Skip to main content
NCBI home page
Drug and Alcohol Dependence Reports logoLink to Drug and Alcohol Dependence Reports
. 2025 Jul 19;16:100362. doi: 10.1016/j.dadr.2025.100362

Factors associated with cannabis use disorder among Australians using prescribed and illicitly-sourced medical cannabis

Llewellyn Mills a,b,c,⁎, Jonathon C Arnold d,e,f, Iain S Mcgregor d,f,g, Nicholas Lintzeris a,b,c
PMCID: PMC12423712  PMID: 40949606

Abstract

Background

People who use cannabis medically do so more frequently than those who use nonmedically, potentially placing them at higher risk of cannabis use disorder (CUD). Prescription involves receiving guidance from doctors how best to administer medication; however, it remains unknown whether prescribed medical cannabis is associated with reduced incidence of CUD compared to illicit.

Methods

Data came from a 2022–23 online anonymous cross-sectional survey of Australians who had used medical cannabis. We examined differences between respondents who use Prescribed medical cannabis and respondents who use Illicit medical cannabis in demographic characteristics, patterns of use, and odds of meeting DSM-5 criteria for Any-CUD (≥2/11 criteria) and Moderate-Severe-CUD (≥4/11). Bayesian penalised regression models were used to identify the most important factors associated with CUD.

Results

Of 1796 respondents, 43 % met Any- and 17 % Moderate-Severe CUD criteria. In bivariate analyses, respondents who sourced illicit medical cannabis were more likely to meet criteria for Any CUD (53 % vs 41 %, OR=1.6 [CI: 1.3, 2.0]) and Moderate-Severe CUD (25 % vs 15 %, OR=2.0 [CI: 1.5, 2.6]) than those who were prescribed; however, with other factors controlled for, age, frequency-of-use, mental health, THC content, route of administration, and proportion of medical vs nonmedical cannabis use were more important correlates than whether medical cannabis was prescribed or illicitly sourced.

Conclusions

CUD is common amongst people who use medical cannabis. While CUD was less prevalent among people who obtained it on prescription than those who obtained it illicitly, other factors such as the concomitant use of cannabis for nonmedical reasons were a more important correlate with CUD.

Keywords: Medical cannabis, Medical marijuana, Substance dependence, Addiction, Epidemiology, Health policy

Highlights

  • •

    It is unknown having medical prescribed affects risk of cannabis use disorder (CUD).

  • •

    We examined whether CUD rates differed between prescribed vs illicit medical cannabis users.

  • •

    Rates of CUD (≥2 symptoms met) were high (43 %) among medical cannabis users.

  • •

    Prescribed users were less likely to meet criteria for CUD than illicit users.

  • •

    However age, mental health, and using nonmedically had stronger associations.

1. Introduction

For most of the 20th century people who wished to treat their medical conditions with cannabis had to do so illegally and without medical guidance. In the last two decades, however, the landscape has changed dramatically (Hall, 2018). Medical cannabis is now legal in 64 countries with great variety across jurisdictions in the products available and how they are used (UNODC, 2023). Yet deregulation has occurred despite: (i) the empirical evidence for the effectiveness of cannabinoids being limited to a handful of conditions, e.g. MS-related spasticity, childhood epilepsy, chemotherapy-induced nausea, and some chronic pain conditions (Abrams, 2018, Inglet et al., 2020) and (ii) an absence of clear clinical guidelines on best-practice prescribing to minimise harms to patients.

Cannabis was legalised for medical purposes in Australia in 2016, allowing any medical practitioner to prescribe a range of cannabis-based products – without special credentialing – as unregistered medicines using the compassionate access regulatory pathways (Special Access (SAS) and Authorised Prescriber Schemes; Australian Government, 2016). Cannabis-based medicines can be dispensed by any community pharmacy and can be delivered by mail or courier from a pharmacy to patients with a valid prescription. Clinical guidance, educational materials and professional development programmes are available to assist clinicians to deliver medicinal cannabis treatment (Arnold et al., 2020, National Prescriber Service, 2020, TGA, 2018), however clinicians can prescribe in any way they see fit once they have obtained the licence to prescribe (Dobson et al., 2024). Uptake of prescribed medical cannabis was initially slow, but efforts by the federal government (Parliament of Australia, 2021), the Therapeutic Goods Administration (TGA) (Therapeutic Goods Administration, 2021) and medical cannabis providers to streamline the process of obtaining prescribed medical cannabis (Department of Health and Aged Care, 2021a, Department of Health and Aged Care, 2021b) have seen a rapid increase in the number of Australian using prescribed medicinal cannabis, with estimates that hundreds of thousands of Australians have obtained prescriptions since 2016 (Leung et al., 2022b, TGA, 2024) and the Australian Institute of Health and Welfare’s (AIHW) National Drug Strategy Household Survey 2022–23 estimating prevalence of past-year medical cannabis use of 3 % (1 % only for medical purposes, 2 % sometimes for medical purposes) compared with 8 % for nonmedical purposes only (AIHW, 2024a, AIHW, 2024b).

Whether prescribed by a doctor or obtained illicitly and self-administered, medical cannabis is typically used to treat chronic conditions, such as chronic pain, mental health, sleep, and neurological conditions (Azcarate et al., 2020, Leung et al., 2022a, Lintzeris et al., 2018, Lintzeris et al., 2022, Lintzeris et al., 2020, Reinarman et al., 2011, TGA, 2024, Turna et al., 2020), often involving prolonged, regular use of medicinal cannabis products. Tetrahydrocannabinol (THC), the main psychoactive ingredient in cannabis, can be addictive and is present in many of these products, putting people who have been using medical cannabis long-term at risk of developing Cannabis Use Disorder (CUD) (Leung et al., 2020).

The Diagnostic and Statistical Manual of Mental Disorders 5th Edition (DSM-5; APA, 2013) defines CUD as a problematic pattern of cannabis use characterised by: repeated unsuccessful attempts to quit; a range of interpersonal, social, financial, physical, occupational, and cognitive issues related to using cannabis; and physiological symptoms of withdrawal and tolerance. The relationship between CUD and nonmedical use has been studied extensively, with a recent meta-analysis estimating prevalence among people who try cannabis at approximately 22 % (Leung et al., 2020) and risk factors including gender, age of initiation of use, other substance use, personality factors (e.g. impulsivity), comorbid mental health conditions, peer influence, exposure to violence, and socioeconomic status (Brook et al., 2011, Butterworth et al., 2014, Courtney et al., 2017, Hayatbakhsh et al., 2009, Rajapaksha et al., 2020).

On average those who use cannabis for medical purposes use on more days per week than those who only use nonmedically, and this appears to be true both for the relatively small proportion of people who use only for medical purposes and the majority who use both medically and nonmedically (Browne et al., 2022, Lapham et al., 2023, Lin et al., 2016, Turna et al., 2020, Wall et al., 2019). Typically medical use is also associated with higher rates of CUD than nonmedical use (Choi et al., 2017, Gendy et al., 2023, Rubin-Kahana et al., 2022, Turna et al., 2020), with a recent meta-analysis of studies from around the world including the US, Israel, Germany, and Australia estimating that 29 % (95 %CI: 21 %, 38 %) of people who used medical cannabis in the previous 6–12 months met criteria for CUD (Dawson et al., 2024) – almost a third higher than the 22 % among people who use nonmedically. However, the relationship between frequency of use and CUD is not straightforward: Recent findings from two US studies conducted in states with legal nonmedical and medical cannabis, found that people who used cannabis for medical reasons only had greater frequency of use but lower incidence of CUD (Browne et al., 2022, Lapham et al., 2023) than people who use nonmedically only. As more jurisdictions around the world legalise medical cannabis, the number of people being prescribed cannabis is growing. It is therefore important to examine the factors associated with CUD in this population.

