Associations between modes of cannabis use and cannabis use disorder: Evidence from the 2022 to 2023 United States National Survey on Drug Use and Health Authors: Amrit Baral, Cerina Dubois, Lakshmi Kumar, Janardan Devkota, Denise C. Vidot, Ryan Vandrey, Johannes Thrul Published: 2026 ABSTRACT Background and Aims:With expanding cannabis legalization, normalization, and diversifying products and delivery methods in the United States (US), cannabis use disorder (CUD) prevalence is rising. Various modes of cannabis use may influence pharmacokinetics, usage patterns, and harm, affecting CUD risk. We measured associations between modes of cannabis use, including multi-modal patterns, and CUD prevalence and severity.Design and Setting:This cross-sectional study analyzed data from a nationally representative sample of US adults using the 2022–2023 National Survey on Drug Use and Health (NSDUH) data. Multivariable logistic regression analyses were employed to estimate the association between modes of cannabis use and past-year CUD, adjusting for potential confounders and covariates. Analyses were stratified by sex, age, and cannabis use frequency. Among multi-modal users, common combinations and their associations with CUD were further examined.Participants/Cases:Respondents 18 years or older who reported past-year cannabis use (unweighted n = 25 549; weighted N = 58 850 309).Measurements:Exposure of interest was the mode of cannabis use, primarily categorized as smoke-only, vape-only, oral/mucosal-only, dab-only, topicals-only, and multi-modal (≥ two modes). The outcome variable was CUD in the past year, and CUD severity, based on Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) criteria. Covariates included age, sex, race/ethnicity, income, education, state cannabis law status, age of cannabis initiation, cannabis use motive, frequency of use, perceived risk of smoking cannabis, illicit drug use, past year mental illness, nicotine dependence, and alcohol use disorder.Findings:Of the total past-year cannabis users, 53.9% reported multi-modal cannabis use. Overall, CUD prevalence was 30.3%, ranging from 4.4% among oral/mucosal-only to 40.5% among multi-modal, and 28.9% among dab-only users (p < 0.0001). Moderate-to-severe CUD affected 13.2% of all users and was concentrated among multi-modal and dab-only users. In multivariable regression, multi-modal users had fourfold higher odds of CUD (adjusted odds ratio [AOR] = 4.14; 95% confidence interval [CI]: 2.91–5.90). Elevated odds were also observed among smoke-only (AOR = 2.98; 95% CI: 2.02–4.39) and vape/dab-only users (AOR = 1.89; 95% CI: 1.09–3.29), compared with oral/mucosal-only users. Analyses of multi-modal combinations showed the highest CUD odds among those using smoke + vape + oral/mucosal + dab (AOR = 19.74; 95% CI: 9.11–42.75), compared with oral/mucosal + topicals users.Conclusions:In the United States, modes of cannabis use appear to be statistically significantly associated with prevalence and severity of cannabis use disorder, with multi-modal and inhaled routes conferring the greatest risk. Findings underscore the importance of considering mode of use alongside frequency and potency in clinical assessment, prevention, and policy strategies aimed at reducing cannabis-related harms. INTRODUCTION Cannabis use disorder (CUD) is a mental health condition characterized by a problematic pattern of cannabis use that leads to negative health consequences, including significant impairment or distress [1]. With approximately 54% of the United States (US) adults living in states where cannabis is legal for adult use [2, 3], concerns about CUD have grown. Estimates of CUD prevalence vary by population studied, ranging from approximately 10% among cannabis users globally [1] to 21% among primary care patients reporting past 30-day cannabis use in the US [4]. This increase parallels the rapid expansion of the cannabis marketplace and the diversification of consumption methods [5] that include inhalation (smoking, vaping and dabbing of concentrates), oral ingestion (edibles and beverages), mucosal absorption and topical application [6]. Understanding the routes of administration and combinations of methods is critical because they influence the pharmacokinetics (i.e. absorption, distribution and metabolism) of cannabinoids [7], patterns of use and perceived harm [8], which may shape CUD risk. Controlled research shows that variation in the route of cannabis administration influences the onset, duration and magnitude of pharmacodynamic effects [9, 10], while reinforcement theory suggests that faster onset and higher potency drug effects