The GLP-1 ROI Question
The only published real-world analysis found no medical cost offsets — costs actually increased $1,338 per member. How the SELECT trial counterargument falls apart, and what honest ROI modeling looks like.
Last updated: April 2026
Context
The question everyone asks, nobody can answer
The GLP-1 ROI question has a simple formulation — do the downstream medical savings from weight loss offset the pharmacy spend? — and a complicated answer. The best available evidence says no, but with important caveats about time horizon, population selection, and the rapidly changing cost landscape.
The distinction between "the data shows no ROI" and "no data exists to show ROI" matters enormously here. For obesity-only commercially insured populations, the data actively shows negative ROI at current prices. For mixed diabetic and obese populations, the picture is slightly better — one large analysis shows medical cost growth bending favorably, though total costs still increase. Two critical structural gaps remain: no published model adjusts ROI projections for real-world persistence (page 3), and no analysis formally models the employer-tenure/time-to-offset mismatch.
This page compiles every published data point on both sides. The goal is not to reach a verdict — it's to give benefits leaders the complete evidence base, including the counterarguments, so they can make an informed decision for their specific population.
The core question
Medical cost-offset evidence
Does GLP-1 spending reduce downstream medical costs enough to offset drug spend? The independent studies say no. The manufacturer-funded studies show medical cost reductions but exclude drug costs from the calculation.
Independent analyses (not manufacturer-funded)
| Study | Population | N | Time horizon | Key finding | Net direction |
|---|---|---|---|---|---|
| Prime Therapeutics | Commercial, obesity without T2D | 3,346 vs. 8,343 | 2 years | Yr 1: TCC +$6,994/user. Yr 2: +$4,206. Medical costs +$1,338 (not significant). No reduction in obesity-related events. | No offset |
| Aon Phase 2 | Commercial, all indications (70% T2D, 30% weight loss) | 192,000 matched | 18–30 months | Medical cost growth rate slowed 3–6 ppts vs. controls. At ≥80% adherence: 7–9 ppts. Total costs still higher with drug spend. | Growth rate only |
| AssuredPartners | Commercial, diabetics | 9,236 vs. 10,625 | 5 years | Medical savings ~$560/yr. Drug cost ~$6,540/yr. | Net negative |
| UnitedHealthcare | Commercial, metabolic conditions | Not disclosed | 12 months | 91% increase in PMPM costs 12 months after initiation. | Cost increase |
| EBRI | Commercial employer simulation | 49.3M modeled | Annual | Premiums up 5.3–13.8%. "No evidence to suggest savings would fully offset GLP-1 prices." | No offset modeled |
| CBO | Medicare, obesity + CV | National | 9–10 years | +$35B net federal spending. By 2034: $7.1B drug spend vs. $1B savings. | No offset |
| Hwang et al. | Medicare simulation, obesity without T2D/CVD | 30M modeled | 10 years | Drug costs $65.9B; offsets $18.2B. Net increase: $47.7B. | No offset |
Manufacturer-funded analyses (Novo Nordisk)
| Study | Population | Key finding | Critical caveat |
|---|---|---|---|
| OFFSET study | T2D + CVD hospitalization | Budget neutral — lower inpatient/outpatient offset higher drug costs | T2D+CVD only; pre-semaglutide era |
| Komodo/Novo — multimorbidity | OW/OB + ≥2 complications | All-cause medical costs 27% lower ($3,870/yr) | Excludes GLP-1 drug cost entirely; 101-day follow-up |
| SHINE-HF | OW/OB + heart failure | Medical costs 28% lower (-$8,544/yr) | Excludes drug cost; N=408 |
| Novo/Value in Health | Medicare, all indications | Net savings of $715M across all indications | Includes T2D (where savings are largest); assumes generic entry year 7 |
The Aon and Prime Therapeutics analyses — the two most important independent datasets — reach different conclusions about medical cost trends. Prime found no medical cost reduction at all in 2 years. Aon found medical cost growth slowing by 18–30 months.
