Optimization of supercritical fluid extraction of steviol glycosides and total phenolic content from Stevia rebaudiana (Bertoni) leaves using response surface methodology and artificial neural network modeling

Stevia leaves include natural, non-caloric sweetening compounds known as steviol glycosides (SGs) mainly stevioside (ST) and rebaudioside-A (Reb-A). Along with sweetness, stevia generates interest because of its dietetic and therapeutic significance related to the presence of other bioactive compounds in its leaves, such as phenolics, which in association with SGs may contribute to human health by exerting anti-inflammatory, antihyperglycemic, anticdrries, chemopreventive, insulinotropic and diuretic properties. In this study, response surface methodology (RSM) and artificial neural network (ANN) modeling were compared in terms of their estimation capabilities for building effective models with maximum response values. A supercritical fluid extraction (SFE) process was optimized by employing a 5-level-3-factor central composite design to achieve maximum target response values for total extract yield, ST yield, Reb-A yield and total phenolic content of 15.85%, 95.76 mg/g, 62.95 mg/g and 25.76 mg GAE/g, respectively. The optimized SFE parameters included a modifier concentration of 40%, an extraction temperature of 45 degrees C, and a pressure of 225 bar. The ANN model proved an attractive alternative to RSM owing to its improved estimation and predictive capabilities. SFE yielded higher target response values than conventional maceration extraction (24 h) and was a faster, lower energy, and greener extraction method with reduced CO2 emissions and lower solvent consumption.