Efficient estimation of grouped survival models

Area of Science:

• Genomics
• Biostatistics
• Pharmacogenomics

Background:

• Experimental designs often lead to imprecise measurement of time- and dose-to-event data.
• Grouped failure time data, where events are recorded in discrete intervals, is common in drug toxicity studies.
• Ignoring this grouping can cause biased results in statistical analyses.

Purpose of the Study:

• To develop a statistically rigorous and computationally efficient R package for genome-wide analyses using grouped failure time data.
• To address the limitations of imprecise or incomplete measurement of time- and dose-to-event phenotypes.
• To enable accurate identification of genetic variants associated with clinical endpoints.

Main Methods:

• Development of the groupedSurv R package for analyzing grouped failure time phenotypes.
• Implementation of methods to adjust for baseline covariates.
• Facilitation of genome-wide association studies at the variant, gene, and pathway levels.

Main Results:

• The groupedSurv package provides a statistically sound and efficient approach for genome-wide analysis.
• Simulations demonstrate the statistical properties and computational performance of the package.
• Reanalysis of a study on taxane-induced peripheral neuropathy identified associated germline variants.

Conclusions:

• The groupedSurv package enables fast and rigorous genome-wide analysis for grouped failure time data.
• Analysis can be performed at the variant, gene, or pathway level.
• The package is publicly available on CRAN.