Key differences: Aon's sample is 57× larger (192,000 vs. 3,346), includes T2D patients (70% of their cohort), uses "digital twin" matching, spans a later time period (2022–2025 vs. 2021–2023), and measures growth rate rather than absolute cost. Prime's sample is exclusively obesity-without-diabetes — the harder population to demonstrate offsets in. Neither is peer-reviewed in a traditional journal; both are from organizations with potential interests (PBM vs. consulting firm).
No independent study has measured total cost of care (medical + pharmacy) showing net savings for obesity-only (non-T2D) commercially insured populations at any time horizon. The manufacturer-funded studies showing medical savings systematically exclude the cost of the GLP-1 drug itself.
The cardiovascular argument
The SELECT trial and employer populations
SELECT is the most important trial for the cost-offset argument: it showed a 20% reduction in major adverse cardiovascular events (MACE). But the trial population and employer populations are fundamentally different in three ways that matter for ROI.
Trial population vs. employer populations
| Characteristic | SELECT trial | Typical employer population |
|---|---|---|
| Mean age | 61.6 years | Working-age (25–55) |
| Sex | 72.3% male | ~50/50 |
| CVD status | 100% established ASCVD (prior MI, stroke, or PAD) | Vast majority have no established CVD |
| Diabetes | Excluded (HbA1c <6.5%) | 30–40% of obese with CVD have T2D |
| Background therapy | 90% on statins, 86% antiplatelets | Much lower |
| Persistence | 82.5% of treatment time | 32–63% at 12 months; 8% at 3 years |
| Eligible fraction | ~31% of post-MI registry patients met SELECT criteria | FDA broadened to ~22.7M adults; most not SELECT-like |
SELECT trial: Lincoff et al., NEJM, 2023. Eligibility analysis: post-MI registry study. FDA CV indication: approved label, March 2024.
A secondary SELECT analysis (Plutzky et al., ECO 2025) found cardiovascular benefits emerging within 3 months — 38% MACE reduction at 3 months, 41% at 6 months — before significant weight loss or target dose was reached. This suggests anti-inflammatory mechanisms beyond weight loss. However, this applies only to the SELECT population with established ASCVD — a population largely absent from typical employer GLP-1 coverage.
Published cost-effectiveness based on SELECT: McEwan et al. (Journal of Medical Economics, Feb 2025; Novo Nordisk-funded) estimated an ICER of $136,271/QALY at U.S. list price and $32,219/QALY with a 48% rebate. The authors explicitly acknowledged: "The generalizability of observations from SELECT to a broader US population is unknown."
Actuarial guidance
Published ROI models and actuarial analyses
EBRI, Mercer, and Milliman — the three organizations employers most rely on for actuarial guidance — all conclude that GLP-1s increase employer costs and that evidence for full cost offset does not exist. None publishes a model projecting positive ROI at current prices within an employer-relevant time horizon.
| Organization | Date | Type | Key assumptions | Projected impact | Adjusts for real-world persistence? |
|---|---|---|---|---|---|
| EBRI | Oct 2025 | Premium simulation | $617–766/30-day supply; 42% persist 12+ weeks; $90 copay modeled | +5.3% to +13.8% premium increase | Yes — uses real-world tiers |
| Mercer | 2024–2026 | Framework | ~$1,000/mo list; 2–5% utilization; 1-in-12 persist at 3 years | "Immediate ROI remains uncertain." Recommends claims analysis. | Acknowledges gap; cites Prime data |
| Milliman | Aug 2023; Sep 2025 | Markov forecast | >68% don't maintain 12-month therapy; 26% drug spend waste | Utilization steady-state 7.9–12.7% of commercial population | Yes — models discontinuation waste |
| Aon | Jan 2026 | Claims analysis | 192,000 users; "digital twin" matching; 35% flat rebate | Medical cost trend slowed 7% by yr 2; MACE hospitalizations down 44% | Yes — reports by adherence tier |
| AssuredPartners | Feb 2025 | 5-year claims | 9,236 diabetic GLP-1 users; medical + Rx costs | Medical savings $560/yr; drug cost $6,540/yr. Net negative. | Implicitly — measures actual users |
| Milliman/FlyteHealth | 2025 | Pilot analysis | CT state employees; 86% persistence at 6 months (managed program) | $430K–$1.2M cost avoidance from rejected claims + switching | N/A — managed program |
Sources: EBRI Issue Brief #644, Oct 2025; Mercer "GLP-1 Considerations for 2026," Nov 2025; Milliman, Aug 2023 and Sep 2025; Aon Phase 2 analysis, Jan 2026; AssuredPartners, Feb 2025; Milliman/FlyteHealth CT pilot, 2025.
