Emergence of Sex Differences in U.S. Adolescent Substance Use

Abstract Background: Adolescence is a common time for experimentation with substance use and the emergence of sex differences in substance use patterns. Although similar in early adolescence, male and female substance use patterns historically diverge by young adulthood, with males using more substances than females. We aim to add to current literature by utilizing a nationally representative sample, assessing a broad range of substances used, and focusing on a sentinel period during which sex differences emerge. We hypothesized that certain sex-specific substance use patterns emerge in adolescence. Methods: Data are from the 2019 Youth Risk Behavior Survey (n = 13,677), a nationally representative sample of high school students. Weighted logistic analyses of covariance adjusting for race/ethnicity evaluated males’ and females’ substance use (14 outcomes) by age category. Results: Among all adolescents, more males reported illicit substance use and cigarette smoking than females, whereas more females reported prescription opioid misuse, synthetic cannabis use, recent alcohol use, and binge drinking. Divergence between male and female use usually occurred at 18+ years. Odds of using most illicit substances were significantly greater among males than females at age 18+ years (aORs 1.7–4.47). Among 18+ year-olds, males and females did not differ in electronic vapor product use, alcohol use, binge drinking, cannabis use, synthetic cannabis use, cigarette smoking, or prescription opioid misuse. Conclusions: Sex differences in adolescent use of most but not all substances emerge by age 18+ years. Sex-specific patterns of adolescent substance use may inform specific prevention efforts and identify peak ages for intervention.


Introduction
Substance use and substance use disorders often emerge in adolescence (Schramm-Sapyta et al., 2009). Studies have examined risks associated with development of substance use including static factors such as sex and race/ethnicity, and modifiable factors. To this end, assessing specific populations for differential risk has potential to target interventions for particularly vulnerable populations. One such differentiation includes biological sex. Children and younger adolescents do not differ significantly in prevalence of substance use based on sex-however, as adolescents age, sex-specific substance use patterns emerge, typically with males using more substances, and more frequently than females in early adulthood (Kuhn, 2015). Additionally, there is evidence for and against females "telescoping" substance use (initiating substance use later but progressing faster to substance use disorders and seeking treatment) (Keyes et al., 2010;Randall et al., 1999). Characterizing the pattern of this divergence in a nationally representative sample has potential implications for development of targeted prevention and intervention, and importantly, can help identify specific substances that are more or less prevalent among sexes. This study aims to provide a snapshot of age-related sex differences in adolescent substance use patterns such that emergence of sex-specific patterns of substance use can be further elucidated.
Most previous literature addressing epidemiology of age-related sex differences has been limited to the young adult population (McPherson et al., 2004;Windle, 2020) or to one or two substances studied (Buu et al., 2015;Keyes et al., 2010;Schepis et al., 2011). It is important to assess trends in sex differences in substance use in younger-age cohorts, while covering the age of risk, because this may be when prevention efforts are most potent. For instance, Windle (2020) described sex differences in alcohol-and cannabis-related outcomes as adolescents entered young adulthood and matured out of young adulthood (Windle, 2020). Though this study highlights sex differences in this age group, with males increasing substance use more than females in transition to adulthood, and females maturing out of certain substance use faster than males, it does not capture a particularly vulnerable time for the emergence of substance use (middle-to late-adolescence) (Windle, 2020). Similarly, another study focused specifically on gender differences in onset timing and quantity/frequency of alcohol co-use with nicotine and cannabis in a high-risk youth sample (youth with low grade point averages in an economically disadvantaged school district) in the Midwest (Buu et al., 2015), finding different patterns for male and female youth. Though this study addresses a vulnerable population, generalizability to the US population is limited due to restrictions in sample selection and number of substances studied (Buu et al., 2015). Keyes et al. (2010) explored if the "telescoping" effect was generalizable to nationally-representative samples, finding that while the full effect was not present, the age of initiation of alcohol use has decreased for women in recent decades (Keyes et al., 2010). Their study highlights the importance of utilizing nationally representative studies to understand sex-specific use patterns, though focuses primarily on alcohol use, therefore leaving other substance use categories unexplored.
Additionally, the last 1-2 decades have seen changes in gender and cultural norms which may affect sex-specific rates of substance use in adolescents (Keyes et al., 2008;Seedat et al., 2009). For instance, data from the National Survey on Drug Use and Health indicate convergence among males' and females' cannabis use rates by the late 1980s (Johnson & Gerstein, 2000). Similarly, data from the Center for Disease Control and Prevention's (CDC's) Youth Risk Behavior Survey (YRBS) indicate overall sex differences among high school students, with males using more illicit substances than females and females reporting higher prevalence of alcohol and prescription opioid misuse than males, though this study did not assess changes in the sex differences across adolescence (Jones et al., 2020). Women and men's alcohol use rates have been converging as well; researchers hypothesize that this may be partly due to changes in social acceptability of drinking and drunkenness for both genders (Keyes et al., 2008). Further, Seedat et al (2009)

