials are conducted for the individualized treatment of patients in the post-genomic era. As a result, the co-development of drug-device has become the foundation for achieving the ultimate goal of personalized medicine. One of the diagnostic devices for detection of molecular targets is the in vitro diagnostic multivariate index assays (IVDMIA) based on the genomic composite biomarker (GCB) classifiers. Thus, the quality of the IVDMIA, innovative designs, and the evaluation of efficacy for targeted clinical trials are vital for achieving this goal. However, before personalized medicine becomes a reality, many challenges for assurance of the accuracy and precision of the IVDMIA and the estimates of the treatment effect in the population with the molecular targets are to be resolved. In this paper, we identify the following issues on the IVDMIA and targeted clinical trials: (i) the selection of the differentially expressed genes, (ii) the optimal representation and algorithm for the genomic composite biomarker (GCB) classifier for the best diagnostic accuracy of the molecular target, (iii) the validation of the IVDMIA, and (iv) the evaluation of effectiveness and sample size estimation for targeted clinical trials. For each issue, the problem and possible resolutions are discussed. An overall assessment and some concluding remarks are also provided.

Duke Scholars

Author Shein-Chung Chow Biostatistics & Bioinformatics, Division of Biostatistics