Abstract

Introduction

Both substance use and mental health disorders contribute substantially to the global burden of disease 1, 2. With legalization of cannabis use (CU) in Canada and many states in the United States, there is considerable interest in how CU affects or is affected by mental health concerns. Recent reviews and meta‐analyses of research show that CU is consistently associated with the onset and severity of psychotic symptoms 3, 4, 5, 6. Research also suggests that CU may have modest effects on mood and anxiety, although findings are inconsistent 4, 5, 7. These inconsistencies relate to differences in age groups sampled, cultural differences in illegality of CU and differences in symptoms assessed. It is also possible that associations between cannabis and mental health indicators are non‐linear, given that experiences of psychotic and mood disorders may be episodic. In addition, concentrations of delta‐9‐tetrahydrocannabinol (THC), age of onset, frequency, duration of use and a host of co‐occurring risks are also related to both acute effects of cannabis and mental health symptoms 4, 5. As a result, the influences of CU and cannabis use disorder (CUD) on mental health have proved difficult to unravel.

Particular concern focuses on CU in adolescents. Several recent reviews suggest that adolescent onset of CU (i.e. 15 years or younger) conveys age‐specific risks for concurrent and adult‐onset mental health disorders 3, 4, 5. Theory relates this age‐sensitivity to ongoing brain maturation and greater reward‐sensitivity and reward‐seeking in adolescence and considerable epidemiological, animal and neuropsychological research shows some support for heightened risks related to adolescent onset of CU (see reviews by 4, 5). However, comparisons with adult samples are lacking and findings differ by symptom type 8, 9. One longitudinal study of Swedish adults aged 20–64 with a 3‐year follow‐up 10 found that CU at baseline was associated with risk for depression and anxiety (as well as the reverse); however, these associations were not significant beyond co‐occurring risks (e.g. alcohol and illicit drug use, education, family tension, place of upbringing) and age of onset of use was not examined.

However, in addition to sensitivities related to brain maturation 11, there are several reasons why adolescents may show stronger associations between CU and mental health symptoms than older adults. First, Winters & Lee 12 showed that risks for CU dependency were greater in recent‐onset users at ages 14–18 than at ages 22–26, although Chen et al. 13 found that adolescents may over‐report the clinical markers of CUD (e.g. used more than intended, spent significant time in activities to obtain cannabis or disrupted family or work relationships). Secondly, more frequent or longer‐term use (also characteristic of older users) may lower sensitivity to the negative acute effects of THC as tolerance increases 14, 15, suggesting that adults could have lower associations between CU and CUD and mental health symptoms. Thirdly, some research also finds that associations between CU and symptoms are not significant after controlling for common liabilities or co‐occurring risks (e.g. alcohol and illicit drug use, education, family tension, place of upbringing, abuse and victimization). Such risks may also be particularly likely to contribute to allostatic load, stress and mental health symptoms in adolescence 11. Fourthly, many mental health disorders onset in adolescence and this temporal co‐occurrence may contribute to the strength of their associations with CU and CUD in adolescence. For example, almost 50% of life‐time cases of any mental health disorder start by age 14 and 75% start by age 24 16, 17. However, Kessler et al. 16 also found marked differences in mean ages of onset of anxiety disorders (11 years), substance use (20 years) and mood disorders (30 years), and the onset of psychotic episodes was rare before 12 or after 30 years, suggesting that correlations may differ by symptom type. The strength of associations between psychiatric diagnoses and CU and CUD may increase from adolescence to young adulthood as new cases increase, and then level off as new cases become less likely. Finally, associations could be stronger in individuals with an early age of onset of CU given longer timing of exposure, or the hypothesized effects on adolescents’ brain development, particularly if use is sustained across adolescence and young adulthood 5. CU and mental health symptoms differ for males and females in adolescence and adulthood 18, 19, 20 and some research shows associations between CU and mental health symptoms for females 21, while others show associations for males 22, warranting further investigation of sex differences in associations.

