Analysis of seven selected cannabinoids in human plasma highlighting matrix and solution stability assessments

Abstract

Cannabis consumption has and continues to increase dramatically, as does its legalization for recreational and/or medicinal use at the state, but not at the federal level. The increased consumption and legalization have spurred significant cannabis focused research, with particular interest in defining the pharmacokinetic characteristics of this complex natural product. Supporting this research requires a bioanalytical method that accurately and simultaneously quantifies the primary cannabinoids and their metabolites. The objective of this method validation was to meet pre-specified sensitivity targets (0.5 ng/mL for most analytes) from a low sample volume (0.2 mL) and a single extraction approach that could quantify Δ9-tetrahydrocannabinol, cannabidiol, and their metabolites. Moreover, we sought to rigorously characterize the stability of included cannabinoid analytes, both in solution and plasma. The developed assay required optimization of extraction and mobile phase solvents, as well as mass transitions to achieve the selectivity required to meet the desired sensitivity targets. Stability experiments indicated solution stability of no more than 6 months when stored in polypropylene at −30 or −80°C and ∼3 years (34.5 months) of plasma stability when stored in polypropylene at −80°C. The assay was successfully applied to ∼1650 samples without a batch failure. This validated LC–MS/MS assay provides unique information on cannabinoid stability and has been utilized to generate novel data on the pharmacokinetics of cannabis constituents and their metabolites.

Introduction

Despite its status as a Schedule 1 federally-controlled substance in the United States, cannabis (Δ9-tetrahydrocannabinol [Δ9-THC] and/or cannabidiol [CBD]) is legalized or decriminalized for medical and/or recreational use in all but four states [1]. Legalization at the state level and the proliferation of marketed products containing cannabis or its derivatives have increased the need for and interest in conducting cannabis-­related research. There is particular interest in an improved understanding of the pharmacokinetics (PK) of cannabis, which vary based on administration route, cannabis composition, and the prior user experience with cannabis.

Cannabinoid concentrations in plasma and other biomatrices have historically been measured using GC–MS [2–4] or GC–MS/MS [2–7]. However, recent advances have made it possible to quantify cannabis constituents on liquid chromatography–tandem mass spectrometry (LC–MS/MS) platforms [8–13] without the need to derivatize the sample prior to analysis. The aim of this study was to transition an existing GC–MS/MS method to determine seven cannabinoids, including: Δ9-THC, Δ8-THC, 11-hydroxy-Δ9-THC (11-OH-THC), 11-nor-9-­carboxy THC (COOH–THC), cannabinol (CBN), CBD, and 7-hydroxycannabidiol (7-OH-CBD) via LC–MS/MS with the following assay objectives. First, we set out to develop an extraction approach that could be applied to the array of physicochemical properties exhibited by cannabinoids and their metabolites. Second, we sought chromatographic conditions that provided reasonable sample throughput while also allowing separation of (1) Δ8- and Δ9-THC and (2) matrix interferences from analytes of interest. Next, we wanted to better characterize the stability of analytes included in the assay under a variety of conditions, and in particular, the long-term matrix stability of these analytes when stored in polypropylene containers at −80°C. Finally, we applied the method to the analysis of over 1650 plasma samples collected across multiple studies.

Experimental

Materials

The seven analytes (1 mg/mL in methanol) and their corresponding deuterated internal standards (100 μg/mL in methanol) were purchased from Cerilliant (Round Rock, TX). These stock solutions were transferred to and stored in polypropylene cryovials at −30°C after opening. Ammonium formate was purchased from Sigma–Aldrich (St Louis, MO), while solvents, including LC–MS grade methanol, hexane, and methyl-tert-butyl ether (MtBE) were procured from Fisher Scientific (Waltham, MA). Type 1 water used in the preparation of reagents and mobile phase was obtained from a Thermo Scientific Barnstead GenPure system (Bedford, MA). Control human plasma (K₂EDTA) was obtained from BioIVT (Westbury, NY) in small volumes from individual donors (male and female) and pre-screened for the analytes prior to purchase in bulk.

