882806

research-article2019
JOP0010.1177/0269881119882806Journal of PsychopharmacologyLake et al.

Original Paper

Does cannabis use modify the effect of

post-traumatic stress disorder on severe
depression and suicidal ideation? Evidence Journal of Psychopharmacology

from a population-based cross-sectional

1­–8
© The Author(s) 2019
Article reuse guidelines:
study of Canadians sagepub.com/journals-permissions
DOI: 10.1177/0269881119882806
https://doi.org/10.1177/0269881119882806
journals.sagepub.com/home/jop

Stephanie Lake1,2, Thomas Kerr1,3, Jane Buxton2,4, Zach Walsh5,

Brandon Marshall6, Evan Wood1,3 and M-J Milloy1,3

Abstract
Background: Post-traumatic stress disorder sharply increases the risk of depression and suicide. Individuals living with post-traumatic stress disorder
frequently use cannabis to treat associated symptoms. We sought to investigate whether cannabis use modifies the association between post-traumatic
stress disorder and experiencing a major depressive episode or suicidal ideation.
Methods: We used data from the 2012 Canadian Community Health Survey-Mental Health, a nationally representative cross-sectional survey of non-
institutionalized Canadians aged ⩾15 years. The relationship between post-traumatic stress disorder and each outcome was modelled using logistic
regression with an interaction term for cannabis and post-traumatic stress disorder, controlling for demographic characteristics, mental health,
and substance use comorbidities. The ratio of odds ratios and relative excess risk due to interaction was calculated to measure interaction on the
multiplicative and additive scales, respectively.
Results: Among 24,089 eligible respondents, 420 (1.7%) reported a current clinical diagnosis of post-traumatic stress disorder. In total, 106 (28.2%)
people with post-traumatic stress disorder reported past-year cannabis use, compared to 11.2% of those without post-traumatic stress disorder
(p < 0.001). In multivariable analyses, post-traumatic stress disorder was significantly associated with recent major depressive episode (adjusted odds
ratio = 7.18, 95% confidence interval: 4.32–11.91) and suicidal ideation (adjusted odds ratio = 4.76, 95% confidence interval: 2.39–9.47) among
cannabis non-users. post-traumatic stress disorder was not associated with either outcome among cannabis-using respondents (both p > 0.05).
Conclusions: This study provides preliminary epidemiological evidence that cannabis use may contribute to reducing the association between post-
traumatic stress disorder and severe depressive and suicidal states. There is an emerging need for high-quality experimental investigation of the
efficacy of cannabis/cannabinoids for the treatment of post-traumatic stress disorder.

Keywords
Cannabis, cannabinoids, post-traumatic stress disorder, major depressive episode, suicide, suicidal ideation, cross-sectional study, epidemiological study

Introduction mental health (including depression and suicidality), and high-

Canada is estimated to have one of the highest prevalence estimates
of lifetime post-traumatic stress disorder (PTSD) worldwide, affect-
ing some 9.2% of the population (Duckers et al., 2016). PTSD is a
psychiatric diagnosis associated with surviving exposure to a trau-
matic life event (Shalev et al., 2017), and involves symptoms of
hyper-arousal and flashbacks, as well as impaired memory, concen-
tration, and sleep (Bremner, 2006). These symptoms are thought to
manifest through stress-induced changes to the brain, resulting in
risk substance use (Gentes et al., 2016). Individuals with severe
or treatment-resistant PTSD symptomologies may use cannabis
extra-medically or therapeutically as a primary or adjunctive
1British Columbia Centre on Substance Use, Vancouver, BC, Canada
2School of Population and Public Health, University of British
Columbia, Vancouver, BC, Canada
3Department of Medicine, University of British Columbia, St. Paul’s

disruptions to the body’s stress response (Bremner, 2006). It has Hospital, Vancouver, BC, Canada
been well established that PTSD, especially when coupled with 4British Columbia Centre for Disease Control, Vancouver, BC, Canada

comorbid depression, greatly increases the risk of suicidality (i.e. 5Department of Psychology, University of British Columbia, Kelowna,