This study has three aims:

1. To assess changes in strength of the associations between CU and CUD and symptoms of psychosis, depression and anxiety at each age in an adolescent and adult sample;

2. To examine differences in these associations for early (versus late) adolescent CU onset; and

3. To examine sex differences in these associations.

Method

Samples

Victoria healthy youth survey

The V‐HYS 23 included 662 Canadian youth assessed biannually across six waves (Table 1). Random digit‐dialing of 9500 telephone listings revealed 1036 households with a youth aged 12–18; 662 youth both had parental consent and agreed to participate. Youth received gift certificates at each interview. Trained interviewers administered surveys and sections dealing with sensitive topics (e.g. substance use) were collected via self‐reported questionnaires to enhance confidentiality. Retention rates at each wave were good (87–72%). Fifty‐three per cent of youth (n = 350) participated in all waves, 16% (n = 107) in five waves, 10% (n = 63) in four, 8% (n = 55) in three, 6% (n = 40) in two and 7% (n = 47) in only one wave. Youth remaining at the sixth wave were more likely to be female, χ² _{(1, 662)} = 8.77, P = 0.003, had higher SES, F _{(1, 655)} = 14.90, P < 0.001 and were less likely to be smokers (F _{(1, 660)} = 3.82, P = 0.05.

Table 1

Demographic and study variable characteristics of the V‐HYS (T1 n = 662).

	Mean (SD) or n (%)
Demographics (time 1)
Sex (% female)	342 (52%)
Age	15.52 (1.92)
SESa	2.88 (1.38)
Ethnicity (% white)	561 (85%)
Time 1
Covariates
Cigarette use	75 (11%)
Heavy episodic drinking (HED)	0.60 (1.00)
Predictors
Cannabis use (CU) frequency (average)	0.73 (1.21)
Never	426 (64%)
A few times a year	109 (17%)
A few times per month	60 (9%)
Once a week	15 (2%)
More than once a week	52 (8%)
Cannabis use disorder (CUD)	–b
Outcomes
Psychotic symptoms	–c
Depressive symptoms	2.79 (2.53)
Anxiety symptoms	5.81 (2.58)
Time 2
Covariates
Cigarette use	107 (19%)
Heavy episodic drinking (HED)	1.14 (1.24)
Predictors
Cannabis use (CU) frequency (average)	1.17 (1.45)
Never	277 (48%)
A few times a year	113 (20%)
A few times per month	74 (13%)
Once a week	22 (4%)
More than once a week	86 (15%)
Cannabis use disorder (CUD)	–b
Outcomes
Psychotic symptoms	–c
Depressive symptoms	3.19 (2.57)
Anxiety symptoms	6.24 (2.62)
Time 3
Covariates
Cigarette use	131 (25%)
Heavy episodic drinking (HED)	1.55 (1.29)
Predictors
Cannabis use (CU) frequency (average)	1.45 (1.54)
Never	207 (39%)
A few times a year	128 (24%)
A few times per month	54 (10%)
Once a week	38 (7%)
More than once a week	105 (20%)
Cannabis use disorder (CUD)	–b
Outcomes
Psychotic symptoms	3.34 (2.54)
Depressive symptoms	3.60 (2.66)
Anxiety symptoms	6.26 (2.51)
Time 4
Covariates
Cigarette use	123 (27%)
Heavy episodic drinking (HED)	1.62 (1.28)
Predictors
Cannabis use (CU) frequency (average)	1.40 (1.56)
Never	191 (42%)
A few times a year	97 (22%)
A few times per month	48 (11%)
Once a week	20 (4%)
More than once a week	95 (21%)
Cannabis use disorder (CUD)	116 (26%)
Outcomes
Psychotic symptoms	2.69 (2.37)
Depressive symptoms	3.07 (2.62)
Anxiety symptoms	5.98 (2.63)
Time 5
Covariates
Cigarette use	127 (28%)
Heavy episodic drinking (HED)	1.52 (1.21)
Predictors
Cannabis use (CU) frequency (average)	1.21 (1.46)
Never	206 (46%)
A few times a year	113 (25%)
A few times per month	39 (9%)
Once a week	24 (5%)
More than once a week	71 (16%)
Cannabis use disorder (CUD)	79 (17%)
Outcomes
Psychotic symptoms	2.41 (2.27)
Depressive symptoms	2.88 (2.58)
Anxiety symptoms	5.61 (2.67)
Time 6
Covariates
Cigarette use	106 (22%)
Heavy episodic drinking (HED)	1.44 (1.21)
Predictors
Cannabis use (CU) frequency (average)	1.22 (1.47)
Never	214 (45%)
A few times a year	116 (25%)
A few times per month	41 (9%)
Once a week	24 (5%)
More than once a week	77 (16%)
Cannabis use disorder (CUD)	73 (16%)
Outcomes
Psychotic symptoms	2.24 (2.34)
Depressive symptoms	2.88 (2.65)
Anxiety symptoms	5.44 (2.78)

