CMJ Kirkpatrick(1), SB Duffull(1), ML Barclay(2), N Patton(3), GA Moore(4), MP Doogue( 1)
(1)School of Pharmacy, University of Queensland, Brisbane, Australia; (2)Clinical Pharmacology, Christchurch Hospital, Christchurch, New Zealand; (3)Haematology, Christchurch Hospital, Christchurch, New Zealand; (4)Toxicology, Canterbury Health Laboratories, Christchurch, New Zealand
Objectives: To undertake a population analysis of high dose oral busulphan for bone marrow conditioning prior to transplantation in adults and children.
Methods: A population pharmacokinetic analysis of busulphan was performed using NONMEM (using FOCE with interaction) on a population of 24 patients [11 adults/13 children, 8F/16M]. Age ranged from 1 to 50 years. Blood samples were available from 3 occasions for most patients. Initial non-compartmental analysis by the investigators suggested a systematic change in AUC during the treatment regimen.
Results: The best base pharmacokinetic model was a one-compartment model with first-order absorption and elimination with between subject variability (BSV) on Ka, CL and V and between occasion variability on Ka, CL and V. The mean population parameters were similar to previous studies. The final model included weight (kg) as a covariate on Vd. For CL, however, two competing covariate models were identified, i) weight0.75 (allometric scaling) and ii) body surface area (BSA). The allometric scaling model had a slightly lower reduction in Obj (-3.5 units) compared with BSA, but both models provided a similar reduction in the unexplained BSV. To assess the best covariate model, 1000 non-parametric bootstrapped datasets were generated, and both competing covariate models fitted to the data. The difference in NONMEMs objective function, ∆Obj, between both models was computed. The density of the distribution of the ∆Obj <0 was 0.75, indicating that the allometric model was preferred (odds = 3). A visual inspection of a predictive check was undertaken to evaluate the full model, this was performed for the full data set and for children and adults separately.
Conclusions: The population analysis provided similar parameter estimates to previous population studies. A possible way to assess biologically plausible competing covariate models is presented.
Reference: PAGE 13 (2004) Abstr 539 [www.page-meeting.org/?abstract=539]
Poster: poster