THE REPRESENTATION OF FEMALES IN CLINICAL TRIALS FOR SUBSTANCE USE DISORDER CONDUCTED IN THE UNITED STATES (2010–19)

Abstract

Background and aims

Women have historically been under-represented in clinical research, but the extent to which this is true for substance use disorder (SUD) trials is unknown. We aimed to determine the ratio of female:male participation in clinical trials for SUDs and describe the reporting of sex-specific outcomes from 2010 to 2019.

Design

A retrospective cohort review of clinical trials involving people with SUD.

Setting

United States.

Participants

Clinical trials including people with SUD registered in clinicaltrials.gov and completed between 1 January 2010 and 31 December 2019 were reviewed. Trials were excluded if they had < 30 participants, focused on SUD prevention, were conducted outside the United States and/or did not report data on participant sex or gender.

Measurements

The following were extracted for each trial: primary outcome, number of participants enrolled, analytical sample size, percentage of participants who were female, inclusion of transgender participants, whether sex-based analyses were performed, funding source, type of SUD and type of intervention. Relative representation in trials was examined using the female:male ratio, reported using median ratios and by year of trial completion. The proportion of females participating was adjusted using the underlying disease prevalence among females using National Survey on Drug Use and Health data.

Findings

A total of 316 trials met inclusion criteria: 274 were mixed-sex, 12 enrolled only males and 30 only females. In 274 mixed-sex trials, 40% of 57 544 participants were female. Only 22 trials (8%) reported any sex-specific analyses; four studies (1.5%) reported inclusion of transgender participants. Females represented 35% of participants in trials targeting illicit drug use disorder, 52% in nicotine use disorder and 29% in alcohol use disorder. Accounting for underlying disease prevalence revealed that women had the lowest relative enrollment in alcohol use disorder trials (median participation to prevalence ratio in 2017: 0.58; 95% confidence interval: 0.13, 0.91).

Conclusions

A review of 316 US clinical trials for alcohol, nicotine and illicit substance use disorders completed between 2010 and 2019 showed that females were enrolled at lower rates than males overall. Only 8% of the trials reviewed reported sex-specific analyses and 1.5% reported transgender participants.

INTRODUCTION

Females have historically been under-represented in biomedical clinical trials [1–3]. Since the 1980s, national organizations and the federal government have promoted the importance of sex (a biological construct) and gender (one’s social identity) to the study of disease prevalence, risk factors and treatments. They have also acknowledged the need to improve the representation of females in research [4]. In 2010, the Institute of Medicine cited concerns that there was ‘inadequate enforcement of requirements that representative numbers of women be included in clinical trials’ and inadequate reporting of sex- and gender-based results [5]. Indeed, sex- and gender-specific data on the effects of interventions are poorly characterized and rarely reported [1]. Certain conditions, including alcohol and drug-related disorders, have substantially impacted women’s quality of life, yet there have been few improvements in reducing the incidence, prevalence, morbidity and mortality associated with such conditions [5], which may be in part attributable to women’s under-representation in clinical trials.

The epidemiology of drug and alcohol use is rapidly changing, with an increasing proportion of those initiating drug use [6] and those with substance use disorders (SUDs) being female [7–9]. Females are more likely to have an accelerated trajectory from initial use to a severe SUD due to factors including a higher prevalence of co-occurring mental health disorders, trauma and stigma [10–14]. The increasing prevalence of females using drugs and alcohol is reflected in outcomes: rates of prescription opioid overdose deaths have increased 400% in females compared to 256% for males since 1999 [15]. Similarly, while rates of alcohol use and alcohol use disorder have remained stable for adult males since 2002, the gap between male and female use has significantly narrowed [16]. Throughout alcohol use disorders, females are also less likely to engage in treatment relative to men: 22% of males and 15% of females with alcohol use disorder receive treatment in their life-time [17].

