Analysis of Cohort Stepped Wedge Cluster-Randomized Trials With Nonignorable Dropout via Joint Modeling

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Longitudinal Data Analysis

Background:

• Stepped wedge cluster-randomized trials (CRTs) involve complex analysis due to intra-cluster correlation and informative dropout, especially in high-mortality cohorts.
• Missing outcomes from mortality can bias treatment effect estimates and reduce statistical power in closed-cohort CRTs.
• Existing methods may not adequately address nonignorable missing outcomes due to death in longitudinal studies.

Purpose of the Study:

• To extend linear mixed-effects models for analyzing closed-cohort stepped wedge CRTs with informative dropout.
• To develop a joint longitudinal-survival modeling framework that explicitly models the dropout process.
• To evaluate the performance of the proposed methodology through Monte Carlo simulations.

Main Methods:

• Joint longitudinal-survival models were developed, linking a time-to-event submodel for dropout with a longitudinal outcome submodel.
• The methodology was extended from linear mixed-effects models to accommodate informative missingness.
• Constant intervention and general time-on-treatment effect parametrizations were considered for the longitudinal component.

Main Results:

• The proposed joint modeling approach effectively accommodates informative dropout in stepped wedge CRTs.
• Monte Carlo simulations demonstrated the methodology's performance under various data-generating scenarios.
• Reanalysis of the Frail Older Adults: Care in Transition (ACT) trial illustrated practical application.

Conclusions:

• Joint longitudinal-survival models provide a robust framework for analyzing stepped wedge CRTs with informative dropout.
• This approach is essential for unbiased estimation of treatment effects in studies with high mortality rates.
• The methodology enhances the validity and power of findings from complex longitudinal trial designs.