Physician Bias and Clinical Trial Participation in Underrepresented Populations

From design to dissemination, including diverse populations in clinical research studies enhances the generalizability of scientific outcomes and enhances the capacity of research to affect policy and practice. But diverse inclusion does more than improve the quality of science; it also is a critical matter of health equity. Profound past ethical failures in medical research, such as Tuskegee and Willowbrook,¹  have led to a focus on the principle of informed consent and the protection of marginalized and vulnerable populations from the potential harms of research. However, the ethical principle of beneficence also requires investigators to ensure that the benefits of research, in terms of both participation and outcomes, are equitably distributed. Purposeful, diverse recruitment balances these principles and ensures that populations are neither disproportionately targeted nor excluded from clinical research.^2,3 

Implicit bias among clinicians and scientists is an important but understudied factor that shapes inclusion in research. Several studies and systematic reviews have shown that unconscious biases among health care providers can lead to differential diagnoses, treatments, and/or referrals of patients by race and ethnicity, sex, age, language, and weight.^4–6  However, less attention has been paid to how unexamined biases among scientists and providers may affect inclusion in research. In this issue of Pediatrics, Coon et al’s secondary analysis of the Bronchiolitis Follow-Up Intervention (BeneFIT) trial⁷  takes an important step in that direction.

BeneFIT was a multicenter, unblinded, noninferiority randomized trial comparing scheduled follow-up for children hospitalized with bronchiolitis (standard of care) to an as-needed (PRN) follow-up with primary care providers.⁸  The primary outcome measure was a decrease in parental anxiety. Initially, 548 patients were determined eligible for recruitment to BeneFIT, but physicians determined that 42 of them were “not appropriate” for participation. The secondary analysis by Coon et al compared the demographic and clinical characteristics of the 42 children deemed ineligible to participate in the BeneFIT trial to the 506 deemed eligible.⁷  They found that only 29% of the children whom physicians determined were “appropriate” to participate were Hispanic, compared to 52% among those considered not appropriate.

Clinicians’ justifications for patient eligibility determinations were often vague or unclear; Coon and colleagues hypothesize that providers’ assumptions and biases about Hispanic families’ health literacy and risks may have shaped the significant differences in ethnicity between children deemed eligible or ineligible. Thus, this secondary analysis emphasizes the importance of clear and specific clinical inclusion and exclusion criteria in medical research, not only for the purposes of rigor and reproducibility, but also because they reduce opportunities for investigators’ implicit biases to shape outcomes.

Importantly, Coon et al also found that there were no racial or ethnic differences in either parental acceptance of the intervention or adoption of PRN follow-up among the 138 families randomized to the intervention arm. Hence, their analysis of the BeneFIT trial illustrates how implicit biases that emerge from structural racism may limit diverse inclusion, which then further contributes to structural inequities by systematically excluding marginalized patients from, and denying them the benefits of, research studies and interventions. These findings underscore how attention to inclusion has importance beyond the generalizability or impact of the scientific enterprise; rather, inclusion is a matter of justice.

Coon et al’s analysis highlights potential entry points for bias, but these can also be viewed as opportunities for enhancing inclusion, equity, and justice. In recent years, the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) and NIH have made efforts to increase inclusion of pregnant and lactating people, children, and people with intellectual disabilities in our vast portfolio of social, behavioral, and clinical research.⁹  The disproportionate toll of COVID-19 infection on African Americans has also highlighted the crucial importance of including people of color in clinical trials of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccines.¹⁰ 

NICHD’s Pediatric Trials Network (PTN)¹¹  recently conducted a global review of the demographic composition of all of its previous and current studies. The analysis showed that the network had no significant exclusions by sex or race and ethnicity. Compared with national averages, however, inclusion of Asian populations was lower than expected. Language barriers to recruitment and consent processes were identified as potential contributing factors. The PTN is developing metrics across all studies to purposefully broaden inclusion.

In NICHD’s 2020 Strategic Plan, 5 scientific research themes and 5 cross-cutting objectives were articulated, including reducing health disparities among underrepresented populations and promoting an inclusive scientific workforce that fosters research training.¹²  The 2021 STrategies to enRich Inclusion and achieVe Equity (STRIVE) initiative¹³  is aiming to implement these goals.

As the Coon et al analysis and these examples illustrate, inclusive research requires a layered, multilevel approach. We encourage scientists and institutions that initiate, sponsor, and conduct clinical research to consider the following principles:

1. Include evidence-based training on implicit bias and structural racism and sexism in required educational material for research project staff at all levels;

2. Increase diversity among investigators who are designing and implementing clinical studies; and

3. Minimize the impact of clinician and researcher bias through clear and specific inclusion and exclusion criteria developed with community input.

Meaningful inclusion in clinical research must begin with training diverse medical and scientific workforces and enhancing the diversity of research and clinical teams. Social and scientific studies have illustrated that these initial forms of inclusion affect the scientific questions that are subsequently asked, how the questions are approached, and which research ideas are pursued.^14,15  In turn, these social and structural changes will generate more inclusive research designs and implementation practices that avoid enrollment biases like those identified in the Coon et al secondary analysis.