You are designing a real-world evidence study, and the sensible first move looks obvious. Pull the current FDA guidance. Read it line by line. Build the design to match its language, cite the right paragraphs in your protocol, and walk into the review able to say you complied with the agency's latest thinking. Reasonable instinct, and on a programme that runs three, five, seven years from first data to filing, it is also a quiet bet that the document you built to will still exist in its current form when you get there.
But it usually will not.
Since the FDA published its Framework for the RWE Program in December 2018, the agency has issued at least ten distinct RWD/RWE-relevant guidance documents or major revisions [1]. That is not a loose impression. Sarri and Hernandez (PMID 39132748), scanning the guidance landscape across the FDA, EMA, MHRA, NICE and four HTA bodies, counted the FDA holding the highest tally of anyone reviewed, ten documents by their March 2024 cut-off, out of forty-six RWE documents in total, all but two of them published after 2018 [1]. Two of the FDA's load-bearing RWE guidances are still drafts years after they opened for comment. A guidance document is a photograph of a field that is still moving.
So here is the discipline that future-proofs your RWE study design, and you can start applying it this week. The durable target is the method, not the paragraph number. Anchor your design choices in the small set of methodological principles that have survived every one of those revisions, evidence them with an auditable trail, and the package holds when the guidance shifts underneath it.
When the FDA critiques an RWE submission, the questions land on the methodology. Reviewers are not checking whether you cited last year's draft, and we know this because the pattern has been catalogued. Arondekar and colleagues (PMID 34667027) reviewed thirteen oncology approvals informed by RWE between 2015 and 2020 and found the agency's critiques clustered on a stable set of issues: the "lack of the following: a prespecified study protocol, inclusion/exclusion criteria matching to the trial, comparability of endpoint definitions, methods to minimize confounding... and plans to handle missing data" [2]. Nothing there about which guidance version was current.
Now the useful part: Vaghela and colleagues (PMID 38475874) ran the same exercise on a completely different, non-overlapping set of twenty rare-disease approvals between 2017 and 2022. The FDA criticised eleven of the twenty, 55%, on the same categories [3]. Two independent review samples, coverage nine years apart, the identical axes of scrutiny both times, across a stretch in which the guidance was rewritten repeatedly. The scrutiny tracked the method, not the documents.
The agency more or less says as much itself. The FDA's final guidance on assessing EHR and medical claims data, finalised in July 2024, "does not endorse any particular set of guidelines or checklists" and instead expects sponsors to evaluate the fitness of their data on a case-by-case basis [4]. That is the same point behind What does regulatory grade RWE mean? regulatory-grade is a property of your process, not the brand name of your dataset.
So what changes on Monday? You stop trying to pass review by paragraph-matching, because there is no paragraph to match against, and start defending each design choice on the axes reviewers actually probe.
Here are the four principles that have not moved, and they are the spec you design to. For each, one concrete case where the guidance changed and the principle did not.
Provenance and traceability. Could you trace every data point back to its origin, and was the record accurate and complete? Riskin and colleagues traced exactly this reliability yardstick, documented step by step, across the 2016 Cures Act, the 2018 Framework, the 2021 draft and the 2024 final, and it survived all four [5]. The cleanest before-and-after is the medical-device RWE guidance: 2017 final, then a 19 December 2023 draft, then an 18 December 2025 final that quietly dropped the expectation of patient-level identifiability, opening the door to de-identified registries, a genuine liberalisation [6]. The provenance principle underneath it did not budge.
Pre-specification. A recurring FDA critique category in both Arondekar and Vaghela, and a discipline that has been demonstrated, not just recommended. RCT-DUPLICATE registered its protocols before the hazard ratios and confidence intervals were computed, first across ten trials [7], then across thirty-two [8]. The FDA's externally-controlled-trials draft, from February 2023 and still a draft, calls for pre-specifying the ECA protocol and statistical analysis plan rather than picking a control after the single-arm trial reads out [9]. Inovia built a whole checklist around that draft in FDA Guidance on External Control Arms, and the pre-specification demand in it is the least likely thing to change when it finally finalises.
