How to write the RWE section of a regulatory briefing document: structure it for the reviewer, not the study report

You did the hard part. You found a data source that is genuinely fit-for-purpose, pre-specified the SAP, thought hard about confounding and missing data, and assembled an RWE package that would hold up against everything regulatory-grade RWE demands. Good. That work is real, and most teams underestimate how much of it there is.

Then the RWE section of the briefing document gets handed to whoever wrote the study report. And it quietly inherits that report's shape: background, methods, results, discussion. Tidy and familiar, and built for a reader who isn't the one about to open it.

A reviewer does not read your briefing document the way you wrote it.

I'll call it the manuscript reflex: letting the study report's own running order decide how the briefing document gets built. It feels like the safe move, and it buries the two things the reviewer opened your section to find.

The reviewer reads your RWE section by question, not by chapter

Start with how these documents are actually used, because there is real evidence on it and not just opinion.

The Frederick National Laboratory's SOP 24405, which draws its content list straight from FDA's own Formal Meetings guidance, sets out the preferred order for a meeting information package. Product and background housekeeping come first, in items one to eleven. Then item twelve: a list of the final specific questions for discussion, each with a brief summary. Then item thirteen: the data to support discussion, organised by discipline and question. Read that again: by question, not by study chronology.

The same SOP is blunt about the mechanics. FDA, it says, "prefers to address the questions raised by the applicant as listed in the information package" [1] rather than sit through a walkthrough. And the package "should be bookmarked to enhance the reviewer's navigation across the different sections." [1] The word navigation, in a document derived from FDA's own guidance, is about as direct a validation as this argument gets.

You see the same instinct in the methodology literature. The STaRT-RWE reporting template (Wang et al. 2021) was built, in its authors' words, to "allow reviewers to quickly orient and find key information," [2] using tabular formats that keep the detail retrievable instead of drowned in prose. Two separate communities, a federal laboratory and a group of pharmacoepidemiologists, arrived at the same design goal: help the reader find the answer fast.

Now the anti-pattern, which is the standard trade-vendor template for briefing documents: background, then data summary, then regulatory status and plan, then questions for the agency, then appendices. The questions come fourth, and the sponsor's narrative comes first. That is the manuscript reflex dressed up as a house style, and it is the inverse of the order FDA's own content list prefers.

One caveat, stated precisely, because it matters: FDA's Formal Meetings guidance is a draft. The current version was published in the Federal Register on 22 September 2023 (docket FDA-2017-D-6530). It superseded a December 2017 version that was itself a draft, which in turn had withdrawn the last final guidance, from May 2009, and no finalisation has appeared as of a Federal Register index checked in November 2025. [3] So the honest attribution for that content order is a federally funded laboratory's operating procedure citing draft FDA guidance, not a final rule. It still tells you how reviewers work just don't oversell where it comes from.

A few years ago I sat in a meeting where the agency's opening question landed squarely on a study limitation, and our briefing book had answered it, thoroughly, except that the answer happened to live in a closing subsection several pages past the results, under a heading nobody would think to jump to. We spent the first stretch of a fixed meeting window flipping through our own document to find our own answer, while the reviewer waited. The evidence was there; the navigation was not. That is a self-inflicted wound.

So the operational move is simple to state and uncomfortable to do: reorganise the section around the questions the agency will ask, not the order in which you ran the study.

The highest-leverage paragraph, and where teams bury it

If a reviewer reads by question, the first question is always the same. Why this data source and this design, for this question? Everything else in the section is downstream of that answer. Which makes the fit-for-purpose justification the single highest-leverage paragraph you will write, and it belongs at the top of the section, before any methodology walkthrough.

Nobody is arguing about the requirement itself. Three separate agencies land on it, independently, in writing.

