No abstract

No abstract

Introduction Aerosol delivery from DPIs could be affected by different factors. This study aimed to evaluate and predict the effects of different factors on drug delivery from DPIs. Methods Modelling and optimisation for both in vitro and in vivo data of different DPIs (Diskus, Turbohaler and Aerolizer) were carried out using neural networks associated with genetic algorithms and the results are confirmed using a decision tree (DT) and random forest regressor (RFR). All variables (the type of DPI, inhalation flow, inhalation volume, number of inhalations and type of subject) were coded as numbers before using them in the modelling study. Results The analysis of the in vitro model showed that Turbohaler had the highest emitted dose compared with the Diskus and the Aerolizer. Increasing flow resulted in a gradual increase in the emitted dose. Little differences between the inhalation volumes 2 and 4 litres were shown at fast inhalation flow, and interestingly two inhalations showed somewhat higher emitted doses than one‐inhalation mode with Turbohaler and Diskus at slow inhalation flow. Regarding the in vivo model, the percent of drug delivered to the lung was highly increased with Turbohaler and Diskus in healthy subjects where continuous contour lines were observed. The Turbohaler showed increased lung bioavailability with the two‐inhalation modes, the Diskus showed a nearly constant level at both one and two inhalations at slow inhalation. The Turbohaler and Aerolizer showed little increasing effect moving from one to two inhalations at slow inhalation. Conclusions Modelling of the input data showed a good differentiating and prediction power for both in vitro and in vivo models. The results of the modelling refer to the high efficacy of Diskus followed by Turbohaler for delivering aerosol. With two inhalations, the three DPIs showed an increase in the percent of drug excreted at slow inhalations.