Introduction
The medicinal use of Cannabis sativa L. (Cannabaceae) products is increasing
all over the world. The most common therapeutic indications of cannabis and
cannabinoids are for the treatment of pediatric resistant epilepsy, chronic
noncancer pain, multiple sclerosis, dyskinesias of Huntington’s and
Parkinson’s diseases, and tics of Tourette syndrome [1 ].
Cannabis oil extracts are prepared from dried Cannabis sativa L.
inflorescences and incorporated in common edible oils (e. g. , olive
or sunflower) or even obtained using these oils as extraction media.
Cannabinoids are terpene phenolic compounds typical of the cannabis plant.
Δ9 -THC is the most psychoactive constituent in cannabis. It
has many diverse pharmacological effects with therapeutic value in the treatment of
different medical conditions [2 ]
[3 ]
[4 ]
[5 ].
Other non-psychotropic cannabinoids, mainly CBD and CBN, are increasingly researched,
showing partially distinctive effects [6 ]
[7 ]
[8 ]
[9 ]
[10 ]
[11 ].
Thus, quantification of these cannabinoids is also important to understanding the
pharmacological properties of cannabis oil. Δ9 -THC and CBD are
present in the plant as THCA and CBDA, respectively [12 ]
[13 ]. Decarboxylation is
temperature-dependent [14 ]
[15 ], and preheating of cannabis samples has
been recommended to potentiate the final cannabis oil extract [16 ]
[17 ].
Medical cannabis oils generally possess high levels of the therapeutic CBD and lower
levels (generally less than 0.3%) of the psychotropic
Δ9 -THC. The FDA has issued warning letters to firms that market
unapproved new drugs that allegedly contain CBD. As part of these actions, the FDA
has determined the cannabinoid content of some cannabis oil products (not approved
by the FDA), and many were found to contain levels of CBD that are very different
from the label claim [18 ].
In Argentina, cannabis and its derivatives are schedule IV-controlled substances
(prohibited use) [19 ], but new regulations
have allowed production for medical purposes through licensed producers [20 ]. Production of commercial cannabis oil in
Argentina must take place in a facility using good manufacturing practices, and
products must be tested for the presence and content of Δ9 -THC
and CBD, using validated analytical methods.
Several methods, based on GC [21 ]
[22 ]
[23 ]
[24 ] or LC [24 ]
[25 ]
[26 ] have been published for the
determination of Δ9-THC, CBD, and other cannabinoids in cannabis oil. GC,
one of the most used techniques for the quantitative analysis of cannabinoids in
plant materials, has been in use for a long time. The high temperature of the
injection port transforms the acid cannabinoids into the neutral cannabinoids. Since
the cannabis oils contain the acidic and neutral forms, a derivatization step is
required to prevent the decarboxylation [27 ]
[28 ], and trimethylsilyl
derivatives have been shown to be suitable for analysis [28 ]
[29 ].
Thereby, the value of total cannabinoids can be measured by determining the acid and
neutral form separately. In contrast to GC, LC-based techniques allow the direct
analysis of cannabinoid (neutral and acid) in the extracted sample.
The main goal of the present work is the development, optimization, and validation of
a method with a simple derivatization coupled to GC-MS for the determination of
Δ9 -THC, CBD, and CBN in cannabis oil.
The GC-MS validated method, according to FDA [30 ] and ICH [31 ] guidelines, has
proven to be very accurate, highly reproducible, and sensitive to determine the
target cannabinoids, with only 10 μL of sample tested. In addition, the
method was successfully applied to the quantitative analysis of 10 different
cannabis oils. The application of this method to differentiate between cannabis oils
with high or low content of Δ9 -THC, CBD, or mixed
Δ9 -THC/CBD will provide physicians with essential
information so that they can carry out a suitable therapy with patient's
pathology.
