A statistically guided optimization plan was developed using a partial factorial design to determine the impact of four parameters used in cannabinoid extractions from dried flowers.
Extraction of tissues
Two levels were selected for each factor: solvent composition (80% methanol, 9:1 methanol/chloroform), extraction technique (sonication, wrist action shaking), extraction time (15 min, 1 h), and solvent volume (10 mL, 25 mL). An initial prescreening of solvents indicated that the extraction efficiency of methanol/chloroform mixture was not different from 80% methanol allowing further studies to use the greener option. The statistical analysis of these data indicated that solvent volume was the most significant factor, followed by solvent composition (Fig. 2). Extraction technique and time did not affect the extraction of cannabinoids. Further studies evaluating the solvent volume to mass and solvent composition using 25 mL extraction solvent with 100 and 200 mg of sample confirmed that 200 mg was equivalent to 100 mg, without saturating the extraction solvent (Fig. 3a). The mass to volume ratios used previously range considerably. In some cases, the mass of sample was as high as 100 mg in 1 mL, up to 500 mg in 100 mL [11, 23]. Although extraction time did not significantly impact the resulting cannabinoid content, it was optimized to increase sample throughput. The factorial design showed slightly lower total cannabinoids at 60 min in comparison with 15 min, potentially indicating some degradation during long extractions. Three time points were assessed: 15, 30, and 60 min. The level of cannabinoids was not significantly different between the time points (Fig. 3b). It was then verified if extraction time could be reduced by evaluating 5, 10, and 15 min in comparison with 15 min with vortexing every 5 min. Again, no significant differences were observed between all four extraction times, while 15 min with vortexing was significantly higher than 5 min (Fig. 3c). These data were used to formulate an optimized standard operating protocol (see Electronic Supplementary Material, ESM) using 200 mg of dried flowers with 25 mL of 80% aqueous methanol for 15 min by sonication with vortexing every 5 min.
Extraction of oils
The extraction method was also optimized for cannabis oil comparing (a) 9:1 methanol/chloroform as used in the UNODC method for cannabis products , (b) 80% methanol, and (c) 100% methanol. These data show that 80% methanol was not sufficient for extracting the cannabinoids in an oil matrix, but 100% methanol was equivalent to the methanol/chloroform mixture. The final optimized standard operating protocol (see ESM) for cannabis oil samples was 50 mg of oil extracted with 10 mL of methanol using the same extraction parameters as the dried flowers.
The stability of cannabinoids was assessed to determine whether losses occur prior to analysis that may significantly affect the quantitative data. Mixed calibration standards prepared in 80% methanol were stored at −20 °C, 4 °C, and 22 °C in the dark. Variations of less than 5% were considered stable and acceptable. Significant changes in the cannabinoids were found after 30 h at room temperature with THCA/CBDA contents decreasing by 6.3 and 9.6%, respectively, while mixed standards stored at −20 °C were degraded after 48 h with THCA and CBDA contents reduced by 8.1 and 10.6%, respectively. The standards were stable at 4 °C for the duration of the 72-h study. Sample extracts were prepared with 80% methanol and stored at 22 °C in light and dark conditions. Under these conditions, THCA and CBDA were considered unstable by 24 h into the study, with reductions of 6.7% for both, resulting in 8–10% increases in the neutral forms of these cannabinoids. Reductions of 11–23% of acidic cannabinoids occurred under light conditions within 24 h. On the basis of these findings an additional study was performed to evaluate extract stability in 80% methanol and 9:1 methanol/chloroform stored in the dark at 22 °C and 4 °C. The 9:1 methanol/chloroform extracts were found to be stable at 22 °C and 4 °C for the duration of the study, 36 and 48 h, respectively, while 80% methanol at 4 ° C was stable for 48 h. The extracts in 80% methanol were not stable at room temperature for 24 h, similar to the previous study. The elimination of chloroform from the extraction solvent improves analyst safety, reduces solvent costs, and maintains cannabinoids contents; therefore 80% methanol is a viable alternative extraction solvent for analytical quantitative analysis with the use of a dark, temperature-controlled autosampler.
