Medicare drug price negotiation, explained.

IRA drug price negotiation in plain words: a selected drug, a government offer and a four-week file

From selection list to a published price: five steps for one selected drug

1. CMS publishes the selected drug listBy February 1. On February 1, 2027, in this guide's fictional example, Tarvessa is on it.
2. The company signs the agreementBy February 28. Data can only be filed once both sides have signed.
3. The team builds and files the dataBy March 1. Every figure found, checked, signed and attested. This is where the $0.55 counted twice is caught, or goes to CMS.
4. CMS makes an offer and the company answersOffer by June 1, answer within 30 days, meetings through the summer, final offer by September 30.
5. The price is published and appliedPublished by November 30, in effect January 1, 2029 for this cycle. The file can be used again if the drug comes back for renegotiation.

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

Why CMS starts from a rival's price, and three more negotiation surprises

Maximum fair price, non-FAMP, Primary Manufacturer and IPAY defined

Fourteen terms are defined in the one-page summary.

How CMS picks selected drugs: spend, years on the market and no generic rival

In shortCMS ranks eligible drugs by Medicare spending and takes the top of the list: up to 20 a year from IPAY 2029. A small-molecule drug, the chemically made kind, qualifies 7 years after approval, and a biologic 11 years after licensure, if no generic or biosimilar is being sold in earnest. Medicare Part B drugs, the ones doctors give in the clinic, joined the list in the third cycle.

Eligibility is a calendar anyone can run: take your approval date, add 7 or 11 years, and compare your Medicare spending with the last list. That is why readiness work starts before selection. Advisers at a 2025 industry meeting called it best practice to be ready by January 1, a month before the list.19

The IPAY 2029 negotiation timeline: February 1, 2027 to the price in effect on January 1, 2029

In shortThe statute fixes the frame: list by February 1, agreement by February 28, data by March 1, offer by June 1, counter within 30 days, negotiation over by November 1, prices public by November 30. IPAY 2029 would be the first cycle CMS runs under regulation, and as of November 3, 2026 that regulation is still a proposal: the final rule is under review and unpublished.

Law, final guidance or proposal: the status of each text behind this file, November 3, 2026

Why four weeks can run a little longer, and why that barely helps

February 1 is a deadline for CMS. It published the last two lists early: January 17, 2025 and January 27, 2026.15,16 Companies do not get more time from that in practice, because the agreement must be fully signed by both sides before data can go in, though CMS opens the template in its Health Plan Management System (HPMS) earlier so teams can gather figures.3

Three other dates sit inside the frame. A company must report a new National Drug Code (NDC) to CMS at least 30 days before first marketing it. It must tell CMS when a figure it filed is restated for another government program, such as restated non-FAMP. And if CMS finds a gap, its Request for Corrective Action sets a five-business-day deadline.3

What the negotiation data elements ask for: non-FAMP, R&D, unit costs, federal support, patents and sales

In shortThe law names five kinds of company data and four kinds of evidence, nine factors in all. Non-FAMP comes on top. Company data is mandatory once you sign; evidence is optional and open to anyone. CMS collects it on form CMS-10849. The file CMS published for one IPAY 2027 drug shows question numbers running to 61, and the form proposed for IPAY 2029 runs to 62.

Who CMS expects to spend the 1,000 hours

How the maximum fair price ceiling is calculated from non-FAMP and the Part D net price

In shortThe law caps every offer. The ceiling is the lower of two amounts: what Part D plans pay after rebates and other price concessions, or a set percentage of non-FAMP. The percentage depends on how long the drug has been on the market: 75%, 65% or 40%. The ceiling is simple arithmetic, and it says nothing about where CMS will start.

