One in eight US adults now takes a GLP-1 drug, and grocery spending in their homes falls 5.3%. An audit of the bodies, budgets and industries it reshapes.
The Scale of the GLP-1 Economy
One in eight adults, a trillion-dollar drugmaker and a falling obesity rate
One in eight American adults now takes a GLP-1 drug, and the adult obesity rate has fallen from 39.9% in 2022 to 37.0% in 2025 ✓ Established [1] [2]. The medicines work, the demand is real and the bills are arriving at once. This report audits what semaglutide and tirzepatide do to bodies, to the budgets that pay for them and to the industries that sold the calories they remove.
The core answer is simple and the details are not. GLP-1 receptor agonists — semaglutide, sold as Ozempic and Wegovy, and tirzepatide, sold as Mounjaro and Zepbound — produce average weight losses of 15% to 21% in trials, cut major cardiovascular events by 20% in people with established heart disease and slow kidney decline in diabetes ✓ Established [3] [4] [5] [6]. They also stop working when patients stop taking them, and most patients outside trials stop within one to three years ◈ Strong Evidence [7] [8]. That combination — a drug that is effective, chronic and expensive, prescribed to a population measured in hundreds of millions — is what makes the GLP-1 class an economic event rather than only a medical one. A November 2025 KFF poll of 1,350 adults found that 12% were currently taking a GLP-1 drug, up 6 points since May 2024, and 18% had ever taken one ✓ Established [1]. Gallup’s National Health and Well-Being Index put self-reported use for weight loss at 12.4% in 2025, more than double the 5.8% of February 2024 ✓ Established [2]. RAND, surveying 8,793 people in spring 2025, found that 11.8% had used one for weight loss and that about 20% of women aged 50 to 64 had done so ✓ Established [9]. Three independent surveys converge on the same order of magnitude.
The population-level signal is now visible. Gallup estimates that the decline from 39.9% to 37.0% means about 7.6 million fewer American adults with obesity, and the largest drop — 5.0 points, to 42.8% — came among adults aged 50 to 64, the group in which use for weight loss is highest at 17.0% ◈ Strong Evidence [2]. Correlation across a national survey is not proof of cause, and self-reported height and weight carry their own error, but no other intervention has moved the American obesity curve in the opposite direction in three decades of measurement. At the same time, diagnosed diabetes reached an all-time high of 13.8% in the same survey, a reminder that the drugs are treating a population whose underlying metabolic risk is still rising ✓ Established [2]. The scale of eligibility is larger still. ICER, the US cost-effectiveness reviewer, estimates that up to 40% of American adults could meet the clinical criteria for treatment ◈ Strong Evidence [10]. The World Health Organization counts more than one billion people with obesity worldwide and projects that GLP-1 medicines will reach fewer than 10% of those who could benefit by 2030 ✓ Established [11].
The money has followed the patients. Eli Lilly became the first pharmaceutical company worth $1 trillion in November 2025, after tirzepatide overtook Merck’s Keytruda as the world’s best-selling drug ✓ Established [12]. Analysts quoted by Reuters expect the obesity-drug market to reach about $150 billion a year by the early 2030s, and Goldman Sachs raised its 2030 forecast to $130 billion from $100 billion ◈ Strong Evidence [13]. Morgan Stanley’s more aggressive scenario has the market doubling to $190 billion by 2035, with about 55 million American users — 15% of the population — compared with its earlier estimate of 33 million ◈ Strong Evidence [14]. On the other side of the ledger, the World Obesity Federation projects that the economic cost of overweight and obesity will exceed $4.32 trillion a year by 2035, almost 3% of global GDP ◈ Strong Evidence [15]. The policy question is therefore not whether the drugs are worth something. It is who pays for a lifelong therapy, at what price, and what happens to everyone else in the economy when a tenth of the population eats less.
Four features distinguish this drug class from earlier blockbusters and explain why its effects spill beyond medicine:
- Breadth: eligibility is defined by body-mass index and common conditions, not by a rare diagnosis, so the addressable population is a large share of all adults [10] [11].
- Duration: the benefit lasts only while the drug is taken; stopping is followed by regain of most of the lost weight within a year [7] [16].
- Price: US list prices of $1,000 to $1,350 a month at launch made annual costs comparable to a used car, per patient, per year [17].
- Behavioural reach: the drugs suppress appetite and appear to dampen other cravings, changing what households buy and drink [18] [19].
This report follows those four features in turn. Sections 2 to 4 set out what the trial evidence shows about bodies — the outcomes beyond weight, the problem of discontinuation and the safety record. Sections 5 and 6 turn to budgets — the American price war, the Medicare experiment and the very different choices made by the United Kingdom, Denmark, Japan, India, China, Canada and Brazil. Section 7 measures the knock-on effects on food, drink, airlines, fitness, medical devices and life insurance. The last section weighs what the evidence supports and where the confident forecasts outrun the data. Throughout, figures are given with their source and date, and claims are graded: established fact, strong evidence or contested. The distinction matters more than usual here, because the GLP-1 story is told simultaneously by manufacturers with multi-billion-dollar stakes, by analysts selling forecasts, by regulators reviewing safety signals and by patients describing their own experience. Much of what circulates as settled — that the drugs cure addiction, that they will shrink the food industry, that they will pay for themselves — rests on evidence of very different quality, and the gaps between those claims and the trial record are where the policy mistakes are most likely to be made.
Beyond the Scale
What the outcome trials show about hearts, kidneys and cravings
The case for GLP-1 drugs no longer rests on weight alone: SELECT found a 20% reduction in heart attacks, strokes and cardiovascular deaths among 17,604 adults with heart disease and obesity ✓ Established [5]. Kidney, heart-failure and sleep-apnoea trials point the same way. Alzheimer’s disease did not.
