If an organization only knows who enrolled, it is starting its equity analysis too late.

Most research organizations are comfortable stating that they value equity. The language is now familiar across the field, appearing in strategic plans, funding applications, recruitment materials, and institutional mission statements. These commitments are important, but they also raise a more difficult question: whether organizations can actually see where inequity is occurring within their own research systems.

Answering that question requires more than reporting who ultimately enrolls in a study. By the time a participant appears in an enrollment table, they have already moved through a series of decisions, eligibility determinations, logistical constraints, and institutional practices that shape whether participation was ever realistically available to them. In that sense, enrollment is not the beginning of the equity story; it is the end of a long and often invisible pathway.

Who actually had the opportunity to participate?

Consider a study that enrolls 50 participants. On its surface, the demographic distribution of those 50 individuals may appear reasonably representative. That information is useful, but it is also incomplete. It tells us very little about how many people were ever in a position to reach that final stage.

Imagine, for example, that 1,000 patients in a health system had the condition under study. Of those, 400 were identified for screening, 150 were deemed potentially eligible, 90 were approached, 60 consented, and 50 ultimately enrolled. These numbers are hypothetical, but the structure they represent is not.

If we only examine the final 50 participants, we lose visibility into where differences may have emerged along the way. We cannot see whether certain groups were less likely to be identified, screened, deemed eligible, or approached. We also cannot distinguish between true differences in willingness to participate and differences in who was ever given a meaningful opportunity to consider participation in the first place.

That distinction is not merely technical; it is foundational to how we interpret equity in research participation.

A 2024 mixed-methods study of older adults with cancer illustrates this point clearly. By examining the pathway from trial availability through eligibility, invitation, and participation, the researchers found that non-invitation itself contributed to underrepresentation. Patients' social circumstances and healthcare professionals' perceptions of insufficient informal support or high home-care needs were associated with whether eligible individuals were invited to participate. The authors concluded that improving inclusion requires attention not only to patients' decisions, but also to physician-patient relationships, professional practices, and institutional resources that shape who is invited into research in the first place (Hagège et al., 2024).

This finding underscores a critical methodological issue: if we only measure acceptance or refusal, we risk overlooking inequities that occur long before a decision is made.

For this reason, organizations concerned with research equity need to be able to ask more upstream questions. These include who is considered eligible, who is screened out and why, who is approached, and who is never approached despite appearing eligible. They also include who declines participation and under what circumstances, who withdraws after enrollment, and who is lost to follow-up. Beyond individual decisions, it is equally important to understand who bears the logistical burden of participation, who experiences repeated scheduling or reimbursement barriers, and which sites consistently struggle to enroll participants, and why.

None of these indicators, on their own, establish causation. A screening dashboard is not a causal model. Rather, these data points help identify where closer investigation is warranted.

When enrollment data obscures the underlying mechanism

Underrepresentation is often framed as a matter of participant willingness. When enrollment rates differ across groups, explanations frequently default to assumptions about trust, interest, or engagement. While these factors can be relevant, they are often invoked too quickly and without sufficient attention to structural conditions.

It is important to distinguish between declining participation and never being offered the opportunity to participate. These are fundamentally different outcomes, yet they are frequently conflated when organizations lack data on earlier stages of the research pipeline.

Even the language of "recruitment" can unintentionally narrow the frame. It encourages a focus on persuading individuals to join studies, rather than examining how and why certain populations encounter research opportunities in the first place.

Site selection provides a clear example. In an analysis of 108 Bristol Myers Squibb-sponsored U.S. oncology trials involving 15,763 participants with available race and ethnicity information, trial sites located in counties with larger non-White populations enrolled more non-White participants. The researchers also found that purposeful recruitment efforts in prostate cancer trials were associated with an 11% increase in Black participant enrollment. Importantly, the authors characterized these findings as hypothesis-generating rather than causal evidence (Kuri et al., 2023).

That qualification matters. The study does not provide a universal formula for achieving equitable enrollment. It does demonstrate that organizations can examine operational features of their own research portfolios and test whether factors such as site geography, investigator characteristics, and recruitment strategies are associated with who participates.

These findings point to a broader reality. Inclusion is not only a matter of outreach; it is also a function of infrastructure. Where research is conducted, who has access to those sites, and what support those sites receive all shape who ultimately appears in the data.

