Cannabis use is associated with changes in psychological and functional well-being during young adulthood: evidence from self-reports and hair analyses

Abstract

Background

Cannabis use in young adulthood is common, yet few studies have explored how it predicts changes in psychopathology and functional well-being in community samples. We assessed these links using both self-reported frequency of cannabis use and hair THC concentrations.

Methods

Data came from a community sample of young adults (N = 863) who reported cannabis use (weekly-to-daily use: n = 150) and provided hair samples at age 20 (cannabis detected: n = 110). Liquid chromatography–tandem mass spectrometry quantified delta-9-tetrahydrocannabinol (THC) and cannabinol (CBN) concentrations in hair. At ages 20 and 24, participants reported psychopathology (psychotic-like experiences, problematic substance use, internalizing symptoms, and aggression) and functional wellbeing (general well-being, delinquency, and not being in employment, education, or training). Multiple linear and logit regression models tested associations between six different continuous and dichotomous operationalizations of self-reported and objective cannabis exposure at age 20 and psychological and functional well-being at age 24, adjusting for sex, sociodemographic characteristics, and the outcomes measured at age 20.

Results

Both self-reported frequency of cannabis use and hair THC concentrations predicted increases in psychotic-like experiences and internalizing symptoms, increased aggression, decreased general well-being, higher odds of not being in employment, training, or education, and more problematic substance use from age 20 to 24, with small effect sizes. Composite exposure scores derived from self-reports and hair data were not more informative than either source alone.

Conclusions

Frequent cannabis use predicted adverse changes in psychopathological outcomes from ages 20 to 24, regardless of how it was assessed.

Introduction

Cannabis is one of the most used psychotropic substances among young people worldwide, next to alcohol and nicotine (UNODC, 2024). In the last decade, Canada and parts of the United States have legalized recreational cannabis, resulting in lower risk perceptions and higher rates of use in adults, though findings for adolescents have been mixed (Mennis, McKeon, & Stahler, 2023; Patrick, Miech, Johnston, & O’Malley, 2023; Pawar, Firmin, Wilens, & Hammond, 2024). In Western Europe, countries are increasingly shifting to decriminalizing or legalizing recreational cannabis use, with some allowing home cultivation for personal use (e.g. Germany and Malta). In some countries, like Switzerland, recreational cannabis use is still illegal but de facto decriminalized, and several trials are testing the sale of cannabis to adults (FOPH, 2024).

Legalization has brought changes in cannabis product availability and diversity (e.g. Hammond et al., 2022; Matheson & Le Foll, 2020). At the same time, cannabis potency (i.e. higher levels of delta-9-tetrahydrocannabinol [THC]) has increased, regardless of legal status (e.g. EMCDDA, 2023). Higher THC levels increase potential harm (Curran et al., 2019; Freeman et al., 2019) while providing stronger intoxicating effects (Curran, Brignell, Fletcher, Middleton, & Henry, 2002). Many people with cannabis use are unaware of the health risks, such as the risk of psychosis and addiction (Goodman & Hammond, 2022). Indeed, evidence of the psychosocial consequences of the use of today’s cannabis products – including the assessment of the best measures of cannabis exposure – has not kept pace with legalization efforts. Further longitudinal research into the mental health and psychosocial implications of cannabis use among young people is needed (Chadi, Weitzman, & Levy, 2018).

Cannabis use in young adulthood

Cannabis use usually peaks between ages 15 and 25, a period of significant brain and social development (Arnett, Žukauskienė, & Sugimura, 2014; Lichenstein, Shaw, & Forbes, 2022; Steinberg, 2007). Developmental transitions, such as moving out of the parental household to start work or pursue professional training or higher education, are typical in the early 20s (Schulenberg & Schoon, 2012). Mental health problems, such as depression and anxiety, are currently at an all-time high in this age group (Keyes, Gary, O’Malley, Hamilton, & Schulenberg, 2019). Accordingly, identifying correlates of such mental health changes is important.

