Introduction

The use of psychoactive substances (e.g., cannabinoids, hallucinogens, stimulants, opioids, including their nonmedical use) is a threat to young people’s health (United Nations, 2018). Risks associated with substance use include physical, psychological, social, and functional impairments. These are multiplied when individuals consume two or more psychoactive substances simultaneously or sequentially, for example during the previous year (i.e., polysubstance use; World Health Organization, 1994). Indeed, compared to the use of a single substance, polysubstance use is associated with more dangerous patterns of substance use (e.g., addiction, overdose), physical health problems (e.g., injury), premature mortality, comorbid risk-taking (e.g., violence, dangerous driving), self-harming behaviors (e.g., suicidal behaviors), psychopathology (e.g., depressive symptoms), cognitive dysfunctions (e.g., impaired executive functions and empathy), and poorer educational and occupational achievements (European Monitoring Centre for Drugs and Drug Addiction, 2002; Conway et al., 2013; Connor et al., 2014; Kroll et al., 2018; Crummy et al., 2020). Encountering these consequences of polysubstance use during early adulthood could be especially harmful, as young people are expected to master important transitions in their educational, professional, social, and identity development (Arnett, 2000).

Community-based research on the developmental course of polysubstance use during adolescence and early adulthood is scarce. This is due, in part, to the lack of assessment of substances other than alcohol, tobacco, and cannabis before late adolescence in previous work (Connor et al., 2014). The first aim of our study was to examine the prevalence and stability of polysubstance use (defined here as the use of at least two psychoactive substances during the previous year) between early adolescence and early adulthood. We leveraged data from a representative, urban community sample (Ribeaud and Eisner, 2010; Shanahan et al., 2020; Ribeaud et al., 2021) with prospective longitudinal assessments from childhood to early adulthood, and substance use assessments beginning in early adolescence.

The different patterns of polysubstance use are also poorly understood. Research has begun to use person-centered analysis, including latent class analysis (LCA), to learn more (for reviews, see Connor et al., 2014; Tomczyk et al., 2016). These studies typically identified a group with no use, groups with limited to medium range use (e.g., use of a single substance, such as alcohol, or additional use of select other substances, especially cannabis), and a broad range group. The latter typically subsumed all users of illicit substances and those engaged in the non-medical use of prescription drugs (Connor et al., 2014). This is not surprising, given that the numbers of individuals using illicit substances were often too low in these studies to disaggregate this group further, or illicit substance use was assessed with summary items not differentiating between particular substances in the first place (Tomczyk et al., 2016; Carbonneau et al., 2021). Our second aim was to better understand heterogeneity within the polysubstance use group, which can only be done based on samples with high rates of substance use assessed with comprehensive substance use questionnaires, as is the case here (Quednow et al., 2021).

The most cost-effective approach to lowering the burden from polysubstance use would be to prevent and intervene before adolescents begin to engage in this pattern of use. Yet, we have a limited understanding of the childhood risk factors that predict (different patterns of) polysubstance use. Based on cross-sectional, retrospective, or short-term longitudinal data, prior research has identified individual-level (e.g., sensation-seeking, psychopathology) and social-environmental (e.g., exposure to others’ substance use, including in family and peer contexts) correlates of polysubstance use (Russell et al., 2015; Tomczyk et al., 2016; Tan et al., 2020; Carbonneau et al., 2021; Crane et al., 2021). The literature on childhood risk factors for any adolescent and early adulthood substance use also suggests putative predictors, including additional individual-level (e.g., self-control, risk-taking and externalizing behaviors), and social factors (e.g., social stress; e.g., Wills and Stoolmiller, 2002; Chapple et al., 2005; Barrett and Turner, 2006; Kelly et al., 2015). The third aim of our study was to identify childhood risk factors (i.e., precursors; Murray et al., 2009) of any polysubstance use and its different patterns in young adulthood.

Polysubstance use likely reflects different motivations for and instrumentalizations of substance use (Müller and Schumann, 2011; Valente et al., 2020). These include, for example, curiosity, craving social connectedness (Ter Bogt and Engels, 2005), enhancing one’s energy or ability to focus, calming and relaxing (LeClair et al., 2015), and enhancing or counteracting the (side) effects and withdrawal symptoms of other substances (Boys et al., 2001; Winstock et al., 2001; Licht et al., 2012). We assumed that some of these motivations may already be reflected in specific childhood precursors. Therefore, we hypothesized that different childhood individual-level (e.g., internalizing symptoms, sensation-seeking, risk-taking behaviors and delinquency) and social environmental factors (e.g., exposure to others’ substance use, social stressors) predict polysubstance use and its different patterns in young adulthood. For example, the inclination to experiment with new experiences and to take risks (e.g., indicated by sensation-seeking, offensive, and risky behaviors) and a low inhibition threshold (e.g., indicated by low self-control) could signal risk of any later polysubstance use and of a broad range of substances used in particular. In turn, childhood social stressors (e.g., victimization experiences) or internalizing symptoms could signal the risk of nonmedical use of prescription drugs for self-medication purposes. Finally, when children are exposed to others’ substance use, they may assume that substance use is common and safe. Thus, exposure to other’s substance use could precede polysubstance use in general. We examine all of these potential associations by making use of the long-term longitudinal design of our data.

Materials and Methods

Participants and Procedures

The data came from the longitudinal Zurich Project on the Social Development from Childhood to Adulthood (z-proso; Ribeaud and Eisner, 2010; Eisner et al., 2019; Ribeaud et al., 2021). Participants were selected using a cluster-stratified randomized sampling approach. In 2004, a sample of 1,675 children from 56 primary schools was randomly selected from 90 public schools in the city of Zurich, Switzerland’s largest city. Stratification was performed by considering the school sizes and socioeconomic backgrounds of the school districts. The sample was largely representative of first-graders attending public school in the city of Zurich. Participants were assessed eight times between 2004 (age 7) and 2018 (age 20).

The current study uses data collected at ages 13 (n = 1,365), 15 (n = 1,446), 17 (n = 1,306), and 20 (n = 1,180), to examine the developmental course of polysubstance use over time. To investigate different patterns of polysubstance use in early adulthood, and the associated developmental precursors (assessed before age 13), we included those who participated at age 20. Of these participants, 51% were male. Consistent with Switzerland’s immigration policies and the city’s diverse population, participants’ parents were born in over 80 different countries. The majority of participants were born in Switzerland (90%). The educational background of their parents was diverse; in 30% of households, at least one parent held a university degree. The mean (M) household occupational status, measured using the International Socio-Economic Index of Occupational Status (ISEI; Ganzeboom et al., 1992), was 47.1 [standard deviation (SD) = 19.7]. This internationally comparable index of socio-economic status was based on occupation-specific income and the required educational level with scores ranging from 16 (e.g., unskilled worker) to 90 (e.g., judge).

This study is consistent with national and international ethical standards and was approved by the relevant ethics committee. Adolescents provided written consent for their study participation, and parents of those aged 15 and younger could decline their child’s participation in the study. Data were collected from groups of 5–25 participants in classroom settings with paper-and-pencil questionnaires up to age 17 and in a computer laboratory setting with computer-administered surveys at age 20. Completing the surveys typically took about 90 minutes. Adolescents received a cash incentive for their participation, which increased from approximately $30 at age 13 to $75 at age 20.

Sample Attrition

In z-proso, the highest participation rate was reached at age 15 (n = 1,446). Of those who participated at age 15, females were more likely than males to participate again at age 20 (84.5% vs. 76.9%, p < 0.001). Those with at least one parent holding a university degree were more likely to participate at age 20 than those whose parents held a lower educational degree (95.0% vs. 79.4%, p < 0.001), and those with at least one Swiss-born parent were more likely to participate than those whose parents were both born abroad (83.9% vs. 77.9%, p = 0.004). Those who responded at age 20 had a higher adolescent family socio-economic status than those who had dropped out of the survey [ISEI score: M = 47.1 (SD = 19.7) vs. M = 40.4 (SD = 16.6), p < 0.001]. Further details on attrition can be found elsewhere (Eisner et al., 2019; Quednow et al., 2021). Such attrition patterns are common in long-term longitudinal research (e.g., Gustavson et al., 2012; Sigurdson et al., 2014; Steinhoff and Keller, 2020). Our handling of missing data is described in the “Analytic Strategy” section.

Variables

Substance Use at Age 20

Participants were asked how often they had used the following substances during the previous 12 months (exempting any use of medical drugs that were prescribed by a physician): (1) tobacco (e.g., cigarettes, shisha/hookah); (2) beer, wine, alcopops; (3) liquor (e.g., vodka, whisky, gin); (4) cannabinoids, including cannabis (e.g., hashish, grass, weed, marijuana, cannabis), cannabidiol (CBD; e.g., CBD-enriched hemp, cigarettes with CBD-enriched hemp, CBD tinctures), synthetic cannabinoids (i.e., cannabis substitutes such as “Dutch Orange,” “Spice,” “K2,” “Ganja Style”); (5) stimulants, including cocaine, and amphetamine/methamphetamine (e.g., “Speed”, “Pepp”, “Ice”, “Crystal Meth”); (6) empathogenes such as MDMA and its analogues (“Ecstasy,” “Molly”); (7) hallucinogens, including LSD/psilocybin (e.g., “Magic Mushrooms,” “Truffles”), 2C substances (e.g., “Bromo,” “Erox,” “Nexus,” “Venus”), and ketamine (“Special K,” “Vitamin K”); (8) opioids, including heroin and the nonmedical use of codeine-based cough medicine and opioid painkillers; (9) the nonmedical use of benzodiazepine tranquilizers; and (10) anabolic steroids. Assessments were made on a six-point scale (1 = never, 2 = once, 3 = two to five times, 4 = monthly, 5 = weekly, and 6 = daily).

Substance Use Between Ages 13 and 17

During adolescence, a limited range of substances was assessed. The list of substances was gradually expanded over time (age 13: alcohol, tobacco, and cannabis; ages 15 and 17: alcohol, tobacco, cannabis, MDMA, cocaine, amphetamine/methamphetamine, and LSD/psilocybin). Assessments of the frequency of use during the previous year were made on the same six-point scale that was used at age 20.

Coding of Polysubstance Use

We created dummy variables indicating whether participants had used specific substances at least once during the previous year. A sum score was computed, counting the number of different substances used. This score was then dichotomized to indicate any polysubstance use (i.e., at least two different substances used) vs. no polysubstance use (i.e., single or no substance use) for the analysis of the prevalence of polysubstance use. Furthermore, individuals were assigned to groups with polysubstance use, single substance use, and no use for the analysis of respective developmental precursors. For descriptive comparisons of the prevalence of polysubstance use over time, three different operationalizations of polysubstance use were applied. First, to compare the prevalence of polysubstance use between ages 13 and 20, we included alcohol, tobacco, and cannabis only because these were assessed during all assessments. Second, we computed an indicator of polysubstance use based on all five illicit substances assessed at ages 15, 17, and 20, excluding alcohol and tobacco. Third, an indicator of polysubstance use at age 20 included all substances (excluding alcohol and tobacco) that were assessed using the extended comprehensive questionnaire, which was administered for the first time at that age. This score mainly represents illicit substance use and non-medical use of prescription drugs. Although several CBD products are freely available in Switzerland, we included them in this score as well, because the effects of CBD are different from those of Δ⁹-tetrahydrocannabinol (THC), which is typically more dominant in cannabis (Freeman et al., 2019), and thus, the motivations underlying the use of CBD products vs. cannabis are likely also different (e.g., to relax vs. get high). Indeed, CBD itself is mildly psychoactive with sedative and anxiolytic effects, at least at moderate doses (Bergamaschi et al., 2011; Zuardi et al., 2017). To adjust the aggregate score of polysubstance use for potential overlap of using CBD products and cannabis, an additional sensitivity analysis of the prevalence of polysubstance use at age 20 was conducted, with CBD products being excluded.

For the identification of different polysubstance use patterns at age 20, the dummy variables of the different substances, excluding alcohol and tobacco, were used in subsequent LCAs (see the section on “Analytic Strategy”). For alcohol and tobacco, we created additional dummy variables indicating frequent alcohol and tobacco use (i.e., categories 5 = weekly and 6 = daily were coded 1, and less frequent or no use was coded 0). These were used as potential correlates of latent class membership (see the “Results” section for the underlying rationale).

Childhood Risk Factors

First, we included psychological factors and indicators of children’s functioning. All factors were self-reported at age 11, except sensation-seeking, which was a behavioral measure assessed at age 7. The descriptive statistics reported here refer to the age-20 participants.

• (a)

  Sensation-seeking: behavioral measure based on an adapted nine-item version of the Travel Game from Alsaker and Gutzwiller-Helfenfinfer (2010), see also Murray et al. (2020a,b), who reported an Omega reliability of 0.80; using a cardboard game, children’s preference for sensational vs. less sensational situations was assessed [e.g., “you must decide whether you want to travel by fast motorbike or funny steam locomotive” (0 = sensational situation not chosen, 1 = sensational situation chosen)]; a sum score was used; recoded scale 0–1 (M = 0.57, SD = 0.25);

• (b)

  Low self-control: 10 items (e.g., “I often act on the spur of the moment without stopping to think”) from Grasmick et al. (1993), Cronbach’s α = 0.75, scale 1 = fully untrue to 4 = fully true (M = 1.94, SD = 0.46);

• (c)

  Aggression: 15 items from the physical, proactive, indirect, reactive, and oppositional aggression subscales of the Social Behavior Questionnaire by Tremblay et al. (1991), α = 0.82; participants were asked to indicate how often during the previous 6 months they had engaged in particular aggressive behaviors (e.g., “physically attacked other people”); scale from 1 = never to 5 = very often (M = 1.48, SD = 0.37);

• (d)

  Internalizing symptoms: eight items from Tremblay et al. (1991); participants were asked how often during the previous month they had particular feelings (e.g., “I was scared, ” “I was sad without knowing why”); α = 0.79, scale from 1 = never to 5 = very often (M = 2.06, SD = 0.65);

• (e)

  Childhood onset of any substance use: three items asked about any previous use of alcohol, tobacco, and cannabis at age 11, which were combined and dichotomized to indicate any substance use—this scale differed from that used in subsequent substance use assessments (9% of the sample reported substance use at age 11);

• (f)

  Risky media use: three items assessed whether the participants had ever watched adult horror movies, adult action movies, or played adult computer games (e.g., “have you ever watched 18+ rated horror movies, that is to say, movies only meant for adults”; yes/no)—items were combined to indicate any use of adult media (42% of the sample reported risky media use at age 11);

• (g)

  Delinquency: nine items assessed particular behaviors during the previous year (e.g., “stolen something from a shop or kiosk that is worth more than 50 CHF” [yes/no]), and a sum score was created (M = 0.97, SD = 1.19).