can increase rewarding effects and addiction risk [11]. Guided by these frameworks, we posit that multi-modal use, particularly patterns that include inhalation of high-potency Δ-9-tetrahydrocannabinol (THC) concentrates, may confer higher CUD risk and greater CUD severity by combining rapid-onset (vaping and dabbing concentrates or smoking) with prolonged effects from ingested products (e.g. edibles), increasing cumulative THC exposure and reinforcement. In parallel, the growing availability of high-potency inhaled THC products such as concentrates (e.g. wax, shatter and rosin) frequently exceed 70% THC, underscoring potency-related concerns such as tolerance, withdrawal and CUD risk [12], although large surveys typically do not measure potency directly. Frequency of cannabis use has been extensively studied and is consistently identified as a dominant correlate of CUD risk [13], with markedly higher risk among daily/near-daily users [1]. Despite the growing diversity of cannabis products in the United States, nationally representative evidence on how different modes of cannabis use relate to CUD prevalence and severity remains limited. Prior research by Choi et al. [8] using National Survey on Drug Use and Health (NSDUH) data, characterized patterns of cannabis consumption and risk perceptions, but primarily examined perceived risk of smoking cannabis rather than directly estimating CUD risk across modes. Although that study identified associations between perceived smoking risk, use of edibles or drinks and correlations between consumption methods and CUD, it did not directly quantify mode-specific and/or multi-modal CUD risk [8]. Beyond NSDUH, large-scale epidemiologic data linking routes of cannabis use to CUD remain scarce, partly because of the lack of validated tools that capture the evolving complexity of cannabis formulations and delivery systems [14, 15]. Prior studies suggest that route of cannabis use and multi-modal use may relate to CUD risk [16–18], but route-specific evidence remains limited in population-based analysis. Much of the CUD literature instead focuses on treatment modalities, including psychosocial and pharmacological interventions [19] and risk perceptions [20]. To address this gap, the present study places CUD as the primary outcome, systematically examining both its prevalence and severity across individual modes of use and multi-modal combinations. Using the most recent (2022–2023) NSDUH data, we applied a granular classification that distinguishes not only common modes (smoking, oral/mucosal), but also high-potency routes, such as vaping and ‘dabbing’ high-potency concentrates, hypothesized to represent higher-risk consumption profiles. METHODS Data source and study population This study used de-identified, publicly available data from the NSDUH 2022 to 2023. Administered annually by the Substance Abuse and Mental Health Services Administration (SAMHSA), the NSDUH uses multi-stage area probability sampling and provides nationally representative cross-sectional estimates of substance use and behavioral health outcomes among the civilian, non-institutionalized US population 12 years and older [21]. Computer-assisted interviewing and audio computer-assisted self-interviewing techniques were used for data collection. Detailed information on survey design and methodology can be found elsewhere [21, 22]. Data used in this study were de-identified by SAMHSA, qualifying the research for exemption from institutional review board (IRB) review. The analytic sample included adults aged 18 years and older who reported past-year cannabis use (unweighted n = 25 578, weighted n = 58 850 309) (Figure S1). Public-use NSDUH data files are available for download at the SAMHSA website: https://www.samhsa.gov/data/. Exposure of interest The primary exposure was the mode of cannabis use. Respondents were first asked whether they had ever used marijuana or any cannabis product [excluding cannabidiol (CBD) or hemp products]. Those answering ‘yes’ were queried about past-year cannabis use, age at first use and time since last use. Past-year cannabis use was defined as any cannabis use during the preceding 12 months. Among past-year users, respondents indicated in which ways they used cannabis during the past 12 months, with separate yes/no items for: smoking (e.g. joints, pipes, bongs, blunts and hookahs), vaping (e.g. vape pens, dab pens, tabletop or portable vaporizers), dabbing waxes/shatter/concentrates, eating or drinking, drops/strips/lozenges/sprays in the mouth, applying lotion/cream/patch to skin (topicals), pills and ‘some other way.’ Each mode was coded as a binary variable (1 = yes, 0 = no). We combined eating or drinking, drops/strips/lozenges/sprays in the mouth and pills into a single ‘oral/mucosal’ category. Therefore, single-mode categories were limited to smoking-only, vaping-only, oral/mucosal-only, dabbing-only and topicals-only. The multi-mode group was defined as the use of two or more modes in the past 12 months. Together, this resulted in a six-category cannabis use variable of (1) smoking-only; (2) vaping-only; (3) oral/mucosal-only; (4) dab-only; (5) topicals-only; and (6) multi-modal. We excluded others-only (n = 29) from the analyses. For subgroup analysis within the multi-modal group, we analyzed the most common unique combinations (n = 12) based on their distribution: (1) ‘smoke + oral/mucosal’, (2) ‘smoke + vape + oral/mucosal’, (3) ‘smoke + vape + oral/mucosal + dab’, (4) ‘smoke + vape’, (5) ‘smoke + vape + dab’, (6) ‘vape + oral/mucosal’, (7) ‘smoke + dab’, (8) ‘smoke + oral/mucosal + dab’, (9) ‘smoke + vape + oral/mucosal + dab + topical’, (10) ‘smoke + oral/mucosal + topical’, (11) ‘oral/mucosal + topical’, and (12) ‘other combinations’ (see details in Table S1). Outcomes of interest The primary outcomes of interest were past-year CUD and CUD severity, based on Diagnostic and Statistical Manual of Mental Disorders (DSM)-5 criteria [23]. The DSM-5 defines CUD as a ‘problematic pattern of cannabis use leading to clinically significant impairment or distress’ in at least two of 11 domains of functioning within 12 months: tolerance; withdrawal; unsuccessful efforts to cut down or control use; failure to fulfill obligations at work, school or home; continued use despite recurrent physical or psychological consequences of use; and continued cannabis use despite having recurrent social or interpersonal problems caused or exacerbated by the effects of cannabis [24]. In NSDUH, each DSM-5 CUD criterion is assessed using dichotomous (yes/no) domains indicating whether the symptom/behavior occurred in the past 12 months. CUD severity count-based is consistent with DSM-5 thresholds. CUD severity was categorized as mild (endorsement of 2–3 domains), moderate (4–5 domains) or severe (≥6 domains). Covariates Key covariates included: age, sex, race/ethnicity, family income, education level, health insurance, residence status (rurality) and state medical cannabis law status at the time of interview. Cannabis-related variables comprised age at initiation (<18 versus ≥18), indicators of medical use (any medical use, all use medical) and a cannabis-use motive (recreation-only, medical-only or both); days of cannabis use in the past year [near-daily/daily (300–365 days) and non-near-daily/daily (<300 days)] [25]; perceived risk of self-harm, physically and in other ways, by smoking cannabis 1 to 2 times per week; illicit drug use (excluding cannabis); past-month nicotine dependence; past-year alcohol use disorder; any mental illness and any psychotherapeutics (i.e. pain relievers, tranquilizers, stimulants and sedatives) use disorder. See Table S2 for details on the operationalization of variables. Statistical analyses We described the socio-demographic, cannabis use, behavioral and clinical characteristics by cannabis use modes (smoke-only, vape-only, oral/mucosal-only, dab-only, topicals-only and multi-modal), comparing distributions using χ2 tests with Rao Scott’s second-order correction and reporting survey-weighted prevalence (Tables 1 and 2). Next, we assessed the association of past-year CUD and CUD severity with cannabis use modes using a χ2 test with Rao Scott’s second-order correction [Figure 1(a,b), respectively]. Survey-weighted univariable and multi-variable logistic regression estimated the association between modes of cannabis use and past-year CUD in the overall sample, with ‘oral/mucosal-only’ as the reference (Table 3). For regression analyses, vape-only and dab-only users were combined into a single vape/dab-only category to improve estimate stability, given small subgroup sizes in dab-only users, as well as conceptual similarity and shared pharmacologic profile as high-potency products administered via inhalation [26, 27]. Topical-only users were excluded because of limited sample size. Both unadjusted and adjusted odds ratios (ORs) and 95% CIs were reported. Covariates included in the model were age, sex, race/ethnicity, family income, education, health insurance, residence, state medical cannabis law, cannabis initiation age, cannabis-use motive, days cannabis used in the past year, perceived risk of smoking cannabis 1–2 times/week, past-month nicotine dependence and past-year alcohol use disorder, mental