Self-report methodology for quantifying standardized cannabis consumption in milligrams delta-9-tetrahydrocannabinol

ABSTRACT

Background: There is currently no format-independent method to determine delta-9-tetrahydrocannabinol (THC) in milligrams for self-report studies.

Objectives: Validate self-report method for quantifying mg THC from commercially available cannabis products using product labeling, which includes both net weight and product potency.

Methods: 53 adult cannabis users (24 M, 29F), 21–39 years of age (M = 28.38, SD = 4.15), were instructed to report daily use via a weekly survey for two consecutive weeks, provide product label photographs, abstain from use for 24 h, submit a urine sample and complete the Cannabis Use Disorder Identification Test – Revised (CUDIT-R) and the Marijuana Craving Questionnaire – Short Form (MCQ-SF). Milligrams of THC were determined by multiplying quantity of product used by its THC concentration. Urine was analyzed for the urine metabolite 11-nor-carboxy-THC (THC-COOH) via liquid chromatography mass spectroscopy. THC and THC-COOH values were log10 transformed prior to correlational analyses.

Results: Median daily THC consumption was 102.53 mg (M = 203.68, SD = 268.13). Thirty-three (62%) of the 53 participants reported using two or more formats over the 2-week period. There was a significant positive correlation between log10 THC-COOH and log10 THC mg (r(41) = .59, p < .001), log10 THC mg and MCQ-SF score (r(41) = .59, p < .001), and log10 THC mg dose and CUDIT-R score, (r(41) = .39, p = .010).

Conclusion: Our label-based methodology provides consumption information across all modalities of cannabis use in standard units that can be combined across products for calculation of dose. It is a viable and valid method for quantifying mg of THC consumed and can be utilized in any region where cannabis is legal, and labeling is regulated.

KEYWORDS:

• Marijuana
• dose
• self-report
• quantification
• bioavailability
• THC-COOH
• THC
• milligram

Introduction

With legalization of recreational cannabis in an increasing number of states and product development within the cannabis industry, standard cannabis use patterns have changed. Smoking cannabis as a joint was the primary mode of consumption previously; however, use of concentrates, edibles, and topical forms of cannabis are becoming increasingly popular, with users mixing-and-matching forms consumed. Simultaneously, the psychoactive compound delta-9-tetrahydrocannabinol (THC) potency has increased over the last several decades. Cannabis flower THC concentration in 1980 was typically 2% but has increased an order of magnitude to over 20% in the intervening years (1–3). Additionally, overall average potency of products consumed has increased predominantly due to concentrates which average over 50% THC (2).

Strategies currently available to quantify cannabis use in observational self-report studies are typically retrospective, and measure frequency of use without accounting for variation in potency or consumption method. What constitutes the dose for “any use” can vary widely depending on the user and their tolerance level. Other strategies have used intoxication as a proxy for quantity of psychoactive THC consumed (4). While intoxication quantification allows for all forms or potencies to be compared, it does not account for development of tolerance, which varies across individuals and is dependent on history of use (5). The importance of factoring in potency of the product used has been recognized by the scientific community for decades. For example, Day (6) applied a correction factor to account for known differences in relative amount of THC in hash cigarettes (1 hash cigarette/bowl = 3 “joints”) and sinsemilla (1 sinsemilla joint = 2 “joints”). However, modern cannabis products require an updated approach. A recently developed retrospective measure, “Daily Sessions, Frequency, Age of Onset, and Quantity of Cannabis Use Inventory” (DFAQ-CU) (7) has made strides in addressing the complexity of modern cannabis use. Estimates of potency are recorded and the quantity of cannabis product used is measured in grams for flower and mg for edibles; however, concentrates are not fully addressed and topicals are not considered at all. To address the need for standardization in cannabis research (8), the National Institutes of Health recently adopted milligrams as the units for THC research. While this is straightforward to adopt for drug administration trials, the implementation is not clear for observational studies.

