Analysis of Female Participant Representation in Registered Oncology Clinical Trials in the United States from 2008 to 2020

Abstract

Background

Female underrepresentation in oncology clinical trials can result in outcome disparities. We evaluated female participant representation in US oncology trials by intervention type, cancer site, and funding.

Materials and Methods

Data were extracted from the publicly available Aggregate Analysis of ClinicalTrials.gov database. Initially, 270,172 studies were identified. Following the exclusion of trials using Medical Subject Heading terms, manual review, those with incomplete status, non-US location, sex-specific organ cancers, or lacking participant sex data, 1650 trials consisting of 240,776 participants remained. The primary outcome was participation to prevalence ratio (PPR): percent females among trial participants divided by percent females in the disease population per US Surveillance, Epidemiology, and End Results Program data. PPRs of 0.8-1.2 reflect proportional female representation.

Results

Females represented 46.9% of participants (95% CI, 45.4-48.4); mean PPR for all trials was 0.912. Females were underrepresented in surgical (PPR 0.74) and other invasive (PPR 0.69) oncology trials. Among cancer sites, females were underrepresented in bladder (odds ratio [OR] 0.48, 95% CI 0.26-0.91, P = .02), head/neck (OR 0.44, 95% CI 0.29-0.68, P < .01), stomach (OR 0.40, 95% CI 0.23-0.70, P < .01), and esophageal (OR 0.40 95% CI 0.22-0.74, P < .01) trials. Hematologic (OR 1.78, 95% CI 1.09-1.82, P < .01) and pancreatic (OR 2.18, 95% CI 1.46-3.26, P < .01) trials had higher odds of proportional female representation. Industry-funded trials had greater odds of proportional female representation (OR 1.41, 95% CI 1.09-1.82, P = .01) than US government and academic-funded trials.

Conclusions

Stakeholders should look to hematologic, pancreatic, and industry-funded cancer trials as exemplars of female participant representation and consider female representation when interpreting trial results.

Female enrollment in clinical trials should adequately reflect their disease burden in the general population. This article evaluates female participant enrollment trends in U.S. oncology clinical trials, identifying associations between proportional female participant representation and trial intervention type, cancer site, and funding.

Implications for Practice.

Our cross-sectional study of 1650 US oncology trials registered on ClinicalTrials.gov from 2008 to 2020 showed female participants were underrepresented compared to their disease burden in surgical oncology, thyroid, bone/joint, kidney, bladder, stomach, and anal cancer trials. They were proportionately represented in medical and radiation oncology, industry-funded, hematologic, and pancreatic cancer trials. Female participant underrepresentation in oncology trials limits the generalizability of results and access to cutting-edge therapeutics. Our study highlights lessons learned from trial types that have achieved proportional female recruitment and retention, and spurs clinicians to interpret trial findings in the context of their characteristics and associations with female representation.

Introduction

Females have historically been underrepresented in biomedical research, including oncology clinical trials.^1-3 In 2002, a widely cited study found that females comprised only 34.7% of participants in cancer prevention and treatment trials from 1990 to 2001.⁴ A 2013 updated study revealed continued sex gaps from 2000 to 2010.⁵ The persistent failure to proportionally enroll female participants relative to their disease burden hinders the generalizability of cancer clinical trial findings and denies females access to novel therapeutics and potentially improved survival through trial participation.

As defined by the National Institutes of Health’s (NIH) Office of Research on Women’s Health (ORWH), “sex” refers to biological differences between females and males, including chromosomes, sex organs, and endogenous hormonal profiles. “Gender” refers to socially constructed and enacted roles and behaviors within the groups traditionally thought of as women and men, which occur in a historical and cultural context and vary across societies and over time.⁶ In their efforts to encourage researchers to consider both sex and gender in their work, the NIH ORWH has developed a series of resources demonstrating how sex and gender can influence health.⁷ For example, those with a biological sex of female are more likely than males to injure their knees when playing sports due to differences in knee and hip anatomy, imbalanced leg muscle strength, and looser tendons and ligaments, with knee injuries ultimately increasing a female’s risk of knee osteoarthritis. Similarly, those with a gender of girl/woman are more likely to walk in high-heeled shoes which stresses the knee joint and can ultimately increase women’s risk of knee osteoarthritis.⁸ Similar examples have been shown in mental health, cardiovascular health, and smoking cessation,⁸ suggesting that further research is needed to understand the interplay of these factors in other areas of health, such as cancer.

