Plant food supplements have shown a rising trend in their consumption in the western world. The extensive marketing campaigns and availability through various different platforms can be rendered as catalysts for their popularity. In regards to the Belgian market, suspicious samples entering the border generate serious concerns, necessitating an efficient screening approach for detection of the regulated plants in plant food supplements.
This study aims at developing a chemometric approach coupled with fingerprinting for detection of four regulated plants which are commonly used in plant food supplements with slimming potential: Aristolochia fanghi, Ilex paraguariensis, Garcinia cambogia and Hoodia gordonii. A chromatographic method was developed using ultra-high performance liquid chromatography with diode array detector (DAD). Furthermore, selection of orthogonal wavelengths was made to extract maximum information from the fingerprints by performing a correlation analysis on the DAD data. Chemometric analysis was carried out using two approaches: 1. binary models for each plant separately and 2. multiclass models. The data were corrected for time shifts using correlation optimized warping and further pretreated using classical techniques like signal noise variate, autoscaling and derivatives), before being subjected to partial least square – discriminate analysis. A minimum of 86% correct classification rate for external test set validation (17 sample for binary dataset and 59 samples for multiclass dataset) was observed for all the plants.
Analysis for twelve plant food supplements having slimming as indication, seized by the Belgian Federal Agency for Medicines and Health Products, was carried out in order to classify the plants using the developed models. Out of the 12 samples, 10 were found to be positive for the regulated plants with the binary models, whereas 11 were found positive when classified using the multiclass model. Even though both models showed good correct classification rates (%), the multiclass model could give problems when two or more of the targeted plants are present, since it is only able to classify with one plant in contrast to using different binary models. Therefore, binary models might be preferred over multiclass models. Furthermore, it can also be concluded that a combination of chemometrics and fingerprinting shows positive feasibility for classifying plants using their characteristic fingerprint profiles.