Relatively few studies have been conducted examining the correlates of CUD among people who use cannabis medically; however, those that have found a range of correlates, including age, frequency of cannabis and tobacco use, age of first regular use, comorbid mental health or chronic pain disorders, route of administration (smoking vs oral vs edible) and using more cannabis for nonmedical reasons (see Dawson et al., 2024 for a review). Our recent analysis of CAMS-22 data revealed that people who have their medical cannabis prescribed have slightly different ‘profiles’ than those who obtain it illicitly, with respondents who mainly use prescribed cannabis more likely to consume their cannabis via oral or vaporised routes (than by smoking); to never have used cannabis nonmedically before commencing medical use; to have commenced medical use later in life; to treat a pain condition (versus a mental health condition); and to use tobacco and nonmedical cannabis less frequently than people who mainly used illicit cannabis (Mills et al., 2024). It is yet to be seen whether these differences extend to rates of CUD.

Harm reduction approaches argue that allowing people access to legal medical cannabis on prescription exposes these individuals – who otherwise would have obtained it illicitly and self-administered – to the healthcare system, allowing them to receive guidance from doctors concerning evidence for efficacy and dosing. By this reasoning rates of CUD and the factors associated with it should be different among people who obtain medical cannabis on prescription and those who mainly source it illicitly. CAMS-22 is the first CAMS survey with large numbers of respondents accessing prescribed cannabis, allowing us to examine this question. In this study we aim to: (i) estimate the prevalence of CUD among people who use medical cannabis, (ii) explore which factors are associated with CUD among people who get their medical cannabis prescribed, and (iii) explore whether the factors associated with CUD differ depending on whether individuals mainly use prescribed or illicit cannabis.

2. Methods

The data we used for the study was a convenience (i.e. non-probability-based) sample, taken from the Cannabis-As-Medicine Survey 2022 (CAMS-22). CAMS-22 was an anonymous, cross-sectional, online survey available for completion from the 16th of December 2022 to the 20th of April 2023. Respondents were adult Australians who had used cannabis to treat a medical condition in the previous 12 months. Detailed study procedures have been published previously (Mills et al., 2024). All respondents gave informed consent. The survey included questions regarding demographics, patterns of cannabis use, the health conditions respondents treat with cannabis, and general health questions (see online materials for full questionnaire), contained 100–250 questions (depending on responses) and took 20–30 min to complete. Respondents were not compensated for taking part. The study was approved by the University of Sydney Human Research Ethics Committee [#2022/433].

2.1. Outcome variables: DSM-5 criteria for cannabis use disorder

Questions used to calculate outcome variables were based closely on the eleven criteria for CUD contained in DSM-5 (APA, 2013; described in Table 2), assessing symptoms over the previous 12 months. The first ten criteria were assessed with a single Yes-No question. To assess the eleventh criteria, cannabis withdrawal, we created eight questions, seven based on the seven criteria listed in the DSM-5 criteria for cannabis withdrawal syndrome (CWS) and an extra question asking respondents if they took cannabis to relieve or avoid withdrawal symptoms (see eTable 1 online materials). DSM-5 guidelines for CUD severity CUD are: No CUD (0–1 criteria), Mild (2–3 criteria), Moderate (4–5 criteria), and Severe (≥6 criteria). As per DSM-5 criteria, if respondents met ≥ 3 of the DSM-5 CWS criteria and/or indicated they used cannabis to avoid withdrawal they were considered to have met the withdrawal criterion.

Table 2.

Proportion of sample who met individual criterion for DSM-5 cannabis use disorder.

Cannabis Use Disorder Individual Symptom Checklist
Instruction: “Please indicate below whether you have experienced any of the following IN THE PAST 12 MONTHS” 		DSM-5 Criterion Source of Medical Cannabis
Prescribed (ref)
n = 1426		 Illicit
n = 370		 Total
N = 1796		 Group differencea
OR (95 % CI)
I often take cannabis in in larger amounts or over a longer period of time than I intended to A1 245 (17 %) 108 (29 %) 353 (20 %) 2.0 (CI: 1.4, 2.6)
I have a persistent desire, or make unsuccessful attempts to cut down or control my cannabis use A2 132 (9 %) 51 (14 %) 183 (10 %) 1.6 (CI: 1.1, 2.2)
I spend a great deal of time in activities necessary to obtain cannabis, use cannabis, or recover from its effects A3 72 (5 %) 40 (11 %) 112 (6 %) 2.3 (CI: 1.4, 3.3)
I have cravings or a strong desire or urge to use cannabis A4 324 (23 %) 122 (33 %) 446 (25 %) 1.7 (CI: 1.3, 2.1)
My cannabis use results in failure to fulfil my major role obligations at work, school, or home A5 21 (1 %) 15 (4 %) 36 (2 %) 2.8 (CI: 1.4, 5.3)
I continue to use cannabis despite having persistent or recurrent social or interpersonal problems related to cannabis use (such as criminal charges, ultimatums of abandonment from spouses/partners/friends, and poor productivity) A6 61 (4 %) 33 (9 %) 94 (5 %) 2.1 (CI: 1.4, 3.3)
I have given up or reduced important social, occupational, or recreational activities because of cannabis use A7 88 (6 %) 33 (9 %) 121 (7 %) 1.5 (CI: 1.0, 2.2)
I recurrently used cannabis in situations in which it was physically hazardous (e.g. driving motor vehicle, operating machinery) A8 210 (15 %) 75 (20 %) 285 (16 %) 1.5 (CI: 1.1, 2.0)
I continued to use cannabis even though it causes problems with emotions, mental, or physical health (e.g. cough) A9 118 (8 %) 63 (17 %) 181 (10 %) 2.3 (CI: 1.6, 3.1)
I continued to use cannabis more than before to get the desired effects or the same amount of cannabis has less effects than before A10 398 (28 %) 118 (32 %) 516 (29 %) 1.2 (CI: 0.9, 1.5)
Withdrawal or using cannabis to avoid withdrawal A11 694 (49 %) 200 (54 %) 894 (50 %) 1.2 (CI: 1.0, 1.6)
Open in a new tab

a: Noteworthy differences in bold. Estimates for categorical variables are odds ratios and for numeric variables mean difference. ‘Prescribed’ is reference group, so OR < 1 means odds of respondents in prescribed group meeting criterion in question are greater than odds of illicit group meeting criterion. For numeric outcomes mean difference < 0 means illicit group had lower mean than prescribed.

Based on these questions we created two binary CUD outcome variables: Any CUD (≥2 criteria met out of a possible 11) and Moderate-Severe CUD (≥4 criteria met/11). Meeting DSM-5 criteria for Moderate-Severe CUD is considered equivalent to an ICD-11 classification of Cannabis Dependence, and can be analysed separately to examine the construct of dependence to cannabis (Degenhardt et al., 2019).

It is important to note that CUD criteria were self-reported and hence do not constitute a formal diagnosis of CUD. It is also worth noting that in 2022–23 high-potency vaping concentrates were extremely rare in Australia, hence in this survey vaping can be assumed to refer to vaping flower/bud.

