We address the problem of design optimization using cost functions in nonlinear mixed effects models with multiple responses. We focus on the relative feasibility of the optimized designs, in term of sampling times and of number of subjects. To do that, we extend the Fedorov–Wynn algorithm—a dedicated design optimization algorithm—to include a cost function that penalizes less feasible designs as well as to take into account multiple responses. We apply this extension to the design optimization of a joint pharmacokinetic model of infliximab and methotrexate administered in rheumatoid arthritis. We show the benefit of such an approach when substantial constraints on the design are imposed.