Although there is a large literature on how early cannabis use predicts later well-being (Copeland, Hill, & Shanahan, 2022; Shanahan et al., 2021), longitudinal work within young adulthood is rare, particularly in community (as opposed to clinical or college student) samples, and previous work typically relied on self-reports of cannabis use only. Data from community cohort samples, with multiple measures of cannabis exposure, can yield new insights into associations of cannabis exposure with negative psychopathological and functional outcomes in young adults from the general population.

Psychopathological and functional correlates of frequent cannabis use

The evidence linking cannabis use and psychopathology is mixed and differs between psychopathologies. However, there is convincing evidence of a robust association between the intensity of cannabis use and psychosis symptoms (Hjorthøj, Larsen, Starzer, & Nordentoft, 2021; Marconi, Di Forti, Lewis, Murray, & Vassos, 2016), which can be a prodrome for a psychosis. In other words, people with psychosis symptoms are at a higher risk of developing a disorder on the psychosis spectrum (Fusar-Poli, 2017; Staines et al., 2022). Long-term use of potent cannabis (i.e. high in THC) has been associated with hospitalization for psychosis (Di Forti et al., 2019; Marconi et al., 2016), and acute administration of THC has been shown to induce psychotic symptoms (Martin-Santos et al., 2012). However, few studies have investigated the associations between cannabis use and the development of prodromal psychotic symptoms in the general population prior to hospitalization for first-episode psychosis.

The evidence for associations between cannabis use and other mental health outcomes, such as depression and anxiety, is less conclusive. A meta-analysis of 11 studies found increased depression but not anxiety after adolescent cannabis use (Gobbi et al., 2019). A large study of 1420 participants aged 9 to 21 found that earlier cannabis use during adolescence and up to age 21 years was not associated with psychiatric outcomes such as anxiety and depression between ages 25 and 30, but that it was positively associated with adverse functional outcomes (e.g. extended unemployment and serious criminal charges) (Copeland et al., 2022). On the other hand, another meta-analysis of longitudinal studies found that heavy cannabis use, compared with light or no use, was associated with an increased risk of depression (Lev-Ran et al., 2014), with more recent evidence from longitudinal studies supporting this (Hengartner, Angst, Ajdacic-Gross, & Rössler, 2020). Previous analyses of the dataset used in this study (up to age 20 only) show that self-reported frequent (weekly to daily) cannabis use between ages 13 and 17 was associated with negative functional outcomes (higher delinquency, lower general well-being, and not being in education or employment), and problematic substance use at age 20, but not with psychosis or internalizing symptoms (Shanahan et al., 2021). Changes in well-being in relation to self-reported and hair-quantified cannabis exposure at age 20 have not been previously investigated in this sample.

Cannabis measurement

Inconsistencies in findings may stem from the fact that there is no standardized measure of cannabis exposure. Frequency of use and potency play key roles in associated harms, and an important methodological limitation in the epidemiological literature is the lack of measures to assess the body’s exposure to THC. Most studies have used self-reported measures of cannabis use, which have limitations, including social desirability and recall bias (Johnson & Fendrich, 2005). Furthermore, cannabis products vary in their composition and cannabinoids content, and due to the illegality of cannabis in many countries, people with cannabis use have little or no way of knowing the composition of the products they purchase on the black market and cannot reliably estimate their exposure to THC (Hindocha, Freeman, & Curran, 2017; Potter, Hammond, Tuffnell, Walker, & Di Forti, 2018; Watkins, Karliner-Li, Lee, Koester, & Ling, 2021). Even in regulated markets where labelling of THC content is mandatory, the label can be inaccurate (Oldfield et al., 2021). Thus, asking participants about the frequency or potency of their cannabis use is not sufficient to estimate the body’s actual exposure to cannabinoids, notably THC. Therefore, objective markers of the body’s exposure to THC could be useful to complement survey data.

Common biological specimens in which exposure to THC can be assayed, such as urine and blood plasma samples, capture a limited exposure period across several days to weeks (Musshoff & Madea, 2006) and can be invasive or uncomfortable, limiting their feasibility for larger scale studies. Hair samples capture exposure to THC over previous months, with each centimeter of hair length representing approximately 1 month, and can be stored for several months to years. While early hair method studies to quantify cannabinoid levels faced skepticism (Moosmann, Roth, & Auwärter, 2015), improved methodological pipelines can now detect continuous levels of cannabinoids in people with weekly to daily use, who are at the highest risk of harm (Kroon et al., 2024; Scholz, Cabalzar, Kraemer, & Baumgartner, 2021; Steinhoff et al., 2023; Taylor et al., 2017).

