Accuracy of labeled THC potency across flower and concentrate cannabis products

Abstract

High total tetrahydrocannabinol (THC) cannabis products are quickly gaining market share across the U.S., but it is unknown whether labeled cannabinoid content accurately reflects true product potency. This study aimed to assess discrepancies between labeled and observed THC content in flower and concentrate products. A representative sample of 277 cannabis flower (n = 178) and concentrate (n = 99) products were purchased from 52 Colorado dispensaries and analyzed for THC content. Products were classified as accurately labeled if observed THC was within ± 15% of labeled THC, and comparative analysis tested observed versus labeled THC content. Labeling accuracy for THC content depended on product type (X²(2, n = 277) = 47.44, p < 0.001), with 96.0% of concentrate products being within ± 15% of labeled THC content compared to 56.7% of flower products. Observed THC potency was significantly lower than labeled potency in both flower (U = 18,971, p = 0.001) and concentrate (U = 6095, p = 0.003) products. Nearly all tested concentrate products met the accuracy threshold for THC content, whereas flower products frequently did not. Both product types had lower observed THC content compared to labeled values. Independent and blinded verification of product labels and testing procedures is needed to protect consumers and patients and ensure rigorous testing standards that apply across various types of cannabis products.

Introduction

As cannabis legalization continues to spread across the United States, new and more potent cannabis products are widely available. The largest retail market share is represented by flower and concentrated forms, which can range up to 95% total tetrahydrocannabinol (THC)^1–3. Accurate potency labeling of these products is important for consumers, who may be medical patients titrating dosage, or recreational users expecting a certain effect based on the label. Cannabis product labels should accurately communicate potency to prevent accidental overdoses⁴ and reduce risk for conditions like psychosis and Cannabis Use Disorder, which are both associated with the use of high potency THC products^5–7. Aside from consumers, regulators and researchers in the public health space rely on the accuracy of product labels to better understand how potency influences potential risks and benefits to more accurately draft public health messages.

While such products are typically required by state laws to be tested by third-party labs and reported via regulated labeling procedures, discrepancies between federal and state cannabis laws have resulted in varied regulation and oversight, leading to previously reported evidence of inaccurate labeling^8–14. An analysis of publicly available cannabinoid potency data from state-certified laboratories in Washington State revealed systematic differences between testing facilities, with some facilities consistently reporting higher or lower potencies than others¹⁵. Further, case studies in Washington State have reported that THC potency inflation by third-party testing laboratories may be a recurring problem¹⁶. A study of 107 recreational flower products collected at random by law enforcement in California, Oregon, and Colorado found over 70% of products fell outside of a 20% accuracy threshold for THC potency. Of these inaccurately labeled products all but one were over-labeled (i.e. the label reported a higher THC potency than was observed via testing)¹¹. Over-labeling of THC potency has also been observed in oral cannabis oil products in Ontario, Canada¹², as well as in medical edible products purchased from San Francisco, California, Los Angeles, California, and Seattle, Washington¹⁴. Importantly, labeling discrepancies observed in the medical edible products differed across states, suggesting that specific state policies can significantly impact label accuracy¹⁴.

Colorado was the first state to legalize adult cannabis use and is often looked to as a model for other states navigating the intricacies of cannabis policy legalization. However, no comprehensive, independent, blinded product label audit has been conducted in Colorado or any other state legal market, and no study has addressed the question of label accuracy for high potency cannabis concentrates. Thus, our primary aim was to quantify THC in cannabis flower and concentrate products systematically purchased from Colorado retail dispensaries to compare with the labeled potency. We hypothesized that there would be deviations from the labeled potency, most prominently in the flower product category. We also sought to quantify the minor cannabinoids not required to be reported on product labels. Cannabidiol (CBD) was selected as the comparator for these minor cannabinoids. Cannabidiol, like THC, is required to be reported on product labels in Colorado, but it is frequently less abundant than THC in flower and concentrate products.

Results

A total of 281 cannabis products, including 182 flower and 99 concentrate products, were purchased between November 29th, 2022, and October 3rd, 2023. Products were purchased from 52 recreational dispensaries located in 19 different counties of Colorado (Fig. 1). Flower products included both loose flower (n = 171) and pre-rolls (n = 11) (Supplementary Figures S1 and S2). Concentrate products included 27 unique forms. Four purchased flower products were excluded from analysis due to misprinted or absent THC potency labeling.

Fig. 1.

