The Centers for Medicare and Medicaid Services finalized the Global Benchmark for Efficient Drug Pricing Model on Sept. 30, and the Federal Register published the rule on Oct. 2. It takes effect Nov. 30 and begins operating Jan. 1, 2027. It is a mandatory model under section 1115A of the Social Security Act, which is the authority Congress gave the agency to test payment changes without new legislation.
The numbers in the final rule are the story. When CMS proposed the model in December 2025, it projected about $2.3 billion a year in Part B savings. The final rule projects about $80 million a year, or roughly $440 million across a payment period that runs through March 31, 2034. Beneficiary out-of-pocket savings are now put at $111 million over the life of the model, split between $50 million in coinsurance and $61 million in premiums. The proposed rule had estimated $1.4 billion in coinsurance savings alone.
A model that was going to move billions now moves tens of millions, and the difference is almost entirely a set of choices the agency made in response to public comments rather than a change in the underlying policy.
The exclusions did most of the work
Three carve-outs account for a large share of the revision. The final rule excludes drugs approved only for orphan indications, certain cell and gene therapies, and plasma-derived products. None of those exclusions appeared in the proposal. The agency also pushed the start date from October 2026 to January 2027, which removes a quarter from the calculation before any rebate is collected.
The exclusions are defensible on their own terms. Orphan-only drugs treat small populations where a rebate can become a disincentive to develop the next product. Cell and gene therapies are one-time treatments whose costs do not fit an annual rebate formula. Plasma-derived products come from a supply chain that behaves unlike a conventional pharmaceutical market.
Together they also remove a substantial part of the high-price, single-source Part B spending the model was designed to reach. The drugs that remain are cheaper relative to the ceiling, and a rebate formula applied to a cheaper baseline produces a smaller number. The agency revised its estimate down because the population it will apply the formula to changed, not because the formula was judged ineffective.
CMS waived the manufacturers it most needed inside the model
The interaction with a separate model is the more consequential change. CMS wrote that manufacturers that signed agreements under the GENEROUS Model by Aug. 17, 2026 will be waived from participating in GLOBE. GENEROUS is the voluntary Medicaid most-favored-nation model announced in November 2025 and launched in January 2026, and it is the vehicle the administration has used to consolidate its discounting deals with manufacturers across the two programs.
The agency's stated reason is methodological. Participation in both models, it wrote, would affect its ability to isolate the effect of each. That reasoning is sound as evaluation design and unusual as policy. A mandatory model is supposed to apply to everyone in its scope so that the results can be read. Waiving the manufacturers that already signed a related agreement means the companies most likely to be affected by an international benchmark are the least likely to be included in the test of it.
CMS did not write a formal exemption into the regulatory text. It noted in the final rule that its estimates assume the waivers and that a change in which manufacturers join GENEROUS would change the GLOBE projection. The practical effect is that the reach of a mandatory model now depends on participation in a voluntary one, and the agency has said so in the same document that sets the mandatory terms.
The drugs left in scope are the ones with the least room to fall
GLOBE covers single-source drugs and sole-source biologics with more than $100 million in annual Original Medicare Part B spending, in classes including oncology, rheumatology, immunology, ophthalmology and endocrinology. The threshold is indexed for inflation. It excludes drugs with a negotiated Medicare price under the Inflation Reduction Act, and it excludes biosimilars along with their reference biologics once a biosimilar reaches the American market.
The logic of that scope is to target products with no competition, where a single seller sets the price and no substitute pulls it down. It is also a description of products whose manufacturers have the strongest reason to protect the price and the least reason to cooperate with a benchmark built partly on information they supply.
That is the design tension at the center of the model. The international benchmark is calculated two ways. One is the lowest adjusted price across the 19 reference countries, which include Canada, France, Germany, Japan and the United Kingdom. The other is a volume-weighted international net price that manufacturers may submit voluntarily. Where both are available, CMS applies the higher figure. The voluntary channel exists because net prices after confidential rebates are not public anywhere, and a manufacturer that declines to submit leaves the agency with gross prices that are higher than what foreign payers pay.
The model also cannot produce a rebate lower than what the manufacturer already owes under the existing Part B inflation rebate program. A GLOBE rebate applies only when the Medicare amount exceeds the international benchmark, and the money goes to the Medicare Supplementary Medical Insurance Trust Fund. Nothing in the rule sets a price. It changes the arithmetic behind a payment the manufacturer already makes.
