Open access Original research

ADULT MENTAL HEALTH

Are reasons for first using cannabis associated with

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
subsequent cannabis consumption (standard THC
units) and psychopathology?
Edoardo Spinazzola ‍ ‍,1,2 Hannah Degen,3 Isabelle Austin-­Zimmerman,4
Giulia Trotta,4 Edward Chesney ‍ ‍,2,5 Zhikun Li,4 Luis Alameda,3,6 Bok Man Leung,3,7
Yifei Lang,3 Andrea Quattrone,8 Diego Quattrone,2,4 Erika Castrignanò,9 Kim Wolff,9
Robin Murray,3 Tom P Freeman,10 Marta Di Forti4

Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
► Additional supplemental ABSTRACT
material is published online Background Reasons for first using cannabis (RFUC) WHAT IS ALREADY KNOWN ON THIS TOPIC
only. To view, please visit the ⇒ Motivations for cannabis use are associated
journal online (https://​doi.​ may influence later use patterns and mental health
org/​10.​1136/​bmjment-​2025-​ outcomes. However, limited research has explored with frequency of use and increased probability
301810). self-­medication versus social RFUCs in depth, and of developing clinical psychosis.
their associations with cannabis use patterns and
For numbered affiliations see WHAT THIS STUDY ADDS
end of article. psychopathology in the general population.
Objectives We examined RFUCs and their associations ⇒ We analysed data from 3389 people with
Correspondence to with (1) reasons for continuing cannabis use, (2) weekly lifetime cannabis use, between March 2022 and
Dr Edoardo Spinazzola, THC (delta-­9-­tetrahydrocannabinol) unit consumption July 2024, from the Cannabis&Me study, the
Psychosis Studies, King’s College and (3) symptoms of paranoia, anxiety and depressive largest independent study of its kind.
London, London, UK; ​edoardo.​ ⇒ We applied and validated the weekly THC
spinazzola@​kcl.​ac.​uk symptoms.
Methods We analysed data from the Cannabis&Me (delta-­9-­tetrahydrocannabinol) units as a
TPF and MDF are joint senior (CAMe) population survey (March 2022–July 2024), standardised measure of consumption.
authors. including 2573 (75.9%) current and 816 (24.1%) past ⇒ First evidence that starting to use cannabis
cannabis users aged 18 years or older. to reduce either physical or psychological
Received 22 May 2025 discomfort, separating the two groups usually
Accepted 24 June 2025 Findings Participants reported a mean weekly
consumption of 206 THC units (SD=268). Initiating analysed together, is associated with (1) high
cannabis use for anxiety (β=36.22, p=3.3e−03), weekly THC unit consumption and (2) higher
depression (β=40.37, p=1.74e−03) or because ’family paranoia, anxiety and depressive symptoms.
members were using it’ (β=87.43, p=1.22e−09) was HOW THIS STUDY MIGHT AFFECT RESEARCH,
associated with higher weekly THC units. RFUC to PRACTICE OR POLICY
relieve physical discomfort (β=8.89, p=4.12e−07), pain
⇒ Specific subgroups of people who use cannabis
(β=7.24, p=5.56e−06), anxiety (β=9.67, p=1.63e−16),
depression (β=9.12, p=1.21e−13) and minor psychotic appear to be at greater risk of developing
symptoms (β=16.46, p=1.2e−04) were linked to higher adverse outcomes and therefore may require
paranoia scores. Similar associations were observed targeted treatment interventions.
⇒ Given the ongoing trend towards legalisation
for anxiety and depression. Conversely, starting for
fun (β=−3.71, p=3.49e−05) or curiosity (β=−2.61, of cannabis for both recreational and medicinal
p=5e−03) was associated with lower paranoia and purposes, particular attention should be given
anxiety. RFUC for ’boredom’ was linked to increased to those who report self-­medicating from either
depression (β=1.09, p=3.8e−03). psychological or physical distress.
⇒ Asking people why they started using cannabis
Conclusions Initiating cannabis use for self-­medication
is associated with higher average THC consumption, and could become an easy and cheap screening
increased anxiety, depression and paranoia. tool for identifying cannabis users who require
Clinical implications Asking individuals why they first monitoring, support and triage to interventions.
used cannabis may serve as a cost-­effective screening
tool to identify those who could benefit from monitoring,
support, or referral to intervention services. with first-­episode psychosis and healthy controls
© Author(s) (or their were more likely to start using cannabis with
employer(s)) 2025. Re-­use friends (75.6%) than for any other reason,3
permitted under CC BY. confirming previous evidence that most people
Published by BMJ Group. BACKGROUND use cannabis in a social context and for hedo-
To cite: Spinazzola E, The reason why an individual starts and continues nistic reasons.1 4 Our path analysis suggested that
Degen H, Austin-­ to use cannabis may predict subsequent patterns those who start using cannabis ‘to feel better’ were
Zimmerman I, et al. BMJ and mental health outcomes.1 2 In a recent significantly more likely to progress to daily use of
Ment Health 2025;28:1–8. case–control study, we found that both people high-­potency cannabis and to suffer a first episode
Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810    1
 Open access
of psychosis later.3 Heavy cannabis use is associated with Measures and assessments
various adverse health outcomes, such as psychosis,5 anxiety6 Participants completed a 40-­ min online survey through the
and major depression.7 However, only a subset of people who http://www.onlinesurveys.ac.uk/ platform. Socio-­ demographic

