David W. Uster (1), Sophie L. Stocker (2,3), Jane Carland (2,3), Richard Day (2,3,4), Sebastian G. Wicha (1)
(1) Dept. of Clinical Pharmacy, Institute of Pharmacy, University of Hamburg, Hamburg, German (2) Dept. of Clinical Pharmacology & Toxicology, St Vincent’s Hospital, Sydney, Australia (3) St Vincent's Clinical School, University of New South Wales, Sydney, Australia (4) School of Medical Sciences, University of New South Wales, Sydney, Australia
Objectives: Currently, the recommendation for therapeutic drug monitoring (TDM) of vancomycin is shifting towards AUC-guided dosing through model-based precision dosing (MIPD) [1]. Selecting an inadequate model from the ‘wrong’ population leads to incorrect prediction of vancomycin exposure and therefore potentially inappropriate dose recommendations [2]. The aim of the present study was to derive and evaluate a model averaging (MAA) and a model selection algorithm (MSA), which automates the model selection process and thus eases the clinical application of MIPD.
Methods: The MAA for Bayesian forecasting comprised two essential steps: First, the individual pharmacokinetic (PK) parameters were estimated with a set of population pharmacokinetic models on observed data, similar to the approach proposed by Aoki et al. [3]. Subsequently, the forecasted, i.e. unobserved concentration-time, data were weighted by the maximum likelihood obtained through the individual objective function value in each model. The MAA averaged the predicted vancomycin concentrations at each forecasted time point using the set of population PK models jointly with the data-derived weighting scheme. In contrast, the MSA algorithm selected the best fitting model (based on Bayesian estimates) and the competing models were discarded.
The exactness of these algorithms was evaluated (i) in simulation studies conducted with six population PK models developed in diverse populations (n=6000 people; extremely obese, critically ill, hospitalized, septic and trauma) [4]–[9] and (ii) in a clinical dataset containing routine vancomycin TDM data of 264 hospitalized and critically ill patients.
The predictive performance of the MAA, the MSA and each individual population PK model was compared on prospectively measured vancomycin concentrations. Relative bias (rBias) and relative root mean square error (rRMSE) were used to determine accuracy and precision of the forecasted concentrations, respectively.
Results: In the simulation study, the MSA detected the correct model (i.e. the model used for simulation) in nearly every second patient with an increasing pattern when samples from more dosing intervals were supplied. Although the individual PK parameter of ca. 50% of the virtual patients were estimated by the population models which were different to the underlying simulation model, the MSA led to similar (i.e. difference rBias and rRMSE < 10%) or more accurate metrics than if the true model was used (2/6 of the simulations).
Overall, the accuracy and precision of the MAA and MSA were in good alignment throughout the 6000 simulated patients with slightly higher imprecision and inaccuracy in the MAA approach (mean difference rBias: 2% and rRMSE 3%). Despite this, the predictive performance of the MAA was better than 66% of the single model approaches and best in the critically ill population simulated with the model by Mangin et al.
Using the clinical dataset, the predictive performance of the MSA was similar to the best single model, but was slightly inferior to the MAA. While the predictive performance of the single models varied substantially (rBias: -16 – 25%, rRMSE: 28 – 62%), MAA using these models simultaneously resulted in unbiased and more precise predictions regarding Bayesian forecasting with samples from both one (rBias: 0%, rRMSE: 30%) and two dosing intervals (rBias 0%, rRMSE: 31%). The MAA performed even better than the model of Goti et al. [10] (rBias: 8%, rRMSE: 34% and 36%) that was recently evaluated best in vancomycin Bayesian forecasting [2].
Conclusions: In conclusion, MAA overcomes one of the major difficulties associated with the implementation of MIPD, which is patient-individual selection of the adequate population pharmacokinetic model. MSA and MAA performed similarly in the simulations, with a slight superiority of the MAA in the clinical dataset and both being better than the single model approach. Especially if the underlying population of the individual patient remains unclear or the patient displays atypical pharmacokinetics, MAA can provide a more reliable Bayesian forecast as compared to using a single model. Therefore, implementation of the MAA into MIPD software (e.g. TDMx [11]) is warranted.