Epidemiologic data already suggest that sex plays a vital role in cancer incidence, prognosis, and mortality.⁹ Data from the National Cancer Institute’s Surveillance, Epidemiology, and End Results Program (SEER) from 2014 to 2016 indicate the lifetime probability of being diagnosed with an invasive cancer is slightly lower for females than males.¹⁰ However, females suffer from lower cancer-specific 5-year survival rates.¹¹ Although overall cancer mortality has declined over the past 20 years, females trail behind males in mortality reduction.¹² Furthermore, sex may modulate cancer therapy efficacy and tolerability.^13-16 Recent research demonstrates sex-specific differences in response to systemic therapy agents,¹⁶ radiotherapy,¹⁷ and immunotherapy.¹⁸ Females may be at higher risk of developing adverse drug reactions and toxicities.^13,19 Outcome variability may reflect differences in environmental exposures,^20,21 endogenous hormones,²² access to care,²³ screening practices,²⁴ health literacy,²⁵ and complex interactions among these factors. Sufficient female enrollment for sex-specific analyses is necessary to better understand underlying mechanisms and develop solutions to sex-specific disparities.

Female enrollment in clinical trials should adequately reflect their disease burden in the general population. Although organizations such as the NIH have recently stated the need for female representation, the current state of sex bias within United States (US) oncology clinical trials remains unclear.²⁶ Therefore, we aimed to characterize female participant enrollment trends in US oncology trials registered on ClinicalTrials.gov from 2008 to 2020 and identify associations between proportional female participant representation and trial intervention type, cancer site, and funding.

Materials and Methods

We conducted a cross-sectional study of oncology interventional clinical trials registered on ClinicalTrials.gov between January 10, 2008 and September 3, 2020. Trials registered prior to 2007 were excluded to coincide with the Food and Drug Administration (FDA) Amendments Act, requiring most clinical trials involving FDA-regulated products and devices to submit registration.

Data Sources

All data were extracted via the Aggregate Analysis of ClinicalTrials.gov (AACT) database from publicly available records for all interventional studies (the term used by ClinicalTrials.gov for clinical trials) submitted to ClinicalTrials.gov from January 10, 2008 to September 3, 2020. Medical subject heading terms were applied to identify clinical trials relevant to oncology, using previously published protocols.^27-31 Trial title, abstract, description, and inclusion/exclusion criteria were manually reviewed to verify content and exclude non-oncology studies. Trials with a focus on reproductive or sex-specific organ cancers, incomplete study status, non-US location, or no report of participant sex data were excluded. As trials reported sex and not gender, we only explored patterns in female participant representation as opposed to women participant representation. Institutional Review Board approval/oversight was not required given publicly available data.

Variables

Oncology Intervention Type and Cancer Site

Oncology intervention type categories included: (1) radiation, (2) surgical, (3) other invasive (non-surgical and non-radiation procedures), (4) medical, and (5) other (physical therapy, counseling, or non-FDA approved therapies like traditional Chinese medicine). If a trial examined one intervention type, that type was assigned. If a trial compared 2 intervention types in separate arms, both types were assigned. If a trial compared 2 or more intervention types, only those that varied among arms were assigned. If a trial examined multiple intervention types in a single arm, all involved types were assigned.

Cancer site was stratified into lung, brain and other nervous system (central nervous system [CNS]), hematologic, skin/melanoma, thyroid, bone/joint, head/neck, other soft tissue, colorectal, anal canal, stomach, liver/biliary tree, pancreas, esophagus, breast, cervix/uterus, ovaries, vulva/vagina, prostate, testicle/penis, kidney/renal pelvis, bladder, other, or unknown site. If a trial focused on multiple sites, all sites were assigned. If a trial focused on more than 7 sites or if no specific site was specified, “other” was assigned. If a trial mentioned a primary and metastasis site, both sites were assigned. If a trial mentioned the metastasis site but not the primary, or vice versa, only the explicitly mentioned site was assigned.

