Clustering and diagnostic modelling of slimming aids based on chromatographic and mass spectrometric fingerprints.

ABSTRACT

Herbal medicines and food supplements intended as slimming aids are increasingly gaining popularity worldwide, especially for treating obesity. In this study, an ultra-performance liquid chromatography coupled to photodiode array detection (UPLC-PDA) and an ultra-performance liquid chromatography mass spectrometry (UPLC-MS) method were developed to analyze 92 slimming aids (confiscated by customs), aimed at acquiring highly informative fingerprints. Three types of fingerprints were acquired (PDA, Total Ion Chromatograms (TIC), and MS fingerprints) which were used in the chemometric data analysis. Both unsupervised (i.e., Hierarchical Cluster Analysis (HCA)) and supervised techniques (i.e., Classification and Regression Tree (CART) and Partial Least Squares - Discriminant Analysis (PLS-DA)) were applied. The aim was to perform an in-depth study of the samples, thereby exploring potential patterns present in the data. HCA was able to generate a clustering which was mainly defined by chemical compounds detected in the samples, i.e., sibutramine, phenolphthalein and amfepramone. PLS-DA generated the best diagnostic models for both PDA and TIC fingerprints, characterized by correct classification rates of external validation of 85% and 80%, respectively. For the MS fingerprints, the best model was obtained by CART (65% correct classification rate of external validation). Despite a lower correct classification rate, exploration of the concerned misclassifications revealed that the MS fingerprints proved to be superior since even very low concentrations of sibutramine could be detected. This study shows that reliable chemometric models can be obtained, based on the presence of prohibited chemical substances, which allow high-throughput data analysis of such samples. Moreover, they generate a prime notion of potential threat to a patient's health posed by these kinds of slimming aids. Copyright © 2016 John Wiley & Sons, Ltd.