Second, we included social-environmental factors, which were also self-reported at age 11, unless otherwise indicated:

• (a)

  Harsh parenting: five items from the Alabama Parenting Questionnaire (Shelton et al., 1996) indicating parents’ response when the child “misbehaves” or is “disobedient” (e.g., “do your parents spank you with their hand”, with scores from 1 = never to 4 = always) were combined and children with scores in the top quartile of the sample were assigned 1 = harsh parenting and compared to 0 = no harsh parenting (Shanahan et al., 2021), 21% of the sample were assigned 1;

• (b)

  Bullying victimization: four items from the Zurich Brief Bullying Scale (Murray et al., 2021); participants were asked to indicate how often in the previous 12 months others had, for example, “ignored or excluded you” or “laughed at, mocked, or insulted you”; α = 0.72, scale 1 = never to 6 = almost every day (M = 1.81, SD = 0.79);

• (c)

  Exposure to friends’ substance use: participants named their two best friends and reported whether these had used alcohol, tobacco, or other substances (e.g., cannabis) during the previous year [i.e., three items for each friend (yes/no)]; we created a binary variable indicating whether at least one friend had used any substances; 9% of the sample reported friends’ substance use at age 11;

• (d)

  Maternal substance use during pregnancy: mother-report, provided in the first assessment wave; three binary items assessing any use of alcohol, tobacco, and any other substances during pregnancy (yes/no)—items were combined to indicate 1 = any substance use during pregnancy vs. 0 = no use (37% of the sample had been exposed to maternal substance use).

Socio-demographics

We included children’s sex (0 = female, 1 = male), socio-economic background assessed as ISEI (Ganzeboom et al., 1992), and parental migration background (0 = at least one Swiss born parent, 1 = both parents born abroad).

Analytic Strategy

First, we calculated and compared the prevalence of any polysubstance use between early adolescence (age 13) and early adulthood (age 20) in an effort to identify typical developmental periods of onset. The maintenance of polysubstance use over time was examined by testing the stability of polysubstance use between adjacent assessments, using binary logistic regression models (i.e., polysubstance use at one assessment was regressed on polysubstance use at the respective previous assessment).

Second, we applied LCA, to identify clusters of participants with different patterns of polysubstance use in early adulthood. In this step of the analysis, we included all participants who reported the use of at least two substances other than alcohol and tobacco at age 20. Our decision on the optimal number of classes was based on relative fit indices [Bayesian Information Criterion (BIC) and Akaike’s Information Criterion (AIC)] and entropy as an indicator of classification precision. Lower BIC and AIC were considered indicative of better model fit; entropy >0.80 and class counts >5% were considered indicative of model accuracy (Muthen, 2004). Furthermore, solutions with varying numbers of classes were inspected graphically for conceptual interpretability (Masyn, 2013). Finally, we tested associations between childhood precursors and individuals’ polysubstance use status (i.e., any polysubstance use vs. no use and single substance use) and their most likely class membership using nominal logistic regression analyses.

In the LCA, missing data was accounted for by applying full information maximum likelihood estimation; in the regression models, we used multiple imputations to handle missing data on predictor variables (Schafer and Graham, 2002; Enders, 2013). For each model that we present, all variables included in that model were involved in the imputation model, and 20 imputed data sets were generated. Parameter estimates were averaged across the imputed data sets. Based on these procedures, we were able to include all age-20 participants (n = 1,180) in the analyses of the precursors of polysubstance use status and all age-20 participants with any polysubstance use (n = 420) in the analyses of the precursors of polysubstance use patterns. Descriptive analyses and tests of stability over time were conducted using SPSS V25; all other analyses were conducted using Mplus V8.

Results

Prevalence and Stability of Polysubstance Use Between Early Adolescence and Early Adulthood

The prevalence of past-year polysubstance use increased between early adolescence and early adulthood (Figure 1). According to our first operationalization (i.e., alcohol, tobacco, cannabis), the increase was especially sharp between ages 13, when one in five adolescents had used at least two of the three substances in the previous year, and 17, when more than two-thirds of adolescents had used at least two of the three. At age 20, more than three-quarters of early adults reported past-year polysubstance use.

Figure 1.

_{Prevalence of polysubstance use between ages 13 and 20 years. Note. The prevalence of polysubstance use at age 20 was 25% when cannabidiol (CBD) was excluded.}

According to our second operationalization (i.e., substances other than alcohol and tobacco, using the limited questionnaire administered starting in mid-adolescence), past-year polysubstance use became increasingly prevalent during late adolescence. At age 17, 1 in 14 adolescents reported illicit polysubstance use; this number increased to one in six at age 20. Importantly, this latter number was more than doubled when using the third operationalization (i.e., extended, comprehensive questionnaire used at age 20 only). Specifically, almost one in three early adults reported polysubstance use according to this operationalization. While the previous numbers are useful for comparison purposes, this latter number, being based on the comprehensive questionnaire, is the most reliable reflection of the true polysubstance use prevalence in our sample. Therefore, our comparative descriptive analyses show that prevalence rates based on narrow assessments of only a small selection of different substances are likely to severely underestimate the prevalence of polysubstance use in a population.

A sensitivity analysis excluding CBD from the age-20 polysubstance use score revealed a polysubstance use prevalence of 25%, reflecting that use of CBD largely overlapped with the use of cannabis (97% of those reporting the use of CBD products also reported cannabis use; 46% of those reporting cannabis use also reported the use of CBD products).

At age 20, the number of different substances used ranged from 0 to 13 (Figure 2), and the average number of substances used among those with any polysubstance use (excluding alcohol and tobacco) was 3.53 (SD = 1.98). This count did not differ between males and females with any polysubstance use (p = 0.381).

Figure 2.

_{Prevalence of counts of substances (based on extended questionnaire and excluding alcohol and tobacco) used during the previous year at age 20. Note. Categories nine and higher were combined due to low prevalence.}

The stability of polysubstance use over time was high, including when the alcohol-tobacco-cannabis coding was used [from age 13 to age 15: odds ratio (OR) = 7.55, 95% confidence interval (CI) = 5.22–10.92; from 15 to 17: OR = 13.52, 95% CI = 9.77–18.71; from 17 to 20: OR = 16.66, 95% CI = 11.87–23.38] and when the narrow coding involving illicit substances and excluding alcohol and tobacco was used (age 15–17: OR = 23.89, 95% CI = 11.35–50.30; age 17–20: OR = 14.78, 95% CI = 8.78–24.88). These findings show that, once initiated, polysubstance use is often continued over time.

Patterns of Polysubstance Use in Early Adulthood

Our investigation of different patterns of polysubstance use in early adulthood focused on illicit substance use, legal drugs other than alcohol and tobacco (in Switzerland, this includes CBD), and the nonmedical use of prescription drugs. We excluded any past-year use of alcohol and tobacco because the latter two were so prevalent in our sample (Quednow et al., 2021) that they would not differ much among the latent classes. However, after identifying the different latent classes, we also examined the prevalence of frequent (i.e., weekly or daily) consumption of alcohol and tobacco within these classes.

The LCA included all participants who reported the use of at least two substances other than alcohol and tobacco during the previous year (36% of the sample, n = 420). Heroin and anabolic steroids were excluded from this analysis due to their very low prevalence (n < 5). Based on a comparison of relative fit indices (Table 1), a three-, a four-, and a five-class solution were selected for further inspection of interpretability and class sizes. The four-class solution, with the lowest BIC, revealed distinct classes with substantially different substance use profiles (see Figure 3) and reasonable prevalence (>10% each). Because this met our criteria for the best solution, we chose the four-class solution for further analysis.

Table 1. Model fit and precision of latent class solutions with one to five classes.

_{Note. Italics indicate the solution chosen for further analyses. BIC, Bayesian Information Criterion; AIC, Akaike’s Information Criterion.}

Figure 3.

_{Polysubstance use profiles at age 20. Note. Prevalence based on estimated model.}

The following four classes were identified (for the prevalence of each class, see Figure 3):

• (1)

  Class 1, “broad spectrum,” included participants who used many substances (e.g., cannabinoids, stimulants, hallucinogens, and opioids)—although this was the smallest class, more than one in 10 early adults with polysubstance use belonged to it;

• (2)

  Class 2, “cannabinoids and club drugs, ” was characterized by the use of fewer substances than class 1, involving cannabinoids plus stimulants, empathogenes, and hallucinogens that are typically consumed in party contexts, and, in some cases, also codeine—one-third of early adults with polysubstance use belonged to this class,

• (3)

  Class 3, “cannabis and medication, ” was also characterized by medium-range substance use but primarily involved cannabinoid use plus nonmedical use of prescription drugs, including those with opioid and tranquilizers—one in eight early adults with polysubstance use belonged to this class;

• (4)

  Class 4, “cannabinoids, ” was primarily characterized by the use of different cannabinoids. Together with class 2, this class comprised the largest group: more than one-third of the sample with polysubstance use was in it.

To further characterize the classes, we investigated their association with weekly or daily alcohol and tobacco use. The likelihood of drinking alcohol frequently was higher among classes 1 [64%, standard error (SE) = 8.2], 2 (61%, SE = 4.4), and 4 (47%, SE = 4.2) than among class 3 (10%, SE = 5.4). The comparisons between class 3 and all other classes were significant (p < 0.001). The difference between classes 2 and 4 was also significant (p = 0.022). Smoking tobacco frequently was more prevalent among members of classes 1 (81%, SE = 6.9) and 2 (69%, SE = 4.2) compared to those of classes 3 (39%, SE = 7.3; p < 0.001 and p = 0.001 for comparison with classes 1 and 2, respectively) and 4 (50%, SE = 4.2; p < 0.001, p = 0.002, respectively). Altogether, these findings show that the spectrum of substances used by members of classes 1 and 2 was supplemented by both frequent alcohol and tobacco use, whereas substances used by classes 3 and 4 were more selective.

Risk Factors for Polysubstance Use

First, we investigated the socio-demographic and childhood precursors of any polysubstance use compared to single and no substance use at age 20. Notably, the majority of individuals reporting single substance use reported cannabis use (n = 259), and only a minority (n = 40) reported the use of another substance as their only drug of choice (excluding alcohol and tobacco). We specified separate models for each precursor, adjusting for socio-demographics, and a multivariable model including all precursors in one model (i.e., a “full model”).

The results revealed precursors of polysubstance use from all three domains: (1) socio-demographics; (2) individual-level factors (i.e., psychological factors and indicators of functioning); and (3) social-environmental factors (Table 2). Although male participants were more likely than females to report polysubstance use rather than single or no substance use, the sex difference was not significant when individual and social-environmental factors were included in the model simultaneously (i.e., the full model). A higher socio-economic background was associated with an increased risk of substance use (i.e., poly- or single use compared to no use). Children with two Swiss parents had a higher risk than those with a parental migration background to engage in polysubstance use compared to single or no substance use. In addition, childhood-onset of risk-taking behaviors, such as substance use and risky media use, was associated with an increased risk of substance use (i.e., single or poly, respectively) in early adulthood compared to no use. Childhood sensation-seeking and exposure to maternal substance use during pregnancy were also associated with an increased risk of polysubstance use compared to no use. Substance use by friends during childhood was uniquely associated with a higher risk of later polysubstance use compared to single substance use.

Table 2. Associations between potential risk factors and early adulthood substance use status (nominal logistic regressions: OR, 95% CI).

Table 3. Associations between potential risk factors and latent class membership, adjusted for socio-demographics (nominal logistic regressions: OR, 95% CI).

Discussion

Our investigation revealed that polysubstance use increases between early adolescence and early adulthood and is often sustained over time. In our urban community sample with high levels of substance use, the range of different substances used simultaneously or consecutively during the previous year was wider, and patterns of substances combined were more heterogeneous than most prior research had suggested. Several socio-demographic factors and childhood precursors signal individuals’ risk of polysubstance use and differentiate, in part, among diverse patterns of polysubstance use reported in early adulthood.

Our data show that polysubstance use is highly prevalent in young people from this urban community. Even when alcohol and tobacco were excluded from the analyses, polysubstance use as defined here was identified in more than one-third of early adults in our sample when CBD was included; and in one out of four young adults when CBD was excluded. Indeed, the prevalence of any substance use is high in urban Switzerland compared to international evidence (Quednow et al., 2021; Shanahan et al., 2021). This might, in part, be a consequence of relatively high drug availability. However, our comparisons of the prevalence rates based on different operationalizations of polysubstance use, using limited and more extensive questionnaires, indicate that prevalence rates from other studies that assessed only a few different substances likely substantially underestimate the prevalence of polysubstance use in youth. To assess the true prevalence of polysubstance use in a community, it is imperative to administer comprehensive lists of substances available on the local market and to consider the potential nonmedical use of prescription drugs.

Our prospective longitudinal study design with substance use assessments that started in early adolescence provides novel evidence of adolescence as an onset period of polysubstance use. At ages 15 and 17, many adolescents may be experimenting with different substances and use them only once. However, the increasing prevalence of polysubstance use over time and the high odds of polysubstance use continuation indicate that the initiation of polysubstance use in adolescence is a major risk factor for prolonged polysubstance use and perhaps also for the progression to increasingly risky patterns of use (Trenz et al., 2012; Olthuis et al., 2013). Like a vicious cycle, the high and increasing prevalence of polysubstance use in mid- and late-adolescence can, in and of itself, also be a risk factor for (more) individuals to engage in polysubstance use, because exposure to peers’ polysubstance use could increase peer pressure and misjudgments about the dangers associated with polysubstance use (Willis et al., 2019). Indeed, our findings show that exposure to peers’ substance use is a unique developmental precursor of one’s own subsequent engagement in polysubstance use.

