A survey of cannabinoids and toxic elements in hemp-derived products from the United States marketplace

Abstract

The 2018 Agricultural Improvement Act removed hemp from Schedule I control, creating a market for hemp products, including cannabidiol-containing products. Due to the market’s rapid growth, little is known about the presence and concentration of cannabinoids in commercial products. Herein, 11 cannabinoids were quantified using liquid chromatography with diode-array detection in a non-representative sampling of 147 products labeled as containing hemp or cannabidiol. A subset of 133 products were analyzed for toxic elements using inductively coupled plasma-mass spectrometry. Cannabinoid content ranged from < LOD – 143 mg/serving, with a median of 16.7 mg/serving. Fewer than half of products surveyed contained cannabidiol concentrations within 20 % of their label declarations. The estimated exposure to lead was below the Interim Reference Level of 12.5 μg/day Pb for women of childbearing age, and most products presented concentrations of Δ⁹-tetrahydrocannabinol below LOQ. These findings emphasize the need for further testing and representative investigation of the cannabidiol marketplace.

1. Introduction

Cannabis sativa L., from the Cannabaceae family, is a flowering plant which has been cultivated in Asian and Middle Eastern countries for centuries, although evidence exists that ancient cultivars were chemically distinct from modern varieties (Russo et al., 2008). Introduced to Western cultures in the 19th century, Cannabis has been used for various purposes including textiles (Klumpers and Thacker, 2019). Although having long been cultivated by humans, the genetic plasticity of Cannabis has made classification difficult and remains a topic of debate. It is now accepted that C. sativa is a single species with cultivars named as C. indica, C. sativa, and C. ruderalis, classified based on geographical origin, morphological characteristics, and chemical composition. Chemotaxonomy has also been used to differentiate between the narcotic “drug-type” (C. indica; marijuana) and non-narcotic “fiber--type” (C. sativa; hemp) cultivars through the concentration ratios of Δ⁹-tetrahydrocannabinol (Δ⁹-THC) and cannabidiol (CBD) (Hazekamp and Fischedick, 2012). More recently, terpenes, secondary metabolites responsible for the characteristic aroma of Cannabis, have also been used in the development of chemovar classifications in conjunction with cannabinoid profiles (Aizpurua-Olaizola et al., 2016; Hazekamp and Fischedick, 2012).

Cannabinoids are terpenophenolic compounds produced as a resinous oil in the glandular trichomes of Cannabis, located primarily on the flowering and fruiting tops of the female plant (Andre et al., 2016). More than 120 phytocannabinoids have been identified and classified into 11 structural subclasses: Δ⁹-THC-type, CBD-type, cannabigerol-type (CBG), cannabichromene-type (CBC), cannabinol-type (CBN), (−)-Δ⁸-tetrahydrocannabinol-type (Δ⁸-THC), cannabicyclol-type (CBL), cannabinodiol-type (CBND), cannabielsoin-type (CBE), cannabitriol-type (CBT), and miscellaneous type (Elsohly and Slade, 2005). Biosynthesized as prenylated aromatic carboxylic acids, almost no neutral cannabinoids are found in the fresh plant (Aizpurua-Olaizola et al., 2016). Despite the vast diversity in known cannabinoid structures, the main cannabinoid components of inflorescence are cannabidiolic acid (CBDA) and Δ⁹-tetrahydrocannabinolic acid (Δ⁹-THCA), which are formed from cannabigerolic acid (CBGA). Non-enzymatic decarboxylation triggered by heat leads to the formation of Δ⁹-THC and CBD from these components, followed by CBC, CBG, and CBN. Decarboxylation occurs rapidly above 150 °C, a condition obtained during smoking or baking. The occurrence of higher levels of neutral decarboxylated cannabinoids, such as CBN, is indicative of oxidative degradation due to extended or inappropriate storage conditions (Pavlovic et al., 2018).

Several cannabinoid-based drugs have been developed. Synthetic Δ⁹-THC, also referred to as Dronabinol, is a Food and Drug Administration (FDA)-approved Schedule III drug in the United States (US) prescribed for the treatment of nausea and vomiting associated with chemotherapy and for the stimulation of appetite in acquired immunodeficiency syndrome patients (Plasse et al., 1991). Epidiolex, a CBD solution, was approved by FDA in 2018 as the first Cannabis-derived drug to treat two rare forms of epilepsy. In 2020, FDA approved an additional indication for Epidiolex for treatment of seizures associated with tuberous sclerosis complex in patients 1 year and older, and Epidiolex remains the only approved CBD-containing product in the US. Epidiolex was initially placed in Schedule V of the Controlled Substances Act, but is no longer controlled (Drug Enforcement Administration, 2020).