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
use cannabis experience such adverse effects.8 Early onset and data were collected using the Medical Research Council Socio-
frequent consumption of high-­potency cannabis appear to be demographic Schedule.16 Participants were assessed using the
linked to such outcomes.9 Generalised Anxiety Disorder Scale (GAD-­ 7),17 the Patient
18
Paranoia, a common adverse effect of heavy cannabis use, Health Questionnaire (PHQ-­9) and trait paranoid ideation was
involves irrational fears of others’ harmful intentions and has assessed with the GPTS.13 For all our cannabis use measures, we
been widely studied in experimental10 and epidemiological used a modified version of the Cannabis Experiences Question-
studies.11 While a symptom of psychosis, paranoia also exists in naire.19 The changes to the questionnaire included expanding
the general population, affecting health, emotional well-­being, the RFUC response options from 4 to 10, enabling us to capture
social functioning and social inclusion.12 The Green et al Para- a broader range of RFUC. This approach aimed to maximise
noid Thoughts Scale (GPTS) was built to measure such trait coverage by reflecting the widest possible variety of motiva-
paranoid ideation in the general population.13 tions while also distinguishing key domains, such as (1) self-­
To investigate whether reasons for first using cannabis (RFUC) medication/coping and (2) enhancement or social use (see also
can be used to identify those cannabis users more likely to online supplemental file 1).

Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
endorse harmful patterns at an early stage, we sought to esti- Participants were asked, ‘Why did you first try cannabis?’,
mate associations between RFUC, subsequent cannabis use and with multiple-­choice responses: (1) my friends were using it; (2)
mental health symptoms in a general population sample. London my family members were using it; (3) to feel better (to get relief
was identified as a key global arena for targeted intervention, from physical discomfort); (4) to feel better (to get relief from
given the high levels of cannabis use regionally and the associ- pain); (5) to feel better (to get relief from anxiety symptoms);
ation with elevated incidence of psychotic disorders.14 To over- (6) to feel better (to get relief from depressive symptoms); (7)
come a lack of standardisation in cannabis research, cannabis use to feel better (to get relief from experiences such as hearing
was assessed using a direct measure of the primary psychoactive voices or feeling suspicious); (8) for curiosity; (9) for fun; (10)
constituent, ‘the standard THC unit’, as recommended by the US to overcome boredom. The following abbreviations are used
National Institutes of Health.15 throughout this paper: (1) ‘Friends’, (2) ‘Family’, (3) ‘Better—
physical discomfort’, (4) ‘Better—pain’, (5) ‘Better—anxiety’,
(6) ‘Better—depression’, (7) ‘Better—psychosis’, (8) ‘Curiosity’,
Objectives (9) ‘Fun’ and (10) ‘Boredom’. Subjects were able to provide up
Using data from a large population sample of cannabis users to 10 RFUC. In addition, participants were asked about their
(Cannabis&Me (CAMe) project), who have never been diag- RCUC: ‘Why do you continue to use cannabis?’, and they could
nosed or treated with a psychotic disorder, we sought to (1) select one of the following responses: (1) I like the effect, it gives
identify the most common RFUCs, (2) investigate the correla- me a buzz; (2) It makes me feel relaxed; (3) It makes me feel
tion between RFUC and reasons for continuing to use cannabis less nervous and anxious; (4) it makes me feel more sociable;
(RCUC) and for the first time, (3) to determine which RFUCs (5) other. The following abbreviations were used: (1) ‘buzz’, (2)
are associated with a standardised measure of cannabis use ‘relaxed’, (3) ‘less nervous and anxious’, (4) ‘sociable’ and (5)
(higher THC units, delta-­ 9-­
tetrahydrocannabinol, consumed) ‘other’.
which takes into account both the frequency and the potency of Frequency of cannabis use was categorised as never or occa-
the cannabis consumed, and (4) explore how RFUCs are linked sional use=0; monthly or less than monthly=1; weekly or
to higher levels of anxiety, depression and paranoia. less=2; more than once a week=3; daily=4.
Self-­
reported data on the frequency and potency of the
cannabis used were collected, as demonstrated to be reliable
METHODS measures of the quantity of THC used.19–21 Participants could
Study design report directly the potency (THC%) of the cannabis used or the
We used data from the CAMe study online survey, a population-­ name of the type used; information on the quantity of cannabis
based, non-­clinical sample of adult current, past (not in the last used g/week was also collected (see online supplemental file 1).
year) and never cannabis users. THC consumption was estimated using standard weekly THC
units (1 unit=5 mg THC), consistent with the National Institutes
of Health mandating reporting.15 Standard THC units provide
Sample a direct measure of the primary psychoactive constituent in
The analyses in this paper are based on data collected between cannabis, THC. To reduce the variability of standard THC unit
30 March 2022 and 31 July 2024, comprising a total sample estimates, winsorisation was applied.22 Quantitative analysis
of N=3389 participants with cannabis use. To achieve a sample using liquid chromatography coupled with mass spectrometry
as representative as possible of the London general population, measured THC and its metabolites.
multiple channels, major social media platforms, targeted efforts Several strategies were applied to ensure data completeness
and collaboration with marketing agency experts in sample (see online supplemental file 1).
recruitment (https://literalhumans.com/) were used.
Participants were aged≥18 years, residents of the London area
or able to travel, English speakers and consented to be contacted Statistical analyses
for a potential follow-­up face-­to-­face assessment. Those with a Statistical analyses were performed using R Studio (V.4.2.1).
current or past diagnosis of psychotic disorders were excluded. Descriptive statistics summarised the sample, with continuous
A subsample of N=88 participants took part in a subsequent variables expressed as means and SD and categorical variables
face-­to-­face assessment. This included a blood sample collection presented as frequencies. Group differences were evaluated
with measurement of THC and its metabolites. using t-­tests or Mann-­ Whitney U tests (online supplemental
2 Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810
 Open access

