AMCP dossier development, explained.

What an AMCP dossier is, told through one medicine and one health plan

How a payer question becomes a formulary decision

1. The medicine changesVeltrimab is approved for ulcerative colitis on September 30. New studies and a new list price follow.
2. The dossier falls behindOne label revision and three other changes touch 14 of 17 sections. A two-page abstract then puts three of them out of date.
3. The team updates it and answers requestsSections are rewritten and signed, and Crestline's four questions are answered from them. This is where stale text is caught, or carried forward.
4. Reviewers approve the packetMedical, legal and regulatory review approve it before the October 29 deadline the company set for itself.
5. The committee decidesCrestline's P&T meets on November 12. The plan's pharmacist uses what she finds useful and may ask again next year.

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

Why payers doubt dossiers, and three more surprises

What P&T, HCEI, unsolicited request and PMPM mean

Fourteen terms are defined in the one-page summary.

Three kinds of AMCP dossier and the six parts of an Approved Product Dossier

In shortAMCP defines three kinds of dossier, set by whether the FDA has approved the product or use. The one most teams spend their years on is the Approved Product Dossier, which has six numbered parts carrying a "B" suffix: executive summary, product and disease, clinical evidence, the economic report, other evidence, and appendices. AMCP recommends a length for each.

Which dossier applies: unapproved product, approved product or unapproved use

The longest allowance goes to the economic report

AMCP warns against padding: makers "should not include overly verbose or superfluous content to meet page recommendations".1 The evidence tables in 3.2B run to less than a page per study, two at most, and a payer's pharmacist often lifts them straight into her own review. AMCP asks for electronic delivery for that reason, with economic models as a Microsoft Excel workbook or another format both sides accept.1

The six parts of an Approved Product Dossier and what moves each one

The rules behind payer communications, with Format 6.0 and FDA's June 2026 draft

In shortThe dossier's layout comes from AMCP, a society. What a company may say comes from the Federal Food, Drug, and Cosmetic Act and FDA guidance. Two things moved in 2026: FDA published a revised draft of its payor guidance (FDA spells it "payor") on June 3, and AMCP released Format 6.0 in October. FDA's 2018 guidance remains the final text until the draft is finalized.

What section 502(a) and section 502(gg) each allow

Two paths in the Act matter. Under section 502(a), health care economic information given to a payer, formulary committee or similar expert body is not treated as false or misleading if it relates to an approved indication, rests on "competent and reliable scientific evidence", and carries a prominent statement of any material differences from the label.4,10 HCEI is defined as "any analysis" that "identifies, measures, or describes the economic consequences" of using a medicine, along with its underlying data, inputs and methods.4

Under section 502(gg), added by the Pre-Approval Information Exchange (PIE) Act, truthful and not misleading information about an unapproved product or use may go to the same audience with required disclosures: that it is not approved, its stage of development, the material aspects and limitations of each study's design, methods and results, and, where applicable, the approved indications and current labeling.8,10 A slide set of this information is often called a PIE deck.

The 2018 guidance suggested follow-up when earlier information went out of date. The June 2026 draft, now that the statute requires it, says firms "must provide updated information to payors if previously communicated information becomes materially outdated".8 Two law firm notes read that as covering events such as a missed primary endpoint or a clinical hold (FDA pausing a study), and one points out that the draft gives no timing, channel or format.10,11

Who may receive which material, by who speaks first

What Format 6.0 proposed in March 2026

AMCP's page says Format 6.0 "was released in October 2026 after being revised to include guidance for cell and gene therapies and biosimilars".3 We read the March request for comment; we could not read the final text in the journal, so treat these as what was proposed and check the published version.7

All quotations in this table are from AMCP's March 17, 2026 draft.7 The 5.0 horizon is from Format 5.0.1

What makes a payer's request unsolicited, and how a company answers it

In shortA request is unsolicited when the requester formed it without any prompt from the company. AMCP tells makers to follow FDA's 2011 draft when they send an Approved Product Dossier, and to mark the dossier as sent on an unsolicited request. The draft asks for a private reply that answers only the question asked, in a scientific tone. Keep the record of how the request arrived.

