Gender Differences in Illicit Drug Access, Use and Use Disorder: Analysis of National Survey on Drug Use and Health Data

Abstract

Although gender differences in the prevalence of substance use disorders (SUD) have been well-characterized, little is known about when gender differences emerge along the continuum of substance use. Understanding the contribution of gender to risk at key transition points across this continuum is needed to identify potential mechanisms underlying gender differences and to inform improved gender-responsive interventions. To characterize gender differences in the progression of cannabis, cocaine, and heroin use, the current study used data from the United States-based 2015–2019 National Survey on Drug Use and Health to quantify gender differences in: (1) perceived access to drugs, (2) lifetime drug use among individuals with at least some access, and (3) past-year SUD among those who had ever used each drug. Logistic regressions were conducted for each drug to examine gender differences across all three stages, controlling for sociodemographic factors and survey year. Compared to women, men had higher odds of reporting access to and lifetime use of all three drug types. Men also had higher odds of past-year cannabis and cocaine use disorders compared to women. Results suggest gender differences emerge in the earliest stage of drug use (access) and may accumulate across the stages of use. The magnitude of gender differences varied across stages, with the largest differences observed for odds of drug initiation among those with perceived access to each drug. Longitudinal data will be needed to confirm these findings and to provide insight into potential contributors to gender-specific risk and intervention targets across the continuum of drug use severity.

Introduction

Population estimates both in the United States and globally suggest that men are 2–3 times more likely than women to meet criteria for a substance use disorder (SUD) in their lifetime (CBHSQ, 2016; Seedat et al., 2010). Importantly, the magnitude of gender¹ differences in SUDs typically exceeds that of substance use (SAMHSA, 2022), suggesting that gender-specific vulnerability may vary across the continuum of substance use to addiction. However, few studies have specifically tested whether gender differences in prevalence emerge across key transition points along the continuum of substance use, such as initial exposure and onset of a SUD. Improved understanding of gender differences across the continuum of substance use can help to inform understanding of why these differences emerge and to identify possible intervention points that may be most critical for gender-responsive SUD interventions.

There are multiple opportunities across the continuum of drug use at which gender differences might emerge. The first is the availability of drugs in the environment, which may vary according to a wide array of factors in the environment (e.g., specific drug, family and community norms, local drug supply, geographic variation including legal status of specific drugs). Gender differences have been identified in the risk for exposure to drugs, with women more likely than men to be introduced to substances by a partner (Mburu et al., 2019).

Once someone has access to a drug, another transition point is the decision to use that drug. Nationally representative data from 1993 found that although men reported significantly greater use of cocaine, hallucinogens, heroin, and marijuana, men did not demonstrate higher odds of using within a year of initial exposure compared to women (Van Etten et al., 1999). Data from nine national samples (1979–1994; Van Etten & Anthony, 1999) and three international samples (Benjet et al., 2007; Caris et al., 2009; Delva et al., 1999), replicated these results, finding no gender differences in odds for use following the initial opportunity to use a drug.

Mixed evidence has emerged for odds of transitioning from initial exposure to regular use or misuse, with some studies reporting no gender differences (e.g., Wells et al., 2011), and others suggesting higher odds in men (Swendsen et al., 2008). Conditional dependence—or the likelihood of developing an SUD among people who use a drug—also appears to be higher in men (Kalaydjian et al., 2009; Wagner & Anthony, 2007), with the possible exception of nicotine (Lopez-Quintero et al., 2011). Taken together, understanding of points along this continuum at which gender differences in substance use emerge, and how these differences may accumulate along this continuum, remain unclear.

The aim of the current study is to characterize recent (2015–2019) gender differences along the drug use continuum. We focused on three drugs with different effects and population base rates: cannabis, cocaine, and heroin. To enhance understanding of the progression of drug use we examined gender differences at three stages: (1) access to drugs (how difficult or easy would it be to get a given drug, and whether a person has been approached to be sold any illegal drug in the past 30 days); (2) lifetime use of drugs among individuals who report having at least some drug access; and, (3) prevalence of past-year SUD among individuals who have ever used that drug (i.e., conditional dependence). This approach was used to attempt to isolate gender differences in key transition points along the continuum of drug use. Specifically, we sought to understand gender differences in each stage of this continuum (access, use, use disorder) once people had completed the prior stage (i.e., among those who reported access to a drug, what were the gender differences in proportion of those with lifetime use). This is akin to conditional dependence (the likelihood of an SUD among people who have used a substance), with the addition of an earlier stage of transition from access to use. To accomplish these aims, we utilized publicly available data from the National Survey on Drug Use and Health (NSDUH). We combined survey data from 2015 to 2019, examining gender differences in access, lifetime use, and prevalence of SUD for each substance with available data (i.e., cannabis, cocaine, heroin) separately.