suicidal ideation (SI), plans, and attempts (Ramsawh et al., 2014))
and completed suicide (Gradus et al., 2010). Treating PTSD is chal-
lenging, particularly due to a substantial heterogeneity in treatment
responses between patients (Shalev et al., 2017).
Cannabis use and cannabis use disorder (CUD) are common
among trauma survivors (Kevorkian et al., 2015). Previous
research among individuals with PTSD suggests that cannabis
use is associated with increased PTSD symptom severity, poorer
BC, Canada
6Department of Epidemiology, Brown University School of Public
Health, Providence, RI, USA
Corresponding author:
M-J Milloy, Department of Medicine, University of British Columbia,
Research Scientist, BC Centre on Substance Use, 400–1045 Howe
Street, Vancouver, British Columbia V6Z 2A9, Canada.
Email: bccsu-mjm@bccsu.ubc.ca
2 Journal of Psychopharmacology 00(0)
therapy (Greer et al., 2014; Ruglass et al., 2017). However, it has
been posited that this type of coping-oriented motive may reflect
an unhealthy pattern of using cannabis to avoid confronting pain-
ful trauma-related memories and experiences (i.e. “avoidance
coping”), which could exacerbate PTSD symptoms in the long
term (Bonn-Miller et al., 2007). Much of the existing evidence in
support of cannabis as a PTSD treatment is anecdotal and posits
a range of therapeutic benefits including improving sleep, main-
taining cognitive control, reducing symptoms of hyper-arousal,
preventing intrusive thoughts, and managing emotional distress
(Passie et al., 2012; Steenkamp et al., 2017)
Discoveries involving the body’s natural endogenous can-
nabinoid (eCB) system (see Discussion) and its role in regulating
a range of PTSD-compromised neurobiological processes (e.g.
fear, memory, stress, and sleep (Passie et al., 2012)), have sparked
interest in eCB modulators (i.e. natural and synthetic cannabi-
noids) as promising PTSD treatment agents (Shalev et al., 2017).
In the wake of these developments, the therapeutic use of can-
nabis for PTSD has gained a level of legitimacy across North
America. For example, cannabis is an approved medication for
PTSD in some states in the United States (US) (Steenkamp et al.,
2017), and Canadian physicians can authorize access to medical
cannabis for the treatment of PTSD under the Access to Cannabis
for Medical Purposes Regulations. In recent years, the number of
Canadians, including veterans, incorporating cannabis into their
PTSD treatment plan has grown dramatically (Veterans Affairs
Canada, 2019). This increase in medical cannabis use for PTSD
has overlapped with growing public awareness of the heightened
risk of suicide among trauma-exposed populations (D’Aliesio,
2017). However, the current state of scientific evidence is insuf-
ficient to conclude whether cannabis is beneficial or harmful to
patients with PTSD (O’Neil et al., 2017).
As Canada moves into the era of legalized cannabis, there are
increasing scientific calls to explore the range of possible clini-
cal applications involving cannabinoids, including the treatment
of PTSD (Serrano, 2018). Canada’s first clinical trial to study
the safety and efficacy of cannabis in managing treatment-­
resistant (i.e. approved medication or empirically supported psy-
chotherapy) PTSD is underway (clinicaltrials.gov, 2015) and
there are several ongoing clinical trials of cannabis for PTSD in
the US (O’Neil et al., 2017). In the interim, there is a valuable
opportunity for epidemiological research to examine the com-
plex relationships between PTSD, cannabis use, and mental
health outcomes including suicidality.
This study draws on epidemiological data from a representa-
tive sample of Canadians to explore whether cannabis use modi-
fies the relationship between PTSD and severe depressive and
suicidal states. If cannabis offers a therapeutic benefit for PTSD,
we would expect the relationship between PTSD and these mark-
ers of severe mental distress to be dependent on cannabis such
that the association would be amplified in its absence and attenu-
ated in its presence.
Methods
Study design
Data were obtained from Statistics Canada’s 2012 Canadian
Community Health Survey – Mental Health (CCHS-MH).
Detailed sampling and interviewing methods for this
cross-sectional study are available from Statistics Canada
(Statistics Canada, 2013). Briefly, the survey covers Canadians
aged 15 or over living the 10 provinces and excludes approxi-
mately 3% of this target population who are: full-time members
of the Canadian Armed Forces; individuals living on reserves or
other First Nations, Inuit, or Métis settlements; and individuals
held in correctional facilities. Multi-stage cluster sampling is
used to obtain a geographically representative sample of the
target population. In total, 25,113 participants completed the
computer-assisted interview over the phone or in person, generat-
ing a combined household and person-level response rate of
68.9% (Statistics Canada, 2013). Ethics approval for this study
was covered by the University of British Columbia’s publicly
available data clause (Policy 89, Item 7.10.3) governing the use
of public release data (2012). All data were kept private and con-
fidential as part of the Statistics Act.
Measures
All respondents who provided valid responses to questions cor-
responding to variables of interest in the present study were eligi-
ble for analysis. The two outcomes of interest were: (a) past-year
major depressive episode (MDE), defined as meeting the criteria
for MDE, as assessed through a modified version of the World
Health Organization Composite International Diagnostic
Interview (WHO-CIDI), and (b) past-year SI, defined as answer-
ing “yes” to the question “In the past 12 months, did you seri-
ously think about committing suicide or taking your own life?”
Both outcomes were binary (yes vs. no). The exposure of interest
was self-reported current PTSD diagnosis, defined as a medical
diagnosis of PTSD that is expected to last, or has already lasted,
6 months or longer (yes vs. no). Cannabis use was defined as hav-
ing used cannabis more than once in the previous year (yes vs.
no). The question did not differentiate between medical and non-
medical use.
Additional socio-demographic variables considered in the
analysis due to their known or a-priori hypothesized relationship
with PTSD and MDE/SI were: sex (male vs. female); age (15–
24 years, 25–44 years, 45–64 years vs. ⩾ 65 years); marital status
(partner vs. single); visible minority (yes vs. no); education (<
high school, high school vs. post-secondary); and annual house-
hold income (<$60,000 vs. ⩾ $60,000). The following past-year
substance use and mental health comorbidities were also consid-
ered in the analyses: non-cannabis illicit drug use (yes vs. no);
alcohol use disorder; and non-cannabis substance use disorder
(yes vs. no). Both substance use disorder variables were defined
as meeting criteria for abuse or dependence through a modified
version of the WHO-CIDI. We also included measures of self-
reported diagnoses of depression (yes vs. no); other mood disor-
der (yes vs. no); and anxiety disorder (including generalized
anxiety disorder, panic disorder, and obsessive compulsive disor-
der; yes vs. no). Similar to the variable for PTSD, these mental
health measures are clinical diagnoses of conditions that have