illness, illicit drug use and any psychotherapeutics use disorder. Given sex- and age-related differences in cannabis use preferences [28], we first tested effect modification by including multiplicative interaction terms (mode × sex; mode × age) in fully adjusted models (Table S3). In the absence of statistically significant interaction (P for joint test >0.05 for both), we report stratified estimates by sex and age for descriptive purposes and to facilitate interpretation. AORs and 95% CIs were reported [Figure 2(a,b)]. Next, to address potential confounding between mode of administration and frequency of cannabis use, we conducted stratified multi-variable logistic regression analyses by past-year cannabis use days, categorized as daily/near-daily (300–365 days) versus not daily/near-daily (<300 days) [25]. Within each stratum, we estimated AORs for CUD comparing mode-of-administration categories, using the same covariate set as in the primary model (Figure S2). Additionally, we performed a subgroup analysis among multi-modal users only. Weighted prevalence estimates of CUD and severity across unique multi-modal combinations were reported [Figure 3(a,b)]. In analyses restricted to multi-modal users, vaping and dabbing were retained as distinct components and combinations were classified using the observed mode-pattern distribution. We reported results for the most frequent unique combinations and grouped less common patterns as ‘other combinations’ (Table S1). Multi-variable logistic regression models estimated AORs and 95% CIs for CUD, with ‘oral/mucosal + topicals’ as the reference (Figure 4). The model was adjusted for sociodemographic, behavioral, cannabis use and mental health covariates. To limit the amount of missing data, multiple-imputed and recoded variables provided by NSDUH were used when available. Missing data across variables ranged from 0% to 0.73% (see Table S2 for details). All analyses were conducted in SAS 9.4 (SAS Institute). Inclusion of variables in the final models was guided by prior research, biological plausibility, univariate analyses and Type-3 analysis. The absence of multi-collinearity was confirmed by the variance inflation factor (VIF) values (all <2.5). All statistical tests were two-sided with an α level of <0.05. Analyses accounted for the NSDUH’s complex survey design. The primary research question and analysis plan for this study were not pre-registered on a publicly available platform, and accordingly, the findings should be considered exploratory. RESULTS Socio-demographic characteristics (Table 1) All characteristics showed statistically significant differences across methods of use (P < 0.0001) (Table 1). Most respondents reported multi-modal cannabis use (53.9%). Younger adults (18–34 years) were more likely to report dab-only (37.3%) or multi-modal use (26.9%), whereas older adults (50+) predominated in smoke-only (38.0%) and topical-only (53.1%) groups. Men predominated across most modes except oral/mucosal and topical use. Non-Hispanic Whites comprised the majority overall (64.2%), while Non-Hispanic Blacks were overrepresented among smoke-only users (21.9%). Higher-income (≥$75 000) and college-educated adults (some college or more) were more common among oral/mucosal users (63.2% and 84.8%, respectively), whereas lower-income groups (≤$49 999) were concentrated among smoke-only users (54.2%). Most respondents lived in large metropolitan areas (56.5%), with rural residence more frequent among smoke-only (13.7%) and dab-only (16.6%) users. Most participants resided in states where a medical marijuana law had been enacted at the time of interview (78.1%). Cannabis use, behavioral and clinical characteristics by modes of use Early initiation of cannabis use (<18 years) was reported by 57.9% overall, highest among multi-modal (62.2%) and smoke-only (61.0%) users (Table 2). Medical cannabis use was relatively uncommon (17.0%), but more frequent among topical-only (38.9%) users. Most participants used cannabis recreationally (83.0%), with medical-only use concentrated among topical users (31.6%). Use frequency varied sharply by mode, approximately one-third of multi-modal (32.6%) and dab-only (34.5%) users reported near-daily/daily use, compared with 7.5% of oral/mucosal-only users. Nicotine dependence was most common among smoke-only (26.0%) users and alcohol use disorder among multi-modal (27.2%) users. Perceived risk of harm from smoking cannabis 1 to 2 times per week was highest among topicals-only (68.2%) and oral/mucosal-only (67.6%) users and lowest among multi-mode (48.8%), dab-only (48.3%) and smoke-only (49.3%) users (vape-only: 