Altogether, these findings underscore the importance of awareness and health education campaigns and tailored polysubstance use prevention programs targeting youth before they reach mid-adolescence. Given the high prevalence and potentially severe consequences of chronic polysubstance use, parents and professionals (e.g., pediatricians) should also be made aware of these issues. Our results show that childhood sensation-seeking, childhood onset of risk-taking behaviors, and premature exposure to others’ substance use were uniquely associated with an increased risk of early adulthood polysubstance use. Therefore, all these factors represent promising targets or markers for the need of early prevention mechanisms tailored to counteract early adulthood polysubstance use in general.

Notably, the group of early adults engaging in polysubstance use was heterogeneous, both in terms of the combinations of substances used and some of the associated risk factors. All groups were characterized by the consumption of cannabis, but they differed in the additional substances used, which is a common finding in similar research (Quek et al., 2013; Connor et al., 2014). Specifically, the use of club drugs and the use of different cannabinoids (especially cannabis and CBD products) were the most prevalent patterns of polysubstance use in our Zurich-based sample. However, the numbers of early adults reporting nonmedical use of prescription drugs in addition to cannabis use, or the use of a broad range of all different kinds of substances, were also considerable. These insights extend the international evidence on latent classes of polysubstance use by adding Swiss data, which had previously been missing (Connor et al., 2014; Tomczyk et al., 2016).

The different polysubstance use classes may reflect different contexts of and motivations for substance use (Valente et al., 2020). For example, using a broad spectrum of substances (class 1) is likely associated with a motivation to experiment with new experiences (including substance use) and a lifestyle that comes with frequent opportunities to try different substances (e.g., in nightlife contexts). The use of cannabinoids plus medical drugs with opioids (class 3) may represent an attempt to self-medicate among some individuals, to ease pain or to calm down and raise one’s mood for those suffering from anxiety or depressive symptoms (Blume et al., 2000; Shehnaz et al., 2014). However, other individuals in class 3 may simply resort to medication instead of using illicit substances because they perceive medical drugs as less harmful, dangerous, and illegal or because these substances are more easily accessible to them.

Our findings of the associations between risk factors and the different classes show that prevention programs may need to address groups of children with different challenges in different ways. For example, the consistent association between low childhood self-control and membership in the broad spectrum class may reflect these individuals’ low inhibition threshold when facing opportunities to try new substances. Thus, those on track towards experimentation with a broad spectrum of substances may need support in channeling low self-control into less risky behaviors. This prevention may not be particularly relevant for those on track toward other polysubstance use patterns.

However, other childhood factors were not differentially associated with particular polysubstance use patterns. For example, the classes did not differ in terms of childhood internalizing symptoms, although prior research based on cross-sectional data found associations between depressive symptoms and a latent polysubstance use class involving, among others, the use of medication to get high (Conway et al., 2013). Together, ours and prior evidence suggest that differences between the classes in the realm of internalizing symptoms, if any, might have developed only during adolescence and early adulthood.

In fact, differences between the cannabis and medication class (class 3) and other classes mainly pertained to socio-demographic factors in our study. Previous research has found that the female gender is associated with a higher likelihood of self-medication (Shehnaz et al., 2014), and indeed, females with polysubstance use in our sample were relatively likely to be in class 3. In addition, parental migration background increased the likelihood of being in the medication class compared to select other classes. Together, these findings indicate that early adults who represent the typically less privileged members of society (especially females and those with migrant backgrounds) tend to engage in more covert substance use than the more privileged groups (especially males and those with native Swiss parents) who tend to engage in more overt substance use indicated by the broad spectrum (class 1) and club drugs (class 2) classes.

Importantly, the time frame we used to define polysubstance use (i.e., the previous year) is common in polysubstance use research (Conway et al., 2013; Connor et al., 2014) but does not allow us to distinguish between individuals who take different substances sequentially (e.g., to counteract other substance effects or, more independently, based on different motivations and at different occasions) vs. those who consume them simultaneously (e.g., to enhance specific substance effects). However, previous research has shown that the simultaneous use of two or more substances is common among early adults with any past-year polysubstance use, and simultaneous use of alcohol or cannabis with other substances is especially frequent (Quek et al., 2013). Indeed, our follow-up analyses of associations between class membership and alcohol and tobacco use indicated that in some of the classes, frequent (i.e., weekly or daily) use of alcohol and tobacco was common. In turn, weekly or even daily consumption of alcohol or tobacco necessarily implies that any other (combinations of) substances must have been used shortly before or after drinking and smoking or at the same time.

Our study has some limitations. First, substance use was self-reported and could have been underestimated due to social desirability. However, the high rates of substance use reported here suggest that underreporting was not a serious issue. Second, although our list of potential childhood precursors of polysubstance use was comprehensive, some potentially relevant factors were not assessed, such as childhood trauma (Martinotti et al., 2009; Tonmyr et al., 2010; Armour et al., 2014; Davis et al., 2021). Third, although the sample was largely representative of young people growing up in the city of Zurich, it is unclear whether the findings are generalizable to the entire Swiss or international populations. It is likely that some of the substance use characteristics in our sample are Zurich-specific (e.g., high prevalence of cocaine, codeine, and CBD), but that the predictors for polysubstance use are generalizable. Finally, our definition of polysubstance use does not distinguish young people who tried each substance only once during the previous year from those who used different substances regularly. Limiting the concept of polysubstance use to regular use of different substances would likely result in a lower prevalence of polysubstance use, and perhaps also different associations between polysubstance use and specific risk factors. More refined assessments of polysubstance use including information on the frequency of use are needed to better understand different polysubstance use patterns, their prevalence, and precursors.

Our study also had important strengths for advancing scientific knowledge about polysubstance use patterns and their developmental precursors. These include a largely representative, prospective longitudinal study design, a high-resolution assessment of early adulthood substance use, and a sample characterized by a high overall prevalence of substance use. The latter facilitated the detection of several patterns of polysubstance use based on advanced statistical methods and their specific developmental correlates. Notably, prior research with adolescents and early adults often identified one polysubstance use class and compared it to single or no substance use classes (Tomczyk et al., 2016; Choi et al., 2018). Our population with high rates of substance use, our consideration of the nonmedical use of prescription drugs, and our use of LCA on only those who exhibited polysubstance use may have contributed to the identification of multiple distinct polysubstance use groups in our sample. Findings regarding different associations among childhood precursors with these classes illustrate that such nuanced insights into polysubstance use patterns are important for developing tailored prevention programs.

In summary, our investigation provides evidence that polysubstance use is a multifaceted phenomenon that affects a considerable proportion of early adults from the community, and this, in part, follows developmental processes that begin in childhood. To adequately assess polysubstance use and its development across an individual’s life span, future studies need to implement comprehensive assessments of the various substances available on local markets. The heterogeneity of polysubstance use and the specific socio-demographic and developmental factors associated with different patterns of substances used need to be more carefully considered in future research to facilitate the design of tailored prevention mechanisms and curb the individual and social burden that often follows polysubstance use.

Introduction

The consumption of psychoactive substances, including nonmedical use of opioids, stimulants, hallucinogens, and cannabinoids, poses a threat to the health of young people. The risks linked to substance use include physical, psychological, social, and functional problems. These risks are amplified when individuals use two or more psychoactive substances at the same time or at different times, often referred to as polysubstance use. Compared to using a single substance, polysubstance use is associated with more dangerous patterns of use, such as addiction and overdose, physical health issues, earlier death, higher rates of risky behaviors like violence and dangerous driving, self-harm, mental health conditions like depression, and cognitive problems, along with poorer achievements in education and careers. Experiencing these consequences in early adulthood can be particularly damaging, as young people are navigating important transitions in their personal, educational, professional, and social development.

Community-based research on the progression of polysubstance use during adolescence and early adulthood has been limited. This is partly due to past studies often not assessing substances beyond alcohol, tobacco, and cannabis until later adolescence. The initial goal of this study was to investigate how common polysubstance use (defined as using at least two psychoactive substances in the previous year) is and how stable it remains from early adolescence through early adulthood. Data from a representative urban community sample with long-term assessments from childhood to early adulthood, including substance use information starting in early adolescence, were utilized for this purpose.

The various patterns of polysubstance use are also not well understood. Research has begun to use person-centered analysis, such as latent class analysis, to gain more insight. These studies typically identify groups with no substance use, groups with limited to moderate use (e.g., only alcohol, or alcohol plus cannabis), and a broad range group. The broad range group often includes all users of illicit substances and those engaging in the non-medical use of prescription drugs. This is common because studies often have too few individuals using illicit substances to divide this group further, or illicit substance use is measured with summary items that do not distinguish between specific substances. A second aim of this study was to better understand the diversity within the polysubstance use group, which requires samples with high rates of substance use assessed with comprehensive questionnaires.

An effective way to reduce the impact of polysubstance use is to prevent it before adolescents adopt this pattern of use. However, there is a limited understanding of the childhood risk factors that predict different patterns of polysubstance use. Previous research, using various types of data, has identified individual factors (e.g., sensation-seeking, mental health issues) and social factors (e.g., exposure to others' substance use in family or peer settings) linked to polysubstance use. The existing literature on childhood risk factors for any substance use in adolescence and early adulthood also suggests potential predictors, including additional individual factors (e.g., self-control, risk-taking, and externalizing behaviors) and social factors (e.g., social stress). The third aim of this study was to identify childhood risk factors, or precursors, for any polysubstance use and its various patterns in young adulthood.

Polysubstance use likely stems from different reasons and serves various purposes. These reasons can include curiosity, a desire for social connection, enhancing energy or focus, seeking calm and relaxation, or boosting or counteracting the effects or withdrawal symptoms of other substances. It was assumed that some of these motivations might be reflected in specific childhood precursors. Therefore, the hypothesis was that different individual childhood factors (e.g., internalizing symptoms, sensation-seeking, risk-taking behaviors, and delinquency) and social environmental factors (e.g., exposure to others’ substance use, social stressors) predict polysubstance use and its various patterns in young adulthood. For instance, a tendency to seek new experiences and take risks, or a low inhibition threshold, could indicate a risk for later polysubstance use, particularly a broad range of substances. Conversely, childhood social stressors or internalizing symptoms might indicate a risk for nonmedical use of prescription drugs for self-medication. Finally, when children are exposed to others' substance use, they may perceive substance use as normal and safe, which could lead to polysubstance use generally. These potential associations were explored using the long-term longitudinal design of the study data.

Materials and Methods

Participants and Procedures

The data for this study were drawn from the longitudinal Zurich Project on the Social Development from Childhood to Adulthood, also known as z-proso. Participants were chosen using a cluster-stratified randomized sampling method. In 2004, a sample of 1,675 children from 56 primary schools was randomly selected from 90 public schools in Zurich, Switzerland’s largest city. The sampling process considered school sizes and the socioeconomic backgrounds of school districts. The resulting sample was largely representative of first-graders attending public school in the city of Zurich. Participants were assessed eight times between 2004 (age 7) and 2018 (age 20).

The current study used data collected when participants were aged 13 (n = 1,365), 15 (n = 1,446), 17 (n = 1,306), and 20 (n = 1,180) to examine the progression of polysubstance use over time. To investigate different patterns of polysubstance use in early adulthood and associated developmental precursors (assessed before age 13), participants who took part at age 20 were included. Of these participants, 51% were male. Consistent with Switzerland’s immigration policies and Zurich’s diverse population, the parents of participants were born in over 80 different countries. Most participants (90%) were born in Switzerland. The educational background of their parents varied; in 30% of households, at least one parent held a university degree. The average household occupational status, measured using the International Socio-Economic Index of Occupational Status (ISEI), was 47.1 (standard deviation = 19.7). This internationally comparable index of socioeconomic status is based on occupation-specific income and required educational levels, with scores ranging from 16 (e.g., unskilled worker) to 90 (e.g., judge).

This study adhered to national and international ethical standards and received approval from the relevant ethics committee. Adolescents provided written consent for their study participation, and parents of those aged 15 and younger could decline their child’s participation. Data were collected from groups of 5–25 participants in classroom settings using paper-and-pencil questionnaires up to age 17. At age 20, data were collected in a computer laboratory setting with computer-administered surveys. Completing the surveys typically took about 90 minutes. Adolescents received a cash incentive for their participation, which increased from approximately $30 at age 13 to $75 at age 20.

Sample Attrition

The z-proso study achieved its highest participation rate at age 15 (n = 1,446). Of those who participated at age 15, females were more likely than males to participate again at age 20 (84.5% versus 76.9%). Participants with at least one parent holding a university degree were more likely to participate at age 20 than those whose parents had a lower educational degree (95.0% versus 79.4%). Similarly, participants with at least one Swiss-born parent were more likely to participate than those whose parents were both born abroad (83.9% versus 77.9%). Individuals who responded at age 20 had a higher adolescent family socioeconomic status than those who had dropped out of the survey (average ISEI score: 47.1 versus 40.4). Such attrition patterns are common in long-term longitudinal research. The handling of missing data is detailed in the "Analytic Strategy" section.

Variables

Substance Use at Age 20 Participants reported how often they had used the following substances during the previous 12 months (excluding medically prescribed drugs): (1) tobacco; (2) beer, wine, alcopops; (3) liquor; (4) cannabinoids (including cannabis, cannabidiol (CBD), synthetic cannabinoids); (5) stimulants (including cocaine, amphetamine/methamphetamine); (6) empathogens (e.g., MDMA); (7) hallucinogens (including LSD/psilocybin, 2C substances, ketamine); (8) opioids (including heroin and nonmedical use of codeine-based cough medicine and opioid painkillers); (9) nonmedical use of benzodiazepine tranquilizers; and (10) anabolic steroids. Responses were on a six-point scale from 1 (never) to 6 (daily).

Substance Use Between Ages 13 and 17 During adolescence, a more limited range of substances was assessed, which gradually expanded over time. At age 13, alcohol, tobacco, and cannabis were assessed. At ages 15 and 17, alcohol, tobacco, cannabis, MDMA, cocaine, amphetamine/methamphetamine, and LSD/psilocybin were included. The frequency of use during the previous year was assessed using the same six-point scale as at age 20.

Coding of Polysubstance Use Binary variables were created to indicate whether participants had used specific substances at least once in the previous year. A sum score was computed by counting the number of different substances used. This score was then converted into a binary variable to indicate any polysubstance use (at least two different substances used) versus no polysubstance use (single or no substance use) for analyzing prevalence. For analyzing developmental precursors, individuals were assigned to groups of polysubstance use, single substance use, and no use.

