PAGE. Abstracts of the Annual Meeting of the Population Approach Group in Europe.
PAGE 27 (2018) Abstr 8722 [www.page-meeting.org/?abstract=8722]
Poster: Drug/Disease modelling - Other topics
Julia Winkler (1), Kosalaram Goteti (2), Tonke Van Bragt (3), Aida Aydemir (2), Hans Guehring (4), Rik Schoemaker (1)
(1) Occams Co÷peratie U.A, Amstelveen, The Netherlands, (2) EMD Serono R&D Institute, Billerica, USA, (3) SDD consulting BV, The Netherlands, (4) Merck KGaA, Darmstadt, Germany
Sprifermin, a recombinant human fibroblast growth factor 18, is under development for the treatment of knee osteoarthritis (OA) . OA is characterized by loss of cartilage thickness. The objective of this analysis was to develop a drug disease model based on the magnetic resonance imaging (MRI) data measuring knee cartilage thickness in osteoarthritis patients.
24 months data were used for this analysis that originated from a Phase II trial (FORWARD study). All subjects were planned to receive 4 cycles (each consisting of 3 weekly injections) of intraarticular (i.a.) treatment with sprifermin or matching placebo at intervals of 6 months. A total of 595 subjects were randomized in equal allocations to one of 5 treatment groups:
Pharmacokinetic blood samples were collected pre-dose and 2 hours after the last dose of each cycle. Population analysis using NONMEM software was performed to identify potential longitudinal dose-exposure-response patterns. Cartilage thickness in total femorotibial joint compartment, as well as lateral and medial femorotibial compartments, as measured by MRI were selected for analysis. A base drug disease model for each of the three MRI measurements was developed and this base model consisted of two components; one for the disease progression, which in the present case also includes placebo-disease model since there was no arm with absence of treatment, and one for the drug effect. As serum sprifermin concentrations were below the lower limit of quantitation following the i.a. injection, the exposure model was replaced with a simplified first order input and output equation (Bateman-type function) for a description of the so-called ‘driving force’ profile, modified through a driving force-effect model. The driving force-drug effect component was described using a linear relationship and the drug-effect profile was added onto the disease progression. The driving force-effect relationship was also explored as a Emax-type model. The models were assessed using goodness of fit plots and visual predictive checks (VPCs). Potential influence of covariates was assessed on drug effect, including age, race, geographical region, gender, minimum joint space width at baseline, Kellgren-Lawrence grade, malalignment, and bilateral/unilateral OA using a stepwise covariate model search (SCM) with a p < 0.01 and 0.001 for inclusion/removal in the forward/backward steps.
MRI data for three measurements from 496 subjects with OA from 5 dosing arms were available.
Best fits were obtained with a linear model describing the placebo-disease effect, a Bateman-type driving force profile, and a linear model for the driving force-effect relationship.
Emax-type models for the description of the driving force-effect relationship instead of a linear relationship, did not result in a significant improvement from the base model.
The SCM did not find any covariates that would pass the set significance criteria for any of the three MRI measurements. Therefore, the base models were considered the final models, and adequacy was confirmed using VPCs.
The final model of the MRI cartilage thickness in total femorotibial joint estimated a baseline thickness of 1.79 mm (95%CI: 1.76/1.81) with 0.258 mm additive inter-individual variability (IIV), slope of placebo-disease progression was -0.0165 mm over 2 years (95%CI: -0.026/ -0.0072) with 0.056 additive IIV. Driving force was given by Dose*(Time/Tmax)*exp(-Time/Tmax), with Tmax estimated at 12.3 months (95%CI: 6.7/17.8), and slope of drug effect was 0.00042 mm/driving force unit (95%CI: 0.00029/0.00055). The results for medial and lateral MRI cartilage thickness were consistent.
A placebo-disease and drug disease model was developed for sprifermin for MRI measurements of total femorotibial joint, lateral femorotibial compartment, and medial femorotibial compartment cartilage thickness. This model will be useful for understanding the gain in cartilage thickness as measured by MRI with different treatment schedules and dose levels of sprifermin in future drug development for osteoarthritis.