A commitment to equity, no matter how clearly stated, cannot compensate for structural decisions that place studies far from the communities they aim to include.

Participation burden as an equity dimension

Equity considerations also extend beyond enrollment. Participation does not end at consent, and the experience of being in a study can vary significantly depending on a participant's resources and circumstances.

A single study visit that appears straightforward in a protocol may involve substantial real-world demands: long travel times, time away from work, childcare or caregiving arrangements, transportation challenges, and out-of-pocket expenses that are later reimbursed, sometimes after significant delay. What is recorded as a brief visit in study documentation may represent an entire day of logistical and financial strain for participants.

These burdens are not evenly distributed. Individuals with flexible employment, reliable transportation, financial stability, and caregiving support experience participation differently than those without these resources, even when enrolled in the same study under identical protocols.

Research on participation burden has demonstrated that these demands can be examined systematically rather than treated simply as unavoidable inconveniences of clinical research. Cameron and colleagues developed the Patient Friction Coefficient as one approach to estimating the burden created by protocol requirements, including travel, time commitments, procedures, and other demands placed on participants (Cameron et al., 2020). The value of the concept is less about adopting one particular metric than recognizing that participation burden can be made visible, assessed, and considered during study design.

Regulatory guidance increasingly reflects the need to consider who can realistically participate in clinical trials. The FDA's December 2025 final guidance on enhancing clinical trial participation recommends approaches to enrolling populations that more closely reflect those likely to use a medical product if approved. The guidance addresses demographic characteristics such as age, race, ethnicity, sex, and residence, as well as non-demographic characteristics including disability, comorbidities, organ dysfunction, extremes of weight, and conditions with low prevalence (U.S. Food and Drug Administration [FDA], 2025).

Taken together, these developments reinforce a central point: representation cannot be reduced to a single demographic table. It must also account for whether study designs create genuine opportunities for participation.

The limits of measurement and the importance of using it well

Increasing the amount of data collected does not automatically improve understanding. Organizations can produce extensive dashboards while still missing the underlying dynamics that shape participation. The challenge is not simply to measure more but to measure more meaningfully.

One key issue is the choice of denominator. Comparing study enrollment to general population demographics may be misleading if the condition under study disproportionately affects certain groups. More appropriate comparisons may include the disease population, the eligible patient pool, the health system population from which participants are recruited, the geographic catchment area of participating sites, or some combination of these.

This is where a superficially impressive diversity percentage can become misleading. If 20% of enrolled participants belong to a particular population, is that good representation? The answer depends on the denominator. If that population represents 8% of people affected by the disease in the relevant catchment area, the interpretation is very different than if it represents 40%.

Equally important is understanding where in the research pathway disparities emerge. A lower enrollment rate in one group may reflect completely different underlying mechanisms depending on whether individuals are less likely to be identified, less likely to be approached, more frequently excluded by eligibility criteria, or more likely to decline after being fully informed.

Those mechanisms require different interventions. A recruitment campaign will do little to correct an upstream screening problem. Community outreach cannot fix unnecessary exclusion criteria. A new consent brochure will not solve a study schedule that requires participants to miss a full day of hourly work every two weeks.

This is precisely why the stage at which inequity appears matters.

Quantitative data also require context. Small numbers can produce unstable estimates, and overly granular reporting may introduce privacy concerns. Moreover, demographic variables such as race, ethnicity, age, disability, geography, language, or insurance status may identify patterns without explaining how those patterns were produced.

For that reason, quantitative analysis is often most useful when paired with qualitative inquiry. Interviews, community-engaged research, participant feedback, observation, and the experiences of site staff can help explain patterns that administrative data alone cannot resolve. Hagège and colleagues' mixed-methods approach provides one example of the value of combining quantitative measures with interviews to understand why eligible patients may never progress to trial participation (Hagège et al., 2024).

A higher withdrawal rate after a particular study visit, for example, may appear as a statistical anomaly until participants describe that visit as requiring prolonged fasting, a 6:30 a.m. arrival, several hours away from work, and weeks of waiting for expense reimbursement.

At that point, the organization has learned something considerably more useful than its retention percentage.

It has identified something it can investigate and potentially redesign.

When data lead to change

The value of equity measurement ultimately depends on whether it informs action. Data collection alone does not improve inclusion if it is not followed by investigation and adjustment.

When certain sites consistently enroll populations that differ substantially from the populations they serve, it is worth examining referral pathways, staffing, infrastructure, site selection, and local engagement.