Relevance and reliability, i.e. exchangeability. Exchangeability is not new regulatory fashion. As a named criterion it traces to Pocock's six conditions for a valid external comparator, published in 1976, and FDA reviewers still lean on it today, fifty years on [10]. The fitness-for-purpose vocabulary that grows out of it runs through the entire FDA suite unchanged [1].
Transparent handling of missing data and confounding. The other category that recurs in both Arondekar and Vaghela. A pre-declared plan to minimise confounding, and a principled approach to missingness, written down before you see the results rather than reverse-engineered after them.
Notice the pattern. The device guidance dropped a patient-identifiability requirement, a real change in what the agency will accept, while the externally-controlled-trials guidance has sat in draft for over three years. Neither touched any of these four. The guidance moved. The design did not have to.
Free download
The RWE Briefing Document Template
The section-by-section structure for the RWE part of a regulatory briefing, built around the questions reviewers actually ask.
Get the template →You might be tempted to wait for it all to settle. Do not. It is not going to settle, and that is by design.
The churn is written into law. The 21st Century Cures Act, section 3022, created the RWE programme in 2016 [11]. FDORA, section 3629, then directed the FDA in 2022 to keep issuing and revising RWD/RWE guidance on an ongoing basis [12]. The agency revises because the statute tells it to, again and again, with no end date attached.
And the evidence that it will keep moving is everywhere you care to look. The externally-controlled-trials draft from February 2023 is still a draft, with no finalisation announced as of July 2026, over three years in limbo [9]. The non-interventional-studies draft from March 2024 is likewise still a draft [13]. On the EMA side the picture is the same: the agency stood up DARWIN EU, its real-world data network, which has run through four periodic reports since the tender was awarded in February 2022 [14]; its Reflection Paper on RWD in non-interventional studies went from draft consultation in 2024 to a CHMP-adopted final on 17 March 2025 (EMA/99865/2025), shipping with a companion roadmap that promises more guidance to come [15]. ICH endorsed its own RWE Reflection Paper on 4 June 2024 and called it, in its own words, "the initial step of an incremental approach" [16]. ICH still has no finalised RWE guideline at all.
Think of the current draft as the waterline. It rises and falls with each consultation cycle, each comment period, each revision. Build your evidence package at the waterline and you get flooded on schedule. The four principles are the bedrock; build above the flood mark.
You can put your hands on a durable design: it is a real, observable object.
Many years ago I worked on a rare-disease registry first sketched against guidance that has since been revised more than once. Nobody at that table knew which way the documents would move. What we did have were endpoints chosen because they were clinically meaningful and would stay that way whatever the regulatory weather turned out to be. When the guidance moved, the registry absorbed it without anyone having to rebuild the thing.
The public record is full of the same shape. The Fabry Outcome Survey (NCT03289065) ran continuously from April 2001 to September 2021 [17]. That start date is fifteen years before the Cures Act even created the RWE programme, and the run closes three years past the 2018 Framework: two decades, straight through the birth of the entire modern RWE apparatus, resting on seven pre-specified, clinically durable endpoints. The Gaucher, Fabry and Pompe registries (NCT00358943, NCT00196742, NCT00231400) are each registered to run to 2034 [18]. The Gaucher registry began in 1991, and its stated objectives name eliglustat, a therapy the FDA did not approve until 2014, twenty-three years after the registry opened. The design absorbed a drug that did not exist when it was written, without being torn up.
The methodologists have made the same move deliberately. STaRT-RWE (PMID 33436424), the reporting template published in the BMJ in 2021 and co-authored with EMA and PMDA personnel, requires "a version history documenting not only what changed, but also why... and when" [19]. Read that again: a discipline built explicitly so a study survives any one agency's document being revised. HARPER, published in 2023, did the archaeology, harmonising the elements already common to four pre-existing templates (EMA's GVP Module VIII-PASS, ISPE's GPP, NEST and STaRT-RWE) [20]. That is about as direct a piece of evidence as you will find for what has stayed constant while the guidance churned.
The same instinct anchors the minimum viable evidence framework: decide which evidence element earns its place before you collect it, rather than after the fact.
"Shouldn't I just follow the current guidance exactly? The agency wrote it for a reason. A reviewer is going to ask why I ignored the current thinking, and 'I designed for principles' sounds an awful lot like a licence to be vague and skip the reading." That is a fair challenge.
