California Counted What the GLP-1 Benefit Cost. It Has Not Published What the Benefit Did.

I. The finding

California is unusually good at counting what it spends. Between 2018 and 2023, GLP-1 prescriptions written in California within a state-constructed weight-loss cohort rose from roughly twenty thousand a year to more than seven hundred thousand, across Medi-Cal and commercial payers together. PROVED In 2023 alone, Medi-Cal cost inside that cohort reached $416.8 million and commercial-payer cost reached $405.2 million, both gross of rebates. PROVED

The Legislature’s nonpartisan analyst put the trend in programme terms. Medi-Cal pharmacy spending nearly doubled between 2018–19 and 2023–24, and nearly half of that growth was associated with a cluster of drugs treating diabetes, obesity and inflammatory disease. PROVED A year later the same office estimated that General Fund pharmacy costs per enrollee had grown by nearly 13 per cent a year since 2017–18 — a figure calculated after federal and state rebates, and one that translates into nearly $4 billion in additional General Fund spending in 2025–26. PROVED

On evidence of that kind, the 2025 Budget Act ended Medi-Cal coverage of GLP-1 drugs used for weight loss, effective 1 January 2026. The enacted spending plan scored the action at $85 million in General Fund savings in 2025–26, rising to $790 million by 2028–29. PROVED The Department of Health Care Services executed the change through a state plan amendment that the federal government approved on 12 March 2026. PROVED

What the published state and federal record does not contain is any Medi-Cal-specific evaluation of what the expenditure achieved. No published series on weight change among the beneficiaries who received the drugs. No published analysis of treatment persistence or discontinuation. No published figures on incident diabetes, cardiovascular events, or downstream medical utilisation and cost. PROVED — as to the absence from the published record examined here.

The asymmetry is the finding. California built a cost record detailed enough to justify ending a benefit, and did not publish an outcomes record capable of testing whether ending it was sound.

That claim is narrower than it sounds, and the narrowness is deliberate. It is not a finding that the drugs failed. It is not a finding that the termination was wrong. It is not a finding that nobody inside the department ever looked. It is a finding about the public evidence architecture: the machinery that counted the money was not matched by machinery that counted the medicine, and the second half was never built into the accountability framework even by the officials who asked for more transparency.

II. What the cost record contains, and what each number is not

The $416.8 million figure is load-bearing in the public argument, and it has to be handled precisely. Three qualifiers travel with it, and each one changes what may be inferred.

It is a gross figure

The Department of Health Care Access and Information states that its cost figures do not reflect manufacturer rebates, coupons or other discounts, because those amounts are not reported to it. PROVED In Medicaid that omission is enormous. Manufacturers participating in the Medicaid Drug Rebate Program owe formula-based federal rebates, and California negotiates supplemental rebates of its own. The analyst’s own pharmacy-cost work is calculated net of both. PROVED

Individual drug rebate amounts are confidential by federal design. That is not a gap in this review; it is a structural feature of the programme. Nobody outside government can compute the net Medi-Cal cost of these five products, and this article does not attempt an estimate by applying an aggregate rebate percentage to a drug-specific gross figure. Rebates vary by product, by pricing history and by inflation penalty; multiplying one by the other would produce a confident number with no evidentiary basis. NOT ESTABLISHED

It is a figure about a cohort, not a diagnosis

The state’s dataset excludes people identified as having diabetes, in order to raise the likelihood that the prescriptions remaining represent weight-loss use. PROVED That is a sound way to build a population-level signal. It is not an adjudication of why any individual clinician wrote any individual prescription.

The cohort method cannot screen for the body-mass-index thresholds under which Wegovy, Zepbound and Saxenda carry approval for chronic weight management. It cannot confirm that a qualifying weight-related comorbidity was present where one was required. It cannot exclude off-label use. It follows that the statement “California spent $416.8 million on FDA-approved obesity treatment” is not supported by this dataset, however natural it sounds. NOT ESTABLISHED The accurate statement is longer and worth the extra clause: California’s all-payer claims data record $416.8 million in 2023 Medi-Cal cost within a constructed non-diabetes cohort intended to approximate weight-loss utilisation.

It is a total-funds figure, not a General Fund figure

Medicaid is jointly financed. The federal share of Medi-Cal cost is 50 per cent in most cases and higher or lower for particular populations — 90 per cent for qualifying childless non-disabled adults, and nil for most non-emergency services provided to people with unsatisfactory immigration status, whose drug costs also generally do not attract the same federal rebate treatment. PROVED Rebate savings are themselves shared between the two governments under the applicable financing rules.