2.2. Correlates of CUD

We examined 21 correlates of CUD. These are listed inTable 1 (20 as rows, one – ‘Source of Medical Cannabis’ – as columns) and described in more detail in eTable 2, online materials. Where possible we used the same covariates used in the CAMS-18 study, chosen based on: (i) a formal literature search, (ii) backwards chaining of bibliographies of papers returned by the literature search, (iii) presence of variables in the CAMS-20 dataset (Mills et al., 2022). The primary covariate – Main source of cannabis – was a binary categorical variable made up of two groups: Prescribed, whose main source is or was prescribed medical cannabis, and Illicit whose main source was or is illicitly-obtained medical cannabis. Respondents were allocated to these groups based on their answers to four questions: (i) “Have you ever accessed legal medical cannabis prescribed by a doctor in Australia?”, (ii) “In the last 28 days on how many days did you use legal (prescribed) medical cannabis?”, (iii) “Have you ever accessed illicit (non-prescribed) medical cannabis?”, (iv) “In the last 28 days on how many days did you use illicit (non-prescribed) medical cannabis?”. If respondents indicated they had used in the last 28 days they were designated ‘currently using’, if they had not, ‘formerly using’ or – if they answered ‘no’ to the question asking if they had ever used – ‘never-used’. These four questions yielded nine groups, which we allocated to the two-group covariate as follows: (i) current prescribed/never illicit (Prescribed group), (ii) ex-prescribed/never illicit (Prescribed group), (iii) never prescribed/never illicit, (iv) current prescribed/ex-illicit (Prescribed group), (v) ex-prescribed/ex-illicit, (vi) never prescribed/ex illicit (Illicit group), (vii) current prescribed/current illicit (viii) ex-prescribed/current illicit (Illicit group), (ix) never prescribed/current illicit (Illicit group). Respondents belonging to groups v and vii were asked which was their main source and included in either the prescribed or illicit groups depending on their answer. Respondents in Group iii were excluded as they indicated they had never used medical cannabis of any kind.

Table 1.

Difference in various correlates of CUD between people who use prescribed vs illicit medical cannabis.
Source of Medical Cannabis
Variable Variable Type Prescribed
(ref)
n = 1426 		Illicit
n =​​​​​​370 		Total
N = 1796		 Group Differencea
Age, in yrs-old Numeric −0.8 (CI: −2.2, 0.7)
 Mean (SD) 41.8 (12.3) 41.0 (14.2) 41.6 (12.7)
 Median (IQR) 40 (32, 50) 40 (29.2, 51.8) 40 (32, 51)
Education, n (%) Binary 0.6 (CI: 0.5, 0.8)
 No tertiary qualification (ref) 285 (20 %) 107 (29 %) 392 (22 %)
 Tertiary qualification 1141 (80 %) 263 (71 %) 1404 (78 %)
Employment, n (%) Categorical
 Employed (ref) 1002 (70 %) 218 (59 %) 1220 (68 %) 0.6 (CI: 0.5, 0.8)
 Not employed 206 (14 %) 74 (20 %) 280 (16 %) 1.4 (CI: 1.1, 2.0)
 Disability pension 177 (12 %) 63 (17 %) 240 (13 %) 1.4 (CI: 1.1, 2.0)
 Other 41 (3 %) 15 (4 %) 56 (3 %) 1.4 (CI: 0.8, 2.5)
Days per week work or study Numeric −0.7 (CI: -0.9, -0.3)
 Mean (SD) 3.3 (2.4) 2.7 (2.4) 3.2 (2.4)
 Median (IQR) 5 (0, 5) 3 (0, 5) 4.5 (0, 5)
Gender, n (%) Categorical
 Male (ref) 950 (67 %) 215 (58 %) 1165 (65 %) 0.7 (CI: 0.6, 0.8)
 Female 434 (30 %) 140 (38 %) 574 (32 %) 1.4 (CI: 1.1, 1.7)
 Non-binary 42 (3 %) 15 (4 %) 57 (3 %) 1.4 (CI: 0.8, 2.0)
Relationship status, n (%) Binary 0.7 (CI: 0.6, 0.9)
 Non-partnered (ref) 486 (34 %) 155 (42 %) 641 (36 %)
 Partnered 940 (66 %) 215 (58 %) 1155 (64 %)
Recruitment source, n (%) Categorical
 Social media (ref) 488 (34 %) 223 (60 %) 711 (40 %) 2.0 (CI: 1.7, 2.5)
 Medical cannabis provider 261 (18 %) 21 (6 %) 282 (16 %) 0.2 (CI: 0.1, 0.3)
 Other healthcare provider 9 (1 %) 3 (1 %) 12 (1 %) 1.0 (CI: 0.3, 2.5)
 Other 668 (47 %) 123 (33 %) 791 (44 %) 0.4 (CI: 0.3, 0.5)
Main condition treated, n (%) Categorical
 Mental health (ref) 535 (38 %) 143 (39 %) 678 (38 %) 1.0 (CI: 0.8, 1.3)
 Pain 498 (35 %) 112 (30 %) 610 (34 %) 0.8 (CI: 0.7, 1.0)
 Sleep 249 (18 %) 60 (16 %) 309 (17 %) 0.9 (CI: 0.7, 1.1)
 Other 144 (10 %) 55 (15 %) 199 (11 %) 1.4 (CI: 1.1, 2.0)
Days per week alcohol use Numeric 0.2 (CI: −0.0, 0.4)
 Mean (SD) 1.3 (1.8) 1.5 (2.0) 1.3 (1.8)
 Median (IQR) 0.5 (0, 1.75) 0.5 (0, 2) 0.5 (0, 1.75)
Days per week tobacco use Numeric 1.3 (CI: 1.0, 1.6)
 Mean (SD) 1.4 (2.7) 2.7 (3.4) 1.6 (2.9)
 Median (IQR) 0 (0, 0) 0 (0, 7) 0 (0, 1.5)
Duration since first tried cannabis – any reason Numeric −0.4 (CI: −1.9, 1.1)
 Mean (SD) 24.4 (12.9) 24.0 (14.1) 24.3 (13.2)
 Median (IQR) 24 (14, 33) 23 (12, 34.7) 24 (13, 33)
Duration since first tried cannabis – medical reasons Numeric 4.6 (CI: 3.5, 5.7)
 Mean (SD) 6.1 (9.7) 10.7 (11.2) 7.0 (10.2)
 Median (IQR) 2 (1, 7) 6 (3, 15) 2 (1, 9)
Duration since first started using cannabis regularlyb– medical reasons Numeric 4.5 (CI: 3.5, 5.6)
 Mean (SD) 5.0 (8.7) 9.4 (10.8) 5.9 (9.4)
 Median (IQR) 1 (1, 5) 5 (2, 13) 2 (1, 6)
Days per week medical cannabis use Numeric −0.4 (CI: -0.6, -0.2)
 Mean (SD) 6.0 (1.7) 5.6 (2.1) 5.9 (1.8)
 Median (IQR) 7 (5.25, 7) 7 (5, 7) 7 (5, 7)
Use before medical, n (%) Categorical
 Had never used non-medically before using medically (ref) 46 (3 %) 20 (5 %) 66 (4 %) 1.7 (CI: 1.0, 2.5)
 Used non-medically but had quit for a year or more 660 (46 %) 136 (37 %) 796 (44 %) 0.7 (CI: 0.6, 0.8)
 Was using non-medically when started using medically 720 (51 %) 214 (58 %) 934 (52 %) 1.4 (CI: 1.1, 1.7)
Proportion medical use Numeric −10.7 (CI: -12.8, -8.7)
 Mean (SD) 89.2 (17.0) 79.0 (22.7) 87.0 (18.8)
 Median (IQR) 99 (84, 100) 85 (66.5, 99.8) 96 (80, 100)
0 % nonmedical use, n (%) Binary 696 (49 %) 93 (25 %) 789 (44 %) 0.4 (CI: 0.3, 0.4)
Nonmedical use ≤ 10%≤ 10 %, n (%) Binary 898 (63 %) 137 (37 %) 1035 (58 %) 0.3 (CI: 0.3, 0.4)
Composition, n (%) Categorical
 Mainly THC 957 (67 %) 273 (74 %) 1230 (69 %) 1.4 (CI: 1.1, 1.7)
 Equal THC/CBD 331 (23 %) 65 (18 %) 396 (22 %) 0.7 (CI: 0.6, 0.9)
 Mainly CBD 138 (10 %) 32 (9 %) 170 (9 %) 0.9 (CI: 0.6, 1.3)
Route of administration, n (%) Categorical
 Vaporised (ref) 697 (49 %) 62 (17 %) 759 (42 %) 0.2 (CI: 0.2, 0.3)
 Smoked 363 (26 %) 240 (65 %) 603 (34 %) 5.5 (CI: 4.3, 6.7
 Oral 366 (26 %) 68 (18 %) 434 (24 %) 0.6 (CI: 0.5, 0.8)
PROMIS Global Mental Health Numeric −1.6 (CI: -2.7, -0.6)
 Mean (SD) 47.1 (9.0) 45.5 (9.4) 46.8 (9.1)
 Median (IQR) 48.3 (48.3, 41.1) 45.8 (38.8, 50.8) 45.8 (41.1, 53.3)
PROMIS Global Physical Health Numeric −1.4 (CI: -2.3, -0.4)
 Mean (SD) 47.0 (8.2) 45.7 (9.3) 46.7 (8.5)
 Median (IQR) 47.7 (42.3, 54.1) 44.9 (39.8, 50.8) 47.7 (39.8, 54.1)
Open in a new tab