Results and Discussion
An accurate and robust analytical method has been developed for the quantification of
3 cannabinoids relevant to the health and safety of cannabis oil users. This process
was optimized for dilution solvent and sonication time. Dilution solvents had been
selected according to the existing literature [23 ]
[25 ]
[32 ]
[33 ]
and to the physical-chemical properties of the studied analytes. The solvents
evaluated were methanol, ethanol, diethyl ether, and petroleum ether. To evaluate
the effect of different sonication times, the samples were sonicated at various
times (5, 10, 20, and 30 min). These experiments were performed in triplicate using
cannabis oil, and the relative peak areas obtained for each cannabinoid were
compared to establish the best dilution solvent and sonication time. An initial
prescreening (methanol and 5 min sonication time) of cannabinoids was made in full
scan mode. Mass spectrometric identification criteria were according to the WADA
[34 ]
[35 ]. The cannabinoids identified were Δ9 -THC, CBD,
CBN, CBC, CBG, THCA, and CBDA.
The statistical analysis of these data indicated that dilution solvent was the most
significant factor ([Fig. 1 ]). Regarding
diethyl ether, the recoveries obtained for Δ9 -THC, CBN, CBC,
THCA, and CBDA were significantly greater when compared to using methanol: [F
(1.4)=10.85, p<0.05], [F (1.4)=36.74, p<0.05], [F
(1.4)=8.37, p<0.05] [F (1.4)=102.94, p<0.05], and [F
(1.4)=107.49, p<0.05], respectively. Subsequently, the recoveries
obtained with diethyl ether for Δ9 -THC, CBN, CBC, THCA, and CBDA
were significantly greater when compared to using ethanol: [F (1.4)=10.53,
p<0.05], [F (1.4)=10.79, p<0.05], [F (1.4)=16.94,
p<0.05] [F (1.4)=157.39, p<0.05], and [F
(1.4)=184.72, p<0.05], respectively. These results were obtained
using Fisher’s test, which evaluates the intra- and inter-group study
variance.
Fig. 1 Relative peak areas and SDs obtained for different dilution
solvents for cannabidiol (CBD), cannabinol (CBN),
Δ9 -tetrahydrocannabinol (Δ9 -THC),
tetrahydrocannabinolic acid (THCA), cannabidiolic acid (CBDA),
cannabichromene (CBC) and cannabigerol (CBG).
Diethyl ether and petroleum ether extractions resulted in similar recoveries, with no
significant differences between them. Lower standard deviations and associated
errors are observed with diethyl ether, which makes diethyl ether the most promising
option ([Fig. 1 ]).
Another relevant parameter, also studied, was the sonication time (5–30 min)
that might result in a greater recovery of the target analytes, as well as influence
signal intensity. All studied times of extraction resulted in similar recoveries,
with no significant differences between them. The sonication appears not to depend
on the time of exposure to the solvent mixture. In order to make the process faster,
5 min was chosen.
The data were used to develop an optimized sample preparation protocol using 10
µL of cannabis oil with 10 mL of diethyl ether, with sonication for 5 min
and vortexing for 30 sec (3 cycle sonication/vortexing).
Full method validation was conducted according to the FDA and ICH guidelines using
olive and cannabis oil as matrix. The selectivity of the method was evaluated by
analyzing of blank samples. No interfering substances were detected at the selected
retention times and m/z windows for all cannabinoids. Using the
above-mentioned criteria for positivity, all the analytes
(Δ9 -THC, CBD, and CBN) were successfully and unequivocally
identified in all the spiked samples at the LLOQ. Therefore, the method was
considered selective for Δ9 -THC, CBD, and CBN determination.
The calibration model was evaluated from a set of 9 calibration points. The
homoscedasticity assumption was tested in a linear regression analysis. The residual
plot clearly showed that error was not randomly distributed around the concentration
axis. The F-test also revealed a significant difference between the variances. There
was evidence that variances were significantly different, thus homoscedasticity was
not met.
Weighted least squares regressions had to be adopted in order to compensate for
heteroscedasticity. Six weighting factors were evaluated for each compound
(1/x0.5 , 1/x, 1/x2 ,
1/y0.5 , 1/y, 1/y2 ). The weighting
factor that resulted in the lower sum of relative errors and, simultaneously, a mean
R2 value of at least 0.99 was chosen. This factor was
1/x2 for all analytes. The method obtained linear
relationships by means of these weighted least squares regressions.
Calibrators’ accuracy [mean relative error (bias) between the measured and
spiked concentrations] was within a±15% interval for all
concentrations. Calibration data are shown in [Table 1 ].