For descriptive comparisons of polysubstance use prevalence over time, three different definitions were used. First, to compare prevalence between ages 13 and 20, only alcohol, tobacco, and cannabis were included because these were assessed at all time points. Second, an indicator of polysubstance use based on all five illicit substances assessed at ages 15, 17, and 20, excluding alcohol and tobacco, was computed. Third, an indicator of polysubstance use at age 20 included all substances (excluding alcohol and tobacco) from the extended comprehensive questionnaire, which was administered for the first time at that age. This third score primarily represents illicit substance use and non-medical use of prescription drugs. Although several CBD products are legally available in Switzerland, they were included in this score because CBD’s effects differ from those of THC, and motivations for using CBD products versus cannabis are likely different. CBD itself is mildly psychoactive, with sedative and anxiety-reducing effects at moderate doses. To account for potential overlap between CBD product use and cannabis use, an additional sensitivity analysis of polysubstance use prevalence at age 20 was conducted, excluding CBD products. For identifying different polysubstance use patterns at age 20, binary variables for various substances, excluding alcohol and tobacco, were used in subsequent latent class analyses. For alcohol and tobacco, additional binary variables were created to indicate frequent use (weekly or daily). These were used as potential correlates of latent class membership.

Childhood Risk Factors The study included psychological factors and indicators of children’s functioning, all self-reported at age 11, except for sensation-seeking, which was a behavioral measure assessed at age 7.

• Sensation-seeking: A behavioral measure based on an adapted version of the Travel Game, assessing children’s preference for sensational versus less sensational situations.

• Low self-control: Measured by items indicating impulsive behavior.

• Aggression: Assessed through items on physical, proactive, indirect, reactive, and oppositional aggressive behaviors.

• Internalizing symptoms: Measured by items reflecting feelings such as fear or sadness without a clear reason.

• Childhood onset of any substance use: A binary variable combining reported use of alcohol, tobacco, or cannabis at age 11.

• Risky media use: A binary variable indicating any reported viewing of adult horror or action movies, or playing adult computer games.

• Delinquency: A sum score of reported behaviors such as theft.

Social-environmental factors, also self-reported at age 11 unless otherwise stated, included:

• Harsh parenting: A binary variable indicating whether parents used harsh responses to misbehavior, based on the top quartile of reported scores.

• Bullying victimization: Measured by items on how often others ignored, excluded, laughed at, mocked, or insulted the participant.

• Exposure to friends’ substance use: A binary variable indicating whether at least one of the participant's two best friends had used alcohol, tobacco, or other substances in the previous year.

• Maternal substance use during pregnancy: A mother-reported binary variable indicating any use of alcohol, tobacco, or other substances during pregnancy.

Socio-demographics Included were the child’s sex (female or male), socio-economic background (measured by ISEI), and parental migration background (at least one Swiss-born parent vs. both parents born abroad).

Analytic Strategy

First, the prevalence of any polysubstance use between early adolescence (age 13) and early adulthood (age 20) was calculated and compared to identify typical onset periods. The persistence of polysubstance use over time was examined by testing its stability between consecutive assessments using binary logistic regression models.

Second, latent class analysis (LCA) was applied to identify groups of participants with distinct polysubstance use patterns in early adulthood. This part of the analysis included all participants who reported using at least two substances other than alcohol and tobacco at age 20. The optimal number of classes was determined based on relative fit indices (Bayesian Information Criterion (BIC) and Akaike’s Information Criterion (AIC)) and entropy, which indicates classification precision. Lower BIC and AIC values suggested a better model fit, while entropy above 0.80 and class counts greater than 5% indicated model accuracy. Additionally, solutions with varying numbers of classes were visually inspected for conceptual interpretability. Finally, the relationships between childhood precursors and individuals’ polysubstance use status (any polysubstance use versus no use and single substance use) and their most likely class membership were tested using nominal logistic regression analyses.

In the LCA, missing data were addressed by using full information maximum likelihood estimation. In the regression models, multiple imputations were used to handle missing data on predictor variables. For each model presented, all variables included in that model were part of the imputation model, and 20 imputed data sets were generated. Parameter estimates were averaged across these imputed data sets. These procedures allowed for the inclusion of all age-20 participants (n = 1,180) in the analyses of polysubstance use status precursors and all age-20 participants with any polysubstance use (n = 420) in the analyses of polysubstance use patterns precursors. Descriptive analyses and stability tests were conducted using SPSS, while all other analyses were performed using Mplus.

Results

Prevalence and Stability of Polysubstance Use Between Early Adolescence and Early Adulthood

The prevalence of past-year polysubstance use increased from early adolescence to early adulthood. Based on the first definition (alcohol, tobacco, cannabis), the increase was particularly sharp between ages 13, when one in five adolescents had used at least two of these three substances, and 17, when more than two-thirds had done so. At age 20, over three-quarters of young adults reported past-year polysubstance use.

According to the second definition (substances other than alcohol and tobacco, using the limited questionnaire), past-year polysubstance use became increasingly common during late adolescence. At age 17, 1 in 14 adolescents reported illicit polysubstance use; this number rose to 1 in 6 at age 20. This latter number more than doubled when using the third definition (extended, comprehensive questionnaire used only at age 20). Specifically, almost 1 in 3 young adults reported polysubstance use under this definition. While previous numbers are useful for comparison, the figure based on the comprehensive questionnaire provides the most reliable reflection of true polysubstance use prevalence in the sample. This comparative analysis shows that prevalence rates based on narrow assessments of only a few substances likely significantly underestimate polysubstance use in a population.

A sensitivity analysis that excluded CBD from the age-20 polysubstance use score showed a polysubstance use prevalence of 25%. This indicated that CBD use largely overlapped with cannabis use, as 97% of those reporting CBD product use also reported cannabis use, and 46% of those reporting cannabis use also reported CBD product use.

At age 20, the number of different substances used ranged from 0 to 13. The average number of substances used among those with any polysubstance use (excluding alcohol and tobacco) was 3.53. This count did not differ between males and females with any polysubstance use.

The stability of polysubstance use over time was high. This was observed both when using the alcohol-tobacco-cannabis coding and when using the narrow coding involving illicit substances and excluding alcohol and tobacco. These findings indicate that once polysubstance use begins, it often continues over time.

Patterns of Polysubstance Use in Early Adulthood

The investigation into different patterns of polysubstance use in early adulthood focused on illicit substance use, legal drugs other than alcohol and tobacco (like CBD in Switzerland), and the nonmedical use of prescription drugs. Alcohol and tobacco were excluded because their high prevalence in the sample would not allow for much distinction among latent classes. However, after identifying the latent classes, the prevalence of frequent (weekly or daily) alcohol and tobacco consumption within these classes was examined.

The latent class analysis included all participants who reported using at least two substances other than alcohol and tobacco in the previous year (36% of the sample, n = 420). Heroin and anabolic steroids were excluded due to their very low prevalence. A four-class solution, which had the lowest BIC, was chosen for further analysis because it revealed distinct classes with substantially different substance use profiles and reasonable prevalence (each greater than 10%).

The following four classes were identified:

• Broad Spectrum: Included participants who used many substances (e.g., cannabinoids, stimulants, hallucinogens, and opioids). Although the smallest class, more than one in ten young adults with polysubstance use belonged to it.

• Cannabinoids and Club Drugs: Characterized by the use of fewer substances than Class 1, involving cannabinoids plus stimulants, empathogens, and hallucinogens typically consumed in party contexts, and in some cases, codeine. One-third of young adults with polysubstance use belonged to this class.

• Cannabis and Medication: Characterized by medium-range substance use, primarily involving cannabinoid use plus nonmedical use of prescription drugs, including opioids and tranquilizers. One in eight young adults with polysubstance use belonged to this class.

• Cannabinoids: Primarily characterized by the use of different cannabinoids. This class, along with Class 2, represented the largest group, containing more than one-third of the sample with polysubstance use.

To further characterize the classes, their association with weekly or daily alcohol and tobacco use was examined. The likelihood of frequent alcohol drinking was higher among Classes 1, 2, and 4 than among Class 3. Comparisons between Class 3 and all other classes were significant, as was the difference between Classes 2 and 4. Frequent tobacco smoking was more prevalent among members of Classes 1 and 2 compared to those of Classes 3 and 4. These findings collectively show that the substance use spectrum for members of Classes 1 and 2 often included frequent alcohol and tobacco use, whereas substances used by Classes 3 and 4 were more selective.

Risk Factors for Polysubstance Use

First, the socio-demographic and childhood precursors of any polysubstance use were investigated, comparing it to single and no substance use at age 20. Most individuals reporting single substance use reported cannabis use, with only a minority reporting another substance as their sole drug of choice (excluding alcohol and tobacco). Separate models were specified for each precursor, adjusted for socio-demographics, and a multivariable model included all precursors simultaneously.

The results identified precursors of polysubstance use across three domains: socio-demographics, individual-level factors (psychological factors and functioning indicators), and social-environmental factors. While male participants were more likely to report polysubstance use than females, this sex difference was not significant when individual and social-environmental factors were included in the full model. A higher socioeconomic background was associated with an increased risk of substance use (either polysubstance or single substance use compared to no use). Children with two Swiss parents had a higher risk than those with a parental migration background for polysubstance use compared to single or no substance use. Additionally, childhood onset of risk-taking behaviors, such as substance use and risky media use, was associated with an increased risk of later substance use (single or polysubstance, respectively) compared to no use. Childhood sensation-seeking and exposure to maternal substance use during pregnancy were also linked to an increased risk of polysubstance use compared to no use. Friends' substance use during childhood was uniquely associated with a higher risk of later polysubstance use compared to single substance use.

Regarding the risk factors for different patterns of polysubstance use, the associations between potential risk factors and latent class membership were examined, adjusted for socio-demographics. The results showed that some childhood factors were differentially associated with specific polysubstance use patterns. For example, lower childhood self-control was consistently linked to membership in the broad spectrum class, suggesting individuals with lower self-control may have a lower inhibition threshold for trying new substances. Other factors, such as childhood internalizing symptoms, did not differ significantly across the classes. Sex and parental migration background were associated with class membership; females with polysubstance use were relatively likely to be in the cannabis and medication class, and a parental migration background increased the likelihood of being in this class compared to others.

Discussion

This investigation revealed that polysubstance use increases from early adolescence into early adulthood and often persists over time. In the urban community sample, characterized by high levels of substance use, the variety of substances used simultaneously or consecutively was broader, and the patterns of combined substances were more diverse than much previous research had indicated. Several socio-demographic factors and childhood precursors signal an individual’s risk for polysubstance use and, in part, differentiate among the various patterns of polysubstance use reported in early adulthood.

The data demonstrate that polysubstance use is highly prevalent among young people in this urban community. Even when alcohol and tobacco were excluded from the analyses, polysubstance use, as defined here, was identified in more than one-third of young adults when CBD was included, and in one in four young adults when CBD was excluded. The prevalence of any substance use in urban Switzerland is notably high compared to international evidence, possibly due to relatively high drug availability. However, comparisons of prevalence rates based on limited versus comprehensive questionnaires suggest that other studies assessing only a few substances likely substantially underestimate the true prevalence of polysubstance use among youth. To accurately assess community prevalence, it is crucial to administer comprehensive lists of substances available locally and consider the potential nonmedical use of prescription drugs.

This prospective longitudinal study, with substance use assessments starting in early adolescence, provides new evidence that adolescence is a critical period for the onset of polysubstance use. While many adolescents may experiment with different substances only once at ages 15 and 17, the increasing prevalence and high odds of continuation indicate that initiating polysubstance use in adolescence is a significant risk factor for sustained polysubstance use and potentially for progression to increasingly risky patterns. This cycle can be self-reinforcing, as high and increasing prevalence in mid- and late-adolescence can itself be a risk factor for others to engage in polysubstance use, possibly by increasing peer pressure and misjudgments about the dangers involved. Indeed, the findings show that exposure to peers’ substance use is a unique developmental precursor to one’s own subsequent polysubstance use. These findings emphasize the importance of awareness, health education, and tailored polysubstance use prevention programs targeting youth before mid-adolescence. Given the high prevalence and potentially severe consequences of chronic polysubstance use, parents and professionals should also be informed about these issues. Childhood sensation-seeking, early onset of risk-taking behaviors, and premature exposure to others' substance use were uniquely associated with an increased risk of early adulthood polysubstance use, making these promising targets or markers for early prevention efforts.

Notably, the group of young adults engaging in polysubstance use was heterogeneous, both in the combinations of substances used and some associated risk factors. All identified groups used cannabis, but they differed in the additional substances consumed, a common finding in similar research. Specifically, the use of club drugs and various cannabinoids (especially cannabis and CBD products) were the most prevalent polysubstance use patterns in the Zurich-based sample. However, a considerable number of young adults also reported nonmedical use of prescription drugs in addition to cannabis, or the use of a broad range of various substances. These insights expand international knowledge on latent classes of polysubstance use by incorporating Swiss data. The different polysubstance use classes may reflect varying contexts and motivations for substance use. For example, using a broad spectrum of substances likely indicates a motivation to experiment and a lifestyle with frequent opportunities to try different substances. The use of cannabinoids plus medical drugs with opioids may represent an attempt at self-medication for some, while others might simply use medication due to a perception of lower harm or easier accessibility compared to illicit substances. The findings on associations between risk factors and different classes suggest that prevention programs may need tailored approaches for children facing different challenges. For instance, the link between low childhood self-control and membership in the broad spectrum class suggests that individuals prone to broad experimentation might benefit from support in channeling low self-control into less risky behaviors, a focus that might be less relevant for those inclined toward other polysubstance use patterns.

However, other childhood factors were not differentially associated with particular polysubstance use patterns. For example, the classes did not differ in terms of childhood internalizing symptoms, even though prior cross-sectional research found links between depressive symptoms and polysubstance use classes involving medication use. This suggests that differences in internalizing symptoms between classes, if any, might develop later in adolescence and early adulthood. In fact, differences between the cannabis and medication class and other classes in this study mainly pertained to socio-demographic factors. Previous research has shown that females are more likely to self-medicate, and indeed, females with polysubstance use in the sample were relatively likely to be in the cannabis and medication class. Additionally, a parental migration background increased the likelihood of being in the medication class compared to other classes. These findings indicate that young adults who are typically less privileged members of society (especially females and those with migrant backgrounds) tend to engage in more hidden substance use, while more privileged groups (especially males and those with native Swiss parents) tend to engage in more overt substance use patterns.