The differences in co-occurring disorders and trauma, increases in use prevalence and lower participation in treatment may similarly influence females’ participation in clinical trials for SUDs. While female participation in clinical trials for some diseases is relatively well-characterized [2, 18, 19], less is known about SUD-related trials. One 2011 study describing gender research in the National Institute on Drug Abuse’s Clinical Trial Network documented that among 24 trials that enrolled 11 500 participants, 41% of participants were female; four of these studies were gender-specific (enrolling only women) [20]. A recent systematic review of alcohol use disorder trials demonstrated that since 1995, approximately half of trials reported the proportion of participants who were female [21]. Given the paucity of recent literature in this area, along with the implementation of a 2015 National Institutes of Health (NIH) policy requiring all funded research to include sex as a biological variable in design, analysis and reporting, we sought to determine the ratio of female to male participation in clinical trials for SUDs and describe the reporting of sex-specific outcomes.

METHODS

Study design

This study uses a retrospective design to examine the representation of female participants in interventional clinical trials for SUDs among trials registered in clinicaltrials.gov. Clinicaltrials.gov is a compendium of clinical trials maintained by the United States National Library of Medicine. All clinical trials conducted in the United States are required to register study details within this system. For this analysis, we identified all interventional studies with a status listed as ‘completed’ between 1 January 2010 and 31 December 2019 and associated with the term ‘substance use disorder’ were identified. This period was chosen to coincide with the Institute of Medicine’s report on inadequate reporting of sex-based clinical trial data in 2010 and previous reviews of female enrollment in clinical trials [20]. This generated 787 trials when extracted on 10 June 2020. This study was not pre-registered as a systematic review.

Study inclusion and exclusion criteria

Following a structured protocol, two study team members (M.P., E.H.) evaluated the 787 potentially eligible trials for inclusion using information from clinicaltrials.gov. Included studies were required to have enrolled at least 30 participants, included people with SUDs, had study results published and reported individual-level characteristics of study participants. We excluded trials outside the United States as all US trials involving human subjects are required to report enrollment in clinicaltrials.gov, which supported the identification of a relatively complete list of eligible trials and associated trial characteristics. Including only the United States also allowed comparisons to underlying national SUD prevalence rates, which vary globally [22]. However, some trials included enrollment outside the United States. Restricting studies to those with > 30 participants ensured that we did not underestimate the prevalence of sex- and gender-based analyses due to power limitations. To minimize erroneous exclusions, if trial results for the primary outcome were not identified through a search of the peer-reviewed literature, we initiated outreach to the principal investigator or study sponsor to solicit results. If no response was received within 30 days and no trial report could be obtained, the trial was excluded. Studies that enrolled healthy individuals to examine the effect of an intervention on the prevention of SUD were also excluded. Trial exclusions were reviewed in weekly meetings to ensure that criteria were consistently applied (M.P., E.H., C.G., S.B.).

Data collection

For each trial, the following data elements were extracted: the study’s primary outcome; participant age eligibility; the number of participants enrolled; the total analytical sample; the percentage of participants who were female; whether transgender participants were included; whether sex- or gender-based analyses were reported; the funding source; type of substance(s) included (nicotine, alcohol, opioids, cannabis, stimulants, multiple, other); trial completion date; and type of intervention. Of note, we report here using the term ‘female’ to denote biological sex. However, some trials report self-described gender, referring to one’s identity. We chose to use the biological term, which has been the primary focus of National Institutes of Health policy, although we recognize the importance of acknowledging both sex and gender as unique constructs. In this study, both sex and gender are represented in reported numbers of ‘females’. Funding sources were categorized as government (e.g. National Institutes of Health, Centers for Disease Control, Veteran’s Affairs), industry (pharmaceutical and biotech companies), research institute (e.g. foundations), internal funding (hospital, university), multiple sponsors including the National Institutes of Health and multiple sponsors excluding the National Institutes of Health. Intervention types were classified as using either drugs (pharmaceutical products), devices (e.g. e-cigarettes), behavioral interventions (e.g. cognitive–behavioral therapy), multiple intervention types (e.g. a drug plus behavioral intervention) or other (e.g. education, collaborative care). Study design was specified as a randomized controlled trial with individual randomization, a single-arm trial, a cluster-randomized trial, multi-level randomization or quasi-experimental. The presence of sex- or gender-specific analyses was determined by any analysis or results that included examining differential effects of the intervention by sex or gender.

At the outset, groups of 10 trials were evaluated by two coders for calibration (M.P., E.H.). Results and discrepancies were reviewed with the study team and consensus was achieved (S.B., C.G., M.P., E.H.). Trials were then individually evaluated, with 15% randomly selected for review by a second coder to ensure consistency in the application of eligibility criteria and extraction of data elements. Weekly consensus meetings with the study team reviewed issues that arose.