Preparation of calibration standards, quality control samples, and internal standards

For calibration standards, an intermediate working solution was prepared in 50% methanol : 50% water from the 1 mg/mL commercial stock solutions at a concentration of 10 µg/mL for all analytes except for COOH–THC, which was prepared at 50 μg/mL. This 10/50 µg/mL combined intermediate standard working solution was stored in a polypropylene container and used to prepare secondary working solutions in 50% methanol:50% water at concentrations of 5/25, 10/50, 25/125, 50/250, 100/500, 250/1250, 500/2500, 750/3750, and 1000/5000 ng/mL, also stored in polypropylene. Nine calibration standards were prepared fresh for each analytical batch at concentrations of 0.5/2.5, 1/5, 2.5/12.5, 5/25, 10/50, 25/125, 50/250, 75/375, and 100/500 ng/mL by fortifying 200 µL of control human plasma with 20 µL of the corresponding standard working solution. The negative controls, double blanks (no analyte or internal standard) and zero samples (internal standard only), were freshly prepared in the same lot of control human plasma as the calibration standards for each analytical batch.

The bulk quality control (QC) sample preparation utilized a separately prepared combined intermediate working solution at 10/50 µg/mL in 50% methanol : 50% water (stored in polypropylene). Low, medium, high, and dilution QCs were prepared at concentrations of 1.5/7.5, 15/75, 80/400, and 200/1000 ng/mL, respectively, in silanized glass volumetric flasks (5 and 10 mL). Aliquots of the QCs were transferred to polypropylene microcentrifuge tubes for frozen storage at −80°C.

The internal standards were d₃-7-OH-CBD, d₃-11-OH-THC, d₃-COOH-THC, d₃-CBD, d₃-CBN, d₃-Δ9-THC, and d₃-Δ8-THC. A combined intermediate internal standard (IS) working solution was prepared in 50% methanol:50% water at 400/2000 ng/mL with d₃-COOH-THC prepared at five times the concentration as the other internal standards. The IS secondary working solution was prepared from the intermediate solution at 20/100 ng/mL. All IS solutions were stored at −30°C in polypropylene containers.

Sample preparation

The double blank, zero samples, and calibration standards were prepared fresh in 200 µL of control human plasma for each analytical run. Human plasma study samples and QCs were stored at −80°C, thawed unassisted, and vortex mixed prior to aliquoting 200 µL for each study sample or QC sample into polypropylene test tubes (13 × 100 mm). A repeater pipette was used to add 50 µL of the 20/100 ng/mL IS solution to all tubes except double blanks, which received 50 µL of IS solvent. The pH was adjusted by the addition of 500 µL of 2 mM ammonium formate (pH 3.5) to each tube and all tubes were mixed for 30 s on a multi-tube vortexer. A 2 mL volume of 5:1, hexane: MtBE was added to all tubes as the extraction solvent. The tubes were capped, vortex mixed for 5 minutes, centrifuged at 1200 g for 15 minutes, and placed in a −80°C freezer for 30 minutes. The organic upper layer was decanted to fresh polypropylene tubes and evaporated in a Turbovap (Zymark, Hopkinton, MA) under ∼15 psi of air from a house compressed air system for ∼18 minutes in a 40°C water bath. The dried extracts were reconstituted in 200 µL of 50% methanol : 50% water. Following a 2-minute centrifugation at 1200 g and then a 30 s vortex mix, the reconstituted extracts were transferred to conical bottom polypropylene autosampler vials (300 µL) for analysis.

LC–MS/MS instrumentation

Method development and validation utilized a Thermo Scientific TSQ Vantage triple quadrupole mass spectrometer interfaced with a Thermo Scientific Accela LC system (Waltham, MA) including a quaternary pump, cooled autosampler, and column heater. Thermo Scientific Xcalibur software (Version 2.2 SP1) was used to acquire and process data including quantitation. Chromatographic separation on the Accela utilized a Luna Omega PS C18, 2.1 × 100 mm, 3.0 µm particle size, column with a SecurityGuard ULTRA C18, 3.0 mm ID guard column (Phenomenex, Torrance, CA), both at a temperature of 40°C. The autosampler was held at 10°C, and the injection volume was 30 µL. The needle wash was 85% methanol : 15% water, mobile phase A was 1 mM ammonium formate (pH 3.5), mobile phase B was methanol, and the flow rate was 250 µL/minutes. Gradient conditions on the Accela LC were: hold at 72% B for 2.25 minutes, linear increase to 85% B from 2.25 to 10.50 minutes, switch to 100% B at 10.51 minutes, hold at 100% B until 12.00 minutes, and return to 72% B at 12.01 minutes. The injection cycle time was 16 minutes. The TSQ Vantage was operated with the following conditions: spray voltage 3000 V, vaporizer temperature 350°C, ion transfer tube temperature 300°C, N₂ sheath gas pressure 30 (arb), N₂ sweep gas pressure 0 (arb), N₂ aux gas pressure 10 (arb), and collision cell argon gas pressure 1.5 mTorr. Mass transitions (precursor ion → product ion), dwell times, declustering potentials, and collision energies are shown in Supplementary Table 1. Nitrogen was from a Peak Scientific Infinity XE nitrogen generator (Westford, MA), and the argon collision gas was ultra-high-purity grade. The resolution was set for unit mass (FWHM ∼0.7 amu) in both mass analysers, which were operated in the Selected Reaction Monitoring (SRM) mode.

