B. Wang1, K. Kowalski2, W. White3, C. Wu1, S. Chapel2, X. Chen1, A. Godwood4, F. Magrini4, C. Kane3, L. Roskos3
1 MedImmune, LLC, Hayward, CA, USA; 2 Ann Arbor Pharmacometrics Group, Ann Arbor, MI, USA; 3 MedImmune, LLC, Gaithersburg, MD USA; 4 MedImmune Ltd., Cambridge, UK
Objectives: Mavrilimumab, a fully-human monoclonal antibody targeting the granulocyte-macrophase colony stimulating factor (GM-CSF) receptor α, demonstrated promising efficacy results in a Phase 2a study in patients with rheumatoid arthritis (RA). The objective of this investigation was to characterize the exposure-response relationship of mavrilimumab using the population modeling approach.
Methods: Adult subjects with active RA received seven subcutaneous administrations of placebo or mavrilimumab (10, 30, 50 or 100 mg) once every other week. Mavrilimumab pharmacokinetics (PK) and efficacy data (DAS28-CRP, ACR20 and ACR50) were analyzed using NONMEM 7. A direct pharmacodynamic model with placebo effect was developed to describe DAS28 response. The dichotomous ACR response was modeled using logistic regression.
Results: Mavrilimumab PK was adequately described by a 2-compartment model with first-order absorption from the subcutaneous dosing site and parallel elimination pathways by the reticuloendothelium system and GM-CSF receptor α. In adult RA patients the maximum placebo effect (Pmax) on DAS28 was estimated to be a 1.29 unit reduction, while the additional drug effect (Emax) was a 0.66 unit reduction. There was a strong correlation between DAS28 and a multi-biomarker disease activity (MBDA) score at baseline. The model predicted probabilities of ACR20 response at Week 12 were 0.40 for placebo and 0.42 to 0.64 for RA patients receiving bi-weekly mavrilimumab administrations. The predicted probability of ACR50 response ranged from 0.12 (placebo) to 0.32 (100 mg). The estimated drug potency (EC50) for ACR20 and ACR50 responses were higher than for DAS28.
Conclusions: Mavrilimumab administrations resulted in significant improvement in RA disease activities, which were directly related to the PK exposure. The MBDA score based on 12 biomarkers was a relevant predictor of DAS28 at baseline. Stochastic clinical trial simulations facilitated the optimal design of late-stage trials of mavrilimumab.
Reference: PAGE 21 () Abstr 2611 [www.page-meeting.org/?abstract=2611]
Poster: Other Drug/Disease Modelling