FDA's revised effectiveness guidance: What it means for evidence strategy and real-world research

In June 2026, as part of HHS's Operation TrialBlazer, the FDA published revised draft guidance on demonstrating substantial evidence of effectiveness that largely codifies existing practice.¹ Once finalized, it will replace the Agency's 1998 guidance. 

Single-trial approvals stopped being the exception some time ago. The average number of studies behind a novel FDA approval fell from 3.41 in 2016 to 1.39 in 2024, and 69.4% of novel products approved in 2024 rested on a single study.²  

Let's be clear about what has not changed. The statutory standard is the same, and the bar for approval has not been lowered. What the guidance does is spell out how a sponsor can clear that bar with one adequate and well-controlled investigation plus confirmatory evidence, and what makes that combination persuasive. 

This is a significant moment for real world research. The guidance confirms that an adequate and well-controlled investigation can be interventional or non-interventional, and it names natural history data, registries and external controls among the potential contributors to the evidence base.¹ That is a real opening. It is also a demanding one. The sponsors who benefit will be those who design their real-world evidence (RWE) with the same discipline they bring to the pivotal trial, and who start on day one. 

One trial is not a shortcut 

Development programs have traditionally met the substantial evidence standard with at least two adequate and well-controlled trials. The revised draft sets out several routes to the same standard, including: 

• One adequate and well-controlled investigation with confirmatory evidence 
• More than one adequate and well-controlled investigation 
• Reliance on the known effectiveness of an approved drug for a new population, dose, regimen, route or dosage form, where scientifically justified and legally permissible 

On the single-investigation route, FDA will look hard at the design, conduct, analysis and persuasiveness of that investigation, and at the source and strength of the confirmatory evidence. Disease context, unmet need and whether a second trial would be practical or ethical all factor in. 

Two points deserve emphasis. First, confirmatory evidence and RWE are not the same thing. Confirmatory evidence can come from earlier clinical studies, trials in related indications, drugs in the same pharmacological class, mechanistic data, and natural history or registry data.¹ RWE is one option among several, and it has to earn its place. Second, the pieces trade off against each other. A highly persuasive trial generally requires less support from confirmatory evidence. As uncertainty increases, greater weight is placed on the credibility, independence and overall strength of the supporting evidence. 

The guidance is an invitation to design a coherent evidence package. Sponsors who interpret it simply as a way to replace two trials with one risk overlooking the amount of planning required to make a single-trial strategy credible. 

Quality still decides the outcome 

Statistical significance on its own will not carry the day. The guidance expects an effect that is clinically meaningful, sufficiently precise and robust to reasonable analytical assumptions. For a single highly persuasive investigation, it points to a representative population, clinically meaningful endpoints, adequate power, supportive results across secondary endpoints and subgroups, comprehensive follow-up and minimal missing data. 

Flexibility will not rescue a weak design or poor execution. If anything, it raises the stakes on getting the design right. 

Don't overload the pivotal trial 

When one trial carries more of the regulatory case, the instinct is to load it with every question the program wants answered. That makes an existing problem worse. A Tufts CSDD and TransCelerate analysis of 105 protocols found that the average Phase 3 protocol now collects around 5.9 million data points, a figure growing 11% a year since 2020. It also found that 46% of non-core procedures in Phase 2 and 35% in Phase 3 gather data for exploratory or future use.³ Every extra procedure adds cost, site and patient burden, and execution risk, at exactly the point where the program can least afford it. 

The better answer is to design the pivotal investigation and the wider evidence package together. Keep the trial focused on the question it must answer. Move durability, generalizability and longer-term outcomes into prespecified follow-up using linked real-world data (RWD), a registry or a companion observational study. And plan the connection before first patient in: consent for linkage, common endpoint definitions and follow-up arrangements cannot be bolted on afterwards. 

Endpoints are where it all connects 

The draft states a preference for clinical endpoints where feasible, with surrogates as an alternative approach.¹ Beyond that, the practical test is whether an endpoint works across the whole package. It should be clinically meaningful and validated in the intended population. Where an external or natural history comparison is planned, it needs to be objective and consistently ascertainable, because assessor-dependent outcomes are the most exposed to bias without randomization. And if real-world follow-up is meant to extend the trial, check early whether the outcome is actually captured in routine care. Often it is not: imaging, specialist assessments, biomarkers and many clinical outcome assessments (COAs) need dedicated collection. 

Burden matters too. More than half of non-core and non-essential trial data comes from clinical and patient-reported questionnaires.³ A strong COA strategy uses the fewest instruments needed to answer the regulatory and value questions. 

What this means for real-world evidence 

The guidance is mainly about clinical trials and refers sponsors to FDA's existing RWE guidance for the detail.¹ It should not be interpreted as observational studies can be a routine replacement for randomized trials. What it does signal is that the role of RWE belongs in the effectiveness strategy from the start. 

That role varies. RWE can describe natural history, inform endpoints and event rates, support trial design and feasibility, provide an external comparator in the right circumstances, test whether results hold in broader or under-represented populations, and extend follow-up. Each of these have different data requirements. A source that works for feasibility may be nowhere near good enough for a comparator. A registry that describes disease progression may lack the endpoint ascertainment needed for causal analysis. A large claims or EHR database may have the breadth but not the clinical depth. Define the role first, then choose the data and the design. 

Two illustrative examples 

A rare disease with a well-characterized course. The sponsor starts a prospective natural history study during Phase 2, using the eligibility criteria, endpoint definitions and assessment schedule planned for the pivotal trial. When the single pivotal study reads out, that cohort provides prespecified context for the untreated course of disease, instead of a comparator assembled after the result is known. Its weight will still depend on comparability, consistent ascertainment and the size of the effect, but prospective alignment makes those questions answerable. 