The EMA Reflection Paper on RWD in non-interventional studies, adopted by the CHMP on 17 March 2025 (EMA/99865/2025), states in §7.2 that the evaluation should "lead to a justification... of why they are deemed fit-for-purpose to answer the research question." [4] The FDA's final guidance on assessing electronic health records and medical claims data (July 2024) names "the appropriateness and potential limitations of the data source for the study question" [5] as one of only a handful of things a protocol must address. The MHRA's draft guideline on RWD external control arms, out for consultation in 2025, puts it plainly in §16: "the choice of external data source should be robustly justified." [6]

No study tests whether moving the fit-for-purpose justification to the top of a document changes a reviewer's decision. Its necessity is thoroughly evidenced; its placement is our argument, not a citation. But the logic follows the way these documents are read: if the reviewer opens the section to ask "why this data," don't make them reconstruct the answer from a methods section three pages down. Hand it to them first.

There is a clean public illustration of a sponsor doing exactly this. The APPEX study (NCT05842486) exists, per its own registry entry, "to contextualize results from the APPOINT-PNH trial with iptacopan." [7] The fit-for-purpose reasoning is written into a public field, in one line, before anyone reaches a method. It is the same move a biotech makes when it uses RWE to give a single-arm pivotal the external context regulators come asking for. AstraZeneca and Daiichi Sankyo did something comparable with ESPERANZA (NCT06973161), stating the tumour-agnostic regulatory logic for a T-DXd external control directly in the registry. [8] If you are building an external control specifically, the FDA external control arm checklist is the companion to this piece.

The Monday version: draft the fit-for-purpose paragraph before you write anything else in the section. If you cannot state it in a few sentences without reaching for the methods, the section is not ready, and possibly the regulatory rationale isn't either.

Limitations are the credibility engine, not the fine print

The manuscript reflex has a second casualty, and it is the more damaging one. Limitations get demoted to a closing disclaimer, a tidy paragraph under the results table that a reader can skim past. To an experienced reviewer, that reads as either naivety or concealment, and neither helps you.

The best evidence for the opposite approach is the deferiprone story. In Jahanshahi et al. 2021, FDA's own statistical reviewer wrote that the study "has several serious limitations including lack of randomization, lack of control group, high rate of missing data and ignoring the variation between studies by simply pooling." [9] A brutal read, and FDA approved anyway. The reviewer concluded the trial "can be considered an adequate and well-controlled trial under the CFR and ICH E10 guidance," [9] because a prospectively planned statistical analysis plan and independent-committee patient selection "minimized the possibility of bias." [9] Sit with what that means. The limitations were named, in the reviewer's own hand, and the approval rested on how they had been anticipated and mitigated, not on their absence.

Now the disclosure gap. Vaghela et al. 2024 looked at twenty rare-disease applications using RWD. Only 15% reported methods for handling bias and missing data, and 85% reported no missing-data methods at all. Yet 45% still drew positive FDA feedback, attributed partly to "the appropriateness and justifiability of the RWD design," [10] while 55% drew concerns about RWD implementation. The teams that engaged with their weaknesses gave the reviewer something to say yes to.

The EMA frames this as normal, not as failure. Its Reflection Paper (§7.1) accepts that some data-quality issues are "difficult or impossible to resolve," [4] and asks only that those uncertainties be "clearly identified in the feasibility assessment and the study protocol." [4] Disclosed, not hidden.

So give limitations their own signposted block in the body of the section, each one paired with what you did about it, not a one-line caveat under a table. A named limitation with a named mitigation is an asset. A hidden one is a question you have handed the reviewer and refused to answer.

The reusable structure: five questions, in order

Here is the template. It is built as five questions in the reviewer's head, in the order they get asked. Lift it into your next briefing book and fill it in.

1. What regulatory question does this evidence answer? State it the way the agency will phrase it, and tie it to the specific meeting question number. This replaces "background." The reviewer wants to know what they are being asked to agree to before they read a single result.

2. Why are this data source and this design fit-for-purpose for that question? The top paragraph from above. Where the data came from and why it fits the question, before any methodology detail. This is the paragraph that earns you the reviewer's attention for everything below it.

3. What does the evidence show, read against the question? Findings interpreted against the question you posed in item one, not a full results dump. Use STaRT-RWE-style tables so the detail stays retrievable without prose burying it, and keep the heavy methodology in appendices, referenced from here.

