Bayesian Estimation of Tobramycin Exposure in Patients with Cystic Fibrosis.

Abstract

Fixed tobramycin (mg/kg) dosing is often inappropriate in patients with cystic fibrosis (CF), as pharmacokinetics are highly variable. The area under the concentration-time curve (AUC) is an exposure metric suited to monitoring in this population. Bayesian strategies to estimate AUC have been available for over 20 years but are not standard practice in the clinical setting. To assess their suitability for use in clinical practice, three AUC estimation methods using limited sampling were compared to measured true exposure by using intensive sampling tobramycin data. Adults prescribed once daily intravenous tobramycin had eight concentrations taken over 24 h. An estimate of true exposure within one dosing interval was calculated using the trapezoidal method and compared to three alternate estimates determined using (i) a two-sample log-linear regression (LLR) method (local hospital practice); (ii) a Bayesian estimate using one concentration (AUC1); and (iii) a Bayesian estimate using two concentrations (AUC2). Each method was evaluated against the true measured exposure by a Bland-Altman analysis. Twelve patients with a median (range) age and weight of 25 (18 to 36) years and 66.5 (51 to 76) kg, respectively, were recruited. There was good agreement between the true exposure and the three alternate estimates of AUC, with a mean AUC bias of <10 mg/liter · h in each case, i.e., -8.2 (LLR), 3.8 (AUC1), and 1.0 (AUC2). Bayesian analysis-based and LLR estimation methods of tobramycin AUC are equivalent to true exposure estimation. All three methods may be suitable for use in the clinical setting; however, a one-sample Bayesian method may be most useful in ambulatory patients for which coordinating blood samples is difficult. Suitably powered, randomized clinical trials are required to assess patient outcomes.