Real-World Evidence in Drug Development: Opportunities and Challenges

Real-world evidence is becoming one of the most useful connective tissues in drug development. It helps teams understand patients before a trial starts, make better design choices during development, and keep learning after approval when a therapy enters routine care.

The reason is simple: randomized controlled trials are essential, but they are not the only place evidence lives. Electronic health records, medical claims, disease and product registries, genomic data, patient-reported outcomes, wearable devices, and post-market safety systems all capture pieces of clinical reality that trials cannot fully reproduce.

The FDA defines real-world data as patient health status or healthcare delivery data routinely collected from sources such as EHRs, claims, registries, and digital health technologies. Real-world evidence is the clinical evidence about a medical product's use, benefits, or risks that comes from analyzing that data. FDA real-world evidence

The opportunity is not to replace trials. It is to use RWE where it makes drug development smarter: earlier decisions, better feasibility planning, stronger external context, faster evidence generation, clearer payer value, and more complete life-cycle management.

RWE is most powerful when it is treated as a development capability, not a post-approval reporting exercise.

What RWE changes in early discovery

In early discovery, the first value of RWE is visibility. Before a team commits to a target, indication, or product profile, real-world data can help answer basic questions that are often hard to see from literature alone: who are the patients, how are they treated today, where does the current care pathway fail, and which subgroups have the highest unmet need?

That matters because a target product profile is not just a scientific document. It is also a bet about where a future therapy can create enough clinical value to justify development. RWE can make that bet less abstract by showing treatment pathways, diagnostic delays, disease burden, comorbidity patterns, healthcare utilization, and gaps in current standard of care.

The opportunity becomes especially important in oncology and rare disease, where development is increasingly biomarker-defined. When genomic profiles can be linked to treatment histories and outcomes, teams can look for subpopulations that respond differently, progress faster, or develop resistance earlier. Those signals can shape biomarker strategy, endpoint selection, patient segmentation, and even the decision to pursue a narrower but more clinically meaningful indication.

Digital health technologies and patient-generated data add another layer. Wearables and remote monitoring can capture symptoms, function, mobility, sleep, activity, and physiologic patterns closer to daily life. Those data streams are not automatically regulatory-grade, but they can help teams discover which patient-centered outcomes may deserve formal validation.

Better trial design and feasibility

RWE has long been used to estimate event rates and sample sizes. Its larger value is now in trial feasibility and design. A team can use real-world data to test how inclusion and exclusion criteria affect the eligible population, identify countries or sites with sufficient patient volume, estimate recruitment difficulty, and understand whether a proposed endpoint is observable in routine care.

This changes the quality of early design discussions. Instead of debating eligibility criteria in the abstract, teams can model how each criterion changes the reachable population. A single lab cutoff, prior therapy requirement, washout period, or comorbidity exclusion can quietly remove a large share of otherwise relevant patients. RWE makes those tradeoffs visible before the protocol is locked.

It can also support patient stratification. Baseline characteristics, disease severity markers, prior treatment patterns, and prognostic variables observed in real-world cohorts can help teams define risk groups or enrichment strategies. In some settings, real-world data can inform Bayesian priors or external context for control outcomes, provided the data source is fit for purpose and the assumptions are explicit.

Externally controlled trials are one of the clearest examples. FDA's draft guidance describes externally controlled trials as studies where outcomes in treated participants are compared with outcomes in patients outside the current trial who did not receive the test treatment. FDA external control guidance This design can matter in rare disease, oncology subtypes, or serious conditions where a conventional randomized control arm may be impractical or ethically difficult.

The bar is high. The external control has to be clinically credible, comparable, and pre-specified. But when the disease context justifies it and the data source is strong enough, RWE can help answer a counterfactual question that a traditional trial may not be able to reach.

More efficient study execution

RWE can also improve the operational side of clinical development. Site selection, recruitment planning, risk-based monitoring, protocol amendment prevention, and drug supply forecasting all benefit from better visibility into real patient populations and care patterns.