lasted or are expected to last 6 months or longer.
Statistical analysis
Sample characteristics were examined using standard descriptive
statistics. We used Pearson’s Chi-square test to analyze bivaria-
ble associations between each independent variable and outcome.
Lake et al. 3
We regressed MDE and SI on PTSD, separately, to calculate the
unadjusted odds ratio (OR) for the association between PTSD
and MDE/SI. Then, to explore whether this association was mod-
ified by cannabis use, we added cannabis and a cannabis-PTSD
interaction term to each model. Wald tests were used to check the
significance of the multiplicative interaction. Next, we conducted
multivariable analyses to estimate the adjusted OR (AOR) for
each outcome across strata of cannabis use. We followed an
a-priori model building strategy in which all listed variables
(except depression for the MDE model) were included as hypoth-
esized confounders in the association between PTSD and MDE/
SI. AORs within each stratum of PTSD and cannabis use were
also calculated by replacing the PTSD, cannabis, and interaction
terms in each multivariable model with a four-category com-
bined PTSD/cannabis variable (i.e. PTSD-/Cannabis-; PTSD+/
Cannabis-; PTSD-/Cannabis+; PTSD+/Cannabis+). The cate-
gory hypothesized to have the lowest probability of SI and MDE
(i.e. PTSD-/Cannabis-) provided the reference.
As per Knol and VanderWeele’s effect modification reporting
guidelines (2012), statistical interaction on the additive scale was
tested by calculating the relative excess risk due to interaction
(RERI) and its 95% confidence interval (CI) using the Delta
method described by Hosmer and Lemeshow (1992). The RERI
can be estimated by approximating risk ratios from ORs in cases
where the outcome is rare. A RERI of 0 means no interaction; a
RERI bigger or smaller than 0 means positive or negative interac-
tion on the additive scale, respectively (Knol et al., 2011).
Measuring interaction on the additive scale is generally consid-
ered more relevant in assessing the public health impact, as it indi-
cates which exposure group would benefit the most from
intervening on the effect modifier, or vice-versa (VanderWeele
and Knol, 2014). The ratio of odds ratios within the strata of can-
nabis use status is also reported as the measure of multiplicative
interaction for each outcome. A ratio of 1 means no interaction,
whereas a ratio bigger or smaller than 1 indicates positive or nega-
tive interaction on the multiplicative scale, respectively.
A sub-analysis was conducted among respondents with PTSD
to estimate the adjusted odds of SI and MDE according to canna-
bis use intensity. We built two multivariable models (one for each
outcome) with a three-level past-year cannabis intensity variable
(low-risk, high-risk, vs. ⩽ one-time use) as the main exposure.
“High-risk” use was defined as meeting the criteria for past-year
CUD, as assessed through a modified version of the WHO-CIDI.
“Low-risk” use was defined as not meeting the criteria for CUD
either through the WHO-CIDI assessment or through not meeting
the CCHS-MH/WHO-CIDI screening threshold (lifetime canna-
bis use ⩾ 50 times). This threshold was based on an expert review
of 2011 CCHS-MH pilot data analyses.
All analyses were conducted in Stata version 14.2 (StataCorp
LLC, College Station, TX) using standard univariable, bivaria-
ble, and multivariable procedures with ‘svy’ commands and
probability weights to account for complex survey sampling
methods. All p values are two sided.
Results
Of the 25,113 Canadians who were interviewed for the 2012
CCHS-MH, 24,089 (95.9%) provided complete data for this
study and were included in the current analysis. The prevalence
of current PTSD diagnosis was 1.7% (n = 420), and cannabis use
more than once in the previous year was reported by 11.4%
(n = 2767) of the overall sample; however, individuals living with
PTSD had a significantly higher prevalence of past-year cannabis
use relative to the rest of the sample (28.2% vs. 11.2%, p < 0.001).
In total, 1261 (4.7%) respondents met the criteria for experienc-
ing a past-year MDE and 875 (3.3%) reported past-year SI.
Across strata of PTSD and cannabis use, the prevalence of each
outcome was lowest among individuals who did not use cannabis
and did not have PTSD.
Table 1 summarizes sample characteristics overall, stratified by
each outcome, and presents bivariable associations. PTSD was
strongly and positively associated with MDE and SI. Cannabis use
in the previous year was also significantly and positively associ-
ated with each outcome. The unadjusted interaction models gener-
ated significant interaction terms (MDE: p = 0.003; SI: p = 0.021).
The cannabis-stratified ORs were split above and below the unad-
justed pooled ORs. Among the sample that did not use cannabis,
having PTSD was associated with 21.06 (95% CI: 13.41–33.07)
times the odds of MDE and 19.91 (95% CI: 11.76–33.72) times the
odds of SI, relative to individuals with PTSD (both p < 0.001). In
contrast, among the sample that reported cannabis use, PTSD was
associated with 6.33 (95% CI: 3.29–12.17) times the odds of MDE
and 7.1 (95% CI: 3.5–14.2) times the odds of SI, relative to indi-
viduals without PTSD (both p < 0.001). These unadjusted results
can be extrapolated from data provided in Tables 2 and 3.
Tables 2 and 3 present the results of the multivariable regres-
sion within combined stratum of PTSD and cannabis use (left)
and within strata of cannabis use (right) for MDE (Table 2) and
SI (Table 3). After controlling for socio-demographic factors, co-
occurring substance use patterns, and comorbid mental health
problems, PTSD remained positively and significantly associated
with MDE (AOR = 7.18; 95% CI = 4.32–11.91) and SI (AOR =
4.76; 95% CI = 2.39–9.47) among cannabis non-users (both
p < 0.001). However, there was no longer a significant associa-
tion between PTSD and MDE (AOR = 1.77; 95% CI = 0.79–
3.97) or PTSD and SI (AOR = 1.66; 95% CI = 0.77–3.56)
among respondents who reported cannabis use more than once in
the previous year (both p > 0.05).
On the multiplicative scale, there was a significant negative
interaction between PTSD and cannabis use, with both ratio of
ORs < 1 (p < 0.05). Tests for additive interaction for MDE pro-
vided strong indication that the estimated effect of PTSD on
MDE among cannabis users was less than the effect of PTSD on
MDE for non-users (RERI = -4.93, 95% CI: -8.49–-1.37;
p = 0.007). The RERI for the PTSD and cannabis on SI was nega-
tive, but fell short of statistical significance (p = 0.125).
In sub-analyses among 420 respondents with PTSD (Table 4),
cannabis use in the previous year was not significantly associated
with MDE or SI (both p > 0.05). Separation of cannabis use into
“low risk” (n = 86; 23.6%) and “high risk” (n = 20; 4.5%) revealed
a negative association between low-risk use and SI shy of signifi-
cance (p = 0.068) and a significant negative association between
low-risk cannabis use and MDE (p = 0.038). High-risk cannabis
use was not associated with either outcome (both p > 0.05).
Discussion
This is the first study, to our knowledge, to document the relation-
ships between PTSD, cannabis use, and severe mental health out-
comes in a representative population sample. Over one-quarter of
4 Journal of Psychopharmacology 00(0)