In response, we developed a format-independent self-report methodology to quantify THC consumption in milligrams. Legalization has resulted in product labeling of cannabis with consumer product information that can be utilized by researchers to quantitatively monitor usage. Product labels typically provide the total net weight in grams, along with potency as a percent of total net weight. This information allows calculation of the dosage, in milligrams, consumed by research participants, enabling comparisons between studies. We additionally assessed the validity of our measure by correlating reported doses with two psychometric inventories, the Cannabis Use Disorder Identification Test-Revised (CUDIT-R) (9), which assess severity of problematic cannabis-related behaviors, and the Marijuana Craving Questionnaire-Short Form (MCQ-SF) (10), which evaluates cannabis craving, as well as to a noninvasive biometric, urine THC metabolite, 11-Nor-9-carboxy-Δ9-THC (THC-COOH). Developing a method to capture user self-dosing quantities is particularly important for observational studies in which investigator-administered cannabis would be unethical, such as dosing at very high levels, investigating use characteristics among participants having an existing psychiatric disorder or a family history of such disorders, or for recording use rates during pregnancy. This labeling approach provides new tools by which to assess and quantify cannabis use in the community and to study related outcomes.

Methods

Participants

A convenience sample was recruited from the Seattle metropolitan area via fliers distributed at local dispensaries from May to August 2022. Participants were eligible if they were 21–40 years of age, used cannabis at least 3 days per week on average, were willing to purchase their cannabis from a regulated dispensary during their participation, had internet access and a way to take photographs of product labeling and upload them to our database, and completed two consecutive diary entries. 85 individuals completed the screening and were eligible to participate. Of these, 57 participants (25 males, 32 females) started the diary portion of the study. Four participants provided only partial diary data and were excluded from further participation (see Table 1).

Table 1. Participant characteristics N = 53 (BMI, body mass index; CUDIT-R, cannabis use disorder identification test-revised; MCQ-SF, Marijuana craving questionnaire-short form).

	M(SD) or count
Sex (M:F)	24:29
Race
Caucasian	37
African American	4
Asian	4
American Indian	1
Bi-Racial	4
Other	3
Ethnicity
Hispanic/Latino	8
Not Hispanic/Latino	45
Age (years)	28.38 (4.15)
Highest Level of Education (years)	15.23 (1.85)
BMI	27.65 (8.94)
CUDIT-R	13.23 (5.72)
MCQ-SF	38.72 (14.10)

Study procedures were approved by the University of Washington Human Subjects Division Institutional Review Board and electronically signed, informed consent was obtained for all participants. Participants were compensated $5 for their initial training on how to estimate and report use, $5 for each weekly cannabis use diary submitted and $25 for an onsite visit which included a urine sample, for a total of $40.

Definitions

We define a “product” as a labeled and branded item purchased at a dispensary, “format” to indicate the categories of flower, concentrate, edible and topical, and “dose” to indicate total amount of THC or CBD consumed on a single day.

Cannabis use questionnaires

Following enrollment, training was provided to participants by the research team to determine their typical amount used of their preferred product format, based on fraction of their standard package Net Weight (Figure 1, Supplemental 1). For example: participant buys 3.5 g of flower, and typically rolls seven joints with that amount. In this case, the unit of use is a joint. They smoke one joint. Entering values into the equation in Figure 1, the participant would calculate 0.5 g for their answer. Participants were also provided a web link to the training sheet which provides instructions on how to determine use amount based on Net Weight for common product formats, conversion factors and other tips that would aid in correct use amount estimation (Supplemental 1). This is a novel estimation method. Prior user quantity estimation strategies rely on users comparing amounts to images of reference amounts (11–13). Training was done via video-conference or phone, typically lasted 5–10 min, and was personalized according to participant cannabis literacy. Participants reviewed the training sheet and talked through examples with staff. Those who had difficulty with the estimation were given additional examples related to their individual use patterns. This training process was repeated until participants could generate use estimates for relevant forms of cannabis. Participants were encouraged to contact team with any estimation questions. Participants tracked their daily use for two weeks, then weekly completed an online diary to record their cannabis use (Supplemental 2). Participants were free to record their per session/daily use in whichever way they found most convenient prior to completing the weekly survey.

Figure 1. Format of equation for calculation of daily use amounts.

The diary captured:

• Number of products used in the past week.

• Format of each product (flower, concentrate, edible, or topical).

• Route of consumption of each product (inhaled, ate, drank, sublingually, topically).

• Amount of each product used each day in grams (or servings if in edible format).

• Uploaded image of product label listing the THC and CBD content in either % or mg.

• Other drug, tobacco, or alcohol consumption.