The weight-loss trials established the headline numbers. In STEP 1, published in the New England Journal of Medicine in February 2021, 1,961 adults without diabetes lost an average of 14.9% of body weight on semaglutide 2.4 mg over 68 weeks, against 2.4% on placebo; 86.4% of those on the drug lost at least 5%, compared with 31.5% on placebo ✓ Established [3]. SURMOUNT-1 then showed tirzepatide producing average losses of 15.0%, 19.5% and 20.9% at its three doses over 72 weeks, against 3.1% on placebo, among 2,539 participants ✓ Established [4]. In a body-composition subset scanned by DEXA, fat mass fell by 33.9% on tirzepatide against 8.2% on placebo ✓ Established [4]. When the two drugs met head to head in SURMOUNT-5, an open-label trial of 751 adults sponsored by Lilly, tirzepatide produced 20.2% weight loss against 13.7% for semaglutide over 72 weeks, with nausea reported by about 44% of participants in both arms ✓ Established [20]. These are the largest average losses any medicine has achieved in randomised trials, and they are roughly half of what bariatric surgery delivers without an operation.
SELECT randomised 17,604 adults with established cardiovascular disease and overweight or obesity but no diabetes, and followed them for a mean of 40 months [5]. The composite of cardiovascular death, non-fatal heart attack or non-fatal stroke occurred in 6.5% of the semaglutide group against 8.0% on placebo, a hazard ratio of 0.80; all-cause death was 4.3% against 5.2% [5]. A prespecified Lancet analysis in October 2025 found the benefit held regardless of baseline adiposity or the amount of weight lost, with only about 33% of it explained by waist reduction [21].
SELECT matters for two reasons. It turned an obesity drug into a cardiovascular drug, which is what moved insurers and regulators from treating weight loss as lifestyle to treating it as risk reduction, and it suggested that the benefit is not simply a by-product of weight lost ◈ Strong Evidence [21]. If roughly two-thirds of the cardiovascular benefit runs through mechanisms other than weight — inflammation, blood pressure, direct effects on blood vessels — then the drug is doing something that diet alone would not replicate. That interpretation remains an inference from a mediation analysis, not a demonstrated mechanism, and the trial population was specific: adults who had already had a heart attack, stroke or symptomatic arterial disease ◈ Strong Evidence [5] [21]. The 20% figure cannot be transferred to a healthy 35-year-old with a BMI of 31, for whom absolute event rates are far lower and the absolute benefit correspondingly smaller. The kidney evidence is similarly specific. FLOW enrolled 3,533 people with type 2 diabetes and chronic kidney disease and stopped early for efficacy after a median of 3.4 years: major kidney events — kidney failure, a large fall in filtration or death from kidney or cardiovascular causes — were 24% lower on semaglutide, 331 events against 410 ✓ Established [6].
The outcome trials now cover most of the conditions that make obesity costly:
- SELECT (cardiovascular): 20% fewer major adverse cardiovascular events over 40 months in 17,604 adults with heart disease [5].
- FLOW (kidney): 24% fewer major kidney events in 3,533 adults with type 2 diabetes and kidney disease [6].
- STEP-HFpEF (heart failure): symptom scores improved by 16.6 points against 8.7 on placebo, with one adjudicated heart-failure event against twelve [22].
- SUMMIT (heart failure, tirzepatide): cardiovascular death or worsening heart failure down 38% and hospitalisation down 56% in 731 patients [23].
- SURMOUNT-OSA (sleep apnoea): breathing interruptions cut by up to 62.8%, leading to the first drug approval for obstructive sleep apnoea [24].
The heart-failure trials target a condition with few effective treatments. STEP-HFpEF tested semaglutide in heart failure with preserved ejection fraction, a form closely tied to obesity, and found that the Kansas City Cardiomyopathy Questionnaire score improved by 16.6 points on the drug against 8.7 on placebo, while weight fell by 13.3% against 2.6% ✓ Established [22]. SUMMIT, published in November 2024, followed 731 patients with a median BMI of 38 and found the composite of cardiovascular death or worsening heart failure 38% lower on tirzepatide, with heart-failure hospitalisation down 56% and C-reactive protein, a marker of inflammation, down 34.9% ✓ Established [23]. SURMOUNT-OSA found that tirzepatide reduced the apnoea-hypopnoea index by 27.4 events per hour against 4.8 on placebo after 52 weeks ✓ Established [24]. Each of these results expands the list of indications under which insurers can be asked to pay, and each represents a cost that the health system would otherwise bear through hospital admissions, devices or dialysis.
The most commercially intriguing signal is in addiction, and it is also the least settled. A phase 2 randomised trial led by Christian Hendershot, published in JAMA Psychiatry in February 2025, gave 48 adults with alcohol use disorder low-dose semaglutide or placebo for nine weeks ✓ Established [19]. Semaglutide reduced alcohol consumed in a laboratory session, drinks per drinking day and craving, and among smokers it reduced cigarettes per day ◈ Strong Evidence [19]. By the second month, drinking-day consumption had fallen by about 30% on the drug against 2% on placebo ◈ Strong Evidence [25]. A Swedish nationwide cohort of 227,866 people with alcohol use disorder, followed from 2006 to 2023, found semaglutide users had 36% fewer alcohol-related hospitalisations ◈ Strong Evidence [26]. Neither study is definitive: the trial was small and short, the cohort observational. But the pattern fits what users report and what the alcohol industry is now measuring in its own sales, discussed in Section 7.