This is not a licence to ignore guidance. Read every relevant draft and final, map each design choice to it, engage with what the agency is signalling. Designing for principles and reading the guidance closely are the same activity, not opposites. Ignoring a genuine guidance signal is exactly as reckless as redesigning your study every time a draft twitches. Both are ways to lose the plot.
Two honest limits:
First, methodological transparency determines whether your flaws are forgivable, not whether you are approved regardless. Both Arondekar and Vaghela found the FDA accepting RWE with acknowledged weaknesses when the effect size was large [2][3]. Transparency buys forgiveness, but it does not buy you a large effect, and it will not rescue a genuinely unfit data source.
Second, this buys durability against one agency revising its own guidance over time. It does not buy you two agencies agreeing on the same day. Jaksa and colleagues put the same seven oncology external-control case studies in front of multiple regulators and HTA bodies and found low agreement between them [21]. Subramaniam and colleagues found the EMA's external-controls position "takes a more cautious stance than the FDA despite similar numbers of approvals" [22], a divergence explored in The EMA vs Single-Arm Trials. Durable design is protection against time, and it is no treaty between agencies.
So, concretely, from your next design meeting:
Map, don't chase. For every material design choice, write down which durable principle it serves and which current guidance passage it currently aligns with, dated. When the passage changes, you re-point one line in a mapping document, not the study.
Keep a version history. STaRT-RWE style: what changed, why, and when. When the guidance moves you can show a reviewer that your reasoning already anticipated the axis it moved on, which is a very different conversation from scrambling to retrofit a justification.
Pre-specify to the principle, then engage the agency on the specifics. Lock the protocol and SAP to the durable spec early, then take the genuinely guidance-specific questions to the FDA or EMA before you are committed to an answer. Early engagement remains the highest-yield thing you can do, as argued in How to use RWE to support regulatory strategy.
Build the audit trail as you go. Provenance captured retrospectively is provenance you cannot prove. Do it at the point of data handling, not at submission, by which time the people who know how a field was derived have often moved on.
None of this is exotic. It is the evidence architecture we help teams build toward a future-proof RWE study design, and the kind of guidance-landscape mapping InovaCS is designed to make faster, so that when a design choice is questioned you can name the principle it rests on and the document it currently maps to in a single move.
A durable design cannot buy you everything. It will not make the EMA and the FDA agree on a Tuesday. That is a different problem, and anyone selling you a clean solution to it is selling. Your package stays standing when the guidance moves, though, which it will, on a schedule Congress has written into law.
One thing would genuinely help the rest of us: the FDA finalising the externally-controlled-trials guidance. Three and a half years is a long time to design against a document that could still change its mind.
Get the monthly digest
The 5 things evidence leads need to know each month: regulatory moves, RWE developments and what they mean in practice. No pitch, one email a month.
[1] Sarri G, Hernandez L. (2024). "The maze of real-world evidence frameworks: from a desert to a jungle! An environmental scan and comparison across regulatory and health technology assessment agencies." J Comp Eff Res. PMID: 39132748. https://pubmed.ncbi.nlm.nih.gov/39132748/
[2] Arondekar B, et al. (2022). "Real-World Evidence in Support of Oncology Product Registration: A Systematic Review of New Drug Application and Biologics License Application Approvals from 2015-2020." Clin Cancer Res;28(1). PMID: 34667027. https://pubmed.ncbi.nlm.nih.gov/34667027/
[3] Vaghela S, et al. (2024). "A systematic review of real-world evidence (RWE) supportive of new drug and biologic license application approvals in rare diseases." Orphanet J Rare Dis. PMID: 38475874. https://pubmed.ncbi.nlm.nih.gov/38475874/