Analysis and quantitation

Peak area ratios (PAR) of each analyte to the corresponding deuterated IS were used for quantitation. The calibration standards for all analytes were regressed using a linear model and 1/x² weighting, which were chosen based on the deviations from nominal concentrations observed during the three accuracy and precision validation runs. Three-fourths of the calibration standards and two-thirds of QC samples were required to have measured concentrations within ±20% of nominal for the run to be acceptable. QC samples were also required to have a CV within 20%.

Method validation

The scope of work for Method Validation (MV) was documented in a protocol prior to the start of experiments to validate this fit-for-purpose assay. The protocol described the objectives, experimental design, specific validation experiments, and acceptance criteria. MV was guided by the 2020 FDA Bioanalytical Method Validation Guidance [14].

The nine calibration standards, along with a double blank and a zero sample, were prepared in duplicate for the three, accuracy and precision runs. The two curves bracketed the low, medium, and high QC samples (n = 6) during analysis. Following these runs, the results were processed, using both calibration curves and using only the first curve, and compared to determine if a single calibration curve could be used during future analytical batches. Acceptance criteria for accuracy was mean QC sample concentrations within ±20% of nominal at all concentrations within and between assays and for precision was a CV of <20%.

Dilution integrity, or partial volume analysis, was included in the first MV run using 10-fold dilution (n = 6) of a QC pool prepared above the upper limit of quantitation (ULOQ). Selectivity was evaluated in the second MV run and was tested by processing and analyzing 10 different lots of cannabinoid-free human plasma as fortifications at the lower limit of quantitation (LLOQ), zero samples, and double blanks, to evaluate potential interference from endogenous components.

Recovery and matrix effect were calculated using PARs from pre-extraction fortifications divided by PARs from post-extraction fortifications and analyte peak areas of post-extraction fortifications divided by analyte peak areas of neat solutions, respectively, at the three QC concentrations in six lots of matrix. Process efficiency (combined recovery and matrix effect) was calculated using analyte peak areas from pre-­extraction fortifications divided by analyte peak areas of neat solutions.

Carryover was evaluated in each analytical run by analyzing an extracted plasma zero sample injected immediately following the calibration standard at the ULOQ. Carryover is calculated for each analyte as the PAR of the zero sample as a percentage of the PAR of the LLOQ for the analyte, with an acceptance criteria of <20%.

Re-injection reproducibility was assessed by re-injecting the entire set of extracts from the third accuracy and precision run, after storage in the autosampler, and comparing the calculated results for the QCs to nominal values.

For matrix stability, all experiments were conducted at low and high QC concentrations (n = 3) exclusively in polypropylene. The short-term stability experiments included freeze/thaw (three cycles of freezing and thawing), and storage at ambient temperature for six and 24 hours. Short-term extract stability was assessed by re-injecting extracts of QCs at low (1.5/7.5 ng/mL) and high (80/400 ng/mL) concentrations from the second MV run in the third MV run. Long-term stability was assessed at −80°C at nominal time points of 1 month, 3 months, 6 months, 12 months, 30 months, and 36 months. For all matrix stability experiments, the mean results were compared to the nominal concentration to determine acceptability.

Short-term stability of the 10/50 µg/mL combined intermediate working solution in polypropylene was assessed after 24 hours at ambient temperature as the ratio of the mean PAR of the solution stored at ambient temperature to the mean PAR of the control solution stored at −30°C. Long-term stability of this solution was assessed at nominal time points of 1 month, 3 months, 6 months, and 12 months at two storage temperatures, −30°C and −80°C. Long-term stability was also calculated as the ratio of the mean PARs of the stored solution to the control solution, a fresh preparation at 10/50 µg/mL.

The criteria for analytical QCs described in Section 2.5 were applied to stability QCs. Solution stability experimental results needed to have a PAR within $\pm$10% difference from the control solution PAR to establish stability. Unacceptable stability data did not invalidate the bioanalytical method. Any restrictions in the method were documented and applied during sample analyses.

