Adult use of highly-potent Δ9-THC cannabis concentrate products by U.S. state cannabis legalization status, 2021

Abstract

Background.

Compared to plant/flower cannabis products, cannabis concentrates have higher average potency of delta-9-tetrahydrocannabinol (Δ9-THC), which may be associated with greater likelihood of cannabis-related harms. Information on factors associated with use of cannabis concentrates is needed.

Methods.

Respondents were 4,328 adult past-7-day cannabis users from all 50 U.S. states and Washington DC (DC) who participated in an online 2021 survey. Using logistic regression to generate adjusted odds ratios (aOR), we investigated whether participants in states that enacted recreational cannabis laws (RCL, 12 states plus DC [treated as a state], n=1,236) or medical cannabis laws (MCL-only, 23 states, n=2,030) by December 31, 2020 were more likely than those in states without cannabis laws (no-CL, 15 states, n=1,062) to use cannabis concentrate products in the prior 7 days.

Results.

Most participants (92.4%) used plant material in the prior 7 days; 57.0% used cannabis concentrates. In RCL, MCL and no-CL states, concentrate use was reported by 61.5%, 56.6%, and 52.5%, respectively. Compared to participants in no-CL states, odds of using cannabis concentrate products were greater among those in RCL states (aOR=1.47; CI=1.17-1.84) and MCL-only states (aOR=1.29; CI=1.08-1.55). Whether states had legally-authorized dispensaries had little effect on results.

Conclusion.

Results suggest that individuals in MCL-only and RCL states are more likely to use cannabis concentrate products. Determining mechanisms underlying these results, e.g., commercialization, could provide important information for prevention. Clinicians should be alert to patient use of concentrates, especially in MCL-only and RCL states. Continued monitoring is warranted as additional states legalize cannabis use.

1. INTRODUCTION

Major changes have occurred in U.S. cannabis laws over the last 25 years. As of November 9, 2022, 39 states had enacted medical cannabis laws (MCL) and 21 states and Washington D.C. (DC) had enacted recreational cannabis laws (RCL), leaving only 11 states without cannabis legalization (no-CL)¹. Many individuals can use cannabis without harm², and legalization helps achieve social justice and financial aims³. However, cannabis use increases the likelihood of various health problems. For example, ~20% of cannabis users develop cannabis use disorder (CUD), with greater likelihood of CUD (~33%) among frequent users, i.e., those using once a week to daily². CUD is also associated with impairment⁴, psychosocial and health problems⁵, impaired driving^6–9, mental disorders^4,10,11, hyperemesis syndrome¹² and cardiovascular events^13,14. Despite these risks, Americans increasingly see cannabis use as harmless^15,16. Frequent cannabis use and related problems have increased in adults in recent years^15–18, and these increases were greater in states that enacted MCL or RCL^16,19–22. However, little is known about whether state cannabis legalization is associated with features of cannabis products that might contribute to problem development, specifically their potency.

Delta 9-tetrahydrocannabinol (Δ9-THC) is the major psychoactive compound in cannabis^23,24. The potency of cannabis products is generally defined by their Δ9-THC content. Using higher-potency cannabis is associated with adverse outcomes, including greater risk of CUD²⁵, more rapid progression to first CUD²⁶, CUD severity²⁷, and cannabis-related mental health problems including depression and anxiety¹⁰, and psychotic disorders^25,28–30. Given these findings, identifying factors that increase the likelihood of using highly-potent cannabis products is an important public health priority.

Cannabis concentrate products are a type of highly-potent cannabis product whose use is becoming increasingly popular^31–33. Cannabis concentrates are made by extracting cannabinoids from plant to form products with Δ9-THC potency ranging from 52%-95%^34–38, which is substantially higher than the average Δ9-THC potency now found in plant/flower cannabis products (14-21%)^36,37,39,40. Cannabis concentrates have several routes of administration, including vaping via a vape pen, or dabbing, i.e., heating a strong cannabis concentrate, usually an oil or wax, and then inhaling the vapor. Determining whether state cannabis legalization status is associated with higher likelihood of using cannabis concentrates is an important aspect of understanding the health impact of changing cannabis laws.