AMCP's wording is direct: the inquiry "must be initiated by the requester (formulated in their own mind) without prompting, suggestion, or solicitation by the manufacturer or its employees".1 FDA's draft treats a request as unsolicited when it comes from people "completely independent" of the firm, and says requests "prompted in any way by a manufacturer or its representatives are not unsolicited requests". A sales representative who mentions an off-label use and invites a follow-up is one of its examples.5

Neither text gives a waiting period or a form. A company may set its own test, such as "no contact from us in the 30 days before". That is a sensible control, and the 30 days is the company's own number.

Payers can send requests by letter or email, or through a portal. AMCP's eDossier system began as a pilot around 2010, built so that decision makers start every request.12 AMCP's Format mentions online platforms as one route for delivery.1

From a letter on file to a packet out the door

Four things FDA's 2011 draft asks of a reply

Standing requests: when a new edition may go to a past requester

AMCP's general answer on sending a new edition to everyone who asked last year is no: makers "should not freely and automatically send updated Approved Product Dossiers to previous requesters without an unsolicited request".1 Three situations differ.

Sorting payer questions: on label, health care economic information, or unapproved use

In shortFDA asks whether an economic analysis relates to the disease and patient population in the approved label. It may model a longer time frame or another care setting than the trials used. It cannot rest only on patients outside the indicated population. Each of Crestline's four asks shows a different answer.

FDA says an analysis relates to an approved indication if it concerns "the disease or condition, the manifestation of the disease or condition, or symptoms associated with the disease or condition in the patient population for which the drug is indicated".4

Its examples of analyses that can still qualify, though the label does not mention them, include modeling use beyond the length of the trials when the indication sets no limit, and care settings that differ from the trial settings. Its examples of analyses that do not qualify include studies "limited to patient populations that are not within the indicated patient population", such as an age group the label excludes.4

Crestline's four questions, sorted

The asks and the label text are fictional sample data. The classification follows FDA's categories.4

Which words change the classification

How one label change and one abstract put AMCP dossier sections out of date

In shortAMCP recommends updating a dossier when something significant changes and checking its technical accuracy once a year when nothing does. A dossier's sections depend on one another, so a new indication spreads widely. In this guide's example one approval touches 14 of 17 sections. The hard part is knowing, section by section, which sources each one rests on.

AMCP's triggers are specific: "changes to the prescribing information, line extensions, new safety information, or any information that materially affects the overall evidence". Examples include a filing for a new indication, whatever the decision, and new evidence on who benefits or on real-world effectiveness.

Without new evidence, dossiers "should be evaluated for technical accuracy on an annual basis (e.g., changes in price, new model assumptions)". If a review finds no revision is needed, that goes on the title page. Every dossier carries its original date of issue and the dates of revisions and reviews.1 The maker also has to help the reader see what changed, by highlighting, an appendix or a summary in a cover letter.1

What moved the example dossier from v5.2 to v6.0

The example dossier uses 17 of the sections the Format describes. A real one may use more or fewer. A section marked "no change" still needs a dated review; AMCP wants that shown.1 v5.2 and v6.0 are this dossier's own edition numbers, separate from AMCP's Format versions.

Why memory and a spreadsheet stop working

A spreadsheet of sections can record a date. It rarely records that the second sentence of 1.1B rests on Table 2 of a study whose follow-up just came out. One consultancy's guide names the usual causes of a poor dossier: evidence scattered across clinical reports, models and internal analyses, and no shared ownership across health economics and outcomes research (HEOR), medical affairs and market access.13 A team then finds a stale section at the annual review, or when a payer's pharmacist notices.

In a 2008 survey of 13 companies, 46% said dossiers were updated at least every six months and 31% yearly.14 That survey is old and small, and it is the only public figure we found.

Working through a budget impact model: from 1,000,000 members to $0.031 per member per month

In shortA budget impact model asks one question: can the plan afford this? It starts with the plan's members, narrows to the patients who could use the medicine, applies an uptake rate (the share who start the medicine), multiplies by the extra cost, and spreads the result across every member and month. The answer looks tiny because everyone shares it. Payers want to change the inputs themselves, so AMCP recommends handing over an unlocked model.