59.4%, P < 0.0001). Illicit drug use (33.6%), any mental illness (45.1%) and psychotherapeutic use disorder (6.8%) were also the highest among multi-modal users. Cannabis use disorder and modes of use Overall, 30.3% of the sample had CUD, with variation across modes (P < 0.0001) [Figure 1(a)]. The prevalence of CUD was highest among multi-modal users (40.5%), followed by dab-only (28.9%) and smoke-only users (24.3%). CUD was lower among oral/mucosal-only (4.4%) and topical-only (4.1%) users. Severity of CUD mirrored these differences [Figure 1(b)]. Moderate-to-severe CUD was concentrated among multi-modal users (11.5% and 7.2%, respectively) and dab-only users (10.5% and 7.1%, respectively), compared to less than 2% severe CUD among vape or oral/mucosal users. The majority of oral/mucosal-only (95.6%) and topical-only (95.9%) users had no CUD. Multi-variable logistic regression analyses of the overall sample Compared to oral/mucosal-only users, odds of CUD were significantly higher for multi-modal (AOR = 4.14; 95% CI = 2.91–5.90), smoke-only (AOR = 2.98; 95% CI = 2.02–4.39), and vape/dab-only users (AOR = 1.89; 95% CI = 1.09–3.29) (Table 3). CUD was independently associated with younger age, male sex and race/ethnicity: 18 to 25-year-olds had nearly fourfold higher odds of CUD than those 50 years or older (AOR = 3.90; 95% CI = 3.12–4.82), and males had greater odds than females (AOR = 1.41; 95% CI = 1.25–1.58). Hispanic and non-Hispanic Blacks had elevated odds of CUD relative to non-Hispanic Whites. Likewise, participants with a high school education had higher odds of CUD (AOR = 1.26; 95% CI = 1.05–1.50) than college graduates. Early cannabis initiation (<18 years) was associated with a greater odds (AOR = 1.32; 95% CI = 1.16–1.50) relative to later cannabis initiation. Near-daily/daily cannabis use (300–365 days/year) was associated with higher odds of CUD (AOR = 6.68; 95% CI = 5.77–7.75) compared to non-near-daily/daily use. Nicotine dependence (AOR = 1.34; 95% CI = 1.09–1.66), alcohol use disorder (AOR = 1.53; 95% CI = 1.35–1.73), any illicit drug use (AOR = 1.45; 95% CI = 1.28–1.65), any mental illness (AOR = 2.43; 95% CI = 2.12–2.79) and psychotherapeutics use disorder (AOR = 1.80; 95% CI = 1.42–2.29) were independently associated with CUD. Stratified multi-variable logistic regression by sex and age In sex-stratified adjusted models [Figure 2(a)] (reference: oral/mucosal-only), multi-mode use and smoke-only use were consistently associated with higher odds of CUD in both males [(AOR = 5.69; 95% CI = 3.29–9.85) and (AOR = 3.25; 95% CI = 1.80–5.88), respectively] and females [(AOR = 8.28; 95% CI = 5.70–12.04] and (AOR = 5.43; 95% CI = 3.54–8.35), respectively], whereas vape/dab-only use was also associated with increased odds in both sexes [(males: AOR = 2.48; 95% CI = 1.11–5.55); (females: AOR = 2.11; 95% CI = 1.29–3.45)]. Likewise, in age-stratified adjusted models [Figure 2(b)] (reference: oral/mucosal-only), multi-mode and smoke-only use were associated with higher odds of CUD across all age groups (AORs ranging from 5.48–10.11 and 3.02–6.44, respectively), whereas vape/dab-only use was associated with higher odds only among those 18 to 25 years old (AOR = 2.84; 95% CI = 1.28–6.31) and 35 to 49 years old (AOR = 4.78; 95% CI = 2.12–10.77). Tests for interaction indicated no significant effect modification of the route-CUD association by sex or age (P for joint test >0.05 for both) (Table S3). Stratified multi-variable logistic regression by days of cannabis use in the past year (near-daily/daily and non-near-daily/daily) With oral/mucosal-only as reference, multi-mode (AOR = 6.27; 95% CI = 2.85–13.80), smoke-only (AOR = 4.15; 95% CI = 1.90–9.07) and vape/dab-only (AOR = 3.32; 95% CI = 1.36–8.15) use were each associated with higher odds of CUD among both near-daily/daily users and non–near-daily/daily users [(AOR = 6.60, 95% CI = 4.45–9.80); (AOR = 3.73; 95% CI = 2.45–5.67); and (AOR = 2.03; 95% CI = 1.13–3.67), respectively], indicating that the mode-CUD associations were consistent across frequency strata (Figure S2). Multi-modal methods of cannabis use and CUD In the multi-modal subgroup, the overall prevalence of CUD was 40.5% [Figure 3(a)]. The highest prevalence was among those combining smoke + vape + oral/mucosal + dabbing (72.2%) and among those using smoke + vape + oral/mucosal + dab + topicals (69.3%). Conversely, prevalence was lower among oral/mucosal + topical (3.3%) or vape + oral/mucosal (16.3%). Severity patterns paralleled these trends [Figure 3(b)]. Moderate-to-severe CUD affected 39.4% of smoke + vape + oral/mucosal + dab users, compared to oral/mucosal + topicals (0.5%) and vape + oral/mucosal (5.7%) users. Users of oral + topicals had the lowest prevalence of CUD (3.3%). In regression analysis (Figure 4), compared to oral/mucosal + topicals users, the odds of CUD were significantly elevated among those using smoke + vape + oral/mucosal + dab (AOR = 19.74; 95% CI = 9.11–42.75), smoke + vape + dab (AOR = 14.70; 95% CI = 6.79–31.84), smoke + vape (AOR = 6.76; 95% CI = 3.16–14.49) and smoke + oral/mucosal (AOR = 4.84; 95% CI = 2.25–10.41). Even combinations lacking dabbing, such as vape + oral/mucosal (AOR = 3.78; 95% CI = 1.97–8.93), carried a significantly increased odds compared to non-inhaled forms. DISCUSSION Our study found that the mode of cannabis use was strongly associated with both the prevalence and severity of CUD. With oral/mucosal-only as the reference, multi-modal use was associated with almost fourfold higher odds of CUD, while smoke-only and vape/dab-only conferred elevated odds, albeit to a lesser magnitude. The mode-CUD associations observed in the overall sample were consistent and statistically significant when stratified by age, sex and frequency of cannabis use. Combinations of multiple modes that involved smoking, vaping and dabbing were associated with the highest odds of CUD, exceeding 10-fold increases relative to the combination of oral/mucosal + topicals. Even less complex combinations, such as smoke + oral/mucosal or vape + oral/mucosal, carried significantly elevated risk, indicating that inhaled and high-potency modes of use may drive much of the observed burden of CUD. Furthermore, frequency of cannabis use, early initiation, co-occurring nicotine dependence, alcohol use disorder and mental health disorders were independently associated with CUD. The strong association between multi-modal use and CUD aligns with a mechanistic interpretation in which combining inhaled (smoking, vaping/dabbing) [26, 29] with ingested routes (edibles) [30] may amplify cumulative psychoactive THC exposure and reinforcement by merging rapid-onset effects from inhalation with prolonged effects from ingestion [17–19]. Multi-modal use may also increase exposure to high-potency products such as concentrates [31], plausibly accelerating tolerance, intensifying withdrawal and fostering compulsive use. These findings are consistent with prior work linking multi-modal use [32] and frequent use to greater dependence [33–35], cannabis-related harms [34] and polysubstance involvement [34, 35]. Similarly, a prospective cohort of adolescents identified multi-modal use as a predictor of CUD [16]. Altogether, our findings highlight multi-modal use as a central behavioral marker of elevated CUD risk, underscoring the need to assess both frequency and diversity of cannabis consumption routes. An important point to note is that NSDUH captures mode of administration, but not cannabinoid composition (THC versus CBD), dose or formulation. This is especially relevant for vaping and oral/mucosal products, which may range from CBD-only to high-THC oils/concentrates, and for topicals, which are often CBD-dominant with little or no THC. Consequently, our oral/mucosal-only and oral/mucosal + topicals reference groups may include CBD-dominant products, potentially contributing to lower observed CUD burden. Sex and age-stratified analyses indicated that, relative to oral/mucosal-only use, multi-modal and smoke-only use were associated with elevated odds of CUD across strata. Although some strata showed larger point estimates (e.g. females versus males for multi-modal and smoke-only; and adults 35–49 years old for several modes), CIs overlapped, so these differences should not be interpreted as definitive between-group contrasts. Consistent with this, formal effect-modification tests in fully adjusted models did not provide evidence of statistically significant interaction, suggesting that the route-CUD associations were broadly similar by sex and age. Together, these findings support prevention and intervention messaging that extends beyond youth and remains relevant across demographic subgroups. Within the multi-modal subgroup, combinations involving dabbing had the highest burden of CUD compared to oral/mucosal + topicals and showed the steepest severity gradient. Dabbing-involving combinations, particularly those paired with both smoking and vaping, showed the strongest associations with CUD. Given that dabbing delivers concentrated extracts containing THC in concentrations of 70% and more [36, 37] and has been linked to psychosis, lung injury and cardiotoxicity [36, 38, 39], these findings underscore potential compounded risks when high-potency