Modelling HIV-RNA viral load in vertically infected children

Area of Science:

• Virology
• Biostatistics
• Pediatric Infectious Diseases

Background:

• Human immunodeficiency virus (HIV) RNA viral load is a key marker for disease progression.
• Understanding viral load dynamics in vertically infected children is crucial for disease management.
• Assay detection limits can result in left-censored viral load measurements.

Purpose of the Study:

• To determine if complex statistical methods are necessary to model HIV RNA viral load patterns in children.
• To assess the impact of repeated measures and data censoring on viral load modeling.
• To compare different statistical approaches for analyzing longitudinal HIV RNA viral load data.

Main Methods:

• Utilized fractional polynomials for modeling viral load dynamics over age.
• Compared complex methods (EM algorithm, Gibbs sampler) with simpler alternatives.
• Investigated approaches for handling left-censored data, including cut-off values and mid-point imputation.
• Employed linear mixed-effects and ordinary least squares models.

Main Results:

• Fractional polynomials significantly outperformed conventional models for viral load dynamics.
• Accounting for repeated measures was necessary to improve statistical power for model selection.
• Simple methods for handling censored data, such as mid-point imputation, did not negatively impact model fit.
• Complex methodologies were not essential for identifying the optimal viral load model.

Conclusions:

• Complex statistical methods are not always necessary for modeling HIV RNA viral load in children.
• Fractional polynomials offer a superior approach to modeling viral load dynamics over age.
• Appropriate handling of repeated measures and censored data is important, but simpler imputation methods suffice.