The objective of this study was to classify samples of tablets containing dipyrone, caffeine and orphenadrine using near infrared (NIR) spectroscopy and chemometric techniques. The data set had 300 spectra of samples from three tablets per batch and four different manufacturers. The pre-processing was accomplished by Savitzky-Golay algorithm with the first derivative, window with 17 points and second-order polynomial. The tablet classification was performed using chemometric models based on principal component analysis (PCA), soft independent modeling of class analogies (SIMCA), genetic algorithm- (GA-LDA) and successive projection algorithm-linear discriminant analysis (SPA-LDA). For PCA analysis, clusters were observed for each group of tablets. The SIMCA model was built using 15 and 30 spectral measures for the training set of similar drugs and reference drugs, respectively. The GA-LDA model used 12 variables, whereas SPA-LDA selected only two wavelengths, 1572 and 1933 nm. The methodology allowed a quick and non-destructive classification of the tablets and without the need for conventional analytical determinations.
chemometrics; drug screening; near infrared spectroscopy; pharmaceuticals; quality control