Few studies have used hair concentrations of THC to examine cannabis use correlates and outcomes, and most of these investigated samples of people with heavy (e.g. daily) use and therefore cannot be generalized to cannabis use in the community (Kroon et al., 2024; Nestoros et al., 2017). A study of 74 Dutch individuals with cannabis use disorder found no associations between hair concentrations of THC and mental health measured with a psychiatric interview (Kroon et al., 2024). In contrast, an earlier Greek study of 48 individuals with heavy cannabis use found higher auditory hallucinations (a symptom of psychosis) in those with higher hair cannabinoid levels (GC/MS) (Nestoros et al., 2017). Furthermore, a study conducted of 410 young people with cannabis use aged 16–24 years found that those with high THC levels in hair had higher anxiety and depression scores than those with low THC levels (Curran et al., 2019); however, there was no comparison group of participants without cannabis use, and this study relied on median splits of hair concentrations and did not investigate linear associations.

Overall, due to the limited evidence, it is still unclear how hair data and self-reports of cannabis exposure are each associated with later well-being in young adults from the community. While some studies have investigated psychopathology, no studies to our knowledge have used cannabis indexed in hair to investigate functional well-being, such as not being in employment, training, or education, or delinquency.

The current study

This study used data from a community sample of young adults in Switzerland (N = 863). Cannabis exposure was operationalized into six different continuous and dichotomous variables based on self-reports, hair data, or a composite of both. This was done to reflect previous operationalizations in the literature and to incorporate hair measures. Multiple linear and binary logit regressions tested associations of the cannabis exposure variables with changes in psychopathology (psychotic-like experiences, problematic substance use, internalizing symptoms, aggression) and general functioning (general well-being, delinquency, employment) from ages 20–24 years.

Methods

Participants

Data came from the Zurich Project on Social Development from Childhood to Adulthood (z-proso) (Ribeaud, Murray, Shanahan, Shanahan, & Eisner, 2022; Shanahan, Steinhoff, Ribeaud, & Eisner, 2024; z-proso Project Team, 2024). In 2004, 1675 children from 56 primary schools were recruited from the 90 public schools in Zurich using a stratified cluster random sampling approach. Among them, N = 1583 (94.5%) participated in at least one wave of data collection. Consistent with Zurich’s diverse population, parents of participants were born in more than 80 countries. In this analysis, data came from the eighth and ninth waves of data collection (in 2018 and 2022, respectively), when participants were 20 and 24 years old, respectively. In the eighth wave, 1002 participants provided hair samples (84.9% of the wave 8 sample). Of these, 915 also participated in wave 9. Participants included in the analyses were those who provided questionnaire and hair data at age 20, and questionnaire data at age 24, and who did not have missing data on the control variables (N = 863). Previous analyses in this sample found no differences in sex, parental migration, socioeconomic background, educational degree, or self-reported 3-month substance use between those who did and did not provide hair samples at age 20 (Steinhoff et al., 2023). For information on attrition patterns within the z-proso cohort, see Ribeaud et al. (2022) and Steinhoff et al. (2023).

Outcomes measured at ages 20 and 24

Psychopathology outcomes

Psychotic-like experiences were assessed using six items adapted from the Community Assessment of Psychic Experiences (Mark & Toulopoulou, 2016). Participants were asked how often they had experienced each symptom during the past month (e.g. ‘heard voices that no one else could hear’), from 1 = never to 5 = very often. Items were averaged (Cronbach’s α = .69).

Problematic (other) substance use was coded as 1 when participants reported either/or daily alcohol use, or any use of illicit substances in the previous 3 months (cocaine, ecstasy, non-medical use of benzodiazepine or opioid medications, 2C drugs, (meth)amphetamine, ketamine, hallucinogens) (Shanahan et al., 2021).

