Thermodynamic analysis and intelligent modeling of statin drugs solubility in supercritical carbon dioxide

Evaluating the solubility of various drugs in supercritical CO₂ is a fundamental step in developing a supercritical process for formulating new pharmaceuticals. Atorvastatin, Lovastatin, and Simvastatin are statin drugs with limited solubility and low bioavailability. Their solubility in supercritical CO₂ has already been experimentally measured. This study provides a thorough theoretical analysis of this topic using novel advanced empirical models, fugacity coefficient-based and liquid thermodynamic models, and various machine learning techniques. Finally, the accuracy of the employed models is validated by comparing their outputs with the experimental data. For all the drugs studied, the empirical models proposed by Bian et al., Si-Moussa et al., Belghait et al., Pitchaiah et al., and Trishen Reddy et al. exhibit the highest accuracy, with AARD values below 10% and R_adj values above 0.99, indicating the least deviation from the experimental data and the best fit. Moreover, both the PC-SAFT and Wilson thermodynamic models demonstrate satisfactory performance in correlating the experimental solubility data for all drugs, with a mean AARD of 6-8% and a R_adj ranging from 0.97 to 0.995. Furthermore, the machine learning model based on an artificial neural network (ANN) showed high accuracy in estimating the solubility of the selected drugs in supercritical CO₂.