Data were from the public use data files of the National Survey on Drug Use and Health (NSDUH) for the years 2014 to 2023. Our sample includes both adolescents (aged 12–17) and adults (aged 18 or older). The survey, sponsored by the Substance Abuse and Mental Health Services Administration (SAMHSA), provides annual estimates on substance use and mental health among the US civilian, noninstitutionalized population aged 12 or older. NSDUH uses a nationally representative multistage probability design with quarterly data collection. Starting in quarter 4 of 2020, NSDUH implemented multimodal data collection methods to accommodate COVID-19 safety protocols, offering both in-person and web-based surveys. Consequently, person-level analysis weights were recalibrated to account for mode-specific response patterns. Because of these changes in survey methodology, data from 2020 should not be compared to data from any other year and data from 2021–2023 are only comparable with each other (Center for Behavioral Health Statistics and Quality, 2024). Given these methodological considerations, this analysis summarizes results across three distinct periods: 2014 to 2019, 2020, and 2021 to 2023.

We assessed three cannabis outcomes: (1) past-month cannabis use; (2) past-year daily/near-daily cannabis use (use on 300 or more days in the past year); and (3) past-year Diagnostic and Statistical Manual of Mental Disorders, 5^th edition (DSM-5) CUD. Prior to 2020, NSDUH collected analogous 4^th edition DSM (DSM-IV) measures, cannabis dependence and abuse. For 2014–2019, we used a proxy for DSM-5 CUD by mapping DSM-IV items to DSM-5 criteria following a previously published method (Compton et al., 2019), which reclassifies symptoms of dependence and abuse into a unified set of DSM-5 criteria based on symptom count thresholds. Sex (female, male) was determined using the binary “IRSEX” variable across all survey years, which is intended to reflect sex assigned at birth.

We used generalized linear models with a binomial distribution and identity link to estimate female–male prevalence differences (PDs) for each study year and present results stratified by age group (12–17, 18–25, and 26 or older). Because we estimated female–male prevalence differences separately for each year without pooling across survey periods, our descriptive comparisons are less likely to conflate methodological changes with sex differences. Significance of PDs was evaluated at an alpha level of 0.05, with 95% confidence intervals (CIs) provided; if the CI included 0, the female–male PD was not considered statistically significant. Positive PDs indicate higher prevalence among females, while negative values indicate higher prevalence among males. All analyses incorporated complex survey weights and were conducted using R version 4.4.0.

From 2014 to 2019, past-month cannabis use among adolescents aged 12–17 years was similar for both females and males, with females generally reporting slightly lower prevalence (Figure 1A, Supplemental Table 1). In 2020, the difference remained non-significant. However, from 2021 to 2023, patterns reversed, with female prevalence surpassing that of males. By 2022, the female-male PD reached 2.3 percentage points (95% CI: 1.0, 3.6), followed by a non-significant PD in 2023 (PD = 1.3; 95% CI: −0.1, 2.7).

Between 2014 and 2019 among adults aged 18–25 years, past-month cannabis use prevalence steadily increased for both females and males, with the PDs narrowing from −7.4 percentage points (95% CI: −9.4, −5.4) in 2014 to –3.4 (95% CI: −5.2, −1.6) in 2019 (Figure 1A, Supplemental Table 1). In 2020, there was no significant difference between females and males (PD = −0.5; 95% CI: −3.9, 2.8). From 2021 to 2023, prevalence continued to increase for both females and males, with PDs further narrowing from –3.8 percentage points (95% CI: −6.7, −1.0) in 2021 to –1.7 (−4.1, 0.8) in 2023.

For adults aged 26 and older, past-month cannabis use prevalence increased steadily for females and males from 2014 to 2019 and 2021 to 2023 (Figure 1A, Supplemental Table 1). Females consistently reported significantly lower prevalence than males. Notably, from 2021 to 2023 the gap widened and peaked at −5.5 percentage points (95% CI: −6.7, −4.3) in 2023.

From 2014 to 2019, past-year daily/near daily cannabis use prevalence among adolescents aged 12–17 years remained low for both females and males, with females generally reporting lower prevalence than males (Figure 1B, Supplemental Table 2). However, in 2022, patterns reversed, with females reporting slightly higher prevalence compared to males, although the difference was not statistically significant (PD = 0.2; 95% CI: −0.4, 0.8). In 2023, the gap widened further, with females reporting significantly higher prevalence (PD = 0.5; 95% CI: 0.0, 1.1).