Table 1.  Bivariable associations with major depressive episode and suicidal ideation in the previous month among 24,089 Canadians aged 15 and
older.

Characteristic 24,089 Major depressive episode Suicidal ideation

total n (%) (100%)
n (%) = yes n (%) = no Odds ratio n (%) = yes n (%) = no Odds ratio
(95% CI) (95% CI)
1,261 22,828 875 23,214
(4.7%) (95.3%) (3.3%) (96.7%)

Demographic characteristics
Sex
Male 10,869 (49.1) 449 (37.6) 10,420 (49.7) 0.61 (0.60–0.74)** 387 (46.4) 10,482 (49.2) 0.89 (0.71–1.12)
Female 13,220 (50.9) 812 (62.4) 12,408 (50.3) 1.00 488 (53.6) 12,732 (50.8) 1.00
Age group
15–24 3884 (15.9) 267 (23.8) 3617 (15.6) 4.48 (3.08–6.52)** 232 (27.7) 3652 (15.5) 4.45 (3.11–6.36)**
25–44 6689 (33.0) 416 (37.7) 6273 (32.7) 3.38 (2.35–4.87)** 282 (38.1) 6407 (32.8) 2.90 (2.00–4.21)**
45–64 7731 (34.2) 464 (32.5) 7267 (34.3) 2.78 (1.95–3.98)** 273 (27.3) 7458 (34.4) 1.98 (1.38–2.84)**
⩾ 65 5785 (16.9) 114 (16.9) 5671 (17.5) 1.00 88 (6.9) 5697 (17.3) 1.00
Marital status
Partner 12,122 (60.1) 427 (42.4) 11,695 (61.0) 0.47 (0.39–0.57)** 274 (40.0) 11,848 (60.8) 0.43 (0.34–0.55)**
Single 11,967 (39.9) 834 (57.6) 11,133 (39.0) 1.00 601 (60.0) 11,366 (39.2) 1.00
Visible minority
Yes 20,095 (76.9) 1051 (78.0) 19,044 (76.9) 1.07 (0.84–1.35) 701 (74.4) 19,394 (77.0) 0.87 (0.66–1.14)
No 3994 (23.1) 210 (22.0) 3784 (23.1) 1.00 174 (25.6) 3820 (23.0) 1.00
Education
< Secondary 5076 (17.9) 265 (18.4) 4811 (17.9) 1.05 (0.83–1.32) 4856 (17.7) 220 (24.5) 1.52 (1.16–2.00)*
Secondary 3877 (15.8) 196 (16.4) 3681 (15.8) 1.05 (0.83–1.35) 3740 (15.8) 137 (15.2) 1.06 (0.79–1.42)
Post-Secondary 15,136 (66.3) 800 (65.3) 14,336 (66.4) 1.00 14,618 (66.5) 518 (60.4) 1.00
Household income
< $60,000 10,819 (34.4) 725 (48.3) 10,094 (33.8) 1.84 (1.52–2.21)** 487 (48.0) 10,332 (34.0) 1.80 (1.43–2.25)**
⩾ $60,000 13,270 (65.6) 536 (51.7) 12,734 (66.3) 1.00 388 (52.0) 12,882 (66.0) 1.00
Substance use
Cannabis use
Yes 2797 (11.4) 332 (27.1) 2465 (10.7) 3.12 (2.52–3.86)** 262 (28.3) 2535 (10.9) 3.24 (2.54–4.14)**
No 21,292 (88.6) 929 (72.9) 20,363 (89.3) 1.00 613 (71.7) 20,679 (89.1) 1.00
Other illicit drug use
Yes 1689 (6.3) 332 (26.0) 1357 (5.4) 6.17 (4.94–7.70)** 224 (22.8) 1465 (5.8) 4.81 (3.72–6.23)**
No 22,400 (93.7) 929 (74.0) 21,471 (94.6) 1.00 651 (77.2) 21,749 (94.2) 1.00
Alcohol use disorder
Yes 771 (3.2) 121 (10.6) 650 (2.8) 4.06 (2.98–5.54)** 112 (12.9) 659 (2.9) 5.02 (3.64–6.92)**
No 23,318 (96.8) 1140 (89.4) 21,178 (97.2) 1.00 763 (87.1) 22,555 (97.1) 1.00
Other substance use disorder
Yes 182 (0.7) 70 (6.2) 112 (0.4) 16.11 (9.98–26.01)** 51 (6.9) 131 (0.5) 15.52 (9.17–26.26)**
No 23,907 (99.3) 1191 (93.8) 22,716 (99.6) 1.00 824 (93.1) 23,083 (99.5) 1.00
Mental health
PTSD
Yes 420 (1.7) 164 (14.8) 256 (1.0) 16.92 (11.65–24.58)** 113 (16.4) 307 (1.2) 16.75 (10.97–25.56)**
No 23,669 (98.3) 1097 (85.2) 22,572 (99.0) 1.00 762 (83.6) 22,907 (98.8) 1.00
Depression
Yes 1,622 (5.8) 579 (44.5) 1,043 (3.9) 19.59 (15.87–24.18)** 338 (34.3) 1,284 (4.9) 10.20 (8.03–12.94)**
No 22,467 (94.2) 682 (55.5) 21,785 (96.1) 1.00 537 (65.7) 21,930 (95.1) 1.00
Other mood disorder
Yes 346 (1.3) 141 (12.0) 205 (0.8) 17.88 (11.87–26.92)** 87 (11.8) 259 (0.9) 14.28 (8.54–23.89)**
No 23,743 (98.7) 1120 (88.0) 22,623 (99.2) 1.00 788 (88.2) 22,955 (99.1) 1.00
Anxiety disorder
Yes 1440 (4.8) 402 (27.3) 1038 (3.7) 9.90 (7.97–12.29)** 257 (27.5) 1183 (4.0) 9.15 (7.09–11.81)**
No 22,649 (95.2) 859 (72.7) 21,790 (96.4) 1.00 618 (72.5) 22,031 (96.0) 1.00

Survey weights were used to calculate weighted proportions.

CI: confidence interval; PTSD: post-traumatic stress disorder.
*p < 0.05; **p < 0.001.
Lake et al. 5

Table 2.  Modification of the effect measure of PTSD on major depressive episode by cannabis use status in a cross-sectional sample of 24,089 Canadians.

