The 510(k) eSTAR, explained.

What a 510(k) eSTAR is: an argument that your device matches one already on the market, filed as an interactive PDF

One infusion set, from design freeze to FDA's decision

1. The design freezesVolterby locks the VLT-Flow 80 design on July 28, 2026, and test reports keep finishing through September.
2. Someone chooses the predicateA person picks one legally marketed infusion set with the same intended use: the Halvern SoftPort 90.
3. The eSTAR is filled and checkedSections drafted from the design file, reports and labeling, then read for facts that disagree. This is where the 2007 citation is caught, or sent.
4. A person signs, and a person sendsThe regulatory director signs that the file is truthful and accurate. The specialist sends it through FDA's portal on October 20.
5. FDA screens, reviews and decidesScreening by November 4. A decision goal of January 18, 2027, if the clock runs clean.

Step 3 is the subject of this guide. The people, company, devices and report numbers in this example are fictional.

The status message, the clock and two more eSTAR surprises

Predicate, substantial equivalence, technical screening, Additional Information

Fourteen terms are defined in the one-page summary.

How FDA decides substantial equivalence, and how Traditional, Special and Abbreviated 510(k)s differ

In shortFDA works through a fixed sequence of questions: is there a valid predicate, is the intended use the same, is the technology the same, and if not, do the differences raise different questions. Almost all files that reach a decision clear it: 95.3% of FY 2024 decisions found substantial equivalence.8 The route you file under changes how much evidence you attach and how long FDA intends to take.

The rule itself is short. FDA finds a device substantially equivalent when it has the same intended use as the predicate and either the same technological characteristics, or different ones that do not raise different questions of safety and effectiveness.6 Technological characteristics means materials, design, energy source and other features.1 A different question, in FDA's words, is one raised by the new technology that did not apply to the predicate and poses a significant concern.5

Which 510(k) route fits a new infusion set, a design change or a standards-based file

The eSTAR rules as of November 2026: version 7.1, the QMSR, new human factors content and the FY 2027 fee

In shortThe regulation behind a 510(k) is old and stable. The practice around it moved repeatedly in 2026: a new quality rule in February, a new template version in June, new human factors expectations in August, a new fee in October. A file built from last year's habits will be wrong in small, findable ways.

Four terms in this section. 21 CFR means Title 21 of the Code of Federal Regulations, where FDA's rules are published. QMSR is the Quality Management System Regulation, the quality rule. Human factors information is evidence that real users can handle the device safely. MDUFA is the Medical Device User Fee Amendments, the program behind FDA's device fees and review goals.

A funding lapse is a calendar risk worth a line in the plan. One law firm, summarizing FDA's plan for a fiscal 2026 lapse, said FDA could not accept new fee-funded submissions, 510(k)s included, until funding returned.23

The eSTAR template section by section, and which of your own documents answers each one

In shortFDA's table of the template has 22 rows. Almost every row is answered by a document that already exists, owned by someone outside regulatory affairs. So filling in an eSTAR is mostly a sourcing job: which revision, which report, which label.

The template follows the memo FDA reviewers use, known as the SMART memo, which is why its order feels like a reviewer's worksheet.1 FDA says the Indications for Use form (Form FDA 3881), the cover sheet (Form FDA 3514) and a declaration of conformity need no separate attachment, because they are built in. Neither does a 510(k) Summary if you let eSTAR create it.2

Grouped from FDA's Table 1.1 The right column is our mapping for a typical device program. Which rows you see depends on your answers, because the template's logic opens and closes questions. EMC is electromagnetic compatibility.

Why five desks feed one eSTAR

Two documents people confuse: the 510(k) Summary and the 510(k) statement

21 CFR 807.87(h) requires one or the other.18 A 510(k) Summary is a public document of the safety and effectiveness information, with a comparison to the predicate, published after clearance.24 A 510(k) statement is a signed promise to supply that information within 30 days of anyone's request.25 The summary you publish is what the next manufacturer reads when it looks for a predicate. That summary is also your own source when you file against your older device.

