Predicting Paediatric Tobramycin Pharmacokinetics with Five Different Methods
JF Standing (1), EG Greening (2), V Holden (2), S Picton (2), N Young (2), H Chrystyn (3), MO Karlsson (1).
1. Uppsala Universitet, Uppsala, Sweden. 2. St James’s University Hospital, Leeds, UK. 3. University of Huddersfield, Huddersfield, UK.
Objectives: An investigation into using adult data to predict paediatric tobramycin pharmacokinetics was undertaken. Tobramycin is a narrow therapeutic index aminoglycoside derived from Streptomyces tenebrarius, which inhibits bacterial protein synthesis. This study had two aims: to evaluate using adult data in modelling paediatric tobramycin pharmacokinetics; and to choose a model for optimising tobramycin dosing for children with febrile neutropaenia.
Methods: Following ethical approval, children with febrile neutropaenia were prospectively recruited to a dose escalation study for once-daily tobramycin, and subsequent data were collected from therapeutic drug monitoring (TDM). Three data sets were used: the first index set from 112 children (age 1-16yrs, CrCl 16-173mL/min) from the escalation study (5-13mg/kg od) and cohort undergoing TDM; the second index set from 97 adults[1] (age 16-85yrs, CrCl 10-166mL/min, dose: 20-140mg tds); and the third, a test dataset, of 54 paediatric patients (age 1-12yrs, CrCl 29-101mL/min) undergoing TDM. In children CrCl was estimated using a physiological method[2], in adults using Cockroft-Gault.
A 2-compartment model with BOV on bioavailability was implemented in NONMEMVI (FOCEI). Linear scaling of CL with CrCl was added a priori, as was linear scaling of V1 and V2 with weight, and non-linear (weight^0.75) scaling of Q[2]. The model was applied to the adult index set, the paediatric index set, pooled adult and paediatric index sets, the paediatric index set analysed using NWPRIOR, and the paediatric index set analysed using TNPRIOR (in the 2 latter cases prior distributions were from the adults). Validation was performed using the test paediatric data.
Results: The NONMEM objective function values (OFV) for the test data estimated from the 5 methods were 317, 296, 304, 312, and 298 respectively, indicating the paediatric only model performed best. This model gave the least biased patient-averaged population prediction error (95%CI) of 9.2%(-5.2%, 23.6%). The paediatric index and test datasets were merged, and the model re-run: CL standardised to a CrCl of 80mL/min was 3.12L/hr and V1 was 17.2L/70kg.
To investigate at what point adding the adult data would become useful, the size of the paediatric index dataset was reduced to 56, 28 and 14 patients. These were analysed alone and pooled with the adult index set and validated against the test data. Paediatric minus pooled adult/paediatric test OFV was -13.1, +0.8 and +19.2 for 56, 28 and 14 children respectively.
Conclusions: A model has been developed that adequately describes paediatric tobramycin pharmacokinetics, and can be used for dose optimisation. Adding adult data to the full paediatric index data marginally worsened predictive performance, but at some point around 28 paediatric patients, it was improved. Investigation into the use of priors with fewer paediatric patients is planned.
References:
[1] Aarons L, Vozeh S Wenk M, Weiss P, Follath F. British Journal of Clinical Pharmacology, 1989;28:305-14. Raw data from: http://www.rfpk.washington.edu/
[2] Anderson BJ & Holford NHG. Annual Review of Pharmacology & Toxicology, 2008;48:12.1-12.30.
