Chromatographic fingerprints recorded for a set of genuine and counterfeit samples of Viagra(®) and Cialis(®) were evaluated for their use in the detection and classification of counterfeit samples of these groups of medicines. Therefore several exploratory chemometric techniques were applied to reveal structures in the data sets as well as differences among the samples. The focus was on the differentiation between genuine and counterfeit samples and on the differences between the samples of the different classes of counterfeits as defined by the Dutch National Institute for Public Health and the Environment (RIVM). In a second part the revealed differences between the samples were modelled to obtain a predictive model for both the differentiation between genuine and counterfeit samples as well as the classification of the counterfeit samples. The exploratory analysis clearly revealed differences in the data for the genuine and the counterfeit samples and with projection pursuit and hierarchical clustering differences among the different groups of counterfeits could be revealed, especially for the Viagra(®) data set. For both data sets predictive models were obtained with 100% correct classification rates for the differentiation between genuine and counterfeit medicines and high correct classification rates for the classification in the different classes of counterfeit medicines. For both data sets the best performing models were obtained with Least Square-Support Vector Machines (LS-SVM) and Soft Independent Modelling by Class Analogy (SIMCA).