PTSD = No PTSD = Yes Adjusted odds

ratios (95% CI)
  n (%) with / Odds ratio n (%) with / Odds ratio within strata of
without MDE (95% CI) without MDE (95% CI) cannabis use

Cannabis = No 811 (3.3) / 1.00 118 (42.1) / 7.18 7.18 (4.32–11.91)

20,167 (96.7) 196 (57.9) (4.32–11.91) p < 0.001
p < 0.001
Cannabis = Yes 286 (9.9) / 1.61 46 (40.9) / 2.86 1.77 (0.79–3.97)
2405 (90.1) (1.25–2.08) 60 (59.1) (1.30–6.28) p = 0.166
p < 0.001 p = 0.009

Measure of effect modification on additive scale: RERI (95% CI) = -4.93 (-8.49, -1.37); p = 0.007.
Measure of effect modification on the multiplicative scale: ratio of ORs (95% CI) = 0.25 (0.10, 0.64); p = 0.004.
OR is adjusted for age, sex, marital status, visible minority, education, household income, other illicit drug use, mood disorder excl. depression, anxiety disorder, alcohol
use disorder, and other substance use disorder.
n is a raw cell count; % is weighted. Unadjusted OR point estimates can be calculated from weighted cell counts (multiply the weighted proportions with and without the
outcome in each cell by the total n in that cell).
CI: confidence interval; MDE: major depressive episode; OR: odds ratio; PTSD: post-traumatic stress disorder; RERI: relative excess risk due to interaction.

Table 3.  Modification of the effect measure of PTSD on suicidal ideation by cannabis use status in a cross-sectional, representative sample of
24,089 Canadians.

PTSD = No PTSD = Yes Adjusted odds

ratios (95% CI)
  n (%) with / Odds ratio n (%) with / Odds ratio within strata of
without SI (95% CI) without SI (95% CI) cannabis use

Cannabis = No 535 (2.3) / 1.00 78 (31.6) / 4.76 4.76 (2.39–9.47);

20,443 (97.7) 236 (68.4) (2.39–9.47); p < 0.001
p < 0.001
Cannabis = Yes 227 (7.0) / 1.43 35 (34.8) / 2.37 1.66 (0.77–3.56);
2464 (93.0) (1.07–1.91); 71 (65.2) (1.13–4.97); p = 0.194
p = 0.016 p = 0.023

Measure of effect modification on additive scale: RERI (95% CI) = -2.82 (-6.42, 0.78); p = 0.125.
Measure of effect modification on the multiplicative scale: ratio of ORs (95% CI) = 0.35 (0.13–0.94); p = 0.038.
OR is adjusted for age, sex, marital status, visible minority, education, household income, other illicit drug use, depression, other mood disorder, anxiety disorder, alcohol
use disorder, and other substance use disorder.
n is a raw cell count; % is weighted. Unadjusted OR point estimates can be calculated from weighted cell counts (multiply the weighted proportions with and without the
outcome in each cell by the total n in that cell).
CI: confidence interval; SI: suicidal ideation; PTSD: post-traumatic stress disorder; RERI: relative excess risk due to interaction.