Study data were collected and managed using REDCap™ electronic data capture tools hosted at University of Washington. The use diary was implemented as a REDCap™ survey requiring participants to enter numeric quantities for amount of product used and upload a picture of the product label. If the user did not have a picture, they were asked to estimate potency of the product used based on typical use. Participants were required to complete all survey fields for two consecutive weeks to progress to the in-person urine collection.

Participants also completed questionnaires pertaining to current and prior use of cannabis at the in-person visit. The CUDIT-R was used to assess severity of problematic cannabis-related behaviors. The MCQ-SF provided a measure of participants’ level of self-reported craving for cannabis at the time of the research visit. To evaluate difficulties with completing the cannabis use estimation and filling out diaries, participants also completed an Exit Interview questionnaire that included a series of questions to rate using a 5-point Likert scale (Supplemental 3).

Urine collection and quantification of metabolites

Participants were instructed to abstain from cannabis use for 24 h prior to submitting their urine sample. This conservative timing was selected to ensure THC-COOH metabolite elimination had reached a stable, constant level that reflected longer term consumption levels, as we are correlating this measure to average cannabis use over a two-week period. THC-COOH metabolite measurements are sensitive to the time interval between when cannabis was last used and sample was obtained, with shorter time intervals yielding higher THC-COOH metabolite measurements. THC-COOH metabolites decay in an exponential fashion as the time interval between last use and measurement increases (14). Six hours after intravenous cannabis consumption, plasma, and tissue have reached a pseudo-equilibrium, and subsequent elimination of metabolites is due to rediffusion from tissue and body fat back into the blood (15). The inhaled route has similar pharmacokinetics as injected, but oral administration adds several hours delay (16). Elimination of metabolites from tissue has a much longer time course and is used here as an estimate of regular use.

Urine samples were placed on wet ice for a maximum of 2 h prior to being frozen at −80 C for storage. Samples were analyzed for the metabolite THC-COOH via liquid chromatography-mass spectrometry (LC-MS) (Xevo-TQXS, Waters) at the University of Washington, School of Pharmacy Mass Spectrometry Center. Briefly, 500 µl of sample was combined with 5 µl of IS mix and vortexed, 50 µl of 10% formic acid in H20 was added and vortexed. 4 ml of diethyl ether:ethyl acetate:hexanes (7:2:1) was then added. Samples were vortexed and centrifuged for 10 min at 4000 rpm. Organic layer was removed, and step repeated. Samples were dried under N2 and reconstituted in 50 µl H2O/AcN 0.2% acetic acid to a 1:1 ratio and added to injection vial. 5 µl of sample was injected into the Water’s Xevo-TQXS. Samples were compared to cannabinoid standards, 1–500 ng. Samples below limit of quantification were excluded (n = 10). Four samples yielded concentrations above 500 ng/mL (prior to creatinine normalization). These samples were quantified using the same linear curve as those within range of the cannabinoid standards, noting the limitation that the assumption of linearity has not been validated for these higher concentrations. For creatinine quantification, samples were prepared by adding 10 µl of urine to 990 µl of K2PO4 mobile phase, UV analysis at 234 nM (TUV, Waters, in-line with Xevo-TQS HPLC,Waters).

Label processing

Cannabis package labeling is regulated by WA State (WAC 314-55-105), requiring Net Weight, trade name, concentration of THC, tetrahydrocannabinolic acid (THCA),Total THC, CBD, cannabidiolic acid (CBDA), and Total CBD. Because of inconsistent labeling practices across products and companies, label information was double entered from image and compared within REDCap to identify discrepancies. Following label validation, total THC was calculated. Total THC is defined as:

(1) $TotalTHC=THC+\left(0.877\right)\times THCA$(1)

Depending on which information was provided (“THC only,” “THC and THCA,” “THCA and Total THC,” “THC, THCA and Total THC,” or “Total THC only”), Eq. 1 was applied using R Statistical Software (v4.2.2; R Core Team 2022) to either calculate or verify Total THC value (metric of interest). If only THC was listed, it was assumed to be Total THC. CBD was handled similarly. The THC dose was then calculated as: (quantity of product used) * (Total THC) = mg THC, where the quantity of product used was obtained via the data diary, and Total THC is a percentage from the labeling.

