Removing uncertainty in real-world evidence and under-represented populations
Under-represented populations introduce uncertainty in real-world evidence. Population coverage determines whether results apply across regions;
How real-world data helps reduce clinical trial risk
When clinical trials start to drift
Development teams across clinical, regulatory, and evidence functions do not need to wait for a postmortem to recognize when a study is in trouble. The signs appear earlier — in recruitment that never accelerates, repeated protocol clarifications, endpoint data that look thinner than planned, and submission dates that keep slipping.
The stakes are high. Delays can push out readouts, extend site costs, increase CRO spend, and force difficult portfolio tradeoffs. Worse, an under-enrolled or operationally unstable trial can leave sponsors with evidence that is hard to interpret or hard to defend with regulators.
That risk is not theoretical. A study in JAMA found that randomized clinical trial discontinuation is common, with poor recruitment the most frequently reported reason. Discontinued trials in that analysis reached a median of only 40.9% of their target sample size.¹ A nationwide analysis of 574 clinical drug trials in the Netherlands reported a 17.8% discontinuation rate, with recruitment failure and efficacy-related futility among the leading causes.²
A 2024 analysis from the Tufts Center for the Study of Drug Development estimated that the direct daily cost of running clinical trials can reach roughly $40,000, even before accounting for the broader value lost when timelines slip.³
Against that backdrop, real-world data can be a practical tool for supporting development programs under pressure. Not every at-risk study can be saved, but in the right context, it can help refine feasibility assumptions, identify missed patient populations, support external comparators, strengthen endpoint context, and keep regulatory packages defensible under review.
Why at-risk clinical trials happen
Most studies do not fail all at once. They drift into risk through a series of issues that gradually stop being manageable.
Recruitment is often the first signal. Eligibility criteria may appear reasonable on paper but prove too restrictive in practice. Protocols may require tests or visit schedules that patients will not accept, or sites may lack access to sufficient eligible patients. In oncology, rare disease, immunology, and advanced therapy programs, even small mismatches between protocol assumptions and real-world care pathways can derail enrollment projections. The literature is consistent: recruitment failure remains a leading cause of trial discontinuation.¹ ²
The second layer of risk is scientific and operational. Endpoint timing may be longer than expected, event rates may be lower, and standards of care can shift mid-study, making assumptions outdated. Protocol amendments often follow, which can strengthen the science but add startup work, retraining, data management burden, and new questions about consistency across enrolled cohorts. Regulatory expectations also continue to evolve, placing greater emphasis on clear justification of trial design, data provenance, and analytical rigor when nontraditional evidence is used.⁴ ⁵
When those pressures combine, the trial may still continue, but it becomes increasingly fragile.
The warning signs of an at-risk trial are usually familiar and align with the operational metrics and risk indicators used in risk-based trial monitoring:⁶ ⁷
- Slow site activation
- Screen failure rates above plan
- Lower-than-expected event accrual
- High protocol deviation volume
- Repeated protocol amendments
- Comparator arm mismatch with current practice
- Submission timelines tied to uncertain data maturity
How real-world data changes clinical trial risks
Real-world data does not fix a struggling trial, but it helps teams understand where assumptions no longer hold and adjust their strategy accordingly.
The FDA has made clear that real-world evidence can support regulatory decision-making for drugs and biologics when the underlying real-world data is fit for use and the study design addresses bias and confounding.⁵ FDA guidance on externally controlled trials also highlights patient-level data from other clinical trials and real-world sources, including registries, electronic health records, and medical claims.⁴ This matters because many at-risk trials face a common problem: the randomized evidence package alone may no longer answer the regulatory or clinical question clearly.
In practice, real-world data is most useful when it addresses a specific evidence gap. It can show where patients are actually treated, which sites see the right phenotype, how the standard of care is evolving, whether an endpoint is realistically observable in routine practice, and whether an external comparator is defensible. It can also help medical affairs and regulatory teams build a more coherent evidence narrative when trial conduct has been uneven.
