The ability to generate predictive models linking the in vitro assessment of pharmaceutical products with in vivo performance has the potential to enable greater control of clinical quality whilst minimizing the number of in vivo studies in drug development. Artificial neural networks (ANNs) provide a means of generating predictive models correlating critical product characteristics to key performance attributes. In this regard, ANNs have been used to model historical data exploring the relative lung bioavailability of salbutamol from several different nebulizers. The generated ANN model was shown to relate urinary salbutamol excretion at 30 min post-inhalation, which is the index of relative lung bioavailability of salbutamol, to specific fractions of the particle size distribution, to subject body surface area and to the methods of nebulization. This model was validated using unseen data and gave good agreement with pharmacokinetic outcomes for 17 data records. The model gave improved predictions of urinary salbutamol excretion for individual subjects compared to the published linear correlation generated using the same data. It is therefore concluded that ANN models have the potential to provide reliable estimates of pharmacokinetic performance that relate to lung deposition, for nebulized medicines in individual subjects.