Internalizing symptoms were averaged from eight statements from the Social Behavior Questionnaire (SBQ) with answers on a 5-point Likert scale ranging from 1 = never to 5 = very often (Cronbach’s α = .86). Four questions captured anxiety symptoms (e.g. ‘I was worried’) and four questions captured depressive symptoms (e.g. ‘I was sad without knowing why’) (Murray, Obsuth, Eisner, & Ribeaud, 2019).

Aggression was measured using 11 items from the SBQ, capturing reactive, physical, and proactive aggression (e.g. ‘You intimidated others to get what you want’), from 1 = never to 5 = very often. Items were averaged (Cronbach’s α = .81) (Tremblay et al., 1991).

Functional outcomes

To measure general well-being, reflecting life satisfaction, participants rated four statements (e.g. ‘I am happy and content’, ‘Life is good’), on a 4-point Likert scale (from 1 = ‘fully untrue’ to 4 = ‘fully true’). A mean score was derived (Cronbach’s α = 0.83) (Murray, Nagin, Obsuth, Ribeaud, & Eisner, 2022).

Delinquency was a sum score of 21 self-reported delinquent acts in the previous 12 months (e.g. ‘stealing at work’, ‘fare dodging’) and captured non-drug-related activities, each coded as yes = 1 (Shanahan et al., 2021).

NEET (not in education, employment, or training) was coded as 1 if participants reported not being in education, employment, or training, and 0 otherwise.

Several of the outcomes measured here were previously investigated in association with frequent teenage cannabis use, measured via self-reports from ages 13 to 17 years in the same sample (Shanahan et al., 2021), but not at age 24 or in relation to hair data.

Indicators of cannabis exposure at age 20

Self-reports

Participants were asked how often they had consumed cannabis (e.g. hash, ‘pot’, marijuana) in the previous 3 months. Self-reported cannabis use was coded as 0 = never, 1 = once, 2 = 2–5 times, 3 = weekly, 4 = (almost) daily (Steinhoff et al., 2023).

Hair data

Trained interviewers collected hair samples from participants who gave at least 3 cm of hair which was proximal to the scalp. Scalp hair was collected (91%) when possible, when not possible, leg or arm hair were collected. THC and cannabinol (CBN) were quantified using liquid chromatography–tandem mass spectrometry. The analytical approaches used for analyzing cannabinoids are described elsewhere (Scholz et al., 2022). THC and CBN were summed to create an estimate of total THC (Huestis, 2007). This total score was log transformed due to high variance and positive skew. THC-COOH was not quantified, as this was not possible within the quantification pipeline at the time. Other psychoactive substances were quantified but not used in this study (Johnson-Ferguson et al., 2023; Steinhoff et al., 2023).

We tested binary predictors of cannabis use/exposure, which have been most common in studies using hair data. To test linear associations, we tested continuous indicators of cannabis exposure, which are more informative about dose–response associations. Finally, to assess whether hair data and self-reports complement each other, we tested composite scores. Except for the Likert scale of self-reported cannabis use, all operationalizations capture frequent cannabis use or high exposure.

Table 1 describes the six different variables that were derived from the hair data and self-reports of cannabis.

Table 1.

Six different operationalizations of cannabis exposure

	Method	Coding
Binary coding
Cannabis detected in hair	Hair	1 = yes
Self-reported frequent cannabis use (daily to weekly)	Self-report	1 = yes
Continuous/ordinal coding
Log of concentration of cannabis detected in hair	Hair	x = log([THC + CBN] + 1)
Likert scale of self-reported cannabis use	Self-report	0–4
Composite binary codes from self-reported and hair data
Self-reported frequent cannabis use (weekly–daily) and cannabis detected in hair	Hair and self-report	1 = yes
Self-reported frequent cannabis use (weekly–daily) and/or cannabis detected in hair	Hair and self-report	1 = yes

Note: Two derived from hair data, two derived from self-reports, and two representing composites of hair and self-report data (four binary variables and two continuous/ordinal variables in total).

Covariates

Sex was defined as sex assigned at birth and was collected from school records when participants were recruited.