used World Health Organization and World Mental
Health Survey data to demonstrate that sex differences in substance use disorders were lower in countries in which there existed less traditional gender roles (i.e. more women in the workforce or achieving higher education, more women using birth control, etc.) (Seedat et al., 2009). Though these studies focus on adults' roles and substance use patterns and gender rather than biological sex, it is possible that gender roles have been changing for adolescents as well and therefore may influence sex-specific substance use patterns in younger cohorts. As many previous studies examining emergence of sex/gender differences in adolescent substance use are now reaching over a decade old (McPherson et al., 2004;Schepis et al., 2011), it is important to reexamine these differences with more recent data.
Using a nationally representative sample of US high school students, this study describes the differences in prevalence of substance use by age and sex to better understand emergence of sex differences in substance use patterns and inform clinicians and future research on this important subject. We hypothesize that certain sex-specific substance use patterns will emerge as adolescents' age increases.

Participants
Data are from the Center for Disease Control and Prevention's 2019 National Youth Risk Behavioral Survey (YRBS) (n = 13,677). The YRBS yields a nationally representative sample of high school youth. Data are obtained from a school-based survey using a three-stage cluster sampling design of US students in grades 9-12 in public, Catholic and other private schools (Centers for Disease, 2018).

Independent variables
Our first independent variable is "sex" as defined by the question "What is your sex?" with the answer options "Male" or "Female." Of note, the YRBS does not provide other options for identifying sex or gender, such as "transgender" or "non-binary," which is a limitation of the sample. Our second independent variable is "age, " defined by the question "How old are you,?" treated as levels of a categorical variable in the analyses. Twelve-and thirteen-year-olds were excluded as they represented 0.3% of the sample.

Dependent variables
We identified 14 substance use-related outcome variables. Most variables were defined as having ever used a substance, or having used a substance at least once in their lifetime. Substances assessed included ever cigarette use, ever cannabis use, ever electronic vapor product use (or ever vaped), ever cocaine use, ever methamphetamine use, ever prescription opioid misuse, ever heroin use, ever inhalant use, ever ecstasy use, and ever synthetic marijuana use. We additionally included current alcohol use (use on at least one of the past 30 days) as the YRBS does not query lifetime alcohol use, binge drinking, frequent electronic vapor product use (or frequently vaped) (i.e. use on 20 or more of past 30 days), frequent cigarette use (i.e. use on 20 or more of the past 30 days), current cannabis use, and current prescription opioid misuse. Please see Supplemental Table 1 for definition of these variables.
Potential confounders available to be assessed in the YRBS include race/ethnicity. Race/ethnicity is defined according to CDC guidelines (Prevention, 2020).

