Additive quantile mixed effects modelling with application to longitudinal CD4 count data

Area of Science:

• Biostatistics
• Epidemiology
• Longitudinal Data Analysis

Background:

• Conventional regression models focus on conditional mean, potentially missing crucial effects.
• Mixed-effects models and computational advances enable practical quantile regression for longitudinal data.
• Additive mixed-effects models extend quantile regression for flexible nonparametric and parametric analyses.

Purpose of the Study:

• To apply additive quantile mixed models to analyze longitudinal CD4 counts in HIV-infected patients.
• To demonstrate the model's ability to capture robust nonlinear and linear effects across different conditional distributions.
• To identify factors influencing CD4 count progression in patients on Highly Active Antiretroviral Therapy.

Main Methods:

• Utilized additive quantile mixed models for longitudinal data analysis.
• Analyzed CD4 count data from HIV-infected patients in a South African follow-up study.
• Examined effects of time, baseline BMI, viral load, residence, and sexual partners on CD4 counts.

Main Results:

• Significant nonlinear effects of time and baseline Body Mass Index (BMI) on CD4 counts were observed across all fitted quantiles.
• Baseline viral load, place of residence, and number of sexual partners were identified as major significant factors influencing CD4 count progression.
• The additive quantile mixed model provided robust insights into factors affecting CD4 counts beyond the central tendency.

Conclusions:

• Additive quantile mixed models are effective for analyzing complex longitudinal data, revealing effects missed by mean-based methods.
• Time and baseline BMI exhibit significant nonlinear associations with CD4 counts in HIV patients.
• Understanding the impact of various covariates on CD4 counts is crucial for managing HIV treatment and patient outcomes.