Parental socioeconomic status was measured using the Socioeconomic Index of Occupational Status (Ganzeboom, De Graaf, & Treiman, 1992). This is an internationally comparable index of socioeconomic status based on occupation specific income and required educational level (range 16 [e.g. unskilled worker] to 90 [e.g. judge]). The score was coded from participants’ self-reports of their parents’ SES at ages 11, 13, and 15. For participants reporting their parents’ SES in more than one wave, the highest score was coded.

Migration background was coded as 1 = both parents born abroad. Participants were asked in which country their biological father and mother were born at ages 11, 13, and 15.

Analytic strategy

Multiple linear regression models and logistic regression models tested associations between cannabis use/exposure at age 20 and young adult psychopathology and functional outcomes at age 24, adjusting for the outcome at age 20, sex assigned at birth, socioeconomic status, and migration background. Each predictor variable (i.e. the six different cannabis use/exposure variables) was tested in a separate model to avoid collinearity. In order to facilitate interpretability of the results, continuous predictors in logistic regressions were standardized.

Analyses were conducted in R, using the stats package (version 3.6.2). Linear models were estimated using the lm() function. Logit models were estimated using the glm() function.

We applied false discovery rate (FDR) correction using the Benjamini–Hochberg method to address multiple comparisons (Benjamini & Hochberg, 1995), using the R function p.adjust (p_values, method = ‘BH’). This procedure controls the false discovery rate (FDR) in multiple hypothesis testing by ranking p values in ascending order, determining a q-value threshold for significance based on a desired FDR level, set here at 0.05.

Results

Sample description

At age 20, 351/863 participants (40.7%) reported any cannabis use in the previous 3 months; of these, 150 reported weekly-to-daily use (17.4% of the sample). One hundred ten participants had cannabis detected in hair, which is indicative of probable frequent use. For descriptive statistics, see Table 2.

Table 2.

Descriptive statistics of the outcomes examined at ages 20 and 24

		TotalN = 863
Outcomes at age 20
General well-being Range: 1–4	Mean (SD)	3.21 (0.58)
Delinquency Sum of 21 items	Mean (SD)	2.41 (1.99)
Not in education, employment, training	% (n)	3.00 (27)
Psychotic-like experiences Range: 1–5	Mean (SD)	1.47 (0.52)
Internalizing symptoms Range: 1–5	Mean (SD)	2.43 (0.83)
Aggression Range: 1–4	Mean (SD)	1.43 (0.39)
Outcomes at age 24
General well-being Range: 1–4	Mean (SD)	3.09 (0.6)
Delinquency Sum of 21 items	Mean (SD)	2.05 (1.6)
Not in education, employment, training	% (n/N)	2.80 (24/863)
Psychotic-like experiences Range: 1–5	Mean (SD)	1.40 (0.49)
Internalizing symptoms Range: 1–5	Mean (SD)	2.49 (0.82)
Aggression Range: 1–4	Mean (SD)	1.34 (0.36)
Demographic characteristics
ISEI	Mean (SD)	47.9 (19.84)
Both parents born abroad	% (n/N)	44.4 (383/863)
Female sex	% (n/N)	50.9 (439/863)
Cannabis use/ exposure at age 20
Detected in hair (pg/mg) (THC + CBN)	% positive (n/N)	12.8 (110/863)
	Mean (SD)	195 (403)
	Median	76.5
Self-reported cannabis use Never to daily use Range: 0–4	% > 0 (n/N)	40.7 (351/863)
	Mean (SD)	2.53 (1.14)
	Median	2
Self-reported frequent cannabis use Weekly-daily use only	% > 2 (n/N)	17.4 150/863

Note: ISEI=Socioeconomic Index of Occupational Status. Self-reported cannabis use: coded as 0 = never, 1 = once, 2 = 2–5 times, 3 = weekly, 4 = (almost) daily.

