• We consider joint selection of fixed and random effects in general low dimensional nonlinear mixed-effects models setting which naturally occur in many applications such as pharmacokinetics. We propose an iterative algorithm that is inspired from stepwise regression strategies and that is based on a BIC model selection criterion whose penalty is adapted to mixed-effects models. We demonstrate the robustness of the algorithm in different simulated experiments and its practical benefits on the clinical study of an antibiotic agent kinetics.