For recruitment, real-world data can identify where eligible patients are likely to be found and which sites have experience with similar populations. That is especially useful when the population is small, underdiagnosed, or fragmented across specialists. The same approach can reduce wasted startup time by avoiding sites that appear attractive on paper but have limited reachable patients after protocol criteria are applied.

For monitoring, RWE can help teams anticipate operational risks. Historical trial performance, real-world clinical workflows, missingness patterns, and prior protocol deviations can inform risk-based monitoring plans. The goal is not more surveillance for its own sake. It is earlier detection of data quality issues that could threaten interpretability.

Even drug supply planning can benefit. Manufacturing timelines are long, shelf life is finite, and overproduction is expensive. If a sponsor can combine historical enrollment and retention data from similar studies with current trial performance, it can forecast supply needs more accurately and reduce both shortage risk and waste.

Regulatory and marketing applications

Historically, regulators used real-world data most heavily for pharmacovigilance: monitoring how products perform after approval, detecting safety signals, and understanding adverse events in broader populations. That remains central. FDA's Sentinel Initiative, launched in 2008, is a national electronic system for monitoring the safety of FDA-regulated medical products and has evolved into a large distributed safety database. FDA Sentinel Initiative

What has changed is the broader regulatory role of RWE. FDA's RWE program now includes guidance on EHR and claims data, registries, data standards, non-interventional studies, external controls, and use of RWD/RWE to support regulatory decisions. FDA real-world evidence In Europe, EMA's DARWIN EU network was established to deliver real-world evidence from healthcare databases for questions about medicine use, safety, and effectiveness across the product life cycle. EMA DARWIN EU

The practical implication for sponsors is that RWE strategy needs to start earlier. If a program may need natural history evidence, an external comparator, broader safety context, or post-approval evidence generation, those data assets cannot be treated as last-minute add-ons.

There are already public examples of RWD supporting regulatory decisions. Pfizer announced in 2019 that the FDA expanded Ibrance's indication to include men with HR-positive, HER2-negative advanced or metastatic breast cancer, with the approval based predominantly on EHR and post-marketing data from multiple sources. Pfizer Ibrance RWD approval

Regulatory-grade RWE is not a pile of real-world data. It is a pre-specified, auditable answer to a decision question.

Product launch and market access

Approval is not the end of the evidence problem. Once a product launches, physicians, payers, HTA bodies, and patients ask questions the pivotal trials may not fully answer: how does the drug perform in older patients, people with comorbidities, underserved groups, different care settings, longer follow-up, or against real-world comparators?

This is where RWE becomes commercially and clinically important. It can support comparative effectiveness, adherence and persistence analysis, treatment sequencing, burden-of-disease studies, patient-reported outcomes, quality-of-life evidence, and healthcare resource utilization. Those outputs are useful for medical affairs, HEOR, payer engagement, field strategy, and clinical decision support.

RWE can also expose underserved patient groups. A therapy may be approved for a broad label but reach patients unevenly because of diagnostic friction, access barriers, physician awareness, formulary restrictions, geography, or care pathway fragmentation. Real-world data can show where eligible patients are not being treated and why the gap exists.

For payers, the issue is value under routine conditions. A product can be effective in a trial and still create uncertainty around budget impact, durability, safety, adherence, and real-world resource use. RWE helps reduce that uncertainty by showing how the therapy behaves after it leaves the controlled environment of the trial.

Life-cycle management

After launch, RWE becomes part of product life-cycle management. It can support label expansion, post-market commitments, comparative safety monitoring, value assessment, treatment pathway optimization, and supply chain planning.

The label expansion use case is especially important when direct trial evidence is limited for a small subgroup. RWE can help characterize the subgroup, estimate outcomes, confirm safety, and support the clinical rationale for broader use. It can also help identify new development hypotheses by showing where outcomes remain poor despite available therapy.

