INTRODUCTION

Clinical trial diversity is critical for the development of safe, efficacious therapies and interventions to advance health equity, yet non-White populations are historically underrepresented in clinical trials.1 For example, racial and ethnic minority groups comprise 40% of the U.S. population,2 yet 75% of the 32,000 patients that participated in clinical trials which approved 53 novel drugs in 2020 were non-Hispanic White.3 Underrepresentation in clinical trials can potentially harm minoritized populations by inappropriate estimation of racial/ethnic differences in outcomes4,5 which may potentially lead to race/ethnicity being used in clinical decision-making rather than individual considerations.6,7 This approach perpetuates inequities that stem from viewing race/ethnicity as a biological rather than a social construct.8

The reasons behind racial underrepresentation in clinical trials are systematic in etiology, but include medical mistrust, lack of access to healthcare, lack of clinical trial availability, and insurance barriers.9,10 In addition, the limited racial diversity within the clinical trial workforce and lack of racial concordance with patients may contribute to providers being less likely to recruit racial/ethnic minority populations for clinical trial participation.11

New efforts seeking to increase diversity in clinical trials are being prioritized by funding agencies. For instance, the National Institutes of Health (NIH) All of Us Research Program specifically aims to recruit a representative sample, with the goal of building a diverse database that can inform thousands of studies and improve precision medicine.12 The NIH further requires sponsored clinical trials to submit progress reports that document progress toward NIH-approved enrollment plans detailing diversity and inclusion.13,14 The Food and Drug Administration has also been attentive to this concern, prompting such responses as providing guidance that new clinical trials can implement to increase the enrollment of populations underrepresented in research.15 This shift in dialog is a necessary first step in enhancing clinical trial diversity.

While these efforts certainly provide a strong policy push to encourage diverse representation in clinical trials, they may be insufficient to translate to participation at the individual level. Existing literature that describes who participates in clinical trials, and explains why, remains underdeveloped, leading to a potential gap between goals of achieving clinical trial diversity and realities of study enrollment. This present study addresses this gap by examining (1) a snapshot of the demographics of clinical trial participation in the U.S. and (2) the association of race/ethnicity with sources of influence on clinical trial participation.

METHODS

Study Population and Design and Setting

The data used in this study was from the Health Information National Trends Survey (HINTS) 5 Cycle 4, fielded from February 24 to June 15, 2020.16 The HINTS survey collects nationally representative data from adults in the U.S. regarding healthcare utilization and prevention behaviors, health information technology use, and demographics. Complete survey administration methodology is published elsewhere.17 Briefly, HINTS uses a random sample of home addresses in the U.S. selected from a database of residential addresses provided by the Marketing Systems Group. Addresses are grouped into two sampling strata: addresses with high racial/ethnic minority populations (i.e., equal to or over 34% of the census tract identified as Hispanic or Black from the 2014–2018 American Community Survey data file) and addresses with low minority populations (i.e., less than 34% of the census tract identified as Hispanic or Black). The high minority stratum was oversampled to increase the precision of estimates for racial/ethnic minority subpopulations.

For HINTS 5 Cycle 4, all surveys were administered by mail, using the next birthday method to identify the appropriate respondent within a household. Survey procedures follow a modified Dillman approach18 including three mailed surveys and a reminder postcard. HINTS 5 Cycle 4 uses the RR4 formula of the American Association of Public Opinion Research to calculate response rates. This formula is adjusted by the rate of unresolved surveys that are either never returned or undeliverable. In total, 3865 surveys were returned (response rates: 27.2% high minority strata; 40.3% low minority strata; 36.7% overall).

HINTS is approved by the Westat Institutional Review Board and classified as exempt by the U.S. NIH Office of Human Subjects Research Protections because the data is deidentified. Our study follows the reporting guidelines of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE).

Study Variables

Exposure Measures

The primary exposure of interest for this analysis was a respondent’s self-reported race/ethnicity (i.e., non-Hispanic White, non-Hispanic Black, Hispanic, non-Hispanic other). The non-Hispanic other category includes American Indian/Alaska Native, Asian, Native Hawaiian/Pacific Islander, and multiple races selected. These categories were combined given small sample sizes that limited analytic model estimation.

Outcome Measures

HINTS survey questions were used to create outcome variables pertaining to three related topics: clinical trial participation history, clinical trial knowledge, and clinical trial influence.