[4] FDA. "Real-World Data: Assessing Electronic Health Records and Medical Claims Data To Support Regulatory Decision-Making for Drug and Biological Products." FINAL guidance, July 2024. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/real-world-data-assessing-electronic-health-records-and-medical-claims-data-support-regulatory
[5] Riskin DJ, et al. (2025). "Implementing Accuracy, Completeness, and Traceability for Data Reliability." JAMA Netw Open. PMID: 40063029. https://pubmed.ncbi.nlm.nih.gov/40063029/
[6] FDA. "Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices." FINAL guidance, 18 December 2025 (supersedes 2017 final; from 19 December 2023 draft). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/use-real-world-evidence-support-regulatory-decision-making-medical-devices
[7] Franklin JM, et al. / RCT-DUPLICATE (2021). "Emulating Randomized Clinical Trials With Nonrandomized Real-World Evidence Studies: First Results From the RCT DUPLICATE Initiative." Circulation;143(10). PMID: 33327727. https://pubmed.ncbi.nlm.nih.gov/33327727/
[8] Wang SV, et al. / RCT-DUPLICATE (2023). "Emulation of Randomized Clinical Trials With Nonrandomized Database Analyses: Results of 32 Clinical Trials." JAMA;329(16). PMID: 37097356. https://pubmed.ncbi.nlm.nih.gov/37097356/
[9] FDA. "Considerations for the Design and Conduct of Externally Controlled Trials for Drug and Biological Products." DRAFT guidance, February 2023 (still draft as of July 2026). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/considerations-design-and-conduct-externally-controlled-trials-drug-and-biological-products
[10] Gray CM, et al. (2020). "A Framework for Methodological Choice and Evidence Assessment for Studies Using External Comparators from Real-World Data." Drug Saf;43(7). PMID: 32440847. https://pubmed.ncbi.nlm.nih.gov/32440847/
[11] 21st Century Cures Act, §3022 (2016). Public Law 114-255. https://www.congress.gov/bill/114th-congress/house-bill/34
[12] Food and Drug Omnibus Reform Act (FDORA), §3629 (2022). https://www.congress.gov/
[13] FDA. "Real-World Evidence: Considerations Regarding Non-Interventional Studies for Drug and Biological Products." DRAFT guidance, 21 March 2024 (still draft; comments due 18 June 2024). https://www.fda.gov/regulatory-information/search-fda-guidance-documents/real-world-evidence-considerations-regarding-non-interventional-studies-drug-and-biological-products
[14] EMA. "DARWIN EU (Data Analysis and Real-World Interrogation Network)." Coordination Centre tender awarded 9 February 2022; four periodic reports through 2026. https://www.ema.europa.eu/en/about-us/how-we-work/big-data/data-analysis-real-world-interrogation-network-darwin-eu
[15] EMA. "Reflection paper on the use of real-world data in non-interventional studies to generate real-world evidence." FINAL, adopted 17 March 2025. EMA/99865/2025 (companion roadmap EMA/100019/2025). https://www.ema.europa.eu/en/documents
[16] ICH. "Reflection Paper on Real-World Evidence." Endorsed by the ICH Assembly, 4 June 2024. https://www.ich.org/
[17] Shire. "Fabry Outcome Survey (FOS)." ClinicalTrials.gov: NCT03289065. https://clinicaltrials.gov/study/NCT03289065
[18] Sanofi Genzyme. "Gaucher / Fabry / Pompe Registries." ClinicalTrials.gov: NCT00358943, NCT00196742, NCT00231400. https://clinicaltrials.gov/study/NCT00358943
[19] Wang SV, et al. / STaRT-RWE (2021). "STaRT-RWE: structured template for planning and reporting on the implementation of real world evidence studies." BMJ;372. PMID: 33436424. https://pubmed.ncbi.nlm.nih.gov/33436424/
[20] Wang SV, et al. / HARPER (2023). "HARmonized Protocol Template to Enhance Reproducibility of hypothesis evaluating real-world evidence studies on treatment effects: A good practices report of a joint ISPE/ISPOR task force." Pharmacoepidemiol Drug Saf;32(1). PMID: 36215113. https://pubmed.ncbi.nlm.nih.gov/36215113/
[21] Jaksa A, et al. (2022). "A Comparison of Seven Oncology External Control Arm Case Studies: Critiques From Regulatory and Health Technology Assessment Agencies." Value Health;25(12). PMID: 35760714. https://pubmed.ncbi.nlm.nih.gov/35760714/
[22] Subramaniam D, et al. (2024). "A Framework for the Use and Likelihood of Regulatory Acceptance of Single-Arm Trials." Ther Innov Regul Sci. PMID: 39285061. https://pubmed.ncbi.nlm.nih.gov/39285061/