Cannabis use/exposure at age 20 predicts psychological and functional changes

Psychopathology

All cannabis predictors from age 20 were significantly associated with greater increases in psychotic-like experiences, internalizing symptoms, and aggression, and with higher odds of problematic substance use from ages 20 to 24 ( Table 3 ). The full table with exact p values is available in Supplementary Table S1. We conducted supplementary analyses, separating the internalizing items into either self-reported anxiety or depression symptoms. All cannabis predictors from age 20 were significantly associated with greater increases in depressive symptoms, while self-report and composite scores indexing cannabis exposure were significantly associated with greater increases in anxiety symptoms. However, these results should be interpreted with caution because the internalizing items of the SBQ has not been validated to investigate anxiety and depression separately. For more information, see Supplementary Table S2.

Table 3.

Associations of cannabis use/exposure at age 20 and young adult psychopathology and functional outcomes at age 24

	Psychotic-like experiences N = 862
	β	B	Lower bound	Upper bound
Binary coding
Cannabis detected in hair (1 = yes)	.089**	.131	.042	.219
Self-reported frequent cannabis use (weekly/daily) vs. occasional or no use	.105**	.135	.057	.213
Continuous/ordinal coding
Log of concentration of cannabis detected in hair Range: 0–8.11	.105**	.034	.014	.053
Likert scale self-reported cannabis use Range: 0–4	.127***	.045	.023	.066
Composite binary codes from self-reported and hair data
Self-reported frequent cannabis use and cannabis detected in hair (1 = yes)	.094**	.149	.053	.246
Self-reported frequent cannabis use and/or cannabis detected in hair	.103***	.126	.052	.201

Note: Associations of cannabis use/exposure at age 20 and young adult functional and psychopathology outcomes at age 24, adjusting for the outcome at age 20, sex assigned at birth, socioeconomic status, and migration background. Each row represents results from a multiple regression or logit regression model using one of the predictors indexing cannabis use (shown on the left). N = number of participants included in the models (i.e. those without missing data on either predictors or outcomes). Lower and upper bounds represent 95% confidence intervals for the unstandardized estimates. β coefficients are derived after standardizing both the predictor and outcome. OR = odds ratio. For logit regressions, upper and lower confidence intervals of the OR are reported, with the estimate reflecting the log odds of the outcome variable associated with a one-unit change in the predictor variable. Problematic substance use encompasses daily alcohol use and/or any illegal drug use in the past 3 months. The significance of bold values are indicated “the exact p values” included in the supplement Table S1.

^* p < .05,

^** p < .01,

^*** p < .001; adjusted using the Benjamini–Hochberg method.

Functioning

All cannabis-related predictors at age 20 were significantly associated with more decreases in general well-being at age 24. Among these, only self-reported frequency of cannabis use was significantly associated with greater increases in delinquency ( Table 3 ).

Notably, the composite measure of self-reported frequent cannabis use combined with cannabis detected in hair (1 = yes) was the only predictor not significantly associated with being ‘not in education, training, or employment’ (NEET) at age 24, whereas all five other operationalizations showed positive associations with this outcome.

Discussion and conclusion

Cannabis is widely used among young adults and is becoming increasingly accessible. This article investigated whether and how six different cannabis use/exposure variables, assessed with self-reports and hair, were associated with changes in psychopathology and functional outcomes in young adults. Adjusting for sex and sociodemographic variables, cannabis use at age 20 was associated with higher increases in psychopathology, problematic substance use, and more decreases in well-being to age 24, regardless of whether frequent cannabis exposure was measured with self-reports or in hair. These findings provide insights into subjective and objective assessment methods of cannabis in large community studies and their associations with changes in young adults’ well-being over several years.

Measuring exposure to cannabis

This study aimed to assess whether hair testing can be useful in detecting associations between cannabis exposure and psychopathology and functional outcomes. We found that cannabis use was robustly associated with changes in mental health in young adulthood across different operationalization, suggesting that either self-report or hair testing can be used to index cannabis use in young adults when investigating these associations. Unlike other psychoactive substances like cocaine, ketamine, and codeine, which demonstrate significantly higher sensitivity and specificity in hair analysis compared to self-reports (Janousch et al., 2025; Steinhoff et al., 2023), the effectiveness of using hair samples to measure cannabis exposure has been a matter of debate due to its limited sensitivity in detecting light or infrequent use (Taylor et al., 2017). However, as it can detect heavy use, which is most associated with risks of adverse outcomes across the literature, it may not be futile. Our findings stand in contrast to a previous study, conducted in a smaller convenience sample, maintaining that hair THC concentration is not able to predict psychopathological burden in people with near-daily cannabis use (Kroon et al., 2024).