Despite its modern medicinal use, Cannabis is perhaps most well-known for its recreational use and psychotropic properties (Hall and Degenhardt, 2009). Narcotic Cannabis plant material, also referred to as marijuana, is currently included in Schedule I of the Controlled Substances Act and is thus considered an illegal drug at the federal level. Despite this illegality, public perceptions of Cannabis have changed over the last two decades, culminating in the legalization or decriminalization of Cannabis for medicinal or recreational purposes in many states (Compton et al., 2017). Recently, the Agriculture Improvement Act of 2018, also referred to as the 2018 Farm Bill, removed C. sativa L. (hemp) and its derivatives with low Δ⁹-THC concentrations (≤ 0.3 % on a dry weight basis) from the list of Schedule I controlled substances, allowing for the cultivation of hemp for non-narcotic uses such as textiles (Agriculture Improvement Act of 2018, United States, 2018). CBD, a chemical constituent of Cannabis which has been anecdotally reported to offer various health benefits beyond the uses for which Epidiolex is approved, has seen a rapid increase in consumer use and interest since the passage of the 2018 Farm Bill. The 2018 Farm Bill explicitly preserved FDA’s authorities over Cannabis products, meaning that FDA-regulated hemp-containing products such as foods, dietary supplements, human and veterinary drugs, and cosmetics must continue to meet any applicable FDA requirements and standards (Food and Drug Administration, 2020). Because of statutory language that generally prohibits the use of drug ingredients in foods and dietary supplements, CBD – which has long been studied as a drug, and which is the active ingredient in the approved drug Epidiolex – cannot legally be used in food or marketed as a dietary supplement under the Federal Food, Drug, and Cosmetic Act (FD&C Act) (Food and Drug Administration, 2020).

The recent influx of hemp and CBD-containing products to the US marketplace has necessitated the development of robust analytical methods for the determination of cannabinoids in these products. Historically, gas chromatography-based methods (GC) have been preferred for the analysis of plant materials and biological samples, and remain widely used (Citti et al., 2018). A major drawback of GC-based methods is that without employing suitable derivatization, carboxylated cannabinoids are decarboxylated in the heated injection port prior to their separation, meaning that only “total” values can be reported for compounds of interest such as CBD and Δ⁹-THC (Citti et al., 2018). High-performance liquid chromatography-based methods (HPLC) with ultraviolet (UV)/visible spectroscopy have increasingly been used for the analysis of cannabinoids (Nahar et al., 2020). These methods advantageously measure both acidic and non-acidic forms of cannabinoids without derivatization, leading to simpler and faster sample preparation (Meng et al., 2018). Single and tandem-mass spectrometry (MS) have also been applied to the determination of cannabinoids, including an LC–MS/MS method for the analysis of hemp-containing oils and cosmetic products in the Canadian marketplace (Meng et al., 2018). Ultra-high performance LC and alternative stationary phases have recently been used to further enhance the separation of difficult critical pairs such as Δ⁹-THC and Δ⁸-THC, CBDA and CBGA, and CBD and CBG (Citti et al., 2018). While less-common, high performance thin layer chromatography (HPTLC) and nuclear magnetic resonance (NMR) have also been successfully applied to the qualitative analysis of hemp and cannabinoids but are generally less sensitive than other techniques (Citti et al., 2018).

The analysis of cannabinoids in food is especially challenging due to the multitude of product types in the marketplace. Most reported methods have focused on the analysis of inflorescence materials, oils, and resins rather than processed foods (Citti et al., 2018, 2019; Vaclavik et al., 2019; Welling et al., 2019). Recently, a validated HPLC/diode array detector (DAD) method was reported by Ciolino et al. (2018) for the quantitative analysis of 11 cannabinoids in foods, beverages, topical preparations, vapes/e-liquids, and over-the-counter (OTC) pharmaceutical products (Ciolino et al., 2018). While lengthy in analysis duration due to a traditional HPLC separation, this method demonstrated acceptable performance in a wide variety of matrices, spanning the range of products commonly seen in the consumer marketplace. Other methods, including those by Vaclavik et al. (2019), have achieved shortened runtimes by using narrower internal diameter columns with smaller particle sizes. However, these methods lack extensive validation in complex food matrices, such as that reported by Ciolino et al. (2018)

The objective of this work was to perform a non-representative sampling of the hemp-containing product marketplace. While large-scale studies of the CBD marketplace have not been previously under-taken, the results of several smaller studies suggest that CBD product labeling may frequently be inaccurate (Bonn-Miller et al., 2017; Digital Citizens Alliance, 2020; Pavlovic et al., 2018; Wakshlag et al., 2020). Several clinical reports have also documented cases of adulteration of CBD products with synthetic cannabinoids (Horth et al., 2018; Rianprakaisang et al., 2019). Due to the propensity of hemp to act as a bio-accumulator of toxic elements (Angelova et al., 2004), it was hypothesized that some hemp products might contain elevated levels of toxic elements such as arsenic (As), cadmium (Cd), mercury (Hg), and lead (Pb). In this work, a total of 147 products were acquired and evaluated for 11 cannabinoids (Fig. 1) and 133 products were analyzed for four toxic elements to provide additional information on products in the American marketplace which are labeled as containing hemp or CBD.

Fig. 1.

Chemical structures of 11 target cannabinoids.

