A.H. Thomson, I. Sharkie, L.S. Murray, R.Farre & M.A. Mangues
Dept of Medicine and Therapeutics, Western Infirmary, Glasgow and Pharmacy Dept, Hospital S. Creu I S. Pau, Barcelona.
Vancomycin is a glycopeptide antibiotic which is active against Gram +ve organisms. As many of these organisms are becoming resistant to penicillins, the use of vancomycinhas increased in recent years. Vancomycin has a narrow therapeutic range and dosage regimens for neonates are not well established. In this study the pharmacokinetics of vancomycin were investigated in a group of neonates and the results were evaluated using data collected from a second group of neonates.
Vancomycin dosage histories, concentration measurements and clinical data were collected from 37 neonates from Glasgow who were receiving vancomycin as part of their therapy. Sample collection and dosage adjustment were performed according to clinical requirements. The neonates were aged 6-61 days (mean 20 days), weighed 0.78-3.56kg (mean 1.32kg) and had postconceptional ages ranging from 26-47 weeks (mean 31 weeks). Creatinine concentration ranged from 25-124 µmol/L (mean 58 µmol/L). Other clinical details recorded were: ventilation status; cardiac abnormalities; nutritional source and 5 minutes APGAR score. There were a total of 463 doses given and 180 concentrations measured of which 75 were peaks, 79 troughs and 26 were taken at other times. These data were analysed using the population pharmacokinetic parameter estimation program NONMEM (Sheiner & Beal, 1985).
Comparison of variance models indicated that the data were best described using additive intersubject and residual error models. The use of a two-compartment structural model offered no advantage over a one-compartment model. Examination of residual plots suggested that postnatal age, sex, cardiac abnormalities, ventilation status, nutritional source and APGAR score had no significant effects. A series of linear and nonlinear models including weight (wt), postconceptional age (PCA), and serum creatinine concentration were therefore examined. Volume was assumed to be related to weight; both linear and nonlinear relationships were investigated. The best models were as follows:
- Clearance (L/h) = 0.071 x wt x (creatinine/58)-0.80 Volume (L/kg) = 0.63
- Clearance (L/h) = 0.0022 x wt x PCA x (creatinine/58)-0.59 Volume (L/kg) = 0.62
There was little to choose between the fits obtained with these models although model 2 appeared to be marginally superior on the basis of variance estimates and the minimum objective value.
These results were evaluated with data collected in Barcelona from a separate group of 42 neonates. Steady state peak and trough concentrations were predicted using clearance and volume estimates derived from the two models then the predictions were compared with the measured concentrations by calculating the prediction error (predicted – observed concentration). The mean (SD) prediction errors for the peak and trough were -1.32 (11.1) mg/L and -1.96 (7.6) mg/L for model 1, and -4.32 (13.6) mg/L and -4.58 (10.2) mg/L for model 2. There was no significant bias for in the predictions for model 1, but model 2 significantly underestimated trough concentrations (unpaired t-test, p<0.01).
These evaluation results indicated that the model which did not take postconceptional age into account was superior. This model will be used to devise dosage guidelines.
Beal S.L. & Sheiner L.B. NONMEM User’s Guide, Version II, University of California, San Francisco, 1985.
Reference: PAGE 1 (1992) Abstr 895 [www.page-meeting.org/?abstract=895]
Poster: oral presentation