4. What does it not show, and how did you handle that? The signposted limitations-and-mitigations block. Every material limitation, paired with its mitigation, in the body, in daylight.

5. What are you asking the agency to agree? The specific ask, closing the loop back to item one. Skip the limp "we welcome the agency's views" and name the precise point of alignment you need out of this meeting.

Set that beside the manuscript order (background, methods, results, discussion) and you can see exactly what moved. The question jumps to the front, fit-for-purpose climbs out of the methods section where it used to hide, and limitations come out of the footnotes and into the body, in daylight rather than small print. Results stop being the destination; they become the answer to a question you already framed for the reviewer.

Two mechanical notes. Bookmark the section, per Frederick's §7.6, so a reviewer can jump straight to any of the five. And treat this as a living document, not a fixed form. A tight Type C meeting on a single question does not need all five sub-blocks at full length. Trim to the window.

The honest objection: they'll find it anyway

Reviewers are trained scientists who read statistical reports for a living, a manuscript structure of background through discussion is precisely the order they were trained on, and they will find your fit-for-purpose logic and your limitations wherever those things sit, because reading past a document's layout to the substance underneath is the entire job. So reordering the section is cosmetic at best, and at worst it looks like you are managing the reader, which a good reviewer resents. Fair challenge.

There are two answers, and the first is that it is not cosmetic to them. The Frederick SOP shows FDA explicitly prefers question-organised packages and bookmarked navigation, and prefers to address the applicant's listed questions rather than sit through a walkthrough, all inside a meeting window that does not stretch. Structure does not flatter the reviewer. It respects their clock.

The second answer is a limit on the claim itself, and I would rather state it than have you find it later. Structure does not manufacture agreement. Jaksa et al. 2022 handed seven identical oncology external-control case studies to FDA, EMA, Health Canada and five HTA bodies, and found that "agreement in critiques... was low," [11] even though everyone converged on the same broad categories of selection bias and confounding. Sarri and Hernandez 2024, scanning 46 guidance documents, found the agencies aligned on the phases of RWE evaluation but differing so much in detail and specificity that they call the landscape a "maze." [12] That said, no structure gets you a uniform yes across all of them. The defensible claim is narrower and true: transparent fit-for-purpose justification and open limitations reduce your risk with every one of them. Structure earns you the reviewer's attention; it does not buy their agreement.

And that attention is scarce: Li et al. 2026 found only 10.8% of FDA approvals between 2016 and 2024 leaned on even one premarket RWE study, with a further 34.5% carrying a postmarket RWE requirement or request. [13] RWE is still a minority-use, high-scrutiny evidence type. When you do bring it, you cannot afford to spend the reviewer's attention making them hunt for your argument. The broader craft of taking an RWE package into a regulatory interaction is its own discipline; this piece is about the one document.

What we are all reverse-engineering

Notice what is missing from all of this. There is no FDA content template for the RWE section of a briefing document, because there is no finalised FDA guidance on briefing documents at all. The current one has sat in draft since September 2023, the version before it was a draft too, and the last final version was withdrawn back in 2017, so that as of a Federal Register check in November 2025 nothing has been finalised and we are all reverse-engineering the preferred structure from a national laboratory's SOP and the way reviewers actually behave in the room. [3]

Until FDA publishes one, the structure of your RWE section is yours to own. So own it. Build it around the reviewer's questions, put fit-for-purpose at the top, give limitations the daylight they earn, and let the study report keep its own shape in the appendix where it belongs.

If you want a second pair of eyes on how an RWE package is structured for a specific FDA or EMA interaction, that's exactly the kind of thing we do at Inovia Bio, and we'd rather look at the section before the meeting than hear how it went afterwards. Bring the evidence; we'll help you make it navigable.