AMCP separates two questions. A cost-effectiveness model asks "Is the technology good value for the money?" A budget impact model asks "Is the technology affordable to the health system?"1

AMCP recommends the second take the payer's own perspective over one to five years and report both the PMPM difference and the overall budget impact. It should also let users enter their own population size, prevalence (how many members have the disease) and costs. "When possible, a stand-alone, electronic, unlocked, modifiable model should be provided".1 Format 6.0's March draft proposed a shorter usual horizon, one to three years.7

The Veltrimab year-one figure, one step at a time

Every input is fictional sample data from the example's model report. "Net" means after the therapies Veltrimab displaces. An advanced therapy is a biologic or another targeted medicine. Crestline's own membership changes the total and leaves PMPM alone until it also changes prevalence, mix or cost.

A second number payers ask for: number needed to treat

In the survey of 20 US payers reported in 2025, 70% found the number needed to treat (NNT) valuable.2 NNT is one divided by the difference in response rates.

In the example's induction study, which covers the first weeks of treatment, 22.8% reached clinical remission on Veltrimab and 7.1% on placebo at week 12: a difference of 15.7 points, so 1 ÷ 0.157, or about 6 to 7 patients treated for one extra remission. Among patients whose earlier advanced therapy had failed, 17.9% against 4.2% gives a difference of 13.7 points, about 7 to 8 patients. Every number here belongs to the fictional example.

Common AMCP dossier errors: stale text, weak models, figures that disagree with tables

In shortDossier errors tend to be errors of connection: a section left behind, a value read from a chart with no table to check it, a label revision cited after it was replaced, material that no ask called for. No one publishes error rates. The closest public evidence is an audit of economic analyses from twenty years ago, and its results were poor.

Eight dossier errors, scored by who they hurt

We found no published FDA warning or untitled letter specific to an AMCP dossier in this research. That was not an exhaustive search, and it is no reason to relax: section 502(a) covers false or misleading labeling, and FDA treats HCEI that meets the definition of labeling as labeling.8

What P&T reviewers read in an AMCP dossier, and what they skip

In shortSmall surveys point the same way across a decade: payers lean on the clinical evidence and doubt the maker's models. Dossiers matter most for rare diseases and costly products. Treat every percentage here as a hint. Where a sample size is given, it runs from 10 to about 60 respondents, and several of these surveys came from firms that sell evidence or dossier services.

Older surveys of dossier use, and why plans still ask

In 2019 interviews, ten senior payers described a "general decline in the utilization of the AMCP Dossier" and rated clinical trial data most useful and the economic section least, and all of them wanted additional content such as comparative effectiveness data.17 Six respondents in the survey reported in 2025 also asked for a stronger focus on comparative effectiveness.2

Payers do say economic evidence plays a part: in 2010 surveys, 74% of managed care respondents and 77% of P&T respondents said health economics and outcomes research was used in their decisions, and 68% of the managed care respondents said their organization had no quality standard for it.18

Medicare drug plans also work to a clock. The P&T committees of Part D sponsors, the insurers that run those plans, are expected to make a reasonable effort to review a new drug or new indication within 90 days of its release and decide within 180, or give a clinical justification.19

Before approval, payers come back again and again

A March 2025 survey of 18 payers shows what they prefer: half wanted pre-approval information up to six months before launch, a third wanted it 6 to 12 months before, and 56% preferred both a pre-approval dossier and a PIE deck. They most wanted efficacy data (17 of 18) and first pricing estimates (10 of 18).21

What a stale dossier or slow answer can cost at formulary time

In shortFormulary decisions are concentrated in a few buyers, and each of the three largest pharmacy benefit managers now keeps more than 600 products off its standard list. A dossier does not win a place alone. A late or outdated one can leave a committee working from older evidence, and that is hard to measure, which is why few companies do.

More new medicines competing for the same meetings

How a stale dossier section reaches a patient

What generative AI changes in AMCP dossier writing, and which rules reach it

In shortEarly published work reports large time savings on a first draft, from small single studies by a services firm. We found no FDA rule that speaks directly to AI-written payer dossiers. AMCP's 2026 draft allowed AI as a writing assistant with a qualified person validating. The rules that do reach the content are the older ones: it must be truthful and not misleading, and rest on competent and reliable evidence.