Few studies have addressed this issue. A 2016 Twitter-based adult survey found an overall prevalence of 48% with lifetime concentrate use⁴¹, which was more likely among those in RCL or MCL-only states; differences between RCL and MCL-only states were not tested. A 2017 Facebook-based survey found that among frequent users, 54.2% used concentrates, which was associated with living in either RCL or MCL-only states, with likelihood of concentrate use greater in RCL states⁴². In 2018-2020 Neilson online surveys⁴³, 30.2%-39.8% of past-year cannabis users age 16-64 had vaped cannabis oil in the past year, with greater likelihood among those in RCL states. However, the two studies of Neilson survey data combined MCL-only and no-CL states, possibly obscuring important distinctions because adult cannabis use and other cannabis outcomes are more likely in MCL-only than in no-CL states^{21,22,44–46}.

Since 2018, additional states enacted MCL and RCL (Figure 1), and adult rates of cannabis use have continued to climb⁴⁷. With adult cannabis use increasingly normative, these changes may have affected the association of RCL and MCL with use of cannabis concentrates, warranting re-examination in more recent data. In addition, previous studies used timeframes from lifetime to past 30-day use. Among current cannabis users, a more recent timeframe, i.e., the prior 7 days, should improve recall and accuracy, and should provide the most specific picture of current use patterns. In addition, no prior study determined whether having state-protected legal dispensaries or retail outlets affected the relationship of state cannabis laws to the likelihood of concentrate use. These aspects of the laws could affect the distributions of concentrate use due to increased availability. We therefore analyzed 2021 data from a large U.S. online survey of adult cannabis users to address two research questions. First, in 2021, was living in RCL or MCL-only states associated with recent use of cannabis concentrates? Second, were these associations altered by whether RCL or MCL-only states had legally-permitted dispensaries or retail outlets?

Figure 1.

RCL and MCL state coding, effective laws by end of 2020

2. METHODS

2.1. Sample and recruitment procedures

Residents of the 50 U.S. states and the District of Columbia (DC) aged ≥ 18 years were recruited to participate in an anonymous online survey that reached potential respondents via keyword-targeted social media (Facebook and Instagram) advertisements. Web-based health surveys are emerging as rapid and cost-effective methods for population-level data collection⁴⁸, particularly suitable for hard-to-reach populations such as current substance users^49,50, including those of interest to the present study. Data were collected between June 3 and August 3, 2021. Further information about our social media recruitment methods, including the two ad campaigns (which differed primarily by the image used to attract potential participants), is provided in Supplementary Appendix 1 and Supplementary Table 1; Supplementary Figures 1a and 1b show the images used. After clicking the social media ad, individuals were directed to an informed consent page that described the study and confidentiality. After consenting and passing initial screening for age and inattention, the questionnaire was presented via the Qualtrics platform⁵¹. The questionnaire covered behaviors related to cannabis use and demographic information. Study procedures were approved by the Dartmouth Committee for the Protection of Human Subjects; no compensation was provided. Those who responded to the ads (n=5,627) were directed to the informed consent page. Of these, 43 did not continue (11 refused consent, 11 were <age 18, and 21 failed initial quality/inattention checks). Of the 5,584 remaining individuals, 53 were excluded due to nonsensical/illogical survey responses, 797 due to survey non-completion, and 406 who did not use cannabis plant or concentrate products in the prior 7 days, leaving an analytic sample of 4,328 (Supplementary Figure 2).

2.2. Measures

2.2.1. Primary exposure: state cannabis laws

State cannabis laws were defined by the effective date of RCL or MCL-only, i.e., when the protections provided by the law became operational. The RAND-USC Schaeffer Opioid Policy Tools and Information Center (OPTIC) marijuana policy data¹, a well-vetted policy dataset, provided the effective dates through 2019. Effective dates for 2020 were determined from legislative and statutory resources from each relevant state. To create a comparable exposure period for all states, each was coded as positive for RCL or MCL-only if the legislation was effective by December 31, 2020 (Figure 1). (Note that all states with RCL had previously had MCL.) Two variables were then created. First, states were categorized into a two-level exposure variable such that participants in states with an effective RCL before 2021 were coded positive and all others coded negative, as was done in a prior study⁴³. The second was a three-level variable that grouped participants by whether they resided in an RCL state, in an MCL-only state, or in a state with neither RCL nor MCL (no-CL). In these analyses, DC was treated as an RCL state.