‘(ref)’ means the level in question was the reference group in the regression models. Noteworthy group differences in bold. a: For continuous outcomes coefficients are mean difference, for binary and categorical outcomes, odds ratios. For the Source of medical cannabis covariate Prescribed was the reference group, therefore OR < 1 means odds of respondents in the illicit group meeting criterion in question are less than odds of respondents in the prescribed group. For numeric outcomes mean difference < 0 means illicit had lower mean than prescribed. b: ≥ 3 days per week.

2.3. Statistical analyses

Bivariate Associations: We estimated the effect of Source of Medical Cannabis on the following outcomes: (i) each of the remaining 20 covariates (ii) each of the eleven individual cannabis withdrawal criteria (iii) meeting criteria for Any CUD (meeting ≥2 criteria), (iv) meeting criteria for Moderate-Severe CUD (meeting ≥4 criteria) using single-level Bayesian regression: linear for numeric outcomes, Bernoulli for binary categorical, multinomial logistic for categorical outcomes with ≥ 2 levels. For multinomial regressions we used estimated marginal means within each global regression model to estimate the difference between people who used prescribed vs illicit medical cannabis in odds of belonging to each category of the outcome. We conducted analyses in R (R Core Team, 2016) version 4.2.2, using the brms (Bürkner, 2017) and emmeans (Lenth, 2020) packages. Priors were the default broad, noninformative or weakly regularising priors specified by the brms package’s ‘brm’ function (see eTable 3 online materials).

Multivariate associations: We modelled the effect of the covariates on Any CUD and Moderate-Severe CUD via Horseshoe logistic regression (Carvalho et al., 2009, Carvalho et al., 2010), a robust data-mining technique for model feature selection akin to ridge, LASSO, and elastic net regression which employs hierarchical Bayesian modelling to shrink coefficients for irrelevant noise variables to near-zero while simultaneously allowing stronger signals to remain relatively unchanged (Blanco and Moris, 2017). Each regression model included the 21 covariates plus terms interacting each of the remaining 20 covariates with ‘Source of Medical Cannabis’, resulting in 61 regression terms in total (see eTable 4 online materials). The Source of Medical Cannabis variable was treatment coded ([0,1]) with Prescribed as the reference category; thus non-interaction coefficients represent the effect of the covariate in question in the prescribed group only and interaction coefficients represent the difference in the effect of this covariate between the prescribed and illicit groups. We used the bayesreg package (Makalic and Schmidt, 2016) to model the data in R, using a “horseshoe+ ” prior for the regression coefficients. Repeating methods from CAMS-18, we considered ‘important’ correlates of CUD to be those covariates whose 95 % credible interval (CI) excluded 0 on the log-odds scale and ‘potentially important’ correlates to be those whose CI did not exclude 0 but where greater than 85 % of posterior probability mass lay either below or above a log-odds of 0 (Mills et al., 2022). Covariates were ranked in order of strength of association based on the amount of variance explained by the corresponding column of the design matrix (Makalic and Schmidt, 2011).

All parameters were estimated via 10,000 samples (+1000 discarded burn-in samples). Bayesian analysis yields no p-values, and inferences are made based instead on posterior credibility intervals (similar to confidence intervals). Hence, any coefficients whose 95 % credibility intervals (CIs) exclude 0 (in the case of linear regression coefficients) or 1 (in the case of odds ratios from logistic regression) will be referred to as ‘noteworthy’ or ‘notable’ rather than ‘significant’.

All analyses were conducted with available cases and no imputation of data. The data and code for this study can be found at the following repository: https://osf.io/7yek8/.

3. Results

Flow of participants through the study is reported in Fig. 1. Of the N = 4453 respondents who remained after initial exclusions, 1796 (40 %) respondents completed all the questions relating to the DSM-5 CUD criteria and all 21 covariates. These respondents’ characteristics are described inTable 1, Table 2.

Fig. 1.

Fig. 1

Open in a new tab

Study flow diagram.

The prescribed group made up 79 % of the sample (1426/1796) and vs 21 % (370/1796) in the illicit group. 43 % of respondents (95 %CI: 40 %, 45 %; 778/1796) met criteria for Any CUD and 17 % (CI: 15 %, 19 %; 306/1796) for Moderate-Severe CUD. 41 % (CI: 38 %, 44 %; 583/1426) of the prescribed group met criteria for Any CUD and 15 % (CI: 13 %, 17 %; 212/1426) for Moderate-Severe CUD compared with 53 % (CI: 48 %, 58 %; 195/370) and 25 % (CI: 21 %, 29 %; 94/370) among the illicit group, notably higher odds among the illicit group in both cases (Any CUD: OR=1.6 [CI: 1.3, 2.0]; Mod-Sev CUD: OR=2.0 [CI: 1.5, 2.6]).

There were many noteworthy between-group differences in levels and frequencies of the 20 remaining covariates, shown in Table 1. The prescribed group were notably more likely to: have a tertiary qualification; be male; be employed; be in a relationship; be treating a pain condition; be using mainly equal THC/CBD cannabis; have heard about the study through a medical cannabis provider; and to administer their cannabis via vaporised or oral routes. The prescribed group were notably less likely to: have already been using nonmedically when they commenced medical use; not be employed or be on a disability pension; use cannabis comprised mainly of THC; and smoke their medical cannabis. They consumed medical cannabis and tobacco on notably more days per week; had a shorter duration from when they first used cannabis for medical reasons; spent more days in work or study; and had notably higher scores – indicating better health – on the PROMIS-29 Global Physical and Mental health scales.

Fig. 2 shows rates of CUD severity and of meeting criteria for Any CUD and Moderate-Severe CUD. The average number of CUD criteria met by respondents was 1.8 ± 2.0 (median=1, IQR=0, 3), with respondents who mainly used illicit meeting notably more criteria on average than prescribed (2.3 ± 2.3 vs 1.7 ± 1.8; mean difference=0.7 [CI: 0.5, 0.9]). 43 % of respondents (778/1796) met criteria for Any CUD and 17 % (306/1796) for Moderate-Severe CUD. 41 % (583/1426) of the prescribed group met criteria for Any CUD and 15 % (212/1426) for Moderate-Severe CUD compared with 53 % (195/370) and 25 % (94/370) in the illicit group, notably greater odds among the illicit group in both cases (Any CUD: OR=1.6 [CI: 1.3, 2.0]; Mod-Sev CUD: OR=2.0 [CI: 1.5, 2.6]).

Fig. 2.

Fig. 2

Open in a new tab

(a) Rates of DSM-5 CUD Severity and (b) Proportion meeting criteria for Any- and Moderate-Severe CUD. Error bars in (b) represent standard error of a proportion: p(1−p)n).

Table 2 shows rates of each individual criteria. The most commonly met CUD criteria were withdrawal (50 %), tolerance (29 %), and cravings (25 %). Respondents who mainly used illicit medical cannabis were notably more likely to have met each of the eleven individual CUD criteria, except for withdrawal, tolerance, or giving up or reducing important social, occupational or recreational activities.