FDA technical screening: what the 15 days check, what a hold means, and how often files fail

In shortIn about 15 days, FDA checks whether your answers describe your device and whether each applicable question has a relevant attachment. Judging your evidence comes later, in what FDA calls substantive review. In FY 2025, 7.2% of 510(k)s failed or were not accepted on the first try, which FDA calls the first cycle, and the preliminary FY 2026 rate is 8.67%.8

FDA's definition gives two examples: an answer that says there are no tissue-contacting components when there are, and a software question with no Software Description attached when software applies.1 FDA's program page adds that an inaccurate answer to "any question", or attachments where none is relevant to the question, can bring a hold of up to 180 days.2 FDA's cybersecurity page repeats the rule for its own section.26 A virus scan runs too, and screening only begins once the user fee, FDA's charge for reviewing the file, is paid.1

The first-cycle failure rate fell by three-quarters when eSTAR arrived, then started to climb

You can check the FY 2025 rate yourself. FDA's table shows 3,584 passed or accepted, 40 with no action by day 15 and 281 failed: 281 ÷ (3,584 + 40 + 281) = 7.20%.8 The 40 count as passes: FDA's definitions say a decision is recorded automatically at the end of day 15 if no other decision has been made, and the file is "deemed" to have passed.8

Most failures recover. Of the 281, 243 were accepted in a later cycle by June 30, 2026, and 38 were closed or not yet accepted.8 Those counts say little about the cost in days, because no public figure gives the average length of a hold.

One file with a November 4 screening date, passing or failing

FDA's documents name inaccurate answers and irrelevant attachments as the triggers, and they do not list "N/A" as its own reason. Our reading: an "N/A" on a question that applies is an inaccurate answer, so give the reason each time you use it.

How to choose a predicate device and build the substantial equivalence comparison table, row by row

In shortStart from intended use, then pick the closest technology among the devices that survive. Name one if you can. Then compare row by row and answer each difference with a named report, an acceptance criterion and a result. FDA requires the comparison and recommends a table.

21 CFR 807.87(f) asks for a statement of how the device is similar to and different from comparable products, "accompanied by data to support the statement".18 The eSTAR asks for the comparison and a discussion of why differences do not affect safety and effectiveness.1 FDA "highly recommends" a tabular format.5 Differences are ordinary: FDA says devices in the program commonly differ from their predicates, and that it rarely finds a device not equivalent at the technology question.5

Finding candidates in FDA's databases

1. Write your own intended use first, in the words you will file. FDA determines intended use from the proposed labeling.5
2. Find the product code. FDA's classification database turns a device type into a code, a regulation and a class. For insulin infusion sets, real clearances sit under code FPA and 21 CFR 880.5440.27
3. List the clearances under that code and read each candidate's public summary, the document you will compare against.
4. Check the history. 21 CFR 807.100 rules out a predicate removed from the market at FDA's initiative or found misbranded or adulterated by a court.6
5. Write down why you chose it, because the testing plan, the table and the labeling all follow from this choice.

What the recall research found about predicate choices

The study rules nothing out, and FDA's draft on predicate selection remains a draft.20 Treat the finding as a reason to write down your reasoning when a candidate has an open recall touching the feature you rely on.

Seventeen rows, six differences: a worked comparison for an infusion set

A row closes with four things: a named report, a criterion set before the test, a result stated as measured, and a sentence tying them to the difference. A reassuring paragraph with no report invites the request for data it implied. FEP and PTFE are two fluoropolymer plastics; sensitization testing looks for allergic reactions.

FDA recognized consensus standards: editions, partial recognition, transition periods and the citation that goes stale

In shortFDA recognizes specific editions of standards, usually publishes changes in lists at least twice a year, and sets an end date for each outgoing edition. Citing a standard is therefore a fact that can expire. The first thing to ask of any wrong edition is whether the testing is out of date, or only the sentence.

A consensus standard is a published method, written by a body such as the International Organization for Standardization (ISO), that an industry has agreed to use. When FDA recognizes one, you can file a declaration of conformity, a signed statement that your device meets that edition. FDA's database has one entry per edition, with a recognition number, the extent of recognition and, when a successor arrives, a transition statement. Three real entries show how this reads.

Old testing or an old sentence: two problems that look identical in a draft

Our framing. The rule behind it: after a transition ends, statements of conformance to expired versions "will generally not be accepted", and superseded standards withdrawn from the list "cannot subsequently be used" in declarations.4

Why eSTAR answers contradict each other, and a shelf-life claim worked through with numbers

In shortThe template asks for the same facts in several places, in different words, because each question has a reviewer behind it. Fix those facts once, early, and reuse them. Shelf life shows why: the claim, the label and the aging data should give one number, and often give two.