When potentially eligible participants from one group are consistently less likely to be approached, screening processes and clinical workflows deserve scrutiny.

When participants disproportionately withdraw after particular procedures or visits, study teams should examine whether protocol burden contributes to the pattern.

When reimbursement delays create financial strain, the reimbursement system becomes part of the equity conversation.

And when sites serving historically underrepresented communities are expected to perform additional outreach, relationship-building, translation, transportation coordination, or participant support without corresponding resources, organizations should examine the resource model before concluding that those sites simply "struggle with recruitment."

In other words, equity measurement should eventually lead back to operations.

That is where the work becomes uncomfortable, because the resulting questions are no longer only about participant behavior. They concern protocol design, budgets, contracting, site selection, eligibility criteria, staffing models, reimbursement processes, investigator decision-making, data systems, organizational priorities, and who has authority to change them.

Large-scale reporting systems demonstrate what can be measured at the portfolio level. In fiscal year 2025, NIH-supported clinical research reported enrollment data for 11,048,927 participants worldwide, including demographic data on sex, race, ethnicity, and age. NIH also requires scientifically appropriate inclusion under its policies concerning women, racial and ethnic minority groups, and inclusion across the lifespan (National Institutes of Health [NIH], 2026).

These data are valuable for understanding broad patterns in clinical research participation. But aggregate reporting still cannot tell an individual research organization where people are disappearing from its own participation pathway.

For that, organizations must examine their own systems more closely:

• Who was represented in the final dataset?
• Who could have been?
• Who was identified?
• Who was screened?
• Who was offered participation?
• Who was excluded, and for what reason?
• Who declined?
• Who enrolled but could not remain?
• Who carried the greatest financial, logistical, physical, or time burden?
• Which sites had the resources to make participation possible?
• And which communities never came close enough to the research system to appear in its data at all?

Conclusion

Data alone will not resolve questions of equity in clinical research. Nor will good intentions. What is required is a willingness to examine the full pathway of participation, from the availability of a research opportunity through eligibility, invitation, enrollment, participation, and retention, and to recognize that each stage contains opportunities for both inclusion and exclusion.

Equity, in this sense, is not a static outcome that can be established with a single percentage. It is something organizations have to keep looking for in the systems they design and operate.

Measurement gives us a place to start. It can show us that a pattern exists, identify where in the process it appears, and tell us what deserves closer investigation. Community knowledge, qualitative research, participant experience, and local context help us understand what those patterns mean.

Then comes the part that matters: deciding whether we are willing to change what produced them.

Equity requires more than believing we are being inclusive.

It requires being willing to find out whether we actually are.

References

Cameron, D., Willoughby, C., Messer, D., Lux, M., Aitken, M., & Getz, K. (2020). Assessing participation burden in clinical trials: Introducing the Patient Friction Coefficient. Clinical Therapeutics, 42(8), e150-e159. doi:10.1016/j.clinthera.2020.06.015.

Hagège, M., Bringuier, M., Martinez-Tapia, C., Chouaïd, C., Helissey, C., Brain, E., Rochette Lempdes, G., Dubot, C., Bello-Roufai, D., Geiss, R., Kempf, E., Gourden, A., Elgharbi, H., Garrigou, S., Gregoire, L., Derbez, B., & Canouï-Poitrine, F. (2024). Disentangling the reasons why older adults do not readily participate in cancer trials: A socio-epidemiological mixed methods approach. Age and Ageing, 53(2), afae007. doi:10.1093/ageing/afae007.

Kuri, L., Setru, S., Liu, G., Moniz Reed, D., Weigand, D., Surampudi, A., Berger, S., Paulucci, D., Rai, A., Sethuraman, V., Vito, B., Kellar-Wood, H., & Micsinai Balan, M. (2023). Data-driven strategies for increasing patient diversity in Bristol Myers Squibb-sponsored US oncology clinical trials. Clinical Trials, 20(6), 585-593. doi:10.1177/17407745231180506.

National Institutes of Health. (2026, May 27). FY25 enrollment data from NIH-supported clinical research now available. NIH Extramural Nexus.

U.S. Food and Drug Administration. (2025, December). Enhancing participation in clinical trials: Eligibility criteria, enrollment practices, and trial designs: Guidance for industry. Center for Drug Evaluation and Research & Center for Biologics Evaluation and Research.