Table 1 Linearity and lower limit of quantification (LLOQ)
data
Compound
Linearity data (n=6)
LLOQ (n=10)
IS
Weight
Lineal range (μg/mL)
Slope (mean±SD)
y-Intercept (mean±SD)
R2 (range)
Nominal conc. (μg/mL)
Mean calculated conc.±SD (μg/mL)
CVa (%)
REb (%)
Recovery±S.D. (%)
CBD
∆9 -THC-d3
1/X2
0.1 – 30
0.236±0.031
0.077±0.015
0.9936–0.9970
0.1
0.109±0.011
10.4
8.6
103±2
∆9 -THC
∆9 -THC-d3
1/X2
0.1 – 30
0.129±0.029
0.030±0.011
0.9924–0.9990
0.1
0.101±0.020
13.2
1.1
100±1
CBN
∆9 -THC-d3
1/X2
0.04 – 11.7
0.770±0.027
-0.018±0.001
0.9911–0.9953
0.04
0.041±0.008
12.8
2.6
95±4
Mean values and standard deviations. a Coefficient of variation
(%). b Relative error (%) =[(spiked
concentration − nominal concentration/nominal concentration)
× 100]. CBD cannabidiol, Δ9 -THC
Δ9 -tetrahydrocannabinol, CBN cannabinol and
∆9 -THC-d3
Δ9 -tetrahydrocannabinol-d3 .
The LLOQs of the compounds were 0.1 µg/mL for
Δ9 -THC, CBD; and 0.04 µg/mL for CBN ([Table 1 ]), and the upper limit of
quantification ranged from 11.7 (CBD) to 30.0 (Δ9 -THC, CBD)
µg/mL with a minimum of 9 calibration points.
These limits were considered satisfactory, especially when compared to those obtained
by other authors. Citti et al. [36 ] used 100
μL (10 times higher) of cannabis oil, and the reported LLOQs for
Δ9-THC, CBD, and CBN were greater than that presented herein. Bettiol et al.
[37 ] and Deida et al. [38 ] applied 40 μL (4 times higher) of
cannabis oil and reported LLOQs of 1.0 μg/mL for Δ9-THC,
CBD, and CBN. The mentioned papers report higher LLOQs than ours when analyses were
carried out using a LC-DAD or LC-MS system. A GC-MS analytical method reported by
Ciolino et al. [23 ] presented a LLOQs of 0.3
μg/mL for Δ9-THC, CBD, and CBN; however, this work, unlike
ours, started from gravimetrically measured samples.
Both intra-day and inter-day precisions for the entire extraction and analysis
process were determined by extracting 2 cannabis oils over the course of 5 days, and
5 times on a single day. These oils were chosen to obtain a broad representation of
analyte profiles. Cannabis oil A contains a high concentration of CBD and a low
concentration of Δ9 -THC and CBN, while cannabis oil B contains a
low concentration of CBD and a high concentration of Δ9 -THC and
CBN.
Intra-day precision was evaluated by analyzing, on the same day, 5 replicates of the
cannabis oils. The obtained% RSD were lower than 13.8% at all
studied concentration levels. ([Table 2 ]).
The evaluation of inter-day precision was made within a 5-day period. The
obtained% RSD were lower than 14.1% for all cannabinoids at the
tested concentrations. ([Table 2 ]).
Table 2 Intra-day and inter-day precision data.
Cannabis Oil
Compound
Intra-day (n=5)
Inter-day (n=15)
Measured±SD (μg/mL)
RSDa (%)
Measured±SD (μg/mL)
RSDa (%)
A
CBD
23.60±0.81
4.5
22.45±0.71
4.7
∆9 -THC
0.42±0.09
3.9
0.52±0.10
10.7
CBN
0.18±0.04
13.8
0.17±0.04
12.5
B
CBD
0.11±0.04
11.3
0.10±0.06
11.7
∆9 -THC
28.20±2.51
10.3
25.81±3.74
10.9
CBN
4.81±0.43
12.9
4.27±0.61
14.1
Mean values and standard deviations. a Relative standard deviation
(%). CBD cannabidiol, Δ9 -THC
Δ9 -tetrahydrocannabinol and CBN cannabinol.