Results

Method development

This bioanalytical method was developed as a fit-for-purpose, research-grade LC–MS/MS replacement for a GC–MS/MS assay that required a 1 mL sample aliquot, involved an extensive sample preparation with both protein precipitation and liquid-liquid extraction to enhance analyte recovery, plus derivatization with N-Methyl-N-(trimethylsilyl) trifluoroacetamide [7]. The target cannabinoids were chosen based on their relevance to human toxicology, pharmacokinetics, and metabolism, as well as for indication of cannabis use. A primary objective for the current method was a relatively simple sample preparation that can accommodate larger batch sizes with reasonable throughput. Because of this objective, sample preparation procedures compromised analyte recovery in exchange for reduced extraction time. All method development steps were completed prior to conducting MV experiments.

The laboratory’s prior GC–MS/MS assay [7] used 9:1 hexane: ethyl acetate as the extraction solvent which was indicative of the requirement for the solvent to cover a range of analyte polarities. For the LC–MS/MS method, mixtures of 9:1 and 5:1 hexane: MtBE were evaluated. The results showed that the 9:1 ratio favored the more lipophilic Δ8- and Δ9-THCs, CBD, and CBN, while the 5:1 ratio was a better compromise for the more polar 7-OH-CBD and 11-OH-THC without completely sacrificing the lipophilic analytes. For COOH–THC, it was preferred to use one set of extraction conditions for all seven cannabinoids and not use a polarity switch or separate LC–MS/MS analysis for this acidic analyte. The 5:1 ratio was selected, and the compromise was to accept low analyte recovery in exchange for a single, combined analysis, with the requirements that there was sufficient and reproducible signal-to-noise with acceptable accuracy at the target LLOQ.

Initial selectivity testing supported five of the seven analytes as acceptable with no significant interferences detected in the negative controls and accuracy at the LLOQ between 88% and 109% across the 10 plasma lots. Results for COOH–THC were not acceptable for four of the 10 lots with significant interference peaks such that accuracy at the LLOQ could not be calculated, and only three of the 10 lots with acceptable accuracy at the LLOQ. For CBN, four of the 10 lots had accuracy at the LLOQ >120%. The strength of the extraction buffer was first investigated to determine if the recovery for COOH–THC could be improved. Comparison between 1 mM and 2 mM ammonium formate (pH 3.5) showed an improvement in response for COOH–THC by a factor of 1.8 with negligible change in the response of the other analytes. Further increasing the concentration of extraction buffer to 10 mM showed greater than seven times the response as the 1 mM for COOH–THC, however, this improvement resulted in response non-linearity for COOH–THC. The next step was to evaluate alternate product ions for COOH–THC and CBN. The SRM transitions that were initially chosen for these two analytes both used the highest abundance product ion (−18 amu) corresponding to the loss of water. Subsequent work identified alternate product ions at m/z 193.2 for COOH–THC and 195.2 for CBN that yielded acceptable method selectivity results across 10 plasma lots.

Method validation

Accuracy and precision

Calibration standards were linearly regressed using 1/x² weighting over the concentration range of 0.5–100 ng/mL for six analytes (2.5–500 ng/mL for COOH–THC). Final determination of regression model and weighting was made after completion of the first three MV runs and based upon inspection of deviations from nominal concentration. All individual calibration standards (18 calibrators/run) were within 12% of nominal for all analytes across the three runs. Representative chromatograms of an extracted double blank (left), zero sample (center), and LLOQ calibration standard (right) are shown in Supplementary Figure 1.

Analytical QC results using duplicate calibration curves were comparable to those generated from a single, nine-point calibration placed at the beginning of the run. Inter-day accuracy and precision data resulting from a single calibration curve are shown in Table 1. Individual QC results were within $\pm$20% of nominal concentration, except one QC High in the first (−24% for 7-OH-CBD, COOH-THC, and CBN, −23% for 11-OH-THC, and −22% for CBD), second (−22% for 7-OH-CBD, −20.2% for CBN), and third (+22% for CBN) MV runs.

Table 1.

Inter-day accuracy and precision experiment results

	Mean accuracy (%CV)
	7-OH-CBD	11-OH-THC	COOH–THC	CBD	CBN	Δ9-THC	Δ8-THC
QC Low (1.5 ng/mL)	95.8 (2.53)	100 (5.99)	96.8 (5.38)	103 (5.48)	96.9 (8.43)	101 (4.31)	101 (5.88)
QC Medium (15 ng/mL)	95.2 (2.26)	95.9 (3.40)	95.9 (2.94)	99.2 (4.70)	98.2 (4.20)	97.2 (3.76)	97.1 (4.61)
QC High (80 ng/mL)	89.3 (5.50)	92.0 (6.52)	89.8 (5.67)	96.0 (8.18)	92.1 (10.3)	95.4 (6.50)	95.8 (7.17)

Data represent three batches of six replicates at each QC concentration.

