Medicare's first negotiated drug prices: what moved in 2026, and what we expect in 2027

A brand-name drug passes through several hands before it reaches a Medicare patient, and money flows back along a different route. There is no single "price": the list price (WAC) is published by the manufacturer; pharmacies buy at close to list; plans pay pharmacies a contracted amount at the counter; and manufacturers later pay plans confidential rebates in exchange for good formulary placement. In 2024, rebates cut the gross Part D cost of the first ten negotiated drugs by nearly 45% (MedPAC, March 2026).

The Inflation Reduction Act of 2022 lets Medicare negotiate a Maximum Fair Price (MFP) for high-spending drugs without generic or biosimilar competition: small molecules at least 7 years past approval and biologics 11 years past licensure when selected. For the rounds taking effect in 2026 to 2028 the law sets a ceiling: at least 25% off the average non-federal price for most drugs, at least 60% for drugs 16 or more years past approval, or the Part D net price if lower. At the pharmacy, plan payment plus patient cost sharing cannot exceed the MFP plus dispensing fees. Manufacturers can provide access to the MFP prospectively or through retrospective refunds. The default refund uses WAC minus MFP, with permitted alternatives. CMS's Medicare Transaction Facilitator exchanges claim data and offers an optional payment service. CMS reports that refunds were being sent within 21 days of service on average; this does not establish that every pharmacy recovered its actual costs.

Prices are adjusted for inflation each year (Eliquis: $231 per 30 days in 2026, $237.25 in 2027). A drug leaves once CMS finds a generic or biosimilar is really marketed: Entresto, Stelara, Xarelto and NovoLog/Fiasp leave on 2027-01-01. The Supreme Court declined to hear the manufacturers' challenges on 2026-05-18.

The design is a before/after comparison against drugs not yet affected: the 15 drugs negotiated for 2027 (no MFP in 2026), the 15 selected for 2028, and about 740 other brand-name products. This describes what happened; it does not by itself separate the MFP from list-price strategy, rebate changes, generic entry or the 2025 Part D redesign.

Between 2025 and 2026 Q1 the median change across the ten drugs was −64%, from −30% for the NovoLog entry to −71% for Enbrel. Holding each group's 2025 drug mix fixed, the ten fell −61%, while drugs negotiated for 2027 changed +1.5%, drugs selected for 2028 +1.2% and other brand-name products +1.1%. The NovoLog entry pools two brands: NovoLog itself (list price already cut in 2024) fell 9%, Fiasp 72%.

This is what the payment rule would produce: the law caps the Part D price of a selected drug at the MFP plus dispensing fees, and the spending file is built from claim amounts. It is consistent with MFP-based payment; it does not prove it for every claim.

NADAC records WAC-based cuts effective 2026-01-01 of 43% for Eliquis, 44% for Jardiance, 37% for Farxiga and 75% for Fiasp. Januvia, and Janumet in the 2027 group, had been cut 42% on 2025-01-01, and Imbruvica (no NADAC) on 2025-12-22 according to NCPA's notice: six of the ten have had list-price cuts since 2024. Most of the other-brand cuts are not independent: 30 of the 33 in January 2026 are Boehringer Ingelheim's empagliflozin combinations, cut by the same 44.4% on the same day as Jardiance.

Why cut list prices? Candidate explanations, not established: under the default refund (list minus MFP) a lower list price shrinks the refunds owed; Bristol Myers Squibb told Reuters a lower Eliquis list price helps avoid Medicare inflation penalties; the cuts also removed the inflation part of the 340B discount, ending "penny pricing" from 2026-07-01 (Eliquis $245 a package, Pennsylvania Office of Rural Health); and for insulin the Medicaid rebate cap ended on 2024-01-01. Their net effect on manufacturer revenue is not measured.

NADAC, the average invoice cost of retail community pharmacies, was 2.2 to 4.0 times the 2026 MFP per unit at the end of 2025 (NovoLog was already at 1.0). After the January cuts it fell to 0.99 for Fiasp, 1.33 for Eliquis, 1.65 for Jardiance and 2.00 for Farxiga. Januvia (2.70), Xarelto (2.84), Entresto (2.22), Stelara (3.32) and Enbrel (3.57) stayed high, some rising with 2026 list-price increases.

CMS's quarterly plan pricing files list each plan's reported unit price for the drugs it covers, in the same billing units as the MFP. For the ten drugs, plan prices were a median 2.77 times the MFP in plan year 2025. In the January 2026 release the median was 1.000: 92.7% of plan-product ratios were within 0.5% of the MFP (99.7% weighting drugs by their claims), 4.9% were below 0.995 (the MFP is a ceiling) and 2.3% exceeded 1.005, all of those unbranded ustekinumab products grouped with Stelara. By October, no ratio exceeded 1.005; the maximum was only about 0.003% above 1.000, consistent with rounding. The fifteen drugs negotiated for 2027 stayed above their future MFP.