FINDINGS
Table 1 Participants’ characteristics in the working sample
(N=3.389 past and current cannabis users)
Sociodemographic characteristics
The characteristics of the sample population of 3389 participants

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
Variable Statistics Descriptor in terms of general sociodemographics, measures of cannabis
Demographics use and psychopathology (GPTSTOTAL, GAD and PHQ-­9) are
 Age M (SD), median (IQR) 30.9 (9.8), 29 (12) summarised in table 1. Data from respondents who reported
 Sex Males (%) 2103 (62.1) past cannabis use 816 (24.1%) or who reported current cannabis
 Ethnicity White/white other (%) 2206 (65.1) use 2573 (75.9%) were considered (online supplemental figure
 Employment status Employed (%) 2802 (85.4) S1, recruitment flow chart).
 Years of education M (SD), median (IQR) 15.9 (3.9), 16 (4) Participants had a mean age of 30.9 years (SD: 9.8); 2103
Cannabis use (62.1%) were male, 2206 (65.1%) were either white British or
 Cannabis use Yes (%) 3389 (100) any other white group, 427 (12.6%) reported mixed ethnicity,
 Cannabis use Current use (%) 2573 (75.9) 380 (11.2%) black British ethnicity and 376 (11.1%) Asian
 Cannabis use Past use only (%) 816 (24.1)
or British Asian ethnicity. The majority were employed, 2802
(85.4%) and reported a mean of 15.9 years of education (SD:
 Frequency of use Daily use (%) 1719 (50.9)
3.9). The average age at first cannabis use was 16.7 years (SD:

Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
 THC unit M (SD), median (IQR) 206 (268), 112 (232)
5.6), and half (1719, 50.9%) of the sample used cannabis daily
 Age first tried M (SD), median (IQR) 16.7 (5.6), 16 (4)
(table 1). The mean weekly consumption of standard THC units
Psychopathology
was 206 (SD: 268). In the subsample of 88 participants with
 GAD-­7 M (SD), median (IQR) 6.1 (5.3), 5 (7)
available THC blood level data, the mean concentration was
 PHQ-­9 M (SD), median (IQR) 7.8 (6.5), 6 (9) 11.5 ng/mL (SD: 10.8).
 GPTS-­A M (SD), median (IQR) 27.4 (11.9), 23 (15) More than two-­thirds of participants (69.8%, 2366) reported
 GPTS-­B M (SD), median (IQR) 23.3 (11.9), 17 (9) starting to use cannabis because of ‘friends’; while 62.3% (2111)
 GPTS-­TOT M (SD), median (IQR) 50.8 (22.5), 42 (24) stated starting for ‘curiosity’ and 52.7% (1787) for ‘fun’. Much
GAD-­7, Generalised Anxiety Disorder assessment; GPTS, Green et al Paranoid lower numbers reported starting use for ‘Better—anxiety’
Thoughts Scale; GPTS-­A, ideas of reference; GPTS-­B, persecution; GPTS-­TOT, GPTS (15.4%, 521); ‘Better—depression’ (13.8%, 469); to reduce
total score; M, mean; PHQ-­9, Patient Health Questionnaire for depression. ‘boredom’ (11.8%, 399); because of ‘family’ members (10.5%,
355); ‘Better—pain’ (7.6%, 257); ‘Better—physical discomfort’
(6%, 204). Online supplemental figure S3 reports overlapping
table S3; table 1). Since participants could provide more than answers for the RFUC variables.
one RFUC, the number of reported RFUC and the overlap
(online supplemental figure S3) were calculated, as well as the RCUC and correlation between RFUC and RCUC
subsample with current cannabis use, followed by a test for the The RCUC distribution for those answering, ‘Why do you
correlation between RFUC and RCUC (online supplemental continue to use cannabis?’ (N=2573) was calculated as a
figure S4). subset with 81.2% (2,089) reporting ‘relaxed’ as their RCUC.
A series of adjusted linear regressions were carried out to About half (51.3%, 1320) reported RCUC ‘nervous’, ‘buzz’
test the association between the following 10 RFUC, age at (47.5%, 1222), while less than one-­third (27.1%, 698) reported
first cannabis use, level of cannabis use in standard THC units, ‘sociable’, or any ‘other’ reasons (19%, 489) (figure 1).
GPTSTOTAL, GAD and PHQ-­9 scores. The Bonferroni correc- There was a moderate positive correlation between RFUC
tion was applied for five independent tests (p<0.01). Missing for ‘fun’ and RCUC to feel ‘relaxed’ (r=0.41), and between
data (25.3%) for models that included the standard THC unit RFUC with ‘friends’ and RCUC ‘relaxed’ (r=0.39) among
were mainly due to ambiguous free-­test entries for g/week. The current cannabis users. We also found a low-­to-­moderate
Multiple Imputation by Chained Equations (MICE, in R soft- correlation between RFUC ‘Better—anxiety’ and RCUC
ware) was used to address non-­random missingness.23 24 ‘nervous’ (r=0.36), between RFUC ‘curiosity’ and RCUC
Predictive Mean Matching was applied for continuous vari- ‘relaxed’ (r=0.38), and between ‘fun’ and RCUC ‘buzz’
ables to ensure realistic imputations. Five imputations (m=5) (r=0.37). RFUC ‘Better—pain’ showed weak but still signif-
and a maximum of five iterations (maxit=5) were used to ensure icant levels of correlation with RFUC ‘nervous’ (r=0.19),
result convergence and stability. All analyses were adjusted for ‘sociable’ (r=0.15) and ‘relaxed’ (r=0.14). Similarly, RFUC
age, sex, ethnicity, education, and employment status. ‘Better—physical discomfort’ was weakly but significantly
To ensure additional robustness, sensitivity analysis using a correlated with RCUC ‘nervous’ (r=0.16), ‘relaxed’ (r=0.13)
series of multinomial logistic regressions adjusted for age, sex, and ‘sociable’ (r=0.12) (online supplemental figure S4).
ethnicity, years of education and employment status was used to
test the association between each of the RFUC and a composite
measure of the frequency of cannabis use (online supplemental Associations between age at first cannabis use and RFUC
table S8). A series of linear regressions adjusted for age, sex, ethnicity,
employment status and years of education indicated that
RFUC with ‘friends’ (β=−1.1; (95% CI: −1.52 to –0.68),
Validation of the standard THC unit measure p=3.17e−07), with ‘family’ (β=−1.58, (95% CI: −2.21 to
Blood THC concentrations, from a subgroup who participated –0.95), p=9.5e−07) and for ‘fun’ (β=−0.67; (95% CI: −1.05
in a later face-­to-­face assessment, were strongly correlated with to –0.28), p=6.57e−04), were associated with lower age of
weekly standard THC units (Spearman’s r=0.77, p=3.08e−13) first cannabis use, while RFUC ‘Better—physical discomfort’
as well as frequency of use (Spearman’s r=0.63, p=1.73e−09) (β=1.72, (95% CI: 0.91 to 2.52), p=2.87e−05) and ‘Better—
(see online supplemental paragraph 2.9). pain’ (β=1.46, (95% CI: 0.73 to 2.19), p=9.42e−05) were
Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810 3
 Open access

