- Introduction
Cannabis is known for a variety of applications e.g. . in the textile industry but more importantly for recreational and medical use. Cannabis and its derived products are available on the Belgium market via both illegal and legal ways. CBD flowers as well as medicinal cannabis can both be purchased in a legal way, but cannabis to be used as recreational drug is illegal. Indeed, cannabis is the most widely consumed illicit drug in Europe and products contain generally a concentration of Δ9-THC up to 15 % (m/m) [1,2]. Δ9-tetrahydrocannabinol is one of the main cannabinoids responsible for the psychotropic effects and the Belgian law authorizes a maximal concentration of 0.2 % (m/m) [2]. However, there are only limited cannabis seizures because it is difficult to discriminate between legal CBD flowers and cannabis flowers with a Δ9-THC concentration higher than 0.2% m/m, at least for police officers on site. There is a need to characterize these products with a rapid, ecological and cheap analytical method. The current methods of reference are GC-MS, GC-FID and HPLC-DAD [4]. These methods are very efficient but slow, expensive and require a thorough sample preparation. Furthermore, they require trained personnel, and aren’t ecological [4,5] and they are not suited for onsite analysis. Near infrared spectroscopy combined with chemometric tools has the potential for quantitative and qualitative prediction of plant natural products compounds [6]. In addition, near infrared spectroscopic tools are easy to employ, green, rapid and relatively cheap and can be used on-site (handheld devices).
- Material and methods
For this study 189 samples, found in Belgium, were used. They were composed of (i) flowers seized on festivals and on the street and supplied by the Belgian authorities, (ii) agricultural hemp from Belgian farmers and (iii) flowers, used or sold as “other tobacco to smoke”, either seized by Belgian authorities, bought on the Belgium market or voluntarily donated by sellers in order to analyze their samples.
Cannabis samples were analyzed by GC-FID for total-tetrahydrocannabinol detection and quantification with an officially validated method.
The analysis of these products was performed by FT-NIR spectrometer (spectra were recorded in reflectance mode with the Near Infrared Reflectance Accessory (NIRA)) and dispersive NIR handheld devices combined with chemometrics.
All pretreated spectra were analysed with chemometrics. Principal component analysis (PCA) is applied as exploratory data analysis. Soft independent modelling of class analogy (SIMCA) was used to build a binary classification model according to the Belgian legislation.
All chemometric treatments were performed with Matlab R2018b (The Mathworks®). The algorithms were part of the ChemoAC toolbox (Freeware®, ChemoAC consortium, version 4.1).
- Results and discussion
Figures : Score plot of the PC1, PC2 and PC3 of spectra pretreated. Samples with >0.2% THC in red and blue and sample with < 0.2% THC in green.
The SIMCA classification model has a correct classification rate of 92 % (for the benchtop) and 93 % (for the handheld device), meaning that 51 of the 56 samples and 52 of 56 samples in the external test set are correctly classified as recreational drug, industrial hemp or cannabis flowers containing less than 0.2% m/m of Δ9-THC.
- Conclusion
Near infrared spectroscopy allows to discriminate various samples. The difference is caused by the totality of compounds of the plant and not only by the Δ9-THC content. Indeed, plant material is a complex matrix and the fingerprint is a combination of the totality of compounds. This preliminary study is a first step to prove that NIR spectroscopy could be used as a preliminary screening method for the authorities to make the decision to seize or not.