A dollar appearing in a gross Medi-Cal pharmacy dataset is therefore not a dollar borne by California’s General Fund, and for some populations it is closer to a dime.

The better anchor for the state-cost conversation

The Legislature’s analyst supplied a narrower and more usable estimate: for 2023–24, Medi-Cal spent in the low-to-mid hundreds of millions of dollars in total funds on Wegovy, the principal anti-obesity brand, and the state share was likely less than half of that, particularly after rebates. DOCUMENTED The analyst characterises much of its pharmacy work as approximate, because it combines sources with differing methodologies; the estimate should be read as an order of magnitude rather than an audited figure.

It is nonetheless a far more defensible anchor than any arithmetic performed on the all-payer cohort total, and it comes from the Legislature’s own office rather than from an advocate on either side of the question.

III. Why the benefit became a target, and what happened to the estimate

The fiscal context was not manufactured. When California’s Medi-Cal costs ran ahead of estimate in the spring of 2025, the administration named three drivers, and the first of them was increased utilisation of high-cost anti-obesity drugs. PROVED The May Revision proposed a package of solutions; the enacted budget adopted a version of it; and ending coverage of GLP-1 drugs used for weight loss was among the largest pharmacy items in the package.

The comparative position mattered too. As of August 2024 the analyst counted twelve other state Medicaid programmes covering anti-obesity drugs, citing a national survey. DOCUMENTED A later count by the same national survey put the number at sixteen states including California as of 1 October 2025. DOCUMENTED Both figures are snapshots with dates attached, and neither describes the position today. What the analyst added is harder to snapshot away: nearly 40 per cent of adults enrolled in Medi-Cal report obesity, a higher share than among Californians with private insurance or Medicare. DOCUMENTED

So California was covering an expensive and fast-growing drug class, for a population with unusually high clinical need, at a moment of acute budget pressure, under a federal rule that left the coverage optional. That combination is a coherent explanation for the decision. It is also precisely the combination that makes an outcomes evaluation more valuable rather than less, because it describes one of the largest publicly financed natural experiments in anti-obesity pharmacotherapy in the country, in a population that clinical trials systematically under-represent. INFERRED

The estimate was revised upward — alone among the pharmacy solutions

There is a detail in the 2026 budget documents that has attracted almost no attention and deserves some. When the Governor’s January 2026 budget updated the savings estimates for the previous year’s enacted solutions, most of the pharmacy items came in below expectation. The rebate aggregator lost $247 million of assumed 2025–26 savings. Step therapy, utilisation management and continuation-of-therapy prior authorisation were all marked down. Across the whole prescription-drug package the administration now expects around $350 million in 2025–26 savings, roughly half of what was assumed at enactment. PROVED

The end of anti-obesity coverage moved the other way. Its 2026–27 estimate rose from $215 million to $364 million, an upward revision of $149 million, and it is the largest single upward revision in the table. PROVED

Swipe the table sideways for the remaining columns

Two things follow. First, on the state’s own updated arithmetic, this is now the most fiscally productive pharmacy action California took in 2025 — the analyst records that most of the savings from ending drug coverage come from the obesity drugs. PROVED Second, and this is the uncomfortable part, the upward revision is a statement about avoided expenditure and about nothing else. INFERRED The estimate got larger because more spending was avoided than expected, which is another way of saying that demand for the therapy was higher than the state assumed. Whether the demand that was turned away was clinically productive demand is exactly the question the record does not answer.

The state now expects the coverage cut to save more than it forecast. It still has no published way to say what the additional saving cost in health.

IV. The legal architecture, stated precisely

The termination was lawful, and the mechanism deserves care, because loose descriptions of it circulate widely.

Under federal Medicaid law, a state that covers outpatient prescription drugs must generally cover nearly all drugs from manufacturers participating in the Medicaid Drug Rebate Program. Section 1927(d)(2) of the Social Security Act, codified at 42 U.S.C. §1396r-8(d)(2), carves out an exception: a state may exclude or otherwise restrict coverage of specified categories, including agents when used for anorexia, weight loss or weight gain. PROVED

That is the correct formulation, and it is the department’s own. The looser version — that federal law makes weight-loss drugs “an optional benefit” a state may decline — obscures the actual structure. The statute does not make a benefit optional. It permits the exclusion of specified uses of covered outpatient drugs. That distinction is the entire reason the same molecule can be covered under one indication and excluded under another, which is exactly what happened here and what still describes the position today.

