In silico methods in stability testing of hydrocortisone, powder for injections: multiple regression analysis versus dynamic neural network/In silico metode u ispitivanju stabilnosti hidrokortizona, liofilizata za infuziju: visestruka Regresiona analiza i dinamicke neuronske mreze

This article presents the possibility of using of multiple regression analysis (MRA) and dynamic neural network (DNN) for prediction of stability of Hydrocortisone 100 mg (in a form of hydrocortisone sodium succinate) freeze-dried powder for injection packed into a dual chamber container. Degradation products of hydrocortisone sodium succinate--free hydrocortisone and related substances (impurities A, B, C, D and E; unspecified impurities and total impurities)--were followed during stress and formal stability studies. All data obtained during stability studies were used for in silico modeling; multiple regression models and dynamic neural networks as well, in order to compare predicted and observed results. High values of coefficient of determination (0.95-0.99) were gained using MRA and DNN, so both methods are powerful tools for in silico stability studies, but superiority of DNN over mathematical modeling of degradation was also confirmed.