To infer patterns of average systemic exposure and to estimate individual exposures in phase III clinical trials of a new anxiolytic, two statistical methodologies were applied and compared: non-linear mixed-effect modelling, and a model-free approach based on quartiles of dose-normalized plasma concentrations of the drug. Although the model-based approach provides more quantitative insight about relationships between average exposure and demographic covariates, the model-free approach provides qualitatively similar results about average clearance and quantitatively similar results about individual exposures, and the model-free approach is easy and inexpensive to implement.