Clinical Trial Participation History

Invitation to participate in clinical trials was assessed by the question: “Have you ever been invited to participate in a clinical trial” with the responses “Yes,” “No,” and “Don’t know.” For this analysis, “Don’t know” responses were considered missing, and a binary variable created to capture responses of “Yes” or “No” (see Supplementary Appendix eTable 1). Within those who responded “Yes” to being invited to participate, they were then asked “Did you participate in the clinical trial” with the response options of “Yes” or “No,” which were used as a binary variable in this analysis. A three-level summary clinical trial participation history variable was created using responses to these two questions (i.e., not invited, invited but did not participate, and participated).

Clinical Trial Knowledge

All HINTS respondents were asked “How would you describe your level of knowledge about clinical trials” and could answer “Don’t know anything,” “Know a little,” or “Know a lot.” For this analysis, “Know a little” and “Know a lot” were combined into one category, know anything, to form a binary variable to compare knowing anything to knowing nothing about clinical trials. How people get information about clinical trials was assessed with two questions: “Imagine you had a need to get information about clinical trials. Which of the following would you go to first to get information about clinical trials?” and “Imagine you had a need to get information about clinical trials. Which of the following would you most trust as a source of information about clinical trials?” with the choices for both including “My healthcare provider,” “Family and friends,” “Government health agencies,” “Health organizations,” “Disease support groups,” “Drug companies,” and “Internet search.”

Clinical Trial Influence

Eight questions regarding which factors would influence participation in clinical trials were asked to all HINTS respondents. All eight questions had the wording “Imagine that you had a health issue and you were invited to participate in a clinical trial for that issue. How much would ‘X’ influence your decision to participate in the clinical trial?” with the eight influential factors including “Helping other people by participating,” “Getting paid to participate,” “Getting support to participate such as transportation, childcare, or paid time off from work,” “Your doctor encouraging you to participate,” “Your family and friends encouraging you to participate,” “Wanting to get better,” “Getting a chance to try a new kind of care,” and “The standard care not being covered by your insurance.” For all eight of these factors, responses included “A lot,” “Some,” “A little,” and “Not at all” to describe the level of influence these each had on participating.

Covariates

Demographic data included as covariates in the analyses included sex (male, female), age (18–34, 35–59, 50–64, 65–74, 75 or older), education level (high school graduate or less, some college, ≥ bachelor’s degree), income (< $35,000, $35,000–$75,000, ≥ $75,000), residence location (metropolitan area, non-metropolitan area), health insurance status (insured, not insured), self-rated general health (excellent/very good/good, fair/poor), number of chronic conditions (0, 1, 2, 3, or more), and ever had cancer (yes, no). The number of chronic conditions variable was created by assigning values of 1 to a response of “Yes” and 0 to a response of “No” to questions asking if respondents were ever diagnosed with five conditions including depression, diabetes, heart problems, high blood pressure, or lung conditions; these values were then summed to calculate the chronic conditions variable.17

Statistical Analysis

All analyses apply survey weights published with the HINTS dataset based on population estimates from the American Community Survey to ensure results are generalizable to the U.S. population, as well as with jackknife replicate weights to provide bias-corrected variance estimates.19 All analyses used listwise deletion for incomplete cases.

Descriptive statistics and chi-square tests were used to examine the 2020 sample, limiting the sample to those who provided a “Yes” or “No” response to the question “Have you ever been invited to participate in a clinical trial?” Frequencies were used to describe the first source and most trusted source of information regarding clinical trials. Separate multivariable logistic regression models were used to examine the association between clinical trial invitation (relative to no invitation), participation (relative to no participation), and knowledge about clinical trials (“Know anything” compared to “Don’t know anything”) and race/ethnicity, controlling for covariates. To assess the clinical trial influence factors, unconstrained partial proportional odds models were estimated to test the association of each variable with race/ethnicity, controlling for covariates. The factors that influence participation were measured on an ordinal scale that violated the parallel lines assumption for some variables, making ordinal regression less appropriate than using the partial proportional odds model for these analyses. Post hoc examination of marginal effects (ME) was performed to provide point estimates comparing the likelihood of each response by race/ethnicity. Throughout the analyses, results were considered significant for p values less than 0.05, and all tests were 2-tailed. Analyses were performed using Stata version 17 (2021, StataCorp LP, College Station, TX).

