Random effects and latent processes approaches for analyzing binary longitudinal data with missingness: a comparison...

Area of Science:

• Biostatistics
• Longitudinal Data Analysis
• Missing Data Methods

Background:

• Analyzing longitudinal data with non-ignorable missingness is a significant challenge in biostatistics.
• Intermittent missing data and dropout are common issues in longitudinal studies, particularly in clinical trials.
• Existing methods may not adequately address the complexities of non-ignorable missing data mechanisms.

Purpose of the Study:

• To discuss and compare random effects and latent process models for longitudinal binary data with non-ignorable intermittent missing data and dropout.
• To evaluate the performance of these models using a real-world clinical trial dataset.
• To provide insights into robust statistical approaches for handling complex missing data patterns.

Main Methods:

• Review of random effects models for longitudinal data.
• Exploration of latent process models to account for missing data mechanisms.
• Application and comparison of these models to an opiate clinical trial dataset with high missingness and dropout.
• Comparative analysis with other methods for non-ignorable missing data.

Main Results:

• Random effects and latent process models offer viable approaches to handle non-ignorable missing data in longitudinal studies.
• These advanced models demonstrated utility in analyzing a clinical trial dataset with substantial intermittent missingness and dropout.
• The study provides a comparative framework for selecting appropriate methods based on data characteristics.

Conclusions:

• Random effects and latent process models are effective for analyzing longitudinal binary data with non-ignorable missingness and dropout.
• These sophisticated statistical techniques are crucial for unbiased inference in the presence of complex missing data.
• The findings support the use of these models in clinical trial data analysis where missing data is prevalent.