State Plan Amendment 25-0029

The instrument was State Plan Amendment 25-0029. DHCS transmitted it to the Centers for Medicare & Medicaid Services on 30 December 2025, proposing to exclude from coverage GLP-1 agonists used for weight loss, seeking an effective date of 1 January 2026, citing Section 1927, and proposing changes to Attachments 3.1.A.1 and 3.1.B.1 of California’s Medicaid State Plan. PROVED CMS approved it on 12 March 2026, effective 1 January 2026. PROVED The submission is posted by DHCS and the approval package is published in the federal state-plan repository; both are public primary documents.

This was not conventional rulemaking

California implemented the change through the enacted budget, a Medicaid state plan amendment, and pharmacy-benefit operational policy. Describing that sequence as a “regulatory process” implies a proceeding under the state Administrative Procedure Act, with the comment structure and procedural protections that go with it. NOT ESTABLISHED The accurate description is a budget-enacted coverage decision implemented through the Medicaid state-plan and pharmacy-benefit administrative process.

Notice and appeal rights existed

The change did not arrive unannounced. DHCS issued tribal notice in November 2025 and an addendum thereafter, and states that the member notification letter would have been received on or before 1 November 2025. PROVED Members receiving GLP-1 medications on or before 1 January 2026 who submitted a written appeal before that date continued to receive them pending a State Hearing decision, until the earliest of the end of their existing authorisation period, a hearing decision, or withdrawal of the hearing. PROVED

V. What actually happened at the pharmacy counter

The operational history has three completed phases and one scheduled fourth. Reporting only the first — as most accounts still do — leaves a reader with a picture that has been superseded for five months.

Swipe the table sideways for the remaining columns

The current position is therefore not “Medi-Cal does not cover GLP-1s.” It is narrower and considerably stranger. Medi-Cal excludes GLP-1 use for adult weight loss as a general pharmacy benefit, while covering the same molecules under other indications — and since April it has covered Wegovy itself, on the formulary and without prior authorisation, for a liver disease that is largely a downstream consequence of the metabolic condition the weight-loss exclusion declines to treat.

That structure is legally coherent and clinically peculiar, and it is the reason the state’s own cohort methodology — excluding diabetes-coded patients to isolate weight-loss use — maps so closely onto the population the cut fell upon. It is also the reason the 1 October 2026 date matters: a defined group of members who have been paying claims continuously since January will need either a diagnosis that fits or an authorisation that holds, and the record contains no published estimate of how many that is.

VI. The gap, and why it is institutional rather than accidental

Set out what California published, and the shape of the omission becomes visible. Prescription counts by year. Gross spending by payer. Cost per member by county. Growth rates. An attribution of pharmacy-cost growth to drug classes. Rebate ranges. A projection of gross pharmacy spending. A scored savings estimate for the termination, and then a revised one. A state plan amendment, a federal approval, member notices, appeal rights, and pharmacy rejection codes precise enough to reproduce at a counter.

What this review did not locate is any Medi-Cal-specific evaluation of clinical outcome. Specifically, no state-published figures for the population that received this benefit on:

• average or median weight change achieved among treated beneficiaries;
• the proportion reaching 5, 10 or 15 per cent weight reduction;
• treatment persistence at six, twelve or twenty-four months;
• rates of discontinuation, or weight regain after discontinuation;
• newly recorded type 2 diabetes among treated beneficiaries against an appropriate comparison group;
• cardiovascular events among treated beneficiaries;
• changes in antihypertensive, antidiabetic or lipid-lowering medication use after initiation;
• any downstream Medi-Cal medical-cost offset attributable to the benefit; or
• net cost per beneficiary after rebates.

The state’s own description of its data source is instructive for what it concedes rather than what it claims. The underlying dataset is administrative claims and encounters generated through transactions between payers and providers. That is a record of what was billed, not a record of what happened to the patient. It is an excellent instrument for utilisation and gross cost. It was not designed to answer a clinical-outcomes question, and the agency does not claim it was.

The transparency recommendations are the strongest evidence for the point

If the omission were a single agency’s oversight, one would expect the Legislature’s own analyst to have flagged it. The analyst did recommend more transparency — twice, in two consecutive budget cycles — and the content of those recommendations is the most telling thing in the record.