Heat map of cannabis product purchase locations. Flower and concentrate products were purchased from 52 recreational dispensaries across Colorado to be representative of current sale locations.

Mean labeled THC potency was 22.5% (SD = 4.67, range = 8.92–39.0) for flower and 73.0% (SD = 6.97, range = 38.0–83.7) for concentrate products. Mean observed THC potency was 20.8% (SD = 4.18, range = 11.7–33.0) for flower and 70.7% (SD = 7.53, range = 17.9–83.6) for concentrate products. The frequencies of labeled and observed THC potencies by product type are presented in Fig. 2. THC label accuracy depended on product type (X² (2, n = 277) = 47.44, p < 0.001), with observed THC for flower products being more frequently outside ± 15% of labeled THC than concentrate products. Of the 178 flower products, 101 (56.7%) were accurately labeled, 54 were over-labeled (30.3%), and 23 were under-labeled (12.9%). Of the 99 concentrate products, 95 (96%) were accurately labeled, 2 (2%) were over-labeled, and 2 (2%) were under-labeled. Observed THC potency was significantly lower than labeled potency for both flower (U = 18,971, p = 0.001) and concentrate (U = 6,95, p = 0.003) products. Mean difference between labeled and observed THC was 1.70% (SD = 4.94, range =  − 8.92–17.9) for flower and 2.35% (SD = 4.90, range = − 12.3–20.1) for concentrate products. Mean percent difference between labeled and observed THC was 5.19% (SD = 20.2, range = − 53.8–58.4) for flower and 3.20% (SD = 8.05, range = − 17.6–52.8) for concentrate products (Fig. 3).

Fig. 2.

Distribution of labeled (a) and observed (.) THC by product type.

Fig. 3.

Percent difference between labeled and observed THC by product type.

Cannabidiol potency was included on the label of 143 (80.3%) flower and 84 (85%) concentrate products. Of the products reporting CBD potency, 124 (86.7%) flower and 67 (80%) concentrate products reported a CBD potency of 0%. Eighty-four percent of flower and concentrate products did not report any values on the label for the minor cannabinoids cannabichromene (CBC), cannabigerol (CBG), cannabigerolic acid (CBGA), cannabinol (CBN), delta-8 THC, and delta-9 tetrahydrocannabivarin-(THCV). Analysis of minor cannabinoids (Table 1) showed CBG and CBGA to be more abundant than CBD in both flower (U = 1179, p < 0.001 and U = 162, p < 0.001, respectively) and concentrate (U = 963, p < 0.001 and U = 3767, p = 0.005, respectively) products.

Table 1.

Observed CBD and minor cannabinoids. Values are mean (SD) [range]. CBD is required to be included on cannabis product labels in Colorado, while the minor cannabinoids are not. Wilcoxon rank sum tests compared CBD potency with the minor cannabinoid potency by product type. ^aSignificantly lower (p < .05) than CBD. ^bSignificantly higher (p < .05) than CBD.

Cannabinoid	Flower (n = 178)	p (Flower)	Concentrate (n = 99)	p (Concentrate)
CBD	0.05 (0.02) [0–0.12]		0.25 (0.56) [0–4.23]
CBC	0.04 (0.04) [0–0.35]a	< .001	0.40 (0.49) [0–1.89]	.68
CBG	0.17 (0.10) [0.05–0.87]b	< .001	1.42 (1.45) [0–6.51]b	< .001
CBGA	0.82 (0.57) [0.02–3.51]b	< .001	1.12 (1.34) [0–5.52]b	.005
CBN	0.02 (0.09) [0–1.10]a	< .001	0.59 (0.88) [0–4.82]	.94
Delta-8 THC	0.03 (0.06) [0–0.29]a	< .001	0.06 (0.20) [0–0.98]a	< .001
THCV	0.005 (0.017) [0–0.126]a	< .001	0.26 (0.30) [0–1.36]	.50

Discussion

The present study is the first to evaluate the accuracy of THC potency labeling in high-potency cannabis concentrate products, and is the largest sample to date examining THC potency labeling in cannabis flower. A representative sample of these products was purchased from recreational dispensaries across Colorado, which is an exemplar state legal cannabis market. Nearly all (96%) concentrate products were found to have accurately labeled THC potency, whereas flower products were more likely to fall outside of the ± 15% accuracy threshold, with only 56.7% being considered accurately labeled. Both product types were found to have lower observed THC potency compared to the labeled values.