Methods

This secondary analysis of publicly available data was pre-registered on the Open Science Framework (https://osf.io/a3u8w). The National Survey on Drug use and Health (NSDUH) is directed by the Substance Abuse and Mental Health Services Administration (SAMHSA) and conducted yearly in all 50 states and the District of Columbia. Roughly 70,000 people are interviewed yearly and identified via random selection within the stratified census tract. A new sample is recruited each year; as such, each survey is cross-sectional. The sample includes individuals 12 years and older living in residential households and noninstitutionalized group housing, such as dorms, shelters, and rooming houses. Excluded individuals included people experiencing homelessness not living in a shelter, active military personnel, and those institutionalized in a jail or hospital. The primary aim of the NSDUH is to report on trends in alcohol, tobacco, and other drug use and misuse in the United States. The NSDUH uses a complex survey design that allows for state and national population estimates of prevalence of different substance use behaviors. Further detail on survey methods can be found on the SAMHSA website (https://www.samhsa.gov/data/data-we-collect/nsduh-national-survey-drug-use-and-health).

The sample used in this analysis combined NSDUH data from 2015–2019, resulting in an unweighted sample of 282,769, corresponding to an estimated weighted sample of over 271,640,450. There was <1% missing data that was assigned missing based on the NSDUH codebook.

Several indicators were extracted from the NSDUH dataset for this analysis. Gender was assessed as a binary variable (male or female). Of note, the NSDUH does not separately assess gender and sex; we use the term gender noting that it cannot be confirmed whether this refers to gender or sex. Drug access was assessed with Likert-type items (i.e., “How difficult or easy would it be to get some [blank], if you wanted some?”) with options of “nearly impossible,” “very difficult,” “fairly difficult,” “fairly easy,” and “very easy” for cannabis, heroin, and cocaine. In addition, we analyzed using a binary item for whether an individual had been solicited to buy an illegal drug in the past 30 days (i.e., “In the past 30 days, has anyone approached you to sell you an illegal drug?”). Lifetime use of cannabis, cocaine, and heroin were assessed using a single self-reported binary item (i.e., “Have ever, even once, used [blank]?”). The NSDUH from 2015–2019 utilized Diagnostic and Statistical Manual of Mental Disorders, 4^th Edition (DSM-IV; American Psychiatric Association, 1994) definitions of substance use disorders, including substance abuse and substance dependence for each drug class. These were combined in the NSDUH dataset to an aggregate binary “substance use disorder” indicator, representing a diagnosis of abuse and/or dependence, roughly corresponding to the DSM-5 (Diagnostic and Statistical Manual of Mental Disorders, 2013) conceptualization of these disorders. Of note, this cannot be fully equated with the DSM-5 definition due to other modifications from the 4^th to 5^th edition (e.g., adding a criterion for substance craving). See Table 1 for mean endorsement of substance use variables by gender. Age (i.e., 12–17,18–25, 26–34, 35–49, 50–64, 65+ years old), county type (i.e., large metro, small metro, rural), annual family income (i.e., <$20,000; $20,000-$49,999; $50,000-$74,999; ≥ $75,000) and survey year (i.e., 2015 – 2019) were extracted for use as covariates.

Table 1.

Substance use behaviors broken down by gender

Measure	Women	Men
Unweighted sample	148,111	134,657
Estimated Weighted sample	139,901,642	131,738,808
Drug Availability
Cannabis	3.43 [3.41, 3.44]	3.59 [3.58, 3.60]
Cocaine	2.27 [2.26, 2.28]	2.36 [2.36, 2.37]
Heroin	2.07 [2.06, 2.08]	2.09 [2.08, 2.10]
Lifetime Drug Use
Cannabis	0.41 [0.41 0.41]	0.49 [0.49, 0.50]
Cocaine	0.18 [0.18, 0.19]	0.18 [0.18, 0.19]
Heroin	0.01 [0.01 0.01]	0.03 [0.02, 0.03]
Past Year Drug Disorder
Cannabis	0.01 [0.01, 0.01]	0.02 [0.02, 0.02]
Cocaine	0.002 [0.002, 0.003]	0.005 [0.004, 0.005]
Heroin	0.001 [0.001, 0.002]	0.003 [0.002, 0.003]

Note. Values are means with weighted 95% confidence intervals in brackets.