Sixteen medical student and resident physician researchers manually sorted trials into the above categories. All labelers achieved 90% or higher agreement in categorizing a training set of trials (correct categorization was pre-determined by co-senior author B.T.) before proceeding to the trials of interest. To ensure agreement, each labeler reviewed a subset (at least 10%) of another’s categorizations.

Clinical Trial Characteristics

Twelve trial characteristics from the ClinicalTrials.gov database were extracted for analysis, including: (1) funding source, (2) phase, (3) number of participants, (4) submission date, (5) number of sites, (6) use of randomization, (7) oversight by a data-safety monitoring committee (DMC), (8) blinding, (9) recruitment status, (10) number of arms, (11) reported results, and (12) early discontinuation. “Early discontinuation” was noted if a study was defined by ClinicalTrials.gov as “terminated,” “withdrawn,” or ‘‘suspended.” Only trials completed on or before September 30, 2017 were included in this portion of analysis to coincide with requirements to report results up to 3 years after completion. The ClinicalTrials.gov Protocol Registration Data Element Definitions for Interventional and Observational Studies was reviewed to understand missing automated data and labeling.³²

Funding source was divided into 3 subcategories: (1) industry, (2) US government, and (3) academic, using the National Library of Medicine designations of sponsors and collaborators. “Industry” was defined as sponsorship/partnership by an industry agency. “U.S. government” was defined as sponsorship/partnership by the NIH or other federal agencies. The remaining trials represented sponsorship by academic institutions and, less so, by individuals, foundations, research institutes, and community-based associations. The heading “academic” was chosen due to prior analyses’ findings that 90.1% of funding sources in this category was defined by US legal code as academic institutions.^33,34

Exposures and Outcomes

Primary exposure variables were cancer site, intervention type, and funding source. Additional analyses compared trial characteristics including DMC presence, randomization, blinding, and status of results reported.

Primary outcome was proportional representation of female participants. Disease prevalence-corrected estimates for female participation were calculated as the percentage of females among trial participants divided by the percentage of females in the disease population per the November 2019 submission of the US SEER*Stat database (containing cancer prevalence data from 2000 to 2017).³⁵ For example, if 50% of females participated in a given clinical trial focusing on lung cancer and SEER data reported the disease prevalence for lung cancer in the United States as 50% female, the disease prevalence-corrected estimate would be 50 divided by 50, or 1. This calculation has been termed “participation to prevalence ratio” (PPR), with the range of 0.8-1.2 reflecting proportional or adequate female trial representation, as established in prior published protocols.^36-41

Statistical Analysis

Statistical analyses were performed using R version 4.4.1 (r-project.org). Data extracted from the AACT database tool were merged with manually labeled trial characteristics using the National Clinical Trial identification number. 2017 SEER database cancer sites were matched to manually labeled trial cancer types. Participation to prevalence ratios were calculated for each cancer type that a trial included. As trials could be classified with multiple cancer or intervention types, these categories were not mutually exclusive. Therefore, category analyses resulted in a single trial’s data repeated for each relevant intervention type. Descriptive statistics were calculated. Repeated measures general linear models were fit using the “ImerTest” package in R to test the association between trial characteristics and adequate representation. P values were calculated using Welch’s modified t test from the “Basic Statistics and Data Analysis (BSDA)” package in R. Statistical significance was set at P value < .05.

Results

After exclusion of ineligible trials, 1650 trials (Table 1) composed of 240,776 participants met inclusion criteria and reported participant sex (Fig. 1). Females represented 46.9% of all participants (95% CI, 45.4-48.4), with an overall PPR of 0.912.

Table 1.

