Bayesian Optimisation of Tobramycin Dosing in Paediatric Patients with Cystic Fibrosis

Abstract

ABSTRACTBackgroundIn children with cystic fibrosis (CF), tobramycin concentrations are monitored via the area under the plasma concentration time curve (AUC) approach. Currently, a minimum of 2 plasma concentrations of tobramycin within one dosing interval are required to estimate the AUC, which is costly, painful and blood collection is often difficult to coordinate.AimTo evaluate the accuracy of tobramycin AUC estimated using 1 plasma concentration compared to AUC based on 2 plasma concentrations calculated using Bayesian software.MethodData were collected from paediatric patients with CF prescribed once daily intravenous tobramycin. Each patient had 2 blood samples taken within a dosing interval (according to usual practice at the hospital) on 2 separate occasions during their admission. Data on tobramycin dosing and concentration, and patient demographics from the first occasion were entered into the Bayesian software, TCIWorks, to establish an individual patient's pharmacokinetic model. Data from the second occasion were then also entered into TCIWorks to estimate 2 AUCs – from 1 (AUC1) and 2 (AUC2) plasma concentrations. The accuracy of each patient's AUCs (AUC1 and AUC2) were then evaluated using a Bland‐Altman analysis to describe bias relative to the mean of the two values. To show the benefits of good quality data when using TCIWorks, data were obtained and analysed from 2 patient groups. For group A patients, accurate sampling and administration times of the tobramycin infusion were obtained prospectively. Group B consisted of patients from group A plus additional patients who had data collected from a retrospective medical chart review.ResultsData were collected for 30 paediatric patients with CF − 14 in group A and 16 additional patients in group B. In group A, there was no significant bias relative to the mean of the values for AUC1 and AUC2 (mean bias 1; 95%CI −4.9–6.9). A similar result was seen in group B (mean bias −0.7; 95%CI −10–8.6) indicating that the 2 estimates of tobramycin AUC (using 1 and 2 blood samples) are not significantly different. Conclusion: Once an individual paediatric patient's tobramycin pharmacokinetic model is determined with two blood samples using TCIWorks, only one blood sample is required to estimate an AUC for tobramycin.