A latent variable approach for modeling categorical endpoints among patients with rheumatoid arthritis treated with golimumab plus methotrexate.

The need for accurate exposure-response modeling is critical in the drug development process. Few methods are available for linking discrete endpoints, especially ordered categorical variables, to mechanistic (e.g., indirect response) models. Here we describe a latent-variable approach that is proposed in conjunction with an inhibitory indirect response model to link the placebo/comedication effect and drug exposure to the endpoints. The model is parsimonious, with desirable characteristics at initial timepoints, and allows simultaneous modeling of multiple endpoints that are categorically ordered. Application of the model is demonstrated with data from a phase 3 clinical trial of golimumab, a human IgG1kappa monoclonal antibody that binds with high affinity and specificity to tumor necrosis factor (TNF)-alpha, in patients with rheumatoid arthritis. The efficacy endpoints were 20, 50, and 70% improvement in the American College of Rheumatology criteria (ACR20, ACR50, and ACR70, respectively) as measures of improvement in disease severity. The modeling results were shown to be consistent by using either a sequential or simultaneous pharmacokinetic/pharmacodynamic modeling approach. The suitability of likelihood profiling and proper use of bootstrap methods in assessing parameter estimation precision are also presented. More accurate, parsimonious models with appropriately quantified uncertainty can facilitate better drug development decisions.