2. Materials and methods

2.1. Standards and chemicals

Certified reference materials for CBDA, CBD, THCA, Δ⁹-THC, Δ⁸-THC, CBN, CBC, CBG, CBGA, cannabidivarin (CBDV), and tetrahydrocannabivarin (THCV) in 1.0 mg/mL concentrations were obtained from Cerilliant Corporation (Round Rock, Texas). Bulk quantities of Δ⁹-THC in ethanol (100 mg/mL) and THCA, CBN, CBD, and CBDA in crystalline powder form were obtained from Lipomed, Inc (Cambridge, MA). Epidiolex was provided by Greenwich Biosciences (Carlsbad, CA). HPLC-grade ethyl alcohol, LC/MS grade-water, LC/MS-grade acetonitrile, and LC/MS-grade acetic acid were purchased from Millipore-Sigma (Billerica, MA). For quantification of toxic elements, multi-element stock solutions were obtained from Inorganic Ventures (Christiansburg, VA). Nitric acid (HNO₃), hydrogen peroxide (H₂O₂), and hydrochloric acid (HCl) used for elemental analysis were double-distilled Optima grade purchased from Fisher Scientific (Fair Lawn, NJ).

2.2. Sample selection

Products marketed as containing CBD or hemp (n = 526) were identified through several channels, including internet searches for the terms “CBD”,” cannabidiol”,” hemp”,” hemp extract”, and “Best CBD”, products listed with online distributors, participants in industry events, advertisements in trade journals, or firms which had previously received FDA warning letters. Identified products were divided into seven categories: tinctures and oils, including oil-containing capsules (tinctures/oils), powders and powder-containing capsules (powders), edibles, gummies, pet products, beverages, and isolates. Products were randomly selected from each of the product categories based on the number of products per category as a proportion of the total. Products included in the final list (n = 201) were purchased online from manufacturers and distributors between February and March 2020, of which up to 147 were analyzed and included in the present survey (Supplemental Table 1.1). The edibles category included honey, chocolate bars, candies, cookies, cake, brownies, and popped corn. Beverages included energy shots, iced tea, green tea, and sparkling water. The pet products category was limited to tinctures, oil-containing capsules, and powder-containing capsules marketed for cats, dogs, or general pet use.

2.3. Composite preparation

Samples were composited prior to sample preparation and analysis. Compositing procedures were adapted from Li and Srigley (2017) and Srigley and Rader (2014) and varied by sample matrix. A portion of each sample composite was split between cannabinoid analysis and toxic element analysis.

2.3.1. Food samples with discrete pieces

A minimum of ten pieces were weighed into a tared grinding chamber (MT40, IKA Works, Inc; Wilmington, NC) and the average weight per-piece was determined. The chamber was transferred to a −80 °C freezer for a minimum of one hr to allow the sample to freeze and harden, then the frozen sample was homogenized using an IKA TubeMill Control at 15,000 RPM for approximately 30 s or until a fine powder was obtained. Products requiring additional homogenization were transferred to an IKA A11 Basic Analytical Mill and ground for approximately 30 s after immersion in liquid nitrogen. Powders were transferred to labeled 15 mL conical bottom tubes (Eppendorf North America; Hauppauge, NY) and stored at −20 °C until use.

2.3.2. Capsules (oil and powder containing)

Ten intact capsules were weighed to determine the average capsule weight. Capsules were carefully cut open and the contents emptied into a labeled 15 mL conical bottom tube to generate a composite sample, which was subsequently shaken or stirred to mix and stored at 4 °C until use. Empty capsule shells were rinsed with hexane and allowed to dry before weighing. Shell weight was subtracted from the intact capsule weight to determine the average content weight of each capsule.

2.3.3. Liquid samples

Liquids, including oils, tinctures, beverages, and honey, were shaken or stirred well to mix, and approximately 10 mL was transferred to a 15 mL conical bottom tube and stored at 4 °C until use. For products specifying a serving by volume, five servings were aliquoted and weighed to determine the average serving weight.

2.4. Sample extraction of cannabinoids

General sample extraction procedures were adapted from Ciolino et al. (2018) and varied based on sample matrix. Where applicable, the composite sample was removed from the refrigerator and allowed to come to room temperature prior to extraction. All samples within a category were initially made up to the same dilution based on an approximate expected CBD concentration range for the category of products, and irrespective of each product’s specific label claims or nomenclature. Samples with CBD concentrations exceeding the calibration range were re-run at higher dilutions.

2.4.1. Tinctures/Oils

Approximately 0.5 mL of sample were added to a tared 50 mL conical bottom tube (VWR International, Radnor, PA) and mass recorded. To the tube, 24.5 mL of 100 % ethanol was added and vortexed for approximately 30 s. An aliquot of 250 μL of extract was added to a 1.5 mL snap-cap centrifuge tube containing 750 μL of 95 % ethanol. The capped tube was briefly vortexed before centrifugation at 5000 RCF for 5 min. The supernatant was passed through a syringe filter (0.22 μm hydrophilic PTFE; Millipore-Sigma) to an amber LC autosampler vial (Agilent Technologies; Santa Clara, CA) under vacuum. The vial was capped and stored at 4 °C until analysis.