The limit of the class was demonstrated in November 2025. Novo Nordisk’s evoke and evoke+ trials gave oral semaglutide 14 mg or placebo to 3,808 people aged 55 to 85 with early Alzheimer’s disease for 104 weeks, and found no significant effect on the primary measure of cognitive and functional decline ✓ Established [27]. Some biomarkers improved, but the planned extension was terminated ✓ Established [27]. The failure matters beyond neurology. Observational data had suggested lower dementia rates among GLP-1 users, and that signal did not survive a randomised test — a caution that applies equally to the other observational benefits now attached to the drugs, from addiction to cancer ◈ Strong Evidence [27] [26]. The randomised evidence for cardiovascular, kidney, heart-failure and sleep-apnoea outcomes is strong. The rest is a set of hypotheses being tested in public, often by the companies that sell the product.
The Stopping Problem
Why a chronic drug meets an acute budget
In the STEP 1 extension, participants who stopped semaglutide regained about two-thirds of the weight they had lost within a year, and their cardiometabolic gains faded ✓ Established [7]. Outside trials, only 8.1% of commercially insured patients were still on a GLP-1 for obesity three years after starting ◈ Strong Evidence [8].
The trial evidence on stopping is unusually clean, because the companies designed withdrawal studies to prove that patients should stay on the drug. In the extension of STEP 1, participants who had lost 17.3% of body weight by week 68 were taken off semaglutide and followed for another year; by week 120 they had regained 11.6 percentage points, leaving a net loss of 5.6% ✓ Established [7]. Improvements in blood pressure, cholesterol and blood sugar reverted towards baseline over the same period ✓ Established [7]. SURMOUNT-4 ran the same test for tirzepatide. After a 36-week lead-in in which participants lost 20.9%, those switched to placebo regained 14% over the following 52 weeks, while those who continued lost a further 6.7%; over 88 weeks the total loss was 25.3% for those who stayed on the drug against 9.9% for those who stopped ✓ Established [16]. The pharmacology is not ambiguous. GLP-1 drugs suppress appetite while they are in the bloodstream, and the body defends its previous weight once they are gone.
STEP 1’s extension found that two-thirds of the weight lost on semaglutide returned within a year of stopping, with cardiometabolic benefits fading in parallel [7]. SURMOUNT-4 found that switching from tirzepatide to placebo after 36 weeks led to a 14% regain over 52 weeks, against a further 6.7% loss for those who continued [16]. Both trials were designed by the manufacturers to show the case for continuous treatment, and both did.
Real-world persistence is far lower than trial persistence. Blue Health Intelligence, analysing claims for about 170,000 Blue Cross Blue Shield members in 2024, found that 58% of patients prescribed a GLP-1 for weight loss stopped before 12 weeks — the point at which clinically meaningful weight loss typically begins — and that more than 30% had stopped by four weeks ◈ Strong Evidence [28]. In the trials, discontinuation ranged from 2% to 14% ◈ Strong Evidence [28]. Prime Therapeutics, a pharmacy-benefit manager, tracked commercially insured patients and found that only 8.1% of those who started a GLP-1 for obesity were still taking one three years later, rising to 14% for the newer high-potency products such as Wegovy ◈ Strong Evidence [8]. Prime’s data also show the picture improving: one-year persistence rose from 33% for patients who started in 2021 to 63% for those who started in the first quarter of 2024, as shortages eased and patients switched between products rather than abandoning treatment ◈ Strong Evidence [8]. About 38% of persistent patients had switched at least once ◈ Strong Evidence [8].
The reasons patients stop are mostly economic and gastrointestinal. KFF found that 56% of current users said the drugs were difficult to afford, and that 27% of insured users had paid the full cost themselves ✓ Established [1]. RAND found that about half of users reported nausea and about a third diarrhoea ✓ Established [9]. Shortages in 2022 to 2024 forced many to interrupt treatment, and employer coverage changes — discussed in Section 5 — have forced others off ◈ Strong Evidence [8] [29]. The Cornell study of household grocery purchases, discussed in Section 7, found that about a third of users stopped within its observation window, and that their food spending then returned to pre-treatment levels ✓ Established [18]. Discontinuation therefore shapes not only health outcomes but also the size of every downstream effect, from food demand to insurance mortality tables.
We don’t have any evidence that these drugs can be taken for a period of time to lose weight and then stopped.
— David Rind, Chief Medical Officer, Institute for Clinical and Economic Review, September 2025 [10]The economic consequence is that the most common pattern of use — start, lose weight, stop within a year, regain — is close to the worst value for money a payer can buy. A payer that funds twelve months of treatment pays most of the cost of the benefit and keeps little of it. ICER’s cost-effectiveness model assumes continuous long-term treatment; under that assumption it rates tirzepatide at about $53,400 and semaglutide at about $61,400 per quality-adjusted life year, both within conventional thresholds ◈ Strong Evidence [10]. Under real-world persistence, the same spending buys far fewer health gains, because many patients are paying for the expensive first months and leaving before the cardiovascular and kidney benefits accrue ◈ Strong Evidence [10] [28]. This is the stopping problem in economic form: a drug whose value depends on lifelong use is being purchased in a system built around annual budgets, annual plan choices and frequent changes of insurer.
American employers and insurers face a timing mismatch that the trials do not capture. They pay for GLP-1 treatment now, while the averted heart attacks, dialysis and knee replacements arrive a decade later — often after the patient has changed jobs, plans or moved into Medicare. Each payer rationally underinvests in a benefit that will accrue to the next one, which is why coverage is retreating even as the outcome evidence strengthens [29] [30].
There is a partial answer in maintenance dosing and in cheaper formats. Lilly and Novo Nordisk are both testing lower or less frequent doses to hold weight after the initial loss, and oral formulations — Novo’s 25 mg semaglutide pill, approved in December 2025, and Lilly’s orforglipron, approved as Foundayo in April 2026 — are priced from $149 a month for self-paying patients ✓ Established [31] [32]. If a cheap pill can maintain what an expensive injection achieved, the economics of long-term use change. That remains a hypothesis: no completed trial has yet shown that switching to a lower-cost oral maintenance regimen preserves the weight loss and the cardiovascular benefit over years ◈ Strong Evidence [33]. Until such data exist, every payer’s decision about coverage is in effect a decision about how long patients will stay on the drug, and the history so far suggests most of them will not.