The study has some limitations. First, substance use was self-reported, which could lead to underestimation due to social desirability, though the high reported rates suggest this was not a major issue. Second, while the list of potential childhood precursors was comprehensive, some relevant factors, such as childhood trauma, were not assessed. Third, although the sample was largely representative of young people in Zurich, the generalizability of the findings to broader Swiss or international populations is uncertain. Some substance use characteristics in the sample might be specific to Zurich, but the predictors for polysubstance use may be more widely applicable. Finally, the definition of polysubstance use, using the previous year as a timeframe, does not distinguish between individuals who tried each substance only once and those who used different substances regularly. Limiting the concept to regular use of multiple substances would likely result in lower prevalence and potentially different associations with risk factors. More precise assessments, including frequency of use, are needed to better understand polysubstance use patterns, their prevalence, and precursors.

Despite these limitations, the study has important strengths for advancing scientific knowledge about polysubstance use patterns and their developmental precursors. These include a largely representative, prospective longitudinal design, high-resolution assessment of early adulthood substance use, and a sample with a high overall prevalence of substance use. The latter facilitated the identification of several polysubstance use patterns using advanced statistical methods and their specific developmental correlates. Prior research with adolescents and young adults often identified only one polysubstance use class, comparing it to single or no substance use classes. The high rates of substance use in this population, along with the consideration of nonmedical use of prescription drugs and the application of latent class analysis only to those exhibiting polysubstance use, likely contributed to the identification of multiple distinct polysubstance use groups in this sample. Findings regarding the different associations between childhood precursors and these classes illustrate that such nuanced insights into polysubstance use patterns are crucial for developing tailored prevention programs. In summary, this investigation provides evidence that polysubstance use is a complex phenomenon affecting a significant proportion of young adults in the community, and its development, in part, originates in childhood. Future studies need to implement comprehensive assessments of various substances available in local markets to adequately assess polysubstance use and its development throughout an individual’s lifespan. The heterogeneity of polysubstance use and the specific socio-demographic and developmental factors associated with different patterns of substance use must be more carefully considered in future research to facilitate the design of tailored prevention mechanisms and reduce the individual and social burden often linked to polysubstance use.

Introduction

Using psychoactive substances like cannabinoids, hallucinogens, stimulants, and opioids, including their nonmedical use, threatens the health of young people. Substance use carries risks such as physical, psychological, social, and functional problems. These risks increase significantly when individuals use two or more psychoactive substances at the same time or one after another, a practice known as polysubstance use. Compared to using a single substance, polysubstance use is linked to more dangerous usage patterns (such as addiction and overdose), physical health issues (like injury), earlier death, other risky behaviors (like violence or dangerous driving), self-harm (including suicidal behaviors), mental health problems (like depressive symptoms), thinking problems (such as impaired executive functions and empathy), and poorer success in school and work. Experiencing these problems during early adulthood can be particularly damaging, as young people are navigating important changes in their education, careers, social lives, and personal identity.

Limited community research exists on how polysubstance use develops through adolescence and early adulthood. This gap is partly because past studies often did not assess substances beyond alcohol, tobacco, and cannabis before late adolescence. A primary goal of the study was to investigate how common polysubstance use is, and how stable it remains, from early adolescence into early adulthood. For this purpose, researchers used data from a representative sample of individuals from an urban community. This data included long-term assessments from childhood to early adulthood, with substance use evaluations starting in early adolescence.

The various patterns of polysubstance use are also not well understood. Recent research has started using methods like latent class analysis (LCA) to gain more insight. These studies typically found groups with no substance use, groups with limited to moderate use (such as only alcohol or alcohol with cannabis), and a "broad range" group. The broad range group often included all users of illegal substances and those using prescription drugs non-medically. This broad classification often occurred because studies had too few individuals using illegal substances to divide this group further, or because illegal substance use was measured generally without distinguishing specific substances. A secondary goal of the study was to better understand the diversity within the polysubstance use population. This requires samples with high rates of substance use, assessed with detailed questionnaires, which this study provided.

The most effective way to reduce the problems caused by polysubstance use would be to prevent and intervene before adolescents start this pattern of use. However, there is limited knowledge about childhood factors that predict various patterns of polysubstance use. Previous research, using different types of data, has identified individual factors (like sensation-seeking or mental health issues) and social-environmental factors (like exposure to others' substance use in family or peer settings) linked to polysubstance use. Research on childhood risk factors for any substance use in adolescence and early adulthood also points to potential predictors, including other individual factors (such as self-control, risk-taking, and externalizing behaviors) and social factors (like social stress). A third objective of the study was to identify childhood risk factors, or precursors, for both general polysubstance use and its distinct patterns in young adulthood.

Polysubstance use likely stems from various motivations and uses for substances. These can include curiosity, a desire for social connection, boosting energy or focus, seeking calm or relaxation, or enhancing or counteracting the effects or withdrawal symptoms of other substances. Researchers believed some of these motivations might be linked to specific childhood experiences or traits. Thus, it was hypothesized that different childhood individual factors (such as internalizing symptoms, sensation-seeking, risk-taking, and delinquent behaviors) and social environmental factors (such as exposure to others' substance use and social stressors) could predict polysubstance use and its various patterns in young adulthood. For instance, a tendency to try new experiences and take risks (seen in sensation-seeking or risky behaviors) and a low ability to control impulses (low self-control) might indicate a risk for later polysubstance use, especially a broad range of substances. Conversely, childhood social stressors (like being a victim of bullying) or internalizing symptoms might suggest a risk for nonmedical prescription drug use for self-medication. Additionally, if children see others using substances, they might believe substance use is common and safe. Therefore, exposure to others' substance use could lead to general polysubstance use. These potential links were investigated using the study's long-term longitudinal data.

Materials and Methods

Participants and Procedures

Data for this study originated from the longitudinal Zurich Project on the Social Development from Childhood to Adulthood (z-proso). Participants were chosen using a specific sampling method. In 2004, 1,675 children from 56 primary schools in Zurich, Switzerland's largest city, were randomly selected from 90 public schools. The selection process considered school size and the socioeconomic status of the school districts, ensuring the sample largely represented first-graders in Zurich's public schools. Individuals were assessed eight times from 2004 (age 7) to 2018 (age 20).

This study specifically utilized data gathered when participants were 13, 15, 17, and 20 years old to track the progression of polysubstance use over time. To study various polysubstance use patterns in early adulthood and their childhood predictors (assessed before age 13), individuals who participated at age 20 were included. Among these, 51% were male. Reflecting Switzerland's immigration policies and Zurich's diverse population, participants' parents originated from over 80 different countries, though most participants (90%) were born in Switzerland. Parental educational backgrounds varied widely, with at least one parent holding a university degree in 30% of households. The average household occupational status, using an international index, was 47.1. This index, comparable across countries, reflects income and required education for different jobs, with scores from 16 (e.g., unskilled worker) to 90 (e.g., judge).

This study adhered to national and international ethical standards and received approval from the appropriate ethics committee. Adolescents gave written consent to participate, while parents of those aged 15 and younger could refuse their child's involvement. Data collection occurred in groups of 5–25 participants. Up to age 17, paper-and-pencil questionnaires were used in classroom settings. At age 20, computer-administered surveys were completed in a laboratory. Surveys typically took about 90 minutes to finish. Participants received cash incentives, which increased from about $30 at age 13 to $75 at age 20.

Sample Attrition

The z-proso study saw its highest participation rate at age 15. Among those participants, females were more likely than males to continue participating at age 20 (84.5% vs. 76.9%). Individuals with at least one parent holding a university degree were more likely to participate at age 20 compared to those whose parents had lower educational qualifications (95.0% vs. 79.4%). Similarly, participants with at least one Swiss-born parent were more likely to continue than those whose parents were both born abroad (83.9% vs. 77.9%). Individuals who responded at age 20 generally had a higher family socioeconomic status during adolescence than those who no longer participated. Such patterns of participant dropout are typical in long-term longitudinal studies. The approach to handling missing data is detailed in the "Analytic Strategy" section.

Variables

Substance use was measured by asking participants how often they used specific substances in the past 12 months, excluding prescribed medical drugs. At age 20, ten categories of substances were assessed: tobacco, beer/wine/alcopops, liquor, various cannabinoids (including cannabis, CBD, and synthetic cannabinoids), stimulants (like cocaine and amphetamines), empathogens (like MDMA), hallucinogens (including LSD/psilocybin and ketamine), opioids (heroin, nonmedical codeine/painkillers), nonmedical benzodiazepine tranquilizers, and anabolic steroids. The frequency of use was rated on a six-point scale, from "never" to "daily."

During adolescence (ages 13-17), a more limited set of substances was assessed, gradually expanding over time. At age 13, only alcohol, tobacco, and cannabis were included. By ages 15 and 17, MDMA, cocaine, amphetamines, and LSD/psilocybin were added. The same six-point frequency scale was used for these adolescent assessments. To define polysubstance use, researchers created indicators for whether participants used each substance at least once in the past year. A total score was calculated by counting the number of different substances used. This score was then divided into "any polysubstance use" (at least two different substances) versus "no polysubstance use" (single or no substance use) for prevalence analysis. For analysis of developmental predictors, individuals were grouped into polysubstance use, single substance use, or no use categories. To track polysubstance use trends over time, three different definitions were used: (1) only alcohol, tobacco, and cannabis (assessed across all ages); (2) five specific illegal substances (excluding alcohol and tobacco), assessed from ages 15 to 20; and (3) all substances from the comprehensive questionnaire at age 20 (excluding alcohol and tobacco). This third definition primarily covered illegal substance use and nonmedical prescription drug use. CBD products were included because their effects and motivations for use differ from typical cannabis. A separate analysis was performed excluding CBD to see its impact on prevalence rates. For identifying polysubstance use patterns at age 20, specific substance use indicators (excluding alcohol and tobacco) were used in statistical modeling. Frequent alcohol and tobacco use (weekly or daily) was also noted as a potential factor related to these patterns.

Childhood risk factors were categorized into psychological and social-environmental factors, mostly self-reported at age 11, with sensation-seeking measured behaviorally at age 7. Psychological factors included sensation-seeking (a behavioral measure of preference for exciting situations), low self-control (acting impulsively), aggression (various aggressive behaviors), internalizing symptoms (feelings of fear or sadness), early onset of any substance use (alcohol, tobacco, or cannabis at age 11), risky media use (watching adult-rated horror/action movies or playing adult video games), and delinquency (behaviors like stealing). Social-environmental factors, primarily self-reported at age 11, included harsh parenting (parents' responses to misbehavior), bullying victimization (being ignored, mocked, or insulted by others), exposure to friends' substance use (whether close friends used alcohol, tobacco, or other substances), and maternal substance use during pregnancy (mother's self-reported use of alcohol, tobacco, or other substances during pregnancy). Socio-demographic variables considered were the participant's sex, their socioeconomic background (measured by the International Socio-Economic Index of Occupational Status), and parental migration background (whether both parents were born abroad or at least one was Swiss-born).

Analytic Strategy

First, the prevalence of any polysubstance use was calculated and compared between early adolescence (age 13) and early adulthood (age 20) to pinpoint typical periods of onset. The continuation of polysubstance use over time was assessed by examining its stability between consecutive assessments, using binary logistic regression models. This involved analyzing if polysubstance use at one assessment predicted its presence at the next.

Second, latent class analysis (LCA) was applied to identify groups of participants with distinct polysubstance use patterns in early adulthood. This analysis included all participants who reported using at least two substances other than alcohol and tobacco at age 20. The optimal number of classes was determined using statistical fit indices like Bayesian Information Criterion (BIC) and Akaike’s Information Criterion (AIC), with lower values indicating a better model fit. Entropy, a measure of classification precision, along with class sizes (preferably over 5%), also guided this decision. Different class solutions were also visually inspected for logical interpretation. Finally, the relationships between childhood predictors and individuals' polysubstance use status (any polysubstance use versus no use and single substance use) and their most probable class membership were examined using nominal logistic regression analyses.

Missing data in the LCA was addressed using full information maximum likelihood estimation. For regression models, multiple imputations were used to manage missing data on predictor variables. For each model presented, all variables involved were included in the imputation model, and 20 imputed datasets were created. Parameter estimates were then averaged across these datasets. These methods allowed for the inclusion of all age-20 participants (1,180 individuals) in analyses of polysubstance use status precursors and all age-20 participants with any polysubstance use (420 individuals) in analyses of polysubstance use patterns. Descriptive analyses and stability tests were performed using SPSS V25, while other analyses were conducted with Mplus V8.

Results

Prevalence and Stability of Polysubstance Use Between Early Adolescence and Early Adulthood

The occurrence of polysubstance use in the past year increased significantly from early adolescence to early adulthood. Using the definition of alcohol, tobacco, and cannabis use, the rise was particularly steep between ages 13 and 17. At age 13, one in five adolescents had used at least two of these three substances in the prior year. By age 17, this figure rose to over two-thirds. At age 20, more than three-quarters of young adults reported past-year polysubstance use.

When considering substances other than alcohol and tobacco (using the questionnaire introduced in mid-adolescence), past-year polysubstance use became more common in late adolescence. At age 17, one in 14 adolescents reported illegal polysubstance use, a number that grew to one in six by age 20. This rate more than doubled when using the most comprehensive questionnaire available only at age 20; nearly one in three young adults reported polysubstance use under this definition. While narrower definitions are useful for comparisons, the comprehensive assessment provides the most accurate picture of polysubstance use in the sample. This suggests that studies using limited substance assessments may significantly underreport the actual prevalence of polysubstance use in a population.

An analysis was performed to see the effect of excluding CBD from the age-20 polysubstance use score, which showed a prevalence of 25%. This indicates that CBD use largely coincided with cannabis use; 97% of those who used CBD products also used cannabis, and 46% of cannabis users also used CBD products.

At age 20, the variety of substances used ranged from none to 13. Among those engaged in polysubstance use (excluding alcohol and tobacco), the average number of substances used was 3.53. This average was similar for both males and females.