2.4.2. Powders and edibles

Approximately 1 g of sample was weighed into a tared 50 mL conical bottom tube, then 19 mL of 95 % ethanol were added and the tube was capped and vortexed for approximately 30 s. An aliquot of 250 μL of extract was added to a 1.5 mL snap-cap centrifuge tube containing 750 μL of 95 % ethanol. The capped tube was briefly vortexed before centrifugation at 5000 RCF for 5 min. Supernatant was passed through a 0.22 μm syringe filter to an amber LC autosampler vial under vacuum. The vial was capped and stored at 4 °C until analysis.

2.4.3. Gummies and candy

Approximately 1 g of frozen sample composite was added to a tared 15 mL conical bottom tube, then the tube was capped and allowed to come to room temperature prior to mass recording. To the tube, 4 mL of 18.2 MΩ water was added, and the tube was capped and subsequently vortexed for approximately 30 s. The tube was transferred to a 50 °C water bath for 30 min, with periodic vortexing until the sample had fully dissolved. An aliquot of 100 μL of dissolved sample was added to a 1.5 mL snap-cap centrifuge tube containing 900 μL of 100 % ethanol. The capped tube was briefly vortexed before centrifugation at 5000 RCF for 5 min. Supernatant was passed through a 0.22 μm syringe filter to an amber LC autosampler vial under vacuum. The vial was capped and stored at 4 °C until analysis.

2.4.4. Beverages

Sample (1 mL) was added to a tared 15 mL conical bottom tube and the mass recorded. To the tube, 1 mL of 100 % ethanol was added, and the tube was capped and subsequently vortexed for approximately 30 s. An aliquot of 100 μL of extract was added to a 1.5 mL snap-cap centrifuge tube containing 900 μL of 100 % ethanol. The capped tube was briefly vortexed before centrifugation at 5000 RCF for 5 min. Supernatant was passed through a 0.22 μm syringe filter to an amber LC autosampler vial under vacuum. The vial was capped and stored at 4 °C until analysis.

2.5. Preparation of stock solutions and standards of cannabinoids

Stock solutions of individual cannabinoids (100 mg/mL or 10 mg/mL in ethanol) were prepared by weighing the standard into a tared, de-staticed 4 mL amber vial and adding an appropriate amount of ethanol to achieve the desired concentration. A mixed cannabinoid calibrator solution was prepared by adding to a 4 mL amber vial 0.1 mL each of the major cannabinoid solutions (CBD, CBDA, Δ⁹-THC, THCA, and CBN; 10 mg/mL) and minor cannabinoid solutions (Δ⁸-THC, CBC, CBG, CBGA, CBDV, and THCV; 1 mg/mL) and diluting with 2.9 mL of ethanol. The mixed cannabinoid calibrator solution was further diluted with ethanol to produce a seven-point calibration curve (Supplemental Table 2.1). Two additional calibration solutions at 1 mg/mL and 0.5 mg/mL CBD were also prepared. Cannabinoid spike recovery solutions were prepared using 10 mg/mL solutions of CBDA, Δ⁹-THC, THCA, and CBN, and a 100 mg/mL solution of CBD (Supplemental Table 3.1). Standards were stored at −20 °C when not in use.

2.6. Spike recovery sample preparation

Representative products intended as matrix blanks (Supplemental Table 3.2) were purchased from a local grocery store, homogenized as described in Section 2.3, and weighed into 50 mL centrifuge tubes. Matrices were spiked with cannabinoid standard solutions at two fortification levels (10 μg/g and 10 mg/g CBD; 10 μg/g and 100 μg/g CBDA, Δ⁹-THC, THCA, and CBN; Supplemental Table 3.1). Matrix blanks were analyzed concurrently with spike recovery samples. Sample test portions were incubated for 15 min at room temperature with shaking to facilitate homogeneous dispersion and adsorption into the matrix. Test samples were then processed according to the appropriate sample preparation procedure (Section 2.4), with all analyses performed in triplicate. Within-batch spike recovery quality control (QC) samples were additionally prepared using spike concentrations of 10 mg/g CBD and 100 μg/g CBDA, Δ⁹-THC, THCA, and CBN for oils/tinctures, or 5 mg/g CBD and 50 μg/g CBDA, Δ⁹-THC, THCA, and CBN for other products.