None of the interaction terms in the multiple regressions were noteworthy. This means: (i) there were no noteworthy differences between respondents who used mainly prescribed and respondents who used mainly illicit in the strength of association between any of the covariates and any of the CUD outcomes, (ii) because we used treatment coding for the Prescribed vs Illicit covariate (Prescribed as the reference level), all non-interaction coefficients represent the estimated effect of the covariate in question in the Prescribed group only.

Fig. 3 shows the top ten ranked covariates from the multiple regressions for each of the two CUD outcomes. Points shaded green were important, grey ‘potentially important’, white not important. Important covariates were: Age (ranked equal 1st for Any CUD, 6th for Moderate-Severe-CUD), PROMIS Global Mental Health (Any 4th, Moderate-Severe 1st), Days per week medical cannabis use (Any equal 1st, Moderate-Severe 4th), Proportion of cannabis use for medical reasons (Any equal 1st, Moderate-Severe equal 2nd), Using mainly CBD products (vs Mainly THC; Any 7th, Moderate-Severe 3rd), Smoking as route of administration (vs Vaporised; Any 6th, Moderate-Severe 5th) and Having been using nonmedical cannabis when started using medically (vs never having used non-medically before using medically; Any 5th). ‘Potentially important’ covariates were the Duration since respondents started using cannabis regularly (Any CUD 8th) and Main condition treated pain (vs Mental Health: Any 9th, Moderate-Severe 8th).

Fig. 3.

Fig. 3

Open in a new tab

Top ten correlates of Any- and Moderate-Severe CUD (in descending order of importance). Green =  important (95 %CI excludes 1), Grey =  potentially important (85 % of log-odds greater or less than 0), White =  not important. *For visualisation purposes, raw values were divided by ten, e.g. age coefficient in decades.

No other covariates met criteria for importance (for full list of coefficients see eTable 4 online materials). Model diagnostics revealed high effective sample sizes (i.e. > 1000) and well-mixed, stationary traceplots.

3.1. Post-hoc comparisons

Medical only versus dual-purpose use: Previous literature has noted differences in frequency of use between people who use for medical reasons only and those who use both medically and nonmedically. To explore differences in these two distinct groups we conducted a Bernoulli regression to estimate the difference in odds of CUD between respondents who indicated – in a question asking respondents to indicate on a 0–100 slider what proportion of their total cannabis consumption was medical – that 100 % of their cannabis use was for medical purposes (medical-only) and those who indicated 1 % or more of their cannabis use was for nonmedical purposes (dual-purpose). Median proportion of total use for medical purposes among these 1007 dual-purpose respondents was 81 % (IQR: 70 %, 90 %). 54 % (539) met criteria for Any CUD and 23 % (233) for Moderate-Severe CUD, compared with 30 % (239) and 9 % (73) among the remaining 789 medical-only respondents. Both differences were notable (Any CUD: OR=0.4 [CI: 0.3, 0.4]; Mod-Sev CUD: OR=0.3 [CI: 0.3, 0.4]).

4. Discussion

The aim of this study was to examine prevalence of, and factors associated with, cannabis use disorder among people who use medical cannabis, and to compare rates of CUD between people accessing prescribed versus illicit medical cannabis.

We observed a high rate of CUD among our respondents. At 43 %, the prevalence of CUD – meeting two or more of the DSM-5 CUD criteria – was considerably higher than the average of 29 % reported in a recent meta-analysis of studies conducted among medical cannabis consumers (Dawson et al., 2024). However, examination of the individual studies in that meta-analysis reveals rates consistent with ours: 47 % (Myers et al., 2023, Rubin-Kahana et al., 2022), 38 % (Bonn-Miller et al., 2014) and 42 % (Wall et al., 2019). Moreover, the studies reporting lower rates – 12 % (Cooke et al., 2023, Gilman et al., 2022) and 10 % (Gilman et al., 2023) – excluded participants with daily use, regular use, or CUD at baseline. Our findings are also consistent with large, stratified, randomly-sampled cohorts within Australia such as the National Drug Strategy Household Survey 2022–23 (N = 22,015), reporting rates of daily use of 53 % among people who use medically versus 12 % among people who use nonmedically (AIHW, 2024a, AIHW, 2024b). While medical use appears to be consistently associated with greater frequency of use than nonmedical (Browne et al., 2022, Lapham et al., 2023, Lin et al., 2016, Turna et al., 2020, Wall et al., 2019), whether this translates to greater rates of CUD appears to be conditional on the concurrent legal status of nonmedical use, with higher rates of CUD associated with any medical use in jurisdictions with legal medical but not nonmedical use (Choi et al., 2017, Gendy et al., 2023, Rubin-Kahana et al., 2022, Turna et al., 2020), but higher rates associated with nonmedical use than only medical use in jurisdictions where nonmedical cannabis is also legal (Browne et al., 2022, Lapham et al., 2023). The relationship between frequency of use and CUD is clearly complex and would merit further research.

The high CUD rate among people who use cannabis medically appears to be especially driven by those individuals who consume cannabis for both medical and nonmedical reasons. In our survey this group of respondents were 2.5 times more likely to meet criteria for CUD than respondents who used exclusively for medical purposes. A 2023 study similarly found that 51 % of people with medical plus nonmedical use met criteria for CUD, compared with 28 % among medical-only users (Gendy et al., 2023).

Tolerance and withdrawal were the most commonly met of the eleven CUD criteria in our study. These symptoms develop through prolonged use, irrespective of the purpose of use (e.g. following doctor’s prescription for symptom relief vs nonmedical purposes of enjoyment or relaxation) and can occur in the absence of the compulsive loss of control that is a central feature of more severe substance use disorders (Ballantyne, 2015). Yet meeting these two criteria alone would qualify as DSM-5 mild CUD, resulting in many patients receiving a diagnosis when simply taking their medication as directed. Future research should examine under what conditions tolerance and withdrawal could be excluded from consideration before making a diagnosis (e.g. if patient is taking medication as prescribed).

In bivariate analyses there were many notable differences between respondents who mainly used prescribed and respondents who mainly used illicit medical cannabis, suggesting a slightly different profile among these two groups. Relative to respondents who obtained their cannabis illicitly, the prescribed group: used products containing less THC; tended to use oral or vaporised products rather than smoking; used less nonmedical cannabis; had used medical cannabis for fewer years; and were less likely to have been using cannabis nonmedically when they commenced medical use. It follows from these differences then that the prescribed group were less likely than the illicit group to meet criteria for both Any- or Moderate-Severe CUD. Our study, being correlational, cannot shed light on the cause of these differences; they may be the result of prescribed consumers receiving medical supervision or may simply reflect pre-existing differences. Interestingly however, when the other covariates were included in the regression model: (i) there was no longer a notable association between Source of Medical Cannabis and CUD, and (ii) the effect of the other correlates on CUD did not differ notably between respondents in the Prescribed and Illicit groups. This does not mean that Source of Medical Cannabis has no association with CUD, but rather that when there is additional information about certain other factors, whether one accesses medical cannabis legally or illicitly is no longer as important correlate of CUD as those other factors.