Tarvessa's ceiling worked through, $612.40 of non-FAMP down to $459.30

Three things make the real calculation harder than this example.3

1. Two non-FAMP figures. The percentage applies to the lower of two figures, so a drug on the market in 2021 also needs its 2021 figure indexed to the year before selection. The example assumes 2026 is the lower.
2. CMS's own Part D figure. CMS uses its own plan data and netting for the Part D amount, the sum of plan-specific enrollment-weighted amounts, which can differ from a company's figure.
3. One price per 30-day supply. The price is set per 30-day equivalent supply and is one figure across strengths and forms. A once-daily bottle of 30 happens to line up. An injection given every eight weeks needs CMS's conversion.

For Part B drugs, the first amount uses Part B payment rules, or blends Part B and Part D where a drug is paid under both.3

How CMS develops the initial offer: alternatives' price first, evidence second, your data third

In shortCMS starts from what therapeutic alternatives cost, moves that figure up or down for clinical benefit using the evidence, moves it again using your five kinds of company data, and caps it at the ceiling. Your cost is one input at the last step. The method is qualitative, and CMS decides how much each factor weighs. These steps are final for IPAY 2028, and CMS proposes to carry them into regulation for IPAY 2029.

How each kind of company data can move the preliminary price

CMS looks at the five together, so one factor's direction may not decide the outcome. Table built from section 60.3.4 of CMS's IPAY 2028 guidance.3

Before the offer, CMS holds one optional spring meeting with each manufacturer. You can give context on your data and share new evidence on alternatives, within 50 pages of material. New company data will not be considered at that meeting.3

Calculating unit cost of production and R&D recoupment to CMS's definitions

In shortThese are the two sections where company accounts and CMS's definitions part ways. Unit cost takes direct and allocated indirect cost, by NDC-11, for US sales only. R&D splits into the drug's own costs and failed products with the same mechanism of action. Each needs a written method, and that method is what CMS reads when it asks a question. The definitions are final in the IPAY 2028 guidance, and the form proposed for IPAY 2029 repeats them.

The Tarvessa double count: where $0.55 a bottle came from

A plant costing system usually works from batches: materials, conversion, testing, packaging and an overhead rate. Finance works from cost centers and journals. In the example, the plant's roll-up for 38 batches charges quality control (QC) release testing directly to each batch at $0.55 a bottle. Journal JE-26-118734 also put the same testing, $902,000 for the year, into the overhead pool. So the roll-up carries it twice.

A difference this small would leave an offer where it was, yet it still has to be found. The officer attests that the data is complete and accurate, and CMS may audit. A team that cannot explain its own $0.55 will struggle with bigger questions in a two-hour meeting with CMS.3

R&D recoupment: failed products, $341.0 million allocated, and revenue that dwarfs it

Recoupment means how far revenue has covered R&D costs, and CMS compares the two sides. The R&D total adds the drug's own program costs to its allocated share of failed products, so both the revenue and the allocation need a method you can explain.3

Evidence on therapeutic alternatives: what CMS prioritizes, and why QALYs never count

In shortAnyone can file evidence by March 1, and CMS runs its own literature review as well. It prefers rigorous research large enough to show a real effect (well-powered), direct comparisons where they exist, and research designed around Medicare-age and disabled people. It will not use quality-adjusted life years. You can add new evidence later, in the spring meeting and in your counteroffer.

How CMS weighs a study, in its own terms

CMS says it considers source, rigor of method, current relevance, peer review, study limits, certainty, bias, time horizon, generalizability and population. It says it will prioritize research that is well-powered and built to avoid false positives from many subgroup analyses, and it treats research on Medicare populations, including disabled people and people with end-stage renal disease, as particularly important.3 In its IPAY 2027 explanations it said it generally prioritized direct comparative evidence, such as head-to-head randomized trials, when available.15

A drug selected in its eighth year or later often rests on the trials that won its approval, which predate its newest rivals. That is our inference, and it is a good reason to check what comparisons exist years before selection. Productivity, independence and quality of life count when they directly affect the people taking the drug and can be measured.3

CMS publishes redacted submissions with its explanations, so a company's and the public's filings can become readable later. Proprietary data stays protected under Exemptions 3 and 4 of the Freedom of Information Act (FOIA) and CMS's confidentiality policy.3,29

Negotiated prices so far: discounts from 38% to 85%, and how the talks ended

In shortIn the first two cycles, every negotiation ended in a price. Second-cycle prices run 38% to 85% below 2024 list, and first-cycle prices 38% to 79% below 2023 list. CMS estimates the second cycle would have cut 2024 net spending by $12 billion. Most talks did not end at the first exchange: CMS reports more than 100 revised offers or counteroffers in the second cycle.