Statistical Analyses

R version 4.1.2 was used for all analyses (R Core Team, 2020) including the “survey” package (Lumley, 2020) to perform analyses accounting for the complex survey design of the NSDUH data. This package uses the complex survey design elements (e.g., stratification, weights) to adjust standard errors and produce weighted population estimates (see Lumley, 2011, 2020 for in-depth information). Gender was analyzed as a binary variable with females as the reference category in all models. All models included age, income, county type, and survey year as covariates. Race and ethnicity were not included as covariates consistent with recent guidance on the use of these variables in biomedical research (see Cardenas-Inigues & Gonzalez, 2024). Age, income, and county type were treated as ordered factors and parameterized using orthogonal polynomial contrasts (Lumley, 2020). For each ordered factor, a series of polynomial contrasts are fit that total one fewer than the number of factor levels (Fox & Weisberg, 2019). More information on the use of polynomial contrasts can be found in supplemental materials. For example, county type has three levels (i.e., rural, small metro, and metro), so there are two polynomial contrasts fitted (i.e., linear and quadratic). Survey year was treated as an unordered factor.

To enhance interpretability of findings, we generalized the concept of conditional dependence across the developmental stages of SUD. To accomplish this, following analyses of access, we removed individuals who denied ability to access drugs for use analyses (aim 2), and removed individuals who denied lifetime use in past-year SUD analyses (aim 3). By removing these subjects from analyses, we are able to more accurately estimate gender differences in odds of lifetime use and past-year SUD, as inclusion of these individuals may artificially lower estimates of lifetime use (if they have no access) or past-year SUD (if they have no lifetime use). Accordingly, each of these analyses reflects a step along the continuum of substance use behaviors: who can access substances; of those who can access substances, who tries them; and of those who try substances, who develops a SUD.

Aim 1: Drug Access.

Our first aim was to examine the effect of gender on markers of ease of access to each drug. Differences in self-reported ease of access to cannabis, cocaine, and heroin were estimated using logistic ordinal regressions with proportional odds (labeled “logistic” in survey package) parameterization (Lumley, 2020). In addition, a logistic regression was used to examine the effect of gender on self-report of being approached to purchase any illegal drug in the past 30 days (binary). Lastly, all analyses in Aim 1 included the full available sample.

Aim 2: Lifetime Drug Use.

Our second aim was to examine the effect of gender on self-reported lifetime use of cannabis, cocaine, and heroin using logistic regressions. For these analyses, the sample only included individuals who reported having at least some access to each drug. This was operationalized by excluding individuals who reported it was “nearly impossible” to obtain the given substance and to include individuals who reported it would be “very difficult,” “fairly difficult,” “fairly easy,” and “very easy” to obtain each drug. Age, income, county type, and survey year were again included as covariates. In addition, access to a given substance was added as a covariate to each substance specific model. These variables were treated as ordered factors that consisted of the remaining four factor levels (i.e., “very difficult,” “fairly difficult,” “fairly easy,” and “very easy”). This was done to control for the relative ease of access among individuals who reported possible access. Lastly, sensitivity analyses were conducted that repeated the above analyses, but only included individuals who reported each substance was “fairly difficult” or easier to obtain. This was done to demonstrate the robustness of our findings to alternative cut points for “potential access”.

Aim 3: Past-year Substance Use Disorder.

Our third aim was to examine the effect of gender on past year cannabis, cocaine, and heroin use disorder using logistic regression. These analyses only included individuals who reported lifetime use of each substance. Age, income, county type, survey year, and drug access were included as covariates. Lastly, sensitivity analyses were conducted for all aims that treated all covariates as nominal to examine how dependent gender effects were on the method for controlling for covariates. Supplementary table 1 presents odds ratios from this analysis and demonstrates that this decision had no substantive effect on the direction or magnitude of estimates of interest.

Results

Aim 1: Drug Access

Table 2 presents the odds ratios for the effect of gender on cannabis, cocaine, and heroin self-reported access. For all substances, men reported significantly greater access compared to women. In fact, being a man was associated with a 16% increase in the odds of a reporting a higher level of access to cannabis and cocaine compared to women and a 6% increase in access to heroin. Men also reported significantly higher odds of being approached to be sold drugs (OR = 1.76 [95% CI, 1.69, 1.82], p<.001). Covariate parameters from the ordinal regressions for drug access are presented in supplemental table 2. Covariate parameters from the logistic regression for being approached to be sold drugs are presented in supplemental table 3.

