Health authority query response, explained.

Health authority queries explained: what happens when a regulator writes back about your medicine

One FDA letter about Veltrimab, from arrival to a published reply

1. The agency sends a letterThe FDA's information request on Veltrimab arrives at 8:12 a.m. on October 6: 12 questions under 7 headings, due October 16.
2. The lead splits it and names ownersTwelve questions go to seven people across chemistry, manufacturing and controls (CMC), clinical pharmacology, immunogenicity, clinical, statistics and labeling.
3. Owners draft, and someone checks what was said beforeEach reply comes from the dossier and earlier answers. This is where the 4.6% told to Health Canada is caught, or missed.
4. Owners review, and the lead signsTwo replies wait on data due October 12 and October 13. Each signature records who signed, when and what it means.
5. Publishing files the sequence, and the answer joins the recordRegulatory operations files the reply as a new numbered package in the electronic dossier format: eCTD sequence 0012. The next agency's question will be checked against this answer.

Step 3 is the subject of this guide. The people, products and figures in this example are fictional.

Why the FDA clock keeps running, and three more query surprises

HAQ, information request, data cut and clock stop

Fourteen terms are defined in the one-page summary.

FDA information requests, EMA Day 120 lists, clarifaxes and section 31: one kind of letter under seven names

In shortEach regulator has its own name, format and clock for the same basic event. Response windows run from a few days to 6 months. Some clocks pause while you answer and some do not. A team with one product in several markets holds all of these at once, and the rules for what you may add in an answer differ too.

What you may add to an answer depends on the agency and the moment

Rules that govern agency replies, and what changed between 2025 and 2026

In shortProcedure guidance sets the clocks, and most of it is old and still current. What moved recently: FDA now publishes complete response letters, the next FDA user fee letter is proposed, FDA's eCTD v4.0 path opened, the EU's new pharmaceutical law is close to final, and rules on AI and on computerized systems are being rewritten.

Abbreviations in this section. PDUFA: the Prescription Drug User Fee Act, the US law under which companies pay fees for FDA review and FDA commits to review goals. NDA and BLA: a new drug application and a biologics license application, the two main US applications. CHMP: EMA's committee for human medicines. GMP: good manufacturing practice, which FDA writes as CGMP. GxP: the good-practice rules for regulated work such as manufacturing and trials.

Controls such as signatures with meaning, audit trails and validated systems support your compliance. No product can be certified as compliant with Part 11 or Annex 11; compliance belongs to the company and how it uses the system.

How an agency letter maps onto the dossier, and what a response package contains

In shortA letter is organized by review discipline, and each question points into one part of the common technical document. Good replies answer each sub-part directly, cite the dossier by module, section and page, and arrive as a package: a cover letter, a question-by-question response, attachments and an electronic sequence.

The five dossier modules every agency question points into

Structure per ICH M4.1 Your region's Module 1 sets the cover letter and form conventions.

Five habits of a reply a reviewer can use

What goes in an agency response package

Working through a data-cut difference and a ten-day clock, with the numbers

In shortTwo rates can differ because the denominator and the follow-up changed. Showing patients over patients, with the cut, tells the reviewer more than the percentage does. A due date works backward: subtract your internal target and the owner's review, and see how little time is left for the first draft.

Why 4.6% and 4.8% are both true

Treatment-emergent means the antibodies appeared or increased after treatment began. Two more patients with antibodies and twelve more patients evaluated explain the whole change. A reviewer can see that in a second if the reply says "31 of 642, Week 52 data cut, database lock March 14, 2026".

The lab test, or assay, matters too. Drug left in the blood can hide antibodies, so FDA's assay guidance treats drug tolerance, the amount of drug a test can cope with, as a sensitivity topic.37 That is why the example's question also asks for it.

A shelf life that changed for a good reason

The July number was right when it was written, because the stability study had not reached 24 months. The Swiss draft is the problem: someone reused July's answer without checking the date on it. The two cases need different treatment. An earlier value that newer data replaced is history. A draft that still carries it is an error to fix.

Counting back from a ten-day date

The Health Canada arithmetic works the same way, with its own counting rule: 15 calendar days, day of issue excluded, weekends and holidays included.7,11 A letter dated September 29 is due October 14.