Data analyses
Data were analyzed using SAS statistical software version 9.4. The SAS procedure PROC SURVEYFREQ was used to generate weighted frequencies of each outcome variable and by sex (male, female) and age. Weighted logistic analyses of covariance (ANCOVA), adjusting for race/ethnicity, were conducted using PROC SURVEYLOGISTIC for each of the 14 outcome variables using the same design and procedures. For each outcome, a 2 × 5 ANCOVA was first fit with sex, age (5 categories), and the sex-by-age interaction, adjusting for race/ethnicity. If the sex-by-age interaction was significant, males and females were compared for each age category. If the sex-by-age interaction was nonsignificant, the ANCOVA was rerun, removing the interaction, to evaluate the main effects of sex and age. Prior to analyses, graphs of each outcome were plotted to visualize the ages at which potential divergence occurred for males' and females' substance use rates, showing that males' and females' use commonly diverged by age 18 (Figure 1). Therefore, sex differences at age 18+ were tested for all outcomes. P-values <0.05 were considered statistically significant in all analyses. Table 1 depicts demographic characteristics of the sample and differences in unadjusted, weighted percent of males vs. females reporting substance use. There were a slightly higher proportion of males aged 18+ years, whereas the proportion of females was higher in the age 14 category. There was no significant difference between males and females with regards to race. Among all adolescents, without consideration of age, a significantly higher percentage of males reported illicit substance (methamphetamine, heroin, ecstasy, and cocaine) and cigarette use (lifetime and current frequent use) than females (0.2%-2.4% point difference). Conversely, a significantly higher percentage of females reported synthetic cannabis, prescription opioid misuse, current alcohol use, and binge drinking (0.2%-5.5% point difference).

Results
Results from weighted logistic ANCOVAs evaluating males' and females' substance use by age are reported in Table 2. Adjustment for race/ethnicity was significant in all models tested (p < 0.0001, not shown). The sex-by-age interaction (sex*age) indicating that the outcomes' relationship with age differed by sex was significant for reporting ever vaped, frequently vaped, and inhalant use. Sex had a significant main effect in the models for frequent vaping, illicit substance use (cocaine, ecstasy, heroin, methamphetamine), prescription opioid misuse, and current alcohol use. The age category was significant in models for all substances except inhalants, heroin, methamphetamine, and prescription opioid misuse. Table 2 also depicts sex differences in substance use in different age categories. For substance use outcomes with significant sex*age interactions (ever vaped, frequently vaped, and ever inhalant use), adjusted odds ratios were reported for each age category. For all other substance use outcomes, adjusted odds ratios for sex were reported. After adjusting for race/ethnicity and age, adolescent males were significantly more likely to report ever cocaine use (aOR 1.85; 95% CI 1.39, 2.46), ever ecstasy use (aOR 2.09; 95% CI 1.56, 2.79), ever heroin use (aOR 2.30; 95% CI 1.56, 3.38), and ever methamphetamine use (aOR 1.89; 95% CI 1.36, 2.62). Conversely, after adjusting for race/ethnicity and age, males were significantly less likely than females to report ever prescription opioid misuse (aOR 0.75; 95% CI 0.64, 0.87), current prescription opioid misuse (aOR 0.69; 95% CI 0.59, 0.81), current alcohol use (aOR 0.75; 95% CI 0.67, 0.85), and binge drinking (aOR 0.83; 95% CI 0.71, 0.96). No significant sex differences were found for cigarette use or cannabis use behaviors. Table 3 depicts sex differences in substance use outcomes, adjusting for race/ethnicity, among 18+ year-olds only. These males had 1.7 times greater odds of reporting frequently vaping (95% CI 1.3, 2.4) and 2.7 times greater odds of reporting ever inhalant use (95% CI 1.7, 4.3) compared to these females. Similarly, these males were significantly more likely than females to report current cannabis use (aOR 1.62; 95% CI 1.12, 2.34), ever cocaine use (aOR 3.18; 95% CI 1.64, 6.14), ever ecstasy use (aOR 3.23; 95% CI 1.69, 6.17), ever heroin use (aOR 4.62; 95% CI 1.85, 11.58) and ever methamphetamine use (aOR 4.47; 95% CI 1.80, 11.07). 18+-year-old males and females did not differ significantly in reports of ever cigarette use, frequent cigarette use, ever vaped, ever cannabis use, ever synthetic cannabis use, ever prescription opioid misuse, current prescription opioid misuse, current alcohol use, or binge drinking.
Figures 1-3 depict weighted percent of males and females reporting substance use outcomes in each age category. For ease of visualization, substance use outcomes were divided into three categories: combustible substances use outcomes ( Figure  1), illicit substance use outcomes ("illicit" as defined by the YRBS Data Users Guide) (Figure 2), and the remainder of substance use outcomes (prescription opioid misuse, alcohol use) (Figure 3). For combustible substances (cigarettes, electronic vapor product use, and cannabis), males and females had similar rates of use throughout adolescence (Figure 1). For most illicit substances, males consistently had higher prevalence of use compared to females, across all ages (Figure 2, panels A-D), except for synthetic cannabis. Females had higher prevalence of current alcohol, binge drinking, and prescription opioid misuse (both lifetime and past-30-day), compared to males throughout most age categories of adolescence, but with converging rates at age 18+ years ( Figure 3).