List-price discounts overstate the change, because list ignores rebates companies already paid. CMS's own measure below compares net spending.

Agreement by revised offer or by final offer: how all 25 talks ended

In the second cycle, every company declined the initial offer and countered, CMS declined every written counteroffer, and three meetings per drug followed. CMS revised its offer upward at least once for each drug, and each company revised downward at least once. CMS says it moved more than the company in 8 of the 15 negotiations.15,32

Under the IPAY 2028 guidance, each meeting runs no more than two hours with up to eight attendees a side, cannot be recorded, and stays on the statutory factors, the alternatives and the price.3

Part B drugs, renegotiation, the court cases and enrollee savings: where the program stands

IRA negotiation penalties: $1 million a day, the $100 million rule's real scope and the section 5000D excise tax

In shortLate, missing or false required data is a violation of the agreement, and carries a daily penalty that the statute sets at $1 million. CMS says it reserves penalties for substantive noncompliance and offers a chance to correct first. Refusing to sign or to agree brings the larger threat: an excise tax on sales for people with Medicare that reaches 1,900% of the price.

When the daily penalty starts and stops

CMS's stated intent is to keep penalties for "substantive noncompliance". Its examples include failing to submit required data and omissions or inaccuracies critical to the process, such as non-FAMP for a partner's sales.3

HHS's January 28, 2026 update set its 2025 amounts, and its October 21, 2026 notice cancelled the 2026 adjustment and keeps the 2025 levels. Its table at 45 CFR 102.3, read for October 30, 2026, has no line for section 1197, so CMS's $1,059,232 stays the latest figure published.36,26,37

The excise tax: what happens if you do not sign or do not agree

The tax applies to sales during a noncompliance period. Those periods begin March 1 if no agreement is signed, November 2 if no price is agreed, and on the date HHS certifies required information overdue.13 Treasury's proposed rules, which are not yet final, tax only units dispensed, furnished or administered to people eligible for the negotiated price, the reading it first gave in Notice 2023-52.14,27

There is a way out: a company may leave the program and avoid the tax for periods when it has no Medicare and Medicaid agreements and none of its drugs are covered by the Part D discount agreements.13,3 CMS reports that all third-cycle makers chose to participate.38,25 The Treasury and IRS administer the tax.3

Ten ways a negotiation data file goes wrong, from a double-counted cost to the wrong year's non-FAMP

In shortCMS publishes no count of errors or findings, so nobody can say how often files go wrong. The list below is our reading of where CMS's definitions and company systems diverge. Most errors come from figures built for another purpose and carried over unchanged.

A published CMS document shows what reviewers can see. In one IPAY 2027 file, the company explained a total acquisition cost, line by line, including its source in a public annual report.29 That level of explanation is what a free-text method field asks for. Teams also have an ongoing duty: when a figure is restated for another government program, CMS must be told, and the proposed certification adds a promise to report anything that has changed or is inaccurate.3,4

Using AI for negotiation data: what CMS, FDA and European rules say, and what stays with people

In shortWe found no CMS rule on AI used to prepare the file. The responsibility stays where it was: an officer attests to it. FDA's AI draft guidance leaves out submission drafting that does not affect the reliability of results, and it addresses FDA decisions anyway. European good manufacturing practice (GMP) drafts cover manufacturing decisions. Treat AI output as an unsigned draft from a new analyst, and check it the same way.