Trial characteristics of all 1650 oncology trials with associated mean participation to prevalence ratios (PPR) and standard deviation (SD).^a

	Number of trials	Mean PPR (SD)	Female proportion (%)
Trial site
Lung	154	0.84 (0.16)	47
CNS	124	1.03 (0.21)	48
Heme	406	0.91 (0.13)	41
Skin/melanoma	102	0.88 (0.21)	40
Thyroid	17	0.58 (0.19)	51
Bone/joint	11	0.79 (0.21)	41
Head/neck	73	0.80 (0.16)	23
Soft tissue	41	1.05 (0.19)	45
Colorectal	98	0.90 (0.18)	46
Anal canal	4	0.21 (0.14)	22
Stomach	29	0.68 (0.14)	29
Liver	61	1.01 (0.17)	36
Pancreas	73	0.88 (0.11)	45
Esophagus	23	1.05 (0.15)	23
Kidney/renal pelvis	38	0.77 (0.14)	29
Bladder	29	1.00 (0.16)	20
Intervention type
Radiation	196	0.83 (0.49)	42
Surgical	76	0.74 (0.25)	44
Invasive	40	0.69 (0.35)	38
Medicine	1948	0.82 (0.16)	46
Other	138	0.88 (0.33)	50
Phase
I	10	1.39 (0.88)	52
II	1380	0.89 (0.37)	66
III	48	0.98 (0.37)	45
IV	18	0.83 (0.45)	54
Other	194	0.90 (0.45)	51
Funding source
Industry	693	0.92 (0.35)	47
Academic	474	0.95 (0.38)	46
US Government	483	0.94 (0.43)	47
Blinding
None/open-label	1439	0.90 (0.39)	45
Single	51	0.90 (0.51)	56
Double	76	0.88 (0.39)	51
Triple	36	0.90 (0.26)	55
Quadruple	45	0.90 (0.40)	52
NA	3	0.59 (0.25)	24
DMC
Yes	921	0.90 (0.41)	45
No	566	0.89 (0.39)	47
NA	163	0.90 (0.32)	51
Randomized
Yes	460	0.90 (0.41)	49
No	1184	0.90 (0.39)	45
NA	6	1.28 (0.63)	32
Results reported
Yes	1504	0.90 (0.39)	47
No	146	0.88 (0.38)	44

^aDescription: trial characteristics of all 1650 trials. Proportional or adequate representation is defined as a PPR of 0.8-1.2. Since many trials focused on more than one cancer site and/or intervention type, the number of trials for these categories total greater than 1650.

Abbreviations: DMC, Data Safety Monitoring Committee; NA, not applicable.

Figure 1.

Flow diagram of clinical trials included in the analysis.

Representation by Cancer Site

Female participant representation varied by cancer site. Females were underrepresented compared to their population disease burden in anal canal (PPR 0.21, 4 trials), thyroid (PPR 0.57, 17 trials), stomach (PPR 0.68, 29 trials), kidney/renal pelvis (PPR 0.70, 38 trials), and bone/joint (PPR 0.79, 11 trials) cancer trials. Females were proportionately represented in head/neck (PPR 0.80, 73 trials), lung (PPR 0.84, 154 trials), bladder (PPR 0.85, 29 trials), skin/melanoma (PPR 0.88, 102 trials), pancreas (PPR 0.88, 73 trials), colon (PPR 0.90, 98 trials), hematologic (PPR 0.91, 406 trials), liver (PPR 1.01, 61 trials), CNS (PPR 1.03, 124 trials), soft tissue (PPR 1.05, 41 trials), and esophagus (PPR 1.05, 23 trials) cancer trials (Table 1). Females were not over-represented in any cancer site trial group (Fig. 2a, 2b).

Figure 2.

(a) Female representation as quantified by participation to prevalence ratio (PPR) for each cancer site. Description: horizontal lines represent the threshold and range for adequate representation of females defined as a PPR of 0.8-1.2. Vertical error bars represent 1 standard deviation from the mean PPR, as also listed in Table 1. (b) Female Proportion in Surveillance, Epidemiology, and End Results (SEER) Program population compared to female proportion in oncology clinical trials in this study.

