Activity Number:
|
477
- Methods in Clinical Trials
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 6, 2020 : 10:00 AM to 2:00 PM
|
Sponsor:
|
Biometrics Section
|
Abstract #312467
|
|
Title:
|
Design and Analysis Considerations for Three-Level Cluster Randomized Trials with a Time-to-Event Outcome
|
Author(s):
|
Ondrej Blaha* and Fan Li and Denise Esserman and Peter Peduzzi and Yize Zhao
|
Companies:
|
Yale University and Yale University and Yale University and Yale University and Yale University
|
Keywords:
|
Cluster Randomized Trials;
Additive Hazard Model;
Pragmatic Trials;
Time-to-event Data;
Multi-level Clustering
|
Abstract:
|
Pragmatic cluster randomized trials (CRTs) may involve three levels of clustering. For example, observations are taken from patients nested within clinics nested within health care systems. Most of the design and analysis strategies for three-level CRTs have focused on continuous and binary outcomes with little methodology development for time-to-event outcomes. Motivated by the STRIDE study, we consider an additive hazard model for the design and analysis of three-level CRTs. The additive hazard model carries a straightforward population-averaged interpretation and dispenses with the proportional hazard assumption. We discuss strategies to estimate model parameters in the context of three-level CRTs and consider improved variance estimators by incorporating finite-sample corrections. We develop a sample size procedure under the additive hazard model and examine its accuracy via simulations. Our numerical results indicate when data are analyzed with an additive hazard model employing a suitable bias-corrected sandwich variance, the empirical power agrees well with the predicted power even in three-level CRTs with a small number of clusters.
|
Authors who are presenting talks have a * after their name.