Discussion
Adolescent sex differences in use of most substances emerge by age 18+ years old, with males reporting 1.6 to 4.5 times greater odds of frequent vaping, inhalant use, and illicit substance use in this age group. However, rates of certain substances used did not differ by sex among 18+ year-olds, including substances that were more likely to be used by females overall (prescription opioid misuse, current alcohol use, binge drinking), as well as cigarette use, ever vaping, and cannabis use.
Previous literature finds adolescent males typically use higher rates of substances than females (Fish et al., 2021;Kuhn, 2015). Our results generally conform with this trend for most substances. However, certain substances were more prevalent in females (prescription opioid misuse and alcohol), and we did not observe sex differences in several substances. Of note, though females used statistically significantly higher rates of synthetic cannabis, the rates were very close between males and females (7.4% for males versus 7.2% for females) and thus, it may be that clinically, this difference is not as significant as others with wider ranges between sexes. Additionally, we observed no significant sex difference in cannabis use either overall or at age 18+ years. Our findings contrast with previous work conducted by Schepis et al. (2011), which found that by 12 th grade, 56% of adolescent males reported lifetime cannabis use compared to 49% of females (7% difference), whereas among 11 th graders, 48.2% of males and 47.8% of females report lifetime cannabis use (0.4% difference) (Schepis et al., 2011). Similarly, Jackson et al. (2002) found that adolescent males reported higher rates of alcohol use than females after age 15, though females had similar rates of tobacco use across age compared to males throughout adolescence (Jackson et al., 2002). Our study finds the converse to be true, wherein females reported higher rates of alcohol use, while males reported higher rates of cigarette use overall and at age 18+ years old. Possibly, the discrepancy between results from previous studies and ours are due in part to the time at which previous studies were conducted (10-20 years ago), and there has been a cultural or societal shift in sex-specific patterns of substance use. Indeed, data from the New Zealand National Alcohol Surveys indicate that between 1995 and 2000, gender convergence occurred in alcohol use (McPherson et al., 2004). The authors suggest that social roles for women in this age group changed in this period (greater percentage of women living alone, being childless, working in nontraditional occupations) (McPherson et al., 2004), which raises the question of changing social roles for adolescent females as well. Further, it is possible that discordance found between our results and those in prior studies may arise from different study designs (including questions and methods of delivery). Ultimately, in the context   3. How can this information inform clinical practice for those working with adolescent populations?

Developmental considerations for divergence/ convergence in sex differences in substance use at age 18+ years old
Several possible explanations exist for why substance use patterns shift by age 18+ years. First, puberty is mostly underway/complete for both sexes by this age, so there may be a hormonal component contributing to differences observed in this age group (Kuhn, 2015). Previous studies demonstrate that puberty is a strong risk factor for initiation and progression of adolescent substance use, particularly for females (Copeland et al., 2010;Windle, 2020). Levels of neurobehavioral disinhibition, sensation seeking, and impulsivity are generally lower in adolescent females compared to males, a difference which becomes exaggerated after puberty (Shulman et al., 2015). Studies also suggest that development of males' prefrontal cortex, a structure facilitating impulse control, plateaus later than females (Heitzeg et al., 2018), which may contribute to higher levels of substance use in males compared to females in adolescence. The combination of hormonal differences and sex-specific timing of brain development are possible explanations for why divergence was observed by 18+ years for most substances. Regardless of potential reasons, we identify that divergence/convergence becomes apparent at this time.

