Pillai Goonaseelan (Colin)1,2, Gieschke Ronald1,3, Goggin Timothy1,2, Steimer Jean-Louis1,3
1Modeling and Simulation Team, 2Clinical Pharmacology and 3Biostatistics, Pharma Development, F. Hoffmann-La Roche, Basel, Switzerland
Introduction & Objectives Ibandronate is a high potency amino-bisphosphonate in Phase III development for osteoporosis. A PK-PD model was developed with a view to explore clinical trial design features. A particular challenge was to simultaneously model rapid PK (hours) and slow PD (weeks) changes.
Methods
Data & Study Design
The model was developed using existing data (Study MF9853). Japanese post-menopausal women with osteopenia (n=50) received 2 administrations of ibandronate approximately 3 months apart. Plasma was collected frequently until 48.0 hours after the first and second dose (9 and 5 samples respectively). On both occasions, urine was collected for 48 hours post-dose. Subjects received either placebo, 0.5mg, 1.0mg or 2.0mg intravenous ibandronate. Urinary CrosslapsR (uCTX – biomarker of osteoclast activity) was measured on Day 1, 4, 8, 15, 29 and 57 after the first administration and on Day 1, 2, 3, 4, 8, 15, 29, 57 and 92 after the second administration.
Subjects did not receive any supplemental background therapy (e.g. Calcium/Vitamin D).
Model
A 4 compartment PK model with an indirect response model attached to the 4th compartment (see Figure) was implemented in NONMEM using ADVAN 9. The 4th compartment was introduced to allow a link to the PD that was not identifiable from the PK alone. Ibandronate was assumed to inhibit osteoclast activity and hence rate of production of the biomarker. Time dependence in Rin was implemented by allowing this parameter to reach a target value (Rtar) at a rate governed by a first order rate constant Kqq.
Results & Conclusions The complex ‘physiological’ PK-PD model was able to adequately describe the pharmacokinetics of Ibandronate and the time course of the urinary CrosslapsR after intravenous ibandronate treatment in this study. A major disadvantage of the model however was the long computer run-times (~ 5 days on a Pentium 800mHz PC). Simulated data from this model was used to evaluate the performance of the simpler K-PD Model (Jacqmin et al PAGE 2001) with respect to schedule dependency. The K-PD model (Gieschke et al PAGE 2001) was used in subsequent simulations.
Reference: PAGE 10 (2001) Abstr 204 [www.page-meeting.org/?abstract=204]
Poster: poster