Radial Basis Functions-Partial Least Squares for Simultaneous Determination of Ethinylestradiol and Levonorgestrel

Resolution of complex binary mixtures of Ethinylestradiol (ST) and Levonorgestrel (LEV) was successfully achieved with minimum sample pre-treatment and without analyte separation. The work is based on the radial basis functions- partial least squares (RBF-PLS) analysis of UV spectral data. For calibration and external test sets, binary mixtures were rationally designed. The results for modeling and subsequent prediction in external test set samples resulted in Q2 values of 94.1 and 99.4% for ST and LEV, respectively. The RBF-PLS models were then successfully applied to determine the analytes in real samples consisting of pharmaceutical preparations. In the analysis of the real samples, we excluded the spectral regions where the unknown interferences are absorbed and matrix effect exists. The mean recoveries for the real samples were between 90.6 and 106.8%. The estimated precisions of the method in terms of RSD% were in most cases below 4%.