Introduction
Piperacillin is a moderately bound, broad-spectrum beta-lactam antibiotic commonly prescribed in intensive care units (ICU) to treat severe infections. The pharmacokinetic/pharmacodynamic (PK/PD) parameter linked to piperacillin antibacterial activity is defined by the time (T) during which the unbound drug concentration (f ) remains above the minimum inhibitory concentration (MIC) of the pathogen of interest (f T > MIC) [1]. As a result, piperacillin unbound concentration is required to assess target attainment. This value is obtained with two methods; by either directly measuring the unbound concentration of piperacillin with a validated bioanalytical method; or by firstly measuring the total concentration to afterwards apply a theoretical unbound fraction value to this concentration to obtain the unbound concentration. The latter is often performed with a value of 70%, the typically assumed unbound fraction in healthy subjects [2-8]. However, some studies have used other values to reflect current findings of piperacillin unbound fractions in ICU patients, as it may vary considerably within this population [7-9].
In our previous work, we noticed that the application of an unbound fraction could alter the predictive performance of a population pharmacokinetic (popPK) model evaluated with an independent dataset [10]. In summary, the independent dataset contained total piperacillin concentrations, whereas the model in question, developed by Klastrup et al., used unbound piperacillin concentrations during the development phase [11]. When calculating the predictive performance of the model with both total and unbound concentrations – retrieved by applying a factor of 70% - the latter concluded in favor of the validity of the model, whereas the model was considered unfit for the studied population when using total concentrations. This indicates that a model may perhaps be inappropriately suitable or unsuitable depending on the unbound fraction applied on a dataset in this type of evaluation. No study has clearly shown whether variations in unbound fraction had an impact on dosing recommendations for piperacillin, nor how it impacted the validity of a popPK model. Additionally, Roberts et al. have stated that more modeling-based research is required to determine whether dosing adjustments would be warranted for drugs that are affected by altered protein binding [12]. The aim of this study was therefore to investigate these potential impacts through simulations by using a previously validated popPK model developed by Klastrup et al. [11].