Quantile regression for survival data with covariates subject to detection limits.

Area of Science:

• Biostatistics
• Biomedical Data Science
• Survival Analysis

Background:

• Biomarkers are crucial for predicting patient survival, but measurements are often censored due to detection limits.
• This dual censoring (survival outcomes and biomarker covariates) presents significant statistical modeling challenges.
• Existing methods like linear regression or accelerated failure time models have limitations in handling these complex censoring scenarios.

Purpose of the Study:

• To propose a novel quantile regression method for survival data with covariates subject to detection limits (DL).
• To provide a more versatile statistical tool for modeling survival outcomes and biomarker effects across different quantiles.
• To address the limitations of existing methods in handling dual censoring in biomedical research.

Main Methods:

• Developed a quantile regression approach for survival data incorporating covariates with detection limits.
• Introduced a novel multiple imputation technique based on quantile regression for censored covariates.
• Avoided stringent parametric assumptions often required by traditional methods for censored covariates.

Main Results:

• The proposed estimation procedure yields uniformly consistent and asymptotically normal estimators.
• Simulation studies confirmed the satisfactory finite-sample performance of the new method.
• The method was successfully applied to analyze sepsis data involving genetic and inflammatory markers.

Conclusions:

• The novel quantile regression method offers a robust and flexible approach for analyzing survival data with censored biomarkers.
• This method enhances the ability to model covariate effects across the distribution of survival times.
• The approach provides a valuable tool for prognostic prediction in biomedical research, particularly in complex datasets with detection limits.