Why are males and females more likely to use specific substances?
Psychological and social differences between sexes also may help explain sex differences in substance use patterns. For  instance, sex-specific risk and protective factors may contribute to the difference in rates between males' and females' substance use. Protective factors for adolescent females against substance use include greater capacity for self-regulation and delayed gratification (Bezdjian et al., 2009;Hosseini-Kamkar & Morton, 2014), possibly contributing to lower observed illicit substance use rates. Sex differences in psychiatric comorbidity also may contribute to differences in substance use rates observed at age 18+ years: in male youth, substance use is most commonly comorbid with conduct disorder and attention deficit/hyperactivity disorder (ADHD) (Deas & Brown, 2006;Latimer et al., 2002), while depression is most commonly associated with substance use in adolescent females (Latimer et al., 2002;Simkin, 2002). Significant post-pubertal increases in female rates of depression may account for the lack of divergence observed in certain substance use rates (Angold et al., 1998;Angold & Worthman, 1993). Further, the timing of the highest prevalence of such disorders (earlier for ADHD/conduct disorder in males, later for depression/anxiety in females) may contribute to our observation of divergence of most substances at age 18+ years. However, this theory does not support our results of females using alcohol and prescription opioids at higher rates earlier in adolescence, with males catching up by age 18+ years old, suggesting the correlation with mental health disorders may only be part of the picture.
Other contributing factors to be considered include sex differences in adverse childhood experiences (ACEs). Though studies indicate that ACEs accumulation is a risk for adolescent substance use (Sheffler et al., 2020), and girls experience ACEs at higher rates than boys (Baglivio et al., 2014;Baglivio & Epps, 2016), evidence of sex/gender affecting the relationship between ACEs and development of substance use has been mixed (Gajos et al., 2023;Leban & Gibson, 2020;Lee & Chen, 2017;Mersky et al., 2013). Nonetheless, understanding root causes of the emergence of sex differences in adolescent substance use in the context of early childhood experiences is an important area for future study. Additionally, by age 18+ years, males and females may experience different sociocultural influences, including differential experiences of deviant peers (Kirisci et al., 2009) and parental supervision (Keogh-Clark et al., 2021). Studies that assessed substance use by pubertal stage found that controlling for peer substance use significantly reduced the association between pubertal stage and substance abuse (Patton et al., 2004). As animal studies often do not find sex differences or, in some instances, find that females have higher substance use rates (Kuhn, 2015), it is likely that beyond biology and hormonal influences, there exist several possible psychological and social explanations for the emergence of sex differences in substance use patterns.
As females historically have reported lower rates of most substance use (Kuhn, 2015), our finding that females were more likely to use certain substances requires further attention. Inspection of Figure 3 shows that for current alcohol use, binge drinking, and prescription opioid misuse, rates of use were higher for females than males throughout adolescence until age 18+-years-old, at which point the rates converge. This is a finding that has been previously described mostly in adults-with adult females reporting higher rates of prescription opioid misuse (Serdarevic et al., 2017) and increasing rates of drinking alcohol (Grucza et al., 2008). Understanding that this trend extends to adolescent females is concerning and helps direct prevention efforts, particularly for these substances. Further, it is possible prescription opioid misuse and alcohol use share commonality in appeal and access for adolescent females that contributes to increased rates. As both are central nervous system depressants, use of either substance may achieve similar effects, or have similar etiologies (for instance, depression or anxiety (Zale et al., 2021) or chronic pain (Witkiewitz & Vowles, 2018)) that are sex-specific (Williams et al., 2021). The reasons why males' prevalence of alcohol and prescription opioid misuse catches up to females' by age 18+ years old is concerning as well. Visualization of the trajectories of males' and females' alcohol and opioid use prevalence across adolescence shows that prevalence peaks or holds steady among females by mid-adolescence, whereas there is an acceleration of prevalence among males. Why this occurs, our results cannot answer, but it is possible that though prevalence may be decreasing among females at transitional ages, those who remain using at this time are at risk for more significant use or adverse consequences. One study found that among young adult females who are using particularly alcohol, risk of use disorder may be higher and consequences of use, more severe (Foster et al., 2015), highlighting the need for future studies exploring the emergence of sex differences in substance use disorders through adolescence.