Open in a separate window

^aSocio‐economic status (SES) was assessed by mother's level of education (0 = did not finish high school to 4 = finished college/university) in the Victoria Healthy Youth Survey (V‐HYS). SD = standard deviation.

^bCUD was not assessed until T4 (2009) in the V‐HYS.

^cPsychotic symptoms were not assessed until T3 (2007) in the V‐HYS.

NSARC‐III

The NESARC‐III (2012–13 25) is a nationally representative sample of 36309 non‐institutionalized US adults aged 18 and older (Table 2). NESARC‐III were sampled from non‐institutionalized civilian populations aged 18 and older, including residents of group quarters (e.g. group homes, dormitories; see 25 for more detailed information regarding data collection procedures). Black, Hispanic and Asian adults were oversampled to provide more reliable estimates; thus, survey weights were incorporated in all analyses. Structured interviews were conducted face‐to‐face. Participants received $90 in compensation.

Table 2

Demographic and study variable characteristics of the NESARC‐III (n = 30999).

	Mean (SD) or n (%)
Demographics
Sex (% female)	17283 (51%)
Age	40.61 (13.41)
SESa	7.25 (4.64)
Ethnicity (% non‐Hispanic white)	15392 (64%)
Covariates
Cigarette use	9350 (30%)
Heavy episodic drinking (HED)	11196 (37%)
Predictors
Cannabis use (CU) frequency (average)	0.31 (0.95)
Never	27360 (89%)
A few times a year	1092 (4%)
A few times per month	656 (2%)
Once a week	370 (1%)
More than once a week	1510 (4%)
Cannabis use disorder (CUD)	965 (3%)
Outcomes
Psychotic symptoms	1.12 (1.66)
Depressive symptoms	2.00 (3.01)
Anxiety symptoms	0.78 (1.81)

Open in a separate window

^aSocio‐economic status (SES) was assessed by participant income (0 = $0 to 17 = $100000 or more) in the National Epidemiologic Survey on Alcohol And Related Conditions (NESARC)‐III. SD = standard deviation.

Results

V‐HYS findings

Changes across age

Figure 1 presents the models showing how CU frequency or the proportion of participants with a CUD varies across age. The y‐axis presents the estimated mean of past year CU frequency (Fig. 1a). CU increased until approximately 20 years (e.g. estimated mean at age 20 = 1.49, 95% CI = 1.36, 1.61) then began to gradually decline. In Fig. 1b, the y‐axis presents the prevalence of CUD which declined across age. Psychotic symptoms were reported at the highest levels in adolescence and declined steadily across age. Depressive and anxiety symptoms increased slightly until approximately 18 years and were then stable (Supporting information, Fig. S1).

Open in a separate window

Figure 1

Victoria Healthy Youth Survey (V‐HYS) data: intercept‐only models for (a) cannabis use frequency (solid line represents estimated mean frequency) and (b) cannabis use disorder (solid line represents proportion of individuals with a past year cannabis use disorder). CI = confidence interval

CU frequency and mental health symptoms

More frequent CU was positively significantly associated (main effect CI > 0) with psychotic symptoms after age 22 (b = 0.13, 95% CI = 0.002, 0.25) (Fig. 2a). Figure 2b shows that CU was significantly associated with more depressive symptoms from ages 16–19 (e.g. age 17: b = 0.21, 95% CI = 0.04, 0.37) and following age 25 (b = 0.17, 95% CI = 0.003, 0.34). CU was not associated with anxiety symptoms in this sample (Fig. 2c). Time‐varying interactions of CU with age of onset (Supporting information, Fig. S2) and sex (Supporting information, Fig. S3) were not significant.