For sensitivity analyses, additional state-level variables were created to indicate cannabis availability through operational, legally protected dispensaries for medical cannabis or retail outlets for recreational cannabis. Dates of operational dispensaries/outlets through 2019 were obtained from the RAND OPTIC dataset. For 2020, we determined these from contemporary news reports and statements released by state agencies. With this information, RCL or MCL-only status was re-coded so that the state was considered positive only if it had operational dispensaries or outlets by December 31, 2020. This resulted in five re-codes: Arizona, DC, and Vermont became “no” for RCL and were recoded as MCL-only, and North Dakota and West Virginia became “no” for MCL-only and were re-coded as no-CL.

2.2.2. Outcome: use of cannabis concentrates

Participants who used cannabis at least once within the prior 7 days were asked about products used and routes of administration. This included questions about using cannabis plant material by smoking or vaping. Questions also covered using cannabis concentrates (“for example, oils or dabs like wax, shatter etc…”) by vaping or by dabbing (using a dab rig). Participants could indicate using both plant and concentrate products. With this information, a binary variable was created indicating any concentrate use vs. all others⁴². The past 7-day period was selected because among this sample of current cannabis users, it targeted a very recent, highly specific period of current behaviors, potentially mitigating recall bias that could occur in lengthier timeframes.

2.2.3. Covariates

Individual-level control covariates were those associated with cannabis use and related outcomes in prior studies. These included age, sex (male; female; other), and race/ethnicity (White; Black; Hispanic; other)^3,52–55, employment status (any vs. none) and education (high school graduate or more vs. all others)^4,56. State-level covariates were included to control for geographically-defined environmental context, as has been done in other studies of cannabis laws^20,57. These were derived from US Census Bureau data⁵⁸. They included age (% of state residents <30 years), gender (% of state residents who were male), race/ethnicity (of state residents who were % non-Hispanic White), % of families within states below the federal poverty level, and state population density.

2.2.4. Statistical analysis

Logistic regression modeled the association between use of cannabis concentrates (outcome) and the cannabis law variables (exposure). We first ran unadjusted models and then models adjusted for the individual- and state-level covariates, generating odds ratios (OR), adjusted odds ratios (aOR), and 95% confidence intervals (95% CI). For the three-level cannabis law variable, contrasts between RCL vs. MCL-only effects determined if these effects differed.

To determine if any states were disproportionally represented in the sample, we compared the proportion of participants who reported living in each state to the proportion of the US population living in each state according to the 2020 Census⁵⁹. For this, we used Spearman correlations and scatter plots. To explore whether participants in no-CL states were under-represented in the sample, we examined whether the strength of the relationship between a state’s proportional representation in the US population and its proportional representation in the sample was modified by the state’s cannabis law status. In robust linear regression models (Stata; Huber/White sandwich estimator), we used the state-specific sample proportions to predict the state-specific Census proportions, adding dummy variable interaction terms representing the state’s medical and recreational cannabis law status (no-CL states as the reference group).

2.2.5. Missing data

Employment status and education were missing for 5.3% (n=229) and 6.4% (n=278), respectively. To address missing data, multiple imputation was used⁶⁰, implemented with SAS 9.4⁶¹. Since missing data were binary, imputation used the fully conditional specification method (multiple imputation by chained equations; MICE), with all variables in the analytical model included in the imputation model. After 100 burn-in iterations, 10 imputed datasets were generated. In each dataset, employment was imputed first, and then education (in this order because employment had fewer missing responses). The regression models were analyzed in each imputed dataset, with parameter estimates averaged over all 10 imputed datasets to yield average effects in the imputed data⁶².