The Safety Ledger
Gastrointestinal harms, a rare eye signal and the muscle question
Gastrointestinal side effects are the most common reason for stopping, and regulators now list a rare optic-nerve condition, NAION, as a very rare side effect of semaglutide ✓ Established [34]. A long-feared link to suicidal thinking was examined by European and American regulators and not supported by the evidence ✓ Established [35].
The common harms are well characterised and mostly transient. In OASIS 4, the trial behind Novo Nordisk’s 25 mg oral semaglutide, 74.0% of participants on the drug reported gastrointestinal adverse events against 42.2% on placebo, mainly nausea, vomiting, diarrhoea and constipation, concentrated during dose escalation ✓ Established [36]. In Lilly’s ATTAIN-1 trial of orforglipron, discontinuation because of adverse events rose with dose, from 5.3% to 10.3%, against 2.7% on placebo ✓ Established [37]. The newest and most potent agents carry higher rates. In TRIUMPH-1, the first phase 3 trial of Lilly’s triple-agonist retatrutide, published in May 2026, 11.3% of participants at the top dose discontinued because of adverse events against 4.9% on placebo, and 12.5% reported dysaesthesia, an abnormal skin sensation not typical of the older drugs ✓ Established [38]. Each step up in efficacy has so far come with a step up in tolerability problems.
The more serious gastrointestinal risks remain debated. A study published in JAMA in October 2023, using insurance claims, compared GLP-1 users with patients taking an older weight-loss drug, bupropion-naltrexone, and found hazard ratios of 9.09 for pancreatitis, 4.22 for bowel obstruction and 3.67 for gastroparesis ⚖ Contested [39]. The study was small for the semaglutide arm — 613 patients — and the absolute risks were low, but it shifted the conversation from nausea to rarer events that require hospitalisation ⚖ Contested [39]. Randomised trials, which are better at establishing cause but too small to detect rare harms, have generally not reproduced large pancreatitis signals. The disagreement is less about whether these harms exist than about how often they occur in a broad population that includes people with fewer reasons to accept risk than the trial participants had.
In June 2025 the European Medicines Agency’s safety committee concluded that non-arteritic anterior ischaemic optic neuropathy, a sudden loss of vision caused by reduced blood flow to the optic nerve, is a very rare side effect of semaglutide, affecting up to 1 in 10,000 users [34]. Its review estimated about one extra case per 10,000 person-years in type 2 diabetes, and the product information now advises stopping the drug if NAION is confirmed [34]. The signal first emerged from a single-centre observational study in JAMA Ophthalmology in July 2024 [40].
The NAION history illustrates how safety evidence for a mass-market drug accumulates. The original study reported hazard ratios of 4.28 among patients with type 2 diabetes and 7.64 among patients with overweight or obesity ⚖ Contested [40]. An invited commentary in the same journal argued that the data showed temporal association rather than causation, and the authors themselves stated that their design did not inform a mechanism ⚖ Contested [40]. A year later, after pooling larger datasets, the European regulator concluded that the association was real but rare ✓ Established [34]. Both readings were partly right: the effect exists, and at about one case per 10,000 person-years it is small relative to the cardiovascular benefit in high-risk patients — but not trivial for a low-risk person using the drug for cosmetic weight loss.
Suicidality followed the opposite path. The European Medicines Agency opened a review of five GLP-1 drugs in July 2023 and concluded in April 2024 that the available evidence did not support a causal link with suicidal thoughts or actions; no change to the product information was required ✓ Established [35]. The US Food and Drug Administration reached a similar preliminary conclusion in January 2024, stating that its evaluation had not found evidence that the medicines cause suicidal thoughts or actions ✓ Established [35]. The episode is a useful reminder that spontaneous reports in a population of tens of millions will always generate alarming signals, and that most of them resolve when tested against proper denominators.
The open question is muscle. Weight lost on any regime includes lean tissue as well as fat, and a 2024 commentary in The Lancet Diabetes and Endocrinology warned that rapid drug-induced weight loss might remove more lean mass than conventional dieting, with no data yet on frailty or falls ⚖ Contested [41]. Samuel Klein of Washington University in St. Louis has argued the reverse, that fat falls proportionally more than lean mass and that the concern about frailty is not supported by any data ⚖ Contested [42]. Both camps agree on two points: lean mass measured by DEXA includes water and organ tissue, not only muscle, and older adults — the group most at risk from sarcopenia and also among the heaviest users — are underrepresented in the trials ◈ Strong Evidence [42] [41]. Resistance exercise and adequate protein are the standard advice, which is also why gym chains and protein-drink makers have started marketing directly to GLP-1 users. The practical consequence for health systems is that the drugs are not a stand-alone treatment. A prescription that comes without dietary guidance and strength training risks producing lighter but weaker older patients, and the trials that would show whether that risk materialises at scale — with falls, fractures and loss of independence as outcomes — have not yet been run ◈ Strong Evidence [41].
The trials established safety in people who had compelling reasons to accept risk — heart disease, diabetes, severe obesity. Mass-market use, including through telehealth and compounded copies, extends the drugs to people with much lower baseline risk, for whom a 1 in 10,000 eye event or a rare pancreatitis carries more weight against a smaller absolute benefit. The benefit-risk balance is not one number; it falls as eligibility widens [34] [39] [1].
The American Price War
List prices, a White House deal and employers heading for the exit
Medicare’s negotiated price for semaglutide, effective in 2027, is $274 a month — 71% below its 2024 list price of $959 ✓ Established [43]. A November 2025 deal set a $245 government price and a $50 copay, while a growing share of employers dropped coverage altogether ✓ Established [44] [29].
