Variability of the gastrointestinal tract is rarely reflected in in vitro test protocols but often turns out to be crucial for the oral dosage form performance. In this study, we present a generation method of dissolution profiles accounting for the variability of fasted gastric conditions. The workflow featured 20 biopredictive tests within the physiological variability. The experimental array was constructed with the use of the design of experiments, based on three parameters: gastric pH and timings of the intragastric stress event and gastric emptying. Then, the resulting dissolution profiles served as a training data set for the dissolution process modeling with the machine learning algorithms. This allowed us to generate individual dissolution profiles under a customizable gastric pH and motility patterns. For the first time ever, we used the method to successfully elucidate dissolution properties of two dosage forms: pellet-filled capsules and bare pellets of the marketed dabigatran etexilate product Pradaxa. We showed that the dissolution of capsules was triggered by mechanical stresses and thus was characterized by higher variability and a longer dissolution onset than observed for pellets. Hence, we proved the applicability of the method for the in vitro and in silico characterization of immediate-release dosage forms and, potentially, for the improvement of in vitro-in vivo extrapolation.
Keywords: PhysioCell; biopredictive dissolution testing; design of experiments; individual dissolution profiles; machine learning; mechanical stress events; motility patterns.