3. RESULTS

3.1. Sample characteristics

Of the participants, 63.9% were female, 33.6% under 30 years old, 79.5% White, 41.4% unemployed, and 46.9% had less than high school education (Table 1). About half (46.9%) lived in MCL-only states, 28.6% in RCL states, and 24.5% in no-CL states. Most reported daily cannabis use in the last 7 days (79.6%; Table 2). Virtually all participants used plant material in the prior 7 days (92.4%), mostly by smoking (90.4%). Cannabis concentrate use was reported by 57.0%. Using both concentrate and plant/flower cannabis products was common (49.4%). Use of plant/flower only was reported by 43.1%, and a small proportion, 7.6%, reported using concentrates only.

Table 1.

sample characteristics, overall and by state cannabis law status

	Whole sample (N=4,328)		Recreational cannabis lawa (n=1,236; 28.6%)		Medical cannabis lawb (n=2,030; 46.9%)		No cannabis law (n=1,062; 24.5%)		Differences by law status
	%	n	%	n	%	n	%	n	χ2 (df), p-value
Sociodemographic
Age									31.72(8), .0001
18-20	07.5	325	08.2	101	07.1	144	07.5	80
21-29	26.1	1,128	30.1	372	25.9	526	21.7	230
30-44	27.5	1,189	25.7	317	28.2	573	28.2	299
45-64	32.1	1,388	29.1	360	31.4	638	36.7	390
65+	06.9	298	07.0	86	07.3	149	05.9	63
Gender									5.32(4), .256
Male	36.1	1,563	36.2	447	36.7	744	35.0	372
Female	59.2	2,563	58.4	722	58.6	1190	61.3	651
Other	04.7	202	05.4	67	04.7	96	03.7	39
Race/ethnicity									53.32(6), <.0001
White	79.5	3,439	75.2	930	80.3	1631	82.7	878
Black	03.3	144	02.4	30	03.6	72	04.0	42
Hispanic	07.9	342	11.8	146	06.2	125	06.7	71
Other	09.3	403	10.5	130	10.0	202	06.7	71
Employment									1.55(4), .818
Any	53.3	2,306	52.8	653	54.0	1097	52.4	556
None	41.4	1,793	42.2	521	40.5	822	42.4	450
missing	05.3	229	05.0	62	05.5	111	05.3	56
Education									7.59(4), .108
High school or less	46.9	2,029	45.9	567	45.7	928	50.3	534
More than high school	46.7	2,021	48.1	595	47.6	966	43.3	460
missing	06.4	278	06.0	74	06.7	136	06.4	68

df = degrees of freedom

^aParticipant lived in a state with a recreational cannabis law (RCL) effective by 12/31/2020: Alaska, Arizona, California, Colorado, DC (treated as a state in this study), Illinois, Maine, Massachusetts, Michigan, Nevada, Oregon, Vermont, and Washington. All these states also had an MCL effective by 12/31/2020.

^bParticipant lived in a state with a medical cannabis law (MCL) effective by 12/31/2020 but no RCL: Arkansas, Connecticut, Delaware, Florida, Hawaii, Louisiana, Maryland, Minnesota, Missouri, Montana, New Hampshire, New Jersey, New Mexico, New York, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, Utah, Virginia, and West Virginia.

Table 2:

Characteristics of cannabis use, prior 7 days, overall and by state cannabis law status

	Whole sample (N=4,328)		Recreational cannabis lawa (n=1,236)		Medical cannabis lawb (n=2,030)		No cannabis law (n=1,062)		Difference by law status χ2(df), p-value
	%	n	%	n	%	n	%	n
Frequency of use, last 7 days c									20.58(8), .0084
1- 2 days	04.3	187	03.6	44	04.2	84	05.6	59
3-4 days	06.6	284	06.0	74	06.0	122	08.3	88
5 days	04.4	192	04.3	53	04.0	80	05.6	59
6 days	04.7	202	04.4	54	04.7	94	05.1	54
7 days	79.9	3,444	81.7	1,006	81.2	1,642	75.4	796
Product potency and method of administration
Low-potency (cannabis plant material)
Vaped plant material	10.8	466	12.5	155	10.5	213	9.2	98	6.82(2), .033
Smoked plant material	90.4	3,914	90.4	1,117	90.3	1,833	90.8	964	0.19(2), .909
Any use of plant material	92.4	3,999	92.4	1,142	92.3	1,874	92.6	983	0.06(2), .970
High-potency (cannabis concentrates)
Vaped cannabis concentrate	50.5	2,186	54.1	669	50.6	1,028	46.1	489	14.95(2), .0006
Dabbed cannabis concentrate	26.3	1,140	32.1	397	25.5	518	21.2	225	36.53(2), <.0001
Any use of concentrates	57.0	2,465	61.5	760	56.6	1,148	52.5	557	19.30(2), <.0001
Combined potency									20.34(4), .0004
Used low-potency plant products only	43.1	1,863	38.5	476	43.5	882	47.6	505
Used high-potency concentrates only	07.6	329	07.6	94	07.7	156	07.4	79
Used both high and low potency products	49.4	2,136	53.9	666	48.9	992	45.0	478

