The Use of the Win Ratio Method in Clinical Trials

Zheng, Sirui (2024) The Use of the Win Ratio Method in Clinical Trials, Thesis (Doctoral), Liverpool School of Tropical Medicine.

Text S Zheng_Thesis_LSTM_Ultimate.pdf - Accepted Version Restricted to Repository staff only until 1 June 2025. Download (1MB)

Abstract

A composite endpoint is referred to as an outcome that combines two or more endpoints of interest into a single variable and is commonly used as a primary outcome in clinical trials. Time-to-first-event analysis, negative binomial regression, and the Andersen-Gill model are commonly used to analyse and report the composite endpoints. However, the conventional methods for analysing composite endpoints have some limitations. The time-to-first event analysis only considers the first endpoint and treats all components of composite endpoints as equally important. The negative binomial regression and the Andersen-Gill model both allocate equivalent weights to each component. These limitations were addressed by Pococket al. in 2012 by introducing the win ratio approach, which gained wide attention. The win ratio approach was proposed and designed for a hierarchy of composite time-to-event outcomes to overcome equal weights to all components. However, there is no systematic evaluation of the statistical performance of the win ratio method in clinical trials in terms of analysis and reporting for various outcomes such as composite endpoints, survival data, ordinal data, etc.

In this thesis, I investigate the use of the win ratio approach for the analysis and reporting of clinical trial data. After a comprehensive review of the win ratio method, I propose the win ratio approach to analyse different types of clinical outcomes. The win ratio method seeks to serve as an alternative to conventional approaches or to outperform conventional methods in situations where assumptions are violated. In the thesis, I assess the performance of the win ratio method through Monte Carlo simulations and real-world trial datasets.
I propose the win ratio method to analyse survival data. Further, I introduce the adjusted win ratio methods to address the imbalanced covariates in composite endpoints. Additionally, I propose the use of win statistics for analysing ordinal data. Lastly, I propose the win ratio
method to evaluate the average bioequivalence for two common trial designs.
Taken together, my results demonstrate that the win ratio method has about the same power as the log-rank test, Cox model, and Agresti’s generalised odds ratio to detect the treatment difference. However, the win ratio method loses its statistical power in small sample-size
clinical trials, which requires further investigation.

Item Type:	Thesis (Doctoral)
Subjects:	W General Medicine. Health Professions > W 26.5 Informatics. Health informatics W General Medicine. Health Professions > W 20.5 Biomedical research WA Public Health > Statistics. Surveys > WA 950 Theory or methods of medical statistics. Epidemiologic methods
Repository link:	Item title Item URI Adjusted win ratio using the inverse probability of treatment weighting https://archive.lstmed.ac.uk/23505/ The association of long-term trajectories of BMI, its variability, and metabolic syndrome: a 30-year prospective cohort study https://archive.lstmed.ac.uk/24093/ A win ratio approach for comparing crossing survival curves in clinical trials https://archive.lstmed.ac.uk/21969
Faculty: Department:	Clinical Sciences & International Health > Clinical Sciences Department
Depositing User:	Lynn Roberts-Maloney
Date Deposited:	19 Jun 2024 13:17
Last Modified:	19 Jun 2024 13:20
URI:	https://archive.lstmed.ac.uk/id/eprint/24778

Statistics

View details

Actions (login required)

Edit Item