A comprehensive comparison among four different approaches for predicting the solubility of pharmaceutical solid compounds in supercritical carbon dioxide

Supercritical technologies have been developed in the food, environmental, biochemical and pharmaceutical product processing during the recent decades. Obtaining accurate experimental solubilities of pharmaceutical compounds in supercritical carbon dioxide (SC-CO2) and their correlations are highly important and essential for the design of industrial operating units. In this study, the solubilities of six pharmaceutical compounds (Anti-HIV, Antiinflammatory and Anti-cancer) in SC-CO2 were correlated using four different models: cubic equation of state (EoS) model (SRK and modified-Pazuki EoSs), empirical and semi-empirical models (Chrastil, Mendez-Santiago-Teja, Spark et al. and Bian et al. models), regular solution model coupled with the Flory-Huggins equation, and an artificial neural network-based (ANN-based) model. In EoS calculations, twin-parametric van der Waals (vdW2) and Panagiotopoulos-Reid (mrPR) mixing rules were used for estimating the supercritical solution properties, with three different sets employed for obtaining critical and physicochemical properties of the solid compounds. To evaluate the capabilities of various approaches, a comprehensive comparison was carried out among the four models based on several statistical criteria, including AARD, R-adj and F-value. Results of the analysis of variance (ANOVA) indicated that the ANN-based model provided the best results in terms of correlating the experimental solubility of the pharmaceutical compounds in SC-CO2.