Polysubstance use demonstrated high stability over time. This pattern held true whether using the alcohol-tobacco-cannabis definition or a narrower definition focused on illegal substances (excluding alcohol and tobacco). These results indicate that once polysubstance use begins, it often continues over time.

Patterns of Polysubstance Use in Early Adulthood

The study explored various polysubstance use patterns in early adulthood, specifically focusing on illegal substance use, legal drugs other than alcohol and tobacco (like CBD in Switzerland), and the nonmedical use of prescription drugs. Alcohol and tobacco use were excluded from the primary pattern analysis because they were so common in the sample that they would not effectively differentiate between groups. However, after identifying distinct groups, the prevalence of frequent (weekly or daily) alcohol and tobacco consumption within these groups was also examined.

The latent class analysis (LCA) included 420 participants, representing 36% of the sample, who reported using at least two substances other than alcohol and tobacco in the previous year. Heroin and anabolic steroids were omitted due to their very low reported use. After comparing statistical models, a four-class solution was selected as the most suitable. This solution showed distinct groups with notably different substance use profiles and had a reasonable number of individuals in each group (over 10%).

Four distinct classes of polysubstance users were identified:

• (1) The "broad spectrum" class included individuals who used a wide variety of substances, such as cannabinoids, stimulants, hallucinogens, and opioids. Though the smallest group, it represented over 10% of young adults with polysubstance use.

• (2) The "cannabinoids and club drugs" class involved fewer substances than the first, primarily cannabinoids combined with stimulants, empathogens, and hallucinogens commonly used in party settings, and sometimes codeine. This class accounted for one-third of young adults with polysubstance use.

• (3) The "cannabis and medication" class also involved a moderate range of substance use, mainly cannabinoid use along with nonmedical use of prescription drugs, including opioids and tranquilizers. One in eight young adults with polysubstance use fell into this category.

• (4) The "cannabinoids" class was primarily defined by the use of various cannabinoids. This class, along with the "cannabinoids and club drugs" class, formed the largest group, representing over one-third of the sample with polysubstance use.

To further describe these groups, their links to weekly or daily alcohol and tobacco use were examined. Frequent alcohol consumption was more common in the "broad spectrum," "cannabinoids and club drugs," and "cannabinoids" classes than in the "cannabis and medication" class. Frequent tobacco smoking was also more prevalent in the "broad spectrum" and "cannabinoids and club drugs" classes compared to the "cannabis and medication" and "cannabinoids" classes. Overall, these findings indicate that members of the "broad spectrum" and "cannabinoids and club drugs" classes often also engaged in frequent alcohol and tobacco use, while the substance use patterns of the "cannabis and medication" and "cannabinoids" classes were more specific.

Risk Factors for Polysubstance Use

First, the study investigated how socio-demographic factors and childhood experiences predicted any polysubstance use at age 20, as compared to single substance use or no substance use. Most individuals reporting single substance use reported cannabis use. Separate statistical models were created for each potential predictor, adjusting for socio-demographic factors, and then a comprehensive model included all predictors simultaneously.

The findings indicated that polysubstance use was predicted by factors across three areas: socio-demographics, individual characteristics (like psychological factors), and social-environmental influences. While males initially showed a higher likelihood of polysubstance use compared to single or no use, this sex difference became non-significant when all individual and social-environmental factors were considered together. A higher socioeconomic background was linked to a greater risk of any substance use (either polysubstance or single substance use). Children with two Swiss parents were at higher risk for polysubstance use compared to those with a parental migration background, when compared to single or no substance use. Additionally, childhood initiation of risk-taking behaviors, such as early substance use and risky media use, was associated with an increased likelihood of substance use in early adulthood. Childhood sensation-seeking and exposure to maternal substance use during pregnancy also increased the risk of polysubstance use. Importantly, having friends who used substances during childhood was uniquely linked to a higher risk of later polysubstance use compared to single substance use. Furthermore, specific risk factors were associated with membership in the different polysubstance use patterns identified in early adulthood.

Discussion

This study found that polysubstance use increases from early adolescence into early adulthood and often continues over time. In the urban sample, where substance use levels were high, the variety and combinations of substances used in the past year were more diverse than typically suggested by previous research. Several demographic factors and childhood experiences were identified as predictors of an individual's risk for polysubstance use, and these factors also helped distinguish between different patterns of use in early adulthood.

Polysubstance use is very common among young people in this urban area. Even excluding alcohol and tobacco, over one-third of young adults in the sample engaged in polysubstance use, a figure that drops to one-quarter if CBD is also excluded. The study's findings suggest that relying on limited substance assessments in other studies likely underestimates the true prevalence of polysubstance use. Comprehensive assessment of all locally available substances, including nonmedical prescription drug use, is crucial for accurate prevalence data. This long-term study also highlights adolescence as a key period for the onset of polysubstance use. The increasing prevalence and high likelihood of continued use suggest that starting polysubstance use in adolescence is a significant risk factor for ongoing and potentially riskier patterns. The high prevalence among peers can itself become a risk factor, increasing peer pressure and leading to underestimation of dangers. Crucially, exposure to friends' substance use during childhood was found to be a distinct predictor of later polysubstance use. These findings emphasize the need for targeted awareness campaigns and prevention programs for youth before mid-adolescence. Parents and professionals should also be informed about these issues, as childhood sensation-seeking, early risk-taking behaviors, and premature exposure to others' substance use are specific factors that signal an increased risk for polysubstance use in early adulthood, making them important targets for early intervention.

The study revealed that polysubstance users are a diverse group, both in the types of substances they combine and their associated risk factors. All identified groups used cannabis, but they varied in other substances consumed. In this Zurich-based sample, the most common patterns included "club drugs" and various cannabinoids. However, significant numbers of young adults also reported nonmedical prescription drug use alongside cannabis, or a very wide range of substances. These distinct patterns might reflect different motivations for use, such as experimentation, seeking social connection, enhancing performance, relaxation, or self-medication for pain or mood issues. It is also possible that some individuals prefer prescription drugs due to a perception of lower harm or easier access compared to illicit substances. The findings suggest that prevention programs may need tailored approaches based on these different patterns. For example, low childhood self-control was linked to using a broad spectrum of substances, suggesting a need for support in redirecting impulsive behaviors. While childhood internalizing symptoms were not distinct predictors across classes in this study, other research suggests such differences might emerge later in adolescence or early adulthood. Demographic factors were important for distinguishing between groups; for instance, females and those with a parental migration background were more likely to be in the "cannabis and medication" group, suggesting that less privileged individuals may engage in more discreet substance use compared to more overt use patterns seen in more privileged groups.

A limitation of the study is that its definition of polysubstance use, covering the previous year, does not differentiate between sequential and simultaneous substance use. However, research indicates that simultaneous use is common among polysubstance users. Additionally, self-reported substance use might be subject to underestimation, though the high rates reported suggest this was not a major issue. While the list of childhood predictors was comprehensive, certain factors like childhood trauma were not assessed. The generalizability of findings beyond Zurich, Switzerland, remains uncertain, although the identified predictors of polysubstance use are likely broadly applicable. Future research would benefit from more detailed assessments of polysubstance use, including frequency information, to better understand patterns and their predictors.

Despite these limitations, the study offers significant strengths. It utilized a largely representative, long-term longitudinal design with detailed assessments of early adulthood substance use. The high overall prevalence of substance use in the sample allowed for the identification of multiple distinct polysubstance use patterns using advanced statistical methods. Unlike previous research that often identified only one polysubstance use class, this study's approach, including nonmedical prescription drug use and focusing LCA on polysubstance users, yielded more nuanced insights into these patterns and their specific developmental links. These nuanced findings are crucial for creating more effective and tailored prevention programs.

In conclusion, polysubstance use is a complex issue affecting many young adults, with roots in developmental processes beginning in childhood. Future studies should employ comprehensive assessments of local substance markets to fully understand polysubstance use and its development over time. Recognizing the diverse nature of polysubstance use and the unique demographic and developmental factors linked to different patterns is essential for future research. This understanding will help in designing targeted prevention strategies and reducing the significant individual and societal challenges often associated with polysubstance use.

Introduction

The use of substances that affect the mind (such as cannabinoids, hallucinogens, stimulants, and opioids, including their non-medical use) poses a risk to young people's health. These substances can lead to physical, mental, social, and functional problems. The risks become much greater when individuals use two or more psychoactive substances at the same time or one after another, a practice known as polysubstance use. Compared to using a single substance, polysubstance use is linked to more dangerous patterns of use, such as addiction and overdose, physical health issues, earlier death, other risky behaviors like violence or dangerous driving, self-harm, mental health problems like depression, issues with thinking and decision-making, and lower achievement in school and work. Facing these problems during early adulthood can be particularly harmful, as this is a time when young people are typically working through important life changes related to education, careers, social connections, and personal identity.

There is not much research available on how polysubstance use develops in communities during adolescence and early adulthood. This gap exists partly because past studies often did not assess substances beyond alcohol, tobacco, and cannabis before later adolescence. A primary goal of one study was to examine how common and stable polysubstance use (defined as using at least two psychoactive substances in the past year) is from early adolescence to early adulthood. Researchers used data from a large group of people in an urban community who were tracked over time from childhood into early adulthood, with substance use information collected starting in early adolescence.

The various patterns of polysubstance use are also not well understood. Research has started using methods like latent class analysis (LCA) to gain more insight. These studies typically identify groups of individuals who do not use substances, groups with limited or moderate use (such as alcohol only, or alcohol with cannabis), and a broad use group. The broad use group often includes all users of illegal substances and those who misuse prescription drugs. This is often the case because the number of people using illegal substances in these studies is too low to create more specific groups, or because illegal substance use was assessed using general questions that did not distinguish between specific substances. A second goal of the study was to better understand the different types of polysubstance use within the group of individuals who use multiple substances, which requires samples with high rates of substance use and detailed questionnaires.

The most effective way to reduce the problems caused by polysubstance use is to prevent it before adolescents begin this pattern of use. However, there is limited understanding of the childhood risk factors that predict different patterns of polysubstance use. Previous research, based on shorter-term data, has identified factors at the individual level (like seeking thrills or mental health issues) and social-environmental factors (like being around others who use substances, including family and friends) that are linked to polysubstance use. Research on childhood risk factors for any substance use in adolescence and early adulthood also points to potential predictors, including additional individual factors (like self-control, risk-taking, and disruptive behaviors) and social factors (like social stress). A third goal of the study was to identify childhood risk factors that predict any polysubstance use and its various patterns in young adulthood.

Polysubstance use likely stems from different reasons and serves different purposes. These reasons can include curiosity, a desire for social connection, boosting energy or focus, seeking calm or relaxation, or trying to enhance or counteract the effects or withdrawal symptoms of other substances. It was hypothesized that some of these motivations might be linked to specific childhood experiences. For example, a tendency to experiment with new things and take risks (like sensation-seeking or risky behaviors) or having low self-control could predict any later polysubstance use, especially using a wide range of substances. On the other hand, childhood social stressors (like being a victim of bullying) or internalizing symptoms (like anxiety or sadness) could predict the misuse of prescription drugs for self-medication. Additionally, if children are exposed to others using substances, they might come to believe that substance use is common and safe, which could lead to general polysubstance use. These potential links were examined using the study's long-term design.

Materials and Methods

Participants and Procedures

The data for this study came from the long-term Zurich Project on the Social Development from Childhood to Adulthood (z-proso). In 2004, a group of 1,675 first-grade children from 56 public schools in Zurich, Switzerland, were randomly selected to participate. This group was largely representative of first-graders in Zurich's public schools. Participants were assessed eight times between 2004 (when they were about 7 years old) and 2018 (when they were about 20 years old). The current study used data collected at ages 13, 15, 17, and 20. Of the participants at age 20, 51% were male. Most participants (90%) were born in Switzerland, and their parents came from over 80 different countries, reflecting the city's diverse population. The educational backgrounds of their parents were varied, with 30% of households having at least one parent with a university degree. The average household occupational status, a measure of socio-economic status, was 47.1 on a scale from 16 to 90. The study followed ethical guidelines and received approval from the relevant ethics committee. Adolescents provided written consent, and parents of those aged 15 and younger could decline their child's participation. Data was collected through surveys, initially paper-and-pencil in classrooms and later computer-administered in a lab. Participants received cash incentives for their involvement.

Sample Attrition

The highest participation rate in the z-proso study occurred at age 15. Among those who participated at age 15, females were more likely than males to continue participating at age 20. Participants whose parents had higher education or who had at least one Swiss-born parent were also more likely to continue in the study. Additionally, participants who remained in the study at age 20 generally came from families with a higher adolescent socio-economic status compared to those who dropped out. Such patterns of participants leaving a study are common in long-term research. The study used specific methods to account for missing data, which are detailed in the "Analytic Strategy" section.

Variables

At age 20, participants were asked how often they had used various substances in the past 12 months, excluding prescribed medical drugs. The substances included tobacco, different types of alcohol, cannabinoids (including cannabis, CBD, and synthetic cannabinoids), stimulants (like cocaine and amphetamines), empathogens (like MDMA), hallucinogens (like LSD/psilocybin), opioids (including heroin and non-medical use of prescription opioids), non-medical use of benzodiazepine tranquilizers, and anabolic steroids. Frequencies of use were reported on a six-point scale from "never" to "daily." During adolescence (ages 13 to 17), a more limited range of substances was assessed, expanding gradually over time, but always using the same frequency scale.

To analyze polysubstance use, researchers created indicators for whether participants had used specific substances at least once in the past year. A total score was calculated by counting the number of different substances used. For some analyses, this score was simplified to show whether any polysubstance use occurred (at least two different substances) versus no polysubstance use (single or no substance use). For comparisons of polysubstance use over time, three different definitions were used: one including only alcohol, tobacco, and cannabis (as these were assessed consistently), another based on five illicit substances (excluding alcohol and tobacco), and a third at age 20 that included all substances from the comprehensive questionnaire (also excluding alcohol and tobacco). The latter score primarily represented illicit substance use and non-medical prescription drug use. For identifying different polysubstance use patterns at age 20, specific substance use indicators, excluding alcohol and tobacco, were used in statistical analyses. Additionally, separate indicators were created for frequent (weekly or daily) alcohol and tobacco use.