Table 2.

Gender effects Odd Ratios (OR)

Aim 1. Drug Access	Cannabis	p-value	Cocaine	p-value	Heroin	p-value
Unweighted N	282,768	-	282,768	-	282,768	-
Weighted N	271,640,450	-	271,640,450	-	271,640,450	-
Gender OR	1.16 [1.14, 1.19]	p<.001	1.16 [1.14, 1.19]	p<.001	1.06 [1.04, 1.09]	p<.001
Aim 2. Lifetime Illicit Drug Use
Unweighted N	232,693	-	169,592	-	147,530
Weighted N	220,146,010	-	159,567,894	-	139,018,437
Gender OR	1.32 [1.29, 1.36]	p<.001	1.79 [1.72, 1.86]	<.001	2.45 [2.22, 2.69]	<.001
Aim 3. Past Year Drug Use Disorder
Unweighted N	121,325	-	33,316	-	4,879	-
Weighted N	122,008,346	-	40,018,689	-	5,237,745	-
Gender OR	1.94 [1.80, 2.10]	p<.001	1.39 [1.13, 1.70]	p<.001	1.10 [0.83, 1.46]	p=.50

Note. Gender is the effect of being male with females as the reference group. OR = Odds ratio.

Aim 2: Lifetime Drug Use

Table 2 presents the odds ratios for the effect of gender on lifetime cannabis, cocaine, and heroin use. Results indicated that males have a significantly higher likelihood of lifetime use of all three substances examined. These differences were substantial for lifetime cocaine and heroin use, with men reporting more than double the odds of using heroin (145% higher odds), and 79% higher odds of using cocaine, compared to women. Men also demonstrated a 32% increase in odds of using cannabis in their lifetime compared to women. Covariate parameters for these analyses are presented in supplemental table 4.

Sensitivity analyses demonstrated there were no substantive differences in results when a different cut off point was chosen to operationalize “potential access” to each substance. These results are presented in supplemental table 5 and covariate parameters are presented in supplemental table 6.

Aim 3: Past-Year Substance Use Disorder

Table 2 presents the odds ratios for the effect of gender on past year cannabis, cocaine, and heroin use disorders. Results indicated that males were significantly more likely to meet criteria for past-year cannabis and cocaine use disorders, but were not more likely to have a heroin use disorder. Specifically, men had 94% higher odds of having a cannabis use disorder, and 39% higher odds of a cocaine use disorder. Covariate parameters for these analyses are presented in supplemental table 7.

Discussion

A growing body of evidence suggests biological, psychological and sociocultural factors might all contribute to the observed gender differences in SUD prevalence (Becker et al., 2016; Greenfield et al., 2007; Lind et al., 2017; see McHugh et al., 2018 for review). In this study, we used US-based national data to characterize gender differences along three transition points in the continuum of cannabis, cocaine and heroin use: access or opportunity to use a drug, lifetime drug use among those with access, and SUD diagnosis among those with lifetime drug use. Although we refer to gender differences throughout this discussion, we acknowledge that the NSDUH methods do not distinguish gender from sex and thus we cannot disentangle the contributions of biological sex and gender identity, or non-binary categories of either of these variables. Consistent with the overall higher prevalence of SUDs in men compared to women, our results indicated a general pattern of higher odds of risk at each stage for men, with the exception of heroin use disorder among people with a lifetime history of heroin use. It is also notable that—cumulatively—these gender differences, characterized by higher likelihood of access, use, and use disorder among men, result in substantial differences in overall prevalence in men and women. For example, men are more than twice as likely as women to have a cocaine use disorder (McHugh et al., 2023).

The magnitude of the observed gender differences varied across the continuum of severity and, to some extent, across the three drugs examined. Particularly notable were the large effects for lifetime drug use and transition to cocaine or cannabis use disorder, which far exceed gender differences in perceived access to cocaine and cannabis. This suggests that women had much lower odds than men of trying cocaine and cannabis if they perceive having access to it and transitioning from cocaine or cannabis use into a disorder, but relatively small differences in perceived access. By contrast, with respect to heroin, among people who had tried heroin in their lifetime, there was not a significant gender difference in the likelihood of past-year heroin use disorder. It is possible that the absence of gender differences in heroin use disorder may have been attributable to the low base rate endorsement of heroin use disorder in men and women (see Table 1).