Why agencies get told different numbers, and what they can see of each other's files

In shortFilings reach agencies at different times, so each dossier freezes data at a different point. Questions arrive in a different order, and different people answer them. Meanwhile regulators have formal ways to share information during review. A difference may be innocent, and the reply should still explain it.

A file sent to Switzerland 569 days after the first filing holds different data from the one sent to the US at the start, and questions on the two run on different calendars. Section 09 shows this overlap growing.

Four ways the same fact drifts apart

The channels regulators have for talking to one another

These are channels. We make no claim about how often reviewers compare any company's answers. Plan as though they might.

Twelve ways an agency reply goes wrong, from an unanswered sub-part to an untracked commitment

In shortReplies fail in four places: in how the question was read, in where the number came from, in what was said elsewhere, and on the way out of the building. Most of these errors are invisible on the page, because the reply reads well. A reader catches a slip only by checking it against a source.

What a weak reply risks: new questions, a later goal date, a complete response or a credibility problem

In shortMost weak replies only cost time, through a new question or a later date. Some contribute to a second review cycle, and the agency's published letters show that quality and manufacturing issues lead those. A false statement is a different category of problem, and an unexplained difference can look like one until someone explains it.

What FDA's published response letters object to

Query replies matter here because an issue can first surface as a question during review. Discipline review letters exist to "convey early thoughts on possible deficiencies" so that the company can start preparing.4 A clear answer to a question can spare a second cycle. A major amendment can also help: the research group CIRS notes that amendments "helped sponsors fix deficiencies early" in 2022.8

Three levels of consequence, from a follow-up question to a refusal

Approval times by agency, and the share of the EU calendar spent answering questions

In shortPublished numbers cover the calendar and leave out the effort behind it. Median approval times range from 292 to 448 days across six agencies. Company response time in the EU has held between 117 and 136 days for five years. We found no public market size for query management software and no public count of queries.

These counts set the scale of the work, since an approval usually follows at least one round of questions.

How many questions an agency asks

Hard counts are rare. One study looked at Day 120 quality concerns across 52 EU applications discussed in 2007 and 2008, and also found 32 major objections from 13 applications.45 The concerns below come to about 38 per application, before any clinical question, though the figures are old and procedures have changed. Our fictional EU list of 86 questions is a sample figure, of a plausible order.

Where agency review and query work are heading

• More overlap between markets. Medicines approved by all six major authorities rose from 37 in 2016 to 2020 to 52 in 2021 to 2025.44
• More sharing. Work-sharing and concurrent review are in regular use: at Swissmedic in 2025, total throughput for Project Orbis applications was 321 days, against 413 overall.9
• A shorter EU timetable. On current drafting, the new law cuts EMA's standard assessment to 180 days.25
• AI on both sides of the table. FDA's AI assistant, Elsa, helps staff read, write and summarize.46 Vendors now describe AI drafting of agency replies.2

AI drafting of agency replies: what FDA, EMA and EU GMP say, and what stays with people

In shortSoftware can split a letter, find passages, write a first draft and compare numbers across agencies. FDA's draft AI guidance leaves out submission drafting unless it affects patient safety, drug quality or study reliability. No rule removes your responsibility for what you send. The strongest control is simple: every number traces to a page, and a named person signs.

What FDA, EMA, the EU and industry guidance say about AI in reply drafting

How you would know a tool catches a difference: plant some and count

Two measures matter. Traceability asks whether each number in a draft is the number on its cited page; a random sample measures that. Catching asks whether the tool flags a difference from what another agency was told; such differences are rare in any one letter, so they have to be planted. A small test proves less than it seems to.

We found no published independent test of any query drafting tool's accuracy, as of November 2026.

Judgment that a regulatory lead keeps

• Whether to answer narrowly or to offer context the agency did not ask for.
• Whether a difference is a data-cut note, a correction or a matter to raise with the agency directly.
• Whether a response is substantial enough to discuss as a major amendment first.16
• Data that do not exist yet: a plant's hold-time study still has to be run.

The company stays responsible for what it files, whatever drafted the words.

