Sample size considerations for matched-pair cluster randomization design with incomplete observations of continuous...

Area of Science:

• Biostatistics
• Clinical Trials
• Health Behavioral Research

Background:

• Matched-pair cluster randomization is increasingly used in clinical and health behavioral studies.
• Incomplete observations (missing data) are a common challenge in these study designs.
• Existing statistical inference methods for this design with missing data are available, but sample size methods are scarce.

Purpose of the Study:

• To propose a closed-form sample size formula for matched-pair cluster randomization designs with continuous outcomes and incomplete observations.
• To provide a flexible formula that accommodates various correlation structures, missing data patterns, and degrees of missingness.
• To offer a more accurate sample size estimation method compared to crude adjustment techniques when dealing with missing data.

Main Methods:

• Development of a closed-form sample size formula based on the generalized estimating equation approach.
• Treatment of incomplete observations as missing data within a marginal linear model framework.
• Utilizing bias-corrected variance estimators to mitigate inflated Type I error rates, especially with few clusters per group.

Main Results:

• The proposed sample size formula is flexible and accounts for missing data complexities.
• Simulation studies demonstrate the method's finite-sample performance across various configurations.
• The new method provides more accurate sample size estimations than crude adjustments in the presence of missing data.

Conclusions:

• The developed sample size formula is a valuable tool for planning matched-pair cluster randomization studies with continuous outcomes and missing data.
• The method enhances accuracy in sample size estimation, addressing a critical gap in research methodology.
• Application to a physical fitness study in adolescents illustrates the practical utility of the proposed formula.