Predictive Performance of Population Pharmacokinetic Models for Amikacin in Term Neonates.

Background and objective: Amikacin is preferred in treating Gram-negative infections in neonates and it has a narrow therapeutic window. The population pharmacokinetic modeling approach can aid in designing optimal dosage regimens for amikacin in neonates. In this study, we attempted to identify the suitable population pharmacokinetic model from the published reports for the study population from an Indian setting.

Methods: Published population pharmacokinetic studies for amikacin in neonates were identified. Data on structural models and typical pharmacokinetic parameters were extracted from the studies. For the clinical study, neonates who met the inclusion criteria were enrolled in the study from the NICU, Kasturba Medical College, Manipal, during Jan 2020 to March 2022. Drug concentrations were estimated, and demographic and clinical data were collected. Identified population pharmacokinetic models were used to predict the amikacin concentrations in neonates. Predicted concentrations were compared against the observed concentrations. Differences between predicted and observed concentrations were quantified using statistical measures. The population pharmacokinetic model, which was able to predict the data well, is considered a suitable model for the study population. Dosing regimens were suggested for neonates using the pharmacometric simulation approach generated by the selected model.

Results: A total of 43 plasma samples were collected from 31 neonates. Twelve population pharmacokinetic models were found for amikacin in neonates. The predictive performance of the 12 studies was performed using clinical data. A two-compartment model reported by Illamola et al. predicted the amikacin concentrations better than other models. Illamola et al. reported creatinine clearance and body weight as the significant covariates impacting the pharmacokinetic parameters of amikacin. This model was able to predict the clinical data with 29.97% and 0.686 of relative median absolute prediction error and relative root mean square error, respectively, which is the best among the published models. The Illamola et al. model was selected as the final model to perform pharmacometric simulations for the subjects with different combinations of creatinine clearance and body weight. Dosage regimens were designed to attain target therapeutic concentrations for the virtual subjects and a nomogram was developed.

Conclusions: The population pharmacokinetic model reported by the Illamola et al. model was selected as the final model to explain the clinical data with the lowest relative median absolute prediction error and relative root mean square error when compared with other models. An amikacin nomogram was developed for the neonates whose creatinine clearance and body weight ranged between 10 and 90 mL/min and between 2 and 4 kg, respectively. A developed nomogram can assist clinicians to design an optimal dosage regimen of amikacin for term neonates.