Real-world data is most valuable when it addresses specific pressure points in at-risk trials:
- Slow recruitment: Refine site selection, identify treating centers, and quantify eligible patient pools from EHR or registry data to improve feasibility assumptions and recruitment recovery plans
- Underpowered single-arm studies: Build an external control arm from patient-level registry, EHR, claims, or prior trial data, aligned with FDA guidance on externally controlled trials⁴
- Protocol amendment pressure: Test inclusion criteria, visit schedules, and endpoint capture against routine care patterns to reduce avoidable amendments and strengthen protocol justification
- Endpoint uncertainty: Evaluate baseline event rates, time to outcome, treatment switching, and follow-up completeness in routine practice to reassess powering and interpretability
- Standard of care variations: Quantify current treatment patterns and outcomes in contemporaneous populations to support comparator relevance and labeling discussions
- Submission risk late in development: Harmonize patient-level data across studies and external sources to close evidence gaps and support submission readiness when timelines compress
Summary: Where real-world data supports at-risk trials
| At-risk trial issue | How RWD can help | Regulatory / development relevance |
|---|---|---|
| Slow recruitment | Identifies treating centers, refine site selection, quantify eligible patient pools using EHR or registry data | Improves feasibility assumptions and recruitment recovery plans |
| Underpowered single-arm study | Builds external control arm using patient-level data from registries, EHRs, claims, or prior trials | Aligned with FDA externally-controlled trial guidance |
| Protocol amendment pressure | Tests inclusion criteria, visit schedules, and endpoint capture against real-world care patterns | Reduces avoidable amendments and strengthens protocol justification |
| Endpoint uncertainty | Evaluates event rates, time to outcome, treatment switching, and follow-up completeness in practice | Supports reassessment of powering and interpretability |
| Standard of care variations | Quantifies current treatment patterns and outcomes in real-world populations | Supports comparator relevance and labeling discussions |
| Submission risk (late stage) | Harmonizes patient-level data across trials and external sources | Supports submission readiness under compressed timelines |
Practical real-world data use cases in trial rescue
The most useful applications of real-world data are tied to specific pressure points in development. A common example is recruitment rescue. When enrollment lags, real-world data can map where the target population is actually diagnosed and treated, how often key inclusion criteria are met in practice, and which referral pathways lead to treatment initiation. This can reshape country prioritization, site selection, and even protocol criteria. In some programs, the issue is not site count but site mix.
Another use case is the external comparator. This is most relevant in rare disease and high unmet need settings, and in single-arm studies where randomization is difficult or no longer feasible. FDA guidance is clear that external control arms can be built from patient-level data, including real-world data, if the data are reliable and the design can distinguish treatment effect from other influences.⁴ That sets a high bar, but is a usable path when internal controls are compromised or recruitment cannot support the original plan.
Real-world data can also help teams determine whether a study is drifting toward futility or simply reflecting different care patterns than expected. If background therapy, treatment switching, or follow-up schedules in clinical practice differ meaningfully from protocol assumptions, the trial may require a revised interpretation strategy rather than a rushed amendment. In practice, these applications tend to fall into a small number of recurring categories:
- Feasibility reset: Re-estimate the eligible population, treatment flow, and event rates using recent patient-level data
- Recruitment recovery: Identify high-yield geographies, referral networks, and treating sites
- External comparator support: Build a contemporaneous control cohort with transparent inclusion rules and confounding adjustment
- Endpoint context: Compare protocol endpoints with real-world clinical milestones and data capture patterns
- Submission stabilization: Close evidence gaps across trials, registries, and observational sources with documented provenance
How to operationalize real-world data in trial rescue scenarios
This is where many programs lose time. Teams know real-world data may help, but they start with the data source instead of the decision they need to make. A better starting point is the clinical and regulatory question: Is the main problem enrollment? Is it a missing comparator? Is it concern about event rates, baseline risk, or endpoint maturity?
The answer should define the data strategy. Fit-for-purpose remains the core principle in FDA’s real-world data guidance.⁵ If the source does not capture the relevant variables with enough completeness and traceability, it will not rescue the study.
The next step is analytic discipline. Rescue work is often urgent, but urgency does not excuse weak design. Eligibility definitions need to be reproducible, and index dates must be explicit. Missing data, treatment switching, immortal time bias, and confounding should be addressed prospectively in the statistical analysis plan. Clinical operations, biostatistics, regulatory, medical affairs, and data management need one shared view of what the real-world data analysis is expected to answer.
Operationally, this is where secure federated access models and trusted research environments become critical. They enable analysis across distributed patient-level datasets while preserving governance and local control, which is essential when time is short and cross-border data movement is constrained.
This is where BC Platforms can help. When trials fall behind, teams often need to quickly reassess whether the target population exists at the expected scale and where those patients are treated. Working with data sourced through our global data partner network, teams can access diverse, multi-modal real-world data across geographies and care settings to support that analysis.
Within a secure, compliant research environment such as BC Mosaic, this data can be analyzed across sites and countries without compromising governance or requiring data movement. Combined with BC Catalyst, teams can then quantify how many patients meet specific inclusion and exclusion criteria, understand where those patients are diagnosed and treated, and iteratively refine cohort definitions based on how those criteria perform in practice.
We can help teams to move from uncertainty to evidence-backed decisions — adjusting feasibility assumptions, refining site strategy, or supporting alternative evidence approaches— without building new data infrastructure from scratch.
