Population pharmacokinetics of gentamicin in paediatrics: quantifying covariate-parameter relationship using the full random effects modelling (FREM) approach

Objectives: To (i) evaluate the suitability of a previous published population pharmacokinetic (PK) model of gentamicin in paediatric oncology patients1 in predicting drug exposure in non-oncology paediatric patients and (ii) investigate the relationship between PK parameters and covariates in both oncology and non-oncology paediatric patients using the full random-effect (FREM) covariate modelling approach.2

Methods: Data were collected retrospectively from non-oncology paediatric patients who had pharmacokinetic and clinical information available. The predictive performance of the population PK model1 in a non-oncology cohort was assessed visually (visual predictive check plots (VPC)) and numerically (root mean squared error (RMSE %) and mean relative prediction error (MRPE %)). The previous oncology and current non-oncology datasets were then combined and a FREM covariate model approach was used to characterise parameter-covariate relationships, to account better for correlated covariates. Model assessment and new model development was undertaken using NONMEM®, with PsN (version 4.7.0). Covariates evaluated in this study, were body weight, fat-free mass (FFM), post-menstrual age (PMA), cisplatin and carboplatin usage within the prior 6 months, serum creatinine and creatinine clearance (calculated using the Flanders metadata equation3, which is an analogous of the Schwartz formula4) on clearance (CL) and apparent volume of distribution of the central compartment (V1) and peripheral compartment (V2). A clinically relevant covariate effect was defined as +/-20% differences in parameter values for the 95th and 5th percentile of the covariate, respectively.

Results: Data from 115 non-oncology patients (median body weight of 12.8 kg and median age of 1.85 years) were used. This data was then combined with data from 475 oncology patients for model development based on the FREM approach. Patients in the combined dataset had a median body weight of 18.4 kg and median age of 4.9 years and typically received a 30-minute gentamicin infusion of 7.3 mg/kg every 24 hours. The previous gentamicin population PK model under-predicted observed gentamicin concentrations in the non-oncology paediatrics, particular the patients younger than 2 years old. The model displayed some bias (MRPE (%) of 6.40) and imprecision (RMSE (%) of 86.3). Based on the FREM approach body weight, PMA and creatinine clearance were correlated with CL with coefficients of 0.014, 0.001 and 0.006, respectively. Therefore, a patient with a body weight of 35 kg (10.4 kg higher than the population mean of 24.6 kg) will have a CL that is 14.6% (i.e 10.4 * 0.014) higher than the typical patient; a patient with a PMA of 400 weeks (38.1 weeks higher than the population mean of 361.9 weeks) will have a CL that is 3.8% higher than the typical patient; and a patient with a creatinine clearance of 110 mL/min/1.73m2 (7.6 mL/min/1.73m2 higher than the typical value of 102.4 mL/min/1.73m2) will have a CL that is 4.6% higher than the typical patient. Body weight, PMA and creatinine clearance influence V1 (coefficients of 0.014, 0.001 and 0.006, respectively). FFM, creatinine clearance and usage of cisplatin and carboplatin influenced V2 (coefficients of 0.04, 0.014 and 0.273, respectively).

Conclusion: A previously developed population PK model of gentamicin for paediatric oncology patients does not predict gentamicin exposure in non-oncology paediatric patients well. Using the FREM approach, results to date show that body weight, PMA and creatinine clearance have a clinically significant effect on CL and V1. Unlike the previous developed model, serum creatinine alone did not have a significant covariate effect on any of the PK parameters and usage of cisplatin and carboplatin was found to have an effect on V2. This abstract will present a ‘stuck-in modelling’ project and any input from the audience would be appreciated.


  1. Llanos-Paez CC et al. (2017) Antimicrob Agents Chemother 61(8).
  2. Karlsson MO et al. (2012) In: PAGE
  3. Pottel H et al. (2010) Pediatr Nephrol 25(5):927-34.
  4. Schwartz GJ et al. (2009) J Am Soc Nephrol 20(3):629-37.