Childhood risk factors included both psychological factors and measures of children's behavior, mostly self-reported at age 11, with sensation-seeking measured at age 7. These factors included sensation-seeking (a preference for thrilling situations), low self-control (acting without thinking), aggression (engaging in aggressive behaviors), internalizing symptoms (feelings like fear or sadness), early substance use (any use of alcohol, tobacco, or cannabis by age 11), risky media use (watching adult-rated movies or playing adult computer games), and delinquency (engaging in illegal behaviors). Social-environmental factors, also mostly self-reported at age 11, included harsh parenting (parents' disciplinary responses), bullying victimization (being ignored, mocked, or insulted by others), exposure to friends' substance use (friends using alcohol, tobacco, or other substances), and maternal substance use during pregnancy (mother's self-report of using alcohol, tobacco, or other substances during pregnancy). Socio-demographic factors included the child's sex, the family's socio-economic background, and parental migration background.

Analytic Strategy

The study employed several analytical steps. First, researchers calculated and compared how common polysubstance use was between early adolescence (age 13) and early adulthood (age 20) to identify typical periods when it begins. They also examined how stable polysubstance use was over time by looking at whether it continued from one assessment to the next. Second, latent class analysis (LCA) was used to identify groups of participants with distinct patterns of polysubstance use in early adulthood. This analysis included only those participants who reported using at least two substances other than alcohol and tobacco at age 20. The best number of groups was determined by statistical fit measures and whether the groups made conceptual sense. Finally, researchers used statistical regressions to examine the connections between childhood risk factors and an individual's polysubstance use status (i.e., any polysubstance use versus single or no use), as well as their likely membership in specific polysubstance use patterns. Missing data was handled using advanced statistical methods to ensure accurate results.

Results

Prevalence and Stability of Polysubstance Use Between Early Adolescence and Early Adulthood

The rate of polysubstance use in the past year increased significantly between early adolescence and early adulthood. When considering only alcohol, tobacco, and cannabis, the increase was particularly sharp between ages 13 and 17, with over two-thirds of adolescents using at least two of these three substances by age 17. By age 20, more than three-quarters of young adults reported past-year polysubstance use based on this definition.

When focusing on substances other than alcohol and tobacco (using the more limited questionnaire available from mid-adolescence), past-year polysubstance use became increasingly common in late adolescence. At age 17, one in 14 adolescents reported using multiple illicit substances; this increased to one in six by age 20. However, this number more than doubled when using the most comprehensive questionnaire, which was administered for the first time at age 20. This indicates that narrower assessments of substance use likely greatly underestimate the true prevalence of polysubstance use in a population. A follow-up analysis showed that excluding CBD from the age-20 polysubstance use score slightly reduced the prevalence, largely because CBD use often overlapped with cannabis use. At age 20, the number of different substances used ranged from 0 to 13, with an average of 3.53 substances among those who reported any polysubstance use (excluding alcohol and tobacco). The stability of polysubstance use over time was high, meaning that once individuals started using multiple substances, they often continued to do so.

Patterns of Polysubstance Use in Early Adulthood

The study investigated different patterns of polysubstance use in early adulthood, focusing on illicit substances, legal drugs other than alcohol and tobacco (like CBD in Switzerland), and the non-medical use of prescription drugs. Alcohol and tobacco were excluded from the initial class identification because their widespread use would not help distinguish different groups. However, the prevalence of frequent alcohol and tobacco use was examined within the identified classes.

The analysis identified four distinct patterns of polysubstance use among participants who reported using at least two substances other than alcohol and tobacco in the past year. These patterns were chosen because they had the best statistical fit and made logical sense, with each group representing a reasonable portion of the polysubstance-using sample. The four identified groups were:

• "Broad spectrum": This group included individuals who used many different substances, such as cannabinoids, stimulants, hallucinogens, and opioids. This was the smallest group, but it still represented more than one in ten young adults who used multiple substances.

• "Cannabinoids and club drugs": This group used fewer substances than the "broad spectrum" group, typically involving cannabinoids along with stimulants, empathogens, and hallucinogens often used in party settings, and sometimes codeine. One-third of young adults with polysubstance use belonged to this group.

• "Cannabis and medication": This group was characterized by moderate substance use, primarily cannabinoid use combined with the non-medical use of prescription drugs, including opioids and tranquilizers. One in eight young adults with polysubstance use belonged to this group.

• "Cannabinoids": This group was mainly characterized by the use of different cannabinoids. Along with the "cannabinoids and club drugs" group, this was the largest group, comprising over one-third of the sample that engaged in polysubstance use.

Further analysis showed that frequent (weekly or daily) alcohol consumption was more common in the "broad spectrum," "cannabinoids and club drugs," and "cannabinoids" groups compared to the "cannabis and medication" group. Frequent tobacco smoking was also more prevalent in the "broad spectrum" and "cannabinoids and club drugs" groups compared to the "cannabis and medication" and "cannabinoids" groups. These findings suggest that the substance use patterns of the "broad spectrum" and "cannabinoids and club drugs" groups often included both frequent alcohol and tobacco use, while the "cannabis and medication" and "cannabinoids" groups were more selective in their substance choices.

Risk Factors for Polysubstance Use

The study investigated socio-demographic and childhood factors that predicted polysubstance use compared to single substance use or no substance use at age 20. Most individuals who reported single substance use reported cannabis use. The results showed that factors from all three areas (socio-demographics, individual characteristics, and social-environmental factors) were linked to polysubstance use. While male participants were initially more likely to report polysubstance use than females, this difference was not statistically significant when all individual and social-environmental factors were considered together. A higher socio-economic background was associated with an increased likelihood of any substance use (either polysubstance or single substance use). Children with two Swiss parents had a higher risk of engaging in polysubstance use compared to those with a parental migration background. Additionally, childhood behaviors like early substance use and risky media use were linked to an increased risk of any substance use in early adulthood. Childhood sensation-seeking and exposure to maternal substance use during pregnancy were also associated with a higher risk of polysubstance use. Importantly, having friends who used substances during childhood was uniquely linked to a higher risk of later polysubstance use compared to single substance use.

Discussion

This investigation revealed that polysubstance use increases from early adolescence to early adulthood and often continues over time. In this urban community sample, which showed high levels of substance use, the variety of substances used together or in sequence over a year was greater, and the combinations of substances were more diverse than previously suggested by most research. Several socio-demographic factors and childhood experiences can indicate an individual's risk of polysubstance use and, in some cases, help distinguish between different patterns of use reported in early adulthood.

The data indicates that polysubstance use is highly common among young people in this urban community. Even when alcohol and tobacco were excluded from the analyses, polysubstance use was identified in more than one-third of young adults when CBD was included, and in one out of four young adults when CBD was excluded. The high prevalence of substance use in urban Switzerland compared to international data might partly be due to the relatively easy availability of drugs. However, comparisons based on different definitions of polysubstance use, using both limited and comprehensive questionnaires, suggest that prevalence rates from other studies that assessed only a few substances likely significantly underestimate the true extent of polysubstance use in young people. To accurately measure the prevalence of polysubstance use in a community, it is essential to use thorough lists of substances available locally and to consider the potential non-medical use of prescription drugs.

This study, with its long-term design starting substance use assessments in early adolescence, provides new evidence that adolescence is a critical period for the start of polysubstance use. While many adolescents may experiment with different substances only once at ages 15 and 17, the increasing prevalence and high likelihood of continued polysubstance use over time suggest that starting polysubstance use in adolescence is a major risk factor for ongoing use and possibly for progressing to more dangerous patterns. This can create a cycle where the high and increasing prevalence of polysubstance use in mid- and late-adolescence itself becomes a risk factor for others to engage in it, as exposure to peers' polysubstance use could increase peer pressure and lead to misjudgments about its dangers. Indeed, these findings show that being exposed to friends' substance use is a unique factor that predicts an individual's later polysubstance use. These findings highlight the importance of awareness campaigns, health education, and tailored polysubstance use prevention programs that target young people before they reach mid-adolescence. Given the high prevalence and potentially severe consequences of chronic polysubstance use, parents and professionals should also be informed about these issues. Childhood sensation-seeking, the early onset of risk-taking behaviors, and premature exposure to others' substance use were uniquely linked to an increased risk of polysubstance use in early adulthood. Therefore, these factors represent promising targets or indicators for early prevention strategies aimed at counteracting early adulthood polysubstance use.

The study also found that the group of young adults engaging in polysubstance use was diverse, both in terms of the combinations of substances used and some of the associated risk factors. All groups were characterized by cannabis consumption, but they differed in the additional substances used, which is a common finding in similar research. Specifically, using "club drugs" and different cannabinoids (especially cannabis and CBD products) were the most common patterns of polysubstance use in this Zurich-based sample. However, a significant number of young adults also reported using prescription drugs non-medically in addition to cannabis, or using a wide range of all different kinds of substances. These insights expand international understanding of polysubstance use patterns by providing data from Switzerland, which was previously limited. The different polysubstance use groups may reflect different situations and motivations for substance use. For example, using a broad spectrum of substances might be linked to a desire to experiment with new experiences and a lifestyle with frequent opportunities to try different substances. Conversely, using cannabinoids plus medical drugs with opioids might represent an attempt at self-medication for some individuals, aiming to ease pain or to calm down and improve mood for those with anxiety or depressive symptoms. However, other individuals in this group might simply use medication instead of illicit substances because they see medical drugs as less harmful, dangerous, or illegal, or because they are more accessible.

The findings on the links between risk factors and the different groups suggest that prevention programs may need to approach children with various challenges in different ways. For instance, the consistent link between low childhood self-control and membership in the "broad spectrum" group might reflect these individuals' low resistance to trying new substances when opportunities arise. Therefore, those on a path toward experimenting with a wide range of substances might benefit from support in redirecting low self-control into less risky behaviors. This type of prevention might not be as relevant for those on other polysubstance use paths. However, other childhood factors were not differently linked to specific polysubstance use patterns. For example, the groups did not differ in terms of childhood internalizing symptoms, even though earlier cross-sectional research found links between depressive symptoms and a polysubstance use group that included misusing medication. Together, these findings and previous research suggest that any differences in internalizing symptoms between the groups might develop only during adolescence and early adulthood. In fact, differences between the "cannabis and medication" group and other groups primarily involved socio-demographic factors in this study. Previous research has shown that females are more likely to self-medicate, and indeed, females in the polysubstance-using sample were relatively likely to be in the "cannabis and medication" group. Additionally, having a parental migration background increased the likelihood of being in the "medication" group compared to certain other groups. These findings suggest that young adults who are typically less privileged in society (especially females and those with migrant backgrounds) tend to engage in more hidden substance use than more privileged groups (especially males and those with native Swiss parents), who tend to engage in more visible substance use, such as the "broad spectrum" and "club drugs" patterns.

One limitation is that substance use was self-reported and could potentially be underestimated, though the high reported rates suggest this was not a major issue. Also, while the study included a wide range of childhood factors, some potentially relevant factors like childhood trauma were not assessed. Additionally, although the sample largely represented young people in Zurich, it is unclear if the findings apply to the entire Swiss or international populations. Some substance use characteristics might be specific to Zurich, but the predictors for polysubstance use may be more broadly applicable. Finally, the definition of polysubstance use (previous year) does not distinguish between individuals who tried different substances only once versus those who used them regularly. More detailed assessments of polysubstance use, including information on frequency, are needed for a better understanding. Despite these limitations, the study has important strengths, including its largely representative, long-term design, detailed assessment of early adulthood substance use, and a sample with a high overall prevalence of substance use. The latter allowed for the identification of several polysubstance use patterns using advanced statistical methods and their specific developmental links. Prior research with adolescents and young adults often identified only one polysubstance use group. This study's population with high rates of substance use, its inclusion of non-medical prescription drug use, and its focus on individuals already engaged in polysubstance use likely contributed to identifying multiple distinct groups. The findings on different links between childhood factors and these groups illustrate that such detailed insights into polysubstance use patterns are important for developing targeted prevention programs. In conclusion, this investigation provides evidence that polysubstance use is a complex issue affecting a significant portion of young adults in the community, and its development is partly rooted in childhood experiences. Future studies need to conduct comprehensive assessments of various available substances to adequately measure polysubstance use and its development throughout life. The diverse nature of polysubstance use and the specific socio-demographic and developmental factors associated with different patterns of use need to be more thoroughly considered in future research to aid in designing tailored prevention strategies and reduce the individual and societal burden often linked to polysubstance use.

Introduction

Using certain drugs (like cannabis, opioids, or stimulants, even if not prescribed by a doctor) can harm young people. When someone uses two or more of these drugs at the same time or one after another, it is called polysubstance use. This can lead to more serious drug problems like addiction or overdose. It can also cause physical harm, earlier death, risky behaviors like violence or dangerous driving, self-harm, sad feelings, trouble thinking clearly, and doing worse in school or at work. It can be especially bad for young adults, who are at a time in life when they are learning important skills for school, work, social life, and figuring out who they are.

Not much research has been done on how using many drugs changes as young people grow up. One main goal of this study was to see how common it is to use many drugs and if people keep doing it over time, starting from their early teen years until early adulthood. The study used information from a large group of people in a city that represents others in that area. This group had been followed since childhood, and their drug use was checked starting in their early teen years.

Another goal was to better understand how different people are within the group that uses many drugs. Some studies have started to group people based on their drug use habits to learn more. These studies often find groups that don't use drugs, groups that use a few drugs (like alcohol or cannabis), and groups that use a wide range of drugs. It has been hard to get a clear picture of how different these groups are because not enough people in past studies used many drugs, or the surveys did not ask about all types of drugs. This study used a group with many people who used drugs and asked about many different drugs.

The best way to save money and reduce problems from using many drugs would be to stop young people from starting this kind of drug use in the first place. Past research, using shorter studies, has found that certain personal reasons (like seeking thrills or having mental health issues) and social reasons (like being around others who use drugs, including family or friends) are linked to using many drugs. Other studies on any drug use have also pointed to possible warning signs from childhood, such as how much a person can control themselves, if they take risks, or if they are stressed. The third goal of this study was to find out what childhood factors might lead to using many drugs as a young adult.