Bioavailability

Accurately determining the amount of THC contained in the quantity of product used is a large step toward determining the bioavailable quantity of THC that enters the blood stream to induce physiological effects. The two primary sources of variation are the process of consumption and the route of administration. The process of consumption is looking at how efficiently the THC in the product is converted to a vapor form by burning flower, vaporizing concentrate, or vaporizing flower (17,18). The route of administration includes inhaled, ingested (eat/drink), and transdermal/topical. Each process and route has a range of efficiencies and rates for THC to enter the blood stream (16,18,19). Our measure captured route of administration but did not differentiate between burning flower or vaporing flower. We scaled formats for bioavailability using the method of administration efficiency constant (MAEC) scale factors described by Budney et al. (11), and other bioavailability scale factors found in the literature (16,18,19), to our per product mg THC data. Specifically a 30% scale factor was applied for flower (11,16,18–20), 50% for concentrate (11), 20% for edibles (19,21), and 10% for topicals (21).

Statistical analysis

The Shapiro–Wilks test was used to evaluate distribution normality. THC dose in milligrams (with and without bioavailability scaling) and creatinine normalized THC-COOH metabolites were log10 transformed to eliminate positive skew. The Pearson’s product–moment correlation and 95% CI between the various measures are reported (self-reported THC dose, CUDIT-R and MCQ-SF, bio-metric urine concentration of THC-COOH). Correlations were run both with and without bioavailability scaling. All analyses were performed using R Statistical Software (v4.2.2; R Core Team 2022).

Results

Product labels

Of the total 401 recorded products, 95% of participants provided photos, and the remaining 5% of users entered potency via recall or estimation based on typical purchasing patterns. Only 1% of photos were coded to have a damaged field (either THC Total, THCA, THC, CBD Total, CBDA, CBD) indicating the image region for that field was torn, smudged, or otherwise unreadable. None of the damaged fields for this sample precluded the determination of Total THC.

Average daily consumption and use patterns

Diary data from 53 participants, who together recorded 608 days of use, were analyzed. On average, participants consumed cannabis 6.18 days of the week (SD = 1.11). Average daily THC and CBD consumption over the two weeks prior to the urine sample (excluding the 14th day on which participants were instructed to abstain) was determined for each participant. Median THC consumption was 102.53 mg (M = 203.68, SD=286.13) across participants, and for CBD the median was 0.72 mg (M = 4.62, SD = 14.79); average consumption, however, varied widely by format (Table 2, Supplemental 4). Thirty-three (62%) of the 53 participants reported using two or more formats over the 2-week period. The most common combination of reported formats was the use of both flower and concentrate (10 out of 53 participants) (Figure 2a). The mean THC potency for concentrates used was 71% (SD = 23.09) and for flower 24% (SD = 11.73). Mean CBD potency was 4% (SD = 12.32) for concentrates and 0.6% (SD = 4.60) for flower. To evaluate typical use amounts of a given format, and how that varies across participants in our study, the average amount consumed of each format on days that the format is used was plotted. Use here is defined as over the span of a day. Days where a format was not used were excluded (Figure 2b).

Figure 2. (a) Venn diagram of recorded cannabis consumption format per participant over the two-week observation period (N = 53). (b) Violin plot depicting distribution of individual user dose by format, i.e., the average amount used of the specific format on the days that use of that format was reported. Zero entries reflect use of a format that contained CBD only; users that did not use the format were not included in the plot. THC consumption (in mg) by format on days used: flower (n = 40, M = 222.79, SD = 207.48), concentrate (n = 26, M = 145.40, SD = 280.04), edible (n = 27, M = 17.12, SD = 24.35), topical (n = 4, M = 5.38, SD = 4.87). (THC, delta-9-tetrahydrocannabinol; CBD, cannabidiol).

Table 2. Average daily THC and CBD consumption N = 53 (THC, delta-9-tetrahydrocannabinol; CBD, cannabidiol).

	THC (mg)		CBD (mg)
Format	Mean	(SD)	Mean	(SD)
All Formats Combined	203.68	(286.13)	4.62	(14.79)
Flower	143.00	(201.03)	1.74	(7.81)
Concentrate	57.91	(192.50)	0.95	(5.47)
Edible	2.66	(4.47)	1.42	(3.79)
Topical	0.12	(0.50)	0.50	(3.53)

Correlation of urine THC metabolites

Of the 53 total samples submitted, 10 (19%) were below level of quantification for MS-LC and were excluded from analyses. Pearson’s product–moment correlation was computed to assess the relationship between calculated log10 THC milligram dose and three correlates: creatinine normalized log10 THC-COOH concentration, MCQ-SF scores, and CUDIT-R scores. There was a significant positive correlation between log10 THC-COOH and log10 THC dose r(41) = .59, p < .001, 95% CI[0.35, 0.75], log10 THC dose and MCQ-SF score r(41) = .59, p < .001, 95% CI[0.35, 0.75], and the log10 THC dose and the CUDIT-R score r(41) = .39, p = .010, 95% CI [0.10, 0.62], See Figure 3a-c (Supplemental 5, Comparative non-parametric analyses).

