Back to: FDA Clinical Investigator Training Course (CITC) 2024
Presenter
Mark Levenson, PhD
Acting Deputy Director
Office of Biostatistics
Center for Drug Evaluation and Research (CDER)
Food and Drug Administration (FDA)
Mark Levenson is currently the acting Deputy Director of the Office of Biostatistics in the Center for Drug Evaluation and Research of the US Food and Drug Administration (FDA). At FDA, he has led many major pre-market and post-market drug safety and efficacy reviews. He contributes to statistical policy and guidance development in the areas of real-world evidence, regulatory evidence, and drug safety. He is a member of the CDER Medical Policy Program Review Committee and the FDA Real-World Evidence Committee. Dr. Levenson received a Ph.D. in Statistics from the University of Chicago, a B.A. from Cornell University in Mathematics, and graduated from the Bronx High School of Science. Dr. Levenson is an elected fellow of the American Statistical Association.
Abstract
Mark Levenson’s presentation, “Statistical Principles for Clinical Development,” provides a comprehensive overview of how statistical concepts are fundamental to good study design, conduct, and analysis. He asserts that design and conduct are more critical than analysis, as analytical methods cannot rectify flaws introduced in earlier stages. A central aim of any study is to minimize bias and variability while ensuring ethical, safe, and feasible execution. Levenson differentiates between variability, which can be measured and is reduced by larger sample sizes, thereby increasing statistical power, and bias, which is unseen and is not mitigated by increasing sample size. Bias is primarily addressed through robust design and conduct elements, such as randomization to control confounding between treatment groups, blinding to combat measurement bias, and careful consideration of selection bias, especially issues arising from post-baseline patient characteristics like adherence. He highlights the intent-to-treat (ITT) principle as crucial for maintaining randomization and addressing certain biases by analyzing patients according to their initial random assignment, regardless of subsequent adherence. Furthermore, Levenson emphasizes pre-specification in protocols and statistical analysis plans as the primary solution to multiplicity, preventing spurious statistically significant findings from multiple comparisons. Ultimately, careful design and conduct are essential for achieving valid and reliable study results.