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
Figure 1 (A) Reasons for first using cannabis as depicted by the number of participants and percentage of sample with lifetime cannabis use.
(B) Reasons for continuing to use cannabis as depicted by the number of people and percentage of sample with current cannabis use.

associated with age of onset of use (figure 2; online supple- indicated a tendency toward higher use (online supplemental
mental table S6). tables S5 and S7).

Association between RFUC and standard THC unit Associations between RFUC and paranoia, anxiety and
Linear regressions tested the association between each of the depression scores
RFUC and THC exposures measured in weekly standard THC Linear regressions were conducted to test the associations
units. RFUC ‘family’ (THC mean: 286.9 (SD: 357); β=87.4 between the 10 RFUC and paranoia (GPTSTOTAL score), anxiety
(95% CI 59.4 to 115.5), p=1.22e−09), ‘Better—anxiety’ (GAD-­ 7 score) and depressive symptoms (PHQ-­ 9 score).
(THC mean: 248 (SD: 294.1); β=36.2 (95% CI 12.1 to 60.3), Including the THC unit measure as a covariate had a marginal
p=3.3e−03) and ‘Better—depression’ (THC mean: 254.7 (SD: impact on the results, with a slight reduction in most effect
291.5); β=40.37 (95% CI 15.1 to 65.7), p=1.74e−03) were sizes. Replacing it with frequency of use had an even smaller
associated with higher average weekly THC units. Those who effect, with no changes in the direction of associations or statis-
started to use for ‘Better—pain’ (THC mean: 250.5 (SD: 337.2); tical significance across all regressions (table 2; figure 2; online
β=20.4 (95% CI: −12.5 to 53.3), p=0.22) and for ‘Better— supplemental table S9).
physical discomfort’ (THC mean: 248.5 (SD: 318.8, median:
136); β=16.2 (95% CI −20.2 to 52.7), p=0.38) were not DISCUSSION
significantly associated with higher weekly THC unit consump- To our knowledge, this is the largest cross-­sectional study of
tion, although both the high mean levels and high β coefficients adult cannabis users to explore how RFUC correlate with reasons
4 Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810
 Open access

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
Figure 2 Adjusted linear regressions to test the associations between RFUC and (1) age at first cannabis use, (2) THC unit, (3) GPTS-­tot, (4) GAD
and (5) PHQ-­9. All the analyses were adjusted for sex, age, ethnicity, employment status and years of education. Additional adjustment for THC is
also reported. We applied a Bonferroni correction for 10 independent reasons for first using cannabis, setting the significance threshold at p<0.005.
Significant p values (below this threshold) are shown in solid lines, while non-­significant p values, following correction, are represented with dotted
lines. GAD, Generalised Anxiety Disorder assessment; GPTS, Green et al Paranoid Thoughts Scale; GPTS-­tot, Green et al Paranoid Thoughts Scale Total
Score; PHQ-­9, Patient Health Questionnaire for depression; RFUC, reasons for first using cannabis.