The differences observed in labeling accuracy between flower and concentrate products may be related to the sampling procedures employed prior to analysis. Plant material by nature would be less homogenous and more susceptible to discrepancies caused by imperfect sampling in comparison to concentrates, which are homogenized during production. The greater magnitude of THC content in concentrates compared to flower products may also account for the observed differences between product types, as greater absolute differences are required to achieve the 15% discrepancy. The threshold of 15% was selected based on current Colorado guidelines for potency labeling accuracy¹⁷.

Our findings that observed THC potency was lower than labeled potency are consistent with those of Schwabe et al.¹³, who sampled 23 flower products in Colorado with an average discrepancy of 23.1% to 35.6%. The smaller observed discrepancy of 5.19% in flower products in our study may be due to the larger (n = 178 flower products) and more representative sample of products purchased across the state. The observed over-reporting of cannabinoid concentrations may be the result of the rapid growth of the cannabis market following state legalization and inconsistent regulatory standards across the state and federal level^14,18, as well as consumer demand for high-potency products.

We also observed that the minor cannabinoid, CBG, was more abundant than CBD in both flower and concentrate products. CBG, like CBD, is non-psychoactive and is the precursor to other cannabinoids. However, unlike CBD, CBG it is not currently mandated to be included in product potency testing or labeling. The present findings warrant consideration for the inclusion of CBG in labeling requirements and expanded research assessing its potential health impacts with acute and long-term use.

In summary, this study found THC potency labeling to be accurate within ± 15% for almost all tested cannabis concentrate products, whereas flower products more frequently did not meet this accuracy threshold. Both product types reported higher THC potency values on the label than what was observed. Continued investigation into labeling inaccuracies and more consistent, rigorous testing moving forward is indicated, particularly for flower products, across various legal cannabis markets.

Methods

This study was a collaboration between MedPharm Research LLC, a licensed cannabis testing facility, and a research team at the University of Colorado Boulder. MedPharm Research, LLC holds a state Marijuana Research and Development license (408C-00001) and City of Denver R&D Co-Location Permit (RDCLP-00001), which are licensing and permitting mechanisms designed to encourage and support industry engaged cannabis research. This project was approved by the State of Colorado’s Medical Marijuana Scientific Advisory Council (SAC) and funded by the Institute of Cannabis Research at Colorado State University Pueblo.

Purchasing

A member of the MedPharm research team, blind to the results of the study’s cannabinoid potency analysis, purchased cannabis flower and concentrate products from licensed recreational dispensaries in Colorado. The purchaser entered product label information and uploaded photos of the product and all layers of its packaging (i.e., outer marking layer, potency labels, and secondary internal packaging) into a Qualtrics survey managed by researchers at the University of Colorado Boulder. The purchaser then transferred the product to a container that was labeled only with a sample number. Labeled samples were stored in the dark in a locked cabinet in a temperature-controlled room. Samples were tested an average of 20.3 days (SD = 15.8, range = 2–62) after purchase.

Cannabinoid potency analysis

A second member of the MedPharm research team, blind to the product label information, conducted triplicate analysis of cannabinoid potencies using high-performance liquid chromatography (HPLC). Certified reference materials of cannabidiolic acid (CBDA), cannabigerolic acid (CBGA), cannabigerol (CBG), cannabidiol (CBD), tetrahydrocannabivarin (THCV), cannabinol (CBN), delta-9-tetrahydrocannabinol (d9-THC), delta-8-tetrahydrocannabinol (d8-THC), cannabichromene (CBC), and delta-9-tetrahydrocannabinolic acid (THCA) were purchased from Restek for calibration curves (Supplementary Table S3). Formic acid (98–100%) was purchased from Sigma–Aldrich. HPLC-grade methanol and water were purchased from Pharmco by Greenfield Global, and HPLC grade acetonitrile was purchased from TEDIA. Flower products were homogenized by placing the whole sample (3.5 g) in a 50 mL conical tube and cooling in a − 20 °C freezer for at least 2 h. The sample was then ground by hand in the conical tube using a scoopula until homogenized. Three aliquots of 100–200 mg each were then collected and weighed in fresh 50 mL conical tubes. Aliquots were extracted by adding 40.00 mL of methanol via an automatic tabletop dispenser to each, vortexed for 10 s and then heated for 30 min at 40 °C in a hot water bath. Samples were vortexed for 10 s every 10 min during the heating process, and once after. Samples were then centrifuged at 4000 RPM for 10 min and an aliquot was transferred off the top to a 2 mL HPLC vial for analysis. Dilutions of 1:4 were also prepared using methanol as the diluent. Concentrates were prepared in the same manner as flower, without the need for homogenization (no freezing or grinding prior to weighing). Dilutions of 1:10 were prepared for concentrates in place of 1:4 dilutions for flower. HPLC analysis was performed on an Agilent 1290 Infinity system (see Supplementary Table S4 for complete methods). A gradient method using 0.1% formic acid in water (mobile phase A) and 0.1% formic acid in acetonitrile (mobile phase B) resulted in sufficient baseline separation for all 10 cannabinoids and was validated according to ISO 17,025 standards for the following parameters: specificity, linearity, range, limit of quantification, limit of detection, and precision. Recovery percentages of d9-THC, THCA, CBD, and CBDA from flower were 93.14%, 96.04%, 95.65%, and 99.15%, respectively. Recovery percentages of d9-THC, THCA, CBD, and CBDA from distillate were 87.98%, 94.03%, 92.82%, and 104.70%, respectively. Results of the cannabinoid potency analysis were transferred to the research team at the University of Colorado Boulder using a secure web platform.