The United States pays the highest prices in the world for GLP-1 drugs and has the largest number of users, which makes it the market where the budget arithmetic is tested first. At launch, monthly list prices ran from about $1,000 to $1,350 for semaglutide and from about $1,080 for tirzepatide ✓ Established [17]. Medicare’s Part D programme spent $15.2 billion gross on semaglutide in 2024 for about 2.3 million users, before the drug was covered for weight loss at all ✓ Established [43]. In October 2024 the Congressional Budget Office estimated that authorising Medicare coverage of anti-obesity medicines would cost about $35 billion between 2026 and 2034, and that in 2034 alone federal spending of about $7 billion would be offset by only about $1 billion in savings from better health ◈ Strong Evidence [30]. The CBO’s conclusion — that the drugs would not pay for themselves within the budget window — has been the single most cited number in the American coverage debate.
Prices then fell faster than any model predicted. In November 2025 the Trump administration announced agreements with Eli Lilly and Novo Nordisk under which Medicare and Medicaid would pay $245 a month for injectable GLP-1s, Medicare beneficiaries would pay a $50 copay, and cash prices through a new federal portal, TrumpRx, would start at or below $350 and fall towards $245 within two years ✓ Established [44]. In the same month, the first round of Medicare price negotiation under the Inflation Reduction Act produced a negotiated semaglutide price of $274 a month, effective January 2027 — 71% below the 2024 list price — as part of a package of 15 drugs expected to save Medicare about $12 billion ✓ Established [43]. Lilly cut the price of Zepbound vials sold through its direct-to-consumer pharmacy to between $299 and $449 a month in December 2025 ✓ Established [45]. TrumpRx launched in February 2026, but in practice it redirected buyers to manufacturers’ own sites or to pharmacy coupons ◈ Strong Evidence [46].
The Medicare coverage experiment began on 1 July 2026. The Medicare GLP-1 Bridge, running until 31 December 2027, covers Wegovy injections and pills, Zepbound and Foundayo for beneficiaries who meet defined criteria — a BMI of 35 or more, or 30 or more with heart failure, hypertension or kidney disease, or 27 or more with prediabetes or a prior heart attack or stroke — at a $50 monthly copay ✓ Established [47]. About 3.8 million beneficiaries are estimated to be eligible ✓ Established [47]. The copay does not count towards the Part D deductible or the $2,100 annual out-of-pocket cap, a detail KFF’s Juliette Cubanski highlighted, and results from the pilot are meant to inform a decision on permanent coverage in 2028 ✓ Established [48] [47]. The design is a deliberate test: a fixed price, a limited eligible population and a defined end date, after which Congress will have real-world persistence and cost data rather than projections.
ICER’s 2025 review judged tirzepatide and semaglutide cost-effective at net prices of about $6,829 and $7,973 a year, but warned of major budget strain [10]. Applying its budget-impact threshold of about $880 million a year, it estimated that fewer than 1% of eligible adults could be treated before the threshold was exceeded, in a country where up to 40% of adults may meet the criteria [10]. Cost-effective per patient and affordable across a population are different tests, and the drugs pass only the first.
Employers, who insure about half of Americans, are moving in the opposite direction from Medicare. KFF’s 2025 Employer Health Benefits Survey found that 19% of large firms covered GLP-1s for weight loss, rising to 43% among firms with 5,000 or more workers, up from 28% in 2024 ✓ Established [49]. Mercer’s 2026 data, which measure a different sample, show the reverse trend: the share of large employers covering GLP-1s for obesity fell from 72% in 2025 to 60% in 2026, 77% named the drugs a top cost concern, and pharmacy costs reached about 25% of total health spending ⚖ Contested [50]. Nearly 30% of large employers had tightened controls such as prior authorisation or lifestyle-programme requirements, and 6% had dropped coverage in 2026 ◈ Strong Evidence [29]. The two surveys disagree on the level of coverage, but agree that the cost pressure is now the main driver of plan design.
Public employers led the retreat. North Carolina’s State Health Plan voted 4 to 3 to end coverage of GLP-1s for weight loss from 1 April 2024 for more than 700,000 members, after the drugs cost $102 million in 2023 — about 10% of its pharmacy spend — and were projected to cost $170 million a year ✓ Established [51]. In 2026, five Massachusetts payers including the state’s Group Insurance Commission and MassHealth dropped weight-loss coverage; commercial users of the drugs fell by about 52% ◈ Strong Evidence [52]. Each of these decisions cut costs within one budget year and pushed the patients either off the drug or onto the cash market, where direct-to-consumer prices are now low enough that some can afford them and many cannot. The result is a two-track system: subsidised access for some Medicare beneficiaries, cash access for the better-off and declining employer coverage for everyone else.
The cash market itself was reshaped by the end of the shortage. When semaglutide and tirzepatide were on the FDA’s shortage list from 2022, compounding pharmacies were allowed to sell copies, and telehealth companies built businesses on them. The FDA declared the semaglutide shortage resolved on 21 February 2025 and set wind-down deadlines of 22 April for state-licensed pharmacies and 22 May for larger outsourcing facilities ✓ Established [53]. Novo Nordisk sued Hims and Hers in February 2026, alleging patent infringement and citing its own tests of compounded products that found impurities of up to 86% ◈ Strong Evidence [54]. The case was dropped within a month, and by late March the two companies had announced a partnership under which Hims would sell Novo’s approved Wegovy pill and injections from $149 a month ✓ Established [55]. The fight over copies ended with the brand-owner selling through the copier’s channel at a copy’s price.
Eight Countries, Eight Bargains
How health systems ration, pay for and copy the same molecules
Semaglutide’s core patents expired in India and China on 20 March 2026, and Indian generic vials went on sale at about $14 a month ✓ Established [56] [57]. The same molecule costs $274 to Medicare, is rationed by queue in England and pays for a measurable share of Denmark’s economic growth.