Analysis

Trials were first segmented into single-sex versus mixed-sex trials and characteristics were descriptively summarized. A trial was considered mixed-sex if inclusion criteria indicated that the intent was to enroll both males and females, even if one sex or gender were not enrolled or included in analyses. For mixed-sex trials, we calculated the proportion of females pooled across all studies and then stratified by funding source, intervention type, study design, type of substance targeted by the intervention and year of study completion.

Next, we descriptively reported female:male enrollment using medians, means and interquartile ranges represented in box-plots, stratified by trial characteristics including intervention type, substance type and funding source, as we anticipated that government-funded trials might have greater representation or better reporting due to National Institutes of Health mandates.

To account for underlying disease epidemiology, we calculated a participation to prevalence ratio (PPR) using a method employed in the cardiology clinical trial literature to examine female representation in clinical trials [19, 23]. This measures the percentage of females among trial participants, divided by the percentage of females affected by a given SUD. If this ratio is greater than 1, it implies that a greater share of females enroll in trials compared to population epidemiology of the disease. Other literature has considered a PPR of > 1.2 to indicate over-representation, while a PPR < 0.8 indicated under-representation [24]. We demarcate these thresholds, although note that they do not definitively define over-/under-representation. Data on the population disease prevalence was generated from the National Survey on Drug Use and Health for each year of trial completion [25]. Consistent with the reporting of disorders in the survey, we collapsed the substances variable to match reported categories of disease prevalence: alcohol, nicotine and illicit drug use disorders. This allowed, for example, all studies involving alcohol to be matched to prevalence rates of alcohol use disorder. Studies including any illicit drug use (e.g. heroin, cannabis, methamphetamines) or multiple substances not associated with a specific use disorder were included in the illicit use disorder category. Confidence intervals (CI) were generated to adjust for sample variance. All analysis activities were conducted using Statistical Analysis Software (SAS) version 9.4 (Cary, NC, USA).

RESULTS

After reviewing 787 clinical trials for inclusion, 316 trials were retained in the analysis (Figure 1). Most trials were excluded for enrolling fewer than 30 participants (n = 146) or being conducted outside the United States (n = 101). Initially, 230 trials had no known published results at the time of review. The outreach resulted in contact with 127 of 230 investigators (response rate: 55%). From this outreach, we obtained data for 89 trials, and the remaining 141 were excluded because the investigator indicated that there were no data or no response was received. Of the 316 included trials, 274 enrolled both sexes, 30 were female-only trials and 12 were male-only trials. Of note, one male-only trial was designed to recruit females, but restricted inclusion to males after being unable to recruit females. Some mixed-sex trials enrolled no females but remain in this category as they intended to recruit females and males. Characteristics of trials designed to enroll only one sex or gender are reported in Supporting information, Table S1.

FIGURE 1.

Consolidated Standards of Reporting Trials (CONSORT) diagram of studies included in analysis

The characteristics of the 274 mixed-sex trials are displayed in Table 1. The most common funding source was the government: 73% as the sole funder, 3% with government as a joint funder. Approximately half (51%) of interventions were behavioral, and 38% tested pharmaceutical interventions. Ninety-one per cent of trials used randomized designs with individual-level randomization. Trials focusing on nicotine use disorder represented 20% of trials, 15% were for alcohol use disorder, and the remaining 65% on illicit drug use disorders. Only 7% of trials included minors and 22 of 274 trials (8%) reported any sex-specific analysis. Four studies reported on the inclusion of transgender participants, although they did so using varying levels of specificity: one reported sex/gender as one category, one reported biological sex only, one reported gender identity only and one reported the number of transgender participants without specifying transgender female or transgender male participants. For this last trial, transgender participants were subtracted from the total analytical sample and the number of male and female participants were reported.

TABLE 1.