This study highlights the most important factors associated with CUD among people who use medical cannabis. Consistent with findings from studies on nonmedical use, being younger (Butterworth et al., 2014, Kosty et al., 2017), using cannabis more frequently (Courtney et al., 2017) and having poorer mental health (Brook et al., 2011, Butterworth et al., 2014) were all associated with notably increased odds of meeting criteria for CUD. Smoking medical cannabis was associated with increased likelihood of CUD compared to vaping. A recent study of college students who used cannabis nonmedically found vaping to be associated with increased risk of cannabis-related problems compared to smoking, however the association disappeared when frequency of use was taken into account (Jones et al., 2018). High-potency vaping concentrates have been available in the US for many years, but were extremely rare in Australia in 2022–23 (though less so now in 2025), thus the discrepancy between our study and Jones and colleagues’ (2018) may reflect a difference in types of products available to consumers. Unsurprisingly, higher THC content in medication was also associated with increased odds of CUD. Concomitant nonmedical use alongside medical use was an important correlate of CUD, even when controlling for the other factors, adding to the growing body of evidence indicating that dual nonmedical and medical cannabis use is associated with significantly heightened risk of CUD (Gendy et al., 2023, Rubin-Kahana et al., 2022, Wall et al., 2019). Motives for nonmedical use differ from those of medical use with nonmedical use oriented towards positive reinforcement (enjoyment, relaxation, celebration, creativity) and medical more towards negative reinforcement (lessening the aversive symptoms of their health condition) (Vedelago et al., 2020). It may be that using cannabis for both reasons simultaneously results in an additive increase in risk of CUD, or that the ready availability of cannabis for those with a prescription removes a barrier to nonmedical use. Our study cannot determine cause; however this finding does highlight a concerning trend among this large subgroup (in fact the majority) of individuals who use medical cannabis.

Our findings have implications for how medicinal cannabis treatment is delivered. Potential patients should be informed of the risks of developing CUD with prescribed medicinal cannabis – with almost 1 in 5 participants meeting criteria for dependence, and approximately half experiencing cannabis withdrawal when discontinuing the medication – important in decision-making and informed consent procedures. Medical practitioners should also consider utilising a universal precautions framework (Gourlay et al., 2005, Lintzeris and AMCA, 2024), originally developed for prescription opioid medications for the treatment of chronic pain. This approach provides a framework for assessing patients, how medications are prescribed and dispensed, and how patient outcomes are monitored – with the aim of minimising dependence and other potential harms associated with medicinal cannabis treatment.

This study had several limitations. Being based on self-report it is likely that our estimates of CUD prevalence are higher than those that would result from diagnostic interviews with qualified clinicians. It is important to note however that several studies that have estimated CUD prevalence via interviews rather than self-report found comparable rates to our study (e.g. 47 % Rubin-Kahana et al., 2022). Future studies should estimate prevalence based on clinical interviews with qualified clinicians. Another limitation is that CAMS-22, being anonymous, was a convenience (i.e. non-probability) sample, and thus may not be representative of the overall population of people who use medical cannabis in Australia and around the world: e.g. being self-selected our respondents may use more heavily than the average person using medical cannabis. The correlational design means we cannot draw any conclusions about the direction of causation between covariates and outcomes. For example, our analysis does not allow us to consider whether increased mental health distress is a risk factor for developing CUD, or is a consequence of CUD. Additionally, anonymous surveys like CAMS are susceptible to errors of recall and insight and subject to social desirability or other biases (Dawson et al., 2024).

5. Conclusion

Our study is the first to examine correlates of CUD among a large sample of people who obtain their medical cannabis on prescription, and to compare this to people using illicitly sourced medical cannabis. Consistent with previous findings among people who use medical cannabis, prevalence of CUD was high, especially among respondents who use nonmedically and medically at the same time. Overall people who have their medical cannabis prescribed were less likely to meet criteria for CUD, suggesting that there may be some benefit to medical cannabis use being supervised by a doctor rather than self-directed use by patients. However, when other factors were taken into account, whether or not medical cannabis was prescribed was no longer an important correlate of CUD; age, frequency of use, mental health, route of administration, THC content had greater correlations with CUD. Proportion of medical use vs nonmedical use remained an important correlate of CUD even when taking all other covariates into accounts. Our study has implications for how medicinal cannabis treatment is delivered, requiring the adoption of strategies – such as the universal precautions framework – that aim to minimise addiction among patients whose medication can cause dependence. Hopefully our findings will help inform future research in probability-based samples, where there can be more confidence that results are representative of the broader population of people who use cannabis to treat their health conditions.

CRediT authorship contribution statement

Mills Llewellyn: Writing – review & editing, Writing – original draft, Visualization, Supervision, Project administration, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Jonathon C Arnold: Writing – review & editing, Project administration, Methodology. Nicholas Lintzeris: Writing – review & editing, Writing – original draft, Project administration, Methodology, Conceptualization. Iain S Mcgregor: Writing – review & editing, Project administration, Methodology.

Ethics approval and consent to participate

This study was approved by the Sydney University Human Research Ethics Committee (2022/433). In order to begin the survey participants were asked to read a linked Participant Information Statement and to tick a check box giving consent for their data to be used in the study

Funding

This research was supported by the University of Sydney Specialty of Addiction Medicine, Faculty of Medicine and Health; and the Lambert Initiative for Cannabinoid Therapeutics, a not-for-profit, philanthropically funded research program at the University of Sydney. It received no specific grant from any funding agency in the public or commercial, or sectors.

Declaration of Competing Interest

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Prof Lintzeris reports grants from the Australian National Health and Medical Research Council (NHMRC) and Camurus for unrelated research, and honoraria from Camurus and Indivior for presenting professional education. Dr McGregor reports grants from NHMRC and grants from Lambert Initiative for Cannabinoid Therapeutics during the conduct of the study for projects unrelated to the submitted work. Dr McGregor is a consultant to Kinoxis Therapeutics and has received speaking fees from Janssen. He reports patents WO2018107216A1 and WO2017004674A1 issued and licensed, patents PCT/AU2020/050941 and PCT/AU2019/903299 pending, a patent WO2019227167 and WO2019071302 issued. Dr McGregor acts as an expert witness and consultant in legal cases involving the use of medical and non-medical cannabis. Dr Arnold reports grants from the NHMRC and grants from Lambert Initiative for Cannabinoid Therapeutics during the conduct of the study for projects unrelated to the submitted work. He reports the patents WO2019227167 and WO2019071302 issued. He has served as an expert witness in various medicolegal cases involving cannabis and cannabinoids and served as a temporary advisor to the World Health Organization on their review of cannabis and cannabinoids.

Acknowledgements

We would like to acknowledge the important contribution of Associate Professor David Allsop to the creation of the original CAMS-16 questionnaire, upon which subsequent CAMS questionnaires have been based. Dave sadly passed away in 2019 and is much missed, but his influence is still being felt in the projects he helped to create.

Footnotes

Appendix A

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.dadr.2025.100362.

Appendix A. Supplementary material

Supplementary material

mmc1.docx (52.2KB, docx)

Supplementary material

mmc2.pdf (457.5KB, pdf)

