Simultaneous spectrophotometric determination of carbidopa and levodopa by partial least squares regression, principal component regression and least squares support vector machine methods

Three chemometric methods; partial least squares regression (PLS), principal component regression (PCR) and least-squares support vector machines (LS-SVM) were applied for simultaneous determination of carbidopa and levodopa in synthetic mixtures and real samples. The simultaneous determination of these drugs is a difficult problem due to spectral interferences. The proposed methods were used for multivariate calibration of the spectrophotometric data. The calibration graphs were linear in the ranges of 1-30 μg mL-1 for both carbidopa and levodopa. The relative standard error of prediction (RSEP %) for applying the three methods to 8 synthetic samples in the linear calibration ranges for carbidopa and levodopa was 11.747 and 7.783 for PLS, 14.764 and 9.869 for PCR, 0.229 and 0.170 for LS-SVM respectively. The root mean square errors of prediction (RMSEP) for carbidopa and levodopa with PLS, PCR and LS-SVM were 1.900, 2.388, 0.037 and 1.314, 1.667, 0.029 respectively. The LS-SVM methods were successfully applied for determination of these drugs in commercial pharmaceutical preparations and human urine samples.