It is thought that people use many drugs for different reasons. For example, some might be curious, want to feel more connected to others, or want more energy or focus. Others might use drugs to calm down, or to help with the effects of other drugs. The study thought that some of these reasons might show up in certain things from childhood. For example, a desire to try new things and take chances (like seeking thrills or doing risky things) and not being able to stop themselves easily (like having low self-control) could mean a risk of using many different drugs later. On the other hand, tough social situations (like being picked on) or feeling sad or worried as a child might mean a risk of using doctor-prescribed drugs without a prescription to feel better. Also, if children see others using drugs, they might think it is normal and safe, which could lead to using many drugs. This study looked at all these possible links using information gathered over many years.

Materials and Methods

Participants and Procedures

The information for this study came from a long-term study called "z-proso" that followed people in Zurich, Switzerland, as they grew up. A careful way of choosing people was used to pick 1,675 children from 56 schools in 2004. These children were a good match for first graders in public schools in Zurich. The people in the study were checked on eight times from age 7 to age 20. This study used information from when people were 13, 15, 17, and 20 years old. About half of the people in the study were boys. Many parents were born in other countries, which is normal for Zurich. The average job status of the parents was about 47 on a scale that goes from 16 (unskilled worker) to 90 (judge). The study followed rules to make sure people were safe and agreed to take part. People answered questions by filling out papers or using computers. They also received money for taking part.

Sample Attrition

Some people did not continue in the study. For example, at age 20, more girls stayed in the study than boys. People whose parents had a university degree or were born in Switzerland were also more likely to stay in the study. Those who stayed in the study at age 20 also had parents with better job status when they were teens. It's normal for some people to leave long studies like this.

Variables

People in the study were asked how often they had used certain drugs in the past 12 months, not counting medicines prescribed by a doctor. The drugs included tobacco, beer/wine, hard liquor, cannabis, CBD, synthetic cannabis, stimulants (like cocaine, speed), MDMA, hallucinogens (like LSD, magic mushrooms), opioids (like heroin, or using codeine cough medicine or pain pills without a prescription), benzodiazepine tranquilizers (used without a prescription), and steroids. People reported how often they used drugs, from "never" to "daily."

When people were younger, fewer drugs were asked about. At age 13, only alcohol, tobacco, and cannabis were asked about. At ages 15 and 17, more drugs were added, such as MDMA, cocaine, speed, and LSD. People reported how often they used these drugs in the past year, using the same "never" to "daily" scale.

To understand polysubstance use, the study looked at whether people had used specific drugs at least once in the past year (yes or no). Then, a total count was made of how many different drugs a person used. This count was then turned into a yes or no answer for "any polysubstance use" (meaning using at least two different drugs) versus "no polysubstance use" (meaning using one drug or no drugs). To compare polysubstance use over time, the study looked at it in three ways. First, only alcohol, tobacco, and cannabis were counted, because these were asked about at all ages. Second, it looked at five illegal drugs (not alcohol or tobacco) asked about from age 15 on. Third, it looked at all drugs (not alcohol or tobacco) from the bigger survey given at age 20. The study found that using a small number of drugs to define polysubstance use can miss a lot of cases compared to using a wide list of drugs. Another check was done without counting CBD products. This showed that using CBD often happened at the same time as using cannabis.

To find different patterns of polysubstance use at age 20, the study used yes/no answers for different drugs, but not alcohol or tobacco. This was because alcohol and tobacco were so common that they wouldn't help sort people into groups. However, after groups were found, the study checked how often people in each group used alcohol and tobacco weekly or daily.

Childhood factors were also included. These were mainly based on what children reported at age 11.

• Sensation-seeking: This was based on a game where children chose exciting things.

• Low self-control: This was about how often a child might say, "I often do things without thinking."

• Aggression: This asked how often a child did aggressive acts, like "hit other people."

• Internalizing symptoms: This asked about feelings like "I was scared or sad."

• Childhood onset of any substance use: This asked if a child had ever used alcohol, tobacco, or cannabis by age 11. A small number had tried drugs by age 11.

• Risky media use: This asked if a child had watched grown-up horror movies, action movies, or played grown-up computer games.

• Delinquency: This asked about breaking rules, like if a child had "stolen something worth more than 50 Swiss Francs."

Social factors from childhood were also looked at:

• Harsh parenting: This looked at whether parents used harsh ways of dealing with a child when they misbehaved, such as "spank you."

• Bullying victimization: This asked how often others had "ignored, left out, or made fun of you."

• Exposure to friends’ substance use: This asked if at least one friend had used drugs.

• Maternal substance use during pregnancy: This was reported by mothers and asked if they used alcohol, tobacco, or other drugs during pregnancy.

The study also looked at basic background information: if the person was a boy or girl, their parents' job status, and if their parents were born outside Switzerland.

Analytic Strategy

First, the study looked at how common it was to use many drugs from early teen years to early adulthood to see when it usually starts. Then, a way to see if one thing predicted another was used to check if using many drugs at one age meant continuing to use them at the next age.

Second, a grouping method was used to find groups of people with different drug use habits in early adulthood. This part of the study included people who used at least two drugs besides alcohol and tobacco at age 20. The study used different ways to decide the best number of groups, making sure each group was a good size. After that, a way to find links between childhood factors and drug use groups was used. Missing information was filled in so that all people could be included in the analyses.

Results

Prevalence and Stability of Polysubstance Use Between Early Adolescence and Early Adulthood

The number of people using many drugs increased from early teen years to early adulthood. When looking at alcohol, tobacco, and cannabis, the increase grew very fast between ages 13 and 17. At age 13, about one in five teens had used at least two of these three drugs. By age 17, more than two-thirds had. At age 20, more than three-quarters of young adults had used many drugs.

When looking at drugs other than alcohol and tobacco using a smaller survey, using many drugs became more common in the late teen years. At age 17, 1 in 14 teens reported using many illegal drugs. This number grew to one in six by age 20. But when using the bigger survey that asked about all drugs only at age 20, the number more than doubled. Almost one in three young adults used many drugs. This shows that studies that ask about only a few drugs can miss a lot of cases of polysubstance use. Another check without counting CBD showed that 25% used many drugs, meaning that using CBD often happened at the same time as using cannabis.

At age 20, people used from 0 to 13 different drugs. On average, people who used many drugs (not counting alcohol or tobacco) used about 3.5 different kinds. Boys and girls who used many drugs used a similar number of different kinds. Once people started using many drugs, they often kept doing it over time. This was true for both the alcohol-tobacco-cannabis group and the illegal drug group.

Patterns of Polysubstance Use in Early Adulthood

The study looked at groups of people who used illegal drugs, legal drugs other than alcohol and tobacco (like CBD in Switzerland), and doctor-prescribed drugs used without a prescription. Alcohol and tobacco were not used to define the groups because they were so common that they wouldn't help sort people into distinct groups. However, the study did check how often people in these groups used alcohol and tobacco weekly or daily.

The grouping method included people who used at least two drugs besides alcohol and tobacco in the past year. Heroin and steroids were not included because very few people used them. The four-group solution, which was the best fit, showed clear groups with very different drug use habits, and each group was a reasonable size.

The four groups found were:

• Class 1, “broad spectrum,” included people who used many different drugs (like cannabis, stimulants, hallucinogens, and opioids). This was the smallest group, but more than one in ten young adults who used many drugs belonged to it.

• Class 2, “cannabinoids and club drugs,” used fewer drugs than Class 1. This group mainly used cannabis plus drugs often used at parties (like stimulants, MDMA, and hallucinogens), and sometimes codeine. One-third of young adults who used many drugs were in this group.

• Class 3, “cannabis and medication,” also used a medium number of drugs. This group mainly used cannabis and doctor-prescribed drugs used without a prescription (like opioids and tranquilizers). One in eight young adults who used many drugs were in this group.

• Class 4, “cannabinoids,” mostly used different kinds of cannabis. This was one of the largest groups, making up more than one-third of the people who used many drugs.

To understand the groups better, the study looked at how often people in each group used alcohol and tobacco. People in Classes 1, 2, and 4 were more likely to drink alcohol often than those in Class 3. People in Classes 1 and 2 were also more likely to smoke tobacco often than those in Classes 3 and 4. This shows that people in Classes 1 and 2 often used alcohol and tobacco in addition to other drugs, while people in Classes 3 and 4 picked their drugs more carefully.

Risk Factors for Polysubstance Use

The study looked at background factors and things from childhood that might predict using many drugs compared to using just one drug or no drugs at age 20. Most people who used only one drug used cannabis. Only a few used a different drug as their only one. The study looked at each risk factor separately, then all together, accounting for background factors.

The results showed that certain background factors and childhood factors were linked to using many drugs. Boys were more likely to use many drugs than girls, but this difference became less clear when other factors were included. Having parents with better jobs meant a higher chance of using drugs (either many or just one). Children whose parents were both Swiss had a higher risk of using many drugs compared to using just one or no drugs, when compared to those whose parents were born outside Switzerland. Also, starting risky behaviors like drug use or watching grown-up media in childhood meant a higher chance of using drugs later. Seeking thrills as a child and a mother using drugs during pregnancy also meant a higher chance of using many drugs later. If friends used drugs in childhood, it was especially linked to using many drugs later, more than just one.

Discussion

This study showed that using many drugs is common among young adults and often continues over time. The different ways people combined drugs were more varied than what most past research had shown. This was especially true in this city group, where many people used drugs. Several background factors and childhood experiences were linked to a person's risk of using many drugs, and these factors also helped to tell the different groups of drug users apart.

The study found that using many drugs is very common in young people from this city. Even when alcohol and tobacco were not counted, more than one in three young adults in the study used many drugs. If CBD was not counted, it was still one in four young adults. In this city, drug use is high compared to other places. The study's comparisons using different ways to define polysubstance use showed that other studies, which only looked at a few drugs, likely missed a lot of cases of polysubstance use in young people. To understand how common using many drugs really is, studies need to ask about all the different drugs available locally, including doctor-prescribed drugs used without a prescription.

This study, which followed people from their early teen years, showed that the teen years are when polysubstance use often starts. Many teens at ages 15 and 17 might be trying different drugs only once. However, the growing number of people using many drugs over time and the high chance of continuing this use mean that starting polysubstance use as a teen is a big risk for long-term use and possibly for using drugs in riskier ways. It can be like a bad cycle: as more teens use many drugs, it can become a risk for even more teens to do so. This is because being around friends who use many drugs can lead to more peer pressure and wrong ideas about how dangerous it is. Indeed, the study found that if a person's friends used drugs in childhood, it was a unique warning sign for that person to use many drugs later.

These findings show how important it is to have health education and specific programs to prevent polysubstance use for young people before they reach their mid-teen years. Parents and health professionals should also be told about these issues. The results show that seeking thrills as a child, starting risky behaviors in childhood, and being around others who use drugs early on were all linked to a higher risk of using many drugs in early adulthood. So, these factors are good targets for early prevention efforts.

The study also found that the group of young adults who used many drugs was quite varied in terms of the drugs they combined and some of the related risk factors. All groups used cannabis, but they differed in the other drugs they used, which is a common finding in similar research. The most common patterns of polysubstance use in this city were using cannabis with club drugs, and using different kinds of cannabis (especially cannabis and CBD products). However, a good number of young adults also used doctor-prescribed drugs without a prescription in addition to cannabis, or used a wide range of all kinds of drugs. These findings add to what is known about groups of polysubstance users by including data from Switzerland.

The different groups of polysubstance users might reflect different reasons and situations for drug use. For example, using a wide range of drugs (Class 1) might mean a desire to try new things and a lifestyle with many chances to try different drugs (like at parties). Using cannabis with certain medications (Class 3) might be an attempt for some people to make themselves feel better, to ease pain, or to calm down if they feel worried or sad. However, others in Class 3 might just use these medications instead of illegal drugs because they think medicines are less harmful or easier to get.

The study's findings on the links between risk factors and the different groups show that prevention programs might need to help different groups of children in different ways. For example, the link between low self-control in childhood and being in the broad spectrum group might mean that these people have trouble stopping themselves when they have a chance to try new drugs. So, those who are likely to try many different drugs might need help finding less risky ways to deal with low self-control. This kind of prevention might not be as important for those who follow other polysubstance use patterns.

However, some childhood factors were not different among the groups. For example, the groups did not differ in terms of childhood feelings of sadness or worry. This means that any differences in these feelings might only show up later, during the teen years and early adulthood. In fact, differences between the cannabis and medication group (Class 3) and other groups were mostly about background factors in this study. Past research has shown that girls are more likely to use drugs to make themselves feel better, and indeed, girls in this study who used many drugs were more likely to be in Class 3. Also, having parents born outside Switzerland increased the chance of being in the medication group compared to other groups. This suggests that young adults who are typically less well-off in society (especially girls and those with migrant backgrounds) tend to use drugs in more private ways than those who are more well-off (especially boys and those with Swiss-born parents), who tend to use drugs in more public ways, seen in the broad spectrum (Class 1) and club drugs (Class 2) groups.

One limitation of the study is that people reported their own drug use, which might be less than what they really used. However, the high numbers reported suggest this was not a major problem. Also, while the study looked at many childhood factors, some other important factors were not included. And while the study group was a good match for young people in Zurich, it is not clear if the findings are true for everyone in Switzerland or other countries. Some drug use details in this study might be specific to Zurich (like many people using cocaine, codeine, and CBD), but the reasons for polysubstance use might be similar everywhere. Finally, the study defined polysubstance use as using drugs in the past year, which does not show if people used different drugs at the exact same time or one after another. More detailed ways of asking about drug use, including how often drugs are used, are needed to better understand patterns of polysubstance use and what causes them.

Despite these points, the study has important strengths. It followed people over many years, used very detailed questions about drug use in early adulthood, and included many people who used drugs. This allowed for finding different patterns of polysubstance use and their specific childhood links. Past research often found only one polysubstance use group. This study, with its high rates of drug use, its focus on doctor-prescribed drugs used without a prescription, and its way of grouping only those who used many drugs, helped to find multiple distinct groups. The findings about how different childhood factors are linked to these groups show that understanding these detailed patterns of polysubstance use is important for creating helpful prevention programs.

In short, this study shows that using many drugs is a complex issue that affects a large number of young adults, and it often starts with things that happen in childhood. To properly understand polysubstance use, future studies need to ask about all the different drugs available. The many different ways people use drugs, and the specific background and childhood factors linked to these patterns, need more careful study. This will help create prevention plans and lessen the problems that often come with using many drugs for individuals and for society.