References

  1. Abrams D.I. The therapeutic effects of Cannabis and cannabinoids: an update from the National Academies of Sciences, Engineering and Medicine report. Eur. J. Intern. Med. 2018;49:7–11. doi: 10.1016/j.ejim.2018.01.003. [DOI] [PubMed] [Google Scholar]
  2. Australian Institute of Health and Welfare (AIHW), 2024. National Drug Strategy Household Survey, 2022-2023, Data. Australian Government, 〈https://wwwaihw.gov.au/reports/illicit-use-of-drugs/national-drug-strategy-household-survey/data〉.〈https://www.aihw.gov.au/reports/illicit-use-of-drugs/national-drug-strategy-household-survey/data〉..
  3. Australian Institute of Health and Welfare (AIHW) Australian Institute of Health and Welfare; 2024. Medical marijuana/cannabis in the NDSHS. [Google Scholar]
  4. American Psychiatric Association, D.S.M.T.F., American Psychiatric, A., 2013. Diagnostic and statistical manual of mental disorders: DSM-5. American Psychiatric Association, Arlington, Va.
  5. Arnold J., Nation T., McGregor I.S. Prescribing medical cannabis. Aust. Prescr. 2020;43:152–159. doi: 10.18773/austprescr.2020.052. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Australian Government, 2016. Narcotic Drugs Amendment Act. Federal Register of Legislation.
  7. Azcarate P.M., Zhang A.J., Keyhani S., Steigerwald S., Ishida J.H., Cohen B.E. Medical reasons for marijuana use, forms of use, and patient perception of physician attitudes among the US population. J. General Intern. Med. 2020;35:1979–1986. doi: 10.1007/s11606-020-05800-7. JGIM. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Ballantyne J.C. Assessing the prevalence of opioid misuse, abuse, and addiction in chronic pain. Pain. 2015;156(4):567–568. doi: 10.1097/j.pain.0000000000000105. [DOI] [PubMed] [Google Scholar]
  9. Blanco F., Moris J. Bayesian methods for addressing long-standing problems in associative learning: the case of PREE. Q. J. Exp. Psychol. 2017:1–52. doi: 10.1080/17470218.2017.1358292. (just-accepted) [DOI] [PubMed] [Google Scholar]
  10. Bonn-Miller M.O., Boden M.T., Bucossi M.M., Babson K.A. Self-reported cannabis use characteristics, patterns and helpfulness among medical cannabis users. Am. J. Drug Alcohol Abus. 2014;40(1):23–30. doi: 10.3109/00952990.2013.821477. [DOI] [PubMed] [Google Scholar]
  11. Brook J.S., Lee J.Y., Finch S.J., Koppel J., Brook D.W. Psychosocial factors related to cannabis use disorders. Subst. Abus. 2011;32(4):242–251. doi: 10.1080/08897077.2011.605696. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Browne K., Leyva Y., Malte C.A., Lapham G.T., Tiet Q.Q. Prevalence of medical and nonmedical cannabis use among veterans in primary care. Psychol. Addict. Behav. 2022;36(2):121–130. doi: 10.1037/adb0000725. [DOI] [PubMed] [Google Scholar]
  13. Bürkner P.-C. brms: an R package for bayesian multilevel models using stan. J. Stat. Softw. 2017;80(1):1–28. [Google Scholar]
  14. Butterworth P., Slade T., Degenhardt L. Factors associated with the timing and onset of cannabis use and cannabis use disorder: results from the 2007 Australian National Survey of Mental Health and Well-Being. Drug Alcohol Rev. 2014;33(5):555–564. doi: 10.1111/dar.12183. [DOI] [PubMed] [Google Scholar]
  15. Carvalho C.M., Polson N.G., Scott J.G. Handling sparsity via the horseshoe. Proc. 12th Int. Conf. Artif. Intell. Stat. 2009:73–80. [Google Scholar]
  16. Carvalho C.M., Polson N.G., Scott J.G. The horseshoe estimator for sparse signals. Biometrika. 2010;97(2):465–480. [Google Scholar]
  17. Choi N.G., DiNitto D.M., Marti C.N. Nonmedical versus medical marijuana use among three age groups of adults: associations with mental and physical health status. Am. J. Addict. 2017;26(7):697–706. doi: 10.1111/ajad.12598. [DOI] [PubMed] [Google Scholar]
  18. Cooke M.E., Potter K.W., Jashinski J., Pascale M., Schuster R.M., Tervo-Clemmens B., Hoeppner B.B., Pachas G.N., Evins A.E., Gilman J.M. Development of cannabis use disorder in medical cannabis users: a 9-month follow-up of a randomized clinical trial testing effects of medical cannabis card ownership. Front. Psychiatry. 2023;14 doi: 10.3389/fpsyt.2023.1083334. -1083334. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Courtney K.E., Mejia M.H., Jacobus J. Longitudinal studies on the etiology of cannabis use disorder: a review. Curr. Addict. Rep. 2017;4(2):43–52. doi: 10.1007/s40429-017-0133-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Dawson D., Stjepanović D., Lorenzetti V., Cheung C., Hall W., Leung J. The prevalence of cannabis use disorders in people who use medicinal cannabis: a systematic review and meta-analysis. Drug Alcohol Depend. 2024;257 doi: 10.1016/j.drugalcdep.2024.111263. -111263. [DOI] [PubMed] [Google Scholar]
  21. Degenhardt L., Bharat C., Bruno R., Glantz M.D., Sampson N.A., Lago L., Aguilar-Gaxiola S., Alonso J., Andrade L.H., Bunting B., Caldas-de-Almeida J.M., Cia A.H., Gureje O., Karam E.G., Khalaf M., McGrath J.J., Moskalewicz J., Lee S., Mneimneh Z., Navarro-Mateu F., Sasu C.C., Scott K., Torres Y., Poznyak V., Chatterji S., Kessler R.C. Concordance between the diagnostic guidelines for alcohol and cannabis use disorders in the draft ICD-11 and other classification systems: analysis of data from the WHO's World Mental Health Surveys. Addiction. 2019;114(3):534–552. doi: 10.1111/add.14482. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Department of Health and Aged Care (DHAC), 2021. Australian Government response to the Senate Community Affairs References Committee report: Inquiry into Current barriers to patient access to medicinal cannabis in Australia. Department of Health and Aged Care..
  23. Department of Health and Aged Care (DHAC) Department of Health and Aged Care; 2021. New streamlined process for medicinal cannabis applications. [Google Scholar]
  24. Dobson O., Barber M., Graham M., Carter A., Savic M. The wild west of medicine’: a qualitative investigation of the factors influencing Australian health-care practitioners' delivery of medicinal cannabis. Drug Alcohol Rev. 2024;43(5):1280–1293. doi: 10.1111/dar.13847. [DOI] [PubMed] [Google Scholar]
  25. Gendy M., Taisir R., Sousa S., Costello J., Rush B., Busse J.W., Mackillop J. Prevalence of cannabis use disorder among individuals using medical cannabis at admission to inpatient treatment for substance use disorders. Addict. Behav. 2023;142 doi: 10.1016/j.addbeh.2023.107667. -107667. [DOI] [PubMed] [Google Scholar]
  26. Gilman J.M., Potter K., Schuster R.M., Hoeppner B.B., Eden Evins A. Cannabis use for medical symptoms: patterns over the first year of use. Addict. Behav. 2023;144 doi: 10.1016/j.addbeh.2023.107719. -107719. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Gilman J.M., Schuster R.M., Potter K.W., Schmitt W., Wheeler G., Pachas G.N., Hickey S., Cooke M.E., Dechert A., Plummer R., Tervo-Clemmens B., Schoenfeld D.A., Evins A.E. Effect of medical marijuana card ownership on pain, insomnia, and affective disorder symptoms in adults: a randomized clinical trial. JAMA Netw. Open. 2022;5(3) doi: 10.1001/jamanetworkopen.2022.2106. -e222106. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Gourlay D.L., Heit H.A., Almahrezi A. Universal precautions in pain medicine: a rational approach to the treatment of chronic pain. Pain. Med. 2005;6(2):107–112. doi: 10.1111/j.1526-4637.2005.05031.x. [DOI] [PubMed] [Google Scholar]
  29. Hall, W., 2018. European Union: Medical use of cannabis and cannabinoids: questions and answers for policymaking. European Monitoring Centre for Drugs and Drug Addiction, Luxembourg.
  30. Hayatbakhsh M.R., Najman J.M., Bor W., O'Callaghan M.J., Williams G.M. Multiple risk factor model predicting cannabis use and use disorders: a longitudinal study. Am. J. Drug Alcohol Abus. 2009;35(6):399–407. doi: 10.3109/00952990903353415. [DOI] [PubMed] [Google Scholar]