In April 2025, reviewing a benefit then under active consideration for elimination, the analyst recommended that the Legislature direct the department to report specified data annually, including utilisation and spending on optional drugs such as anti-obesity drugs, so that their fiscal impacts could be tracked. PROVED In March 2026, after the benefit was gone, the analyst repeated the theme and sharpened it: the Legislature should direct the administration to provide more detailed pharmacy data on users, utilisation, and net unit costs by therapeutic class and drug. PROVED

Both recommendations are sound on their own terms, and the second is genuinely useful — net unit cost by drug is the figure the public record most conspicuously lacks. But read them together and the frame is unmistakable. The nonpartisan office charged with telling the Legislature what its money is doing asked, twice, for better numbers about money. Clinical outcome did not appear in either list. INFERRED

Had every transparency recommendation been adopted in full, California would now have considerably better numbers about what the drugs cost, and still none about what they did.

That is why this article treats the gap as institutional rather than as a lapse by any official. Every component of the apparatus had a field for cost or coverage. Pharmacy claims generated utilisation counts. Allowed amounts generated spending. Rebate data supported gross-to-net analysis. Budget models generated savings projections. State-plan law generated a yes-or-no coverage decision. Pharmacy systems generated rejection codes. There was no corresponding public field for benefit, and no institution whose job it was to ask for one. INFERRED

VII. What could, and could not, have been measured from data California already holds

An honest account has to concede what a claims-based evaluation cannot deliver, because the concession is where the argument either earns its authority or loses it.

The hardest outcome is the most obvious one

A pharmacy transaction carries no body weight. A diagnosis claim may carry an obesity code without a measurement. Body-mass-index codes exist but are neither universally recorded nor a substitute for a series of measured weights. A rigorous analysis of percentage weight change therefore requires clinical records — electronic health record data held by managed-care plans and clinic systems, not by the payment layer. Building that linkage means data-use agreements, privacy safeguards and plan cooperation. It is genuinely difficult, and it is the one item on the list that could not have been produced from claims alone.

Three of the questions need no clinical record at all

Persistence is a refill-pattern calculation. How many members initiated therapy, how many filled again at three and six and twelve months, how many had extended gaps, how many switched products. Claims answer this directly. A persistence analysis would not prove clinical benefit; it would tell policymakers whether the state purchased sustained treatment or mostly brief exposure, which for a therapy whose effect depends on continuation is close to half the value question.

Newly recorded diabetes is a diagnosis-code question. Claims cannot prove progression to diabetes in the pathophysiological sense merely because a code appears. They can identify newly recorded diagnoses after a sufficiently long clean baseline, subject to continuous-enrolment requirements, coding accuracy, confounding and the gap between a recorded diagnosis and biological onset. The correct sentence is neither “claims can tell us who developed diabetes” nor “claims cannot measure outcomes”; both extremes are wrong. It is: incident diabetes coding can be studied longitudinally in claims, subject to stated limits.

Downstream utilisation is a claims-linkage question. Hospitalisations for myocardial infarction and stroke can be identified through validated claims algorithms, imperfectly. Emergency visits and admissions are plainly observable. So is total paid medical expenditure. A state evaluation could ask a practical fiscal question without any clinical measure at all: did the treated cohort’s non-pharmacy utilisation evolve differently from an appropriately constructed comparison cohort?

And the capability was demonstrated, on this exact population

California consolidated Medi-Cal pharmacy benefits into a single fee-for-service system under Medi-Cal Rx, so pharmacy exposure for the whole programme — including members otherwise enrolled in managed care — runs through one claims environment. PROVED The analyst notes that this makes pharmacy one of the better-documented parts of Medi-Cal even as it complicates year-over-year comparison.

More to the point, the state performed a considerably more elaborate exercise than any of the three analyses above in order to produce the cost figures it did publish: it constructed a diagnosis-excluded cohort across five products and six years, and mapped cost per member by county. PROVED

The analytical capability was demonstrated on this population, with this drug class, in this period. It was pointed at what the benefit cost rather than at what it did.

VIII. Why the answer matters in both directions

The temptation, having established a gap, is to treat it as proof that the termination was wrong. It is not, and the argument is stronger for saying so plainly.

Both readings are consistent with everything California has published. INFERRED That is the problem. A programme of this size should not be able to end without the public learning which description is true, and the state should not have to argue for its own decision using only the price tag.

