J.-L. Steimer, B. Fotteler, R. Gieschke, N. Buss, F. Hoffmann-La Roche
Ltd., Basel, Switzerland & Welwyn, UK
In studies in healthy volunteers, the protease inhibitor saquinavir (SQV) showed overproportional increase of AUC with dose, and dependency of exposure on external factors such as food. In a retrospective meta-analysis of dose-AUC data with mixed-effects modeling, the dependency of AUC (and therefore bioavailability) of SQV on dose and cofactors was further investigated. We report preliminary results on a data set consisting of 726 (dose-AUC) observations from 277 healthy subjects enrolled in 17 single dose, mostly crossover, experimental studies. SQV had been given intravenously (4 studies, 29 subjects) and orally (16 studies, 265 subjects), in doses ranging from 6 to 72 mg iv and from 75 to 1800 mg po. The oral formulations used were hard (172 subjects) and soft (120 subjects) gelatin capsules, and a suspension (17 subjects) in early trials. Coadministration of the protease inhibitor ritonavir (RTV) was investigated in one study (58 subjects). SQV plasma concentrations for AUC calculations were determined by different analytical techniques, depending on the study. The primary goal of the present model-building work is to prospectively predict exposure (AUC) to SQV under varying conditions of drug administration. The intermediate objectives addressed herein were: 1) to account for special features of the AUC vs dose relationship of SQV (over-proportional increase, food effect, interaction with RTV, etc…); 2) to estimate inter- and intra-individual variability in the AUC vs dose relationship and in the parameters that describe it.
We developed an integrated mathematical pharmacokinetic model for the first-pass loss and the disposition clearance of SQV. The model is partly physiologically-based and partly empirical. It involves the following sequential steps: dissolution, back-transport by P-glycoprotein, gut wall metabolism, liver metabolism. To characterize the dose-AUC relationship, multiple influential factors were considered including route of administration, formulation, co-administration with RTV, food,… Data evaluation proceeded stepwise as well for data compilation as for model-based analysis. At each step up to the current step #6, increase in model complexity was triggered through inclusion of additional studies into the data set. The data-fittings were achieved with NONMEM Version 4, programmed to return the population and individual (posterior) parameters and predictions. Graphical display of “by study” observations and predictions, as well as residuals plots were used as primary diagnostics. The model in its present version quantitatively reproduced the pattern of the data over a range of individual AUCs spanning five orders of magnitude, from 1.4 to 62620 ng/mL*h. Some parameters with physiological meaning were fixed at realistic values (e.g. 120 L/h for hepatic blood flow after food). Bioavailability of SQV was predicted to range from less than 1% (75mg dose, fasted) up to more than 80% (600mg dose, with RTV).
Based on the model, preliminary computer simulations were carried out in order to predict exposure (AUC) under various conditions (e.g. in paediatrics and for co-administration of other protease inhibitors). The work is currently being pursued with inclusion of multiple dose patient data. A further extension will be to take advantage of the model for dose recommendations for mono- and especially combination therapy through linkage of AUC to effect (surrogate) data, e.g. viral load (HIV RNA) in plasma.
Acknowledgment: The authors thank Dr. Ann Hsu (Abbott) for making the AUC data of SQV in the SQV+RTV study available to them.
Reference: PAGE 6 (1997) Abstr 594 [www.page-meeting.org/?abstract=594]
Poster: oral presentation