Some participants did not report using cannabis, but cannabis was detected in their hair. It is unclear whether their self-reports were biased or whether this was due to external contamination of the hair samples as we did not measure a THC-specific metabolite such as THC-COOH (Moosmann, Roth, & Auwärter, 2016). Furthermore, some participants reported frequent cannabis use but did not have cannabis detected in their hair. Creating composite scores of participants who both reported frequent use and had cannabis detected in their hair, or who had one or the other, yielded very similar effect sizes for predicting outcomes as the individual hair/self-report predictors. Moreover, the hair data did not detect any associations that were not detected by self-reports. Therefore, the findings suggest that hair analysis is not superior to less expensive and less invasive self-reports for measuring cannabis exposure.

Cannabis use and associated changes in psychopathology within young adulthood

Consistent with the literature, all six cannabis use variables were associated with higher increases in psychotic-like experiences (Marconi et al., 2016). This proof of concept shows that THC variables derived from hair testing replicate associations previously identified with self-reported or diagnostic data, extending previous work on people with heavy cannabis use (Nestoros et al., 2017) in the general population at different cannabis use levels.

Our study found robust associations between cannabis use and increased odds of later other problematic substance use, aligning with previous research (Blanco et al., 2016; Degenhardt, Hall, & Lynskey, 2001; Shanahan et al., 2021; Silins et al., 2014). One explanation could be the ‘gateway hypothesis’, suggesting that cannabis use increase young people’s openness to also trying other substances (Kandel & Kandel, 2015; Wilson et al., 2022). Other explanations include shared risk factors for different substances, including genetic and/or environmental vulnerabilities or the general transfer of reward-seeking behavior from one substance to another (Degenhardt et al., 2001; Lüscher, Robbins, & Everitt, 2020).

The cannabis use/exposure variables were associated with higher increases in our score of general aggression. Cannabis use disorder has been linked to reactive aggression (Blair et al., 2021; Bortolato, Braccagni, Pederson, Floris, & Fite, 2024), which was included in our general score, but several studies found no association of frequent cannabis use in adolescence with physical aggression or the perpetration of violence (Loher et al., 2024; Shanahan et al., 2021). Overall, findings on the association between cannabis use and aggression have been mixed, perhaps due to heterogeneity in the measurement of aggression and the number and types of covariates included.

Cannabis use/exposure was robustly linked with internalizing symptoms. Several meta-analyses have shown inconsistent findings, with some indicating that only heavy, not recreational, cannabis use is associated with increased depression (Degenhardt, Hall, & Lynskey, 2003; Moore et al., 2007). Self-reported frequent teenage cannabis use between the ages of 13 and 17 was not associated with internalizing symptoms at age 20 in this sample after controlling for nearly 20 childhood covariates (Shanahan et al., 2021). However, the early to mid-20s might be a sensitive period for cannabis use to impact mental health, as the transition to adulthood creates new vulnerabilities. Furthermore, with peak use in young adulthood and higher frequencies and concentrations, there may be more power to detect an effect.

Cannabis use and associated changes in functioning within young adulthood

In all six models, cannabis use/exposure was associated with reduced well-being, reflecting participants’ life satisfaction. This supports limited findings that illegal drug use is negatively associated with life satisfaction in young adults (Kang, 2023), though research is scarce.

Cannabis was associated with delinquency in only one of the six models. As we adjusted for delinquency at age 20, this suggests that cannabis does not consistently predict changes in delinquency between the ages of 20 and 24. In other words, a possible interplay between cannabis use and delinquency, as found in other studies (Rocca, Verde, & Gatti, 2019; Shanahan et al., 2021), may have occurred to age 20, but not thereafter. As with other aggressive and antisocial behaviors, delinquency generally declines after the teenage years (Wittenberg, Reinecke, & Boers, 2009), and our participants had lower levels of delinquency at age 24 than at age 20.

