Statistical resolution of missing longitudinal data in clinical pharmacogenomics

Area of Science:

• Genomics
• Statistical genetics
• Pharmacogenomics

Background:

• Clinical pharmacogenomics integrates genomic data for predicting drug response, a key public health area.
• Longitudinal drug response data in trials often suffer from missing values due to patient dropout.
• Analyzing this missing data is a significant challenge for clinical pharmacogenomics.

Purpose of the Study:

• To develop a statistical algorithm for detecting genetic control of longitudinal responses with non-ignorable dropout.
• To jointly characterize genetic influences on both longitudinal drug response and dropout events.
• To address the challenge of missing longitudinal data in clinical pharmacogenomics.

Main Methods:

• Developed a statistical algorithm incorporating a selection model into a dynamic functional mapping model.
• The selection model assumes dropout is dependent on the drug response outcome.
• The model jointly analyzes genetic control of longitudinal responses and dropout events.

Main Results:

• The developed model can jointly characterize genetic control of longitudinal responses and dropout events.
• Simulation studies demonstrated the model's statistical properties and practical utility.
• The algorithm effectively handles non-ignorable dropout in longitudinal pharmacogenomic data.

Conclusions:

• The new statistical model addresses a critical challenge in analyzing missing longitudinal pharmacogenomic data.
• This approach facilitates the discovery of genetic variants impacting drug response and dropout.
• The findings have significant implications for advancing personalized medicine through clinical pharmacogenomics.