The same discipline applies to the savings figure

The $790 million is a projected General Fund saving from eliminating covered drug use, later revised upward for the budget year. PROVED This review did not identify a public model presenting it as a comprehensive long-term cost-effectiveness estimate incorporating weight regain, incident diabetes, cardiovascular events and other downstream utilisation.

It would be equally wrong to assert that the projection assumed downstream costs to be zero. Without the underlying fiscal model, that cannot be established. NOT ESTABLISHED It is entirely possible for downstream medical costs to rise after a pharmacy cut and for the policy still to produce a substantial net saving. How much of the projection survives those effects is unknown — and it is precisely the question an outcomes evaluation would have informed.

And to the compounded-drug question

Loss of coverage for an expensive medicine can push demand somewhere else. This review identified no California dataset showing that Medi-Cal beneficiaries who lost coverage moved to compounded semaglutide or tirzepatide, and that causal proposition is therefore not asserted here. NOT ESTABLISHED

What is established is that the alternative market carries a materially different safety framework. The federal regulator states that compounded drugs are not FDA approved, meaning they do not undergo premarket review for safety, effectiveness and quality, and it has accumulated well over a thousand adverse-event reports involving compounded semaglutide and tirzepatide, while cautioning that a report does not establish causation and that totals are likely undercounts because many state-licensed pharmacies are not required to report. DOCUMENTED Our patient guide on compounded semaglutide and tirzepatide exists because of that dynamic.

The distinction between a plausible consequence and a measured consequence is exactly the discipline this article asks of the state. It would be incoherent to relax it here.

IX. The people who were mid-treatment on 31 December 2025

Every account of this decision has been fiscal. There is a clinical account too, and it has had almost no attention.

GLP-1 receptor agonists are not a course of antibiotics that is finished and put down. They suppress appetite and slow gastric emptying for as long as they are taken, and the physiological drivers of weight regain — lowered resting energy expenditure, altered satiety signalling, restored hunger — reassert themselves when they stop. The clinical literature on discontinuation is consistent about the direction of travel: a substantial share of lost weight returns, over months rather than years.

So on 31 December 2025, when existing authorisations expired, a population of Medi-Cal members who had been treated — some for years, some of whom had reached clinically meaningful reduction — reached the end of coverage without a taper. Some had a clinician who could document a covered indication. Some paid cash. Some appealed and continued pending a hearing. Some stopped. How many fell into each group is not published.

That matters fiscally as well as clinically, because a savings projection assumes the spending simply ceases. INFERRED If a share of those members regain weight and progress to diabetes, the state pays for that too — through the diabetes GLP-1 coverage federal law effectively requires it to maintain, and through the rest of the disease. The state has, in effect, chosen to pay later for defined disease states while declining to pay earlier for pharmacological weight management. That may be the right call. It is not a call anyone can currently evaluate.

The April 2026 restoration of Wegovy for noncirrhotic MASH makes the point concrete rather than theoretical. Weight status, dysglycaemia, diabetes, cardiovascular disease and metabolic liver disease are not isolated administrative boxes in a human body merely because a pharmacy claims system codes them separately. If the state wants to know whether paying for an earlier intervention prevents a later covered disease, it has to measure the relationship. The claims system will not answer it by accumulating invoices.

X. The legislative sequel, stated at its actual strength

In the same session in which the benefit ended, the Legislature sent the Governor a measure addressed to the same drug class from the opposite direction. Senate Bill 1089, the Preventive Treatment Health Care Act, was introduced on 13 February 2026 by Senator Laura Richardson and amended on 24 March 2026. DOCUMENTED

The amended March version carried a CalPERS coverage requirement: commencing 1 January 2027, a health benefit plan contracting with the CalPERS board would have had to offer optional coverage for chronic weight disease management, including at least one FDA-approved anti-obesity medication, as part of one of its plan options. The California Health Benefits Review Program analysed that version and estimated it would reach the health insurance of approximately 1,224,000 enrollees, or 3.2 per cent of Californians. DOCUMENTED The CalPERS board adopted an oppose position, arguing the bill would mandate coverage of high-cost medications without a cost-containment mechanism. DOCUMENTED Later amendments removed the CalPERS language. DOCUMENTED

What survived, and the difference between may and shall

The enrolled measure is substantially narrower than its history suggests, and the distinction between what it permits and what it requires is load-bearing.