A scoping scan of about 180 papers and grey-literature items (2026-10-20) found no published measure of realized 2026 spending across the ten drugs, and no weekly product-level NADAC-to-MFP series through 2026. The closest work covers Stelara alone (Castle et al. 2026, JAMA Network Open) or compares price benchmarks at one point in time (PORTAL, 2024). The list-price cuts are documented by NCPA, 46brooklyn and Drug Channels but not yet in academic work.

The 2026 results were seen before any test was designed, so they cannot confirm a hypothesis. The fifteen drugs whose MFPs start on 2027-01-01 can. On 2026-10-20 we registered five predictions in git (commit b24d4c8, before examining any plan-year 2027 pricing or outcome data); clarifications added after review are dated amendments that never change a registered number.

The rule. If the MFP binds, a drug's gross spending per claim should fall by about 1/r − 1, where r is the plan price divided by the coming MFP in the last plan pricing file before it starts. It has no fitted parameter. Applied retrospectively to the 2026 drugs, the rule missed by at most 2.1 percentage points for eight of ten. The two large misses were identified after examining the results: Entresto's brand claims fell sharply after generic entry, and the NovoLog family pools differently priced NovoLog and Fiasp products. These are plausible explanations, not an identified decomposition of the errors. Across all ten, mean absolute error was 4.7 points; the fixed-2025-claim-mix prediction was −62.2%, compared with −61.4% observed.

Amendments after review

Amendment 1 (2026-10-20, after the Codex review). Codex reproduced all 221 baseline numbers. The amendment fixes how each prediction is scored without changing any registered number: every clause is scored separately as pass, fail or not evaluable, with no overall verdict; the primary plan-price test uses the first CMS release containing plan-year 2027 unit prices, expected in January 2027, with later releases as replications; rows are contract x plan x segment x NDC with a 30-day supply, at full precision against the frozen 2026-10-20 MFP file after dropping superseded rows; missing products are listed, never counted as passes. P2 fixes each drug's weight at its 2026 Q1 claims: the observed index is the sum of 2026 Q1 claims times 2027 Q1 spending per claim over the sum of 2026 Q1 claims times full-year 2026 spending per claim, minus 1, scored against the registered −68% to −58%. The semaglutide exception is withdrawn: its median ratio must lie between 0.995 and 1.005, inclusive, like every other 2027 drug; compliance with the MFP ceiling is reported separately (Medicare's $50-copay GLP-1 Bridge operates outside Part D coverage and payment through 2027). It also corrects the registration's data note: the monthly 2027 release CMS posted on 2026-10-08 has no pricing table, so the plan-price test waits for the January 2027 quarterly release. The calibration is reported in full: across all ten drugs the mean absolute error is 4.7 points and the rank correlation 0.72, so P2's rank clause (0.8) is a harder test than the rule has passed so far.

Amendment 2 (2026-10-20, after the Agy review). Each Agy finding was checked against the data before use. Applied: the missing Fiasp row was restored to one baseline table and Farxiga's 2026 MFP corrected to $178.50. Five sensitivities were specified: a Q4-only 2026 baseline for all 15 drugs (Q4 year-to-date minus Q3 year-to-date spending and claims); separate Ozempic (r = 3.66) and Wegovy (r = 3.33) results, keeping the pooled semaglutide prediction primary; brand-only plan-price shares, keeping every listed NDC against its own MFP in P1; a 2026 Q1 baseline for P4 and P5; and a 2028 comparator excluding drugs with WAC-based cuts of at least 20% effective 2026-12-20 through 2027-02-28. Xifaxan's ratio rose from 2.80 to 3.25 during 2026, so its decline against the full-year baseline is expected to be smaller than the registered October-based −69%. Trulicity is 42% and Anoro Ellipta 13% of the 2028 comparator's 2026 Q1 claims. Not applied: Agy listed wrong values for Gabriel et al.'s discounts while calling them an exact match, described a scoring band that does not exist, and summarized a paper with figures not in its text. The rank-correlation threshold stays at 0.8.

Codex (OpenAI GPT-6 Astra, high effort) independently recomputed the project's analysis outputs from raw files, in its own folder, before reviewing the analysis code and wording. Agy (Google Gemini 4 Argon, high) recomputed headline figures and reviewed methods; its findings were treated as leads and checked against the data or sources before any change. Literature findings were checked against publications, not re-created from their underlying study data. The table records the reviews completed for each component; those reviews do not by themselves certify every later summary sentence. A final Codex fact-check of this page compared 2,491 embedded values with the outputs and corrected its wording.