The United Kingdom chose rationing by phased rollout. In December 2024 NICE set out a 12-year plan to give tirzepatide to people with the highest clinical need through primary care, starting with about 220,000 patients in the first three years at a cost of about £317 million a year ✓ Established [58]. The plan made the National Health Service one of the first public systems to commit to large-scale obesity-drug provision, and also one of the slowest in practice. Most British users buy privately. When Lilly raised the UK private list price of the top Mounjaro dose from £122 to £330 a month from 1 September 2025, an increase of up to 170%, private buyers — about 90% of UK users — bore it, while NHS patients were protected by the negotiated price ◈ Strong Evidence [59]. Britain’s experience shows the price gap that opens when a public system negotiates hard and a private market absorbs the manufacturer’s response.
Denmark has experienced the GLP-1 economy as a macroeconomic force. Novo Nordisk’s growth made it, for a period, the most valuable company in Europe, and its fortunes moved Danish national accounts: Danish GDP fell 0.5% in the first quarter of 2025 as Novo’s growth slowed, with its shares down 55% from their July 2024 peak ◈ Strong Evidence [60]. In September 2025 Novo announced 9,000 job cuts, about 5,000 of them in Denmark, out of 78,400 employees, aiming for annual savings of $1.25 billion ✓ Established [61]. In February 2026 it guided to a 5% to 13% fall in 2026 sales at constant exchange rates, the first decline in more than a decade, and its shares fell about 17% ✓ Established [62]. Yet by September 2026 the Danish central bank had raised its 2026 growth forecast to 4% from 1.8%, citing American demand for the drugs ◈ Strong Evidence [63]. A country of six million has become a leveraged bet on one drug class.
Japan chose strict clinical gatekeeping. Wegovy was approved for obesity in March 2023 and listed for national health insurance reimbursement in November 2023, but only for patients with a BMI of 35 or more, or 27 or more with at least two obesity-related conditions, after diet and exercise had failed, with the 2.4 mg dose priced at ¥10,740 ✓ Established [64]. Tirzepatide for obesity, sold as Zepbound, was approved in December 2024 and launched in April 2025 ✓ Established [64]. Japan’s low obesity rate and its insurance rules mean that the drugs are reimbursed for a narrow clinical population, while cosmetic off-label use of diabetes products through private clinics has been the main regulatory worry. The contrast with the United States is structural: Japan sets one national price and limits who qualifies; America lets thousands of payers set their own rules around very high list prices.
India and China are where the price falls fastest. In India, Natco launched semaglutide vials at ₹1,290 a month — about $14 — when the patent expired, against about ₹10,480 for Novo’s Wegovy pen; more than 40 companies and 50 brands were expected to enter ✓ Established [57] [56]. In China, at least ten companies, including Hangzhou Jiuyuan, Huadong, Livzon and Qilu, filed for approval of semaglutide copies around the March 2026 expiry ◈ Strong Evidence [65]. Online prices for small doses fell below 300 yuan, more than 65% lower in some cases, and analysts warned of a price war that would compress margins for domestic and foreign producers alike ◈ Strong Evidence [66]. Canada became the first G7 country to approve a generic semaglutide in April 2026, from Dr. Reddy’s, with generics expected at 45% to 90% below brand prices, and Brazil’s EMS launched its own version at about 30% below Ozempic in May 2026 ✓ Established [67] [68].
The World Health Organization has moved to legitimise the class globally. In September 2025 it added semaglutide, dulaglutide, liraglutide and tirzepatide to its Essential Medicines List for type 2 diabetes with cardiovascular disease, kidney disease or obesity ✓ Established [69]. In December 2025 it issued its first guideline on GLP-1 medicines for obesity, with conditional recommendations, noting that more than a billion people live with obesity and that the condition was linked to 3.7 million deaths in 2024 ✓ Established [11]. Dr Tedros Adhanom Ghebreyesus, its Director-General, described obesity as a major global health challenge that WHO was committed to addressing effectively and equitably ✓ Established [11]. The essential-medicines listing matters most where it interacts with patent expiry: it gives low- and middle-income health systems a reason to procure generics that, in India, already cost a fraction of the American price.
Within a year of patent expiry in India, China, Canada and Brazil, the global floor price for semaglutide has fallen to roughly 1% of the American list price [57] [17]. Patents in the United States, Europe and Japan run longer, but the existence of cheap, approved generics in G7 Canada changes every future price negotiation. The relevant comparison for payers is shifting from the brand’s list price to the generic’s cost of production [67].
The Knock-On Economy
Food, drink, airlines, gyms, devices and life insurance
Households that start a GLP-1 drug cut grocery spending by 5.3% within six months, savoury snacks by about 10% and fast food by about 8% ✓ Established [18]. Whether that becomes an industry-wide contraction depends on how many people stay on the drugs, and the evidence on that is contested ⚖ Contested [70].
The best evidence on food comes from purchase data rather than surveys. Sylvia Hristakeva and Jura Liaukonyte of Cornell University, publishing in the Journal of Marketing Research in December 2025, matched GLP-1 adoption to the transaction records of about 150,000 households in a Numerator consumer panel ✓ Established [18]. Grocery spending fell by 5.3% within six months of starting, and by more than 8% in higher-income households; spending on savoury snacks fell by about 10%, and spending at fast-food restaurants, coffee shops and other limited-service outlets by about 8% ✓ Established [18]. Household adoption rose from about 11% in late 2023 to more than 16% by mid-2024 in the same panel ✓ Established [18]. Circana, tracking a larger panel, reported in November 2026 that 37% of US households now reported at least one user and that these households bought 32% of all food and beverage units, a share it projects to reach 40% by 2030 ◈ Strong Evidence [71].