It authorises the California Health and Human Services Agency — if needed, and subject to an appropriation by the Legislature — to enter partnerships to increase competition, lower prices and address supply shortages for at least one FDA-approved GLP-1 anti-obesity medication. Separately and more narrowly, it requires the agency, if needed, to establish distribution partners and to make its best effort to negotiate pricing at or below the cost to Medi-Cal beneficiaries in 2025. DOCUMENTED

Most of the machinery is therefore discretionary and appropriation-contingent, and even the mandatory clause retains an if needed qualifier and frames the pricing obligation as best effort rather than result. Reporting that the bill “directs the state” to build a low-cost GLP-1 pathway overstates a permissive grant. In a statute, may and shall are not stylistic variants.

The vehicle contemplated is CalRx, the programme that brought a state-branded insulin to market and now also distributes albuterol inhalers and naloxone. That effort took roughly three years, and a GLP-1 equivalent would likely be a multi-year endeavour. DOCUMENTED State health officials have said GLP-1s are among the drugs under consideration for the programme. DOCUMENTED

California determined that it could not afford this drug class at prevailing prices for its Medicaid population, and in the same session contemplated obtaining the same class more cheaply through state procurement. Both propositions can be true. Neither required knowing whether the drugs worked in the population the state had been treating.

The timing has a consequence that deserves stating. INFERRED On a three-year arc, the members who lost coverage on 1 January 2026 are not the members a CalRx product would serve. The interval between termination and any replacement is a period in which the affected population has neither the benefit nor its cheaper successor — and nobody is measuring it.

Falling prices make the missing evidence more valuable, not less

Suppose the state does eventually obtain substantially lower prices. At what price should Medi-Cal reconsider broader coverage? That is a fiscal question with a clinical denominator. How many patients would be treated, for how long, with what persistence, at what durable weight reduction, with how many diabetes diagnoses delayed, how many cardiovascular events changed, how much medical spending avoided, and in which subgroups?

Every one of those questions becomes more consequential as the price falls, because a lower price shifts the decision from we cannot afford this to is this worth what it now costs — and only the second question requires knowing what the therapy accomplishes. Without California outcomes data, the state will meet that moment with an increasingly precise local price on one side of the ledger and clinical-effect estimates imported from other populations on the other.

XI. The federal turn, which makes the timing worse

One development since the termination changes the stakes and appears in none of the earlier accounts of this policy.

The federal Centers for Medicare & Medicaid Services has launched a voluntary model, announced as the BALANCE Model, to expand access to GLP-1 medications, with a Medicaid launch as early as May 2026 and a Medicare Part D launch in January 2027. Ahead of it, CMS planned a Medicare GLP-1 payment demonstration beginning in July 2026 under which eligible beneficiaries would pay $50 for a month of therapy at prices negotiated federally. Participation is voluntary for manufacturers, states and plans. DOCUMENTED

Set that beside California’s position and the asymmetry sharpens. INFERRED A state deciding whether to participate in a voluntary federal model for this drug class needs to know what the drug class did for its own beneficiaries when it was paying for it. California will be making that decision with a well-documented account of what the therapy cost between 2018 and 2023, a revised savings estimate for having stopped, and no published account of what any of it achieved.

The relevant question for the next round is no longer what does this cost. It is at what price is it worth buying, and that is not a question a price can answer by itself.

XII. What this article does not claim

The claim is narrow, and it survives scrutiny because it is narrow. Set out explicitly, this article does not establish and does not assert:

• that every prescription in the state’s cohort represented obesity treatment — the cohort excludes diabetes-coded patients to raise the likelihood of weight-loss use; it does not adjudicate indication at patient level.
• that California government spent the combined Medi-Cal and commercial total — the commercial share was paid by commercial insurers and employer plans.
• that Medi-Cal’s net cost was $416.8 million — that figure is gross and pre-rebate within a defined claims cohort.
• that $416.8 million was General Fund expenditure — Medi-Cal financing includes federal participation and rebates, both varying by population.
• that every treated patient met an FDA-approved obesity indication — the methodology cannot answer that in either direction.
• that the 2026 coverage exclusion was unlawful — the state had a statutory pathway, filed a state plan amendment, and obtained federal approval.
• that GLP-1 drugs are clinically ineffective — the trial literature is substantial and is not in dispute here.
• that they are cost-effective for Medi-Cal — no California programme-specific analysis establishing that was identified.
• that the savings projection assumed downstream costs of zero — the underlying model does not permit that conclusion from the public material.
• that former beneficiaries migrated to compounded products — no California migration dataset was identified.
• that the department holds no internal analysis — the finding concerns the published record.