A quarter of Original Medicare, chosen by ZIP code
The model applies to a randomly selected subset of geographic areas covering about 25 percent of beneficiaries who have Original Medicare as their primary coverage. Medicare Advantage enrollees are outside it. Beneficiaries in the selected areas may see reduced coinsurance beginning April 1, 2027, with their out-of-pocket costs tied to the international benchmark rather than the domestic one. Medicare's payment to the provider or supplier is adjusted downward by the same amount, so the reduction in what the patient owes is not a reduction in what the drug costs the program.
The geographic design creates a patchwork. Two patients with the same diagnosis on the same drug can face different coinsurance depending on which ZIP Code Tabulation Area they live in, and that difference will persist for the five years the model runs. Beneficiaries who want to avoid it can move into a Medicare Advantage plan, which the model does not touch. Cost containment models generally try to avoid giving the population a route out.
The timing lands in the middle of an enrollment decision. Medicare open enrollment runs from Oct. 15 through Dec. 7, and CMS released its 2027 Part D projections two days before announcing the GLOBE rule. Those projections show Medicare Advantage drug premiums falling about 38 percent to roughly $7 a month while stand-alone prescription drug plan premiums rise to about $36 a month. The two halves of the drug benefit are moving in opposite directions at the moment beneficiaries are choosing between them.
Both of the agency's own figures on American prices are in this rule
CMS Administrator Mehmet Oz said Part B patients and taxpayers have "paid significantly more for prescription medications than people in comparable countries." The agency's news release puts the gap at four times as much for brand-name prescription drugs, and its model page puts the same gap at three times. The rule does not reconcile them.
The discrepancy is minor next to the savings revision, and it is the kind of inconsistency that shows up in any large rulemaking assembled across offices. It also matters for how the model's results will be read later. If the premise is a threefold gap and the projected savings are $80 million a year, the model is testing whether a rebate formula can capture a fraction of a difference that the agency itself describes with two different numbers.
The most-favored-nation lineage runs through this rule without being named in it. The concept dates to the end of the first Trump administration, when it was applied specifically to Part B. A May 2025 executive order set a target price tied to the lowest price in Organisation for Economic Co-operation and Development countries with a gross domestic product per capita at least 60 percent of the American figure. GLOBE's benchmark is defined separately, using its own reference countries, and reaches the same objective through a rebate rather than a set price.
What the model can still measure
Five performance years run from April 1, 2027 through March 31, 2032, with rebate calculation, invoicing, collection and reconciliation continuing to March 31, 2034. The evaluation question the agency set out is whether an international benchmark lowers spending for beneficiaries and the program without reducing quality of care.
That question requires a treatment group and a comparison group, and the waivers make the treatment group small and self-selected in a specific way. Manufacturers that signed Medicare and Medicaid discounting agreements early are outside the model. Manufacturers that did not are inside it, and the agency has said it cannot separate the two programs' effects for the ones in both.
The savings figure tells you what to expect from the answer. A test projected to move about $80 million a year across a program that spends hundreds of billions on Part B drugs will produce a result that is difficult to distinguish from noise, in a population that was randomly assigned by geography rather than by patient. The agency can narrow a mandatory model almost to nothing through exclusion lists and waiver language, and this rule shows how that is done: not by withdrawing the model, but by revising the arithmetic and letting the participants remove themselves.
Primary sources
- Centers for Medicare and Medicaid Services, CMS finalizes new mandatory drug payment model to deliver lower drug prices for beneficiaries in Original Medicare Part B, news release, Sept. 30, 2026, for the model's purpose, the administrator's statement and the agency's comparison of American and international drug prices.
- Centers for Medicare and Medicaid Services, GLOBE (Global Benchmark for Efficient Drug Pricing) Model, model page updated Sept. 30, 2026, for the performance period, the payment period, the geographic scope, the covered drug classes and the coinsurance mechanics.
- Federal Register, Global Benchmark for Efficient Drug Pricing (GLOBE) Model, final rule, 91 FR 62936, Oct. 2, 2026, for the regulatory text, the effective date, the exclusions added in the final rule, the interaction with the GENEROUS Model and the basis of the benchmark.
- Government Publishing Office, Federal Register full text of the GLOBE final rule, for the complete rule as published.
- Managed Healthcare Executive, Final GLOBE drug model projects much smaller savings while 2027 Part D premiums split, for the revised savings estimates, the beneficiary cost-sharing figures, the Part D premium projections and the agency's explanation of the GENEROUS waiver.
- AJMC, CMS Finalizes Mandatory GLOBE Model to Test Lower Part B Drug Costs, for the launch date, the reference countries, the scope thresholds, the Inflation Reduction Act rebate background and the two published figures for the American price gap.