From 36 months to 24: the arithmetic behind an accelerated aging claim

Shelf life is usually claimed from accelerated aging, where packaged devices sit in a warm chamber to stand in for months on a shelf. The common formula is the accelerated aging factor, AAF = Q10^{(T_aging − T_ambient) ÷ 10}, where Q10 says how many times faster aging runs for every 10 °C of added heat.

With a chamber at 55 °C (the temperature in the fictional report), an assumed room temperature of 25 °C and an assumed Q10 of 2, the factor is 2³ = 8. To stand in for 24 months, or about 730 days, the samples need 730 ÷ 8 ≈ 91 days in the chamber. A 36-month claim would need about 137 days.

Illustrative inputs. Use the values in your own aging protocol and your own standard's method; many teams also run real-time aging alongside.

In the fictional file, the 36-month figure came from the predicate's summary and the report only supported 24. The team had a choice: claim 24 now or wait for data. Choosing 24 meant changing four places together: the description, the labels, the Shelf Life row and the 510(k) Summary. A change made in three of the four creates a new contradiction.

How long a 510(k) takes: FDA days, applicant days and the total time to decision

In shortFDA's goal of 90 FDA days counts only days at FDA. Files received in FY 2024, which FDA calls the FY 2024 cohort, averaged about 75 FDA days, 69 applicant days and 144 total.8 From the 20th to the 80th percentile of files, FDA days run from 57 to 90 while applicant days run from 0 to 157.8 So the part of the calendar that differs between a fast file and a slow one is mostly on your side.

FDA days are the calendar days a submission is under review at the agency.3 When FDA sends an Additional Information request, the review clock stops, and starts again when your complete response arrives.7 Those waiting days count as applicant days. The Government Accountability Office (GAO) described the gap in 2012: FDA review time "excludes this waiting time", and that review time alone decides whether FDA met its goals.31

Where slow 510(k)s lose their time

Read the 80th percentile. One file in five took more than 243 total days from receipt to decision in the FY 2024 cohort. At the same percentile, applicant days (157) run well above FDA days (90), each counted in its own ranking.8 Files averaged 1.69 review cycles, each one a round of FDA review.8 At the other end, one file in five reached a decision within 58 total days.

The shared goal FDA and industry have both missed

FDA and industry also share a second goal: the average total time to decision, FDA days and applicant days together.

The FY 2025 goal was 112 days, with FY 2026 and FY 2027 at 112 or 108 depending on a performance adjustment.3 FDA's report explains the FY 2024 miss in one line: FDA met its own goals, but "submissions placed on hold increased the total review time".3

The draft MDUFA VI letter restarts at 128 days in FY 2028 and steps down to 112 by FY 2032, keeping the 90-day goal.16,17 FDA discussed the draft at a public meeting on August 5, 2026.33 It is still a draft, and Congress must act by September 30, 2027.17

No goal covers the weeks before you send the file. No regulator publishes them, and they are the one block of the calendar nobody else can shorten.

Why FDA sends Additional Information requests, and the common 510(k) errors behind holds and second cycles

In shortTwo in three first reviews draw a request. FDA does not publish a breakdown of reasons, so anyone quoting exact shares is using a private sample. Most errors teams can find before sending fall into a handful of patterns, and most of them come from reusing an old file or from documents moving at different speeds.

The rate comes from FDA's own table: for a 510(k) at the substantive interaction stage, the counted action is an Additional Information request.3 Both years are preliminary. Earlier cohorts ran in the same band: FDA's own chart shows 63% to 68% on the first review cycle for FY 2018 through FY 2022.34 GAO saw the same pattern over a longer period: FDA asked for more information on a majority of more than 62,000 device submissions from 2001 through 2016.35

Eight eSTAR errors that come from reused files and moving documents

The patterns are our analysis of FDA's definitions and guidance. FDA publishes no ranking of them.

How the FY 2024 cohort ended: cleared, not equivalent, withdrawn or deleted

Withdrawn plus deleted comes to 425 of 3,462, or 12.3%.8 Those files ended with no finding on equivalence either way. FDA's rule for requests: a complete response must arrive within 180 calendar days of the request, or FDA considers the file withdrawn.7

What a 510(k) costs in user fees, how many FDA reviews, and what nobody publishes

In shortThe FY 2027 fee is $28,653, or $7,163 for a small business. FDA receives about 4,000 510(k)s a year and clears most of them. The bigger costs are unpublished: the people who prepare the file, the second cycle and the launch that waits.

