Statistical inference for sequential designs of randomized clinical trials with binary responses

Jayasooriya, Apsara Pathum (2023) Statistical inference for sequential designs of randomized clinical trials with binary responses. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (1MB)

Abstract

Sequential designs of Randomized Clinical Trials (RCT) allow repeated significance testing based on cumulative data over time. The sequential testing method enables early termination of the study using a pre-defined stopping rule when preliminary results show a clear superiority of one treatment over the other. Over the decades, researchers have presented several techniques for determining the stopping rule, mainly for continuous data. However, clinical trial data are not necessarily continuous. In certain cases, data can be dichotomous, containing only two distinct values. Some researchers have proposed special sequential testing procedures to analyze binary data considering individual data points at each stage. With the in uence of those approaches, we are more focused on a method which can be used to analyse groups of binary data. The thesis considers the implementation of three main approaches, namely, Pocock [32, 34], O'Brien and Fleming [29] and Haybittle-Peto [31, 15] methods for computing the critical values required for controlling the size and power of tests at various stages of sequential analysis. Critical values are obtained using an iterative Markov chain approach to satisfy the alpha spending at each stage. Considering the discrete nature of the data, a likelihood ratio test statistic is used for testing the proportions. Examples of two-stage and three-stage analysis were used to illustrate the computation of the critical values, size and power of tests of proportions, and then the outcomes based on Pocock, O'Brien & Fleming and Haybittle-Peto methods are compared.

Actions (login required)

View Item