M. Rosario1,4, C.A. Sharp2, D.I. Jodrell3, A.H. Thomson1
1Department of Medicine & Therapeutics, University of Glasgow, 2Stirling Royal Infirmary, 3ICRF Medical Oncology Unit, University of Edinburgh,4Programa ciencia, Praxis XXI Lisboa
The population pharmacokinetics of gentamicin were investigated in patients with cancer using NONMEM. Data were collected prospectively from 210 patients whose ages ranged from 15 to 81 years (median 50 years) and weights from 38 to 113 kg (median 66 kg). The data were split into two sets: a population data set comprising 140 patients and an evaluation set comprising 70 patients. Both monoexponential and biexponential disposition models were fitted to the data and the following covariates were examined for a possible influence on the pharmacokinetics of gentamicin: gender; weight; age; height; body surface area; lean body mass; ideal body weight; body mass index; creatinine concentration; creatinine clearance; urea; albumin; platelets; haemoglobin; white cell count; and pyrexia. Creatinine clearance (CrCl) was estimated using measured creatinine concentration and with low creatinine concentrations set to 60 umolL-1, 70 umolL-1 or 88.4 umolL-1. Interindividual variability was assumed to correspond to an exponential model and additive, exponential and combined (additive plus exponential) error models were used to describe residual error.
The data were best fitted with a biexponential disposition model with a combined residual error structure. Preliminary examination of posthoc clearance estimates versus clinical covariates indicated possible relationships between clearance and weight, ideal body weight, lean body mass height, gender, body surface area, creatinine concentration and creatinine clearance. These covariates were then included in the model in a stepwise fashion. Investigation of the optimal creatinine concentration found that setting concentrations < 60 umolL-1 to 60 umolL-1 or concentrations < 70 umolL-1 to 70 umolL-1 was superior to using either actual creatinine concentration or a minimum value of 88.4 umolL-1. The best covariate model related clearance to estimated creatinine clearance (minimum creatinine value 60 umolL-1) and volume of the central compartment to body surface area and albumin concentration, i.e.
CL (L h-1) = 1.02 * (1 + 0.037 * CrCl)
V1 (L) = 8.49 * body surface area * (albumin/34)-0.612
Intercompartmental clearance (L h-1) = 1.61
V2 (L) = 8.22
Coefficient of variation was 17.7% on clearance and 29.5% on intercompartmental clearance. Residual error expressed as a coefficient of variation was 30.6% at 1 mgL-1 and 16.7% at 8 mgL-1. The mean posthoc estimate of clearance was 4.3 L h-1 and volume of distribution (Vss) was 24.0 L (0.37 Lkg-1). The mean posthoc estimates of distribution and elimination half-life were 1.5 h and 4.5 h respectively.
The best model and the variance in parameter estimates were evaluated by simulating 1000 concentrations for each data point in the test data set. Ninety per cent of measured concentrations lay within the 95% confidence interval of the simulated concentrations. The mean percentage prediction error was -25%. These results confirm the wide interpatient variability in aminoglycoside pharmacokinetics and the need for dosage optimisation of this drug.
Reference: PAGE 6 (1997) Abstr 666 [www.page-meeting.org/?abstract=666]
Poster: poster