2.7. HPLC/DAD/FLD analysis

Analysis was performed according to Ciolino et al. (2018) with modifications using an Agilent 1290 Infinity series LC (Agilent Technologies; Santa Clara, CA) comprised of a binary pump, high-performance autosampler with thermostat, column compartment, DAD equipped with a 60 mm path length/4 μL flow cell, and fluorescence detector (FLD) equipped with a standard 8 μL flow cell. An ACE Excel 3 C18-AR column (150 mm × 2.1 mm ID, 3 μm particle size; Advanced Chromatography Technologies LTD; Aberdeen, Scotland) was used for analysis at 25 °C with pre-heating. The autosampler was held at 8 °C, with 2 μL sample injections followed by a 5 s wash with 70 % ethanol. Isocratic conditions of 33 % water with 0.5 % acetic acid (A) and 67 % acetonitrile (B) at 0.3 mL/min were held for 16 min before a 1 min ramp to 95 % B with a 1 min hold before re-equilibration at 67 % B for 6 min. UV/Vis absorbance was monitored at 220 nm, 240 nm, 270 nm, and 307 nm, with absorption at 240 nm and 270 nm used for quantification. Fluorescence data were acquired across four excitation/emission pairs: 234/311 nm (0–8.5 min), 261/378 nm (8.5–11 min), 234/315 nm (11–13.5 min), and 272/346 nm (13.5–24 min). A cannabinoid calibration check standard and blank were analyzed every ten samples to monitor system performance. Every twenty samples, a within-batch spike recovery QC sample and triplicate analysis of a test sample were analyzed to monitor extraction performance and method precision. A full external standard curve was run at the start of each batch of samples.

2.8. Data analysis

Acquired data were exported from Agilent Chemstation into Agilent MassHunter Quantitative Analysis using the Agilent LC-SQ Translator tool. Data were integrated and unweighted linear standard curves were generated across seven points ranging from 0.05 to 25 μg/mL for Δ⁸-THC, CBC, CBG, CBGA, CBDV, and THCV; across seven points ranging from 0.5 to 250 μg/mL for THCA, CBDA, CBN, and Δ⁹-THC; and across nine points ranging from 0.5 to 1,000 μg/mL for CBD. Peak integrations were subsequently quantified against generated external standard curves. Cannabinoid concentration within products was calculated by multiplying the measured concentration of an extract against the sample dilution factor.

2.9. Determination of toxic elements by ICP-MS

Total As, Cd, Hg, and Pb concentrations were determined using microwave assisted digestion followed by inductively coupled plasma mass spectrometry (ICP-MS) as described in FDA Elemental Analysis Manual (EAM) method 4.7 (Gray et al., 2020). EAM method 4.7 is a multi-lab validated method across all sectors of the AOAC food triangle for multiple elements, including toxic elements in food products (Gray and Cunningham, 2019). Separate aliquots of previously homogenized sample composites, described in section 2.3, were used for toxic element analysis. Homogenized composites were brought to room temperature and vortexed or mixed by hand prior to sample preparation. Duplicate analytical portions were prepared by weighing 0.25–3.0 g analytical portions, dependent upon the sample type, into Teflon microwave digestion vessels with 5 mL of HNO₃ (67 %) and 1 mL of H₂O₂ (30 %). Samples were digested in a Milestone UltraCLAVE microwave system (Milestone Inc; Shelton, CT) by ramping the temperature to 250 °C for a minimum of 30 min and holding the temperature at 250 °C for an additional 15 min. After samples cooled, 2.5 mL of 10 % (v/v) HCl was added before samples were gravimetrically diluted to 50 g with 18.2 MΩ water (Millipore system; Bedford, MA).

Concentrations of As, Cd, Hg, and Pb were quantified using an Agilent 8800 ICP-QQQ-MS (Agilent Technologies; Santa Clara, CA) using external calibration and kinetic energy discrimination (He mode) for polyatomic interference attenuation (Supplemental Table 4.1). Internal standard solutions (Ge, Rh, Ir, and Bi) and six working calibrators ranging from 0.05 to 25.0 μg/kg (As, Cd, and Pb) and 0.005 to 2.5 μg/kg (Hg) were gravimetrically prepared from stock multi-element solutions. Method performance was verified through the use of standard reference materials (SRM; specified in Supplemental Table 4.2) purchased from the National Institute of Standards and Technology (NIST; Gaithersburg, MD), initial calibration verification, continuing calibration verification, relative percent difference of duplicate sample portions, method blanks, and fortified sample portions falling within acceptable ranges when above LOQ (Gray et al., 2020). SRM recoveries and limits of detection (LOD) and quantification (LOQ) can be found in the supplemental materials (Supplemental Tables 4.2 & 4.3).

3. Results and discussion

3.1. HPLC method selection and modifications

Recognizing the immense public interest around hemp-containing products and the proliferation of methods for the analysis of hemp-containing products, we determined to adopt an established method requiring only minimal matrix extension validation, rather than validating a new in-house method. Although numerous methods are available for the analysis of Cannabis plant material and extracts, few have been extensively validated for complex food matrices such as gummies and edibles (Welling et al., 2019). A notable exception is the method of Ciolino et al. (2018), in which the authors developed and validated an HPLC/DAD method for the analysis of a wide range of cannabinoid-containing products. This method chromatographically separated eleven cannabinoids over 50-min using a 4.6 × 250 mm MacMod ACE C18-AR column packed with 5 μm particles. To improve throughput, column dimensions were scaled to a 2.1 × 150 mm MacMod ACE C18-AR column packed with 3 μm particles and a flow rate of 0.3 mL/minute, enabling a runtime of 24 min/sample while maintaining resolution of ≥ 1.2 for the two critical pairs of CBG-CBD and Δ⁹-THC-Δ⁸-THC (Fig. 2).

