Sample size for comparing linear growth curves

Area of Science:

• Biostatistics
• Clinical Trial Design
• Longitudinal Data Analysis

Background:

• Comparing treatment efficacy often involves repeated measurements over time.
• Accurate sample size calculation is crucial for the validity and efficiency of clinical trials.
• Longitudinal data analysis requires specialized statistical methods to account for within-subject correlations.

Purpose of the Study:

• To develop a sample size computation method for comparing two treatment groups in longitudinal studies.
• To address scenarios where repeated measurements follow exponential family distributions.
• To provide a framework for sample size determination under a linear growth curve model.

Main Methods:

• Utilized a linear growth curve model with a suitable link function.
• Employed generalized estimating equations (GEE) to estimate regression parameters.
• Derived sample size based on the asymptotic normality of the difference between treatment slopes (delta).
• Incorporated dispersion matrices under null and alternative hypotheses using a working correlation matrix.

Main Results:

• A formula for sample size calculation was derived for comparing two treatment groups.
• The method is applicable to repeated measures designs with binary and count data.
• Numerical examples illustrate the practical application of the sample size computation.

Conclusions:

• The proposed method offers a statistically sound approach to sample size determination in longitudinal studies.
• This methodology facilitates robust clinical trial design for interventions with repeated outcome measures.
• The findings are particularly relevant for trials involving binary or count data analyzed using generalized linear models.