Statistical Methods for Composite Endpoints

Win Ratio and Beyond

Author

Lu Mao

Course Info

This is a companion site for the same-titled workshop for the 44th annual Applied Statistics Workshop hosted by the Southern California Chapter of the ASA.

Time and Place

  • Friday, April 25 · 9am - 4pm PDT
  • Platt Conference Center, Cooper Auditorium
    City of Hope, 1500 East Duarte Road, Duarte, CA 91010

Instructor Profile

Lu Mao, PhD
  • Associate Professor of Biostatistics at UW-Madison
  • Methodologic research
    • R01HL149875: Novel Statistical Methods for Complex Time-to-Event Data in Cardiovascular Clinical Trials (12/01/2019 – 07/31/2028)
    • DMS2015526: Randomized Trials with Non-Compliance (07/01/2020 – 06/30/2024)
  • Collaborative research
    • Cardiovascular disease, cancer, radiology, behavioral health interventions
  • Teaching
  • Editorial service

Learning Outcomes

  • Identify statistical and regulatory challenges with composite endpoints
  • Apply key methods such as hypothesis testing, power analysis, and regression
  • Gain hands-on experience with real data using R packages

Syllabus

  • 1. Introduction (30 min)
    • 1.1 Examples & guidelines
    • 1.2 Traditional methods and limitations
    • 1.3 Win ratio methods and limitations
  • 2. Hypothesis Testing (40 min)
    • 2.1 Win ratio by Pocock et al. (2012)
    • 2.2 Statistical properties
    • 2.3 Handling recurrent events (R-package WR)
    • 2.4 Sample size calculation (R-package WR)
  • 3. Nonparametric Estimation (60 min)
    • 3.1 Restricted win ratio
    • 3.2 Average win time analysis (R-package rmt)
      • 3.1.1 Restricted mean time in favor of treatment
      • 3.1.2 Estimation, inference, and graphics
      • 3.1.3 Real Examples
    • 3.3 While-alive loss rate (R-package WA)
      • 3.2.1 Definition, interpretation, and estimation
      • 3.2.3 Real Examples
  • 4. Semiparametric Regression (45 min)
    • 4.1 Proportional win-fractions model (R-package WR)
      • 4.1.1 Model assumptions
      • 4.1.2 Estimation, inference, and model diagnostics
      • 4.1.3 Real Examples
    • 4.2 Variable selection and risk prediction
      • 4.2.1 Elastic net approach (WRNet)
  • 5. Discussions (25 min)
    • 5.1. Open problems
      • 5.1.1 Covariate adjustment
      • 5.1.2 Interim monitoring
      • 5.1.3 Meta analysis
    • 5.2. Course summary

R-Packages

You will need the following R packages for this workshop:
Wcompo, WR, rmt, and WA, along with base packages like tidyverse and survival.

install.packages(c("Wcompo", "WR", "rmt", "WA"))
library(tidyverse)
library(survival)
library(Wcompo)
library(WR)
library(rmt)
library(WA)