The Prediction of Serum C-Reactive Protein Concentration Using Nonlinear Mixed-Effects Model

Abstract

Serum C-reactive protein (CRP) is a useful biomarker reflecting the efficacy of clinical treatments for infectious and autoimmune diseases. Accurate prediction of the serum CRP concentration of a patient through accurate initial clinical evaluation must be preceded to promptly cope with inflammatory diseases. In general, serum CRP concentration rises sharply right after hip surgery and then falls down at a certain period of time. Such patterns can be used as a meaningful indicator to estimate recovery tendency of individual patients. This study proposes a nonlinear mixed-effects (NME) model to describe nonlinear patterns of serum CRP concentration over time for patients suffering from hip arthroplasty through an observational study. The bi-exponential model with random effects is applied to predict temporal CRP concentrations in patients after hip surgery. Analytical results show that the proposed model accurately predicts serum CRP concentrations over time by effectively capturing individual variation in serum CRP concentrations through random effects. Based on the estimated model, we derive the distribution of normalized concentration times using the Monte Carlo (MC) simulation. The NME model will be expected to support future research on best practices for intraoperative and postoperative management of patients with hip surgery, based on various levels of predicted risks of infection.