When each cancer site was compared to all others, hematologic ([OR] 1.78, 95% CI 1.09-1.82, P < .01) and pancreatic (OR 2.18, 95% CI 1.46-3.26, P < .01) cancer trial focuses were significantly associated with higher odds of proportional female representation (with ORs reflecting the odds of a given trial type having a PPR 0.8-1.2, not to be confused with the PPR itself or indicating odds of having a higher PPR). Bladder (OR 0.48, 95% CI 0.26-0.91, P = .02), head/neck (OR 0.44, 95% CI 0.29-0.68, P < .01), stomach (OR 0.40, 95% CI 0.23-0.70, P < .01), and esophageal (OR 0.40 95% CI 0.22-0.74, P < .01) cancer had lower odds of adequate female representation when compared with all other cancer sites (Table 2).

Table 2.

The association of multiple trial characteristics with adequate representation of females.^a

	Odds ratio (95% CI)	P value
Intervention type
Radiation	0.73 (0.49-1.10)	.13
Surgery	0.77 (0.42-1.43)	.41
Invasive	0.57 (0.23-1.41)	.23
Medicine	1.13 (0.88-1.45)	.32
Other	0.86 (0.54-1.39)	.55
Funding source
Industry	1.41 (1.09-1.82)	.01
Academic	1.27 (0.96-1.67)	.10
Cancer site
Lung	1.35 (0.95-1.91)	.09
CNS	0.85 (0.60-1.19)	.34
Hematologic	1.78 (1.42-2.25)	<.01
Skin/melanoma	0.73 (0.51-1.06)	.10
Thyroid	0.43 (0.17-1.10)	.08
Bone/joint	0.60 (0.20-1.76)	.35
Bladder	0.48 (0.26-0.91)	.02
Head/neck	0.44 (0.29-0.68)	<.01
Soft tissue	0.83 (0.47-1.46)	.52
Colorectal	1.35 (0.95-1.91)	.09
Anus	0.15 (0.02-1.23)	.08
Stomach	0.40 (0.23-0.70)	<.01
Liver	0.66 (0.43-1.02)	.06
Pancreas	2.18 (1.46-3.26)	<.01
Esophagus	0.40 (0.22-0.74)	<.01
Kidney/renal pelvis	0.62 (0.37-1.02)	.06
DMC
Yes	0.94 (0.75-1.18)	.59
Randomization
Yes	1.62 (1.30-2.01)	<.01
Blinding
Single	0.90 (0.52-1.57)	.72
Double	1.26 (0.80-1.99)	.33
Triple	4.33 (1.64-11.41)	<.01
Quadruple	2.13 (1.13-4.01)	.02
Results reported
Yes	1.00 (0.71-1.40)	.99

^aThe reference variable for each intervention type was all other intervention types. The reference variable for each funding source was funding by the United States government. The reference variable for each cancer site was all other cancer sites. The reference variable for data-safety monitoring committee (DMC) was trials without a DMC. The reference variable for randomization was all trials without randomization. The reference variable for blinding was open label trials. The reference variable for results reported was all trials without results reported. Significance is defined as P value < .05.

Abbreviations: CNS: central nervous system; DMC: data-safety monitoring committee.

Representation by Oncologic Intervention Type and Funding Source

Female participants were underrepresented in clinical trials of surgical (PPR 0.74, 76 trials) and other invasive (PPR 0.69, 40 trials) oncology trial types, but proportionately represented in radiation (PPR 0.83, 196 trials), medical (PPR 0.82, 1948 trials), and “other” (PPR 0.88, 138 trials) oncology trial types (Table 1). When compared with each other, oncology intervention types were not significantly associated with proportional representation of females (Table 2).

Less than half of the clinical trials were funded by industry (N = 693, 42%), followed by academic institutions (N = 474, 29%) and the US government (N = 483, 29%). Overall female representation was adequate across all funding sources (Table 1). Industry-funded trials had greater odds of proportional female representation (OR 1.41, 95% CI 1.09-1.82, P = .01) compared to government-funded counterparts (Table 2). Academic-funded trials did not show a statistically significant difference in odds of proportional female representation compared to government-funded trials (OR 1.27, 95% CI 0.96-1.67, P = .10; Table 2).

Representation by Trial Methodology

Females were over-represented within phase I trials as PPR was greater than 1.2 (PPR 1.39), and proportionally represented within phase II (PPR 0.89), phase III (PPR 0.98), phase IV (PPR 0.90), and trials without FDA-defined phases (PPR 0.90) (Table 1).

