The Regulatory Backdrop
FDA's April 2022 draft guidance on diversity action plans for clinical studies — followed by finalized guidance and the requirements introduced under the FDORA (Food and Drug Omnibus Reform Act) in December 2022 — moved diversity in clinical enrollment from a voluntary aspiration to an explicit expectation for sponsors of certain new drug applications and biologic license applications. Sponsors of Phase III programs in many therapeutic areas are now expected to develop and submit a Race and Ethnicity Diversity Action Plan that outlines how they intend to enroll diverse populations and what protocol and operational barriers they have assessed.
The guidance is clear that FDA is not setting mandatory quotas — it is requiring sponsors to articulate a thoughtful approach and demonstrate that they have addressed known structural barriers to participation. But the regulatory momentum is real, and sponsors who treat diversity planning as a checkbox exercise rather than an operational design problem are likely to find their action plans subject to scrutiny during review.
More importantly for clinical operations: meeting diversity targets is not primarily a regulatory problem. It is a site selection and patient identification problem. The upstream choices — where sites are located, which patient populations they serve, how patients are identified and referred — determine the demographic composition of the trial population before a single patient is consented. No recruitment advertising campaign can fully compensate for a site network that structurally excludes diverse communities.
How Site Selection Creates Demographic Bias Before Enrollment Begins
The typical Phase III site selection process, focused on investigator experience and patient volume at academic medical centers and large specialty practices, introduces demographic bias through geographic concentration. Academic medical centers are disproportionately located in urban cores and suburban research corridors. The patient populations they serve, while often diverse in aggregate, skew toward patients who have the resources, insurance coverage, and proximity to navigate a research-oriented healthcare system.
Community health centers, federally qualified health centers, Historically Black Colleges and University-affiliated medical programs, and regional specialty practices in underserved areas serve patient populations with high representation of the demographic groups that are chronically underrepresented in clinical trials — but they are frequently overlooked in site selection processes that prioritize prior trial experience and academic affiliation. Eligibility for trial participation and proximity to a research site are not equivalent criteria, but they are often treated as such in practice.
EHR phenotype analysis applied to a geographically diverse site candidate list can help address this by identifying which sites in underserved regions have patient populations matching the trial phenotype — without requiring the site to have prior Phase III trial experience. The patient population data is the selection criterion; the research track record is a secondary qualification that can be developed with appropriate sponsor support and CRO co-monitoring.
Referral Design and Its Effect on Who Gets Referred
Even within a diverse site network, referral design choices affect which patients are actually surfaced as trial candidates. Passive identification approaches — coordinators flagging patients as they come through clinic — are more likely to identify patients with higher engagement rates, more frequent care-seeking behavior, and stronger relationships with healthcare providers. These characteristics correlate with socioeconomic status, insurance type, and health literacy in ways that are not neutral with respect to demographic representation.
Proactive EHR-based identification, by contrast, operates on the full patient panel of the site — not just the patients who are actively in the care cycle at the moment a coordinator is looking. It surfaces patients regardless of how frequently they seek care, regardless of whether they've been recently reminded about their condition, and regardless of whether the provider has them top of mind. This makes the identification process structurally more equitable, because the mechanism of surfacing candidates is not filtered through the recency and availability biases that affect passive referral.
The operational implication is that phenotype-based identification at sites serving diverse populations is not just an efficiency improvement — it has a diversity enrollment benefit that passive referral at the same site cannot replicate. Sponsors who use proactive identification exclusively at their high-volume academic sites while relying on passive referral at community health center sites will underutilize the diversity potential of their community site network.
Protocol Design Factors That Create Enrollment Barriers
Site selection and referral design are the most tractable levers on diversity, but protocol design creates barriers that operate independently. Several common protocol features disproportionately limit participation from populations that are underrepresented in trials:
- Frequent in-person visit requirements: Protocols requiring 8-12 in-person clinic visits over a 12-month period create participation barriers for patients who work hourly, have caretaking responsibilities, or lack reliable transportation. These barriers are not demographically neutral.
- Exclusion criteria for comorbidities disproportionately prevalent in specific populations: Hypertension, type 2 diabetes, and chronic kidney disease exclusions, when applied in cardiovascular or metabolic disease trials, can structurally exclude patient populations with higher rates of these comorbidities if those criteria are not medically required for safety. Protocol Medical teams should evaluate whether comorbidity exclusions are safety-driven or convenience-driven, and remove the latter.
- English-language consent and documentation requirements where not legally mandated: Providing informed consent only in English is not required by regulation in most circumstances and creates a documented barrier to participation in multilingual communities. FDA guidance on informed consent under 21 CFR Part 50 contemplates consent in languages understood by participants.
We're not saying every protocol can or should eliminate all visit burden or all comorbidity exclusions — the clinical and safety justification for specific criteria is real and important. We're saying that a systematic review of eligibility criteria and visit schedules through a participation-barrier lens, early in protocol development, can identify restrictions that are not scientifically necessary and that create disproportionate access barriers.
Measuring Diversity Against Meaningful Benchmarks
A persistent problem in diversity reporting for clinical trials is that demographic representation is measured against overall US population demographics — a benchmark that may not reflect the demographics of the disease-affected population. A trial in a therapeutic area where the affected population is disproportionately older, or disproportionately male, or has different racial and ethnic distribution than the general population, should measure diversity against the disease epidemiology, not against general population figures.
FDA's 2022 guidance acknowledges this: diversity goals should be based on available epidemiological data for the relevant condition. This is a meaningful standard shift — it asks sponsors to characterize the disease burden by demographic group and design enrollment to be proportional to that burden. For sponsors, this means disease epidemiology research is a required input to the Diversity Action Plan, not just a nice-to-have context section.
Operational Accountability: Who Owns the Diversity Target?
One organizational challenge that clinical teams frequently encounter is that diversity targets are set by regulatory affairs, enrollment design is driven by clinical operations, and site selection is managed by a CRO — three separate organizational functions with different incentive structures and different timelines. The diversity target doesn't translate into enrollment outcome unless the site selection process and the patient identification strategy are explicitly designed to deliver it, which requires coordination across those functions.
The most effective approach assigns explicit accountability: a named clinical operations lead owns the diversity enrollment metric, with authority to make site selection and referral design recommendations based on population diversity analysis. Without that accountability, diversity goals remain aspirational documents and enrollment demographics reflect the structural defaults of the site network rather than any intentional design choice.