  31. Inglet S., Winter B., Yost S.E., Entringer S., Lian A., Biksacky M., Pitt R.D., Mortensen W. Clinical data for the use of cannabis-based treatments: a comprehensive review of the literature. Ann. Pharmacother. 2020;54(11):1109–1143. doi: 10.1177/1060028020930189. [DOI] [PubMed] [Google Scholar]
  32. Jones C.B., Meier M.H., Pardini D.A. Comparison of the locations where young adults smoke, vape, and eat/drink cannabis: implications for harm reduction. Addict. Behav. Rep. 2018;8:140–146. doi: 10.1016/j.abrep.2018.09.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Kosty D.B., Seeley J.R., Farmer R.F., Stevens J.J., Lewinsohn P.M. Trajectories of cannabis use disorder: risk factors, clinical characteristics and outcomes. Addiction. 2017;112(2):279–287. doi: 10.1111/add.13557. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Lapham G.T., Matson T.E., Bobb J.F., Luce C., Oliver M.M., Hamilton L.K., Bradley K.A. Prevalence of cannabis use disorder and reasons for use among adults in a US state where recreational cannabis use is legal. JAMA Netw. Open. 2023;6(8) doi: 10.1001/jamanetworkopen.2023.28934. -e2328934. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Lenth, R.V., 2020. emmeans: Estimated Marginal Means, aka Least-Squares Means. R package Version 1-5-3.
  36. Leung J., Chan G.C.K., Hides L., Hall W.D. What is the prevalence and risk of cannabis use disorders among people who use cannabis? a systematic review and meta-analysis. Addict. Behav. 2020;109 doi: 10.1016/j.addbeh.2020.106479. [DOI] [PubMed] [Google Scholar]
  37. Leung J., Chan G., Stjepanović D., Chung J.Y.C., Hall W., Hammond D. Prevalence and self-reported reasons of cannabis use for medical purposes in USA and Canada. Psychopharmacology. 2022;239(5):1509–1519. doi: 10.1007/s00213-021-06047-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Leung J., Lim C.C.W., Chiu V., Chung J., Mekonen T., Dawson D., Hall W.D., Chan G.C.K. Prevalence and correlates of cannabis use for medicinal reasons – an Australian cross-sectional study. Addict. Behav. Rep. 2022;15 doi: 10.1016/j.abrep.2022.100436. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Lin L.A., Ilgen M.A., Jannausch M., Bohnert K.M. Comparing adults who use cannabis medically with those who use recreationally: results from a national sample. Addict. Behav. 2016;61:99–103. doi: 10.1016/j.addbeh.2016.05.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Lintzeris, N., AMCA, W.G., 2024. Interim Guidance regarding THC based Medicinal Cannabis Treatment for Patients with Cannabis and/or other Substance Use Disorders. Australian Medicinal Cannabis Association.
  41. Lintzeris N., Driels J., Elias N., Arnold J.C., McGregor I.S., Allsop D.J. Medicinal cannabis in Australia, 2016: the cannabis as medicine survey (CAMS-16. Med. J. Aust. 2018;209(5):211–216. doi: 10.5694/mja17.01247. [DOI] [PubMed] [Google Scholar]
  42. Lintzeris N., Mills L., Abelev S.V., Suraev A., Arnold J.C., McGregor I.S. Medical cannabis use in Australia: consumer experiences from the online cannabis as medicine survey 2020 (CAMS-20) Harm Reduct. J. 2022;19(1):1–88. doi: 10.1186/s12954-022-00666-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Lintzeris N., Mills L., Suraev A., Bravo M., Arkell T., Arnold J.C., Benson M.J., McGregor I.S. Medical cannabis use in the Australian community following introduction of legal access: the 2018–2019 Online Cross-Sectional Cannabis as Medicine Survey (CAMS-18) Harm Reduct. J. 2020;17(1) doi: 10.1186/s12954-020-00377-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Makalic E., Schmidt D.F. Springer Berlin Heidelberg; Berlin, Heidelberg: 2011. A simple bayesian algorithm for feature ranking in high dimensional regression problems. [Google Scholar]
  45. Makalic, E., Schmidt, D.F., 2016. High-dimensional Bayesian regularised regression with the BayesReg package. arXiv preprint arXiv:1611.06649.
  46. Mills L., Arnold J.C., Suraev A., Abelev S.V., Zhou C., Arkell T.R., McGregor I.S., Lintzeris N. Medical cannabis use in Australia seven years after legalisation: findings from the online Cannabis as Medicine Survey 2022-2023 (CAMS-22) Harm Reduct. J. 2024;21(1):104–111. doi: 10.1186/s12954-024-00992-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Mills L., Lintzeris N., O'Malley M., Arnold J.C., McGregor I.S. Prevalence and correlates of cannabis use disorder among Australians using cannabis products to treat a medical condition. Drug Alcohol Rev. 2022;41(5):1095–1108. doi: 10.1111/dar.13444. [DOI] [PubMed] [Google Scholar]
  48. Myers M.G., Ganoczy D., Walters H.M., Pfeiffer P.N., Ilgen M.A., Bohnert K.M. Assessing the diagnostic utility of the Cannabis Use Disorder Identification Test – Revised (CUDIT-R) among veterans with medical and non-medical cannabis use. Drug Alcohol Depend. 2023;247 doi: 10.1016/j.drugalcdep.2023.109876. -109876. [DOI] [PubMed] [Google Scholar]
  49. National Prescriber Service, 2020. Medicinal cannabis: what you need to know, Information and resources for consumers and health professionals. 〈https://www.nps.org.au/professionals/medicinal-cannabis-what-you-need-to-know〉. (Accessed 23 June 2025).
  50. Parliament of Australia, 2021. Australian Government response to the Senate Community Affairs References Committee report: Inquiry into Current barriers to patient access to medicinal cannabis in Australia.
  51. R Core Team, 2016. R: A language and environment for statistical computing. R Foundation for Statistical Computing.
  52. Rajapaksha R.M.D.S., Hammonds R., Filbey F., Choudhary P.K., Biswas S. A preliminary risk prediction model for cannabis use disorder. Prev. Med. Rep. 2020;20 doi: 10.1016/j.pmedr.2020.101228. -101228. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Reinarman C., Nunberg H., Lanthier F., Heddleston T. Who are medical marijuana patients? Population characteristics from nine California assessment clinics. J. Psychoact. Drugs. 2011;43(2):128–135. doi: 10.1080/02791072.2011.587700. [DOI] [PubMed] [Google Scholar]
  54. Rubin-Kahana D.S., Hassan A.N., Sanches M., Le Foll B. Medical cannabis and past-year cannabis use disorder among adult recreational users in the United States: results from a nationally representative sample. Front. Psychiatry. 2022;13 doi: 10.3389/fpsyt.2022.836908. -836908. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. TGA, 2018. Medicinal Cannabis - guidance documents. 〈https://www.tga.gov.au/medicinal-cannabis-guidance-documents〉. (Accessed 15th Jan 2024).
  56. TGA, 2024. SAS-B Notifications Dashboard. 〈https://dashboard-data.health.gov.au/single/?appid=1066afbe-2b37-427d-8c47-2caa5082cccc&sheet=088f611b-10de-4d72-be68-ccf8d12c54e9&select=clearall〉. (Accessed 17th January 2024).
  57. Therapeutic Goods Administration, 2021. New streamlined process for medicinal cannabis applications, in: Australian Government, C. (Ed.).
  58. Turna J., Balodis I., Munn C., Van Ameringen M., Busse J., MacKillop J. Overlapping patterns of recreational and medical cannabis use in a large community sample of cannabis users Overlapping patterns of recreational and medical cannabis use in a large community sample of cannabis users. Compr. Psychiatry. 2020;102 doi: 10.1016/j.comppsych.2020.152188. [DOI] [PubMed] [Google Scholar]
  59. UNODC . United Nations Office on Drugs and Crime; 2023. World Drug Report 2023. [Google Scholar]
  60. Vedelago L., Metrik J., Amlung M. Differentiating medicinal and recreational cannabis users via cannabis use motives. Cannabis. 2020;3(1):52–63. [Google Scholar]
  61. Wall M.M., Liu J., Hasin D.S., Blanco C., Olfson M. Use of marijuana exclusively for medical purposes. Drug Alcohol Depend. 2019;195:13–15. doi: 10.1016/j.drugalcdep.2018.11.009. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary material

mmc1.docx (52.2KB, docx)

Supplementary material

mmc2.pdf (457.5KB, pdf)

Articles from Drug and Alcohol Dependence Reports are provided here courtesy of Elsevier

RESOURCES