Selectivity

Significant method development was required to achieve a selective method, as described in Section 3.1. LLOQ data from ten distinct plasma lots during the final selectivity experiment (MV2) are shown in Supplementary Table 2. All LLOQ samples were within ±20% of nominal for all analytes. No co-eluting interferences from endogenous compounds were observed in either the double blank or zero sample in any lot.

Partial volume analysis

Dilution integrity was acceptable for all analytes with individual dilution QC accuracy within ±8.2% of nominal.

Recovery and matrix effects

Results from the recovery and matrix effect experiment are shown in Supplementary Table 3 where the analytes are presented in elution order. Analyte recoveries trended downward, along with process efficiency, while the matrix effect trended generally upward, as the cannabinoid polarity decreased (later elution in reversed-phase chromatography). The notable exception is COOH–THC, which is driven by the unoptimized acid/base extraction characteristics for this analyte. Relatively low recovery values, combined with matrix effects, resulted in low process efficiencies for CBN, Δ9-THC, and Δ8-THC. This is due primarily to extraction conditions that favor the more polar hydroxy metabolites, 7-OH-CBD and 11-OH-THC, and the significant matrix effects of over 70% ionization suppression for these less polar analytes.

Stability aspects

Re-injection reproducibility was demonstrated from the re-injection of the third MV run after storage in the autosampler at 10°C for 76 hours. The mean accuracies for the re-injected QC Low, Medium, and High ranged from 92.6% to 101%, 93.5% to 99.2%, and 90.2% to 101%, respectively. Short-term extract stability was established for 103 hours at 10°C in the autosampler. The QC Low and QC High mean accuracies were between 93.3% and 106%, and 89.8% and 96.1%, respectively.

Stability in plasma

Results from short-term matrix stability experiments are shown in Table 2. The mean accuracy for ambient temperature stability (6 and 24 hours) ranged from 88.3% to 105% with precision from 0.32% to 13.2%. Freeze/thaw testing for three cycles (−80°C to ambient) showed mean accuracy from 83.1% to 104% with precision from 0.16% to 9.46%.

Table 2.