Fig. 2.

UV chromatogram (240 nm) demonstrating separation efficiency for the targeted cannabinoids in an ethanolic standard (top; 100 μg/mL major cannabinoids, 10 μg/mL minor) and in a representative oil sample containing 62.8 mg/g CBD, 0.91 mg/g Δ⁹-THC, and 0.48 mg/g CBC.

3.2. Fluorescence detection of neutral cannabinoids

Numerous studies have combined LC with optical absorbance spectroscopy or MS for the determination of cannabinoids, however the use of fluorescence detection (FLD) for this application has only been sparsely reported (Citti et al., 2018). Neutral cannabinoids are known to fluoresce under the acidic conditions commonly used in reverse-phase LC analysis of cannabinoids (Hazekamp et al., 2005), suggesting that FLD could offer a complementary, low-cost, and selective detection method without the expense and complexity of MS instrumentation and analysis. However, FLD is limited by its inability to detect cannabinoid acids such as CBDA and THCA under low pH conditions (Hazekamp et al., 2005). This limitation restricts the use of FLD to that of a complementary detection method except in cases where cannabinoid acids are not of concern, or where analysis is conducted at basic pH. Within this study, fluorescence data were collected as a complement to the primary UV/DAD data and demonstrated acceptable linearity for the major neutral cannabinoids, albeit with a narrower linear range than DAD due to saturation at high concentrations (Supplemental Table 3.5). This limited linear range was most notably seen with CBN. The challenge of FLD saturation for high concentration cannabinoids could likely be rectified by de-tuning the emission channel if FLD were to be adopted as a primary detection technique. Demonstrating the application of this alternative detection technique, FLD successfully resolved DAD interferences for Δ⁹-THC in a handful of samples without requiring orthogonal separation, further sample clean-up, or analysis by MS, and was used for quantification in cases where clear interferences were present in DAD signal (Supplemental Fig. 5.1).

3.3. Method verification for cannabinoids

Method performance was verified prior to the analysis of marketplace samples. Spike recovery experiments were performed for CBD, CBDA, Δ⁹-THC, and THCA, and CBN at two concentration levels (Supplemental Table 3.1) in seven representative matrices. Representative matrix blanks (Supplemental Table 3.2) were analyzed concurrently with spike recovery samples and determined to be devoid of cannabinoids except for hemp seed, which contained trace-level CBDA (7.93 ± 0.11 μg/g) at concentrations below the method LOQ of 10 μg/g (determined by Ciolino et al. (2018). Average recovery of spiked cannabinoids was 103.3 % (range: 95–123 %) for high-concentration spiked samples, and 95.6 % (range: 59–119 %) for samples spiked at the method LOQ (Supplemental Tables 3.3 and 3.4). Average percent relative standard deviations were 2% (range: 0–9 %) for high-concentration spiked samples, and 11 % (range: 2–54 %) for samples spiked at the method LOQ across all cannabinoids in each test matrix. Several samples, most notably coffee beverage, demonstrated recovery values outside an acceptable range of performance (70–130 %) at 10 μg/g due to co-eluting matrix interferences. As this level was far below typical labeled cannabinoid claims and a local LOQ re-determination was not performed, the potential for interference in some matrices, especially coffee brew, at levels near the LOQ was noted. Due to the randomized nature of product survey sample selection, no coffee beverages were analyzed during the course of the product survey, and thus recovery from coffee beverage did not impact the reported data. Gummy and hemp seed matrices were seen to have poor precision at the method LOQ, which was theorized to be due to homogeneity issues during sample extraction and chromatographic interferences. Precision was monitored with replicate samples during analysis of survey samples, with relative standard deviations observed to be below 10 %.

Additional method verification parameters included linearity, accuracy, absence of carryover, and selectivity. DAD analysis of CBD over the concentration range of 0.5–1000 μg/mL and CBDA, THCA, Δ⁹-THC, and CBN over the range of 0.5–250 μg/mL was highly linear (R² ≥ 0.9990; Supplemental Table 3.5). Linearity of the minor cannabinoids (Δ⁸-THC, CBC, CBG, CBGA, CBDV, and THCV) was also verified over the concentration range of 0.05–25 μg/mL (data not presented). As an addendum to spike recovery studies above, accuracy was additionally verified by quantification of CBD in a sample of Epidiolex. The analyzed content of CBD was 104.9 mg/mL, corresponding to a measurement error of 5.4 % from the certified value of 99.5 mg/mL. The absence of carryover was verified by injection of a 250 μg/g standard of major cannabinoids (equivalent to the analysis of a 50 mg/g oil sample), followed by the analysis of a blank sample. No cannabinoid carryover peaks were detected. Carryover performance was additionally verified through the analysis of a blank sample every ten samples during the survey, with no carryover observed in the acquired blanks. Selectivity was verified by qualitative LC–MS analysis of a subset of samples with above-average concentrations of CBD or Δ⁹-THC, or in which coelutions were present (n = 18; greater than 1.5 mg/g Δ⁹-THC or 24 mg/g CBD). The presence of CBD and Δ⁹-THC were confirmed based on matching retention times and product ion ratios between the samples and standards injected at time of analysis, and the method was deemed to be appropriately selective (Supplemental 6).

