Classificação simultânea de fármacos contidos em comprimidos de analgésico por espectroscopia NIR e ferramentas quimiométricas

Abstract

The focus of this work was to build a chemometric algorithm for classifying samples of tablets, the basis of dipyrone (300 mg), caffeine (50mg) and orphenadrine (35mg) using NIR spectroscopy. The database has 300 spectra of samples taken from three tablets, twosided, per lot, in 50 different lots belonging to four different manufacturers. The data were processed in advance by selecting a spectral range from 1100 to 2499,5 nm, and then was employed the filter Savitzky-Golay with first derivative, window 17 points and second-order polynomial. The characterization of the tablets was done using chemometric models based on Principal Component Analysis (PCA), Hierarchical Cluster Analysis (HCA), and Independent Modeling of Class Analogy for Flexible (SIMCA) method and the K-th Nearest Neighbor (KNN) , prepared for each group of samples tablets. For PCA and HCA, the formation of clusters for each class of tablets was observed. For models SIMCA and KNN, the training set was constructed using 20 and 40 spectral measures, while the test set made use of 10 and 20 spectra, both, of similar and reference drugs , respectively. All models achieved 100% correct classification without errors type I or type II. Using this strategy it was possible to classify the tablets rapidly and non-destructively without the need for multiple analytical determinations.