The recovery was evaluated at LLOQ for Δ9 -THC, CBD, and CBN. The
recoveries for the compounds were between 95% and 103% ([Table 1 ]). According to our results, the
method can be considered a powerful technique, revealing a fast and efficient
extraction of the target analytes with a lower sample volume.
After validation of this analytical method, in order to demonstrate the
applicability, it was successfully applied to routine analysis of 10 cannabis
oils.
In controlled/regulated production cannabis oils (cannabis oil #1 and #2),
CBD was detected at levels consistent with the product labeling, and
Δ9 -THC levels were very low, as expected for products derived
from hemp oil ([Table 3 ]). In addition,
cannabis oil #1 and #2 showed high ratios of CBD to Δ9 -THC and
CBN ([Table 3 ]).
Table 3 Cannabinoid content of tested authentic samples.
Cannabis Oil
Declared CBD*(mg/mL)
CBD (mg/mL)
Δ9-THC (mg/mL)
CBN (mg/mL)
CBD / Δ9-THC ratio
# 1
20
22.0
1.1
1.1
20
# 2
20
22.0
1.0
1.0
22
# 3
Not declared
0.4
29.0
3.4
0.01
# 4
Not declared
1.5
6.1
0.3
0.25
# 5
Not declared
0.3
2.0
ND
0.15
# 6
Not declared
ND
1.3
0.6
–
# 7
Not declared
ND
10.3
2.6
–
# 8
Not declared
ND
4.3
ND
–
# 9
Not declared
ND
2.4
0.2
–
# 10
Not declared
ND
ND
ND
–
* CBD declared on labels. ND: Not detected. CBD
cannabidiol, Δ9 -THC
Δ9 -tetrahydrocannabinol and CBN cannabinol.
Uncontrolled/unregulated production cannabis oils had CBD,
Δ9 -THC, and CBN concentrations that differed notably. Our
analysis revealed that 2 preparations (samples oils #3 and #7) exhibited high levels
of Δ9 -THC and low or undetectable concentration of CBD, while in
another 3 (samples oil #6, #8, and #9), the CBD content was not detectable and
Δ9 -THC ranged from 1.3 to 4.3 mg/mL ([Table 3 ]).
CBD appears not to have adverse consequences at high doses; however,
Δ9 -THC concentrations observed in cannabis oils (especially
oil # 3, # 4, # 7, # 8) could be enough to produce intoxication, especially among
children [39 ].
Interestingly, the web site from the CBD oil #4 producer reports a
CBD/Δ9 -THC ratio of 1:1, while we found a
CBD/Δ9 -THC ratio of 0.25:1.
Finally, in the cannabis oil #10, CBD, Δ9 -THC, and CBN were not
detected, which would indicate that it was falsely sold as cannabis oil.
CBN was quantifiable in most samples (except oil #5, #8, and # 10). CBN is formed by
Δ9 -THC oxidation during plant aging or inappropriate storage
conditions [40 ]. Therefore, its determination
may assist in the evaluation of the quality of cannabis oils.
Taken together, the results highlighted the extreme variability of the
uncontrolled/unregulated production of cannabis oils, and these results are
in agreement with those obtained from products available on the United States and
Italy markets [41 ]
[42 ]. Bonn-Miller et al. [41 ] reported that 26% of tested
products contained less CBD than declared on the label, while Pavlovich et al. [42 ] reported 9 out of 14 tested samples had
concentrations that differed notably from the declared amount.
In conclusion, a GC-MS method was developed, optimized, validated, and applied for
the simultaneous detection and quantification of CBD, Δ9 -THC, and
CBN in cannabis oil. The analyses were carried out using small sample volumes (10
μL of cannabis oil), and the method was successfully applied to real samples
derived from Argentina’s market.
The issues of variability of cannabinoid content in preparations and inaccurate label
claims in the global market justify the need to have a method like the one that has
been developed and validated to provide concentration data for each preparation.
To our knowledge, this is the first method available in Argentina validated according
to international guidelines for quantification of CBD, Δ9 -THC,
and CBN in cannabis oil.
