January 12, 2021 - Jing Ning, PhD, Associate Professor, Department of Biostatistics, Division of Quantitative Sciences, The University of Texas MD Anderson Cancer Center, "Statistical Modeling and Adjustment for Sampling Biases" [download flyer]
Zoom Link: https://bit.ly/FIBJan12
Bias sampling mechanisms are commonly encountered in applications where the subjects in a target population are not given an equal chance to be selected, either accidentally, by natural circumstances, or intentionally by design. Statistical methods not properly accounting for such a challenge often lead to invalid inferences. For example, evidence combined from published studies may lead to overly optimistic conclusions due to publication bias, and the well-known length bias can cause the screening to appear to be more successful than it really is. In this talk, I will present our recent work to adjust the sampling biases in diverse applications such as the survivorship bias in prevalent cohort, the self-reporting bias in longitudinal analysis and the publication bias in meta-analysis.
February 09, 2021 - Keaven Anderson, PhD, Scientific AVP, Methodology Research, Biostatistics at Merck, "Group Sequential Design Assuming Delayed Benefit" [download flyer]
Zoom Link: https://bit.ly/FIBFeb21
We consider an asymptotic approach to design of group sequential trials with a potentially delayed effects. Logrank, weighted logrank tests and combination tests are of primary interest, but we also consider restricted mean survival. The asymptotic approach allows both quick derivation of study design properties that are also easily verified using simulation. We rely heavily on work done by Tsiatis, 1981 published while he was at the HSPH. The impact of a potential delay in treatment effect on timing of analyses, study boundaries and sample size are demonstrated. The value of a robust design that is well powered under a variety of assumptions is emphasized. Open source software is provided for implementation. Given the current regulatory climate, logrank testing may still be preferred for these trials, but we hope that efficiencies and Type I error control associated with alternatives may make other options more acceptable for future consideration.