A Bayesian quantile joint modeling of multivariate longitudinal and time-to-event data.

Area of Science:

• Biostatistics
• Clinical Research
• Computational Biology

Background:

• Traditional linear mixed models struggle with non-Gaussian longitudinal data and event-time outcomes.
• Quantile regression offers a more appropriate approach for non-Gaussian data, allowing analysis across different outcome levels.
• Understanding how time-varying covariates influence event-time across various outcome quantiles is crucial for complex diseases.

Purpose of the Study:

• To develop and apply a Bayesian quantile joint model for simultaneously analyzing longitudinal biomarkers and time-to-relapse in Acute Lymphocytic Leukemia (ALL) patients.
• To investigate the impact of longitudinal biomarker levels on relapse risk across different quantiles.
• To assess the effects of treatments (6MP and MTx) on biomarkers and their relationship with relapse time.

Main Methods:

• A Bayesian quantile joint model was developed for three longitudinal biomarkers (lymphocyte, neutrophil, platelet counts) and time-to-relapse.
• The Asymmetric Laplace Distribution (ALD) was used for outcomes, with its mixture representation enabling a Gibbs sampler algorithm for parameter estimation.
• The model allows for different quantile levels for each biomarker while simultaneously estimating regression coefficients for specific quantile combinations.

Main Results:

• Higher lymphocyte counts were found to accelerate relapse risk, whereas higher neutrophil and platelet counts jointly reduce relapse risk.
• The drug 6-mercaptopurine (6MP) was inferred to reduce lymphocyte counts across most quantiles.
• The drug methotrexate (MTx) was inferred to increase neutrophil counts across most quantiles.

Conclusions:

• The proposed Bayesian quantile joint model effectively analyzes complex longitudinal data and event-time outcomes in cancer research.
• Biomarker levels have a significant, quantile-dependent impact on relapse risk in ALL patients.
• Specific chemotherapy drugs exhibit distinct effects on biomarker trajectories and, consequently, on relapse risk.