Figure 3. Relationship between mean daily THC dose in milligrams and (a) creatinine normalized THC-COOH concentration, n = 43, r(41) = .59, p < .001, 95% CI[0.35, 0.75], (b) MCQ-SF scores, n = 43, r(41) = .59, p < .001, 95% CI[0.35, 0.75], and (c) CUDIT-R scores, n = 43, r(41) = .39, p = .010, 95% CI [0.10, 0.62]. To evaluate sample collection time sensitivity, a secondary analysis was performed, restricted to only compliant participants who reported that they abstained from cannabis use for at least 24 h prior to providing the urine sample (d-f), relationship between mean daily THC dose in milligrams and (d) creatinine normalized THC-COOH concentration, n = 32, r(30) = .64, p < .001, 95% CI[0.37, 0.81], (e) MCQ-SF scores, n = 32, r(30) = .66, p < .001, 95% CI[0.40, 0.82], and (f) CUDIT-R scores, n = 32, r(30) = .47, p = .006, 95% CI[0.15, 0.71]. There was a significant positive correlation in all cases. Gray shading indicates 95% CI. Raw mean daily THC dose plotted on log10 scaled axis presented for illustrative purposes, Note that all Pearson's correlations reported above were computed using log10 THC. (THC, delta-9-tetrahydrocannabinol; THC-COOH, 11-Nor-9-carboxy-Δ9-tetrahydrocannabinol; MCQ-SF, Marijuana Craving Questionnaire-Short Form; CUDIT-R, Cannabis Use Disorder Identification Test-Revised).

Secondary analyses were conducted to address the influence of time since last use, a potential source of error (14) in our data. Eleven of the 43 study participants (26%) with valid THC-COOH levels did not comply with abstaining from cannabis use 24 h prior to providing their urine sample. Non-compliant times ranged from 10.5–23.8 h with median of 17.2 h. Correlational analyses were restricted to participants who reported that they abstained from cannabis use for at least 24 h prior to providing the urine sample (n = 32; abstinence period Median = 36.2, range 24–140 h). There was a significant correlation between log10 THC-COOH and log10 THC dose r(30) = .64, p < .001, 95% CI[0.37, 0.81], between log10 THC dose and MCQ-SF r(30) = .66, p < .001, 95% CI[0.40, 0.82], and between the log10 THC dose and the CUDIT-R r(30) = .47, p = .006, 95% CI[0.15, 0.71]. See Figure 3d-f for restricted sample results.

For comparison purposes, we ran correlational analyses between the THC metabolites and our clinical measures on the restricted sample (n = 32). Significant correlations between THC metabolites and the cannabis use questionnaires (Figure 4) were observed, indicating that our self-report methodology has similar predictive validity to our biological marker of THC use. There was a significant positive correlation between log10 THC-COOH and MCQ-SF score r(30) = .63, p < .001, 95% CI[0.36, 0.80], and between log10 THC-COOH and the CUDIT-R score r(30) = .35, p = .005, 95% CI[0.01, 0.63].

Figure 4. Relationship between the creatinine normalized THC-COOH concentration and (a) MCQ-SF scores, r(30) = .63, p < .001, 95% CI[0.36, 0.80], and (b) CUDIT-R scores, r(30) = .35, p = .005, 95% CI[0.01, 0.63]. There was a positive correlation in both cases, n = 32. Gray shading indicates 95% CI. Untransformed THC-COOH/Creatinine plotted on log10 scaled axis to assist with interpretation, Note that all Pearson's correlations reported above were computed using log10 THC. (THC-COOH, 11-Nor-9-carboxy-Δ9-tetrahydrocannabinol; MCQ-SF, Marijuana craving questionnaire-short form; CUDIT-R, Cannabis use disorder identification test-revised).