df = degrees of freedom

^aParticipant lived in a state with a recreational cannabis law (RCL) effective by 12/31/2020: Alaska, Arizona, California, Colorado, DC (treated as a state in this study), Illinois, Maine, Massachusetts, Michigan, Nevada, Oregon, Vermont, and Washington. All these states also had an MCL effective by 12/31/2020.

^bParticipant lived in a state with a medical cannabis law (MCL) effective by 12/31/2020 but no RCL: Arkansas, Connecticut, Delaware, Florida, Hawaii, Louisiana, Maryland, Minnesota, Missouri, Montana, New Hampshire, New Jersey, New Mexico, New York, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, Utah, Virginia, and West Virginia.

^c19 were missing frequency of use

Participants lived in all 50 states and in Washington DC (Figure 1). Comparing the proportion of respondents who reported living in each state to the proportion of the US population living in each state according to the 2020 Census yielded a Spearman coefficient of 0.90. None of the interaction terms in the regression models were statistically significant, so representation in the sample by state legal cannabis status did not appear to be disproportionate. Using log-transformed versions of the two proportion variables yielded highly similar results.

3.2. Cannabis law status and cannabis concentrate use

Comparing participants in RCL states to all others (two-level predictor; Table 3), those in RCL states were more likely to use cannabis concentrates in unadjusted (OR=1.30; 95% CI=1.14, 1.49) and adjusted models (aOR=1.20; 95% CI=1.01, 1.43). Using the three-level predictor, unadjusted and adjusted results were very similar: participants in RCL states had greater odds of using concentrates than participants in no-CL states (aOR= 1.47; 95% CI=1.17, 1.84), as did participants in MCL-only states (aOR= 1.29; 95% CI=1.08, 1.55). When participants in RCL and MCL-only states were compared (Table 3), unadjusted results showed that those in RCL states were more likely to use cannabis concentrates (OR=1.23), but in the adjusted model, the association weakened and became null. Using the imputed sample, results were essentially unchanged (Supplementary Table 2). Results were also very similar when just RCL and MCL-only states with legally-protected operational dispensaries/outlets were considered positive (Table 4).

Table 3.

Recreational cannabis laws (RCL), medical cannabis laws (MCL-only) and use of high-potency cannabis concentrate products in the prior 7 days

	Participants with use of high-potency products	Unadjusted	Adjusted a
2-level cannabis law variable	% (SE)	Odds ratio (95% CI)
RCLb	61.5 (1.38)	1.30 (1.14, 1.49)	1.20 (1.01, 1.43)
No RCL	55.1 (0.89)	reference	reference
3-level cannabis law variable
RCLb	61.5 (1.38)	1.45 (1.23, 1.71)	1.46 (1.17, 1.83)
MCL-onlyc	56.6 (1.10)	1.18 (1.02, 1.37)	1.29 (1.07, 1.55)
No-CL (no RCL or MCL)	52.5 (1.53)	reference	reference
RCL vs MCL-only	--	1.23 (1.06, 1.42)	1.13 (0.95, 1.36)
Contrasts between RCL and MCL effects		Wald χ2 (dof=1), p-value
		7.67, 0.0055	1.87, 0.1717

CI=confidence interval; RCL=recreational cannabis law; MCL=medical cannabis law; dof = degrees of freedom

N=4,328 for unadjusted, 4,050 for adjusted models

^aAdjusted for individual-level (age, gender, race/ethnicity, employment, education) and state-level (age, gender, race/ethnicity, poverty, population density) characteristics. ORs with 95% CI above 1 are significant at the p<0.05 level.

