Methods for Optimising Neonatal Antimicrobial Use: Time- and Concentration-Dependent Agents
Joseph F Standing (1), Jake Morris (2), Eva Germovsek (3), Irja Lutsar (4), Mario Cortina-Borja (5)
(1) Department of Pharmacy, Great Ormond Street Hospital, London, UK (2) Department of Statistical Science, University College London, UK (3) Faculty of Pharmacy, University of Ljubljana, Slovenia (4) Department of Clinical Microbiology, University of Tartu, Estonia (5) MRC Centre of Epidemiology for Child Health, UCL Institute of Child Health, London, UK.
Objectives: Pharmacokinetic studies in neonates are increasingly common; it is important their results are used rationally. This study arose from two questions on treating sepsis in neonates. Firstly (time-dependent): what infusion length maximises time above MIC (T>MIC) for meropenem? Secondly (concentration-dependent): what is the optimum dose of gentamicin? We aimed to answer the above questions and in doing so develop methods for finding optimal treatments.
Methods: A meropenem pharmacokinetic model was derived from premature neonates receiving infusions of 0.5 or 4h [1], and a gentamicin model from a freely available dataset [2]. For meropenem utility the aim was 100% T>MIC, and penalties for increasing infusion length were explored. For gentamicin utility, a model for clinical response was fitted to literature data [3] and target set to 100%; the penalty function was a model for uptake kinetics into renal cortical cells [4] and set to 0%. Uncertainty on benefit and risk functions was incorporated. For meropenem a fixed dose was used, and infusion length optimised; for gentamicin dose was optimised. EUCAST MIC distributions [5] for E. coli were used to assign values to simulated subjects. Optimal infusion length/dose was derived for a range of fixed MIC values.
Results: For meropenem pharmacokinetics a 1-compartment model provided best fit and optimal infusion length for a 20mg/kg dose with randomly assigned MIC was 0.99h, although this was sensitive to individuals with outlying MIC values. At fixed MIC values below 0.08mg/L optimal infusion time was 5min, rising steeply to 6.07 and 6.98h for the sensitivity and resistance breakpoints of 2 and 8mg/L. For gentamicin a 2-compartment model was chosen, and the optimal dose for randomly assigned MIC was 2.25mg/kg. Optimal doses for sensitivity and resistance breakpoints of 2 and 4mg/L were 4.54 and 6.74mg/kg respectively.
Conclusions: Optimising utility functions provides a more efficient method than exploring dose recommendations by simulation. Details of the approach used will be discussed to show how these methods can be applied.
References:
[1] Padari H et al, ESPID Abstract 2011
[2] Thomson AH et al, http://www.rfpk.washington.edu
[3] Moore RD et al Am J Med. 1984;77:657-62
[4] Giuliano RA Am J Kidney Dis. 1986;8:297-303
[5] www.eucast.org