Basic trial characteristics of 274 mixed-sex clinical trials, 2010–19

Trial Characteristic	Total % (n)
Funding source
Government	73% (200)
Internal funding	11% (31)
Industry	9% (25)
Multiple sponsors (including NIH)	3% (8)
Research institute	2% (6)
Multiple sponsors (excluding NIH)	1% (4)
Intervention type
Behavioral	51% (139)
Drugs	38% (103)
Multiple interventions	7% (19)
Other	4% (10)
Devices	1% (3)
Study design
Randomized controlled trial with individual randomization	91% (250)
Single-arm trial	4% (12)
Quasi-experimental (non-randomized design)	3% (7)
Multi-level randomization	1% (3)
Cluster-randomized trial	1% (2)
Substance type
Nicotine	20% (55)
Alcohol	15% (42)
Illicit drugsa	65% (177)
Age
Adults only	93% (254)
Includes ages < 18	7% (20)
Trial Size
Quartile 1 (30–54)	24% (67)
Quartile 2 (55–102)	26% (70)
Quartile 3 (103–229)	25% (69)
Quartile 4 (≥ 230)	25% (68)

^aIllicit drugs include multiple drugs (n = 87 trials), opioids (n = 42), stimulants (n = 35), cannabis (n = 11) and other (n = 2). These categories have been collapsed to reflect National Survey on Drug Use and Health reporting categories across all study years.

NIH = National Insitutes of Health.

Table 2 reports the proportion of female participants in mixed-sex trials by selected trial characteristics. Overall, 40% of participants among all 274 trials including 57 544 individuals were female. Female participation among funding sources ranged from 23% (research institutes) to 41% (government). Trials involving multiple interventions had the highest rates of female participation at 55%, with the lowest rates in trials administering drug interventions (35%). Females represented 35% of participants in trials targeting illicit drug use disorder, 52% in those involving nicotine use disorder and 29% in alcohol use disorder studies. Additional data on the distribution of female:male participation by intervention type, substance type and funding source can be found in the Supporting information, Figure S1. Results for each use disorder adjusted by underlying use disorder prevalence are described in detail below. Detailed tables reporting the calculation of PPR by year for each use disorder are shown in Supporting information, Table S2 and summarized in Figure 2.

TABLE 2.

Representation of female participants in mixed-sex clinical trials by trial characteristics (n = 57 544)

	Female participants % (n)
Overall	40% (23035)
Funding source
Government	41% (19215)
Internal funding	40% (1502)
Industry	37% (1619)
Multiple sponsors (including NIH)	35% (375)
Research institute	23% (209)
Multiple sponsors (excluding NIH)	40% (115)
Intervention type
Behavioral	39% (13923)
Drugs	35% (4201)
Multiple interventions	55% (3703)
Other	37% (920)
Devices	53% (288)
Study design
Randomized controlled trial with individual randomization	38% (19829)
Single-arm trial	45% (576)
Quasi-experimental (non-randomized design)	43% (911)
Multi-level randomization	28% (169)
Cluster-randomized trial	77% (1550)
Type of substance (targeted by intervention)
Illicit drugs	35% (11322)
Nicotine	52% (10098)
Alcohol	29% (1615)
Age
Adults only	39% (20762)
Includes ages < 18	56% (2273)
Year of trial completion
2010	40% (805)
2011	42% (1412)
2012	31% (2195)
2013	42% (3098)
2014	46% (4578)
2015	49% (3602)
2016	36% (2103)
2017	33% (2438)
2018	36% (1472)
2019	46% (1332)

NIH = National Insitutes of Health.

FIGURE 2.

Participation to prevalence ratios for each use disorder by year of trial completion. Dots = Mean participation to prevalence ratio (PPR); whiskers = 95% confidence intervals for the mean. The dashed lines denote the thresholds used by prior literature as cut-offs to represent under- or over-representation.

Note: There were no alcohol use disorder trials reported as completed in 2015

Alcohol use disorder (AUD)

Forty-two mixed-sex trials administered interventions for AUD (15%), with trial enrollment ranging from 30 to 614 participants (mean = 135). Of the 5656 participants in these trials, 29% were female. Examining the participation to prevalence ratio by year (Figure 2a), the median PPR ranged from 0.27 in 2016 to 1.58 in 2010. In 2017, the mean PPR was 0.52 (95% confidence interval for the mean 0.13, 0.91), suggesting possible under-representation of females. No trials were reported as completed in the year 2015. Of note, all median PPRs greater than 1 occurred before 2014, and all median PPRs after 2015 were less than 0.65, suggesting relative under-representation of females in later years. However, these are not definitive, as wide confidence intervals suggest uncertainty in these estimates due to the small numbers of trials within each year.