Small business status needs total gross receipts or sales, affiliates included, of no more than $100 million in the most recent tax year. It expires on September 30 of the fiscal year it is granted, and FDA's review aims to finish within 60 calendar days. If you file before FDA approves the request, you pay the full fee with no refund of the difference.36 A file planned for October needs a request lodged with time to spare.

How many 510(k)s FDA reviews each year

Can AI prepare a 510(k)? What FDA has said, who still signs and which records Part 11 reaches

In shortNo FDA regulation forbids drafting a 510(k) with software, and FDA has not issued guidance on it for devices. The firm limits come from older rules: a named person states the file is truthful and accurate, a person's electronic signature must show a name, a time and a meaning, and each claim must rest on evidence.

21 CFR 807.87(l) requires a statement that the submitter believes, "to the best of his or her knowledge", that all data and information are truthful and accurate and that no material fact has been omitted.18 That belief is a person's. The practical test of any drafting tool is whether the person signing can still check what it wrote, so a citation that opens at the page matters more than fluent prose.

Which 510(k) jobs software can carry and which stay with a person

Measure a tool on a file whose answers you already know. Take a past cleared submission, plant known defects (a changed model number, a withdrawn edition, a missing report) and count what it finds, where its evidence points, and what it raises that you did not plant. Part C sets this out as a pilot.

How device regulatory teams assemble a 510(k) today: who does what, and where ten weeks go

In shortOne regulatory specialist assembles what four other people own, by hand, then reads it twice. By this guide's planning estimate, a Traditional 510(k) takes about 50 working days from the start of drafting to a ready file. Nobody publishes a measured figure, so time your own.

Specialist, design engineer, test lab lead, labeling owner, director

From predicate decision to the CDRH Portal, step by step

Where the working days on a Traditional 510(k) go

In house, with a consultant, or with submission software

Most teams mix these. Third-party review, where an FDA-accredited outside reviewer handles the file, changes who reads it and leaves preparation to you; FDA counts those files separately.8

The VLT-Flow 80 eSTAR: 22 sections drafted, 31 checks run, one citation fixed and a person's signature

In shortThis is the example from the top of the page, inside the product. Five AI agents, each with one job, draft sections, compare the device with its predicate, match standards and check the file. Nadia Brenholt, the regulatory specialist, works down what is left and applies the fix. Marcus Feld, the director, approves and signs. Nadia sends the file herself.

1

The whole file and what is left, on one screen

It is the morning of October 13, 2026. Nadia owns the VLT-Flow 80 eSTAR: a Traditional 510(k) for six models, product code FPA, 21 CFR 880.5440, Class II, against the Halvern SoftPort 90. All 22 working sections are drafted, with 46 attachments and a target of October 20.

The preflight is the product's run of 31 checks before a file goes to FDA. It last ran on October 12, and the draft has had 14 edits since, so all 31 checks sit gray. Nothing counts until it has been checked on the current draft.

2

The device beside its predicate, each difference with its report

A person chose the predicate. The predicate comparer then read its public summary and the design file and laid out 17 rows: 11 the same, 6 different. Each difference names what answers it. For the FEP cannula against the predicate's PTFE, that is biocompatibility report TR-0412 and kink report TR-0418, and the sentence for the equivalence section is already drafted.

One row did not hold up. The draft claimed 36 months of shelf life, like the predicate, and report TR-0433 supports 24. On October 9 Nadia set the claim to 24 months, and the drafter changed it in four places.

3

Thirty lines turn green and line 22 stays red

Nadia presses Run preflight. The agents read 22 sections and 46 attachments, cross-check the facts that repeat across 31 documents, tie 6 differences to their reports and match 14 standards to FDA's list.

Thirty lines pass. Line 22 does not: Section 11 and Section 4 cite ISO 11135-1:2007, a withdrawn edition, where FDA recognizes ISO 11135:2014 + Amd 1:2018. The panel shows where, shows the evidence and offers a fix. The light reads "Not ready yet".

4

The lab tested to the right edition all along

One click opens TR-0430, the ethylene oxide validation from Lakeview Sterilization Services, dated July 21, 2026, at page 3. It says the validation was performed to ISO 11135:2014 + A1:2018, using the overkill half-cycle approach, a common method that builds in a safety margin.

So no test has to be repeated. The wrong edition came from a 2019 file for an older model, reused as a starting point. The test report stays exactly as the lab signed it.