Matrix and solution stability results

n = 3 replicates	Mean accuracy (%CV)
	7-OH-CBD	11-OH-THC	COOH–THC	CBD	CBN	Δ9-THC	Δ8-THC
Conditional stability, plasma
Ambient Temperature (6 Hour)
QC Low (1.5 ng/mL)	90.2 (0.316)	94.4 (2.62)	90.7 (1.99)	98.9 (2.36)	90.6 (7.43)	94.8 (2.84)	93.4 (2.65)
QC High (80 ng/mL)	90.5 (1.68)	96.1 (3.22)	88.3 (1.14)	95.4 (0.588)	93.5 (3.41)	95.5 (4.04)	94.3 (2.63)
Ambient Temperature (24 Hour)
QC Low (1.5 ng/mL)	91.8 (1.28)	99.7 (8.97)	98.0 (5.83)	97.5 (7.36)	94.7 (13.2)	98.8 (4.12)	96.2 (13.2)
QC High (80 ng/mL)	89.5 (1.34)	90.1 (3.68)	89.7 (4.81)	95.4 (4.75)	105 (11.4)	96.1 (3.50)	99.4 (6.91)
Freeze-Thaw (3 cycles)
QC Low (1.5 ng/mL)	95.4 (1.86)	99.2 (9.46)	91.5 (6.50)	89.5 (3.01)	83.1 (5.73)	101 (7.54)	101 (7.28)
QC High (80 ng/mL)	91.9 (0.481)	98.0 (9.44)	91.8 (0.158)	93.2 (6.25)	104 (6.38)	97.7 (2.45)	99.5 (0.821)
Conditional stability, solution, 10/50 μg/mL
Ambient Temperature (24 Hour)	97.2 (0.692)	95.6 (1.06)	98.7 (0.595)	95.8 (0.775)	97.8 (0.894)	94.7 (0.419)	94.7 (0.846)
Long-term stability, −80°C, plasma
42 Days
QC Low (1.5 ng/mL)	94.8 (1.85)	97.1 (0.773)	94.8 (2.52)	100 (1.03)	102 (6.04)	98.8 (1.42)	103 (2.41)
QC High (80 ng/mL)	91.3 (1.54)	95.9 (1.98)	93.4 (0.930)	97.0 (2.13)	92.7 (5.72)	96.4 (2.59)	96.0 (1.40)
95 Days
QC Low (1.5 ng/mL)	95.1 (1.98)	92.2 (3.58)	103 (4.08)	109 (2.82)	86.7 (1.46)	98.4 (1.20)	99.2 (2.88)
QC High (80 ng/mL)	89.3 (1.13)	91.2 (2.36)	86.4 (2.02)	93.2 (0.450)	89.4 (2.91)	97.5 (4.03)	96.3 (1.67)
187 Days
QC Low (1.5 ng/mL)	100 (1.81)	93.5 (3.38)	104 (5.66)	112 (3.40)	95.2 (2.30)	108 (2.91)	105 (8.04)
QC High (80 ng/mL)	94.6 (2.52)	91.3 (0.313)	89.2 (1.96)	97.2 (0.185)	94.4 (0.331)	101 (0.379)	95.5 (1.47)
368 Days
QC Low (1.5 ng/mL)	92.9 (8.48)	91.9 (2.46)	95.4 (2.40)	94.3 (2.83)	97.4 (3.94)	106 (3.49)	89.1 (4.80)
QC High (80 ng/mL)	91.2 (0.552)	92.7 (1.46)	96.9 (4.15)	99.1 (0.961)	94.7 (1.17)	98.6 (2.34)	89.2 (0.535)
909 Days
QC Low (1.5 ng/mL)	92.1 (3.68)	93.0 (1.53)	107 (1.91)	99.6 (0.942)	99.1 (1.39)	106 (3.58)	110 (10.9)
QC High (80 ng/mL)	95.0 (5.82)	98.2 (6.28)	110 (3.91)	97.9 (8.27)	104 (5.65)	110 (7.85)	109 (3.92)
1050 Days
QC Low (1.5 ng/mL)	85.7 (0.899)	89.6 (2.87)	109 (5.87)	92.0 (0.779)	103 (2.73)	101 (1.12)	95.4 (5.12)
QC High (80 ng/mL)	88.5 (0.802)	91.1 (2.09)	109 (1.76)	91.1 (3.10)	99.6 (1.07)	103 (1.01)	104 (0.200)
Long-term stability, −30°C, Solution, 10/50 μg/mL
43 Days	108 (1.33)	107 (1.68)	104 (0.516)	106 (1.44)	100 (1.29)	110 (0.747)	107 (0.524)
95 Days	102 (1.95)	101 (1.11)	103 (0.738)	101 (1.40)	94.4 (1.57)	101 (1.16)	101 (1.06)
184 Daysa	99.8 (1.51)	101 (1.36)	100 (0.627)	95.6 (1.11)	89.5 (1.01)	94.1 (0.714)	91.1 (0.377)
184 Daysb	103 (1.51)	101 (1.36)	102 (0.627)	101 (1.11)	97.8 (1.01)	95.7 (0.714)	95.2 (0.377)
365 Days	99.4 (0.652)	99.7 (1.82)	99.9 (0.291)	88.8 (1.20)	86.5 (0.458)	83.9 (0.529)	81.9 (1.04)
Long-term stability, −80°C, Solution, 10/50 μg/mL
43 Days	108 (0.894)	104 (0.541)	101 (1.06)	105 (0.671)	96.9 (1.22)	108 (1.09)	106 (0.614)
95 Days	103 (0.866)	100 (0.375)	102 (0.815)	103 (0.825)	94.9 (0.866)	103 (0.871)	104 (1.47)
184 Daysa	97.1 (1.23)	96.4 (1.36)	97.0 (0.519)	93.9 (0.973)	87.1 (0.281)	91.5 (0.502)	88.8 (1.07)
184 Daysb	100 (1.23)	96.3 (1.36)	98.6 (0.519)	99.6 (0.973)	95.2 (0.281)	93.0 (0.502)	92.8 (1.07)
365 Days	99.8 (0.373)	101 (1.64)	102 (0.240)	92.7 (0.931)	91.5 (0.335)	88.8 (0.560)	86.8 (0.760)

^aStock solutions used for control working solution were previously stored in polypropylene cryovials.

^bStock solutions used for control working solution were from freshly opened ampules.

The long-term storage (−80°C) matrix stability results for up to 34.5 months (nominal 3 year stability) are shown in Table 2. The mean accuracy across all six timepoints ranged from 85.7% to 112%, while the precision ranged from 0.185% to 10.9%.

