More Winners, or Fewer Losers? Mastering Win Ratios in Statistical Analysis: Innovative Methods and Illuminating Case Studies

Hope Knuckles Chair
Abbott Laboratories
 
Tyson Rogers Discussant
NAMSA
 
Hou-Cheng Yang Organizer
Edwards Lifesciences
 
Luoxi Shi Organizer
 
Zhaoxun Hou Organizer
Edwards Lifesciences
 
Shiyu Wang Organizer
Edwards Lifesciences
 
Wednesday, Aug 6: 8:30 AM - 10:20 AM
0640 
Topic-Contributed Paper Session 
Music City Center 
Room: CC-101D 
In the field of clinical trials, win ratios have emerged as a pivotal metric for evaluating treatment efficacy, providing a more nuanced comparison of patient outcomes than traditional methods. This presentation offers an in-depth exploration of win ratios, focusing on their innovative methodologies and practical applications within clinical research. This session explores the advanced methodologies behind win ratios, highlighting their application in the rigorous environment of clinical research. We will delve into the techniques, including sophisticated adjustments for ties and censored data, and discuss the integration of win ratios into various statistical models to enhance their interpretative power.

Through a series of detailed case studies, we will illustrate the practical use of win ratios in clinical trials. Our speakers' background from different therapeutic areas, and academia will bring some case studies, and their insight. These case studies will showcase how win ratios can provide deeper insights into treatment effects, beyond traditional metrics. These case studies will cover a range of therapeutic areas, illustrating how win ratios can provide deeper insights into treatment effects and patient outcomes. For example, we will explore a clinical trial where win ratios were used to compare the efficacy of two competing treatments, by analyzing composite endpoints based on clinical priorities, and a user-friendly R-package. In addition to showcasing successful applications, we will discuss the common challenges and limitations associated to clinical trials. Strategies to overcome these challenges will be discussed, ensuring that attendees are equipped to apply win ratios effectively and accurately in their own research. Attendees will learn about the challenges and limitations encountered in real-world applications and discover strategies to address these issues, ensuring robust and reliable results.

Finally, we will look towards the future of win ratios in clinical trials, discussing potential innovations in methodology and expanding applications. Mastering the application of win ratios in clinical trials, uncovering innovative methods and gaining valuable insights from illuminating case studies. These presentations aim to equip clinical researchers, biostatisticians, and practitioners with the knowledge and tools to leverage win ratios, ultimately advancing the field of clinical research and improving patient outcomes. This comprehensive exploration promises to enhance audiences' understanding and application of these powerful statistical tools, driving forward the quality and impact of clinical research.

Keywords

Win Ratio

clinical trial 

Applied

Yes

Main Sponsor

Section on Medical Devices and Diagnostics

Co Sponsors

Biopharmaceutical Section
Stats. Partnerships Among Academe Indust. & Govt. Committee

Presentations

An IPCW adjusted Win Statistics Approach in Clinical Trials Incorporating Equivalence Margins to Define TiesPresentation

In clinical trials, multiple outcomes of different priorities commonly occur as the patient's response may not be adequately characterized by a single outcome. Win statistics are appealing summary measures for between-group difference at more than one endpoint. When defining the result of pairwise comparisons of a time-to-event endpoint, it is desirable to allow ties to account for incomplete follow-up and not clinically meaningful difference in endpoints of interest. In this paper, we propose a class of win statistics for time-to-event endpoints with a user-specified equivalence margin. These win statistics are identifiable in the presence of right-censoring and do not depend on the censoring distribution. We then develop estimation and inference procedures for the proposed win statistics based on inverse-probability-of-censoring (IPCW) adjustment to handle right-censoring.  We conduct extensive simulations to investigate the operational characteristics of the proposed procedure in the finite sample setting. A real oncology trial is used to illustrate the proposed approach. 

Keywords

Win Statistics

Inverse Probability of Censoring Weighting 

Speaker

Ying Cui, Stanford University

Enhancing the Win Ratio Method for Composite Endpoints: The Win Ratio with Multiple Thresholds (WR-MT)

Composite endpoints combining terminal (e.g., death) and non-terminal (e.g., hospitalization) events are widely used in cardiovascular clinical trials. The Win Ratio (WR) method, introduced by Pocock et al. (2012), prioritizes fatal events in a strict hierarchical manner, which may reduce power when treatment effects are primarily on non-fatal outcomes. We introduce the Win Ratio with Multiple Thresholds (WR-MT), which relaxes this strict prioritization by incorporating additional stages with non-zero thresholds. Our weighted adaptive approach selects these thresholds dynamically, maintaining the desirable statistical properties of the standard WR while improving the capacity to detect effects on non-fatal outcomes. Through simulations accounting for follow-up time, event associations, and treatment effect levels, as well as an application to the Digitalis Investigation Group trial, we demonstrate WR-MT's overall more favorable performance. This enhancement provides an additional statistical tool for analyzing composite endpoints in clinical trials. 

Keywords

Win ratio

Win statistics


Composite endpoints

Pairwise comparison

Clinical trials 

Speaker

Yunhan Mou

Operating Characteristics of Hierarchical Composite Endpoints in Noninferiority Settings

The Win Ratio (WR) is a hierarchical composite endpoint which has been used in controlled trials in order to compare efficacy of two treatments across multiple components ordered by clinical priority (Pocock, 2012). Pairwise comparisons are performed between all patients in the treatment and control arms, and the total number of "treatment wins" may be divided by the number of "treatment losses" to calculate the WR. The non-parametric nature of this statistic, as well as its ability to incorporate clinical endpoints of varying modalities (time-to-event, continuous, dichotomous) has made it popular for superiority trials, but the operational characteristics of this statistic have not been adequately investigated when the goal of the trial is to declare noninferiority (NI). Due to suboptimal performance of the WR in the event of a large number of ties, another statistic, the Win Odds (WO) has been proposed as an alternative (Peng 2020) This simulation study examines a hypothetical NI trial where the hierarchical composite endpoint has three levels. Operating characteristics for WR and WO will be compared across multiple choices of NI boundary and event rate in order to demonstrate utility of these statistics in NI trials. 

Keywords

Win Ratio

Noninferiority 

Speaker

Shirley Liao, Shirley Liao

Restricted Time Win Ratio: From Estimands to Estimation

The win ratio is a novel approach to analyzing prioritized composite endpoints in clinical trials. The win ratio has grown increasingly popular and has been applied in many ongoing cardiovascular clinical trials. Despite the intuitive interpretation of win ratio statistics, the traditionally used win ratio estimands are heavily influenced by the pattern of censoring. Treating missing values or censored data as ties is generally not a reliable way of handling missing values even in the estimation procedure. To avoid this complication, we propose the restricted time win ratio, which is the win ratio at a pre-specified time point and can be written in terms of the expectations of potential outcomes. To estimate the proposed estimand, we develop an approximately fully conditional specification algorithm which imputes the missing data of longitudinal and clinical outcomes jointly up to the pre-specified time point. The imputation algorithm can be easily revised to adjust a different hierarchical structure of the sequential comparisons specified by the users. Simulation studies in realistic settings show that the proposed imputation algorithm provides an unbiased and robust estimation of the restricted time win ratio. Finally, real data from a cardiovascular clinical trial are used to illustrate the proposed method. 

Keywords

Win Ratio

Estimand

Hierarchical composite endpoint 

Speaker

Tuo Wang, Eli Lilly and Company