Bioavailability

For this study 89% of participants reported consuming cannabis via inhalation, which has pyrolytic loss and side stream loss, 51% reported consuming cannabis orally, which decreases bio-availability with liver first pass metabolism, and 8% reported topical application for which the skin reduces bioavailability (16). There was a significant positive correlation between log10 THC-COOH and log10 of bioavailability adjusted THC dose r(41) = .59, p < .001, 95% CI[0.35, 0.76], log10 of bioavailability adjusted THC dose and MCQ-SF score r(41) = .59, p < .001, 95% CI[0.35, 0.76], and the log10 of bioavailability adjusted THC dose and the CUDIT-R score r(41) = .43, p = .004, 95% CI [0.15, 0.64].

Participant attitudes

Response from participants was predominantly favorable. (Table 3 and Figure 5) However, three participants reported that using the diary or quantifying their use was hard. Clarification provided in additional comments revealed that high-frequency use made keeping track challenging. Comments received included “the only difficult part was estimating the number of puffs, I vape all day long,” and “I wish I was able to record my use daily rather than at the end of each week.”

Figure 5. Histogram of respondents reported diary completion times from the Exit Survey (see Supplemental Information). The most frequently reported completion time was 5 min (Mode = 5).

Table 3. Participant feedback on survey instrument.

Exit Survey (Scale Range: 1 = no/hard to 5 = yes/easy)	Mean	(SD)	Median	Interquartile Range
Did you have the information needed to feel confident in completing your cannabis use survey	4.86	(0.48)	5	0
Did you feel the amount of time spent completing Diary was reasonable?	4.81	(0.59)	5	0
How easy to use was the Cannabis Use Diary?	4.49	(0.85)	5	1
How hard was it to estimate your quantity of cannabis used?	4.25	(0.98)	5	2
Was the Consumption Guide helpful?	4.62	(0.95)	5	0

Discussion

Capturing information from cannabis product labels is a valid method for quantifying THC and CBD dosage and provides improved precision and versatility over currently available cannabis self-report measures. Label-based dose quantification is a feasible strategy. Users felt estimation of their use quantity was easy, not overly burdensome with respect to time (<5 min), and 99% of labels submitted were fully usable for quantification. Using this label-based methodology, we were able to determine average daily doses of our participants over the duration of the study, despite varying product potencies, format use patterns, and varied routes of administration. The THC potencies of both flower (24%) and concentrate (71%) for products used by our convenience sample of regular cannabis users in Seattle, Washington are consistent with those previously reported (3), indicating our results are relevant to other regions with legalization.

Validation

Self-reported use quantities were validated against both a quantitative bio-metric as well as clinical measures of cannabis use. Our study saw a significant, moderate-to-strong correlation between THC dose calculated, using our self-report method, and urine THC-COOH metabolites. These results are consistent with prior studies that used self-reported frequency measures (22) and is compelling, particularly since urine-based metrics have their own limitations. Hyperhydration can dilute samples below the detection level of LC-MS (23). Reduced delay between last cannabis use and urine sample is another limitation. Participants were asked to abstain from cannabis use for 24 h. However, 11 of 43 participants who had THC metabolite levels within quantification range reported use less than 24 h prior to sample submission. An inability to abstain from cannabis use for long periods may be inherent to observational studies of community-dwelling heavy cannabis users.

Quantification of THC dose in mg also demonstrated a stronger correlation to the CUDIT-R than the best current alternative self-report approach for measuring quantity that utilized the CUDIT-R as a measure of predictive validity (7). This is likely because the DFAQ-CU provides separate quantities for flower, concentrate, and edibles. As illustrated in Figure 2, the participants in our sample did not exclusively use a single format, with 62% using two or more formats over the two-week period. In addition, typical dose appears to be different for different formats. The differences in format-specific dosing observed in our study may reflect differences in patterns of use – with concentrate users being the most likely to combine formats (see Figure 2). Concentrate-specific dosing was associated with relatively low levels of THC consumption despite the very high concentrations of THC in this product, possibly because users may combine it with other formats to boost the THC (e.g., “spiking” flower with a concentrate before smoking).