Stability in solution

The results from short-term (24 hours) ambient temperature and long-term storage (43 and 95 days at −30 and −80°C) stability for the 10/50 µg/mL combined intermediate working solution (50% methanol: 50% water) appear in Table 2. The results for 24 hours ambient storage and up to 95 days of storage at either −30 or −80°C were within 5.6% difference (mean, n = 5 injections) across all seven analytes.

At the 184-day time point, stability for the 10/50 µg/mL combined intermediate working solution was within ±4.8% relative to a control solution that was prepared from fresh 1 mg/mL stock solutions for every analyte. Notably, comparing the 184-day old 10/50 µg/mL combined solution to a control solution made from the original 1 mg/mL stock solutions resulted in two analytes (CBN and delta-8-THC) exceeding the ±10% criteria for acceptance as stable, and a third (delta-9-THC) approaching the criteria. At the 365-day timepoint, differences for delta-9-THC (−30 and −80°C), delta-8-THC (−30 and −80°C), CBD (−30°C), and CBN (−30°C) exceeded the $\pm$10% acceptance criteria when compared to a control solution made from fresh 1 mg/mL stock solutions for every analyte. A final experiment demonstrated deviations between the secondary solutions at 1/5 µg/mL (ULOQ) and 5/25 ng/mL (LLOQ) and freshly prepared solutions near −10% following 8 months of storage. Combined, these data indicate that the 10/50 µg/mL combined solution and any solutions derived from this intermediate solution are stable for six months but lose integrity by 12 months. As a result, expiration dates for all solutions stored in polypropylene at either −30 or −80°C were set at 6 months. Finally, the 1 mg/mL stock solutions lose integrity at ∼6 months, and should be used with caution or avoided at that timepoint.

Application to human plasma sample analysis

The TSQ Vantage method was successfully applied to the analysis of ∼1650 human plasma samples across several studies and more than 30 analytical batches. All batches met the acceptance criteria for calibration standards and QCs. The analytical ranges for the seven cannabinoids were appropriate with <10% of samples requiring re-analysis for concentrations above the ULOQ (primarily COOH–THC), and the LLOQ was sufficient for the research objectives that were supported with the assay. Selectivity issues for COOH–THC were manageable due to the high concentrations found in the samples. Fifteen samples had measurable Δ8-THC. Figure 1 summarizes the combined data from two large studies (n = 1105 samples from 763 participants) of ad libitum recreational cannabis use analysed using the described LC–MS/MS method.

Figure 1.

Summary concentration data from two studies of recreational cannabis. A total of 1105 samples were analysed from 763 participants using cannabis ad libitum.

Discussion

Cannabis products vary in composition, administration routes, and use frequency, complicating studies targeting an understanding of their PK. To support research on cannabis PK, our laboratory developed and validated an LC–MS/MS method that simultaneously quantifies seven cannabinoids, including Δ9-THC, CBD, and their metabolites, in plasma. The validated method met all planned objectives and allowed a detailed characterization of stability of the target analytes in both solution and matrix. Approximately 1650 samples were analysed using the assay without a single batch failure.

Several manuscripts detailing analytical methods for cannabis or its constituents have been published, with a recent review by Karschner et al. highlighting methods from 2000 to 2019 [15]. Though GC–MS analysis following derivatization have been used historically, recent methods have primarily utilized LC–MS/MS platforms [2–6, 13, 15–19]. This method achieves the targeted LLOQ of 0.5 ng/mL (akin to many published methods) for all analytes other than COOH–THC from a 200 μL plasma aliquot, which is a lower volume than many other published assays (which utilize up to 1 mL of plasma) [15]. Moreover, further work in our laboratory demonstrated the application of this method when aliquot volumes as low as 50 μL were assayed in support of a study in a preclinical mouse model [20]. In addition to plasma, the method has been adapted for use in hair (preceded by washing the hair with 1:1 acetone: dichloromethane and a 0.1 N hydrochloric acid digestion), urine (including addition of a β-glucuronidase), and tissue (most commonly brain, homogenized to a concentration of 200 mg/mL in type 1 water). Though full validation (i.e., stability, matrix effects, selectivity, etc.) was not pursued for these additional matrices, accuracy/precision experiments obtained similar results to those described for plasma within this manuscript. Including these additional matrices, over 3000 samples have been analysed.