3.4. Product survey sample selection

Of the 526 products initially identified, 201 were randomly selected for purchase, with only 147 of the selected products analyzed for cannabinoids and 133 products analyzed for toxic elements due to restricted laboratory access during the SARS-CoV-2 pandemic. The majority of analyzed products (56 %) belonged to the tinctures/oils category (Fig. 3). Gummies, edibles, and pet products were also common whereas beverages and powders made up a smaller proportion of the product list. Isolates were not included among the products selected for analysis. While not necessarily a representative sampling, this distribution is similar to a recent survey of CBD consumer preferences, in which 48 % of consumers surveyed reported using gummies and edibles, 43 % reported using tinctures, and 29.5 % reported using capsules and softgels (The CBD Insider, 2019).

Fig. 3.

Breakdown of products by category. Tinctures & oils comprised the majority of tested samples at 56 % of the total.

3.5. Cannabinoids

The total cannabinoid content across all products ranged from < LOD – 143 mg/serving (LOD was matrix dependent; 0.004–0.04 mg/g), with a median content of 16.7 mg/serving (Fig. 4; tabular data as Supplemental 7). Products in the tinctures/oils category contained the highest median content of total cannabinoids at 23 mg/serving. Research on CBD consumption habits of young adults suggests consumers are confused regarding CBD dosage (Wheeler et al., 2020), a finding which may be generalizable to CBD consumers as a whole. While further research into CBD may clarify effective intake levels, at present it is speculated that the wide range in cannabinoid contents observed in this study are due to product vendors responding to consumer demand for an expansive variety of CBD concentrations and quantities.

Fig. 4.

Cannabinoid composition of surveyed products. CBN, CBC, CBGA, CBDV, THCA, Δ⁸-THC, and THCV were grouped together as other cannabinoids, as they were present only infrequently and at low concentrations.

As expected, CBD was the most abundant cannabinoid by mass, averaging 94 % of the total (15.7 mg/serving), although seven samples (5%) contained higher levels of CBDA than CBD, and the cannabinoid profile of one edible sample containing 51 mg total cannabinoids per-serving was almost entirely CBG. The cannabinoid composition of 46 samples (31 %) was comprised of >98 % CBD, suggesting that these products may have been produced using CBD isolates or highly purified extracts rather than crude extracts. While 21 of the 46 presumed-isolate samples belonged to the tinctures/oils category, nearly half of the gummy and edibles samples (47 % and 53 %, respectively) contained a cannabinoid profile of >98 % CBD. Although not a representative sampling of the marketplace, these findings suggest that the use of CBD isolates may be more common in these product types.

On average, Δ⁹-THC comprised 2% of the cannabinoid content by mass. Due to historic limitations of GC around the analysis of cannabinoid acids (Citti et al., 2018), and to account for conversion of THCA to THC during heating, THC content is often reported as “total THC,” the combined mass of THC and 0.877*THCA (accounting for decarboxylation during heating). In a majority of samples the concentration of total THC measured below LOQ (matrix dependent; 0.01–0.1 mg/g); for samples above LOQ, the median total THC content was 0.6 mg/serving. The highest total THC samples tested were primarily oils/tinctures and pet products, however a fruit gummy and a cookie edible were also found to contain above-average concentrations of total THC per serving, at 1.0 and 2.3 mg/serving total THC, respectively. A hemp powder sample was found to contain the highest total THC content, at 3.2 mg/serving, however this was made up almost entirely of THCA. The highest Δ⁹-THC sample, an oil, contained 2.7 mg/serving total THC, below the typical recreational intake of 5–25 mg (Curran et al., 2002). Forty-three products (29 %) claimed to be THC-free or broad-spectrum extracts; of these, 8 samples (19 %) contained quantifiable amounts of total THC with a median concentration of 0.18 mg/serving. One purportedly THC-free tincture sample contained 1.0 mg total THC per serving. While above the stated claim, this level was still far below the levels typically used recreationally. Based on the samples selected in this study, these results suggest that marketed hemp and CBD products generally conform to guidelines regarding Δ⁹-THC content laid out in the 2018 Farm Bill, likely due to USDA and DEA regulation of hemp production and the legal importance of compliance.