Five of our six cannabis variables were associated with higher odds of not being in education, employment, or training at age 24. Notably, only a small group of young adults (n = 24) reported this outcome at age 24, which is consistent with the Swiss educational and training system, and the relatively low unemployment rate in Switzerland (Staatssekretariat für Wirtschaft [SECO], 2024). Nevertheless, the robust effects of this association should be of concern in a country that may soon legalize cannabis for 18+ year olds, as we know from the United States that such legalization may lead to more use and lower risk perceptions in this age group (Mennis et al., 2023; Patrick et al., 2023).

Study limitations

Our study has several limitations. First, the generalizability of results might be limited to young adults in urban Switzerland, where cannabis use has become a social norm. In populations where cannabis use is criminalized, hair testing might be superior in detecting cannabis-related declines in well-being, but this was not the case here. Second, while controlling for the outcomes of interest at age 20 did allow us to parcel out some between-persons variance and assess changes in these outcomes over 4 years, our analyses did not allow us to establish causation. Third, we did not consider the interplay between different outcomes. For example, some individuals at high risk of psychosis report using cannabis to treat symptoms such as anhedonia and anxiety (Gill et al., 2015). Furthermore, internalizing and externalizing symptoms are likely to co-occur and/or transition from one to the other (Murray et al., 2021). Moreover, psychotic-like experiences, internalizing symptoms, and problematic substance use are often comorbid (Staines et al., 2022). Fourth, our models did not control for a wide range of covariates, such as sensation seeking, which could explain some of the variance found (Shanahan et al., 2021). Additionally, our instruments do not have diagnostic cut-offs, and we cannot draw inferences about clinical diagnoses. Our scale for internalizing symptoms has not been validated for measuring anxiety and depression separately, which is a limitation as findings for these distinct psychopathologies have differed. Finally, our effect sizes were relatively small. However, this is to be expected with complex psychological phenomena (Götz, Gosling, & Rentfrow, 2022), and most effects were robust across different measures of cannabis use and exposure, allowing us to identify associations with some confidence.

Conclusion

Longitudinal investigations of the time windows in which cannabis may be a risk factor for worsening mental health and functioning are of utmost importance. Our results show that, first, hair data, as well as self-reports, can be used to detect associations between cannabis exposure and changes in various young adult outcomes. Second, after adjusting for several sociodemographic factors and symptom burden at baseline, young adults who use cannabis at age 20 showed higher increases in internalizing symptoms, psychotic-like experiences, aggression, and problematic substance use, and more decreases in general well-being from ages 20 to 24 years compared with participants without exposure to THC; self-report and hair data detected these associations equally well. Linear effects for both self-reported cannabis use and hair THC concentrations suggest a possible dose–response relationship. Further research is needed to identify those most at risk in the general population, given the possible increases in use and decreases in risk perception following the legalization of cannabis.

Supporting information

Johnson-Ferguson et al. supplementary material

Johnson-Ferguson et al. supplementary material

Supplementary material

The supplementary material for this article can be found at http://doi.org/10.1017/S003329172510144X.

Funding statement

This research was supported by the Swiss National Science Foundation (NFP-52 Grant 405240-69025, Grants 100013_116829, 100014_132124, 100014_149979, 10FI14_170409, 10FI14_198052, 10FI15_229647, 10531C_189008, 214979, 198052), the Federal Office of Public Health (Grants 2.001391, 8.000665), the Educational Directorate of the Canton of Zurich, the Swiss State Secretariat for Migration (previously IMES; Grant 03–901), the Federal Commission for Foreigners (Grant E-05-1076), the Julius Bär Foundation, the Foundation Visana Plus, the Foundation for Research in Science and the Humanities at the University of Zurich (Decision of the Board of Trustees: 3.02.2004), the Zürcher Kantonalbank, The University of Zurich’s Jacobs Center for Productive Youth Development, and the Jacobs Foundation (Grants 2007-720, 2010-888, 2013-1081-1).

Competing interests

The authors declare none.