The timepoints along the continuum of use may also correspond to differing degrees of gender vs. biological sex effects, including effects that may vary across the lifespan or over time (i.e., cohort effects). Differences in perceived access to drugs may be driven in part by environmental and role norms (e.g., likelihood of partner-influenced use), whereas differences in the risk for SUD onset in people exposed to a drug may be influenced by biological differences in the metabolism of drugs, their pharmacokinetic effects, gonadal hormones levels, or in the availability and distribution of certain receptors in the brain (e.g., Becker & Chartoff, 2019; Weinberger et al., 2015). Furthermore, comparable mechanisms cannot be assumed, even in cases of modest or absent gender differences. The risk factors for development of a SUD are varied and may take multiple potential pathways. For example, some people with SUD have a high genetic risk, whereas others may have had increased risk as a consequence of trauma exposure or chronic pain. Although broad gender differences are present such factors (e.g., women are more likely than men to have a diagnosis of posttraumatic stress disorder; Kilpatrick et al., 2013), the variability within each gender also must be considered. Effects of culture and environment may also intersect with biological differences. As such, there is potential for substantial interaction between environmental, cultural and biological differences, potentially modifying biological stress related SUD vulnerability, that may have a dynamic impact on observed gender differences across the stages of substance use. Future research may seek to identify the biological, social, and environmental factors that contribute to the observed elevated risk for men across the stages of substance use.

Of note, these analyses used cross-sectional data and thus conclusions about any individual’s trajectory along these time points cannot be drawn. For example, gender differences in risks may vary based on the age at which use is initiated (e.g., adolescence vs. early adulthood) or the contexts in which this occurs (e.g., introduced by a partner vs. by a friend). In addition, although these results characterize gender differences in prevalence, they cannot directly provide information on the mechanisms underlying these differences. Furthermore, our analysis did not consider the time between any of these stages; prior research has shown some evidence for gender differences in the rapidity of transitioning along the continuum of substance use (Hölscher et al., 2010; Khan et al., 2012). Similarly, both age effects and cohort effects on gender differences in substance use have been previously noted (McHugh et al., 2021; Warren et al., 2023). Our study adjusted for age, but cannot disentangle the possible effects of cohort differences, such as the possibility for narrowing gender differences in younger cohorts. Longitudinal research will be needed to further clarify these findings.

There are several additional limitations in the current study related to the design of the NSDUH. First, there was a lack of representation or underrepresentation of people who are incarcerated or experiencing homelessness, which may underestimate certain substance use behaviors, such as severe SUDs. Second, the NSDUH assesses gender as a single variable, with a binary definition of male or female. Accordingly, the ability to disentangle biological sex from gender identity or to consider the full variability and intersectionality of sex and gender is not possible. Third, as the NSDUH was designed to capture population-level trends in substance use and SUDs, base rates for some issues such as heroin use disorder remain low, which potentially obscures any possible gender differences. In addition, the NSDUH does not assess illicit fentanyl use, which has increased in prevalence during the current study’s time frame (e.g., Martinez et al., 2020).

Conclusions

Women have historically been underrepresented in research in SUDs (Wetherington, 2007) and a concerted effort to increase representation continues to reveal new information about the ways in which biological and sociocultural factors may differentially influence substance use in men and women. Our results suggest that gender differences in the prevalence of cannabis, cocaine and heroin use disorder emerge at the very earliest stages of the continuum (i.e., potential access or opportunity to try a drug) and accumulate across this continuum from lifetime use to SUDs. At any of these points, potential sociocultural (e.g., gender norms), psychological (e.g., impulsivity, depression) and biological (e.g., metabolic) gender and sex differences may contribute to the observed differences in prevalence. These results should not be assumed to generalize to other drug types (e.g., prescription drugs, tobacco, alcohol), which may exhibit distinct patterns of gender differences (see Ford et al., 2014). Future research examining how gender differences across the continuum of use to SUDs in prescription drugs will be needed to determine whether a similar process occurs for this drug type. Further research also will be needed to elucidate the contributors to these differences at each stage to ultimately inform our understanding of sex/gender and substance use, and to contribute to prevention and treatment interventions that are best targeted to the unique and common needs of men and women.

Highlights.

• It is unclear when gender differences emerge in the progression of drug use.

• Men have higher odds of perceived access to cannabis, cocaine, and heroin.

• Of those with access, men have substantially higher odds of drug initiation.

• Men have higher odds of a past-year cocaine or cannabis use disorder.

• Men and women have similar odds of past-year heroin use disorder.