Collecting, tying and citing: where software helps on a negotiation file

How market access teams prepare today: who touches the file, the hand-offs and the twenty working days

In shortOne director runs a program across finance, the plant, pricing, legal and HEOR, usually in spreadsheets, shared folders and email, often with a consultancy or law firm helping. Advisers say to allow at least nine months and to be ready before selection. Nobody publishes how the time splits.19

Medicare negotiation director to CFO: who carries the file

Advisers describe readiness reaching "from market access teams, to corporate affairs, all the way to the CFO".19

Six hand-offs from the form map to the officer's attestation

Where a selected drug's twenty working days go

Consultancy, spreadsheets or a system of record for the file

Tarvessa's file: 41 of 44 elements ready, one $0.55 difference found, signed and on to the CFO

In shortThis is the example from the top of the page, inside the product. Veltara Therapeutics' Tarvessa was selected on February 1, 2027. On Wednesday, February 17, Nadia Brennan opens a board with the statutory clock and 44 tiles. The amber tile is unit cost of production. The Number reconciler finds the duplicate and proposes a figure, the plant controller signs it, and the CFO's certification stays locked until the last two elements are signed.

1

The statutory clock over 44 data elements

Nadia's home is a countdown board: 12 days to the data due date, internal cut-off February 24; 104 days to CMS's initial offer; 134 to the counteroffer, 7 of 8 arguments drafted; 256 to the end of negotiation. Below it, the 44 elements sit in seven form sections, each filling in three bands: sourced, reconciled, signed.

All 44 sources were re-collected overnight. 43 came back unchanged and one difference was found. Three elements are open.

2

One amber tile: $0.55 off the general ledger

Nadia clicks the amber tile. The number trace opens beside the board: plant costing system, extract PCS-ROLLUP-2026 across 38 batches, and a roll-up of $14.82 against $14.27 in the general ledger, over the $0.01 tolerance.

The other two open elements are waiting for a person: pending FDA applications for Owen Castillo's signature, and the real-world study for Dr. Meera Anand's review on Thursday.

3

QC release testing, counted twice

Nadia runs the reconciler. It reads the roll-up for batches VR-2611 to VR-2648, then cost center 4410. It compares the seven components: six agree to the cent. It searches 412 lines of the overhead pool's journals and finds JE-26-118734, "QC release testing, 2026", $902,000. The same testing is already charged direct to each batch at $0.55 a bottle.

It proposes $14.27, which ties to the cent, and leaves the decision to the owner.

4

The owner signs, and the tile turns green

Nadia accepts the proposal and sends it to Ji-woo Park, the plant controller at Raleigh, who owns the unit-cost section. Ji-woo signs at 9:16 a.m. The board reads 42 of 44.

The trace now carries a prepared answer, component by component, including the note that an earlier roll-up counted QC testing twice and was corrected before certification.

5

Certification waits for every signature

Helen Marsh, the CFO, is due to certify on Friday, February 26. The certification screen lists each section with its owner and signature. Two sections are still open, each with a Remind button: real-world comparative effectiveness with Dr. Meera Anand, and pending FDA applications in Leah Morgan's patents section, waiting for Owen Castillo's legal signature.

Until all 44 are signed, the Certify button stays locked and reads "2 open". The file goes to CMS through the company's own submission.

The solution's design targets are zero days late against any CMS deadline, with the file complete before the internal cut-off, and 100% of figures traced to their system, general ledger tie and signer. They are targets, to be measured on your own drug.

Ten more screens, from the full data file to a counteroffer drafted argument by argument

In shortThe example follows one figure. These screens show the rest: the data file, a figure broken into components, the failed-product answer, the ceiling and offer simulator, the evidence, the statutory calendar and the rules your team sets.

Select any screen to open it larger. The company, drug, trials, people and figures are fictional. Wording on these screens follows the sample data; Part A gives each rule as CMS states it.