How can our results inform clinical practice for those working with adolescent populations?
This study's findings highlight nuances of sex differences in adolescent substance use. Such nuances indicate the need for continuous updating of the landscape of adolescent sex differences in substance use. Healthcare workers will benefit from information presented by considering sex and age in providing more targeted preventions and interventions: with females reporting more prescription opioid misuse and alcohol use/binge drinking, we demonstrate that previously held understanding of males using more substances than females may not be entirely accurate. Clinicians especially should consider adolescent females a vulnerable group for use of these demonstrably harmful substances, be careful to consider adolescent females when screening for substance use (particularly for prescription opioid misuse, alcohol use), and refer to treatment accordingly. Information regarding gender differences in screening and response to brief intervention in the adolescent population is lacking. Though our study cannot inform such work, it may be that standards of screening and intervention for adolescent substance use need to be tailored based on gender. As substance use typically occurs in the context of a myriad other risky behaviors, screening for adolescent females' substance use should be accompanied by screening for risky sexual behaviors that predispose adolescent females to adverse health outcomes. Moreover, as we see increasing trends in substance use from age 14 to 18+ years old, our study highlights the need for improved substance use prevention for all adolescents, regardless of sex. Additionally, our results indicated that race was a significant predictor of all substance use outcomes. One previous study explored race as an independent predictor of substance use in the YRBS, similarly finding that race significantly predicted substance use, though without a clear pattern to be found. Future studies may explore the role that race/ethnicity plays in the emergence of sex differences. Beyond healthcare workers, our study provides framework for additional research to be conducted with the goal of disentangling the hormonal, biological, and societal factors contributing to the emergence of sex differences in substance use.
Though our study has several strengths, a few important limitations exist. First, the YRBS does not include a full spectrum of sex/gender by which to assess sex/gender differences in substance use. As such, we may be missing information from an at-risk population that identifies as gender non-binary. Secondly, the YRBS is limited in its ability to account for other potentially important confounders, such as socioeconomic status or neighborhood availability of substances. Third, because the study is survey-based, assessment of substance use is based on self-report, which may introduce recall or acceptability bias. Fourth, because the YRBS is a school-based sample, our results may not generalize to youth not attending school. Fifth, though race was a significant predictor in the models analyzed, the incorporation of race-by-gender analysis in the emergence of sex differences was outside the scope of this study and should be considered as an important separate future study exploring the emergence of sex differences by race. Sixth, though this is a large, nationally representative sample, use of certain substances (namely, illicit substances, frequent use of cigarettes) is relatively rare (<10% prevalence), possibly limiting the power of conclusions drawn regarding sex differences in rates of use. However, the value of including these substances in our analyses is that we can provide comprehensive understanding of the emergence of sex differences in adolescent substance use.
Our study strengths include the generalizability of our results, comprehensive nature of substances studied, and scientific importance of assessing sex differences in substance use patterns with recent data and in an understudied age range. We demonstrate that emergence of sex differences for most substances occur in late adolescence and that females are more likely to misuse prescription opioids and report alcohol use/binge drinking. Our work contributes to scientific literature by setting the stage for healthcare workers to provide more targeted interventions and for future researchers attempting to understand the underlying mechanisms by which substance use sex differences emerge.