^bParticipant lived in a state with a recreational cannabis law (RCL) effective by 12/31/2020: Alaska, Arizona, California, Colorado, DC (treated as a state in this study), Illinois, Maine, Massachusetts, Michigan, Nevada, Oregon, Vermont, and Washington. All these states also had an MCL effective by 12/31/2020.

^cParticipant lived in a state with a medical cannabis law (MCL) effective by 12/31/2020 but no RCL: Arkansas, Connecticut, Delaware, Florida, Hawaii, Louisiana, Maryland, Minnesota, Missouri, Montana, New Hampshire, New Jersey, New Mexico, New York, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, Utah, Virginia, and West Virginia.

Table 4.

Cannabis laws with legally protected dispensaries/retail outlets and use of high-potency cannabis concentrate products in the prior 7 days

	Participants with use of high-potency products	Unadjusted	Adjusted a
2-level cannabis law variable	% (SE)	Odds ratio (95% CI)
RCLb	62.3 (1.46)	1.34 (1.17, 1.55)	1.21 (1.01, 1.44)
No RCL	55.1 (0.88)	reference	reference
3-level cannabis law variable
RCLb	62.3 (1.46)	1.51 (1.28, 1.79)	1.52 (1.21, 1.90)
MCL-onlyc	56.7 (1.08)	1.20 (1.04, 1.39)	1.33 (1.11, 1.59)
No-CL (no RCL or MCL)	52.2 (1.49)	reference	reference
RCL vs MCL-only	--	1.26 (1.09, 1.47)	1.14 (0.95, 1.37)
Contrasts between RCL and MCL effects		Wald χ2 (dof=1), p-value
		9.31, 0.0023	2.05, 0.1520

CI=confidence interval; RCL=recreational cannabis law; MCL=medical cannabis law; dof = degrees of freedom

N=4,328 for unadjusted, 4,050 for adjusted models

^aAdjusted for individual-level (age, gender, race/ethnicity, employment, education) and state-level (age, gender, race/ethnicity, poverty, population density) characteristics. ORs with 95% CI above 1 are significant at the p<0.05 level.

^bParticipant lived in a state with a recreational cannabis law (RCL) effective by 12/31/2020: Alaska, Arizona, California, Colorado, DC (treated as a state in this study), Illinois, Maine, Massachusetts, Michigan, Nevada, Oregon, Vermont, and Washington. All these states also had an MCL effective by 12/31/2020.

^cParticipant lived in a state with a medical cannabis law (MCL) effective by 12/31/2020 but no RCL: Arkansas, Connecticut, Delaware, Florida, Hawaii, Louisiana, Maryland, Minnesota, Missouri, Montana, New Hampshire, New Jersey, New Mexico, New York, North Dakota, Ohio, Oklahoma, Pennsylvania, Rhode Island, Utah, Virginia, and West Virginia

4. DISCUSSION

Among participants age ≥18 years in a 2021 U.S. online survey who reported using cannabis in the past 7 days, those in states with RCL were more likely than others to use cannabis concentrates, and participants in both RCL and MCL-only states were more likely to use concentrates than those in states with no cannabis legalization. These results were robust to whether RCL or MCL states had legally protected dispensaries. When RCL and MCL effects were directly compared, unadjusted results showed that RCL effects were stronger than MCL-only effects, although after adjustment, RCL and MCL-only effects no longer differed significantly, indicating that each type of cannabis legalization was associated with use of concentrates. Cannabis concentrates have substantially higher mean Δ9-THC potency than plant/flower products^34–38, and higher-potency cannabis products are associated with greater risk of harmful effects^{10,25–29,63–65}. Therefore, living in a state with legalized cannabis use may increase the likelihood of recent use of cannabis products with greater risk of harm. Although the present study is cross-sectional, findings are consistent with Canadian findings that cannabis concentrate use increased post-legalization in 2018⁶⁶.

