Note: Poster #18 is cross-listed in the poster competition category of Innovative Solutions to Public Health Challenges.
Poster # 18
Title: Group Sequential Design Without Proportional Hazards Assumption
Presenting Author: Yiming Chen (UMD SPH Doctoral Student in Department of Epidemiology and Biostatistics)
Faculty Mentors: Mei-Ling Ting Lee
Primary Category: Data Analytics, Surveillance, Community Needs Assessment, Pedagogy
Secondary Category: Innovative Solutions to Public Health Challenges
View Yiming Chen's Abstract
Background: The Non-Proportional Hazard (NPH) phenomenon is not uncommon in medical research. Traditional methods such as Cox model, Log-Rank test lose power significantly when PH assumption is violated.
Goal: We would like to apply TR – a survival model based on first hitting time to address the NPH challenge, especially for group sequential design.
Objectives: A treatment is considered as beneficial using TR, in three cases. We design clinical trials for each of these cases and derived sample size formula for Case 1/2.
Approach/Methods: The group sequential design of TR following the parametric GSD convention.
Results: TR can easily address the NPH challenge with physical meanings. It further protects the overall study power regardless of the PH assumption.
Importance to Public Health: As the PH assumption is difficult to assess before the data are collected, using Cox PH model that relies on the assumption but fail to justify it at the end of study may lead to wrong conclusion. Hence, it’s beneficial to use TR based tests that do not require the assumption at all.