How Should Parallel Cluster Randomized Trials With a Baseline Period be Analyzed?-A Survey of Estimands and Common...

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Parallel cluster randomized trials with baseline (PB-CRTs) are common, but informative cluster sizes pose analytical challenges.
• Two key estimands, individual-average treatment effect (iATE) and cluster-average treatment effect (cATE), address different hypothesis levels.

Purpose of the Study:

• To theoretically derive the convergence of various treatment effect estimators in PB-CRTs with informative cluster sizes and continuous outcomes.
• To evaluate the consistency and robustness of different statistical models under these conditions.

Main Methods:

• Theoretical derivation of convergence for unweighted and weighted independence estimating equations (IEE), fixed-effects (FE) models, exchangeable mixed-effects (EME) models, and nested-exchangeable mixed-effects (NEME) models.
• Analysis focused on PB-CRTs with informative cluster sizes and continuous outcomes.
• Simulation study and re-analysis of a real-world PB-CRT.

Main Results:

• Unweighted and weighted IEE and FE models provide consistent estimators for both iATE and cATE.
• Mixed-effects models generally yield inconsistent estimators for these estimands with informative cluster sizes.
• The EME model demonstrated unexpected robustness to bias in PB-CRTs, unlike NEME models or analyses in standard P-CRTs.

Conclusions:

• Independence estimating equations and FE models are reliable for estimating iATE and cATE in PB-CRTs with informative cluster sizes.
• The EME model's robustness offers a potentially valuable alternative, contrasting with NEME models and standard P-CRT analyses.
• Findings have practical implications for the design and analysis of cluster randomized trials.