Open in a separate window

Figure 2

Victoria Healthy Youth Survey (V‐HYS) data: the time‐varying main effect (b) of cannabis use frequency on (a) psychotic, (b) depressive and (c) anxiety symptoms. Models are adjusted for sex, socio‐economic status (SES), cigarette use and heavy drinking. Results are presented as regression coefficients (b); 95% confidence interval (CI) (dashed lines) that do not include 0 indicate periods of significance

CUD and mental health symptoms

CUD was associated with psychotic symptoms after age 23 (b = 0.51, 95% CI = 0.01, 1.01) (Fig. 3a). CUD was also associated with higher depressive symptoms at ages 19–20 (e.g. age 20: b = 0.70, 95% CI = 0.02, 1.39) and following age 25 (b = 0.71, 95% CI = 0.001, 1.42) (Fig. 3b) and with higher anxiety symptoms at ages 26–27 only (e.g. age 27: b = 0.96, 95% CI = 0.10, 1.82) (Fig. 3c). Time‐varying interactions with age of onset were not significant (Supporting information, Fig. S4). However, sex differences were significant for associations between CUD and psychotic and depressive symptoms. For psychotic symptoms, the interaction was significant following age 26 (b = 1.12, 95% CI = 0.02, 2.21) (Fig. 4a), when females showed stronger associations than males (Fig. 4b). For depressive symptoms, the interaction was significant at 19–20 (e.g. age 20: b = −1.22, 95% CI = 0−2.42, −0.01) (Fig. 4c), when males showed stronger associations than females (Fig. 4d), and following age 25 (e.g. age 27: b = 2.39, 95% CI = 0.91, 3.87), when females showed stronger associations than males (Fig. ​(Fig.4d).4d). Sex did not moderate the association between CUD and anxiety symptoms (Fig. 4e).

Open in a separate window

Figure 3

Victoria Healthy Youth Survey (V‐HYS) data: the time‐varying main effect (b) of cannabis use disorder on (a) psychotic, (b) depressive and (c) anxiety. Models are adjusted for sex, socio‐economic status (SES), cigarette use and heavy drinking. Results are presented as regression coefficients (b); 95% confidence interval (CI) (dashed lines) that do not include 0 indicate periods of significance

Open in a separate window

Figure 4

Victoria Healthy Youth Survey (V‐HYS): the time‐varying interaction (b) of cannabis use disorder by sex on psychotic (a,b), depressive (c,d) and anxiety (e; non‐significant) symptoms. The time‐varying interaction is plotted on the left. The sex difference in the main effect is plotted on the right; the area inside the box represents the approximate ages when the sex interaction is significant (P < 0.05). Models are adjusted for socio‐economic status (SES), cigarette use and heavy drinking. b = regression coefficient; CI = confidence intervals

NESARC‐III findings

Changes across age

Figure 5 presents the intercept‐only models for the NESARC‐III CU data. Frequency of CU and prevalence of CUD declined from ages 18 to 40 and then were stable (Fig. 5a,b, respectively). Mental health symptoms were relatively stable over time (Supporting information, Fig. S5).

Open in a separate window

Figure 5

National Epidemiologic Survey on Alcohol and Related Conditions (NESARC)‐III data: intercept‐only models for (a) cannabis use frequency (solid line represents estimated mean frequency) and (b) cannabis use disorder (solid line represents proportion of individuals with a past year cannabis use disorder). CI = confidence interval