References:
[1] M. Rybak et al., “Therapeutic monitoring of vancomycin: A revised consensus guideline and review of the American Society of Health-System Pharmacists, the Infectious Diseases Society of America, the Pediatric Infectious Diseases Society and the Society of Infectious Diseases Pharmacists,” ASHP, 2019. [Online]. Available: https://www.ashp.org/-/media/assets/policy-guidelines/docs/draft-guidelines/draft-guidelines-ASHP-IDSA-PIDS-SIDP-therapeutic-vancomycin.ashx. [Accessed: 17-Feb-2020].
[2] A. Broeker et al., “Towards precision dosing of vancomycin: a systematic evaluation of pharmacometric models for Bayesian forecasting,” Clin. Microbiol. Infect., vol. 25, no. 10, pp. 1286.e1-1286.e7, Oct. 2019.
[3] Y. Aoki, D. Röshammar, B. Hamrén, and A. C. Hooker, “Model selection and averaging of nonlinear mixed-effect models for robust phase III dose selection,” J. Pharmacokinet. Pharmacodyn., vol. 44, no. 6, pp. 581–597, Dec. 2017.
[4] E. D. Adane, M. Herald, and F. Koura, “Pharmacokinetics of Vancomycin in Extremely Obese Patients with Suspected or Confirmed Staphylococcus aureus Infections,” Pharmacother. J. Hum. Pharmacol. Drug Ther., vol. 35, no. 2, pp. 127–139, Feb. 2015.
[5] O. Mangin, S. Urien, J.-L. Mainardi, J.-Y. Fagon, and C. Faisy, “Vancomycin Pharmacokinetic and Pharmacodynamic Models for Critically Ill Patients with Post-Sternotomy Mediastinitis,” Clin. Pharmacokinet., vol. 53, no. 9, pp. 849–861, Sep. 2014.
[6] S. E. Medellín-Garibay, B. Ortiz-Martín, A. Rueda-Naharro, B. García, S. Romano-Moreno, and E. Barcia, “Pharmacokinetics of vancomycin and dosing recommendations for trauma patients,” J. Antimicrob. Chemother., vol. 71, no. 2, pp. 471–479, Feb. 2016.
[7] N. Revilla, A. Martín-Suárez, M. P. Pérez, F. M. González, and M. del M. Fernández de Gatta, “Vancomycin dosing assessment in intensive care unit patients based on a population pharmacokinetic/pharmacodynamic simulation,” Br. J. Clin. Pharmacol., vol. 70, no. 2, pp. 201–212, Aug. 2010.
[8] J. A. Roberts, F. S. Taccone, A. A. Udy, J.-L. Vincent, F. Jacobs, and J. Lipman, “Vancomycin Dosing in Critically Ill Patients: Robust Methods for Improved Continuous-Infusion Regimens,” Antimicrob. Agents Chemother., vol. 55, no. 6, pp. 2704–2709, Jun. 2011.
[9] A. H. Thomson, C. E. Staatz, C. M. Tobin, M. Gall, and A. M. Lovering, “Development and evaluation of vancomycin dosage guidelines designed to achieve new target concentrations,” J. Antimicrob. Chemother., vol. 63, no. 5, pp. 1050–1057, Mar. 2009.
[10] V. Goti, A. Chaturvedula, M. J. Fossler, S. Mok, and J. T. Jacob, “Hospitalized Patients With and Without Hemodialysis Have Markedly Different Vancomycin Pharmacokinetics,” Ther. Drug Monit., vol. 40, no. 2, pp. 212–221, Apr. 2018.
[11] S. G. Wicha, M. G. Kees, A. Solms, I. K. Minichmayr, A. Kratzer, and C. Kloft, “TDMx: A novel web-based open-access support tool for optimising antimicrobial dosing regimens in clinical routine,” Int. J. Antimicrob. Agents, vol. 45, no. 4, pp. 442–444, Apr. 2015.
Reference: PAGE () Abstr 9305 [www.page-meeting.org/?abstract=9305]
Poster: Oral: Methodology - New Modelling Approaches