All ETDs from UAB

Advisory Committee Chair

Jeff M Szychowski

Advisory Committee Members

George Howard

Leslie A McClure

Charles M Wilcox

John Owen

Document Type


Date of Award


Degree Name by School

Doctor of Philosophy (PhD) School of Public Health


Study designs that allow sample size review may potentially save a study by identifying misspecified assumptions, modifying the planned sample size, and reinforcing the study’s power. Most sample size re-estimation (SSR) methods in non-inferiority trials have been developed for continuous or binary outcomes. In many medical domains studying cancer, heart failure, and other chronic diseases, most phase III trials use censored endpoints (e.g. survival or event-free survival). Non-inferiority designs have inherent complexity by the introduction of the non-inferiority margin and therefore understanding of strategies in these methods is essential. In our first paper, we detail the issues related to misspecification of the hazard ratio (HR) in non-inferiority trials with censored survival outcomes and investigate interim monitoring strategies and current sample size re-evaluations to control the type I error and power. In our second paper, we review four different methods for the re-evaluation of sample size in superiority designs with censored outcomes and extend them to the context of non-inferiority designs with censored outcomes. We find that the sample size re-estimation method using conditional power with a weighted log-rank test statistic proposed by Cui, Hung, and Wang (CHW) outperforms the other methods. We detail the differences and similarities between the methods through simulation studies in terms of type I error rate control, power, and additional sample size to be recruited. In our third paper, we apply this CHW method to data collected from the CREST trial and illustrate an example of a non-inferiority trial that could benefit from sample size recalculation at the interim monitoring stage.

Included in

Public Health Commons