Labeled THC

Products purchased in the present study were found to be labeled with the terms “Total THC” (37.5% of products), “THC” (43.7% of products), or both “Total THC” and “THC” (18.8% of products). “Total THC” was used as the labeled THC value for statistical analysis when present on the label, and “THC” was used when “Total THC” was not present. Labeled THC values included both single values (e.g., “18.5%”; 91.3% of products) and ranges of values (e.g., “16.7–22.4%”; 8.7% of products); see Statistical Analysis section for information on how this was handled for accuracy and comparison purposes.

Observed THC, CBD, and minor cannabinoids

Total THC was calculated using the industry standard equation Total THC = d9-THC + (0.877 X THCA) following cannabinoid potency analysis. Total CBD was calculated using the industry standard equation Total CBD = CBD + (0.877 X CBDA). The mean of the triplicate values for the calculated Total THC and Total CBD were used as the observed values for THC and CBD, respectively, for statistical analysis. The mean of the triplicate values for the minor cannabinoids was used as the observed values for each respective minor cannabinoid.

Statistical analysis

Consistent with the state of Colorado standards that allow a 15% deviation between labeled and actual potency, products were classified as accurately labeled if observed THC was within ± 15% of the labeled value, over-labeled if observed THC was more than 15% below the labeled value, or under-labeled if observed THC was more than 15% above the labeled value. For products labeled with a range of THC potency, products were considered accurately labeled if observed THC potency was less than or equal to 15% of the maximum of the labeled range and greater than or equal to 15% of the minimum of the labeled range, over-labeled if observed THC was more than 15% below the minimum of the labeled range, or under-labeled if observed THC was more than 15% above the maximum of the labeled range. Although Colorado was the first to develop a legalized state market, a deviation threshold of 15% or less is a common labeling regulation across many legal state markets.

The cannabinoid potency data were found to be non-normally distributed, so Wilcoxon rank sum tests were used to compare observed and labeled THC for each product type, and to compare cannabidiol (CBD) with the minor cannabinoids cannabichomene (CBC), cannabigerol (CBG), cannabigerolic acid (CBGA), cannabinol (CBN), Delta-8 THC, and tetrahydrocannabivarin (THCV). A chi-squared test was used to evaluate the effect of product type on label accuracy. For products labeled with a range of THC potency, the mean of the reported range was used for descriptive statistics and for the rank sum tests. Percent difference between labeled and observed THC was calculated as ((labeled THC – observed THC) / labeled THC) * 100. Statistical significance was determined by a p value of less than 0.05. All analyses were conducted in R¹⁹ version 4.3.2 using the ggplot2²⁰ package for figures.

Author contributions

L.C.B. and D.I.M. designed the research. G.G., C.B., M.O.T., G.M., and D.M. performed the research. G.G. analyzed the data. All authors wrote and reviewed the manuscript.

Data availability

The data analyzed for the present study are available on the CU Scholar repository (https://scholar.colorado.edu/concern/datasets/d504rn329).

Declarations

Competing interests

D.I.M. and C.B. were employees of MedPharm Holdings (dba Bud & Mary’s) during the study period. MedPharm Holdings is a U.S. Drug Enforcement Administration (DEA) and State of Colorado licensed cannabis research company and a bulk manufacturer of cannabis products for the Colorado adult-use marketplace. MedPharm Holdings has no stake in any products tests in this study. G.G., M.O.T., G.M., C.S., J.L., and L.C.B. declare no competing interests.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-03854-3.