Comparison of the predictive ability of several second-order multivariate methods in the simultaneous determination of two therapeutic drugs in human urine

Abstract

The aim of this paper is to study the applicability of second-order multivariate methods in the simultaneous determination of two therapeutic drugs in human urine samples. The studied drugs, irinotecan and thalidomide, are used in the treatment of malignant tumours. Irinotecan (CPT-11) is used to treat colon cancer; recent studies have shown the benefits of using thalidomide in combination with CPT-11 in the treatment of this disease. CPT-11 is highly fluorescent, but the native fluorescence of thalidomide is very weak. The second-order methods assayed were parallel factor analysis (PARAFAC), unfolded partial least-squares (U-PLS) and multidimensional partial least-squares (N-PLS), both combined with the residual bilinearization procedure (RBL). The excitation-emission matrices (EEMs) of the samples were recorded as analytical signal. The accuracy and precision of the algorithms were evaluated through the root mean square error of prediction (RMSEP) and the elliptical joint confidence region test (EJCR), obtaining better results with PARAFAC, which was successfully applied to the determination of thalidomide and CPT-11 in human urine samples, after a previous liquid–liquid extraction with chloroform.