Testing homogeneity of two zero-inflated Poisson populations

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Pharmacovigilance

Background:

• Assessing treatment differences in safety parameters is crucial for clinical trials.
• The zero-inflated Poisson (ZIP) distribution is suitable for modeling count data with excess zeros, common in safety monitoring.

Purpose of the Study:

• To develop statistical methods for testing treatment differences in safety parameters using the ZIP distribution.
• To derive likelihood ratio tests (LRT) for homogeneity between two ZIP populations.
• To provide sample size calculation formulas for detecting clinically meaningful differences.

Main Methods:

• Derivation of likelihood ratio tests (LRT) for homogeneity of two ZIP populations.
• Formulation of hypotheses regarding differences in inflation parameters and non-zero means.
• Development of approximate sample size formulas for power calculations.

Main Results:

• The study presents LRTs for three distinct hypotheses concerning ZIP distribution parameters.
• Approximate formulas for sample size calculations are derived to ensure adequate statistical power.
• An illustrative example using gastrointestinal erosion counts is provided.

Conclusions:

• The developed LRTs and sample size formulas offer a robust framework for analyzing safety data in clinical trials.
• These methods are applicable to randomized parallel-group studies where safety counts follow a ZIP distribution.
• The findings aid in the efficient design and analysis of clinical trials, particularly for safety assessments.