Given potential health risks from high-potency cannabis products and the predictions of policy experts that legal cannabis markets emerging in RCL states would expand access to more potent cannabis products^67,68, surprisingly few studies have examined the relationship of state cannabis laws to use of cannabis concentrates. However, from the studies that did and that differentiated between no-CL and MCL^41,42 states, results are consistent with our findings on past-7-day cannabis users, thus adding to accumulating evidence that among cannabis users, state legalization is associated with greater use of highly-potent cannabis products.

Some individuals titrate their Δ9-THC exposure from high-potency cannabis products through self-administration behaviors^69,70, which can mitigate potential harms from use of such products⁶⁵. However, this does not always occur, and therefore, greater Δ9-THC concentrations in cannabis products can deliver higher Δ9-THC doses to the user³⁸. Further, many concentrate users select high-potency products deliberately for “stronger intoxication effect”, “effects last longer” and “fewer hits are necessary”⁷¹, suggesting that concentrate users are not only exposed to higher doses of Δ9-THC but in fact, seek out concentrates because of their potency. Such drug-seeking behavior is of public health concern if the potent forms are more likely cause health harms. Some of the literature on this is mixed, for example, whether high-potency cannabis use is more likely to cause memory problems, which was found in one study⁶⁴ but not in two others^65,72. More refined knowledge on patterns and effects of high potency cannabis use within the legal landscape could elucidate mixed findings on cannabis-related harms. Acute and long-term neurocognitive effects of high potency product use, including memory impairment, are of particular interest.

While our 2017 survey and the present study differed in important methodological respects (e.g., a 30-day timeframe in 2017 and a 7-day timeframe in 2021), comparison of the results of the two surveys suggested that during the four years between the surveys, when 5 additional states adopted more permissive cannabis laws (four changing from MCL-only to RCL; one changing from no-CL to MCL-only), the relationships remained similar. However, closer inspection of the results also suggests some differences. The strength of the association descriptively appeared stronger in 2017 (aOR=2.61) than in 2021 (aOR=1.47). This could be due to diffusion of positive cannabis norms across U.S. adults, with diminishing influences by state laws, or a diffusion of products across state borders⁷³. Another difference in 2017 and 2021 results is that in 2017, RCL and MCL-only effects differed strongly from each other in unadjusted and adjusted analyses. However, in 2021, RCL effects were stronger than MCL-only effects in unadjusted but not adjusted analyses. This is also consistent with a general diffusion of positive cannabis-using norms in U.S. adults, regardless of their specific state laws.

Most studies on state cannabis legalization and use of highly-potent cannabis products have utilized cross-sectional designs, precluding knowledge of the direction of effect, i.e., whether enactment of cannabis laws is followed by increases in cannabis outcomes. Analyzing data from repeated annual surveys that include states that change their laws and those that do not using difference-in-difference methodology would provide more information. We know of no existing national datasets with such information. Large-scale annual national surveys should measure specific cannabis products so that the influence of cannabis legalization on the likelihood of using certain types of cannabis products can be better understood.

Cannabis concentrates are rapidly gaining in popularity and sell well in both medical and recreational cannabis outlets^31–33. These products have high profit margins^74,75, and the burgeoning cannabis industry favors them, incentivizing “providers” (in this case, “budtenders’/sales associates) to increase such sales⁷⁶, and subsidizing lobbying aimed at increasing consumption and profits, including opposition to potency restrictions^77,78. Such cannabis industry practices are reminiscent of earlier pharmaceutical industry practices aimed at promoting opioid prescriptions^79–81, and may be contributing to the increasing popularity of highly potent cannabis products.

The present study adds to an accumulating body of evidence that in the U.S., current cannabis users are more likely to use high-potency products if they live in states that legalized cannabis use for medical or recreational purposes, where the cannabis industry can enter and potentially make these products more available. However, the role of industry practices and other potential mechanisms for increasing use of cannabis concentrates requires further investigation. In the meantime, policymakers in states and local communities should consider the evidence on use of high-potency products when formulating laws, policies and public health messaging about the potency of permitted products. Policies to consider include minimum unit pricing and taxes based on potency rather than sales price^36,82,83.

