Sample selection

A total of 107 dried cannabis flower samples were obtained from state Law enforcement personnel, the National Cannabis Initiative (NMI). These samples were randomly selected by the state Law enforcement team from each state. Samples were received from three states: Colorado (23 samples), Oregon (16 samples), and California (68 samples, with 47 from San Diego and 21 from the Central Valley region). The plant samples were comprised of different brands. Each product was randomly assigned a study identifier to blind researchers to product identification. Upon receipt, product packaging and seals were inspected to ensure product integrity. The lot numbers were recorded, and the claimed amounts on the products were acquired from a label on the product. The products were stored according to packaging instructions or at room temperature in a dry space if instructions were not provided. All products were tested immediately after opening.

For product label accuracy, an acceptance criteria of ± 20% was applied (Sarma et al. 2020). If Δ9-THC concentration is more than 120% of the labeled value, the product was under-claimed, but if the value is less than 80% of the labeled value the product was over-claimed. Products within ± 20% (i.e., 80–120% of labeled value) are categorized as accurately labeled.

Cannabinoid standards and calibration curves

Standard solutions of seven pure cannabinoids (THCV, CBD, CBC, Δ8-THC, Δ9-THC, CBG, and CBN) were isolated in our laboratory with a purity greater than 95% (Ahmed et al. 2008; Husni et al. 2014). The analysis was carried out following the GC-FID method previously described (ElSohly et al. 2016). The cannabinoids standards were used to prepare the calibration curves used for the quantification of the individual cannabinoids.

Sample preparation for GC-FID analysis

Two samples (100 mg each) from each product were analyzed and the average content was used for the label accuracy (the results of the two analysis cannot differ by more than ± 20%; otherwise analysis is repeated again in duplicate). Each sample was extracted with 3 ml of internal standard (I.S.) solution, which contained 1 mg/ml of 4-Androstene-3,17-dione in CH3OH/CHCl3 (9:1), for one hour. The resulting mixture was then filtered to create a working solution for GC analysis where 1 μL of the extracted material is injected on the GC/FID. Analysis was carried out using a Varian 3380 gas chromatography system, which was equipped with a Varian CP-8400 automatic liquid sampler, dual capillary injectors, and dual flame ionization detectors (GC/FID). The instrument parameters were as follows: air at 30 psi (300 mL/min), hydrogen at 30 psi (30 mL/min), the carrier gas is helium with column head pressure of 14 psi (1.0 mL/min), the split ratio at 15:1, septum purge flow rate at 3 mL/min, makeup gas (helium) pressure at 20 psi (30 mL/min), injector temperature at 240 °C, detector temperature at 270 °C, oven temperature was programmed starting at 170 °C (hold for 1 min) and ramping up to 250 °C at 10 °C/min (hold for 3 min), with a total run time of 12 min. The GC column used was J&W DB-1 Agilent with dimensions of 15 m (length) × 0.25 mm (diameter) and a 0.25 μm wall thickness (ElSohly et al. 2016).

Method validation

The method used for this analysis was validated as per the ICH method validation guidelines for the following parameters; specificity, linearity and range, sensitivity, and precision (ICH Guideline 1996).

Specificity

The method specificity was determined as no interfering peaks were found at the retention time of any of the target cannabinoids. Moreover, the qualitative and quantitative analysis of samples was performed by comparing their retention times with the reference standard of each analyzed cannabinoid. The method provided baseline separation for all of the seven cannabinoids. The chromatogram of a sample prepared at the lower limit of quantification (LLOQ) of all cannabinoids is shown in Fig. 1. Furthermore, a chromatogram of one of the commercial samples here is shown in Fig. 2.

Fig. 1

Chromatogram of a sample prepared at the lower limit of quantification (LLOQ) of all of the analyzed cannabinoids

Fig. 2

Chromatogram of a representative commercial sample

Linearity and range

Linearity and range samples were prepared by spiking placebo plant material with different amounts of the target cannabinoids. For THC and CBD, the calibration range was 0.01–70% w/w, while for THCV was 0.01–0.5% w/w. On the other hand, Δ8-THC, CBC, CBG, and CBN calibration range was 0.01–3% w/w. The selection of the calibration range was based on the expected prevalence of the tested cannabinoids in the plant material. Calibration curves were constructed by plotting the concentration of each cannabinoid against the peak area ratio (peak area of each cannabinoid/peak area of the I.S.). The correlation coefficient of the regression line(R2) for each cannabinoid ranged from 0.997–0.9999 (Fig. 3). The detailed results of regression parameters are shown in Table 1.

Fig. 3

Calibration curves of the analyzed cannabinoids

Table 1 Regression equations, R2, Calibration range, LOD, LOQ and retention time (min.) of the tested cannabinoidsMethod sensitivity (limit of detection (LOD) and limit of quantification (LOQ))

Method sensitivity was assessed by LOD and LOQ. The LOD and LOQ concentrations were 0.005 and 0.01%w/w for all the tested cannabinoids, respectively. The LOD and LOQ were evaluated based on signal-to-noise ratios greater than 3 and 10, respectively.

Precision

Method precision was evaluated using six different sample preparations of cannabis plant material (High THC chemovar). The intermediate precision was evaluated by data generated on three different days. The %RSD was less than 8% for all the tested cannabinoids indicating the method is precise (Table 2).

Table 2 Precision (intra-day and inter-day) resultsStatistical analysis

Cannabinoid profiling was performed for each product sample. Descriptive statistics were used to report the mean (SD) claimed amounts and observed amounts of THC in the products from each regional/state area.

Classification was based on a ± 20% acceptance range, with products classified as under-claimed, over-claimed, or accurately labelled. All calculations were completed using Microsoft Excel version 16.66 and JASP version 0.16.4.

If the claimed value is higher than the observed value, the % variance is calculated as follows:

$$\left\\;\mathrm\Delta^9\mathrm\;\mathrm\;(\%)\;-\;\mathrm\;\mathrm\Delta^9\mathrm\;\mathrm\;(\%)\right\}\times100/\mathrm\;\mathrm$$

If the observed value is higher than the claimed value, the % variance is calculated as follows:

$$\left\\;\mathrm\Delta^9\mathrm\;\mathrm\;(\%)\;-\;\mathrm\;\mathrm\Delta^9\mathrm\;\mathrm\;(\%)\right\}\times100/\mathrm\;\mathrm$$