Female representation was proportional within all blinding categories: open-label trials PPR 0.90, single-blinded trials PPR 0.90, double-blinded trials PPR 0.88, triple-blinded trials PPR 0.90, and quadruple-blinded trials PPR 0.90 (Table 1). Compared to open-label trials, adequate representation was significantly associated with triple-blinded (OR 4.33, 95% CI 1.64-11.41, P < .01) and quadruple-blinded trials (OR 2.13, 95% CI 1.13-4.01, P = .02) (Table 2).

Regarding participant randomization, females were proportionately represented in randomized (PPR 0.90) and non-randomized trials (PPR 0.90) (Table 1). Female representation was also proportional in trials with (PPR 0.90) and without (PPR 0.89) DMC oversight (Table 1). The presence of a DMC was not significantly associated with increased odds of adequate representation.

Association Between Representation, Trial Completion, and Results Reporting

Within our already strict inclusion criteria, no trials were discontinued early, and 1504 (of 1650 trials, 91%) reported results. PPR among trials reporting results was 0.90 compared to PPR 0.88 among trials not reporting results. Results reporting did not have statistically significant higher odds of proportional female representation compared to not reporting results (Table 2).

Representation Over Time

The overall number of clinical trials decreased from 1082 (2007-2012) to 568 (2013-2018). Female participant representation remained relatively static between 2007-2012 compared to 2013-2018 (PPR 0.89-0.91, P = .51). Over the same period, female representation remained without statistically significant change in medical (PPR 0.89-0.89, P = .87), surgical (PPR 0.73-0.82, P = .45), other invasive (PPR 0.71-0.75, P = .84), and radiation (PPR 0.89-0.83, P = .55) oncology trials. Female representation also remained stable in ­industry-funded (PPR 0.89-PPR 0.88, P = .52), academic-funded (PPR 0.92-0.93, P = .84), and US government-funded (PPR 0.87-0.91, P = .86) trials.

By cancer site, trials displayed statistically significant increases in female representation in stomach and esophageal cancer (PPR 0.55-0.86, P = .02 and PPR 0.64-1.52, P = .01, respectively). Trials in bone/joint and bladder cancer demonstrated non-statistically significant decreases in female representation (PPR 1.02-0.82, P = .34; PPR 0.92-0.79, P = .45, respectively).

Discussion

Despite decades of national efforts to increase female inclusion in US clinical trials and the known importance of proportionate representation for female health,^2,26 our study demonstrates continued female participant underrepresentation in certain oncology clinical trials. Although overall female enrollment in cancer clinical trials has increased compared to previous studies,^4,5 sex bias persists within trial subtypes. Specifically, females were underrepresented in surgical and other invasive oncology trials, with statistically significant sex gaps in bladder, head/neck, stomach, and esophageal cancers. Conversely, females were adequately represented in industry-funded, hematologic and pancreatic cancer trials.

The mean proportion of females enrolled in oncology trials in this study was 46.9% (95% CI, 45-48.4), modestly higher than the roughly 41% reported in 2 contemporary studies.^42,43 This demonstrates an upward trend in oncology clinical trial female enrollment, with previous studies documenting 34.7% enrollment in the 1990s and 38.6% in the 2000s.^4,5 However, our study reveals variations in female representation still exist among oncology intervention types, cancer sites, and funding sources, consistent with studies across other specialties: female enrollment varies among individual diseases and funding sources.^44,45 This lack of uniform proportional female representation challenges the ability of the medical field to move toward a culture of inclusivity and develop evidence-based recommendations to address disease-specific disparities.

