An exposure–response model was used to characterize the pharmacokinetic‐pharmacodynamic relationship of GW406381, a COX‐2 inhibitor, based on data from ex vivo prostaglandin E2 inhibition collected in a phase 1 study in healthy subjects. The final model was then used to simulate a proof‐of‐concept study and to explore suitable dosing ranges in case of hepatic dysfunction or metabolic induction. Trough concentrations in the range of IC80 to IC95 were used as target therapeutic concentrations. Symptom relief in a subsequent phase 2b study in 400 patients with rheumatoid arthritis receiving GW406381 was then analysed to support the design of a phase 3 study in which doses in the range between 10 to 400 mg were explored. The exercise also allowed the evaluation of correlations between the biomarker and clinical end point. A 2‐compartment model described the pharmacokinetics of GW406381, whereas the pharmacodynamics was described by an Imax model. In patients with normal organ function, the predicted median therapeutic dose was between 100 and 150 mg. The time course of symptom relief was fitted by a Weibull model, with an Imax model describing the drug effect. Simulations showed that dose‐response discrimination occurred at doses higher than 150 mg, with predicted 60%–80% target engagement in the dose range of 150–400 mg.