Nicotine use disorder (NUD)

Fifty-five mixed-sex trials (20%) focused on interventions for NUD, with 52% of 19 524 participants identified as female. Mean trial size in NUD trials was 355 participants (range = 30–2538). The median PPR ranged from 0.75 in 2018 to 1.13 in 2011. All the confidence intervals for mean PPRs in nicotine use disorder spanned 1, suggesting that there was no significant difference in proportionate enrollment when accounting for uncertainty in the PPR point estimate (Figure 2b).

Illicit drug use disorder

Throughout all years, 177 trials focused upon illicit drug use disorders (65%), with a mean trial enrollment of 183 (range = 30–1285). A large number of trials in this category is probably a function of including an array of use disorders and those that addressed multiple use disorders. Overall, 32 364 participants were enrolled into these trials, and 35% were female. Median PPRs ranged from 0.67 (2013) to 1.33 (2012) (Figure 2c). Mean PPRs suggest that women were enrolled at a rate relatively proportionate to disease epidemiology per the National Survey on Drug Use and Health, except for trials in 2013 (mean PPR = 0.74, 95% CI = 0.59, 0.88) and 2016 (mean PPR = 0.79, 95% CI = 0.63, 0.95) where females represented proportionately fewer participants relative to what the epidemiology of SUD would predict.

DISCUSSION

In a review of clinical trials for alcohol, nicotine and illicit substance use disorders that were registered in the National Library of Medicine’s clinicaltrials.gov registry and completed between 2010 and 2019, we found that females were enrolled at lower rates than males overall. After adjusting for the underlying population prevalence of SUDs, females were under-represented in alcohol use disorder studies, while there was no significant difference in enrollment for nicotine and illicit drug use disorders. Only 8% of all trials reviewed reported any sex-specific analyses and 1.5% reported transgender participants. Below we explore the implications of these findings for the conduct of clinical trials involving people with SUDs.

This study is not the first to document that females are enrolled in clinical trials at lower rates than their male counterparts, including in trials involving SUDs. The National Institute on Drug Abuse-funded Clinical Trials Network previously documented rates of female enrollment from 2000 to 2010 at 41% among 24 clinical trials [20]. This is similar to the overall rate of female enrollment in this study (40%), although across use disorders we observed variation in representation. The Clinical Trials Network has been a leader in advancing the inclusion of females in addiction-related research, not only publishing on the results of studies’ gender-specific findings, but also in developing educational material to guide researchers to be more inclusive of females and women [26]. Given the efforts within the National Insitute on Drug Abuse to enhance the inclusion of females within clinical research, and the increasing prevalence of SUDs among females in the past 10 years, we might expect to see increasing rates of female participation over time, particularly in government-funded trials. However, our findings did not suggest increasing rates of female participation in SUD-related trials, either overall and when accounting for underlying disease prevalence. In fact, female participation in AUD research appeared to decrease. Further outreach, advocacy and dissemination of the importance of sex and gender representation and reporting to the scientific advancement of SUDs and their treatment are warranted.

The InWomen project, a multi-disciplinary group of more than 200 researchers and advocates, has identified the ongoing need to improve terminology in the literature related to both sex (a biological categorization) and gender (an identity) through their extensive collation of literature on women with SUDs [27]. In our review of recent clinical trials, sex and gender were often used interchangeably, or definitions were not made explicit. This was further highlighted by only four of 274 studies reporting the number of transgender participants. In one of those studies, there was simply a category ‘transgender’, a label that alone can be stigmatizing and exclusionary to those who identify with either a specific gender or as non-binary. In many of the other trials, the proportion of ‘women’ was reported, although whether this represented a biological or social construct was not defined. For research to advance its understanding of the unique contribution of sex and gender to biological, social, psychological and cultural outcomes, more transparent and detailed reporting is critical. The National Institutes of Health has made clear standards for reporting, supported by guidance for inclusive and accurate data documentation [28, 29], yet our study finds that these standards are infrequently met. While this study was conducted in the United States, future research might explore trends in female participation and sex- and gender-based reporting in global settings, where disease epidemiology, treatment availability and research infrastructure may vary. Similar reporting requirements are being implemented across European clinical trials as of 2018 as defined in Regulation EU No. 536/2014 [30], the effects of which are yet to be determined. These regulations also establish a mandatory clinical trial database in Europe, which was recently implemented in 2021, and should facilitate opportunities to assess female participation across a broader geographic area.

