Population pharmacokinetics of piperacillin-tazobactam in the plasma and cerebrospinal fluid of critically ill patients

Abstract

Ventriculitis in neurocritical care patients leads to significant morbidity and mortality. Antibiotic dose optimization targeting pharmacokinetic/pharmacodynamic (PK/PD) exposures associated with improved bacterial killing may improve therapeutic outcomes. We sought to develop and apply a population PK model in infected critically ill patients to determine optimal piperacillin-tazobactam (PTZ) dosing regimens to achieve target cerebrospinal fluid (CSF) exposures. Neurosurgical patients with external ventricular drains and receiving PTZ treatment were recruited and had plasma and CSF samples collected and assayed. A population PK model was developed using plasma and CSF piperacillin and tazobactam concentrations. Eight patients were recruited. Median age was 59 years, median weight was 70 kg, and five patients were female. The median creatinine clearance was 84 mL/min/1.73 m² (range 52-163). Substantial inter-individual PK variability was apparent, particularly in CSF. Piperacillin penetration into CSF had a median of 3.73% (range 0.73%-7.66%), and tazobactam CSF penetration was not predictable. Dosing recommendations to optimize CSF exposures for the treatment of ventriculitis were not possible due to substantial PK variability and very low drug penetration. High plasma PTZ exposures may not translate to effective exposures in CSF.

Keywords: antibiotics; cerebrospinal fluid; critical illness; pharmacokinetic/pharmacodynamic; piperacillin-tazobactam; ventriculitis; ventriculostomy-associated infection.

Fig 1

Schematic diagram of the final PK model for piperacillin after intravenous infusion. C_c, plasma concentration in central compartment; C_CSF, concentration in CSF; V₁, volume of central compartment; Cl, clearance from the central compartment; V₂, volume of peripheral compartment; Q, inter-compartment clearance between the central and peripheral compartment; Q₃, inter-compartment clearance between the central and CSF compartment; V₃, volume of CSF compartment.

Fig 2

Observed plasma and CSF piperacillin and tazobactam concentrations.

Fig 3

Goodness-of-fit plots for plasma piperacillin concentrations. (A) Observed vs population-predicted concentrations; (B) observed vs individual-predicted concentrations. The solid black line represents the line of identity, and the solid red line represents the spline line. (C) NPDE vs time; (D) NPDE vs population-predicted concentrations. The solid red line represents the spline line.

Fig 4

Goodness-of-fit plots for CSF piperacillin concentrations. (A) Observed vs population-predicted concentrations; (B) observed vs individual-predicted concentrations. The solid black line represents the line of identity, and the solid red line represents the spline line. (C) NPDE vs time; (D) NPDE vs population-predicted concentrations. The solid red line represents the spline line.

Fig 5

Goodness-of-fit plots for plasma tazobactam concentrations. (A) Observed vs population-predicted concentrations; (B) observed vs individual-predicted concentrations. The solid black line represents the line of identity, and the solid red line represents the spline line. (C) NPDE vs time; (D) NPDE vs population-predicted concentrations. The solid red line represents the spline line.

Fig 6

Prediction-corrected visual predictive check plots for (A) plasma concentrations and (B) CSF concentrations of piperacillin and (C) plasma concentrations of tazobactam. Symbols represent observations. Lines represent empirical percentiles (5th, 50th, and 95th percentile). Shaded area represents 90% prediction interval around each predicted percentile.