Vaccines, Blood & Biologics
Statistical and Bioinformatics Tools for the Analysis of Blood Products and Pharmacogenomic and Proteomic Data
Principal Investigator: Boris Zaslavsky, PhD
Office / Division: OBE / DB
Public Health Issue: Our ability to provide sponsors with sound counsel on design of studies is directly related to our public health mission to evaluate the safety, efficacy and manufacturing of biological products.
Regulatory Contribution: It is important to be able to provide authoritative advice to sponsors about methods to determine the best sampling strategy in diagnostic methodology and to calculate patient sample size for tolerance intervals for binary outcomes. It becomes crucial to evaluate bioinformatics tools for assessing pharmacogenomic and proteomic data submitted for regulatory submissions. Providing appropriate statistical advice to clinical and product reviewers is a major regulatory contribution.
Research Approach: Three major areas are addressed: first, to provide new methods to determine whether to use a paired or independent sampling scheme to evaluate a diagnostic scheme with dichotomous endpoints; second, to determine patient sample size to compute tolerance intervals for dichotomous endpoints; third, to evaluate software for pharmacogenomic or proteomic data that is submitted to the FDA. This research program provides the theoretical basis for clinical design strategy for dichotomous data issues including pairing or independent sample, and study size for tolerance intervals with dichotomous data. It will provide a detailed evaluation of current software for microarrays and proteomics submissions.
Mission Relevance and Outcomes: This research provides the basis for appropriate advice to sponsors and FDA regarding dichotomous variables. The evaluation of current software is a key aspect of the review process for pharmacogenomics and proteomics applications.
Stat Med 2005 Aug 30;24(18):2837-2855
Assessing equivalence or non-inferiority in screening tests with paired and independent binomial endpoints through confidence intervals for the odds ratio.