When done well, the sequence typically looks like this:
- Define the rescue question and decision point
- Assess whether available real-world data is fit for purpose
- Lock the cohort logic, endpoints, and bias-control plan
- Produce regulator-ready outputs with full traceability
Evidence from practice: recovering a submission under pressure
A real-world example comes from a kidney transplant therapy submission we supported where risk emerged late in development. A biopharma company preparing a Biologics License Application (BLA) needed to integrate data across 21 clinical trials. However, delivery issues in the pivotal Phase III study put the submission timeline at risk only three months before database lock, creating uncertainty around whether the evidence package could be completed in time.⁸
The challenge was not collecting additional data, but aligning and interpreting heterogeneous datasets across studies, vendors, and specialized transplant variables, and turning them into a coherent, regulator-ready submission under compressed timelines. By standardizing and integrating patient-level data across all studies, and delivering the required analyses and documentation, our client was able to stabilize the program and support submission. The BLA was completed four months after database lock and accepted for FDA review.
This reflects a common pattern in at-risk programs: recovery often depends less on generating new data and more on the ability to structure and validate existing evidence so it can support clear clinical and regulatory decisions under pressure.
How to apply real-world data across the trial portfolio
For clinical development leaders, the practical question is not whether real-world data is interesting, but where it can reduce avoidable trial risk.
Start by scanning the portfolio for studies under clear pressure: persistent recruitment underperformance, single-arm designs with limited internal context, endpoints exposed to shifting care patterns, or programs likely to face compressed submission timelines. Then identify which of those issues could be informed by existing registries, EHR data, claims, imaging, or prior trial datasets.
A simple operating model helps focus efforts:
- Protocol teams: Validate feasibility assumptions against real treatment patterns before the next amendment
- Clinical operations: Use real-world data to reset site strategy based on where eligible patients are actually treated
- Biostatistics and RWE leads: Pre-specify external comparator methods before urgency turns into improvisation
- Regulatory and medical affairs: Build the evidence narrative early, with provenance and fit-for-purpose standards documented from the start
The strongest programs treat real-world data as part of development planning, not just as a recovery tool. Its value is most visible when a trial comes under pressure—but its impact is greatest when it is used early to prevent avoidable risk.
Reduce avoidable risk in your clinical trials
When a trial is under pressure, you need fast answers: where eligible patients are treated, whether your assumptions still hold, and which evidence gaps could affect recruitment, interpretation, or submission readiness. BC Platforms helps you find those answers with secure, fit-for-purpose real-world data and evidence support.
Request a consultation to see where real-world data can reduce risk in your trial portfolio.
Frequently asked questions
How can real-world data reduce clinical trial risk?
Real-world data can reduce clinical trial risk by helping teams validate feasibility assumptions, identify eligible patient populations, assess endpoint maturity, understand current standard of care, and support evidence strategies such as external comparators when trial conduct becomes challenging.
When should sponsors use real-world data in at-risk clinical trials?
Sponsors should consider real-world data when a trial faces persistent recruitment delays, high screen failure rates, endpoint uncertainty, comparator mismatch, protocol amendment pressure, or submission timelines that depend on evidence gaps being resolved quickly and credibly.
Can real-world data support an external control arm?
Real-world data can support an external control arm when patient-level data are reliable, cohort definitions are transparent, and the study design can address bias, confounding, and differences between treated and comparator populations.
How does real-world data improve clinical trial recruitment?
Real-world data can improve clinical trial recruitment by showing where eligible patients are diagnosed and treated, which sites see the right patient populations, how inclusion and exclusion criteria perform in practice, and which geographies or referral pathways may improve enrollment.
References
- Kasenda B, von Elm E, You J, et al. “Prevalence, Characteristics, and Publication of Discontinued Randomized Trials.” JAMA. 2014;311(10):1045–1052. Source: JAMA Network.
- Van den Bogert CA, Souverein PC, Brekelmans CTM, et al. “Recruitment failure and futility were the most common reasons for discontinuation of clinical drug trials. Results of a nationwide inception cohort study in the Netherlands.” Journal of Clinical Epidemiology. 2017;88:140–147. Source: Journal of Clinical Epidemiology.
- Smith Z, DiMasi J, Getz K. “Quantifying the Value of a Day of Delay in Drug Development.” Tufts Center for the Study of Drug Development. 2024. Source: Tufts CSDD white paper.
- U.S. Food and Drug Administration. “Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products.” Draft guidance for industry. February 2023. Source: FDA guidance document.
- U.S. Food and Drug Administration. “Considerations for the Use of Real-World Data and Real-World Evidence To Support Regulatory Decision-Making for Drug and Biological Products.” Guidance for industry. August 2023. Source: FDA guidance document.
- International Council for Harmonisation. “ICH E6(R3) Guideline for Good Clinical Practice.” Final guideline. 2025. Source: ICH guideline.
- TransCelerate BioPharma. “Risk Based Monitoring.” Source: TransCelerate BioPharma resources.
- BC Platforms. Accelerating FDA submission for kidney transplant therapy. Case study. 2026. Source: BC Platforms.
Last updated