The data show clear changes in food spending following adoption. After discontinuation, the effects become smaller and harder to distinguish from pre-adoption spending patterns.
— Sylvia Hristakeva, Assistant Professor, Cornell SC Johnson College of Business, December 2025 [18]The food industry has responded with products, write-downs and price cuts. Nestlé launched Vital Pursuit, a range of 12 frozen meals aimed at GLP-1 users and priced at $4.99 or less, in 2024 ✓ Established [72]. Conagra began labelling 26 Healthy Choice products as GLP-1 Friendly in January 2025 without changing the recipes ✓ Established [73]. J.M. Smucker wrote down the value of Hostess, the snack-cake maker it had bought for $5.6 billion, by about $1 billion in early 2025, with sweet baked snack sales down 7% in the quarter ✓ Established [74]. PepsiCo cut prices on Lay’s, Doritos and Cheetos by up to 15% in 2026 and still reported flat North American snack sales ◈ Strong Evidence [75]. Restaurant chains have redesigned menus: Olive Garden introduced lighter-portion dishes in December 2025, and Subway launched high-protein snacks in January 2026, with Subway’s North American president naming GLP-1 use as a key driver ✓ Established [76]. Morgan Stanley’s base case projects that by 2035 American consumption of soft drinks, baked goods and salty snacks will fall by about 3% and ice cream by 5.3% ◈ Strong Evidence [14].
The counter-evidence is substantial and should not be ignored. Circana’s David Portalatin has argued that although trial rates are high, active 30-day use is a single-digit share of households, and that as a dieting cohort GLP-1 users are slightly larger than keto and slightly smaller than intermittent fasting ⚖ Contested [70]. USDA data presented in February 2026 showed snack unit sales holding at about 10 billion a year in 2024, comparable with 2020, and the department’s chief economist, Justin Benavidez, said the long-term trend towards less food consumption was not yet visible in the data ◈ Strong Evidence [77]. Circana’s own figures describe reallocation more than shrinkage: GLP-1 households started from a spending index of 108, above average, and their food and beverage spending fell 2.5 points in the first year and 1.9 points in the second, while spending on refrigerated weight-control drinks and protein powders surged ◈ Strong Evidence [71]. The more plausible reading is that the drugs move money between categories within a food budget that shrinks only modestly.
Alcohol shows the clearest change outside food. A Drinkaware study published in October 2026 found that GLP-1 users in Britain reduced drinking days from 3.1 to 2.2 a week, a 29% fall, cut weekly consumption by 16% and spent 22% less on alcohol at home ◈ Strong Evidence [78]. Heineken advised UK pubs to promote half-pints and low- and no-alcohol drinks and said it would monitor how behaviour evolved ✓ Established [78]. Circana’s American data show the alcohol spending index of GLP-1 households falling 6.5 points in the first year and 10.2 in the second, larger declines than for food ◈ Strong Evidence [71]. That fits the trial and cohort evidence on craving discussed in Section 2 [19] [26]. For brewers and distillers already facing generational declines in drinking, the drugs are an accelerant rather than a cause.
The effects on other sectors range from measurable to speculative:
- Bariatric surgery: procedures fell 46.4% between 2022 and the third quarter of 2025 in Epic Cosmos records covering 31.68 million eligible adults, while 24.17% had a semaglutide or tirzepatide prescription [79].
- Sleep-apnoea devices: Inspire Medical cut its 2025 revenue guidance and its shares fell 32.4% after tirzepatide’s approval for sleep apnoea, while ResMed reported that GLP-1 users were 10% to 11% more likely to start CPAP therapy [80] [81].
- Weight-loss services: WeightWatchers filed for Chapter 11 in May 2025, erasing $1.15 billion of debt, after its diet-programme model lost ground to prescription drugs [82].
- Gyms: Planet Fitness says about half of GLP-1 users consider a gym membership and has shifted marketing towards strength training to preserve muscle [83].
- Airlines: Jefferies estimated in 2023 that if average passenger weight fell by 10 lb, United Airlines would save about $80 million a year in fuel — an analyst scenario, not an observed saving [84].
- Clothing: Impact Analytics estimates 400 million clothing units mismatched to demand by 2027 as size curves shift downwards [85].
Life insurers have the longest horizon and the most cautious models. Swiss Re estimated in 2025 that GLP-1 drugs could reduce cumulative US all-cause mortality by up to 6.4% by 2045 in an optimistic scenario, by 4% in its baseline and by 2.3% in a pessimistic one ⚖ Contested [86]. RGA’s central scenario is a 3.5% reduction over 20 years, while Munich Re attributes only 0.2% to 0.5% of future annual mortality improvement to the drugs, stressing that about two-thirds of patients stop within the first year ⚖ Contested [87]. The range is wide because the inputs — persistence, the population that gains access and the size of the benefit in lower-risk users — are the same uncertainties discussed in Sections 3 and 4. At the macroeconomic level, Goldman Sachs estimated in 2024 that US GDP could be up to 1% higher if 60 million Americans were taking the drugs by 2028, noting that obesity reduces output per person by more than 3% ⚖ Contested [88].
The Case That the Disruption Is Real
Cornell’s panel of 150,000 households recorded 5.3% lower grocery spending and about 10% lower snack spending after adoption [18].
Smucker’s $1bn Hostess impairment and PepsiCo’s 15% price cuts are management decisions, not forecasts [74] [75].
Circana finds GLP-1 households buy 32% of food and beverage units, heading to 40% by 2030 [71].
The Case That It Is Overstated
Circana’s Portalatin puts active 30-day use at a single-digit share of households [70].
USDA data show snack units steady at about 10 billion a year, comparable with 2020 [77].
Cornell found that households returned to pre-drug spending after discontinuation [18].
