Statistical Influence on Regulatory ProcessesJohn W Green, DuPont
*Jennifer H. Van Mullekom, DuPont
Keywords: regulations, power, design, animal welfare
Examples from development work leading to four newly issued OECD Test Guidelines and one in progress will be presented. These examples will demonstrate technical issues and presentation skills that were found to be useful, and some less successful efforts, gained over a long career. Technical issues addressed will include (1) poorly proposed experiment acceptance criteria that were successfully refuted; (2) an analysis using “normalized” data that violated important model assumptions and produced unreliable estimates, together with a scientifically sound alternative approach; (3) a reduced experimental design that save cost and animals and maintains adequate statistical power by selecting the right design and analysis; (4) a new statistical test with comparisons to previous tests that resolved issues for the subject matter specialists; (5) the use of statistical power to drive improved experimental design and statistical test selection; (6) a new statistical test proposed for an emerging new guideline contrasted with earlier methods. All examples are presented in a way accessible to non-statisticians and, in the regulatory framework, resulted in acceptance of the proposed methods.