Several HPLC columns, phases, dimensions, mobile phases, and gradients were compared in preliminary experiments to determine a method for baseline separation of as many cannabinoids as possible while maintaining a short separation time. The optimal column used for separation of cannabinoids was a Phenomenex Kinetex core shell C18 column with sub-2.0-micron particle size. This required an HPLC system capable of pumping above 400 bar, but a 600-bar instrument was sufficient. Mobile phase pH appeared to be a significant factor in cannabinoid separation; as pH decreases, retention of cannabinolic acids increases, while at higher pH the retention decreases with poor peak shape. By optimizing the pH of the aqueous mobile phase for the separation to 3.6, it was possible to separate THCA from other cannabinoids, while maintaining adequate peak shapes. The final methodology was a 15-min analytical run time with 10 mM ammonium formate pH 3.6 and acetonitrile as the mobile phase as shown in Fig. 4. Resolutions of major cannabinoids were greater than 2.0, while minor components were greater than 1.70 using the mixed calibration standards. Sample extracts were used to confirm resolution during the validation study.
The seven-point calibration curves used on each day of the validation were linear on visual inspection. The correlation coefficient (r
2) for each cannabinoid was greater than 99.5% for all calibration curves on each day of the analysis, as summarized in Table 1. The plots of residuals were random, confirming that linear functions were suitable for cannabinoid concentrations up to 250 μg/mL. Concentrations of CBDA in the high-THC products CAN004 and CAN007 were lower than the lowest concentration standard used in the validation study. In this case, the materials are outside the calibration range of the method. If samples are known to have lower concentrations, the calibration curves can be adjusted to match the materials, therefore allowing for quantitation within the calibration range. In this case, a calibration range from 2.5 to 50 μg/mL would be sufficient for quantitation. One additional standard would be necessary to run in this case, and a lower concentration range curve could be generated independent of the typically employed curve from 5 to 250 μg/mL.
The chromatographic profiles of Cannabis extracts at 220 nm were used to evaluate peak resolution. The peak resolutions for the eight cannabinoids quantified with this method ranged from 1.64 to greater than 2.0 as specified in Table 1; in accordance with AOAC guidelines for dietary supplements and natural products, a resolution greater than 1.5 is sufficient for quantitation given the complexity of natural products . Sample extracts were used to evaluate peak purity and confirm resolution with minor peaks. Peak purity was greater than 99% for all cannabinoids evaluated in this assay.
Repeatability was assessed by quantifying the eight minor and major cannabinoids for which standards were available. The quantitative data from the four replicate samples on day 1 were used to determine method repeatability. All precision measurements used authentic Cannabis materials with a range of cannabinoid concentrations. The repeatability data are summarized in Table 2. Repeatability RSDs ranged from 0.78 to 10.08%. For materials with higher than 0.5% w/w cannabinoid content, the % RSDs ranged from 0.78 to 7.64%; only two of these materials had RSDs greater than 5%. Given that the precision for seven of the nine materials evaluated was less than 5% for the analytes above 0.5% w/w, it is possible that the results for the two materials with greater than 5% RSDs are due to inherent variability of the cannabinoids within these two test samples, rather than an indication of method performance. This is likely due to a higher inherent variability compared with the other strains used in the study. The RSDs over 5% for all other materials were observed for low level cannabinoids, for which small variations in the quantitative data will have more of an impact on the % RSDs. These values are within acceptable validation limits based on their concentrations .
The quantitative data from the four replicates on the 3 days of analysis were used to calculate the within-day, between-day, and total standard deviations to determine the intermediate precision of the method. Intermediate precision ranged from 2.07 to 11.67% RSD, as summarized in Table 2. The HorRat ratios used to determine the acceptability of the % RSDs based on concentration ranged from 0.5 to 2.0, which is the acceptable range as specified by AOAC International guidelines . The HorRat ratios ranged from 0.3 to 0.7 for the oils (Table 3), indicating that there is improved precision with homogeneous materials where the cannabinoids are not bound to the plant material. The minor cannabinoid concentrations in the oils were below the quantitation and detection limits; therefore only data pertaining to the major cannabinoids was obtained.
The cannabinoid recovery was evaluated for the following four major cannabinoids: CBDA, THCA, CBD, and THC. Three concentration levels were evaluated to represent a high, medium, and low concentration material with stinging nettle as the matrix blank. Recoveries are summarized in Table 4 and are within the acceptable ranges as specified by AOAC guidelines .
Limits of detection and quantitation
The method detection limit (MDL) and limit of quantitation (LOQ) were determined using the EPA’s method detection limit procedure . A test sample extract was diluted to very low concentrations to account for issues with closely eluting compounds, which will make detection more difficult for cannabinoids with closely eluting compounds. The detection and quantitation limits for each cannabinoid are summarized in Table 5. The limits for CBD are much higher in comparison with the other cannabinoids because of the number of close eluting peaks in the chromatogram. Most other cannabinoids have sufficient resolution from other unknown peaks which do not impact their quantitation and detection.