CU frequency and mental health symptoms

More frequent CU had a significant and stable association with more psychotic symptoms from age 18 (b = 0.21, 95% CI = 0.10, 0.33) to age 65 (b = 0.36, 95% CI = 0.16, 0.56) (Fig. 6a). Frequent CU was positively associated with depressive symptoms from age 18 (b = 0.22, 95% CI = 0.05, 0.40) to age 64 (b = 0.25, 95% CI = 0.01, 0.48) (Fig. 6b). CU was also positively associated with anxiety symptoms from age 20 (b = 0.07, 95% CI = 0.01, 0.13) to age 63 (b = 0.13, 95% CI = 0.01, 0.25) (Fig. 6c). Time‐varying interactions of CU with early age of onset were not significant beyond chance (Supporting information, Fig. S6). Time‐varying interactions of CU with sex were significant for psychotic symptoms between ages 20 (b = 0.15, 95% CI = 0.01, 0.29) to age 24 (b = 0.11, 95% CI = 0.01, 0.21) and between ages 47 (b = 0.12, 95% CI = 0.003, 0.25) to age 61 (b = 0.24, 95% CI = 0.01, 0.47), for depressive symptoms between ages 18 (b = 0.50, 95% CI = 0.17, 0.82) to age 25 (b = 0.17, 95% CI = 0.001, 0.34) and between ages 32 (b = 0.17, 95% CI = 0.01, 0.34) to age 41 (b = 0.22, 95% CI = 0.001, 0.44) and for anxiety symptoms between ages 18 (b = 0.19, 95% CI = 0.003, 0.39) to age 23 (b = 0.11, 95% CI = 0.01, 0.21) and ages 32 (b = 0.11, 95% CI = 0.001, 0.23) to age 46 (b = 0.14, 95% CI = 0.003, 0.28) (Fig. 7a,c,e). During these periods, females showed stronger associations between CU frequency and symptoms than males (Fig. 7b,d,f).

Open in a separate window

Figure 6

National Epidemiologic Survey on Alcohol and Related Conditions (NESARC)‐III data: the time‐varying main effect (b) of cannabis use frequency on (a) psychotic, (b) depressive and (c) anxiety symptoms. Models are adjusted for sex, socio‐economic status (SES), cigarette use and heavy drinking. Results are presented as regression coefficients (b); 95% confidence interval (CI) (dashed lines) that do not include 0 indicate periods of significance

Open in a separate window

Figure 7

National Epidemiologic Survey on Alcohol and Related Conditions (NESARC)‐III data: the time‐varying interaction (b) of cannabis use frequency by sex on psychotic (a,b), depressive (c,d) and anxiety (e,f) symptoms. The time‐varying interaction is plotted on the left. The sex difference in the main effect is plotted on the right; the area inside the box represents the approximate ages when the sex interaction is significant (P < 0.05). Models are adjusted for socio‐economic status (SES), cigarette use and heavy drinking. b = regression coefficient; CI = confidence intervals

CU disorder and mental health symptoms

CUD was associated with more psychotic symptoms from age 18 (b = 1.21, 95% CI = 0.66, 1.76) to age 64 (b = 1.09, 95% CI = 0.05, 2.12) (Fig. 8a), more depressive symptoms from age 18 (b = 0.96, 95% CI = 0.19, 1.73) to age 61 (b = 1.11, 95% CI = 0.01, 2.21) (Fig. 8b) and more anxiety symptoms from age 18 (b = 0.45, 95% CI = 0.02, 0.88) to age 62 (b = 0.75, 95% CI = 0.08, 1.43) (Fig. 8c). Time‐varying interactions of CUD with early age of onset were not significant beyond chance (Supporting information, Fig. S7). Time‐varying interactions of CUD with sex were significant for psychotic symptoms between ages 18 (b = 1.21, 95% CI = 0.18, 2.25) to age 23 (b = 0.56, 95% CI = 0.01, 1.12), for depressive symptoms between ages 18 (b = 1.81, 95% CI = 0.30, 3.31) to age 26 (b = 1.00, 95% CI = 0.01, 2.00) and for anxiety symptoms between ages 18 (b = 1.46, 95% CI = 0.36, 2.56) to age 22 (b = 0.59, 95% CI = 0.003, 1.18) (see Fig. 9a,c,e). During these periods, females showed stronger associations between CUD and mental health symptoms than males (Fig. 9b,d,f).

