PII-071 - INTEGRATED MODEL BASED ANALYSES OF MILVEXIAN PHASE 2 SSP TRIAL DATA SUPPORTS PHASE 3 DOSE SELECTION

H. Back¹, A. Ajavon-Hartmann¹, D. Li¹, W. Chen¹, W. Zhou², E. Bozenhardt², S. Merali¹; ¹Bristol Myers Squibb, Princeton, NJ, United States, ²Janssen Research & Development, LLC, a Johnson & Johnson Company, Raritan, NJ, United States.

Background: Factor XIa (FXIa) is a promising target for the development of safer anticoagulants with better efficacy and reduced risk of clinically significant bleeding versus other antithrombotics. Milvexian, a direct, reversible, high affinity, intrinsic human coagulation FXIa inhibitor, was tested in multiple phase 1 and 2 studies and is being evaluated in large phase 3 trials with patients at risk for thrombotic events due to atrial fibrillation, ischemic stroke (IS), or acute coronary syndrome. The current study characterized milvexian disposition and exposure-response (E-R) of clinical efficacy (time to IS or undetermined stroke during treatment) and safety (bleeding) in the AXIOMATIC-SSP phase 2 study to inform potential clinically viable dose range in the presence of 21-day dual anti-platelet therapy (APT) followed by 90-day single APT.

Methods: Population pharmacokinetic (PPK) and PK/PD models were developed using pooled data from 6 phase 1 studies and 2 phase 2 studies. Activated partial thromboplastin time (aPTT) was evaluated as a PD marker. Semi-parametric Cox Proportional Hazards models described the relationships between milvexian exposure and time to efficacy (IS or undetermined stroke) or safety (all bleeding and ISTH major/CRNM bleed) events. Covariates were explored via stepwise covariate modeling; the final models were evaluated by visual predictive check (VPC).

Results: The PPK of milvexian in patients treated for Secondary Stroke Prevention (SSP) was well described using a previously developed model with relevant covariates. The PK/PD of milvexian was adequately characterized using a direct effect Emax/Imax model. Milvexian was modeled to have relationships with efficacy and safety with some key covariates (race, age, milvexian treatment duration, etc.). The VPC of the final models (PK, PK/PD, and E-R) showed adequate model performance, capturing the observed data well. The final E-R models provided a quantitative description of the risk over time of efficacy and safety events by milvexian exposure and dose.

Conclusion: The unique PPK/E-R relationship of clinical efficacy and safety endpoints favorably distinguishes milvexian from previously studied direct oral anticoagulants. These analyses constitute a component of the dose selection for the milvexian phase 3 program.