Oliver Scherf-Clavel (1)
(1) Institute of Pharmacy and Food Chemistry, Julius-Maximilians-University D-97074 Würzburg, Germany
Objectives: Vancomycin is today a last-resort antibiotic in the infection with Methicilin Resistant Staphylococcus Aureus (MRSA) and other gram-positive bacteria. The narrow therapeutic index and high interindividual variability warrant therapeutic drug monitoring of vancomycin. There is increasing evidence that the ratio of area under the curve (AUC) and minimal inhibitory concentration (MIC) is a superior PK/PD-target compared to through concentration [1, 2]. However, the determination of the AUC is associated with practical problems. The withdrawal of multiple samples in order to calculate the AUC accurately is not achievable in clinical practice. Thus through concentrations are often used as a surrogate for AUC, although they might not correlate very well with the AUC [3]. We aimed at developing an easy to use tool, to assist the AUC/MIC guided dosing of vancomycin in critically ill, adult patients by using a limited number of PK samples. The software should feature:
- An easy to use and intuitive user interface (UI)
- The ability to import and export data
- A Module for manual and automatic dosage adaptation to reach the PK target
- The possibility to export the results of the dosage adaptation in the form of a report for the clinician
Methods: VancoSim was developed under R version 3.5.3 using ‘shiny’ [4] as a frontend. The pharmacokinetic model by Goti et al. [5] was chosen because of the recently demonstrated suitability [6] for this purpose. Stochastic simulations are used for a PK prediction without available TDM data, whereas Gibbs sampling is used to obtain Bayesian estimates of the individual pharmacokinetic parameter when TDM data is provided. Gibbs sampling is performed using the open software JAGS via the package ‘R2jags’ [7]. All other calculations are performed in R using the closed form solution of a two-compartment model with zero order infusion and first order elimination.
Results: The open source project VancoSim [8] offers options to perform necessary tasks of an AUC/MIC guided TDM while being open to modifications and further development. Prediction intervals are included to visualize the uncertainty of the estimated PK profiles. A simple algorithm for automatic dosage adaptation and a clinical report module are included.
Parameters for stochastic simulations, Gibbs sampling and error model parameters can be modified for experimental use. Import and export of NONMEM-like datasets in the form of Microsoft Excel files allows for an easy transfer of data to and from the software tool. Covariates used in the simulation according to the PK model are creatinine clearance, dialysis and body weight. As an experimental feature, time variant covariates were implemented in the tool in order to consider a change in renal function and patient condition.
The mode of the simulated posterior distribution of PK parameters was demonstrated to be equivalent to the result of commonly applied maximum a posteriori Bayesian (MAP) estimation. The results of Bayesian estimation can be reviewed and compared to the stochastic simulation results in order to assess the plausibility of the simulations.
Conclusions: By outsourcing the ‘Bayesian Forecast’ to JAGS, sampling from the posterior distributions is efficient and quick allowing for using the posterior distribution rather than only using the mode of the distribution. This approach enables the presentation of prediction intervals for the Bayesian estimate as well as for the stochastic simulation without using TDM data. The provided modules for simulation and clinical report enable the personalization of Vancomycin dosing based on patient covariates either alone or in combination with TDM samples.
References:
[1] Rybak MJ et al Pharmacotherapy (2009) 29:1275-1279 doi: 10.1592/phco.29.11.1275.
[2] Liu C et al Clin Infect Dis (2011) 52:e18-e55 doi: 10.1093/cid/ciq146.
[3] Hamberg AK and Keizer R. Ther Drug Monit (2017) 39:303 doi: 10.1097/FTD.0000000000000390.
[4] https://cran.r-project.org/web/packages/shiny/index.html
[5] Goti V et al Ther Drug Monit (2018) 40:212-221 doi: 10.1097/FTD.0000000000000490.
[6] Broeker A et al Clin Microbiol Infect (2019) 25:1286 e1-1286 e7 doi: 10.1016/j.cmi.2019.02.029.
[7] https://cran.r-project.org/web/packages/R2jags/index.html
[8] https://github.com/OlWaWue/VancoSim
Reference: PAGE () Abstr 9244 [www.page-meeting.org/?abstract=9244]
Poster: Methodology - Other topics