Zhigang Wang (1), Chenyan Zhao (1), S. van den Berg (2), Robin Otto (2), Johan Mouton (2), Elisabet I. Nielsen (1), Lena E. Friberg (1)
(1) Department of Pharmaceutical Biosciences, Uppsala University, Sweden (2) Department of Medical Microbiology and Infectious Diseases, Erasmus Medical Center, Rotterdam, The Netherlands
Introduction: Pharmacokinetic-Pharmacodynamic (PKPD) models built on preclinical data can be a useful tool for efficient exploration of antibiotic effects. Polymyxin B (PMB) is an antibiotic mainly used for the treatment of infections caused by resistant Gram-negative bacteria and is regarded as an antibiotic of last resort.
Aim: The aim of this work was to explore if an in silico semi-mechanistic PKPD model for PMB and Klebsiella pneumoniae (K. pneumoniae) built on in vitro data from static concentrations can be extrapolated to predict in vivo data in mice by incorporating an in vivo PK model.
Methods: Plasma concentrations were sampled 0-8 h post-dosing from thigh infected mice models with PMB (0.5-64 mg/kg) administered subcutaneously. A PK model was developed and applied to predict the unbound PMB concentration-time profiles (fu=0.086) following the dose regimens of 4-32 mg/kg q6h and 8-64 mg/kg q12h in five K. pneumoniae isolates (MIC 0.5-2 mg/L) in mice. Thighs were homogenized and plated at 24 h post-dosing, colony-forming units (CFU) were counted after overnight incubation. Growth controls in untreated mice were also included. An in vitro PD model developed from a resistant K. pneumoniae strain (MIC 16 mg/L) linked with the developed in vivo PK model was applied to predict in vivo observations. Bacteria growth rate constant (kgrowth) and maximal bacterial amount (Bmax) were re-estimated based on CFUs observed in growth controls to adjust to in vivo conditions. Slope, describing the PMB killing effect, was first scaled by MIC and then allowed to be re-estimated to explain differences between the in vivo and in vitro systems. Similarly, the hill factor γ, describing the PMB killing effect, was first fixed to in vitro value and then allowed to be re-estimated. CFU counts predicted by the PKPD model were compared to the observed.
Results: Overall 139 PMB concentrations were available, and well described by a PK model comprising of a saturable absorption pathway with the maximum absorption rate constant (Amax) decreasing with higher doses, and parallel linear and capacity-limited elimination. In total, 84 in vivo CFU counts were available, including 20 from growth control experiments and 64 from experiments following various PMB exposures. The estimated in vivo kgrowth was 51% lower than the earlier estimated in vitro value (0.676 vs. 1.37 h-1). Bmax was estimated to be 9.34 log10 CFU/thigh. The model predicts the PD data reasonably well when the PMB killing effect (kdrug) is scaled by MIC, an in vitro measurement of strain susceptibility. Model predictions were further improved (dOFV=-118.5, df=3) after re-estimation of both Slope and γ of PMB killing effect (kdrug) compared to scaling by MIC.
Conclusions: The PK of PMB in mice was well described by a PK model comprising of a saturable absorption pathway and parallel linear and capacity-limited elimination. The PKPD model built on in vitro data could predict the effect data of PMB against in vivo K. pneumoniae infectious model data reasonably well after re-estimation of kgrowth and Bmax. The model fit improved further when Slope and γ were allowed to be re-estimated. The predictive performance of the in vitro PKPD model will be further tested in combination with rifampicin and minocycline in in vivo models in the future.
References:
[1] Nielsen EI and Friberg LE. Pharmacol Rev (2013) 65: 1053-90.
[2] Zhao C et al PAGE 27 (2018) Abstr 8745 [www.page-meeting.org/?abstract=8745
[3] Landersdorfer CB et all. Antimicrob. Chemother (2017) 73: 462–468.
Reference: PAGE 28 (2019) Abstr 9187 [www.page-meeting.org/?abstract=9187]
Poster: Drug/Disease Modelling - Infection