Numerous stability experiments were conducted to fully characterize appropriate storage conditions. Historically, cannabinoid assays have utilized silanized glass to prevent analyte loss due to binding to the container observed when using untreated glass [7, 21–25]. However, where possible it is preferable to limit the use of glassware, to prevent the risk for personal injury if accidental breakage occurs. This is of particular concern in our laboratory, where ∼1100 of the described study samples were collected from individuals suspected or known to be infected with either HIV or hepatitis C. We therefore opted to use polypropylene storage containers, extraction test tubes, and autosampler vials throughout the assay. For stock solutions (50% methanol : 50% water) stored in polypropylene at either −30 or −80°C, all analytes were stable for up to 184 days, however, the more lipophilic analytes (Δ9-THC and Δ8-THC) exceeded differences of $\pm$15% at 365 days. As a result, working solutions are freshly made every 6 months. Critically, QC samples in plasma stored at −80°C were found to be stable for ∼3 years (34.5 months), with longer term matrix stability studies ongoing. In summary, the exclusive use of polypropylene in our assay demonstrated acceptable (6 month) solution and excellent (∼3 years) matrix stability for the studied cannabinoids. To our knowledge, this is the first manuscript describing cannabinoid stability in polypropylene, and importantly, our data suggest similar or greater stability duration of these analytes relative to prior publications [7, 9, 10, 12, 19, 21, 26, 27].

A critical factor in the success of the described assay is the use of chromatography to separate analytes of interest from matrix interferences, as well as to baseline resolve Δ9- and Δ8-THC. Recent work found Δ8-THC in ∼10% of samples (n = ∼1000) found to contain Δ9-THC [28] supporting the necessity to separate these two analytes within LC–MS/MS methods despite the findings in our work that ∼0.2% of study samples contained detectable Δ8-THC. During method development, we observed interference peaks within the COOH–THC and CBN mass transition windows, which compromised signal-to-noise at low concentrations. Significant effort was then invested to optimize extraction and chromatographic conditions that would separate the interference peak from the analyte peak. Though this work decreased the observed interferences, eliminating these interferences required the use of mass transitions other than the loss of water for these analytes. While these transitions generated a less abundant signal than the loss of water transition, their use successfully eliminated the interference peaks in the mass transition windows, resulting in improved signal-to-noise. This finding serves as a valuable reminder that optimizing signal-to-noise, rather than absolute signal intensity, represents the more critical parameter to attain the desired assay sensitivity.

We selected the main cannabinoids (CBD and Δ9-THC) and metabolites being studied in the field at the beginning of method development for inclusion in the described LC–MS/MS assay. However, minor cannabinoids, such as cannabigerol (CBG), represent an emerging research interest. Work is therefore ongoing to add minor cannabinoids (CBG, cannabidivarin) and additional metabolites (7-carboxy cannabidiol, 6-hydroxy cannabidiol, as well as the glucuronides of Δ9-THC and COOH–THC). While the validated core assay described in this publication will be utilized in support of most studies encountered by our laboratory, adding these analytes will enable a broader understanding of cannabis PK and provide support for important research on minor cannabinoids.

Conclusion

The described LC–MS/MS assay was successfully validated and utilized to quantify cannabinoid concentrations in ∼1650 plasma samples without a single batch failure. Novel information on cannabinoid stability in both solution and plasma when stored in polypropylene is presented. There is a critical need to define cannabis PK, particularly in the setting of greater potency cannabis and newer routes of administration, such as vaping.

Author contributions

David J. Anderson (Formal analysis, Investigation, Methodology, Supervision, Validation, Writing—original draft), Tia S. Freeman (Methodology, Validation, Writing—review & editing), Kalii S. Caldwell (Methodology, Validation, Writing—review & editing), Logan R. Hoggard (Investigation, Writing—review & editing), Christopher A. Reilly (Funding acquisition, Project administration, Resources, Supervision, Writing—review & editing), and Joseph E. Rower (Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing—original draft, Writing—review & editing)

Funding

Method development, validation, and analysis of ∼200 samples were supported by the National Institute on Drug Abuse under contract 75N95019C00016 (now 75N95024C00014). Analysis of ∼700 samples to determine cannabinoid concentrations was supported by R01 DA057351 (Rower, co-I), while another ∼400 samples were analysed with support from an ALSAM Foundation Skaggs Scholar Award (Rower, co-PI).

Data availability

The described sample analysis data are owned by the investigators who generated funding for and collected samples. These investigators will publish study specific manuscripts and store data in appropriate repositories. Method specifics or validation data can be shared after a reasonable request to the corresponding author (JER).

Ethics statement

All samples were de-identified prior to shipment to our laboratory for analysis and thus considered exempt human research.