Disease prevalence estimates in the general population were used to evaluate the number of females in the trial compared to the number of females that should be enrolled, as opposed to simply providing a percentage of female participation without disease burden contextualization. According to these prevalence-corrected estimates, recent oncology trials overall were within the defined range for adequate representation of female participants, trending toward increased equity. However, deeper analysis is vital because overall adequate representation can result from over-representation in one area coupled with under-representation in another.^46,47 Overlooking sex-specific disparities within certain trial types or cancer types can negatively affect the millions of females living with cancer in the United States

Analysis by clinical trial type revealed that female representation was below the prevalence estimate for surgical and other invasive oncology intervention types, despite improvement since 2008. Females were also underrepresented in bladder, head/neck, stomach, and esophageal cancer trials, although representation in the latter 2 groups appears to be gradually improving. Improvements in representation are crucial as underrepresentation can contribute to outcome disparities. Previous studies demonstrate lower 5-year survival rates for females than males in anal canal and bladder cancer, with the latter seen especially following cystectomy.^48-50 Our study highlights female underrepresentation in bladder cancer relative to disease burden overall; improving representation in bladder cancer clinical trials could help evince strategies for reducing this mortality gap. Concerning bladder cancer particularly, one study cites a lack of targeted recruitment effort toward female participants despite well-known historical trial enrollment disparities, but does not provide reasoning on why this occured.⁵¹ In non-sex-specific surgical oncology, a general paucity of research,⁵² coupled with limited funding, training, and support for surgical clinical investigators specifically⁵³ may exacerbate female trial participation disparities. Previous studies on barriers to clinical trial participation have cited fear, transport difficulties, economic considerations, and interference with family responsibilities, with such barriers often disproportionately affecting women, as reasons for female participant underrepresentation.^54-56 There is a dearth of information to confidently identify the causes of participation disparities of females among non-sex-specific esophageal and stomach cancer trials, although it likely results from a combination of factors described above. Further research to understand the root causes of such disparities is necessary to aid the development of targeted solutions.

One potential solution to increase female participation and retention in trials would be the use of a skilled clinical trial team member to assist potential trial participants in identifying and negotiating barriers to care/participation. A 2020 pilot study found that trial participants heavily utilized these “trial navigators” to assist with logistical, housing, and transportation planning, as well as to provide referrals to relevant social, financial, and medical support systems that enabled their sustained trial participation and well-being during the trial.⁵⁷

Clinical trials with industry funding or in hematologic and pancreatic cancer demonstrate these barriers can be overcome and can serve as potential models for fields with sex enrollment disparities. Addressing the above concerns with potential participants represents a path toward increased female recruitment and retention in oncology trials.

Of all funding sources, industry-funded trials most adequately represented female participants. This may speak to greater success in FDA regulations for nearly all industry-funded trials compared to NIH regulations for most government and academic-funded trials not overseen by the FDA.^58,59 Interestingly, the FDA does not mandate the participation of females in clinical trials, whereas the NIH has required female participation since 1986.⁶⁰ Although both organizations attempted to bolster female enrollment through various advocacy and outreach programs throughout the 1990s and 2000s, in 2013, the FDA alone was congressionally directed to investigate female participation in clinical trials and develop a plan for addressing barriers to female enrollment, which included increasing efforts to enroll cohorts who reflect disease prevalence and incidence.^61-64 Our study’s finding of increased odds of proportional female representation in industry-funded trials compared to US government and academic trials suggests that relying on strict mandates for female participation has limited efficacy; resources may be better spent identifying and addressing root causes of sex representation imbalance.

The only trial types with over-representation of female participants were phase I trials (PPR 1.4) and trials where randomization status was not applicable or not disclosed (PPR 1.28). This could raise the concern of whether female participants are shouldering an increased burden when it comes to evincing efficacy and side effects of newer/experimental treatments. However, these are also 2 of the smallest trial groups within our entire study, with 10 or less trials each, so we caution against over-interpretation of this trend.

This study has several strengths. Combining 12 years of ClinicalTrials.gov data with manual trial categorization, our study represents an extensive, verified data extraction and can therefore comment on both general and trial-subtype trends. Furthermore, recent research detailing statistically lower female enrollment rates in oncology trials⁶⁵ has drawn responses calling for representation analysis that considers not just enrollment rate, but also the different disease prevalence between males and females.⁶⁶ Our study does exactly that by utilizing PPR instead of percentages alone, rendering its conclusions more generalizable. Many cardiovascular disease studies started utilizing prevalence-corrected estimates in the form of PPR in the 2000s,^37-40 but the field of oncology has only more recently begun to utilize PPR.^46,47 To our knowledge, this study is among the first few in oncology to utilize PPR, hopefully paving the way for continued contextualized analysis. Although some studies in oncology note that incidence better reflects the individuals eligible for frontline therapy and thus they have opted not to use PPR,⁴⁷ our study aimed to examine not only novel therapy trials but also the many trials that focus on the quality of life/supportive interventions for patients already living with cancer and their loved ones, therapies for those who have relapsed, and the entire spectrum of those affected by oncologic disease. As we continue to make gains in life expectancy for patients diagnosed with cancer who continue to need evidence-based interventions and support, this broad spectrum of trial focus is better represented by a prevalence-based tool such as PPR.