A common condition with a strong pivotal trial. The trial is lean and focused on its primary endpoint. Participants consent to linkage with routine-care records, and a prespecified observational follow-up tracks durability and outcomes in patients the trial under-represented. That evidence may not be central to approval, but it answers the questions regulators, payers and prescribers will ask next. 

Fit for purpose, not just available 

There has never been more healthcare data, but availability is not regulatory fitness. The right starting question is "What evidence do we need, and which data can generate it?" 

Before a data source becomes a dependency in the regulatory strategy, sponsors should know whether it represents the intended population, captures exposure accurately, and records the outcome reliably and consistently. For comparative questions, they need to know whether follow-up can be aligned across groups, whether key prognostic factors and confounders are captured, how much is missing, and whether care pathways or data capture have shifted over time. Provenance must be transparent, and the analysis must be capable of prespecification and independent reproduction. 

Do this work before the pivotal trial starts. Data sources found after results are known carry limited confirmatory value and invite questions about selection and analytical flexibility. 

External controls: a high bar and an early decision 

The guidance accepts that externally controlled investigations can provide evidence of effectiveness in some circumstances, while being clear about the potential for bias.¹ Without randomization, differences in baseline characteristics, concomitant treatment, start of follow-up, endpoint ascertainment and knowledge of treatment can all produce differences that have nothing to do with the therapy. 

FDA describes the conditions that strengthen an external control: a well-defined natural history, a control population that closely resembles the treated one, comparable concomitant therapy, an objective and clinically meaningful endpoint measured the same way, aligned start of observation, and a prespecified analysis showing an effect large enough that bias is unlikely to explain it. That is a high bar, and it is exactly why external controls need to be evaluated before the principal investigation begins. An external control should be part of the plan from the outset. Used as a rescue option for a struggling trial, it will rarely convince. 

Registries and natural history studies as planned assets 

Natural history studies and registries are particularly valuable in rare diseases and small or heterogeneous populations. Where the untreated course of a disease is well established, the guidance notes that natural history evidence can, in some situations, support the conclusion that an outcome would not have occurred without treatment.¹ 

Their value depends on how they were built. A registry set up for disease awareness or surveillance often lacks the variables a regulatory question needs. If a registry may one day support a submission, design it that way from the start: population, eligibility, index date, endpoint definitions and ascertainment, visit schedules, source data verification, retention, governance and the statistical analysis plan. Prospective design does not guarantee confirmatory value. It does stop the asset from becoming an expensive dataset that cannot answer the question that matters. 

Talk to FDA early, and be specific 

The draft encourages sponsors to discuss their approach with FDA early, potentially at the pre-IND stage and no later than the end-of-Phase 2 meeting, with a clear scientific justification.¹ That conversation has to go further than whether a data source is acceptable. Sponsors should be ready to explain what each component contributes, how independent the confirmatory evidence is, how the design controls bias, which assumptions could change the interpretation, and why the pieces address the same clinical question. A set of individually credible studies does not automatically add up to a coherent evidence package. 

Early agreement does not guarantee approval; FDA still judges the totality of evidence at review. But a strategy tested with the Agency early is far less likely to fail late over a fundamental disagreement about what the evidence is for. 

Five things to do now 

1. Define the regulatory question early: the claim, population, endpoint and level of certainty required. 
2. Map the whole evidence package: what the pivotal investigation must establish, and what each confirmatory source adds. 
3. Test RWD before you depend on it: coverage, outcome capture, follow-up, confounders, provenance and analytical feasibility. 
4. Build observational assets prospectively: aligned to their intended regulatory use, with consent, linkage and common definitions in place from the start. 
5. Go to FDA with a specific rationale: why the combination is sufficient, and how the design deals with bias and uncertainty. 

Looking ahead 

The revised guidance leaves the bar where it was. It widens the routes to clearing it and puts the onus on sponsors to show that their evidence package hangs together. Access to data will matter far less than the ability to generate regulatory-grade evidence for a clearly defined purpose, and to bring clinical, epidemiological and real-world expertise together early enough to shape the program. 

How Parexel can help 

Parexel brings clinical development, regulatory strategy, epidemiology, biostatistics, clinical outcome assessment and real-world research together in one evidence team. We work with sponsors to define the regulatory question, test whether RWD is genuinely fit for purpose, and design and deliver the prospective registries, natural history studies, external comparator strategies and observational follow-up the question requires. The result is a focused pivotal trial, a wider evidence package that anticipates the next set of questions, and a strategy that is rigorous, realistic and aligned from the outset. 

To discuss how the revised guidance could shape your development program, contact the Parexel Real-World Research team. 

References 

1. US Food and Drug Administration (2026). Demonstrating Substantial Evidence of Effectiveness for Human Drug and Biological Products: Revised Draft Guidance for Industry. Notice of availability, Federal Register, 24 June 2026 (91 FR 37994), docket FDA-2019-D-4964. federalregister.gov 
2. Narayan A, Irvin VL, Koong AJ, Song S, Kaplan RM (2026). Changes in evidence used for FDA novel drug approvals: 2016–2024. PLoS One 21(8): e0342878. doi.org/10.1371/journal.pone.0342878 
3. Tufts Center for the Study of Drug Development and TransCelerate BioPharma (2025). Insights informing strategies for optimizing the collection of clinical trial data. Therapeutic Innovation & Regulatory Science. doi.org/10.1007/s43441-025-00899-4 

Return to Insights Center