Canadians with PTSD reported past-year cannabis use, which is
remarkably high compared to the prevalence of recent use in the
general Canadian population (estimated at 11.4% in the present
study), and the prevalence of recent use among US adults with
lifetime PTSD (14.0%) (Cougle et al., 2011). Through the novel
conceptual and methodological treatment of cannabis as an effect
measure modifier, this study provides evidence that cannabis use
may reduce the effect of PTSD on MDE and SI. Relative to
Canadians without either exposure, the odds of MDE and SI were
elevated among individuals with PTSD and individuals who use
cannabis. Yet, exposure to both these risk factors produced a
smaller effect on MDE (multiplicative and additive scales) and SI
(multiplicative scale) than expected based on their individual effect
estimates. A more straightforward way of interpreting these find-
ings is within strata of cannabis use: whereas PTSD was strongly
linked to MDE and SI among the cannabis non-using population,
PTSD was not associated with either outcome among Canadians
who used cannabis. Furthermore, sub-analyses involving respond-
ents with PTSD suggested that lower-risk cannabis users are even
less likely than non-users to experience MDE or SI (although the
estimate for SI was shy of reaching significance). These benefits
were not observed for individuals engaging in high-risk cannabis
use (see discussion below).
In general, PTSD can be classified according to three symp-
tom groups: re-experiencing (e.g. flashbacks, reliving the trau-
matic experience), hyper-arousal (e.g. irritability, difficulty
concentrating, lowered tolerance to stressful stimuli), and avoid-
ance/numbing behaviors (e.g. attempting to avoid aversive
situations or emotional thoughts/feelings) (Pai et al., 2017).
Cross-sectional studies have observed significant correlations
between cannabis use and higher PTSD symptom severity
(Bordieri et al., 2014; Earleywine and Bolles, 2014; Gentes et al.,
2016). Research by Bonn-Miller et al. (2007) suggests this asso-
ciation is explained by increased motivation to use cannabis for
coping purposes among those with higher symptom severity.
However, whether coping-motivated cannabis strategies effec-
tively manage symptoms or simply facilitate avoidance/numb-
ing—potentially contributing to intensified symptoms (Pineles
6 Journal of Psychopharmacology 00(0)
Table 4.  Associations between cannabis use and suicidal ideation (left) and major depressive episode (right) among 420 Canadians with PTSD.
Cannabis use Major depressive episode
AOR (95% CI)
Model Series 1
No 1.00
Yes 0.47 (0.19–1.15)
Model Series 2
None 1.00
Low-risk use 0.38 (0.15–0.95)
High-risk use 1.56 (0.25–9.76)
Major depressive episode model is adjusted for age, sex, marital status, ethnicity, education, income, other illicit drug use, mood disorder (excluding depression), anxiety
disorder, alcohol use disorder, and other substance use disorder. Suicidal ideation model is adjusted for all of the above as well as depression.
PTSD: post-traumatic stress disorder; AOR: adjusted odds ratio; CI: confidence interval.
et al., 2011)—is not entirely clear. In a recent longitudinal study
among PTSD patients initiating non-pharmacological treatment,
patients who used cannabis weekly reported higher symptom
severity at treatment initiation, but experienced a 75% reduction
in all assessed symptoms, including avoidance, over 12 weeks
(Ruglass et al., 2017), providing some evidence for the symptom-
management hypothesis. Similarly, a >75% reduction in symp-
tom severity was recorded in a retrospective analysis of 80 PTSD
patients in New Mexico’s medical cannabis program (Greer
et al., 2014). In contrast, Wilkinson et al. (2015) conducted a ret-
rospective chart review of over 2000 veterans referred for PTSD
treatment and noted that patients who started using cannabis over
the 4-month treatment period experienced an increase in treat-
ment severity, whereas patients who stopped using cannabis
experienced improvements.
The relationship between problematic cannabis use (i.e.
CUD) and PTSD symptom severity has been the focus of several
observational studies (Boden et al., 2013; Bonn-Miller et al.,
2013; Bordieri et al., 2014). One hypothesis is that CUD is impli-
cated in a feedback loop with PTSD symptomology whereby
cannabis is used for symptom management, but symptoms inten-
sify during cessation periods via withdrawal, leading to a relapse
of cannabis use for symptom management (Boden et al., 2013).
There is also some evidence that increased PTSD symptom
severity is predictive of CUD only when cannabis is used as a
strategy of avoidance (Bordieri et al., 2014). Through a sub-anal-
ysis of participants with PTSD, our study considered how PTSD
could be distinctly linked with mental health outcomes according
to higher- and lower-risk use patterns. Although larger observa-
tional and experimental studies are warranted, the results of this
sub-analysis suggest cannabis may be used as an effective form
of PTSD symptom management among lower-risk cannabis
users (>80% of all cannabis users in this sub-sample), but not
among those exhibiting higher-risk usage patterns. One possible
explanation is that high-risk use is a marker of unhealthy avoid-
ance coping in this sub-sample. It is also possible that problem-
atic patterns of cannabis use may trigger certain negative social
and psychological effects common to PTSD (e.g. paranoia
(Freeman et al., 2013)), which could contribute to intensified
PTSD symptoms and degradation of longer-term mental health
(Tsai et al., 2014). Individuals using cannabis to treat trauma-
associated symptoms should therefore be carefully monitored for
the development of CUD.
Suicidal ideation
p value AOR (95% CI) p value
1.00  
0.100 0.59 (0.25–1.41) 0.237
1.00  
0.038 0.42 (0.17–1.06) 0.068
0.634 2.50 (0.57–10.92) 0.223
A growing pre-clinical research base has identified a potential
modulating role of the eCB system in the manifestation of PTSD.
The eCB system is comprised of type-1 cannabinoid receptors
(CB-1), predominantly in the central nervous system; CB-2, pre-
dominantly in the peripheral nervous system; and eCB ligands
including N-arachidonoyl ethanolamine (anandamide) and 2-ara-
chidonoyl glycerol that bind to these receptors to regulate neuro-
transmitter release (Hillard, 2015). CB-1 is abundant in
stress-related regions of the brain including the amygdala, hip-
pocampus, and prefrontal cortex (Trezza and Campolongo, 2013;
Zer-Aviv et al., 2016). These regions are involved in emotional
and memory processes that tend to be disrupted in PTSD, such as
the detection and regulation of emotions, and storage of fear
memories (Passie et al., 2012; Trezza and Campolongo 2013; Zer-
Aviv et al., 2016). Human brain imaging studies have shown that
patients with PTSD exhibit higher levels of CB-1 and lower levels
of the eCB anandamide in these regions compared to individuals
without PTSD (Neumeister et al., 2013), providing a possible neu-
ropharmacological rationale for cannabinoid supplementation in
the treatment of PTSD. In animal models, exposure to cannabi-
noids following exposure to trauma prevents the development of
certain effects from trauma, including impaired fear extinction
and stress-induced anxiety (Zer-Aviv et al., 2016). Rigorous
experimental research is needed to understand if humans exhibit a
similar response.
Sleep is a commonly reported therapeutic motive for medical
cannabis use (Reinarman et al., 2011), including among PTSD
patients (Bonn-Miller et al., 2014). A small number of pilot clini-