Open in a separate window

Figure 8

National Epidemiologic Survey on Alcohol and Related Conditions (NESARC)‐III data: the time‐varying main effect (b) of cannabis use disorder on (a) psychotic, (b) depressive and (c) anxiety symptoms. Models are adjusted for sex, socio‐economic status (SES), cigarette use and heavy drinking. Results are presented as regression coefficients (b); 95% confidence interval (CI) (dashed lines) that do not include 0 indicate periods of significance

Open in a separate window

Figure 9

NESARC‐III data: the time‐varying interaction (b) of cannabis use disorder by sex on psychotic symptoms (a,b), depressive (c,d) and anxiety (e,f) symptoms. The time‐varying interaction is plotted on the left. The sex difference in the main effect is plotted on the right; the area inside the box represents the approximate ages when the sex interaction is significant (P < 0.05). Models are adjusted for socio‐economic status (SES), cigarette use and heavy drinking. b = regression coefficient; CI = confidence intervals

Discussion

This study is the first, to our knowledge, to examine the strength of the associations of CU and CUD with symptoms of psychosis, depression and anxiety at each age in adolescence (V‐HYS) and in adults (NESARC‐III). We used TVEM, which does not assume linear relations among the variables and allows for changes in the strengths of associations at each age 30. Our findings are consistent with past research showing associations between CU and psychosis in late adolescence and early young adulthood 3, 4, 5, 6. The associations with depression and anxiety is less consistent in past research 4, 5, 7. We also accounted for SES, cigarette use and HED and examined the moderating effects of age of onset and sex. Interactions with age of onset were not significant. Sex differences were significant.

In the V‐HYS, CU and CUD were associated with symptoms of psychosis (after age 22) and depression (ages 16–19 and after 25), but not anxiety. Findings were not significant for anxiety for the V‐HYS sample and this association was weakest for the NESARC‐III sample, suggesting power limitations due to the smaller sample size of the V‐HYS sample may have decreased our ability to detect associations. Associations with anxiety peaked in ages 40s and 50s for the NESARC‐III. Findings of associations between CU and anxiety disorders are also inconsistent in past research 32. Individuals experiencing anxiety symptoms may avoid CU if anxiety is experienced as an effect 10.

In the V‐HYS, the associations with CUD and psychosis and depressive symptoms were stronger for females than males after age 23–25. Associations for depressive symptoms were stronger for males than females at ages 19–20. In the NESARC‐III, CU and CUD were associated with all three mental health symptoms across age and associations were stronger for females, in young adulthood (approximately ages 18–25). These findings are consistent with a growing body of research 33, 34 showing that associations between CUD and mental health symptoms are stronger in females relative to males in young adulthood. Females appear to have a shorter or ‘telescoped’ trajectory from experimentation to addiction 35. In the V‐HYS, males showed significantly higher associations between depressive symptoms at ages 19–20 compared to females. This may be related to males’ more frequent use of greater amounts of cannabis, and greater impulsivity, sensation‐seeking and avoidant coping strategies at this age 22. Differences may also reflect sex differences in neural systems responsible for motivation and addiction or variations in contexts of CU by males (e.g. with peers or in sports) compared to females, who may be more likely to use substances in isolation for self‐medication or stress reduction 36.

Contrary to expectations from past literature, the interaction of CU frequency or CUD and mental health symptoms with early age of onset (15 years or younger) were not significant. Early onset of CU may have cascading effects that affect other age‐specific areas of concern including contributions to legal, work and relationship problems that are subsequently associated with increasing mental health symptoms 5, 11, 12. Longitudinal data following early‐onset users well into adulthood is needed to unravel whether early onset has a distinctive age‐sensitive association with mental health.

As with all correlational research, the causal direction of effects cannot be determined; however, the consistent association of CU frequency and CUD with symptoms of psychosis and depression across age groups suggests that more research with adults is needed. The strength of associations was not specific to the typical ages of onset of psychosis or depression 16 or attributable to early onset of CU. Adolescents and adults experiencing psychosis or depression may also experience aggravated symptoms if they begin or continue to use cannabis.

Declaration of interests

None.

Supporting information

Acknowledgements

Notes

Leadbeater, B. J. , Ames, M. E. , and Linden‐Carmichael, A. N. (2019) Age‐varying effects of cannabis use frequency and disorder on symptoms of psychosis, depression and anxiety in adolescents and adults. Addiction, 114: 278–293. 10.1111/add.14459. [Europe PMC free article] [Abstract] [CrossRef] [Google Scholar]

References