Between 2014 and 2019 among adults aged 18–25 years, past-year daily/near daily cannabis use increased slightly for both females and males, and PDs remained stable around −3.7 percentage points during this period and in 2020 (Figure 1B, Supplemental Table 2). After 2020, PDs narrowed to −1.3 percentage points in 2022, which was not statistically significant (95% CI: −2.7, 0.2). In 2023, the PD widened to −2.9 percentage points (95% CI: −4.6, −1.1), indicating a statistically significant difference.

For adults aged 26 and older, prevalence of past-year daily/near daily cannabis use increased steadily from 2014 to 2019 for females and males, with females consistently reporting lower prevalence (e.g., PD = −1.9 in 2019; Figure B, Supplemental Table 2). After 2020, prevalence continued to increase; the PD peaked at −3.2 percentage points in 2022 (95% CI: −4.2, −2.2) before narrowing to −2.5 in 2023 (95% CI: −3.1, −1.8).

From 2014 to 2019 among adolescents 12–17 years, past-year CUD prevalence for both females and males remained relatively stable, with females consistently reporting similar prevalence compared with males (e.g., PD = −0.7 [95% CI: −1.6, 0.2] in 2019; Figure 1C, Supplemental Table 3). However, this started to shift in 2020, and from 2021 to 2023, the gap steadily increased, reaching a PD of 2.4 percentage points (95% CI: 1.4, 3.4) in 2023.

Between 2014 and 2019 among adults aged 18–25 years, past-year CUD prevalence steadily increased for females and remained relatively stable for males, resulting in narrowing PDs from –6.3 percentage points (95% CI: −7.7, −4.9) in 2014 to −3.5 (95% CI: −5.0, −2.1) in 2019 (Figure 1C, Supplemental Table 3). After 2020, the PD narrowed to a low of −1.8 percentage points (95% CI: −3.8, 0.1) in 2022 as prevalence increased for females more than males. However, the PD widened in 2023 to −3.5 (95% CI: −5.8, −1.2) as male prevalence increased while female prevalence slightly decreased.

For adults aged 26 years and older, past-year CUD prevalence generally increased from 2014 to 2019, with females consistently reporting lower prevalence (e.g., PD = −1.4 [95% CI: −1.8, −1.0] in 2019; Figure 1C, Supplemental Table 3). After 2020, prevalence increased for both females and males, but to a larger degree for males, resulting in widening PDs of −3.1 percentage points in both 2022 and 2023.

Estimated population counts (in thousands) that underlie all prevalence calculations are provided in Supplemental Table 4.

NSDUH relies on self-reported data which may introduce reporting biases. For example, increased legalization and normalization of cannabis use could make individuals more willing to disclose their consumption than in previous years. The survey also does not account for hemp-derived cannabis products or cannabis potency, factors that are critical for understanding the rising prevalence of CUD and observed sex-based differences. These omissions limit the ability to fully capture evolving cannabis use patterns and their implications. Additionally, methodological changes to NSDUH in 2020 and 2021, including a transition to multimodal data collection due to the COVID-19 pandemic, complicate comparisons across the full study period. However, we do not believe these changes explain the observed sex-based patterns. Sex differences were estimated separately for each year, and the narrowing gap among 18–25-year-olds began before 2020 and continued afterward; a similar convergence among adolescents appears in other national surveys such as the YRBS. Lastly, approximating DSM-5 CUD criteria from DSM-IV items for the years 2014–2019 (Compton et al., 2019), may introduce inconsistencies when comparing estimates across years due to differences in diagnostic criteria. However, because estimates were generated separately for each year using the same method for both sexes, these inconsistencies are unlikely to explain the observed changes in sex differences over time.

The reversal of historical female versus male patterns in adolescent cannabis use, including higher CUD prevalence among adolescent females, and narrowing gaps among younger adults highlight the need for interventions addressing emerging sex-based disparities in cannabis use and mitigating associated risks like CUD. These could address specific motivations for use, such as mental health management, and consider consumption methods like the growing appeal of edibles. As cannabis legalization and normalization continue, it is critical that public health prevention and treatment efforts prioritize addressing the increasing prevalence of CUD. Lastly, more research is needed to understand the underlying reasons for these shifts and to inform effective intervention strategies.