Pharmacovigilance remains one of the strongest life-cycle applications. Rare adverse events may not appear in pre-approval trials because the treated population is too small or follow-up is too short. Claims, EHRs, registries, safety databases, and active surveillance systems can help detect signals earlier and evaluate whether they persist after adjustment for patient risk factors and treatment context.

RWE also has operational value. Demand forecasting, inventory planning, delayed shipment analysis, site-of-care trends, adherence patterns, and competitor uptake can all inform better commercial and supply decisions. This is less visible than a regulatory submission, but it matters because evidence affects how a product is made, distributed, reimbursed, and used.

The challenges are real

The enthusiasm around RWE is justified, but it can become dangerous if teams ignore the limits. Real-world data is not collected for the convenience of researchers. It is collected during care delivery, billing, registry participation, device use, or safety reporting. That means the data can be incomplete, inconsistent, delayed, biased, miscoded, or missing the variable that matters most.

Data access is the first barrier. Privacy rules, contracts, institutional governance, country-specific legal requirements, and reluctance to share data can make multinational evidence generation slow. Even when data exists, it may not be linkable across settings, current enough for the decision, or available with the granularity needed for a specific endpoint.

Data quality is the second barrier. FDA's final guidance on EHR and claims data asks sponsors to assess whether the data are relevant and reliable for the regulatory question, including provenance, accrual, missingness, linkage, and validation. FDA EHR and claims RWD guidance A claims code may be excellent for reimbursement and weak for clinical severity. An EHR field may be clinically rich but inconsistently populated. A registry may be curated but narrow.

Technology is another barrier. RWE often requires integration across structured claims, semi-structured EHR fields, unstructured notes, imaging, lab data, genomics, registries, and device streams. Common data models and AI can help, but they do not remove the need for validation. A transformed variable still has to preserve clinical meaning.

The hardest challenge is methodological

In randomized trials, randomization protects the comparison by balancing measured and unmeasured factors across groups. In observational data, treatment choice is not random. Patients receive therapies because of disease severity, physician preference, access, prior response, comorbidities, contraindications, and other factors that also affect outcomes.

That creates confounding. Statistical adjustment, matching, weighting, propensity scores, instrumental variables, negative controls, and sensitivity analyses can all help, but none of them make observational data magically equivalent to randomization. They work only when the required assumptions are plausible and the important variables are measured with enough accuracy.

This is why target trial emulation has become so important in RWE. The team should define the trial it wishes it could run: eligibility, treatment strategies, time zero, assignment, follow-up, outcome, causal contrast, and analysis plan. Then it should use real-world data to emulate that design as closely as possible. The discipline is not cosmetic. It prevents teams from turning a database query into a causal claim it cannot support.

Talent is the final constraint. Good RWE requires domain experts, epidemiologists, biostatisticians, data engineers, clinicians, regulatory strategists, and product teams who can work from the same evidence question. Data science alone is not enough. Clinical intuition alone is not enough. The output has to be scientifically defensible, technically reproducible, and relevant to the decision-maker.

The practical takeaway

RWE creates opportunities across the drug development life cycle: discovery, feasibility, trial design, study execution, external controls, regulatory strategy, launch, payer value, safety, label expansion, and life-cycle management. But every opportunity comes with the same condition: the evidence has to be fit for purpose.

That means starting with the decision, not the dataset. What question needs to be answered? Who will use the answer? What action depends on it? Which data source can observe the right population, exposure, comparator, outcome, timing, and covariates? What assumptions are required? How will bias be assessed? How will the analysis be reproduced and inspected?

The organizations that answer those questions well will not use RWE as a loose supplement to traditional development. They will use it as an operating layer for evidence generation - one that makes drug development more patient-centered, more adaptive, and more connected to what actually happens in care.

That is the real opportunity. Not more data. Better decisions.