References

[1] Frederick National Laboratory for Cancer Research. Standard Operating Procedure 24405, "Preparing a Meeting Information Package for a Pre-IND (Type B) Meeting" (Rev 06, effective 2 November 2020) — §7.1 content list (items 1–13) and meeting-conduct provisions §7.5, §7.6. The SOP states in §7.1 that its content list follows FDA's "Guidance for Industry: Formal Meetings Between the FDA and Sponsors or Applicants (of PDUFA Products)."

[2] Wang SV, Pinheiro S, Hua W, et al. (2021). "STaRT-RWE: structured template for planning and reporting on the implementation of real world evidence studies." BMJ;372:m4856. PMID: 33436424. https://pubmed.ncbi.nlm.nih.gov/33436424/

[3] FDA. "Formal Meetings Between the FDA and Sponsors or Applicants of PDUFA Products" — DRAFT guidance, Federal Register 22 September 2023, docket FDA-2017-D-6530. Supersedes a December 2017 draft, which withdrew the last final version (19 May 2009). No finalisation identified as of a Federal Register index dated 18 November 2025.

[4] EMA. "Reflection Paper on the use of Real-World Data in non-interventional studies to generate real-world evidence" — FINAL, adopted by CHMP 17 March 2025, EMA/99865/2025 (§7.1, §7.2). Read in full via EMA. https://www.ema.europa.eu

[5] FDA. "Real-World Data: Assessing Electronic Health Records and Medical Claims Data To Support Regulatory Decision-Making for Drug and Biological Products" — FINAL, July 2024, docket FDA-2021-D-2038. Verified against the primary FDA guidance PDF (fda.gov/media/152503/download): the quoted string is item (1) of the four data issues the guidance states "should be addressed in the protocol" (Section III).

[6] MHRA. "Guideline on the use of real-world data in external control arms" — DRAFT, out for public consultation 20 May–14 July 2025 (§16). No finalised version published as of the research date.

[7] APPEX — "Hematological Response in Patients With Paroxysmal Nocturnal Hemoglobinuria Treated With Anti-C5 Antibody: an External Control Arm Study for Iptacopan Use in Anti-C5 naïve Patients." Sponsor: Assistance Publique – Hôpitaux de Paris (collaborator: Leeds Cancer Centre); the study contextualises Novartis's iptacopan pivotal, APPOINT-PNH. ClinicalTrials.gov: NCT05842486. https://clinicaltrials.gov/study/NCT05842486 (paired pivotal: APPOINT-PNH, sponsor Novartis, NCT04820530, https://clinicaltrials.gov/study/NCT04820530)

[8] AstraZeneca / Daiichi Sankyo. "ESPERANZA — external control arm study for trastuzumab deruxtecan (T-DXd)." ClinicalTrials.gov: NCT06973161. https://clinicaltrials.gov/study/NCT06973161

[9] Jahanshahi M, Gregg K, Davis G, et al. (2021). "The Use of External Controls in FDA Regulatory Decision Making." Ther Innov Regul Sci;55(5):1019–1035. PMID: 34014439. https://pubmed.ncbi.nlm.nih.gov/34014439/ (deferiprone/Ferriprox reviewer quotations)

[10] Vaghela S, Tanni KA, Banerjee G, 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;19(1):117. PMID: 38475874. https://pubmed.ncbi.nlm.nih.gov/38475874/

[11] Jaksa A, Louder A, Maksymiuk C, 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):1967–1976. PMID: 35760714. https://pubmed.ncbi.nlm.nih.gov/35760714/ (seven oncology external-control case studies reviewed by three regulators — FDA, EMA, Health Canada — and five HTA bodies; cross-agency critique agreement low)

[12] Sarri G, Hernandez L. (2024). Environmental scan of 46 RWE guidance documents across FDA/EMA/NICE/HTA bodies. J Comp Eff Res. PMID: 39132748. https://pubmed.ncbi.nlm.nih.gov/39132748/

[13] Li LY, Ramachandran R, Ross JS, Wallach JD. (2026). "Premarket and postmarket real-world evidence studies supporting U.S. Food and Drug Administration regulatory decision-making, 2016–2024." Clinical Trials. PMID: 41641793. https://pubmed.ncbi.nlm.nih.gov/41641793/