Concentration–response studies and modelling of the pharmacodynamics of linezolid: Staphylococcus aureus versus Enterococcus faecium

• A new PD metrics, the RBR, was introduced and derived.
• A versatile PK/PD modelling approach in freely available software ‘R’ was introduced.
• RBR PD analysis was successfully applied comparing LZD effect against 2 strains.
• The PK/PD relationship was precisely described by a flexible and robust PK/PD model.
• The model provides easy-to-interpret quantitative measures as decision-making tool.

For assessing antibiotic effects, the minimum inhibitory concentration (MIC) neglects that most antibiotics display different rates of bacterial killing. Time–kill curves, on the other hand, provide details on these killing rates but their interpretation is more complex and hardly standardised. The aim of the present study was to develop an analysis method to easily compare the pharmacodynamics of linezolid (LZD) against Staphylococcus aureus and Enterococcus faecium via in vitro time–kill curve experiments and to describe it by mathematical modelling. The effect of LZD against both organisms was investigated in a static in vitro infection model using 0.5–32.0 μg/mL LZD over 24 h. LZD concentrations were quantified by a validated HPLC assay. A modified sigmoidal maximum effect (Emax) pharmacokinetic/pharmacodynamic (PK/PD) model that accounts for time-dependent effects was developed in ‘R’. As a continuous, growth-control-normalised pharmacodynamic measure, the relative bacterial reduction (RBR) was introduced and derived. LZD was more effective against S. aureus than against E. faecium (Emax 1.8-fold higher) at a comparable potency (EC50, 3.02 μg/mL vs. 1.80 μg/mL). The time delay of the maximum effect was predominantly observed within 6 h of exposure. Model evaluation demonstrated its precision, robustness and predictive performance. In conclusion, the presented PK/PD analysis method provides quantitative measures (EC50, Emax) for the antibacterial drug effect as easy to interpret as point estimates, but more informative than the MIC since time- and concentration-dependent effects were considered. Application of the presented model developed as a flexible, robust tool in the free software ‘R’ appears promising.