Objective: Our objective was to individualize tobramycin dosing regimens in neonates of various gestational ages with use of early therapeutic drug monitoring.
Methods: This study was performed in neonatal patients with suspected septicemia in the first week of life. All patients received tobramycin, 4 mg/kg per dose, as a 30-minute intravenous infusion, with a gestational age-related initial interval of 48 hours (<32 weeks), 36 hours (32-36 weeks), and 24 hours (> or =37 weeks). The target serum peak and trough serum concentrations were 5 to 10 mg/L and 0.5 mg/L, respectively. Serum trough samples and 1- and 6-hour samples were taken after the first dose. Tobramycin concentrations were used to obtain gestational age-dependent population models with nonparametric expectation maximization software. To investigate the effect of timing of sampling in a second group of patients, serum trough samples and 3- and 8-hour samples were taken after the first dose of tobramycin was administered. Serum trough concentrations were predicted by use of linear pharmacokinetics in both groups and by use of the population models with bayesian feedback of 1 or 2 serum concentrations in the second group. These predicted concentrations were compared with actual serum trough concentrations. The predictive performance of the 1- to 6-hour and 3- to 8-hour models and the population models were compared with a gestational age-related model without therapeutic drug monitoring.
Results: A total of 247 patients were analyzed: 206 with 1- to 6-hour serum samples and 41 with 3- to 8-hour serum samples. Peak serum concentrations were above 5 mg/L in 90.8% of cases, and trough serum concentrations were above 1 mg/L in 25.5% of cases. The 3- to 8-hour linear model had a bias of -0.31 mg/L and a precision of 0.48 mg/L, and it performed significantly better than the 1- to 6-hour model. The best nonparametric expectation maximization model had a bias of -0.11 mg/L and a precision of 0.45 mg/L. None of the models yielded a significant improvement of predictive performance over the model without therapeutic drug monitoring.
Conclusions: Routine early therapeutic drug monitoring does not improve the model-based prediction of initial tobramycin dosing intervals in neonates in the first week of life.