Which texts reach dossier drafting

What software can take, and what stays with people

How HEOR and market access teams build dossiers and answer payer requests today

In shortHEOR writes the economics, medical writers and medical information write the clinical parts, a dossier owner holds the whole, and reviewers approve. Consultancies often share the writing.14 Expect the hand-offs and rounds of review to take much of the calendar, and time your own, because nobody publishes how much. Three ways of working are common: in house, outsourced, and increasingly with software.

From HEOR lead to national account director: who touches a dossier

Six steps and five hand-offs in one unsolicited packet

Reviews and rework are where a hand-built packet spends its days

Public data is thin. In a 2008 survey of 13 companies, 61% shared dossiers between medical information and health economics, every respondent used consultants as well as in-house staff, 81% of outsourced dossiers needed revision, and 54% said a quarter to half of the content did, adding 25 to 50 hours. A third of respondents spent 25 to 50% of their time on dossiers, and 31% reported more than 100 dossier submissions a year for a single product.14

In house, consultancy, or software on your own evidence

The Veltrimab binder: three sections turn amber, then Crestline's packet comes from the same text

In shortThis is the example from the top of the page, inside the product. A new abstract enters the evidence library and three sections are marked out of date within seconds. One click drafts cited updates for the medical reviewer. The same binder then answers Crestline's four questions: three in the packet, one handed to Medical Information. Two reviewers approve before it can be sent.

1

Seventeen sections, each with its evidence and its date

Priya Ranvelle, the dossier owner, opens the binder for the v6.0 draft: 17 sections, 90.7 pages, sign-off target October 24. Each row has an owner, a page count, how many sources it rests on and a freshness track from v5.2 to today.

Four are current and signed, three were reviewed with no change, nine are being updated, and §4.2B waits for a cost-effectiveness report due October 20. The library feed on the right lists every source as it arrives.

2

Medical Affairs sends the week-100 abstract

At 9:20 a.m. the VLT-302 extension results arrive as a congress abstract: two pages, one table, one Kaplan–Meier figure (a curve of how many patients stay in remission over time). Priya presses "Add and check the binder". The solution's AI agents, each with one job, read it down to individual tables and figures and check every section that rests on VLT-302.

3

Three sections turn amber

§1.1B Clinical benefits, §3.1B Study summaries and §3.2B Evidence tables rest on VLT-302 and were last updated before today. Each row turns amber, the track shows a tick past the last update, and the feed marks the abstract "3 out of date". The library now holds 413 sources.

The other sections stay as they were. Nobody had to remember which sentence cited which study.

4

Three cited updates, sent for medical review

"Draft the 3 updates" adds the new data to each section: of patients in clinical remission at week 52, 78.4% were still in remission at week 100, as observed, n=208. Each value carries its citation. The references gain one entry, and all three sections go to Omar Haddad, the medical reviewer.

The words "as observed" stay in, because the source uses them.

5

Crestline's four asks, matched to the sections that answer them

The packet panel on the binder holds REQ-2417 from Dr. Hannah Brooks, PharmD, sent through the eDossier portal on October 6 for Crestline's November 12 meeting. A1 is the whole dossier. A2 is §4.3B and an Excel model preset to 1,400,000 members. A3 is the two subgroup rows of §3.2B as a 3-page excerpt.

A4, adolescents aged 12 to 17, is held out: an unapproved use under label §8.4.

6

The letter, the unsolicited check and the edition

The request opens beside the answers. Each ask is highlighted in the letter in the payer's own words. The unsolicited check records the channel, the requester's role on the P&T committee, and no company contact in the 30 days before, which is this company's own test.

The dossier owner chooses the edition. v6.0 is due for sign-off October 24, before the October 29 packet deadline, so v6.0 goes.

7

The budget model preset to the plan

A2 opens the model at Crestline's 1,400,000 members: 27 patients and about $522,000 in year 1, 80 patients and $1.46 million in year 3. PMPM holds at +$0.031 and +$0.087, as in the 1,000,000-member base case. Members, prevalence and uptake stay editable, and the Excel workbook goes with its inputs unlocked.

8

The packet: four files, nothing unrequested

Priya hands A4 to Dana Okafor in Medical Information, with the label passage and a reply date of October 9. Then the packet builds: a cover letter that answers A1 to A3 one by one and says A4 is answered separately, the v6.0 dossier, Crestline's model and the evidence excerpt.

