Population approach for exposure-response modeling of golimumab in patients with rheumatoid arthritis

Golimumab is a human immunoglobulin G1κ monoclonal antibody that binds with high affinity and specificity to tumor necrosis factor-α. The objective of this study was to establish an approach for exposure-response modeling for golimumab in patients with rheumatoid arthritis using the American College of Rheumatology index of improvement (ACRN) as a measure of change in disease severity. Data were collected from 302 patients in the phase III GO-FORWARD trial who received golimumab or placebo plus methotrexate (background therapy) every 4 weeks through week 52. A latent-variable (unobservable) approach was used with an inhibitory indirect response model to link the placebo (methotrexate) effect and golimumab concentrations to ACRN scores. A model parameterization was proposed to allow deterioration beyond baseline and maintain mechanistic interpretability of the population-based indirect response model. The modeling was conducted using a sequential pharmacokinetic/pharmacodynamic approach. None of the covariate factors evaluated (demographics, disease characteristics, comorbidities, or concomitant medications) significantly improved the model fits. Likelihood profiling and a bootstrap analysis were used to assess parameter estimation precision, with their suitability discussed. The approach can be readily extended to model other types of clinical (efficacy or safety) scores with either an upper or a lower boundary.