How regulatory teams answer agency letters today: tracker, shared drive, Word and memory

In shortA regulatory lead splits the letter into a tracker and emails owners. Owners search submitted sequences and a shared drive of past answers, then write in Word. Reviews go round, the lead signs, and regulatory operations publishes. By our estimate, finding sources and writing first drafts take about half the effort. The check against other agencies depends on who remembers.

From the regulatory lead to the plant: who touches a reply

Five hand-offs between an agency's letter and its reply

Spreadsheet tracker, outsourced writing or regulatory software

Veltrimab's agency letters: twelve cited drafts, two differences caught, and a signed answer

In shortThis is the example from the top of the page, inside the product. Dr. Maya Chen, the regulatory lead at Northwind Bio, opens the letter and starts the agents with one click. They split it, assign seven owners, find sources, write twelve drafts and compare every number with what other agencies were told. Two differences surface before any person has written a word.

1

A new FDA letter, already compared with the past

On Tuesday, October 6, 2026, nine agency letters sit in Maya's query inbox, each with its clock. The FDA's information request on BLA 761904 arrived at 8:12 a.m.: 12 questions, due Friday, October 16.

Before anyone splits it, the panel on the right lists what the letter touches and what agencies have already been told: nine facts, with two marked as told differently. The antibody rate reads 4.6% for Health Canada, and the shelf life differs in a Swiss draft.

2

One click runs the split, the search and the drafts

Maya clicks "Split and draft replies". The question splitter finds 12 numbered questions under 7 review headings and keeps the agency's wording. The owner router assigns 7 owners from the approval rules. The dossier retriever finds 16 passages in Modules 2, 3 and 5, and 7 related past answers. The response drafter writes 12 drafts with 24 citations. Then the consistency checker compares every number.

3

Each question arrives with its reply, owner and source

The header counts seven owners, none of the 12 signed, 1 waiting on data from a plant and 2 told differently elsewhere. A second reply, question 12, stays in drafting until its statistics tables arrive on October 13.

Question 1 asks the company to justify its limit for host cell protein, material left over from the cells that make the drug: no more than 100 parts per million (NMT 100 ppm). Its draft cites 14 clinical and 3 process qualification batches at 8 to 41 ppm, and it carries a green chip: the same as the answer to the EMA on July 30.

4

Question 7: Health Canada was told 4.6%

The draft says 4.8%, 31 of 642, Week 52 cut, the same as the EMA reply of July 30. Health Canada was told 4.6%, 29 of 630, on August 21, from the Week 40 interim cut. The checker flags it in amber and explains it: "Both are right for their data cut — the reply should say which cut it uses."

It proposes a fix, and it does not change the reply on its own.

5

A one-click note explains both numbers

Maya clicks "Add a data-cut note". The reply now ends: "All figures use the Week 52 data cut (database lock March 14, 2026); our response to Health Canada of August 21, 2026 used the Week 40 interim cut (4.6%, 29 of 630 patients)." The flag turns to Explained, and the owner, Dr. Priya Ranvelle, reviews next.

6

The owner reviews, then the lead signs and says what the signature means

The FDA replies are still with their owners, so the story moves to a second letter that morning. Health Canada's clarification request on Veltrimab is 8 days from its due date, and its question 3 asks for the Week 52 update. Dr. Ranvelle has reviewed the answer. Maya signs with her name, her password and the meaning of her signature: "I approve this answer for submission to the agency."

7

The answer says what it updates

Signed at 8:17 a.m., and the letter moves to 4 of 5 signed. A banner states plainly that this reply updates Health Canada from 4.6% (Week 40 interim cut, sent August 21) to 4.8% (Week 52), as the agency asked. Regulatory operations publishes once the package is signed.

The solution's design targets are 4 hours from an agency letter to cited first drafts, 9 days left for expert review on a 10-day clock, and no conflicting answers sent. They are targets, to be measured on your own letters. Agency clocks vary, and 10 days here is an example.

Nineteen more screens, from the fact-by-agency grid to who owns each topic

In shortOne letter is the story. These screens show the rest: the second difference and its fix, the grid of every fact told to every agency, a reply with its sources at the passage, a reply waiting on data, the tracker and response package, the library, the dashboard and your settings.

