Comparison of Bayesian and First-Order Estimations of Vancomycin Area Under the Curve in Pediatrics Using One- and...

Current consensus guidelines recommend therapeutic drug monitoring of vancomycin to target a 24-h area under the curve (AUC₂₄) of 400-600 mg*hr/L. Recommended methods for estimation of AUC₂₄ include the use of first-order (linear) pharmacokinetic equations or Bayesian software. Limited data exist to compare these methods in pediatric patients. This study aimed to compare calculated AUC₂₄ using first-order equations with two drug concentrations at steady state to Bayesian one- and two-concentration estimations.

This was a single center retrospective review of hospitalized pediatric patients receiving intravenous vancomycin. Patients with two concentrations collected at steady-state within 96 h of vancomycin initiation were screened. Pharmacokinetic parameters and AUC₂₄ were estimated using first-order equations and InsightRx Bayesian software. Pearson's correlation and clinical agreement were used to compare methods. Bland-Altman plots were used to assess mean difference (MD) and 95% limits of agreement (LOA) to assess variability.

Overall, 196 patients (22 neonates, 17 infants, 114 children, and 43 adolescents) were included in the final analysis. Reasonable agreement was observed between linear and Bayesian two-concentration methods (84.2%; R² = 0.723) and between Bayesian two- and one-concentration methods (82.7%, R² = 0.839). Some variability was noted with 95% LOA of -123 to 178 (MD = 27 mg*hr/L) and -88 to 110 (MD = 11 mg*hr/L), for the respective comparisons. Lower agreement was noted between linear and Bayesian one-concentration methods (73.9%, R² = 0.544), and the greatest amount of variability was demonstrated with 95% LOA of -179 to 211 (MD = 16 mg*hr/L). In subgroup analysis, neonates demonstrated significantly lower clinical agreement (54.5%) and correlation (R² = 0.2986) between linear and Bayesian one-concentration methods.

Linear and Bayesian two-concentration methods demonstrated reasonable agreement with acceptable variability and may be considered comparable for estimating vancomycin AUC₂₄ in pediatric patients. Similarly, Bayesian two-concentration and one-concentration methods demonstrated reasonable agreement with acceptable variability, supporting the comparability of these methods to estimate AUC₂₄. As poor agreement was demonstrated between linear and Bayesian one-concentration methods in neonatal sub-analysis, additional investigation is warranted to determine the comparability of these methods in neonates.