Within our sample, self-reported craving measured using the MCQ-SF had a stronger correlation with THC consumption than the CUDIT-R (see Figure 3e,f), and a similar correlation strength to THC urine metabolites (see Figure 4). It is possible that measures capturing physiological adaptive responses to addiction (withdrawal and craving) inherently show a stronger relationship with quantitative measures of use than screening tools for addiction in general. However, our study indicates that measures of craving and THC metabolite measures are not interchangeable, showing a correlation to each other that is similar (but not stronger) to their correlation with THC consumption (see Figure 4).

Adjusting our THC quantification by bioavailability scale factors only slightly changed the correlation values with our THC-COOH urine biometric, i.e., r values fell within the 95% CI of the analyses conducted with THC dosing based on the weight of the product prior to consumption. It is important to note that the scale factors we tested are based on estimates derived from a wide range of potential values. For example, 30% was used as a MAEC efficiency adjustment for cannabis flower, however studies for burned flower have ranged between 2 and 56% bioavailability (16,19,20). Thus, we interpret the lack of significant correlation improvement after adding a bioavailability correction to indicate that the efficiency factors themselves are a potential source of error and require further validation.

Limitations

Quantification of cannabis use requires estimating use amount correctly. Participant carelessness or low motivation to report accurately may impact accuracy of results. We have minimized these factors through designing an individualized approach for quantification based on typical number of uses per purchased product, detailed training, and easily accessible resources. The current study did not formally assess/document participants’ ability to accurately calculate dose, this will be incorporated in future studies.

Regulation varies widely among states with some level of cannabis legalization (24). This lack of regulatory standardization creates missing data challenges for studies spanning multiple states. Potency is the critical piece of information needed from labels and is a limitation for CBD studies given that only 87% of legal states require CBD content on the label (24). Even within Washington State, prior pilot work showed that inconsistent labeling practices resulted in high error rates for participant self-entry of label information, requiring labor intensive hand entry of label data by our research team. Consistent product labeling across the legal market would allow consumers to make more accurate dosing decisions, and researchers to understand the impacts of various dosing patterns. This methodology is not appropriate where product labeling is not implemented, or retrospective studies for which other methods may be more appropriate.

This study was limited to 21–40-year-olds to capture the highest use segment of the population who can legally purchase cannabis while minimizing potential confounds associated with aging, therefore the conclusions cannot currently be extended to older participants. Further work is needed to evaluate if this trend holds with older cannabis users, and if not, potentially tease apart impacts of age and other potential factors such as duration of regular cannabis use.

We implemented weekly diary entries over daily entries to provide balance between accuracy and potential burnout. Based on the challenges reported by several high-frequency users, future directions include development of an app-based diary which would allow users to enter data at a frequency that is most convenient for them yet also providing a reminder mechanism to prevent excessive delay.

Conclusions

Our label-based self-report methodology provides detailed consumption information across all modalities of cannabis use for calculation of dose. It can be utilized in any region where cannabis use is legal, and labeling is regulated. Obtaining more accurate community use patterns can inform the dosage levels chosen for basic science, pre-clinical, and clinical trials. Further, understanding dose levels above which harms may occur can also help inform public health regulation and outreach.

Authorship contribution

Sarah Larsen: Conceptualization, Methodology, Software, Validation, Formal analysis, Data Curation, Writing – Original Draft, Visualization Allegra Johnson: Investigation, Project administration, Software, Validation, Writing – Review & Editing Mary Larimer: Writing – Review & Editing Stephen Dager: Writing – Review & Editing, Funding acquisition Natalia Kleinhans: Conceptualization, Funding acquisition, Formal analysis, Resources, Supervision, Writing – Review & Editing

Acknowledgments

We acknowledge Valentina Apostol-Maughan for her help with designing the REDCap™ weekly diary and Josefin Koehn and the University of Washington Mass Spectrometry Center for THC metabolite analysis. In addition, we would like to thank the retail staff at cannabis dispensaries in the Seattle area for their insights on typical purchasing patterns, information on the variety of products available, and descriptions of the different ways newer formats are used. In particular, we would like to thank the staff at The Joint in Burien, WA for their support and willingness to share their expertise with our research staff.

Disclosure statement

The authors report no relevant disclosures.

Supplementary data

Supplemental data for this article can be accessed online at https://doi.org/10.1080/00952990.2023.2232525

Additional information

Funding

This work was supported by the National Institutes of Health / National Institute on Drug Abuse [R21DA046696] and by an internal grant from the University of Washington Department of Radiology