Modeling C-reactive protein kinetic profiles for use as a clinical prediction tool in patients with Staphylococcus aureus bacteremia

Abstract

Aim: We hypothesized that C-reactive protein (CRP) kinetics can be accurately modeled and might have clinical utility in a cohort of patients with Staphylococcus aureus bacteremia.

Materials & methods: We constructed and validated a nonlinear mixed effects model using CRP values obtained during the first week of illness.

Results: Hematological malignancy, prosthetic heart valves and metastatic seeding were identified as major covariates that influenced CRP kinetics. When considering the presence of metastatic infection as an 'unknown', the model could predict its presence through analysis of the observed CRP profile with an Area-under-the-Receiver-Operator-Characteristic curve of 0.81, indicating some diagnostic accuracy.

Conclusion: We conclude that early CRP kinetics can be accurately modeled and can help identify patients with metastatic seeding in S. aureus bacteremia. Further validation is required.

Keywords: C-reactive protein; Staphylococcus aureus bacteremia; biomarkers; nonlinear mixed effects modeling.