Nor is this an argument that obesity is a lifestyle condition undeserving of pharmacological treatment, or that beneficiaries who received these drugs did so improperly. None of the primary sources examined here makes any such claim, and this review adopts none of them. NOT ESTABLISHED

The positive claim is this. A government that spends at this scale on a discretionary clinical benefit, tracks the spending in granular administrative detail, and then withdraws the benefit for cost reasons owes an accounting — to the Legislature that appropriates the money and the public that bears it — of what the expenditure purchased in health. California has met the first obligation with unusual rigour. On the public record examined here, it has not met the second.

California’s GLP-1 benefit, by the record

Swipe the table sideways for the remaining columns

The final row is a finding about the public record. It is not an error state, and it is not a claim that the benefit failed.

Sources and what remains open

Figures were taken from the named California and federal publications rather than from secondary summaries.

Swipe the table sideways for the remaining columns

Reported, and not confirmed against a primary record for this page

Each item below appears in identifiable secondary reporting or in a source this page did not open directly. They are listed so a reader can see what carries a lighter grade and why, rather than finding them stated as fact: the specific adverse-event counts for compounded semaglutide and tirzepatide; the count of other state Medicaid programmes covering anti-obesity drugs, in both the August 2024 and October 2025 snapshots; the Medi-Cal adult obesity rate; the LAO’s Wegovy-specific estimate for 2023-24; the 40 to 50 per cent federal rebate range and the roughly 3 per cent supplemental rebate estimate; the share of pharmacy-spending growth attributable to GLP-1 agonists alone, as distinct from the verified cluster finding; the gross pharmacy projections of $19.4 billion and $20.6 billion; the CalPERS board’s premium projection and oppose position; the CHBRP enrollee estimate; the amendment removing the CalPERS provision; the enrolled text’s operative verbs; the presentment date and time; and the CalRx insulin timeline.

Figures deliberately excluded

• Any cumulative multi-year Medi-Cal GLP-1 total. None was located in verifiable form, and a plausible round number is not a substitute.
• Any net or General Fund figure specific to these drugs. Drug-level rebates are confidential; the figure does not exist publicly and applying an aggregate rebate percentage to a drug-specific gross number would manufacture it.
• Any interior-year values for 2019 through 2022. Drawing a curve through points this review did not extract is fabrication, not visualisation.
• Any average annual cost per member. Circulated in draft materials, not confirmed against the published output.
• Any statement that beneficiaries who lost coverage had achieved meaningful weight reduction. That is precisely the outcome this article establishes California did not publish; asserting it would be the article contradicting itself.
• Any claim that the termination caused migration to compounded products.

What would close this

Four things, in ascending order of difficulty. Three require no new data collection.

1. Extend the published series through 2024, 2025 and the post-termination months. The claims exist and the cohort definition is already built. This alone would establish what happened to utilisation after 1 January 2026.
2. Publish treatment persistence from the same claims. Refill patterns answer how long members actually stayed on therapy without touching a clinical record.
3. Link pharmacy to medical claims for newly recorded diabetes, cardiovascular events and downstream utilisation, with the confounding and attribution limits published alongside.
4. Report net cost after rebates by therapeutic class and drug — which is, in substance, what the Legislature’s own analyst has now asked for twice.

A fuller reconstruction would also need direct extraction of California rows from the federal State Drug Utilization Data files for 2024 and 2025, isolating the three weight-management brands by national drug code; a formal application to the state’s data-access programme for a diagnosis-linked file extending the methodology past 2023; and a public records request for any internal analysis that exists and has not been published.

None of that changes the central finding. Reconstructing later utilisation would sharpen the spending side of the ledger. It would not supply the missing half, because reimbursement data, however current, cannot say whether a beneficiary who received Wegovy in 2024 lost weight, kept it off, or avoided a diabetes diagnosis as a result. That is the accounting California has not published, for any year of the programme, and no volume of additional utilisation data closes it.

Written by Kanwar Partap Singh Gill, MD, family medicine physician in Fresno, California, USA · Approved for publication by Kanwar Partap Singh Gill, MD · 8 September 2026 · primary sources verified through 8 September 2026 · Editorial standards