cal trials have been conducted to determine the efficacy, safety,
and tolerance of cannabinoids for relief of PTSD-associated
insomnia and nightmares. Two studies noted cessation or reduced
severity of nightmares after treatment with nabilone (a synthetic
analogue of tetrahydrocannabinol (THC)) (Fraser, 2009; Jetly
et al., 2015). Roitman et al. (2014) observed significant improve-
ments in PTSD symptoms and sleep quality with oral THC.
Cannabinoids were relatively safe and well tolerated, with a
small number of mild side effects (e.g. dizziness, short-term
memory impairments) (Fraser, 2009). However, the plausibility
of sleep management as a pathway underlying the current find-
ings is challenged by other Canadian research demonstrating that
the relative impact of insomnia on SI is reduced in the presence
of comorbid mental health problems including PTSD (Richardson
et al., 2017). Future experimental studies will need to address the
Lake et al. 7
role of sleep in the relationship between cannabinoid therapy and
PTSD outcomes.
The major strength of this study is its external validity through
the use of a representative sample of Canadians from the 10 prov-
inces. However, a number of limitations are also worth consider-
ing. First, the cross-sectional study design prevents the ability to
confirm the sequence of exposures for PTSD, cannabis use, and
MDE/SI, thus preventing causal interpretations. In particular,
cannabis use in this study was measured over the previous year,
whereas PTSD referred to a current diagnosis (i.e. has lasted or is
expected to last 6 months or longer). It is possible that some
PTSD patients engaged in cannabis use ahead of their diagnosis
but not after. Second, as this was a secondary analysis of a
Statistics Canada survey, we had no role in questionnaire devel-
opment and were limited to the variables available in the dataset.
As such, it was not possible to ascertain information on a number
of details. For example, the study did not provide information on
licensed medical cannabis patient status, or whether cannabis use
was considered medicinal or recreational by the participant. The
data collection period preceded the creation of a commercially
licensed industry for medical cannabis, meaning any licensed
medical use at the time of study would have occurred through
the more restrictive 2001–2013 “Marihuana [sic]” for Medical
Access Regulations system. We also did not have access to infor-
mation on dose, potency, strain, or mode of cannabis administra-
tion. Although this is a common limitation of observational
research involving cannabis, awareness is growing amongst
clinical and public health researchers that these details are critical
to understanding the health and social risks and benefits of can-
nabis use (Solowij et al., 2016), particularly against the evolving
cannabis policy backdrop. Because the study was limited to vari-
ables available through the CCHS-MH, it may be susceptible to
residual confounding. Third, although only 3% of the target pop-
ulation was excluded from the survey sample pool, certain
excluded groups are likely to have a high prevalence of PTSD
(e.g. full-time members of the Canadian armed forces, indige-
nous populations living on settlements or reserves). The rele-
vance and generalizability of this work would be optimized
through future studies using a similar methodological approach
with data from these sub-populations in the era of Canada’s com-
mercial medical cannabis system. Fourth, many outcomes were
self-reported and subject to recall inaccuracies and reporting
according to perceived social norms. This could bias the study
outcomes if there were systematic differences in the accuracy of
reporting according to exposure status. However, this is unlikely
to be the case in the present study, because these responses were
extracted from a larger survey assessing a broad set of health-
related factors among Canadians, and participants were not
aware of specific research questions to be explored with this data.
Finally, given a low prevalence of PTSD in the study population,
the trends revealed through the current sub-analysis will need to
be assessed in future studies with larger representative samples of
individuals with PTSD.
In sum, this study provides compelling preliminary evidence
from a general population sample that cannabis use modifies the
association between PTSD and depression and suicidality, such
that the associations appear to be strengthened in the absence of
cannabis use and tempered in the presence of lower-risk cannabis
use. However, a number of limiting factors prevent the ability to
determine a causal connection. Despite the overall findings of
this study suggesting a possible therapeutic benefit of cannabis
use in the treatment of PTSD, cannabis use is not without risks,
including the development of CUD. High-quality randomized
controlled trials that can measure the possible effect of therapeu-
tic cannabis use on PTSD-specific symptoms as well as these
broader mental health outcomes over time are urgently needed to
elucidate whether the current finding is in fact indicative of a
biological interaction.
Acknowledgements
This analysis is based on the 2014 public use microdata file for Statistics
Canada’s Canadian Community Health Survey – Mental Health cycle
(2012). All computations, use, and interpretation of these data are entirely
those of the study authors. The views expressed do not represent the
views of Statistics Canada. SL is supported by doctoral scholarships from
the Canadian Institutes of Health Research (CIHR) and the Pierre Elliott
Trudeau Foundation. TK is supported by a foundation grant from CIHR
(FDN-148476). M-JM and BDLM are supported by the US National
Institutes of Health (U01-DA021525). M-JM is also supported through a
New Investigator Award from CIHR and a Scholar Award from the
Michael Smith Foundation for Health Research. We would like to thank
Huiru Dong for providing statistical support for this research.
Declaration of Conflicting Interest
The author(s) declared the following potential conflicts of interest with
respect to the research, authorship, and/or publication of this article:
M-JM is a professor of cannabis science at the University of British
Columbia, a position established by arms’ length gifts to the university
from Canopy Growth, a licensed producer of cannabis, and the
Government of British Columbia’s Ministry of Mental Health and
Addictions. The University of British Columbia has received unstruc-
tured funding from NG Biomed, Ltd. (an applicant to the Canadian fed-
eral government for a license to produce medical cannabis) to support
M-JM. ZW is coordinating principal investigator on a clinical trial of
cannabis for PTSD sponsored by Tilray, a licensed producer of medical
cannabis. All other authors report no conflicts of interest.
Funding
The author(s) received no financial support for the research, authorship,
and/or publication of this article.
ORCID iD
M-J Milloy https://orcid.org/0000-0003-3821-6221
References
Boden MT, Babson KA, Vujanovic AA, et al. (2013) Posttraumatic stress
disorder and cannabis use characteristics among military veterans
with cannabis dependence. Am J Addict 22: 277–284.
Bonn-Miller M, Boden M, Vujanovic A, et al. (2013) Prospective investi-
gation of the impact of cannabis use disorders on posttraumatic stress
disorder symptoms among veterans in residential treatment. Psychol
Trauma 5: 193–200.
Bonn-Miller MO, Babson KA and Vandrey R (2014) Using cannabis
to help you sleep: Heightened frequency of medical cannabis use
among those with PTSD. Drug Alcohol Depend 136: 162–165.
