Nonlinear hierarchical modeling of experimental infection data

Area of Science:

• Biostatistics
• Infectious Disease Modeling
• Veterinary Epidemiology

Background:

• Longitudinal experimental infection (EI) data analysis often relies on traditional methods like repeated measures analysis of variance (RM-ANOVA) and linear mixed models (LMM).
• These methods may not fully capture the complex biological dynamics of infection progression.
• There is a need for advanced statistical models to analyze EI data more effectively.

Purpose of the Study:

• To introduce and evaluate a nonlinear hierarchical model (NLHM) for analyzing longitudinal EI data.
• To demonstrate the NLHM's advantages over RM-ANOVA and LMM in characterizing infection dynamics.
• To assess the NLHM's performance in detecting differences in key infection features between populations.

Main Methods:

• Development of a nonlinear hierarchical model (NLHM).
• Application of the NLHM to EI data from equine arteritis virus studies.
• Comparative analysis using simulation studies against RM-ANOVA and LMM.
• Evaluation of model bias and statistical power for detecting differences in infection features.

Main Results:

• The NLHM enables comparison of biological properties such as peak intensity, duration, and time to peak of infection.
• Simulation studies showed the NLHM substantially reduces bias and improves power for detecting differences in infection response features.
• For instance, detecting a 20% difference in response duration, NLHM achieved 58% power compared to 11% (RM-ANOVA) and 12% (LMM) with n=15, while controlling Type I error rate.

Conclusions:

• The NLHM is a powerful and flexible tool for analyzing longitudinal EI data.
• It provides deeper biological insights into infection dynamics than conventional methods.
• The NLHM demonstrates superior performance in detecting subtle differences in infection characteristics between populations.