RESULTS

There were 2888 respondents who provided a yes or no answer to the question “Have you ever been invited to participate in a clinical trial?” and were not missing any demographic data used in the analysis. There were significant differences in who had or had not been invited to clinical trials by race/ethnicity, education, metro status, self-rated general health, number of conditions, and ever having cancer (Table 1). Among the 2888 who answered the clinical trial participation question, 324 respondents were invited to a clinical trial and answered whether or not they participated: no differences were observed in demographic characteristics. A total of 2956 respondents answered whether they knew anything about clinical trials, and there were significant differences in knowing about trials by race/ethnicity, education, income, insurance status, and general health.

Table 1 Analytic Sample Population from HINTS 5 Cycle 4 (2020) Data
Full size table

Clinical Trials Participation History

Of the 2888 respondents with complete data, 331 (11.46%) had been invited to participate in a clinical trial (Table 2). Respondents identifying as non-Hispanic Black (relative to non-Hispanic White) (adjusted odds ratio [AOR]: 1.99; 95% confidence interval (CI): 1.08, 3.69), those with a bachelor’s degree or higher education (relative to those with a high school education or less) (AOR: 3.66; 95% CI: 1.92, 7.00), those in fair or poor health (relative to those in excellent, very good, or good health) (AOR: 1.75; 95% CI: 1.04, 2.93), those with at least one chronic condition (relative to those without chronic conditions) (Table 2), and those who have had cancer (relative to those who have not had cancer) (AOR: 2.08; 95% CI: 1.30, 3.33) had higher odds of being invited to participate in a clinical trial, while respondents living in non-metro areas (relative to those living in metro areas) (AOR: 0.43; 95% CI: 0.19, 0.98) had lower odds of being invited to participate. Of the 331 respondents who had been invited to participate in a clinical trial, 324 answered the question about whether they did participate, with 154 (47.53%) selecting “Yes,” and no significant differences were observed in terms of clinical trial participation by demographic variables.

Table 2 Odds of Clinical Trial Invitation, Participation, and Knowledge by Demographic Characteristics
Full size table

Clinical Trial Knowledge

Of the 2956 respondents with complete data, 1932 (65.35%) responded that they “know anything” about clinical trials (Table 2). Respondents with some college (AOR: 1.75; 95% CI: 1.26, 2.44) or bachelor’s degree or higher education (AOR: 4.08; 95% CI: 2.67, 6.24) compared to high school graduates had higher odds of answering that they know anything about clinical trials, while those who were uninsured (AOR: 0.41; 95% CI: 0.21, 0.81) had lower odds of knowing anything about clinical trials.

The sources of information that respondents most frequently reported going to first for information about clinical trials followed the same pattern within and across races/ethnicities: personal healthcare provider, then internet searches, and then health organizations (Fig. 1). Respondents from all races/ethnicities most frequently reported that personal healthcare providers were the most trusted source of clinical trial information, followed by health organizations and government health agencies (Fig. 2).

Fig. 1
figure 1

Frequency of first source for clinical trial information, total and by race/ethnicity.

Full size image
Fig. 2
figure 2

Frequency of most trusted source for clinical trial information, total and by race/ethnicity.

Full size image

Clinical Trial Influence

Marginal effects for race/ethnicity of partial proportional odds models for each of the eight clinical trial influence factors are presented in Table 3 (see Supplementary Appendix eTables 2–4). Respondents identifying as Hispanic compared to non-Hispanic White had lower odds of saying that helping people would influence their decision “A lot” (ME: − 0.10; 95% CI: − 0.16, − 0.05), but significantly higher odds of saying that it would influence their decision “Some” (ME: 0.11; 95% CI: 0.02, 0.21).

Table 3 Differences in Attitudes About Sources of Influence of Participation in Clinical Trials by Race/Ethnicity
Full size table

Respondents identifying as non-Hispanic Black (ME: − 0.11; 95% CI: − 0.18, − 0.024), Hispanic (ME: − 0.09; 95% CI: − 0.16, − 0.03), and non-Hispanic other (ME: − 0.11; 95% CI: − 0.19, − 0.02) had lower odds than those identifying as non-Hispanic White of saying they would be influenced “A lot” by their doctor encouraging them to participate, and higher odds of reporting that their doctor’s encouragement would influence them “A little” (Table 3) or “Not at all” for both non-Hispanic Black (ME: 0.05; 95% CI: 0.01, 0.10) and Hispanic (ME: 0.04; 95% CI: 0.01, 0.08) respondents.