5

Swapped in two places, approved, signed, and sent by a person

Nadia presses "Swap the citation in 2 places". The product checks line 22 again and records her name against the change. With 31 of 31 passed, she sends the five open sections to Marcus with a note.

Marcus approves them and signs the truthful and accurate statement. The eSTAR is ready to download, and the record says Nadia submits it through the CDRH Portal. No agent can sign it or send it.

The solution's design target is about three weeks to a ready eSTAR against an estimated ten by hand, and no technical screening holds. These are targets, to be measured on your own devices. The sample's preflight is the product's own list of 31 checks: FDA does not publish a checklist for screening, and a withdrawn edition goes beyond FDA's written definition of it (section 05). Passing every line cannot promise that FDA's screen or review will pass.

Seventeen more screens, from the old citation in Section 11 to the checks and approvers your team sets

In shortOne submission is the story. These screens show the rest: the citation as it stood before the fix, the standards list, reasons for each not-applicable answer, the reports, the plan to FDA's Day 90, the record of past submissions and the settings.

Five agents draft, compare, match and check; a regulatory director approves and signs; a person sends

In shortEach agent has one job and limits it cannot cross. None can choose a predicate, change a test report, sign a statement or send a file to FDA. A named director approves every section, the lab lead co-approves the test sections, and every step lands on an audit trail.

Section drafter to cover letter writer: what each 510(k) agent does

Who approves each eSTAR section, who signs the statement and who submits

Signatures show the name, time and meaning, the elements 21 CFR 11.50 requires of a signed electronic record.38 The product provides controls that support your Part 11 compliance; validating and using them for your purpose stays with you. Botminds is SOC 2 Type II audited, and its information security is managed to ISO 27001. A settings change, like an edit or a fix, goes on the audit trail with who, what and when, and earlier text stays readable.

Nadia, Marcus, Aisha and Tomasz in the screens

Design file, test reports and labeling in; a filled eSTAR with every claim sourced and every check evidenced out

In shortThe solution reads the records you already keep and works beside them. It leaves behind sections with citations, the comparison with its reports, a standards check, a preflight with evidence and a complete audit trail. Your quality system stays the system of record.

Records and systems a 510(k) rollout connects to

The design targets for a VLT-Flow sized submission

What a technical screening hold, an Additional Information cycle and a late launch cost you

In shortFDA's fee is the one number you can budget to the dollar. The larger costs are people's weeks and whatever the file holds up, and nobody publishes them. FDA's own counts frame the risk: how often files fail the first screen, how often they draw a question, and how many days applicants spend answering.

Timing your last ten 510(k)s to find your own baseline

Launch windows, your signer's exposure and other 510(k) costs no calculator shows

What faster, cleaner 510(k)s return when you count specialist days and avoided holds against the tool's yearly cost

In shortPut in your own numbers. The calculator counts two things: the specialist days you get back and the expected delay you avoid. Only the hold rate comes from a published source, and every other default is an assumption to replace.

Drafting, rereading and second cycles: where a 510(k) team saves

510(k)s a year, specialist days and the value of a month

A team filing two 510(k)s a year will rarely pay for a tool on time alone, and should say so before buying anything. The case for that team rests on the value of a month of delay, which only your product owner can give you.

How ready your regulatory team is to automate 510(k)s: filings, records, reused files 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.

A low score means fixing your record-keeping before your first 510(k) pilot

What to ask 510(k) submission software or an AI drafting tool about citations, predicates, standards and signatures

In shortAsk every vendor the same questions, us included. The first three test what the software may never do, and the last two ask for a test on your own file.

Asking the same questions of your current process is the fairest comparison, and the same test measures it.

Test 510(k) software on a cleared file with planted errors, run one live file in parallel, then decide on three numbers

In shortMeasure first, then test on a file whose answers you know, then run one live file alongside your normal process, then decide. None of these steps commits you to a vendor.

1. Pull dates from your last ten files. That is your baseline: first draft to file sent, holds, requests, response days.
2. Read last year's FDA requests. Mark each that your own file could have answered.
3. Pick a cleared file, ideally one that drew a request. Copy it and plant known errors: a changed model number, a withdrawn edition, a report removed from a difference, an attachment under the wrong question.
4. Run one live file in parallel with your people approving every section as usual.
5. Decide on three numbers: planted errors caught with the right evidence, false alarms, and specialist days to a ready file.

Catch rate, false alarms and citation accuracy on a 510(k) with planted errors

510(k) eSTAR questions answered: screening holds, fees, review times, predicates, standards and AI

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