RETRACTED ARTICLE: A Bayesian joint model for multivariate longitudinal and time-to-event data with application to...

Area of Science:

• Pediatric Oncology
• Biostatistics
• Translational Cancer Research

Background:

• Acute lymphocytic leukemia (ALL) is the most common childhood cancer.
• Longitudinal biomarkers and drug effectiveness in ALL relapse prediction require further investigation.

Purpose of the Study:

• To identify longitudinal biomarkers associated with time-to-relapse in pediatric ALL patients.
• To assess the effectiveness of 6-mercaptopurine (6MP) and methotrexate (MTx) in relation to relapse probability.
• To develop and validate a Bayesian joint model for analyzing longitudinal biomarkers and time-to-relapse.

Main Methods:

• A Bayesian joint model was developed, integrating a linear mixed model for longitudinal biomarker data (white blood cell, neutrophil, platelet counts) and a semi-parametric proportional hazards model for time-to-relapse.
• The model was applied to data from 236 pediatric ALL patients treated at TTCRC, Kolkata.
• Extensive simulation studies were conducted to assess the model's effectiveness.

Main Results:

• Neutrophil count and platelet count were significantly associated with time-to-relapse, while white blood cell count was not.
• A lower dose of 6MP and a higher dose of MTx were associated with a lower relapse probability.
• Patients classified as 'high-risk' at presentation exhibited the lowest relapse probability.

Conclusions:

• The developed Bayesian joint model effectively assesses the impact of covariates and biomarkers on ALL progression and relapse.
• The model efficiently imputes missing longitudinal biomarker data.
• Findings provide insights into pediatric ALL relapse prediction and potential therapeutic strategies.