Select any screen to open it larger. The companies, people, letters and figures are fictional. The dashboard's median of 3.9 hours to a first draft and its 27 differences caught are sample figures, shown to explain the screen.

Where 412 sample answers came from

Six agents split, route, find, draft, check and assemble; a topic owner and the regulatory lead decide

In shortEach agent has one job and limits it does not cross. None of them signs, resolves a difference or sends anything. The topic owner reviews each reply, and the regulatory lead signs each answer and the package. Regulatory operations sends the sequence. Every step goes on the audit trail with who, what and when.

From question splitter to package assembler: the six reply agents

The owner reviews each reply; the regulatory lead signs

Each signature records name, date, time and meaning with that version of the answer, and can only be withdrawn on the record. These are controls that support your compliance with 21 CFR Part 11 and EU Annex 11; no product is certified against them. Botminds' own credentials are SOC 2 Type II audited and information security managed to ISO 27001.

The regulatory lead, CMC author, immunogenicity owner and labeling lead in the screens

Agency letters and submitted sequences in; cited replies, consistency flags and a signed response package out

In shortThe solution reads the letters you already receive, the dossier you already filed and the answers you already sent, and leaves behind a cited reply for each question, every difference flagged with its fix, and a package built from signed answers. Your regulatory information system and publishing tools stay where they are.

Systems a query rollout connects to

For each, check whether it has an API, a file feed, or only screens, and who owns it.

Your dossier, regulatory system, publishing tool and owners stay

Cited, compared and signed by a named person

What a hand-built reply costs you in expert days, review time and repeat questions

In shortStart from your own costs. No regulator or study publishes hours per question, cost per reply, the share of replies that run late, how often replies conflict or how many reviewer questions those conflicts cause, so the numbers that matter are yours. A few published figures frame them: how much of the EU calendar goes to answering, how far apart filings land, and what a day of delay can be worth.

Timing five letters to find your own HAQ baseline

Reviewer trust, lost review time and other query costs no calculator counts

What faster first drafts return when you count expert hours against the tool's yearly cost

In shortPut in your own numbers. The calculator asks how many letters and questions you handle, the hours a first draft takes now and after, a loaded hourly cost, and what the tool and your testing cost a year. Every default is an assumption to replace, and time saved only counts if you decide what it is for.

Searching, drafting and checking: where query work saves

Letters a year, questions per letter and your hourly cost

A portfolio company answering a few hundred letters a year clears all three. A company with one product under review in one market probably does not, on hours alone, and should say so before buying anything. For it, the argument, if there is one, is the value of a clean file.

How ready your regulatory team is: letter volume, past answers, dossier access and sign-off

In shortPick the answer closest to your team today. Eight questions score 0 to 2; the last is for reflection. The meter adds up your score as you go, and each question says why it matters and what a strong answer sounds like.

What your readiness score means: a low one starts with gathering past answers

What to ask HAQ software or an outsourced writer about citations, past answers and cross-agency checks

In shortAsk every vendor the same questions, us included. The answers show how a reply is built: what it reads, how each number traces back, how a difference gets found, who decides, and how you would measure all of it on your own letters.

With no published independent test to lean on (section 10), a planted-difference test on your own letters is the fairest comparison, and the same test measures your current process.

Test query drafting on your own letters with planted differences, then decide on three numbers

In shortMeasure first, then test on your own closed letters with differences you planted, then run alongside a live one. Decide with three numbers: catch rate by type, false alarms on clean files, and expert hours to a signed reply. None of these steps commits you to a vendor.

1. Audit one product: pick five facts and list what every agency was told. You will learn something from it whichever way you go.
2. Time five recent letters from your tracker: days to first draft, hours per question, days left for review.
3. Gather 15 closed letters with their sequences and final answers, and an approval path to share them. Copy five and plant known differences in earlier answers.
4. Run alongside live letters for four weeks, with your owners reviewing and your lead signing as usual.
5. Decide on those three numbers, set against the baseline from step 2.

Catch rate, false alarms and citation accuracy per planted difference

HAQ questions answered: FDA information requests, EMA Day 120, Health Canada clarifaxes and AI drafting

In shortThe questions regulatory teams and buyers search for most, each answered on its own.