In addition, CBD and Δ9 -THC concentration data in medicinal
cannabis oil are crucial for physicians to be able to properly adapt the prescribed
dose to the available preparation.
Further studies are needed to evaluate the impact of different cannabis oils on
cannabinoids pharmacokinetics and clinical outcomes.
Materials and Methods
Reagents and standards
Analytical standards: Δ9 -THC (purity 99.4%), CBD
(purity 99.8%), CBN (purity 99.5%), and ISTD:
Δ9 -tetrahydrocannabinol-d3
(Δ9 -THC -d3 , purity 98.8%) were
purchased from Cerilliant (Round Rock, TX, USA) as 1.0 and 0.1 mg/mL in
methanol solutions.
Methanol, ethanol, diethyl ether, petroleum ether, and ethyl acetate were
provided from Merck Chemistry (Buenos Aires, Argentina); all chemicals were
analytical ACS or chromatography grade. MSTFA was acquired from Thermo Fisher
Scientific.
Calibrators and internal standard
A working solution (A) was prepared by proper dilution of stock solutions (1
mg/mL) with methanol to the final concentrations of 10.0
μg/mL for Δ9 -THC and CBD. Additionally,
working solutions of 400 and 4.0 μg/mL for CBN were prepared. As
ISTD (Δ9 -THC -d3 ) stock solutions of 0.1
mg/mL was used. All primary and working solutions were stored at
−20°C into amber glass vials.
Working calibrators (0.1, 2.0, 4.0, 6.0, 8.0, 10.0, 15.0, 20.0, and 30.0
μg/mL Δ9 -THC and CBD; and 0.04, 0.80, 1.50,
2.40, 3.20, 3.90, 5.90, 7.80, and 11.70 μg/mL CBN) were made
daily by adding of each standard to 1.0 mL (final volume) of ethanol. For the
0.1 μg / mL Δ9 -THC and CBD and 0.04
μg / mL CBN points, working solution (A) and 4.0 μg
/ mL CBN were used, respectively. For the rest of the points, stock
solutions and a working solution of 400 μg / mL CBN, were
used.
Tested material
The cannabis oil extracts used in this study were obtained from subjects who
attended the Analytical Toxicology Advisory Laboratory (CENATOXA) with a request
for quantification of Δ9 -THC and CBD.
Eight samples were obtained from uncontrolled/unregulated production
cannabis oil (n =8) and 2 samples from controlled/regulated
production cannabis oil (n =2).
Gas chromatography and mass spectrometry analysis
The samples were analyzed using an HP 6890 N gas chromatograph, combined with an
HP 5973 quadrupole mass spectrometer and an HP 6890 Series injector (all from
Hewlett-Packard). Data were acquired and analyzed using Agilent Enhanced
ChemStation G1701DA software (Agilent Technologies). The separation of the
analytes was achieved using a capillary column (30 m ×0.25 mm I.D., 0.25
μm film thickness) with 5% phenylmethylsiloxane (HP-5 MS),
supplied by J & W Scientific. Carrier gas (helium) was set at a constant
flow rate of 1.0 mL/min. The volume of injection was
2 μL on split mode (split ratio of 1:10); the injection port and
transfer line temperatures were set at 280°C and 280°C,
respectively. The oven temperature started at 60°C, followed by a
temperature ramp of 10°C/min to 300°C held for 2 min.
Total separation run time was 26.0 min. The ion source was maintained at
220°C and the quadrupole at 150°C. The mass spectrometer was
operated with a filament current of 300 mA and an electron energy of 70 eV in
the positive electron ionization mode. Selected ion monitoring mode was used
with a dwell time of 80 ms. Three ions for each analyte and 1 ion for ISTD were
chosen based on selectivity and abundance in order to maximize the
signal-to-noise ratio in matrix extracts ([Table 4 ]). Agilent Enhanced ChemStation G1701DA software (Agilent
Technologies) was used for data acquisition, data processing, and instrument
control. The mass spectra were obtained by collecting the data at rate of 1.38
scan/s over the m/z range of 50–600. Compounds were
identified by comparing the retention times of the chromatographic peaks with
those of authentic compounds analyzed under the same conditions when available
and through the National Institute of Standards and Technology (NIST 1998);
Pfleger/Maurer/Weber: Mass Spectral Library of Drugs, Poisons,
Pesticides, Pollutants and Their Metabolites (2011), and Scientific Working
Group for the Analysis of Seized Drugs (SWGDRUG 2019), MS spectral database.
