Mid-infrared spectroscopy as process analytical technology tool for estimation of THC and CBD content in Cannabis flowers and extracts

Tetrahydrocannabinol (THC) and cannabidiol (CBD) are the most notable Cannabis components with pharmacological activity and their content in the plant flowers and extracts are considered as critical quality parameters. The new Medical Cannabis industry needs to adopt the quality standards of the pharmaceutical industry, however, the variability of phytocannabinoids content in the plant material often exerts an issue in the inconsistency of the finished product quality parameters. Sampling problems and sample representativeness is a major limitation in the end-point testing, particularly when the expected variation of the product quality parameters is high. Therefore, there is an obvious need for the introduction of Process Analytical Technology (PAT) for continuous monitoring of the critical quality parameters throughout the production processes. Infrared spectroscopy is a promising analytical technique that is consistent with the PAT requirements and its implementation depends on the advances in instrumentation and chemometrics that will facilitate the qualitative and quantitative aspects of the technique. Our present work aims in highlighting the potential of mid-infrared (MIR) spectroscopy as PAT in the quantification of the main phytocannabinoids (THC and CBD), considered as critical quality/material parameters in the production of Cannabis plant and extract. A detailed assignment of the bands related to the molecules of interest (THC, CBD) was performed, the spectral features of the decarboxylation of native flowers were identified, and the specified bands for the acid forms (THCA, CBDA) were assigned and thoroughly explained. Further, multivariate models were constructed for the prediction of both THC and CBD content in extract and flower samples from various origins, and their prediction ability was tested on a separate sample set. Savitskzy-Golay smoothing and the second derivative of the native MIR spectra (1800–400 cm−1 region) resulted in best-fit parameters. The PLS models presented satisfactory R2Y and RMSEP of 0.95 and 3.79% for THC, 0.99 and 1.44% for CBD in the Cannabis extract samples, respectively. Similar statistical indicators were noted for the Partial least-squares (PLS) models for THC and CBD prediction of decarboxylated Cannabis flowers (R2Y and RMSEP were 0.99 and 2.32% for THC, 0.99 and 1.33% for CBD respectively). The VIP plots of all models demonstrated that the THC and CBD distinctive band regions bared the highest importance for predicting the content of the molecules of interest in the respected PLS models. The complexity of the sample (plant tissue or plant extract), the variability of the samples regarding their origin and horticultural maturity, as well as the non-uniformity of the plant material and the flower-ATR crystal contact (in the case of Cannabis flowers) were governing the accuracy descriptors. Taking into account the presented results, ATR-MIR should be considered as a promising PAT tool for THC and CBD content estimation, in terms of critical material and quality parameters for Cannabis flowers and extracts.