Respondents identifying as non-Hispanic Black had significantly lower odds of answering that family encouragement would influence their clinical trial participation decision “A lot” (ME: − 0.09; 95% CI: − 0.14, − 0.03) or “Some” (ME: − 0.05; 95% CI: − 0.09, − 0.00) and had higher odds of saying that family encouragement would influence them “A little” (ME: 0.06; 95% CI: 0.02, 0.10) or “Not at all” (ME: 0.07; 95% CI: 0.01, 0.13). Respondents identifying as non-Hispanic other had higher odds of reporting that wanting to get better would influence their participation decision “Some” (ME: 0.07; 95% CI: 0.00, 0.14), and lower odds of responding “A lot” (ME: − 0.09; 95% CI: − 0.17, − 0.01) and higher odds of responding “Some” (ME: 0.01; 95% CI: 0.00, 0.02), “A little” (ME: 0.03; 95% CI: 0.00, 0.05), or “Not at all” (ME: 0.06; 95% CI: − 0.00, 0.11) to being influenced by the standard of care not being covered by insurance. No differences were observed across races/ethnicities for the influence of trying a new kind of care, getting paid, or getting support on clinical trial participation decisions.

DISCUSSION

Our analysis found that while non-Hispanic Black respondents were more likely to have been invited to take part in clinical trials, they were neither more nor less likely to participate than other racial/ethnic groups. This finding may be a consequence of the recent focus on the lack of diversity in clinical trials20 and the resulting efforts from healthcare providers and funding agencies to rectify the disparity.21 However, the lack of a corresponding greater likelihood of trial participation by non-Hispanic Black individuals may point to deeper issues such as medical mistrust,22 perceptions of discrimination,23 and lack of health literacy.9

Our findings align with those of previous studies that have found mistrust in the medical system may play a larger role than trust in their personal providers in influencing racial/ethnic minorities’ non-participation in clinical trials.23,24,25 Specifically, while racial/ethnic minority patients may trust their own medical providers, thereby going to them for information about clinical trials, their distrust in the overall healthcare system, and pharmaceutical companies, may play a larger role in influencing clinical trial non-participation despite encouragement.26,27,28

Practice Implications

Identifying sources of influence in clinical trial participation can help orient efforts to support participation and overcome medical mistrust. For instance, doctor’s encouragement was not found to be positively associated with influencing participation for non-Hispanic Black, Hispanic, or non-Hispanic other respondents. Future efforts to improve clinical trial diversity may need to (1) reach outside of healthcare systems to build trust directly in communities through community engagement29 and (2) leverage collaborative partnerships between academia, community organizations, government, and industry.30 Many community organizations have established trust among racial/ethnic minority populations that can serve as a foundation for academic, government, and industry healthcare organizations to commence the trust-building process.30 These approaches may ultimately lead to community-informed, community-engaged, and community-based participatory research approaches that are shown to increase racial/ethnic diversity.31

Clinical trials related to COVID, for instance, highlight how applying intentional approaches can improve diversity.32,33,34 Intentional outreach efforts have also been used successfully in the Black Impact trial, a 24-week health and wellness study to promote attainment of cardiovascular health in Black men,35 and in FAITH!, a study to improve cardiovascular health in Black church participants.36,37 Moreover, using a wider range of media that are culturally targeted, such as video education interventions, may provide an alternate means of changing attitudes and increasing intention to enroll and actual trial enrollment.38

Limitations

Our study used self-reported data, which are subject to survey respondents’ recall bias from not remembering details about their healthcare or whether they have been invited to participate in a clinical trial. Additionally, our analysis of racial differences in participation has a relatively small sample size (n = 324), limiting our ability to detect differences between groups. Thus, the non-significant point estimates for Black and non-Hispanic other may reach statistical significance with a larger sample size. This same limitation due to sample size exists in exploration of important questions related to the interaction of knowledge with race/ethnicity and the decision to participate for those individuals that were invited to participate in a clinical trial.

CONCLUSIONS

The gap in clinical trial participation persists despite a growing recognition of its presence and initiatives by funders to address this disparity. Increasing diversity in clinical trial participation will require deliberate efforts to earn the trust of non-White populations and overcome other social barriers that contribute to underrepresentation in medical research by non-White populations.