What a $318 offer and a $402 counteroffer do to Part D revenue

Six agents that read and draft, with owners signing every element and an officer certifying the file

In shortEach agent has one job and limits it does not cross. None of them changes a figure in a source system, moves a statutory date or sends anything to CMS. Section owners sign, counsel signs legal elements and reviews every draft, the director approves, and the certifying officer certifies only when all 44 are signed.

Form mapper to Deadline keeper: the six negotiation agents

Owners sign, counsel reviews, the officer certifies

Signatures record the name, the role and the time. Agent steps and human decisions are kept apart on each figure's history, settings are saved as new versions, and your own people submit in CMS's system. Security: SOC 2 Type II audited, with information security managed to ISO 27001. Controls like these support your own compliance program; a product cannot be certified to GxP or Part 11, and the claim would mean nothing for a CMS filing.

The director, controller, HEOR lead, pricing lead and CFO in these screens

Selection notice and six systems in, a traced data file, an evidence dossier and counteroffer drafts out

In shortThe solution reads the CMS form and your systems, and writes nothing back to them. It leaves behind a complete file with each figure traced and signed, a dossier on alternatives and draft arguments, plus a record you can use again for renegotiation or the next selected drug. Your systems of record stay where they are.

Systems a negotiation rollout reads from

For each, check whether it has an API, a file feed or an export, and who owns it. Read access is enough.

Your costing policy, HPMS submission and systems of record stay

What a negotiation file costs you in hours, rework and exposure before CMS ever makes an offer

In shortCMS gives a planning figure for the hours, and the statute gives the penalties. The numbers that matter most are yours, and a week of counting will find them.

Rehearsing one selected drug to find your own baseline

Offers you cannot defend, and the costs outside the calculator

What traced data elements return, counted in hours against the yearly cost of the tool

In shortPut in your own numbers. On hours alone, one selected drug rarely pays for a tool. Two or more drugs, a large portfolio on the watch list, or a single avoided rebuild during negotiation changes that. The stronger reasons are in the costs no calculator counts, and you should say so plainly in your business case.

Collection, reconciliation and follow-up: where negotiation files save

Selected drugs a year, CMS's hours and your hourly cost

A company with one selected drug and a settled finance team may well decide a program office and an adviser are enough. A company with a portfolio approaching 7 and 11 years, or partners in the data, is closer to that break-even.

How ready your team is for the data deadline: watch list, definitions, records, evidence 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.

Rehearse by hand, pilot on a watch-list drug or plan the whole program: what each score calls for

What to ask IRA negotiation software or a readiness adviser about data elements, traces and certification

In shortAsk every vendor the same questions, us included. The answers show how the file gets built: whose definitions apply, what a figure is tied to, who can change it, how the evidence is checked, and what you would keep afterward.

We found no published independent test of any negotiation software or readiness service, as of November 2026. A rehearsal on one of your own drugs is the fairest comparison, and it measures your current process too.

Rehearse a selected drug before the next February list, then decide on three numbers

In shortChoose the drug most likely to be selected, rehearse the file by hand for a sample of elements, then test a tool on the same drug with planted errors. Decide with three numbers. None of this commits you to a vendor, and the rehearsal pays off even if you buy nothing.

1. Rank your watch list by approval date, spending and competition; pick the likeliest drug.
2. Name an owner per section and time five figures by hand, from system to signature.
3. Gather the source extracts and your costing policy, and plant known errors in copies: a duplicate, a transfer price, a wrong period.
4. Run the rehearsal with your owners signing every element, and bring in your evidence library.
5. Decide with three numbers: catch rate by error type, false alarms on clean figures, and hours from extract to signature.

Catch rate, false alarms and hours to a signed figure decide it; citations and answer time back them up

Medicare drug price negotiation questions answered: maximum fair price, the data due March 1, penalties and AI

In shortQuestions market access teams and buyers ask about the program, each answered on its own.