Sebastian G. Wicha (1), Alexander Solms (2), Wilhelm Huisinga (2) and Charlotte Kloft (1)
(1) Dept. of Clinical Pharmacy and Biochemistry, Freie Universitaet Berlin, Germany, (2) Institute of Mathematics, University of Potsdam, Germany
Objectives: In time-kill curve (TKC) modelling, precise estimation of PD parameters is crucial, e.g. to perform reliable simulations. Conventional designs (CD) for choosing drug concentrations to be evaluated in TKC studies are usually based on multiples of the minimal inhibitory concentration (MIC) in two-fold increments. The objective of the present study was to assess the performance of such a CD by a simulation-based study using a common TKC model [1], and to compare it with a design using D-optimal concentration tiers (OD). Further, we intended to propose an adaptive optimal design (aOD) algorithm as a potential link to experimental practice.
Methods: Modelling and simulations were performed in ‘R’[2]. 500 parameter sets of the TKC model were randomly sampled. Time points were 0, 2, 4, 8 and 24 h in all designs. The CD included 9 antibiotic concentration tiers of 0-16x MIC. In the ODs, 0x and 16x MIC was fixed whilst the other 7 concentration tiers were D-optimal. For the aODs, in the first stage, a reduced CD (0, 0.5, 1, 2 and 16x MIC) was used. Upon the estimates of the reduced CD, two D-optimal concentration tiers were added to the reduced CD to obtain the aOD. Based on 500 simulations each, distributions of relative error (RE; 2.5th-97.5th percentile) for bias and relative standard error (RSE; 95th percentile) for precision were used to compare the designs.
Results: All designs allowed for precise estimation of the bacterium-specific parameters of the TKC model. For flat concentration-effect relationships (Hill factor; H < 4), the drug-specific parameters (EC50 and H) were accurately and precisely estimated for both CD and OD (RE [-13.8%; 18.3%], RSE < 12.3%). For steep concentration-effect relationships (4 < H < 10), the ODs were superior to the CD (CD: RE [-35.1%; 74.3%], RSE < 133% vs. OD: RE [-12.9%; 16.7%], RSE < 10.8%). An OD with a minimal number of concentration tiers is currently assessed. The aODs performed comparably to the CD for H < 4 (aODs: RE [-14.2%; 16.4%] , RSE < 12.7%) and also superior for 4 > H > 10 (aCDs: RE [-22.0%; 31.2%], RSE < 17.8%) by requiring only 7 instead of 9 concentration tiers.
Conclusion: For antibiotics with a considerably steep concentration-effect relationship, individual adaption of the concentration tiers in TKC studies by optimal design techniques might be beneficial for accuracy and precision of PD parameter estimates. Further research is necessary to confirm this in silico approach in experimental settings.
References:
[1] Treyaprasert W, Schmidt S, Rand KH, et al. Pharmacokinetic/pharmacodynamic modeling of in vitro activity of azithromycin against four different bacterial strains. Int. J. Antimicrob. Agents, (2007) 29(3): 263–70.
[2] R Core Team. R. A language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
Reference: PAGE 23 () Abstr 3202 [www.page-meeting.org/?abstract=3202]
Poster: Methodology - Study Design