Whether the association of residing in RCL or MCL states with use of high-potency products is modified by important individual characteristics is relatively unknown. Exploration of our data suggested that use of high-potency products was more likely among participants age 18-20 and 21-29 (aOR=1.73 and 1.89, respectively), and much less likely among those age 65 and older (aOR=0.52; Supplementary Table 3). While investigating moderators is beyond the scope of the present study, an important area of future research is to determine whether the effects of changing cannabis laws are modified by individual sociodemographic and also clinical characteristics such as psychiatry comorbidity or pain.

Study limitations are noted. 1) Participants were recruited through social media advertisements and the algorithms that disseminate these advertisements are not available, leaving potential selection biases unclear. However, similarity of the proportion of participants by state to the proportions of state residents in the U.S. 2020 Census suggested that the sample was proportionally representative of US state populations and we found no evidence for differential participation by state cannabis law status. While a national probability sample of U.S. adults would alleviate other sampling bias concerns, such data are not available. Present findings are interpreted in light of the sample design. 2) Reasons for use of higher- and lower-potency products were not ascertained; future studies should do so. 3) Future studies should address whether other individual-level variables (e.g., alcohol, tobacco or nicotine use, perceptions of risk in cannabis concentrates, presence of cannabis use disorder, general health issues) or geographically-defined variables (cannabis outlet density; COVID-19 protections of cannabis outlets as essential businesses) modify results. 4) While cannabis concentrates could theoretically contain mainly cannabidiol rather than Δ9-THC, cannabidiol-dominant products constitute a very minor portion of the legal cannabis concentrate market⁸⁴, and the cannabidiol concentration in illicit cannabis has been very low since the 1990s³⁸. The survey gave instructions to only consider cannabis products known to contain Δ9-THC. 5) A few states (Connecticut, New Mexico, Virginia) had 1-3 months of exposure to newly-passed RCL during our survey. Whether these brief exposures affected the results is unknown; future studies will be needed to address the effects of all changes in cannabis laws that occurred during 2021. 6) Quantities of cannabis were not analyzed in this report. Measuring quantity accurately is challenging, especially in non-laboratory studies, but studies should implement validated quantity measures when they become available. 7) Cannabis potency was inferred from the type of cannabis product used. Cannabis consumer knowledge of product Δ9-THC levels is low⁸⁵. When more accurate product labeling becomes standard, enabling consumers to be more knowledgeable about the potency of the products they use, surveys should question product potency directly. 8) Edible cannabis products are also gaining in popularity and can have relatively high Δ9-THC concentrations. Future studies should examine the use of these products by MCL and RCL state status. 9) This study addressed dispensaries by whether states had legally authorized them or not. Future studies could include the density dispensaries within a given radius of participants’ zipcodes, thereby ignoring state boundaries, since residents of some states (e.g., within New England) can easily drive to a nearby neighboring state with a more permissive law to purchase cannabis products through a dispensary. 10) The present study utilized a single year of cross-sectional data. Future studies should apply difference-in-difference methods to surveys that are repeated over time to be able to address the direction of the relationship between state cannabis law status and use of highly-potent cannabis products.

5. CONCLUSION

Although further studies are needed in nationally representative samples, the present study suggests that recent (past-7-day) cannabis users in states with legalized recreational or medical cannabis use are more likely than those in other states to use cannabis concentrates, which are generally more potent than cannabis plant products. Assumptions about the harmlessness of cannabis based on the low-potency forms used before 2000 may no longer be tenable given the increasingly widespread use of high-potency forms such as concentrates. Clinicians in outpatient and emergency settings, especially in states with legalized medical or recreational cannabis, should be alert to these issues, screening regularly for use of potentially harmful cannabis products in conjunction with frequency and other potentially relevant user behaviors and characteristics, analogous to the screening for harmful/binge drinking that is now widely recommended^86,87. The U.S. is currently in a period of rapidly expanding public support for legal cannabis use, and additional states are likely to legalize recreational or medical cannabis use soon. How these further changes in the legal status of cannabis will affect the use of highly potent forms of cannabis, their potential health consequences and which policies best promote public health requires continued monitoring.

Highlights.

• In a 2021 survey of 4,328 adult past-7-day cannabis users, 57.0% used concentrates

• Concentrate use was higher in states with medical or recreational marijuana laws

• Determining mechanisms of the differences by state law is important for prevention