Of the 147 products analyzed, 102 (69 %) carried a label claim for CBD. As a regulatory standard for labeling accuracy of CBD does not exist, a threshold of ±20 % of the claimed amount was used to facilitate discussion but does not carry regulatory significance. More than half of the products (55 %) contained CBD at concentrations outside ±20 % of the indicated CBD content, with a bias towards under-labeling (Fig. 5). These results are similar to those of several smaller studies. A recent survey of 29 veterinary CBD products found 34 % of products were within 90–110 % of the CBD concentration indicated, with 31 % of products containing more than 110 % of the claimed amount, and 34 % with less than 90 % of the label claim (Wakshlag et al., 2020). In another survey of 84 oils, tinctures, and vaporization liquids purchased online, 43 % of product labels were found to under-report CBD quantities, with 31 % over-reporting, and 26 % accurate to within 10 % of the labeled value (Bonn-Miller et al., 2017). A study of 15 European CBD oils purchased via the internet found 36 % of product labels to be accurate within 10 % of the labeled value, with 29 % of products over-reporting the amount of CBD, and 36 % of products under-reporting the amount of CBD (Pavlovic et al., 2018). Finally, a 2020 survey of 59 CBD-containing products found 58 % of products to be outside of ±20 % of their indicated CBD content, with a bias towards over-labeling and an average labeling discrepancy of 30 % (Digital Citizens Alliance, 2020). From these studies, and from the results presented here, it can be postulated that CBD content claim accuracy of some commercial products may be variable, to the potential detriment of consumers seeking consistent intakes.

Fig. 5.

CBD content relative to product label claim.

3.5.1. Toxic elements

Hemp has been reported as a bio-accumulator of toxic elements (Angelova et al., 2004) and has previously been used for phytoremediation of contaminated soil (Linger et al., 2002). Because of its propensity to uptake toxic elements, there is concern that products produced from hemp grown in areas with contaminated soil have the potential to contain high levels of toxic elements (Mead, 2017), especially in concentrated products such as CBD oils. To assess this risk, 133 products were additionally analyzed for toxic elements, including 72 tinctures/oils, 20 pet products, 17 gummies, 15 edibles, 8 beverages, and 1 powder.

Of the analyzed products, 30 (23 %) samples contained levels of one or more toxic elements above the LOQ. Table 1 shows the number of samples by analyte which contained levels < LOD, trace levels between the LOD and LOQ, and samples > LOQ. Only 5 of the 133 (3.8 %) samples analyzed contained quantifiable levels of As, including three edible products, one tincture/oil, and one powder sample. Arsenic concentrations in those five samples ranged from 12.3 to 33.0 μg/kg, with a median concentration of 18.3 μg/kg for samples > LOQ. Cadmium concentrations were quantifiable in 15 of the 133 (11.3 %) samples, including eleven edibles, three tinctures/oils, and one powder with a median value of 17.0 μg/kg and range from 4.45–377 μg/kg for the 15 samples > LOQ. The highest Cd concentrations found were in edible products, including chocolate which commonly contains Cd (Abt et al., 2018). Mercury was not quantifiable in any analyzed sample. Lead concentrations were quantifiable in 22 samples (16.5 %), with a median content of 23.3 μg/kg and ranged from 2.72–3360 μg/kg for the 22 samples > LOQ. Lead was found in the largest number of samples, including in each product category, and showed the highest maximum, mean, and median concentrations out of the four elements tested. One tincture contained 3360 μg/kg Pb; however, based off the suggested serving size, the calculated ingestion was 3.12 μg Pb/day, below the interim reference level of 12.5 μg Pb/day limit for women of childbearing age (Flannery et al., 2020). All other samples found to contain Pb were below 1000 μg/kg Pb.

Table 1.

Results of trace metals analysis. < LOD: Not detected. Trace: Trace amounts between LOD and LOQ. LOQ and LOD values are found in Table S3.3.

	As	Cd	Hg	Pb
n	133	133	133	133
<LOD	92	103	131	63
Trace	36	15	2	48
>LOQ	5	15	0	22

Lead was the predominant contaminant found in the analyzed samples, similar to results observed in several smaller studies. A US study of 29 commercially available veterinary hemp supplements found four samples to contain toxic elements above the LOD, with Pb being the prevalent contaminant (Wakshlag et al., 2020). A study of 29 over-the-counter CBD products available in the UK market also found low levels of toxic elements, with Pb having the highest concentrations (Liebling et al., 2020). Due to its limited sample size, data from the present study are not fully representative of hemp and CBD containing products in the US market, however it was found that levels of As, Cd, Hg and Pb were below the limit of quantification in the vast majority of samples in this convenience sampling.

4. Conclusions

A non-representative survey of the US CBD marketplace was performed, analyzing 147 hemp and hemp-derived products for cannabinoids and 133 products for toxic elements. A wide range in CBD concentrations was observed in profiled products, and it was found that fewer than half of the tested products which presented label claims contained CBD at concentrations within 20 % of their claimed amount. Toxic element and THC concentrations were below the limits of quantification in a majority of tested samples, with Pb concentrations below the interim reference level (IRL) for women of childbearing age across all tested samples, although more research is needed to confirm these trends. This work, along with the results of prior studies, further emphasizes the variability in labeling accuracy and concentration of CBD products in the US and supports the need for larger-scale representative studies to better characterize the marketplace.