To evaluate the predictive performance of population pharmacokinetic models for piperacillin (PIP) available in the software MwPharm, TDMx and ID-ODs for initial dosing selection and therapeutic drug monitoring (TDM) purposes.
This is a prospective observational study in adult patients with severe infections receiving PIP treatment. Plasma concentrations were quantified by ultra-high performance liquid chromatography coupled to tandem mass spectrometry. The differences between predicted and observed PIP concentrations were evaluated with Bland-Altman plots; additionally, the relative and absolute bias and precision of the models were determined.
A total of 145 PIP plasma concentrations from 42 patients were analysed. For population prediction, MwPharm showed the best predictive performance with a mean relative difference of 34.68% (95% CI –197% to 266%) and a root mean square error (RMSE) of 60.42 µg/mL; meanwhile TDMx and ID-ODs under-predicted PIP concentrations. For individual prediction, the TDMx model was found to be the most precise with a mean relative difference of 7.61% (95% CI –57.63 to 72.86%), and RMSE of 17.86 µg/mL.
Current software for TDM is a valuable tool, but it may also include different population pharmacokinetic models in patients with severe infections, and should be evaluated before performing a model-based TDM in clinical practice. Considering the heterogeneous characteristics of patients with severe infections, this study demonstrates the need for therapy personalisation for PIP to improve pharmacokinetic/pharmacodynamic target attainment.