The relative rarity of reported sex- or gender-specific analyses within this group of clinical trials was lower than expected, given recent policy changes. Eight per cent (n = 22) of the trials reported any sex-specific analyses, and of these only nine occurred after the 2015 National Institutes of Health changes in mandated reporting of sex/gender analyses. There are known sex-based, biological female–male differences that merit consideration in clinical trial design. For example, females metabolize alcohol differently than males and demonstrate more signs of impairment with equivalent consumption [31]. Our data show under-representation of females in trials for alcohol use disorder. However, only two studies reported on sex-specific effects in trials focused upon alcohol behaviors. These findings suggest either publication bias, where data are available but not reported; or alternatively, a lack of consideration of sex and gender effects in the study of AUD interventions. This is aligned with a recent systematic review that demonstrated consistently poor sex- and gender-related reporting in treatment for alcohol use disorder over time [21]. Given the known sex- and gender-related factors that contribute to differences in both the development of an AUD and engagement in treatment, study design, data collection and analyses should consider these as potential mediators and moderators of intervention effects. Such steps could help to identify key biological or social differences that influence the treatment and outcomes of AUD and SUD through building a body of knowledge generated from clinical research [21, 32].

There are limitations to this study. Because inclusion criteria specified the reporting of results, there is the potential for publication bias. To combat this, the study team made every effort to identify reported findings for each study, including individual outreach to sponsors and investigators, and allowed for the inclusion of studies that had results posted on clinicaltrials.gov. Also, due to delays between the completion date of a study and results publication, studies completed more recently are more likely to be excluded based on having no published findings. This is potentially exacerbated by data and publication delays prompted by the COVID-19 pandemic in 2020. By excluding these more recent studies, the number of studies reporting sex-specific analyses may represent an undercount of what will appear in the literature over time. We also excluded small trials (< 30 participants) and non-US-based trials. The relative enrollment of women in smaller trials may differ from that of larger trials, and generalizability to pilot and feasibility studies should be limited given the exclusion of such studies in this analysis. The magnitude of potential bias generated from these exclusion criteria is unknown. The use of PPRs was intended to provide some insight beyond crude recruitment rates by accounting for the underlying representation of females in the population with specific use disorders. There is inherent uncertainty in the National Survey on Drug Use and Health rates used for this calculation, and there are no established criteria for over/under-representation. We cite guidelines used in the cardiology literature, but further methodological development to reliably measure over/under-representation is warranted. The reliance upon reported data without access to primary data collection instruments did not allow for separate analyses of sex and gender as unique constructs. Finally, studies were not assessed for quality, and therefore no conclusions about the rigor and reproducibility of these trials or the relationship between quality and participation can be drawn.

CONCLUSION

This retrospective review of clinical trials delivering interventions for SUD demonstrated that there is variation in female participation across use disorders. Factors influencing representation are not characterized here, although it is known that females enter SUD treatment at rates that are relatively lower than males [33], which may also contribute to willingness to participate in SUD intervention research. Improvements in efforts to include females, especially in alcohol-related clinical trials, are necessary but not sufficient to address possible outcome disparities. More importantly, better reporting of sex- and gender-specific effects across all SUD types is urgently needed. As the sex/gender differences in SUD prevalence decline, identifying how to optimize trials that seek to reduce SUD harms such that outcome disparities do not widen as a result of lack of female recruitment and/or lack of reporting sex- and gender-based effects of interventions is a critical issue for public health. Enhancing and disseminating best practices for both conducting research that is inclusive of females and women and reporting sex- and gender-specific effects offers opportunities to improve the treatment of substance use disorders.

Supplementary Material

Supplementary Materials

Figure S1: Female to male enrollment ratios for 274 mixed-sex clinical trials, stratified by selected trial characteristics

Table S1: Basic Characteristics of Single-Sex Clinical Trials, 2010–2019

Table S2: Trial Level Participation to Prevalence ratio (PPR) by Substance Use Disorder Type and Year