Only 8.1% of insured patients remain on treatment at three years [8].
Protein drinks and weight-control products are growing as indulgent categories decline [71].
What the Evidence Tells Us
A lifelong drug in a short-term economy
The randomised evidence that GLP-1 drugs reduce heart, kidney and heart-failure events in high-risk patients is strong ✓ Established [5] [6] [23]. The economic case depends on two things the trials did not settle: how long people stay on the drugs, and what they will cost once generics arrive in the largest markets ◈ Strong Evidence [8] [67].
Three findings are now settled enough to plan around. First, the drugs produce large and durable weight loss while they are taken, and the outcome trials in cardiovascular disease, kidney disease, heart failure and sleep apnoea show benefits that justify treatment in high-risk patients ✓ Established [5] [6] [22] [24]. Second, stopping reverses most of the effect, so the drugs behave like statins or blood-pressure medicines rather than like a course of antibiotics ✓ Established [7] [16]. Third, prices are falling steeply — through negotiation in the United States, through generics in India, China, Canada and Brazil, and through cheaper oral formulations — and the price at which the drugs pay for themselves is being approached from above ✓ Established [43] [57] [32]. Each of these findings is supported by randomised trials, regulatory decisions or published prices, not by forecasts.
The unsettled questions are mostly about behaviour and time. Real-world persistence is improving but remains far below trial persistence, and the gap determines whether payers buy long-term health gains or short-term weight loss that is regained ◈ Strong Evidence [8] [28]. The observational signals for addiction, dementia and other conditions have produced one prominent failure in Alzheimer’s disease and one promising but small trial in alcohol use disorder ◈ Strong Evidence [27] [19]. The muscle question and the rare-harm questions will be answered only by longer follow-up in older and lower-risk users ⚖ Contested [42] [41] [34]. The industry effects depend on all of these: a food company forecasting its 2035 volumes is, in effect, forecasting GLP-1 persistence and price. So is a life insurer pricing a 20-year policy, and so is a finance ministry estimating the long-run cost of Medicare. The breadth of those dependencies is the strongest argument for treating the next three years of real-world data — from the Medicare Bridge, the NICE rollout and the generic markets — as the evidence base that the original trials could not provide ◈ Strong Evidence [47] [58].
| Risk | Severity | Assessment |
|---|---|---|
| Coverage retreats faster than prices fall | Employers and public plans dropping coverage push patients off the drug before benefits accrue, wasting the spending already made [29] [52]. | |
| Low persistence erodes the health return | With 58% stopping before 12 weeks and 8.1% persistent at three years, most spending buys regained weight [28] [8]. | |
| Rare harms scale with mass use | NAION, pancreatitis and lean-mass loss matter more as eligibility widens to lower-risk users [34] [39]. | |
| Generic price war in emerging markets | Indian and Chinese price collapses compress margins and may outpace regulatory oversight of quality [66] [56]. | |
| Overstated industry forecasts | Food and insurance forecasts built on high adoption may misallocate capital if active use stays low [70] [87]. |
For payers, the evidence points towards targeted, durable coverage rather than either of the current extremes. Paying for everyone who qualifies at American list prices is unaffordable on ICER’s arithmetic, and dropping coverage altogether forfeits the benefit already demonstrated in high-risk groups ◈ Strong Evidence [10] [51]. The Medicare Bridge — a fixed $245 price, defined clinical criteria, a modest copay and an end date that forces evaluation — is the closest any large American payer has come to a deliberate experiment, and its persistence data will matter more than any model ✓ Established [47] [44]. England’s phased NICE rollout and Japan’s strict eligibility rules offer the same logic in different forms: treat first the patients for whom the trials showed the largest absolute benefit, and widen access as the price falls ✓ Established [58] [64].
For industries, the lesson of the first three years is that the effect is real but smaller and slower than the most dramatic forecasts suggested. Grocery spending falls by about 5% in a user household, not by half; alcohol falls further; and most of the change reverses when people stop ✓ Established [18] [78]. The companies that have responded best — protein-focused product lines, smaller portions, telehealth partnerships — have treated GLP-1 users as a new consumer segment rather than as an existential threat ◈ Strong Evidence [72] [76] [55]. Lilly’s pledge to invest more than $50 billion in American manufacturing and Novo’s $16.5 billion purchase of Catalent, the contract manufacturer, show the producers themselves betting on volume rather than on price ✓ Established [89] [90]. The companies that have struggled, such as WeightWatchers and Inspire Medical, sold a product that the drug directly replaced ✓ Established [82] [80].
Every major forecast in this report — Medicare’s net cost, insurers’ mortality tables, food companies’ volumes, the obesity rate itself — rises or falls with one number: the share of patients who stay on treatment for years rather than months. Persistence rose from 33% to 63% at one year between 2021 and 2024 cohorts [8]. If cheaper pills and generics push it higher, the GLP-1 economy becomes the largest change in population health economics since statins; if not, it becomes an expensive cycle of loss and regain [33] [7].
The GLP-1 economy is, in the end, a test of whether health systems built for short-term budgets can pay for long-term prevention. The drugs have passed the clinical test in the patients who need them most ✓ Established [5] [6]. The economic test is still running — in the Medicare Bridge, in NICE’s rollout, in Indian pharmacies selling $14 vials and in the grocery receipts of tens of millions of households ◈ Strong Evidence [47] [58] [57] [71]. The answer will not come from forecasts. It will come from the persistence data of the next three years, and from whether the price falls far enough, fast enough, to make lifelong treatment something more than a privilege of the insured and the well-off ◈ Strong Evidence [1] [11]. Until then, the most effective medicines ever developed for obesity will reach a minority of the people who could benefit, and the knock-on economy built around them will remain an American story with a Danish balance sheet and an Indian price list.