A GEE Approach to Determine Sample Size for Pre- and Post-Intervention Experiments with Dropout.

Area of Science:

• Biostatistics
• Medical Research Methodology
• Social Behavioral Sciences

Background:

• Pre- and post-intervention studies commonly involve paired observations from each subject.
• Missing data in post-intervention measurements due to subject dropout is a frequent challenge.
• Existing sample size calculations, like McNemar's test, do not adequately handle missing data.

Purpose of the Study:

• To derive a sample size formula for pre- and post-intervention studies with missing data.
• To account for the impact of partial observations on sample size requirements.
• To compare the proposed method with traditional approaches.

Main Methods:

• Development of a closed-form sample size formula using the generalized estimating equation (GEE) approach.
• Investigation of sample size requirements under scenarios with missing post-intervention data.
• Validation through simulation studies and a practical example.

Main Results:

• The GEE-based sample size formula effectively accommodates missing data in paired observations.
• When no data is missing, the GEE estimate closely aligns with the McNemar's test calculation.
• The proposed method demonstrates substantial sample size savings in the presence of missing data.

Conclusions:

• The GEE approach provides a robust method for sample size calculation in longitudinal studies with dropouts.
• This method offers practical advantages by reducing the required sample size when data is incomplete.
• The derived formula is valuable for researchers in medical and social behavioral fields facing data attrition.