This study has several limitations. ClinicalTrials.gov does not include all oncology trials. Despite amended recommendations by both the NIH and FDA, only 1650 of 2893 trials (57.0%) eligible for analysis reported participant sex, which is considerably lower than a similar study of US clinical trials across all specialties.⁶⁷ Of note, only NIH phase III clinical trials are required to report participant sex.⁶⁸ Additionally, our study does not include demographics such as participant age or race/ethnicity, which are often not reported and difficult to extract. Nor did we have access to individual patient-level data that might provide a more granular understanding of differences in sex-specific trial participation. Ideally, future research will investigate this and barriers that limit the participation of other underrepresented groups such as ethnic minorities, the elderly, and non-binary gender identities. As we did not have access to participant gender identity, we cannot comment on trends in the representation of trial participants who identify as women, which is of utmost importance to help further the field’s understanding of how both sex and gender influence cancer care and outcomes. Similarly, sex data were provided by each trial’s reporting investigator without the description of whether patients self-reported their sex or if sex was determined by other means, again limiting any discussion on sex versus gender in our study. Furthermore, our study does not factor in prognosis within each disease category. For example, females have lower 5-year cancer-specific mortality rates than males in stomach and esophageal cancer.⁴⁸ Perhaps this more favorable prognosis justifies the proportional underrepresentation of females in stomach and esophageal cancer trials that our study found. Lastly, ClinicalTrials.gov has several independent limitations in reporting, data change over time, and data entry enforcement.⁶⁹ One example is that clinical investigators are required to report participation only by biological sex rather than in combination with gender; this precludes examination of gender differences distinct from sex differences, a commonplace limitation throughout the literature.⁷⁰

Sex differences should not be ignored when conducting clinical trials, especially with recent studies demonstrating sex differences in cancer risks, treatment response rates, and incidence of adverse drug reactions.⁷¹ Our cross-sectional study shows that contemporary oncology trials underrepresent females compared to their disease burden in fields with sex-specific disparities in outcomes such as surgical oncology, bladder, head/neck, and stomach cancer trials. Fortunately, challenges to improve representation can be overcome, as seen in hematologic, pancreatic, and industry-funded trials, where mandates have been replaced with nuanced reviews and solutions to improve female representation. This data suggest that sex bias persists within oncology trials and calls for targeted and collaborative efforts to address this issue. Clinical researchers, industry, federal agencies, and patient-advocacy groups could cooperate to ensure representative patient populations are enrolled, enhancing our ability to provide informed, relevant, safe, and efficacious recommendations for every patient with cancer.

Funding

This research received no grant from any public, commercial, or not-for-profit funding agency.

Conflict of Interest

The authors have no conflicts of interest, financial or other, to disclose.

Author Contributions

Conception/design: N.D.P., T.R.B., H.K.L., J.R.S., B.E.T., A.M. Provision of study material or patients: J.R.S., B.E.T., A.M. Collection and/or assembly of data: N.D.P., T.R.B., H.K.L., M.Shyu, M.A.S., M.M.W., A.B., M.Saleki, K.L.W, R.I., J.J.T., S.J., R.M.O., M.P., J.M.T. Data analysis and interpretation: N.D.P., T.R.B., W.S., R.G., B.E.T., A.M. Manuscript writing: N.D.P., T.R.B., W.S., H.K.L., M.Shyu, M.A.S., J.R.S. Final approval of manuscript: All authors.

Data Availability

The data underlying this article will be shared on reasonable request to the corresponding author.