products are combined with multiple delivery methods. Near-daily/daily users had more than sixfold greater odds of CUD compared to non-daily users, a pattern consistent with dependence models [40, 41]. Similarly, analyses stratified by frequency of cannabis use (near-daily/daily versus non–near-daily/daily) showed consistent mode-CUD associations in both strata, supporting that observed mode-related differences were not solely attributable to use frequency. Co-occurring nicotine dependence, alcohol use disorder, other illicit drugs and past-year mental illness also predicted CUD independently, underscoring the role of polysubstance use and mental health comorbidity. Lack of perceived harm similarly tracked with greater risk [8, 42–45]. Additionally, we observed clear sociodemographic disparities, including higher risk at younger ages, early initiation and disproportionate burden among racial/ethnic minority groups. These disparities likely reflect structural and social determinants, such as greater exposure to unlicensed retail environments [46] rather than biological differences [47–49], highlighting critical targets for prevention and policy. Although DSM-5 does not specify a minimum frequency threshold, NSDUH assesses cannabis use disorder criteria primarily among respondents reporting cannabis use on 6 or more days in the past year (or with missing frequency) [50], and discordant cases because of imputation are uncommon. Frequency was a dominant correlate in our data: 62.8% of daily/near-daily users met criteria for CUD versus 18.9% of non-daily/near-daily users. Therefore, mode-of-use associations should be interpreted in the context of substantial differences in use frequency, underscoring the value of reporting both frequency and mode when characterizing CUD risk. Strengths and limitations Strengths include the use of recent, nationally representative US data, detailed classification of cannabis use modes and comprehensive adjustment for covariates. Stratified analyses by sex, age and frequency further tested the robustness of CUD-mode associations. Limitations include the cross-sectional design, precluding causal inference and do not rule out reverse causality (e.g. individuals with emergent CUD shifting toward specific modes). All measures were self-reported, introducing potential recall or social desirability bias. Although we adjusted for a wide range of covariates, residual measurement limitations may affect our estimates. NSDUH does not capture THC potency, dose, product formulation, device characteristics or mode-specific dose/frequency, which could dilute or amplify mode-specific associations. Consequently, our combined vape/dab-only category likely groups heterogeneous exposures, introducing potential misclassification that may attenuate potency-related associations with CUD. In addition, CUD in NSDUH is a survey-based DSM-5 classification rather than a clinician-administered interview; criteria are assessed dichotomously, and severity reflects the number of criteria endorsed, not graded symptom intensity, which may introduce misclassification and limit nuance in interpreting severity. Our results suggest that the risks associated with cannabis use differ considerably by product type and route of administration, and therefore, regulators should consider a framework that classifies cannabis products into differential risk categories. For example, if cannabis was to be rescheduled, topical and edible products could be assigned to lower-risk tiers compared with inhaled products. In the context of expanding legalization and product diversity, the strong association between multi-modal use and CUD underscores the need for targeted prevention and harm-reduction efforts. Clinicians should screen not only for frequency, but also for modes and product types, while public health campaigns should highlight cumulative exposure risks from poly-route use and the disproportionate harms linked to concentrates. Policy measures could include potency labeling, product standardization and marketing restrictions that discourage multi-modal consumption. CONCLUSION In conclusion, our findings underscore that CUD risk is shaped not only by frequency, but by the mode and complexity of use, with multi-modal and inhaled routes posing the highest risk. As cannabis markets evolve, integrating product-type differentiation into regulation, clinical screening and public health messaging will be essential to mitigate harm and guide safer patterns of use. Supplementary Material Additional supporting information can be found online in the Supporting Information section at the end of this article.