Semi-mechanistic PK/PD Model of the Effect of Odanacatib, a Cathepsin K Inhibitor, on Bone Turnover to Characterize Lumbar Spine and Distal Forearm Bone Mineral Density in A Phase IIb Study of Postmenopausal Women
Stefan Zajic, Julie A. Stone, David Jaworowicz, Albert Leung, Le Thi Duong, Julie Passarell, Jill Fiedler-Kelly, Dosinda Cohn, Nadia Verbruggen, and S. Aubrey Stoch
Merck Research Laboratories, Cognigen Corporation
Objectives: Odanacatib (ODN, MK-0822), a potent oral inhibitor of cathepsin K, is being developed to treat osteoporosis. A semi-mechanistic model of bone turnover was developed to account for responses in cortical and trabecular bone, as described by creatinine adjusted urinary aminoterminal crosslinked telopeptides of Type I collagen (uNTx), a bone resorption biomarker, and lumbar spine and distal forearm bone mineral density (lsBMD, dfBMD) data from a Phase IIb study.
Methods: Data from 391 postmenopausal women (PMW) receiving placebo or 3 to 50 mg weekly ODN for up to 2 years were used. Patients who completed 2 years of treatment were re-randomized to placebo or 50 mg weekly ODN and followed for an additional year. ODN concentration, biomarker, and BMD data were collected. A population PK model was used to estimate individual exposures and an indirect response model to characterize the timecourses of lsBMD and dfBMD as functions of bone formation and resorption rates. The PK/PD model describes the action of ODN through an inhibitory sigmoid Emax function applied to the bone resorption rate and the release rate of uNTx, which is a function of resorption. Transient elevation of bone resorption biomarkers after cessation of treatment is described by incorporating active and inactive osteoclasts as system variables and including an inhibitory sigmoid Emax function describing ODN inhibition of osteoclast apoptosis rate to reflect an increase in osteoclast numbers during therapy. Effects on bone formation from treatment with placebo or ODN were included using an empirical, time-dependent term to better account for the BMD response profile.
Results: The population PK/PD model was simultaneously fit to uNTx, lsBMD and dfBMD data from all treatments. Goodness of fit diagnostics and visual predictive checks indicate that the model well characterizes the uNTx, lsBMD and dfBMD data. Only underlying bone formation and resorption rate parameters need to be adjusted between bone sites with primarily cortical (dfBMD) versus trabecular (lsBMD) bone.
Conclusions: The model supports that a combination of drug effects on bone resorption and osteoclast cycling can generate the behaviors observed in the Phase II data, including a non-monotonic dose-response relationship and enhanced bone resorption post-cessation of therapy in both cortical and trabecular bone.