Bonn-Miller MO, Vujanovic AA, Feldner MT, et al. (2007) Posttrau-
matic stress symptom severity predicts marijuana use coping motives
among traumatic event-exposed marijuana users. J Trauma Stress
20: 577–586.
Bordieri MJ, Tull MT, McDermott MJ, et al. (2014) The moderating role
of experiential avoidance in the relationship between posttraumatic
8 Journal of Psychopharmacology 00(0)
stress disorder symptom severity and cannabis dependence. J Con-
textual Behav Sci 3: 273–278.
Bremner JD (2006) Traumatic stress: Effects on the brain. Dialogues Clin
Neurosci 8: 445–461.
Clinicaltrials.gov. (2015) Evaluating Safety and Efficacy of Cannabis in
Participants With Chronic Posttraumatic Stress Disorder. Avail-
able at: https://clinicaltrials.gov/ct2/show/results/NCT02517424
(accessed 30 January 2018).
Cougle JR, Bonn-Miller MO, Vujanovic AA, et al. (2011) Posttraumatic
stress disorder and cannabis use in a nationally representative sam-
ple. Psychol Addict Behav 25: 554–558.
D’Aliesio R (2017) Veterans Face Much Higher Suicide Rate than Civil-
ians. Available at: https://www.theglobeandmail.com/news/national/
veterans-dying-by-suicide-at-significantly-higher-rate-than-general-
population-study/article37237508/ (accessed 3 April 2018).
Duckers ML, Alisic E and Brewin CR (2016) A vulnerability paradox in
the cross-national prevalence of post-traumatic stress disorder. Br J
Psychiatry 209: 300–305.
Earleywine M and Bolles JR (2014) Marijuana, expectancies, and post-
traumatic stress symptoms: A preliminary investigation. J Psychoac-
tive Drugs 46: 171–177.
Fraser GA (2009) The use of a synthetic cannabinoid in the management
of treatment-resistant nightmares in posttraumatic stress disorder
(PTSD). CNS Neurosci Ther 15: 84–88.
Freeman D, Thompson C, Vorontsova N, et al. (2013) Paranoia and post-
traumatic stress disorder in the months after a physical assault: A
longitudinal study examining shared and differential predictors. Psy-
chol Med 43: 2673–2684.
Gentes EL, Schry AR, Hicks TA, et al. (2016) Prevalence and correlates
of cannabis use in an outpatient VA posttraumatic stress disorder
clinic. Psychol Addict Behav 30: 415–421.
Gradus JL, Qin P, Lincoln AK, et al. (2010) Posttraumatic stress disorder
and completed suicide. Am J Epidemiol 171: 721–727.
Greer GR, Grob CS and Halberstadt AL (2014) PTSD symptom reports
of patients evaluated for the New Mexico Medical Cannabis Pro-
gram. J Psychoactive Drugs 46: 73–77.
Hillard CJ (2015) The endocannabinoid signaling system in the CNS: A
primer. Int Rev Neurobiol. 125: 1–47.
Hosmer DW and Lemeshow S (1992) Confidence interval estimation of
interaction. Epidemiology 3: 452–456.
Jetly R, Heber A, Fraser G, et al. (2015) The efficacy of nabilone, a syn-
thetic cannabinoid, in the treatment of PTSD-associated nightmares:
A preliminary randomized, double-blind, placebo-controlled cross-
over design study. Psychoneuroendocrinology 51: 585–588.
Kevorkian S, Bonn-Miller MO, Belendiuk K, et al. (2015) Associations
among trauma, posttraumatic stress disorder, cannabis use, and can-
nabis use disorder in a nationally representative epidemiologic sam-
ple. Psychol Addict Behav 29: 633–638.
Knol MJ and VanderWeele TJ (2012) Recommendations for presenting
analyses of effect modification and interaction. Int J Epidemiol 41:
514–520.
Knol MJ, VanderWeele TJ, Groenwold RHH, et al. (2011) Estimating
measures of interaction on an additive scale for preventive expo-
sures. Eur J Epidemiol 26: 433–443.
Neumeister A, Normandin MD, Pietrzak RH, et al. (2013) Elevated brain
cannabinoid CB1 receptor availability in post-traumatic stress disorder:
A positron emission tomography study. Mol Psychiatry 18: 1034–1040.
O’Neil ME, Nugent SM, Morasco BJ, et al. (2017) Benefits and harms of
plant-based cannabis for posttraumatic stress disorder: A systematic
review. Ann Intern Med 167: 332–340.
Pai A, Suris AM and North CS (2017) Posttraumatic stress disorder in the
DSM-5: Controversy, change, and conceptual considerations. Behav
Sci 7: 7–14.
Passie T, Emrich HM, Karst M, et al. (2012) Mitigation of post-traumatic
stress symptoms by Cannabis resin: A review of the clinical and neu-
robiological evidence. Drug Test Anal 4: 649–659.
Pineles SL, Mostoufi SM, Ready CB, et al. (2011) Trauma reactivity,
avoidant coping, and PTSD symptoms: A moderating relationship?
J Abnorm Psychol 120: 240–246.
Ramsawh HJ, Fullerton CS, Mash HB, et al. (2014) Risk for suicidal
behaviors associated with PTSD, depression, and their comorbidity
in the U.S. Army. J Affect Disord 161: 116–122.
Reinarman C, Nunberg H, Lanthier F, et al. (2011) Who are medical
marijuana patients? Population characteristics from nine California
assessment clinics. J Psychoactive Drugs 43: 128–135.
Richardson JD, Thompson A, King L, et al. (2017) Insomnia, psychi-
atric disorders and suicidal ideation in a national representative
sample of active Canadian Forces members. BMC Psychiatry 17:
211–221.
Roitman P, Mechoulam R, Cooper-Kazaz R, et al. (2014) Preliminary,
open-label, pilot study of add-on oral Delta9-tetrahydrocannabinol
in chronic post-traumatic stress disorder. Clin Drug Investig 34:
587–591.
Ruglass LM, Shevorykin A, Radoncic V, et al. (2017) Impact of can-
nabis use on treatment outcomes among adults receiving cognitive-
behavioral treatment for PTSD and substance use disorders. J Clin
Med 6: E14.
Serrano A (2018) As Vets Demand Cannabis for PTSD, Science Races
to Unlock Its Secrets. Available at: https://www.scientificamerican.
com/article/as-vets-demand-cannabis-for-ptsd-science-races-to-
unlock-its-secrets/ (accessed 3 April 2018).
Shalev A, Liberzon I and Marmar C (2017) Post-traumatic stress disor-
der. New Engl J Med 376: 2459–2469.
Solowij N, Lorenzetti V and Yücel M (2016) Effects of cannabis use on
human behavior: A call for standardization of cannabis use metrics.
JAMA Psychiatry 73: 995–996.
Statistics Canada (2013) Canadian Community Health Survey - Mental
Health (CCHS). Available at: http://www23.statcan.gc.ca/imdb/p2SV.
pl?Function=getSurvey&SDDS=5015 (accessed 10 October 2017).
Steenkamp MM, Blessing EM, Galatzer-Levy IR, et al. (2017) Marijuana
and other cannabinoids as a treatment for posttraumatic stress disor-
der: A literature review. Depress Anxiety 34: 207–216.
The University of British Columbia Board of Governors (2012) Research
Involving Human Participants (Policy 89). Available at: http://
universitycounsel.ubc.ca/files/2012/06/policy89.pdf (accessed 2
December 2017).
Trezza V and Campolongo P (2013) The endocannabinoid system as
a possible target to treat both the cognitive and emotional features
of post-traumatic stress disorder (PTSD). Front Behav Neurosci 7:
100.
Tsai J, Harpaz-Rotem I, Pilver CE, et al. (2014) Latent class analysis
of personality disorders in adults with posttraumatic stress disorder:
Results from the National Epidemiologic Survey on Alcohol and
Related Conditions. J Clin Psychiatry 75: 276–284.
VanderWeele TJ and Knol MJ (2014) A tutorial on interaction. Epide-
miol Methods 3: 33–72.
Veterans Affairs Canada (2019) Cannabis for Medical Purposes. Avail-
able at: http://www.veterans.gc.ca/eng/news/vac-responds/just-the-
facts/cannabis-medical-purposes (accessed 8 October 2019).
Wilkinson ST, Stefanovics E and Rosenheck RA (2015) Marijuana use
is associated with worse outcomes in symptom severity and violent
behavior in patients with posttraumatic stress disorder. J Clin Psy-
chiatry 76: 1174–1180.
Zer-Aviv MT, Segev A and Akirav I (2016) Cannabinoids and post-trau-
matic stress disorder: Clinical and preclinical evidence for treatment
and prevention. Behav Pharmacol 27: 561–569.