Table 4 Optimized identification parameters for compounds
Compound
Retention Time (min)
Time window (min)
Quantification ion (m/z)
Qualifier ion 1 (m/z)
Qualifier ion 2 (m/z)
Dwell time (ms)
Peak width (m/z units)
CBD
20.944
20.50 – 21.30
390
337
458
80
0.5
Δ9 -THC
21.807
20.50 – 21.30
21.30 – 22.10
315
386
80
0.5
Δ9 -THC-d3
21.787
21.30 – 22.10
374
-
-
80
0.5
CBN
22.564
22.10 – 22.90
367
238
310
80
0.5
CBD cannabidiol, Δ9 -THC Δ9 -tetrahydrocannabinol, Δ9 -THC-d3 Δ9 -tetrahydrocannabinol-d3 and CBN cannabinol.
Sample preparation
Cannabis oil was mixed by inversion prior to sample preparation. Then 10
μL of oil was diluted thousand-fold (1:1000 dilution) in 10 mL of
diethyl ether and vortexed for 30 sec. Extracts were sonicated for 5 min and
vortexed for 30 sec. The sonication and agitation cycles were performed twice
more.
One hundred microliters (100 μL) of extract with 4 µL of ISTD
(0.1 mg/mL) were evaporated to dryness under a gentle stream of nitrogen
at 45°C. For the derivatization procedure, 50 μL of MSTFA were
added to the dried residue and vortexed for 10 sec. The tubes were heated on a
thermo block at 60°C for 20 min. A 2 μL aliquot of the resulting
solution was injected into the GC–MS system.
Validation procedure
The analytical method validation was performed in accordance with the guidelines
of the FDA [30 ] and ICH [31 ]. The validation was performed following
a 5-day validation protocol and included selectivity/specificity,
linearity, limits, intra- and inter-day precision, and recovery.
Selectivity
Since it is not possible to obtain cannabis oil that is devoid of cannabinoids,
blank samples were prepared using olive oil (n =10). Samples were
extracted and analyzed according to the previously described procedure.
Peaks at the retention time of interest were compared with those from olive oil
samples spiked with analytes at the LLOQ.
The acceptance criteria for compounds identification were according to the WADA
[34 ]. The method would be considered
selective if no analyte could be identified in the blank samples by applying
those criteria.
Calibration curves and limits
The linearity of the method was established on aliquots of ethanol (100
μL) spiked with the corresponding working solution to obtain calibrator
samples. Replicates (n =9) at each concentration were analyzed as
described Fernandez et al. [35 ].
The lowest point of the calibration curve was the LLOQ of the method. The LLOQ
was determined by analyzing 10 replicates of spiked blank olive oil samples
(independent from those of the calibration curve). It was tested whether the
signal-to-noise ratios (S/N) of all analytes was greater than 10.
Furthermore, precision and accuracy data with a coefficient of variation
(CV%) less than 20% and a relative error (RE%)
within±20% of the nominal concentration were obtained.
Intra- and inter-day precision
In order to evaluate intra- and inter-day precision, different authentic samples
containing different cannabinoid profiles were evaluated on 5 separate days as
well as 5 times on the same day.
Precision, expressed as % RSD, was determined by calculating the percent
ratio of the standard deviation divided by the calculated mean concentration
times 100. Data were evaluated using a 1-way analysis of variance (ANOVA) with
day as the grouping variable. RSD values below 15% and at LLOQ below
20% were acceptable for quantitative analysis.
Recovery
For the analysis of recovery, 2 sets of samples (n =5) were prepared at
LLOQ: sample set 1 representing the neat standard/ISTD and sample set 2
consisting of blank olive oil spiked before dilution. The ISTD were added to
sample set 2 after dilution. The recovery results were obtained by comparison of
peak areas ratio of sample set 1 to those of the corresponding peaks in sample
set 2.
