Objectives: Our objectives were: 1) to develop a mechanism-based pharmacodynamic (PD) model for the time course of total bacterial titer (CFU/mL), obtained in a 1-compartment in vitro infection model (IVIM), in which bacteria (as colony forming units, CFU) are exposed to dynamic concentrations of drug simulating the free-drug profiles predicted for selected candidate regimens in humans and, 2) to explore the PK/PD/Pharmacogenomic relationship observed in two strains of MRSA 3) to evaluate the potential benefit of front loading.
Methods: The IVIM studies were of 48 hour duration and tested the PD of daptomycin against two isolates of MRSA (USA 300, MIC = 0.5 mg/L, and MU50 MIC =2 mg/mL), studied at an initial inoculum of 108 CFU/mL. Regimens evaluated were traditional (4, 6, 8 mg/kg q24h) versus front loading (8 then 4 mg/kg at 24h, 12 then 6 mg/kg at 24h, 16 then 8 mg/kg at 24h. Samples were obtained at the following times: 0, 2, 4, 24, 26, 28, and 48h. All log10 (CFU/mL) data were simultaneously modeled using the MC-PEM algorithm, in parallelized SADAPT (and the SADAPT-Tran) 1. A log-normal likelihood distribution was assumed for all model parameter values; residual error models were additive in the log10 domain; the Beal M3 method was used for censored observations. The candidate structural PD models considered included models for bacterial growth, natural death and loss of bacteria via the flow of the media through the open system. Models with one, two, and three subpopulations of bacteria, differing (at least) in susceptibility to drug, were considered; drug effect were modeled as inhibition of growth or stimulation of bacterial killing. Model discrimination was based on Akaike’s Information Criterion2, the visual inspection of the goodness of fits, and improvements in the overall objective function.
Results: The mechanism-based PD model was able to fit all of the experimental data precisely. The final structural PD model included 2 bacterial sub-populations, differing by susceptibility to drug, capacity-limited replication, first order natural death, with the modeled drug effect as an increased rate constant for bacterial death. The interexperimental variability of the core parameter estimates from the population fits had a CV < 0.18. Model estimates precise and unbiased for both population and individual fits (individual overall R= 0.923, population R = 0.977.
Conclusions: The proposed model successfully described the data at all the regimens tested for daptomycin. Front loaded (first dose at two times the maintenance dose) regimens were most active for these isolates. This final model can be used to simulate responses to candidate dosing strategies, for a range of MIC values. To our knowledge this is the first mechanism based model developed for daptomycin.