Table 2 Adjusted linear regressions to test the associations between RFUC (Reasons for first using cannabis) and (1) GPTS-­tot, (2) GAD-­7 and (3)
PHQ-­9
GPTS-­TOTAL GAD PHQ-­9
β (SE) P value β (SE) P value β (SE) P value
Friends −0.33 (0.83) 0.69 −0.24 (0.2) 0.23 −0.02 (0.24) 0.93
Adjusted for THC unit 0.25 (0.96) 0.79 −0.17 (0.23) 0.45 0.05 (0.28) 0.85
Family 3.16 (1.24) 1.1e−02 0.76 (0.3) 1.02e−02 1.11 (0.36) 1.88e−03
Adjusted for THC unit 2.15 (1.41) 5e−03 0.56 (0.3) 9e−02 0.82 (0.41) 0.04
Better—physical discomfort 8.63 (1.59) 6.51e−08 0.99 (0.38) 8.43e−03 1.29 (0.46) 5.04e−03
Adjusted for THC unit 8.89 (1.75) 4.12e−07 1.07 (0.41) 0.01 1.42 (0.51) 5.03e−02
Better—pain 8.14 (1.44) 1.63e−08 1.41 (0.34) 3.94e−05 1.82 (0.41) 1.07e−05
Adjusted for THC unit 7.24 (1.59) 5.56e−06 1.31 (0.38) 4.75e−04 1.57 (0.46) 6.41e−04
Better—anxiety 9.56 (1.04) 8.1e−20 2.69 (0.25) 2.42e−27 3.2 (0.3) 1.56e−26
Adjusted for THC unit 9.67 (1.16) 1.63e−16 2.66 (0.27) 4.35e−22 3.07 (0.3) 7e−20
Better—depression 9.7 (1.10) 1.6e−18 2.38 (0.26) 9.11e−20 3.72 (0.31) 4.25e−32
Adjusted for THC unit 9.12 (1.22) 1.21e−13 2.28 (0.29) 3.65e−15 3.60 (0.35) 1.38e−24
Better—psychosis 20.89 (3.9) 9.14e−08 4.35 (0.93) 2.68e−06 5.26 (1.12) 2.74e−06
Adjusted for THC unit 16.5 (4.26) 0.0001 4.02 (1) 6.26e−05 5.12 (1.22) 3.06e−05
Curiosity −3.49 (0.79) 9.32e−06 −0.47 (0.19) 0.01 −0.23 (0.23) 0.3
Adjusted for THC unit −2.61 (0.93) 0.005 −0.44 (0.22) 4.18e−02 −0.19 (0.27) 0.48
Fun −3.88 (0.76) 3.39e−07 −0.73 (0.18) 5.39e−05 −0.55 (0.22) 1.23e−02
Adjusted for THC unit −3.71 (0.9) 3.49e−05 −0.73 (0.21) 6.04e−04 −0.55 (0.26) 3.29e−02
Boredom 3.66 (1.17) 1.79e−03 0.65 (0.28) 1.97e−02 1.34 (0.34) 6.96e−05
Adjusted for THC unit 3.06 (1.31) 0.019 0.52 (0.31) 9.12e−02 1.09 (0.38) 3.8e−03
All the analyses were adjusted for sex, age, ethnicity, employment status and years of education. Additional adjustment for THC is also reported. We applied a Bonferroni
correction for 10 independent reasons for first using cannabis, setting the significance threshold at p<0.005.
GAD-­7, Generalised Anxiety Disorder assessment; GPTS-­tot, Green et al Paranoid Thoughts Scale Total Score; PHQ-­9, Patient Health Questionnaire for depression; RFUC, reasons
for first using cannabis; β, beta coefficient.

Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810 5

 Open access
for continuing use, and for the first time to investigate if RFUC Among these, ‘anxiety’ (15.4%) and ‘depression’ (13.8%) were
were associated with standard weekly THC units consump- the most common, followed by ‘pain’ (7.6%) and ‘physical
tion, paranoia, anxiety and depressed mood, in cannabis users discomfort’ (6%). Although using different methodologies, a

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
who had not been diagnosed or treated for psychotic disor- recent survey conducted in the USA and Canada on self-­reported
ders. Participants (N=3389 reporting cannabis use; N=2573 reasons for medical cannabis use found that pain (53%) was the
current, N=816 past) were young adults (mean age 30.9 years, most common physical health motive, while anxiety (52%) and
SD: 9.8), mostly male (62.1%) and predominantly white British/ depression (40%) were the most common mental health ones.2
white Other (65.1%). The majority were employed (85.4%) In that study, physical health motives slightly outweighed mental
and spent an average of 15.9 years in education. The average health ones overall. This might be the result of the demographics
age at first cannabis use was 16.7 years, and half (50.9%) used of the age at first cannabis use in our RFUC sample (16.7 years).
cannabis daily (table 1). The mean weekly standard THC unit At this age, people are less likely to start using cannabis to get
consumption was 206 THC units, significantly higher than the relief from physical health problems. However, as reported
60–80 units reported in the CannTeen study.22 This is likely due previously, while the average age at first use does vary among the
to the stringent enrolment criteria of the CannTeen study, which different RFUC, these variations are not particularly pronounced
excluded, for instance, participants with a diagnosis of cannabis (online supplemental tables S4 and S6) and, more importantly, in
use disorder or any physical health condition deemed problem- all categories related to using cannabis ‘to feel better’, scores for

Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
atic by a medical doctor, as well as those using cannabis more GPTSTOTAL, GAD and PHQ-­9 are higher compared with cate-
than once a week during adolescence (for the adult group). In gories like ‘friends’, ‘fun’ and ‘curiosity’. Interestingly, a small
contrast, our study included a broader range of participants, US-­based survey conducted in 2020 showed that 76% of their
making our sample more representative of the demographics and sample reported using cannabis to self-­ medicate from either
clinical characteristics of the London cannabis-­using population. physical or psychological discomfort.26 Our study, conducted in
Starting to use cannabis for anxiety or depression was signifi- a country where recreational cannabis is still officially illegal,
cantly correlated with continuing to use cannabis to be ‘less shows that the majority of people report starting to use cannabis
nervous and anxious’, showing low-­to-­moderate correlations of in the social context and for enhancement rather than ‘to
r=0.36 and r=0.31, respectively. These correlations are some- feel better’. Notably, the RFUC categories ‘Better—pain’ and
what lower than expected, suggesting that reasons for cannabis ‘Better—physical discomfort’ showed weak correlations with all
use change over time. This aligns with findings from a previous reasons for continued use and were both associated with higher,
study conducted on patients with first-­episode psychosis, which though not statistically significant, weekly THC unit consump-
reported a reduction in the strength of endorsed reasons for tion. A US-­based epidemiological survey suggested that people
cannabis use at 3 months and 12 months compared with base- using medical cannabis for pain were more likely to develop
line.25 Age at first cannabis use differed across various RFUC. patterns of use similar to recreational cannabis users. Addition-
Participants reporting starting use for pain and ‘physical discom- ally, participants with pain were more likely to develop frequent
fort’ as RFUC initiated use marginally later than those who cannabis use compared with people without pain. The authors of
used it to alleviate psychological distress. Interestingly, the the study concluded that adults with pain are a group at higher
earliest onset was observed among those who reported starting risk of developing adverse cannabis outcomes.27 Indeed, we
for minor ‘psychotic symptoms (mean: 15.3 (SD: 2.7)) or with found that starting for ‘pain’ and ‘physical discomfort’ was both
‘family’ (mean: 15.2 (SD: 3.8)). associated with higher levels of psychopathology.
All RFUC groups had average GAD scores within the mild
anxiety range (5–10), except for ‘Better—psychosis’ (mean:
11.1 (SD: 5.3)), which fell within the moderate anxiety cate- Limitations and strengths
gory, exceeding the threshold for follow-­up or symptom moni- Participants were recruited through advertisement, potentially
toring.17 Similarly, most RFUC groups had mean depression leading to a non-­entirely representative sample of the London
in the mild range (5–9), except for ‘Better—pain’ (mean: 10.1 general adult population (≥18 years). Nevertheless, compared
(SD: 7.2)), ‘Better—anxiety’ (mean: 11.1 (SD: 7.2)), ‘Better— with the ethnic breakdown against the latest London census
depression’ (mean: 11.7 (SD: 7.2))) and ‘Better—psychosis’ data,28 our sample showed an ethnic distribution (online supple-
(mean: 14.1 (SD: 7)), which fell within the moderate depression mental figures S8 and S9) representative of the ethnic diversity
severity range (10–14), meeting criteria for referral counselling and distribution of the London area. Also, while the census data
for follow-­up18 (online supplemental table S10). refers to the overall London population, our sample focused on
cannabis users, making it plausible that some of the unrepre-
sented groups were simply less likely to use cannabis, as reported
Comparison with previous research by previous studies, which included data on cannabis use among
In a previous case–control study, we found that 75.6% of first-­ participants from the London area.29
episode psychosis patients and 86.1% of controls reported The cross-­sectional nature of our study does not allow us to
starting using cannabis because of ‘friends’.3 In November 2018, draw any conclusions in terms of causality, particularly regarding
in the UK, medicinal cannabis became legal, and this might be the association between RFUC and psychopathology. Neverthe-
reflected in the slight shift in RFUC, with the percentage of less, it clearly highlights that people who use cannabis to seek
RFUC because of ‘friends’ being slightly lower (69.8%) and a relief from either physical or psychological discomfort present
higher proportion of RFUC to feel better compared with the with more psychopathology. Furthermore, while sensitivity anal-
above study. Nevertheless, given that curiosity (62.2%) and fun yses indicated moderate to high agreement (ranging from κ=0.68
(52.8%) are, respectively, the second and third most common for RFUC ‘fun’ to κ=0.99 for RFUCs for physical discomfort
RFUC, our findings confirm that people are more likely to and psychosis) between responses in the online survey and face-­
report using cannabis in a social context and for enhancement.1 to-­face assessments for all RFUC variables, recall bias cannot be
A novel feature of this study is the ability to distinguish different ruled out, as RFUC data were collected retrospectively through
categories of self-­medication motives (RFUC ‘to feel better’). self-­report (online supplemental table S2).
6 Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810
 Open access
Despite these limitations, this study has strengths. Self-­ and editing, funding acquisition. ES is the guarantor for the study. TPF and MDF
administered surveys offered greater anonymity for individuals share senior authorship.
using illicit substances.30 Furthermore, we included 10 distinct Funding The Cannabis&Me project was founded by the UKRI MRC (IRAS project

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
RFUC options, which allowed us, compared with previous ID 301405). MDF, IA-­Z, GT and ES were supported by the MRC SRF Fellowship
(MRC MR/T007818/1). ES was also supported by Lord Leverhulme’s Charitable Trust
studies, to capture a broad range of RFUC, including self-­ and the Velvet Foundation. LA thanked the Adrian and Simone Frutiger Fellowship
medication motives for both psychological and physical discom- and the Carigest SA Foundation for their support. TPF is funded by a UKRI Future
fort (namely, the ‘to feel better’ categories). Leaders Fellowship (MR/Y017560/1). The funders were not involved in the design
Additionally, a face-­to-­face assessment was also used for blood and conduct of the study; collection, management, analysis and interpretation of the
for THC measurement and validated our measurement of weekly data; preparation, review or approval of the manuscript; and decision to submit the
manuscript for publication.
THC consumption. This is an important strength of our study,
which helped us to use a comprehensive and direct standardised Competing interests MDF reports personal fees from Janssen outside the
submitted work. RM reports personal fees from Janssen, Lundbeck, Sunovion and
measure of THC consumption. Moreover, the standard weekly Otsuka outside the submitted work. The other authors declare no conflict of interest.
THC units showed excellent validity when correlated with THC
Patient consent for publication Not applicable.
blood concentrations (r=0.77, p=3.08e−13) (online supple-
mental figure S7). THC units were also associated with RFUC. Ethics approval The study was approved by the Research Ethics Committee
at King’s College London (IRAS project ID 301405) and adhered to General Data
Those who reported first using cannabis to feel relief from

Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
Protection Regulation (GDPR) g​ uidelines.​rec@​kcl.​ac.​uk. Participants gave informed
psychotic symptoms had the highest THC consumption (mean: consent to participate in the study before taking part.
307.4 (SD: 353.6) THC units), while the group with the lowest Provenance and peer review Not commissioned; externally peer reviewed.
THC weekly consumption had started for ‘curiosity’ (mean:
Data availability statement Data are available upon reasonable request. Data
193.6 (SD: 259.7) THC units) (online supplemental table S5). are available from the corresponding author upon reasonable request under the
Another strength of this study is that it was conducted in a condition of approval of the Cannabis&Me steering committee.
general population sample of non-­help-­seeking cannabis-­using Supplemental material This content has been supplied by the author(s). It
individuals living in an area with high levels of cannabis use and has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have
an elevated incidence of psychosis.14 19 This design helps to iden- been peer-­reviewed. Any opinions or recommendations discussed are solely those
tify people for targeted interventions and offers valuable clinical of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and
responsibility arising from any reliance placed on the content. Where the content
insights. includes any translated material, BMJ does not warrant the accuracy and reliability
of the translations (including but not limited to local regulations, clinical guidelines,
terminology, drug names and drug dosages), and is not responsible for any error
Clinical implications and/or omissions arising from translation and adaptation or otherwise.
Our findings, for the first time, provide evidence that people Open access This is an open access article distributed in accordance with the
who start using cannabis to reduce psychological or physical Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits
discomfort differ from those who start for reasons related to others to copy, redistribute, remix, transform and build upon this work for any
socialising, curiosity and having fun. The former report higher purpose, provided the original work is properly cited, a link to the licence is given,
and indication of whether changes were made. See: https://creativecommons.org/​
average weekly THC unit consumption (and blood levels of licenses/by/4.0/.
THC) and higher levels of paranoia, anxiety and depression.
Asking people why they started using cannabis could become ORCID iDs
an easy and cheap tool, in clinical and non-­clinical settings to Edoardo Spinazzola http://orcid.org/0000-0002-8070-0618
Edward Chesney http://orcid.org/0000-0003-2851-5252
identify users, including those who are prescribed cannabis, who
might benefit from monitoring, support or referral to interven-
tion services to prevent a possible transition from moderately REFERENCES
1 Kolliakou A, Joseph C, Ismail K, et al. Why do patients with psychosis use cannabis
severe psychopathology to potentially disabling anxiety, depres- and are they ready to change their use? Int J Dev Neurosci 2011;29:335–46.
sion and paranoia. 2 Leung J, Chan G, Stjepanović D, et al. Prevalence and self-­reported reasons of
cannabis use for medical purposes in USA and Canada. Psychopharmacology (Berl)
Author affiliations 2022;239:1509–19.
1 3 Spinazzola E, Quattrone D, Rodriguez V, et al. The association between reasons for
Psychosis Studies, King’s College London, London, UK
2 first using cannabis, later pattern of use, and risk of first-­episode psychosis: the EU-­
South London and Maudsley NHS Foundation Trust, London, UK
3 GEI case-­control study. Psychol Med 2023;53:7418–27.
Psychosis Studies, King’s College London Institute of Psychiatry, Psychology and
Neuroscience, London, UK 4 Dekker N, Linszen DH, De Haan L. Reasons for cannabis use and effects of
4 cannabis use as reported by patients with psychotic disorders. Psychopathology
Social, Genetic, and Developmental Psychiatry Centre, King’s College London
Institute of Psychiatry Psychology and Neuroscience, London, UK 2009;42:350–60.
5 5 Marconi A, Di Forti M, Lewis CM, et al. Meta-­analysis of the Association Between the
King’s College London Department of Addictions, London, UK
6 Level of Cannabis Use and Risk of Psychosis. Schizophr Bull 2016;42:1262–9.
South London and Maudsley Mental Health NHS Trust, London, UK
7 6 Xue S, Husain MI, Zhao H, et al. Cannabis Use and Prospective Long-­Term Association
Department of Psychiatry, The University of Hong Kong Li Ka Shing Faculty of
Medicine, Hong Kong, Hong Kong with Anxiety: A Systematic Review and Meta-­Analysis of Longitudinal Studies:
8 Usage du cannabis et association prospective à long terme avec l’anxiété: une
University of Porto Institute of Biomedical Sciences Abel Salazar, Porto, Portugal
9 revue systématique et une méta-­analyse d’études longitudinales. Can J Psychiatry
King’s College London Department of Forensic and Analytical Science, London, UK
10 2021;66:126–38.
Psychology, University of Bath, Bath, UK
7 Gobbi G, Atkin T, Zytynski T, et al. Association of Cannabis Use in Adolescence and
Contributors ES: conceptualisation, data collection, dataset cleaning, study design, Risk of Depression, Anxiety, and Suicidality in Young Adulthood: A Systematic Review
data analysis, manuscript writing. HD: data collection, dataset cleaning, study design, and Meta-­analysis. JAMA Psychiatry 2019;76:426–34.
data analysis, manuscript writing. IA-­Z: dataset cleaning, data analysis, manuscript 8 Henquet C, Di Forti M, Morrison P, et al. Gene-­environment interplay between
reviewing and editing. GT: data collection, dataset cleaning, manuscript reviewing cannabis and psychosis. Schizophr Bull 2008;34:1111–21.
and editing. EC: manuscript reviewing and editing. ZL: data collection, manuscript 9 D’Souza DC, DiForti M, Ganesh S, et al. Consensus paper of the WFSBP task force on
reviewing and editing. LA: manuscript reviewing and editing. BML: manuscript cannabis, cannabinoids and psychosis. World J Biol Psychiatry 2022;23:719–42.
reviewing and editing. YL: data collection, manuscript reviewing and editing. AQ: 10 D’Souza DC, Perry E, MacDougall L, et al. The psychotomimetic effects of intravenous
manuscript reviewing and editing. DQ: manuscript reviewing and editing. EC: data delta-­9-­tetrahydrocannabinol in healthy individuals: implications for psychosis.
collection, manuscript reviewing and editing. KW: manuscript reviewing and editing. Neuropsychopharmacology 2004;29:1558–72.
RM: study design; manuscript reviewing and editing. TPF: study design, manuscript 11 Freeman D, Morrison PD, Murray RM, et al. Persecutory ideation and a history of
reviewing and editing. MDF: conceptualisation, study design, manuscript reviewing cannabis use. Schizophr Res 2013;148:122–5.

Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810 7

 Open access
12 Freeman D, Garety PA, Bebbington PE, et al. Psychological investigation of the 22 Lees R, Lawn W, Petrilli K, et al. Persistent increased severity of cannabis use disorder
structure of paranoia in a non-­clinical population. Br J Psychiatry 2005;186:427–35. symptoms in adolescents compared to adults: a one-­year longitudinal study. Eur Arch
13 Green CEL, Freeman D, Kuipers E, et al. Measuring ideas of persecution and Psychiatry Clin Neurosci 2025;275:397–406.

BMJ Mental Health: first published as 10.1136/bmjment-2025-301810 on 26 August 2025. Downloaded from https://mentalhealth.bmj.com on 5 September 2025 by guest.
social reference: the Green et al. Paranoid Thought Scales (GPTS). Psychol Med 23 White IR, Royston P, Wood AM. Multiple imputation using chained equations: Issues
2008;38:101–11. and guidance for practice. Stat Med 2011;30:377–99.
14 Di Forti M, Quattrone D, Freeman TP, et al. The contribution of cannabis use to 24 van Buuren S, Groothuis-­Oudshoorn K. mice: Multivariate imputation by chained
variation in the incidence of psychotic disorder across Europe (EU-­GEI): a multicentre equations in R. J Stat Softw 2011;45:1–67.
case-­control study. Lancet Psychiatry 2019;6:427–36. 25 Kolliakou A, Castle D, Sallis H, et al. Reasons for cannabis use in first-­episode
15 Freeman TP, Lorenzetti V. Moving forwards with the standard THC unit. Addiction psychosis: does strength of endorsement change over 12 months? Eur Psychiatry
2020;115:1222–3. 2015;30:152–9.
16 Mallett R, Leff J, Bhugra D, et al. Social environment, ethnicity and schizophrenia. A 26 Wallis D, Coatsworth JD, Mennis J, et al. Predicting Self-­Medication with Cannabis
case-­control study. Soc Psychiatry Psychiatr Epidemiol 2002;37:329–35. in Young Adults with Hazardous Cannabis Use. Int J Environ Res Public Health
17 Spitzer RL, Kroenke K, Williams JBW, et al. A brief measure for assessing generalized 2022;19:1850.
anxiety disorder: the GAD-­7. Arch Intern Med 2006;166:1092–7. 27 Hasin DS, Shmulewitz D, Cerdá M, et al. U.S. Adults With Pain, a Group Increasingly
18 Kroenke K, Spitzer RL, Williams JB. The PHQ-­9: validity of a brief depression severity Vulnerable to Nonmedical Cannabis Use and Cannabis Use Disorder: 2001-­2002 and
measure. J Gen Intern Med 2001;16:606–13. 2012-­2013. Am J Psychiatry 2020;177:611–8.
19 Di Forti M, Morgan C, Dazzan P, et al. High-­potency cannabis and the risk of 28 Office for National Statistics. Ethnic distribution in greater london: census 2021 data
psychosis. Br J Psychiatry 2009;195:488–91. from Nomis. 2023. Available: https://www.nomisweb.co.uk
20 Bharat C, Webb P, Wilkinson Z, et al. Agreement between self-­reported illicit drug 29 Di Forti M, Marconi A, Carra E, et al. Proportion of patients in south London with first-­

Protected by copyright, including for uses related to text and data mining, AI training, and similar technologies.
use and biological samples: a systematic review and meta-­analysis. Addiction episode psychosis attributable to use of high potency cannabis: a case-­control study.
2023;118:1624–48. Lancet Psychiatry 2015;2:233–8.
21 Buchan BJ, L. Dennis M, Tims FM, et al. Cannabis use: consistency and validity of self‐ 30 Dodou D, de Winter JCF. Social desirability is the same in offline, online, and paper
report, on‐site urine testing and laboratory testing. Addiction 2002;97:98–108. surveys: A meta-­analysis. Comput Human Behav 2014;36:487–95.

8 Spinazzola E, et al. BMJ Ment Health 2025;28:1–8. doi:10.1136/bmjment-2025-301810