The payer deck stays out. Crestline did not ask for slides. The dossier itself follows once v6.0 is signed on October 24, ahead of the deadline.

9

Two approvals, then the portal

Omar Haddad approves for medical review and Lena Ortiz for regulatory and legal, who keeps the unsolicited record. Only then can Priya send through the portal. The activity record shows every step, from "split REQ-2417 into 4 asks" to the send.

The redline, disputed values, the P&T calendar and the evidence library

In shortOne abstract and one request are the story. These screens show what sits around them: a section as a redline (every edit marked against the last edition) with a reason on every change, a value that differs between two places, a section waiting on a report, the calendar of P&T meetings, the library with every source read, the dashboard and your own rules.

Select any screen to open it larger. Every plan, person, medicine and figure is fictional.

Six agents read, map, draft and assemble; the dossier owner and MLR sign every section and packet

In shortEach agent has one job and limits it does not cross. None signs a section, sends a packet or answers an unapproved-use question. The dossier owner and section owners sign, and each signature states what it means. Medical review and regulatory and legal review both approve a packet. Every agent step and human decision is on the activity record.

From request intake to packet building: the six dossier agents

Who signs a section, who approves a packet

Signatures record the name, the meaning, the decision and the time, with controls that support your compliance with 21 CFR Part 11 where your records fall under it. Roles, maker-checker approval (one person prepares, another approves) and the full activity record are settings you own. Botminds is SOC 2 Type II audited and manages information security to ISO 27001.

The dossier owner, HEOR lead and reviewers in the screens

Labels, study reports, model reports and payer letters in; a current binder and an approved packet out

In shortThe solution reads the evidence a dossier already rests on and the requests payers already send, and leaves behind a binder of sections with a date and sources on each, cited redlines, tailored packets and decks drawn from the signed dossier. Your document, medical information and portal systems stay where they are.

Systems a dossier rollout connects to

A 2.4-day design target for one payer packet

What stale sections and slow payer packets cost your team

In shortStart from your own costs, because public studies do not give them. What exists is old or small, and it points at rework: outsourced dossiers coming back for revision, and staff spending a large share of their time on this one document.

Finding your own hours per packet and per refresh in a week

Formulary positions, reviewers' trust and a thin record

What faster packets and shorter refreshes return against the yearly cost of a tool

In shortPut in your own numbers. We found no published wage or benchmark we could use as a default, so every input here is an assumption to replace. The calculator counts staff time on packets and refreshes. Formulary outcomes are too uncertain to put in a formula, and we leave them out.

Packets, refreshes and review rounds: where dossier work saves time

Packets a year, refresh hours and your hourly cost

A company with three products on the market and a launch coming, answering requests all year, may reach 104 or even 208 packets a year. A company with one mature product and a few dozen requests a year probably will not, and should say so before buying anything.

How ready your dossier team is: volume, baselines, section maps, Format 6.0 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 a score of 0–6, 7–11 or 12–16 means for your dossier team

What to ask dossier software or an HEOR consultancy about citations, freshness and unsolicited records

In shortAsk every vendor the same questions, us included. The answers show how the work gets done: what it reads, how it knows a section is stale, what evidence you keep, whose rules apply, and how you would measure it on your own dossier.

We found no published independent test of any dossier software or service, as of November 2026. A test on your own dossier, with sources you hold back, is the fairest comparison, and the same test measures your current process.

Test AMCP dossier software on one dossier with held-back sources, then decide on three numbers

In shortMeasure first, then test on one dossier, with errors you planted and sources you held back, then decide with three numbers. None of these steps commits you to a vendor.

1. Time five packets and your last refresh, by step. That is your baseline.
2. Choose one dossier and 15 to 20 past requests. Pick one with a recent label change.
3. Hold back three recent sources. Plant known errors in copies: an old price, a superseded label cited, a value from a figure that disagrees with its table, an unrequested file.
4. Run it alongside your current process on live requests for six weeks, with your people signing every section and packet.
5. Decide with three numbers: sections correctly found when a source arrives, false alarms, and staff hours to an approved packet.

Sections found, false alarms and hours per approved packet

AMCP dossier questions answered: Format 6.0, unsolicited requests, HCEI, PIE and budget models

In shortThe questions people search for most about AMCP dossiers, each answered on its own in a few sentences.