Rapid and robust quantification of ten cannabinoids in cannabis oils using FT-NIR spectroscopy coupled with PLS-R...

Fourier transform near-infrared (FT-NIR) spectroscopy, when coupled with chemometric modeling, offers a promising alternative to conventional chromatographic methods for the rapid quantification of cannabinoids. However, its application to cannabis oils - particularly for simultaneous multi-cannabinoid profiling - remains largely unexplored. In this study, we evaluated the performance of FT-NIR spectroscopy combined with partial least squares regression (PLS-R) models for the quantitative prediction of ten cannabinoids, including both neutral and acidic forms, in cannabis oils. A total of 102 spectra from 17 distinct oil formulations were acquired and modeled against high-performance liquid chromatography-photodiode array (HPLC-PDA) reference data. All PLS-R models demonstrated good to excellent predictive capabilities, with cross-validation and external prediction R² values exceeding 0.75, residual predictive deviation (RPD) values exceeding 1.8, and ratio of performance to inter-quartile range (RPIQ) values exceeding 2.0. Notably, seven cannabinoid models are reported here for the first time in cannabis oils. The present oil-based models exhibited superior or comparable performances to those previously obtained for dry and wet cannabis inflorescences, a finding attributed to the greater compositional homogeneity of oil matrices. These results establish FT-NIR spectroscopy as a rapid, non-destructive, and cost-effective analytical platform for cannabinoid quantification in cannabis oils, enabling high-frequency, on-site testing for quality control, stability assessment, and shelf-life monitoring.