The missing analysis
The persistence-adjusted ROI problem
Every published cost-effectiveness model uses clinical trial persistence assumptions (75–85%). Real-world persistence is 8–63% depending on cohort and time horizon. No published analysis reconciles this gap — and it changes everything.
| Time point | 2021 cohort | 2024 cohort | Clinical trial |
|---|---|---|---|
| 6 months | ~46% | ~61% | ~90%+ |
| 12 months | 32% | 63% | ~82–85% |
| 24 months | ~15% | Not yet available | ~75–80% |
| 36 months | 8.1% | Not yet available | N/A |
Source: Prime Therapeutics (JMCP 2024; 3-year white paper June 2025). The 2024 improvement (from 32% to 63% at one year) reflects resolution of supply shortages and newer formulations, but no data yet exists confirming whether this cohort maintains higher persistence at 2–3 years. Full analysis on page 3.
No published analysis formally incorporates real-world persistence rates into GLP-1 ROI or cost-effectiveness projections for employers. This is the single most important gap in the evidence base. As Prime Therapeutics noted: "Without current real-world research describing obesity GLP-1 treatment persistency…forecasting cost-effectiveness evaluations and use will rely on the RCT persistency rates."
The implication: all published CEA results are likely overstated relative to what employers would actually experience, because they assume 75–85% persistence while real-world persistence is 8–63%.
The weight regain data makes this worse. The STEP 1 extension study found participants regained approximately two-thirds of prior weight loss within one year of stopping semaglutide. At week 120, only 48.2% still had ≥5% weight loss (vs. 86.4% at end of treatment). Cardiometabolic improvements also reverted toward baseline. Full regain analysis on page 4.
Structural problem
Time horizon mismatch
Cardiovascular benefits take years. The employer who pays may not be the employer who benefits. No published analysis formally models this mismatch.
| Parameter | Value | Source |
|---|---|---|
| Median private-sector employee tenure | 3.5 years | BLS, January 2024 |
| Ages 25–34 tenure | 2.7 years | BLS |
| Time to demonstrated medical cost offset (GLP-1s) | >2 years (none demonstrated at any horizon for obesity-only) | Prime Therapeutics |
| Time to medical cost growth reduction | ~18 months (adherent patients only) | Aon |
| 3-year persistence rate | 8.1% | Prime Therapeutics |
| Bariatric surgery break-even | 2–4 years | Crémieux et al. |
| DPP break-even | ~3 years | ICER 2016 |
An employer invests ~$6,540/year net drug cost per GLP-1 member. At 3.5 years median tenure, with 8% persistence at 3 years, the employer's cost exposure is front-loaded while any potential savings are back-loaded — and likely accrue to the next employer's plan.
As UMass Lowell's Cherniack stated: "CVD expression may occur at a later point in life history, beyond the tenure of an incumbent employer or health plan."
Value assessment
Cost-effectiveness analyses: opposite conclusions
Two major 2025 analyses reached opposite conclusions about GLP-1 cost-effectiveness. The divergence is driven primarily by whether cardiovascular outcomes from SELECT are incorporated. Both agree that near-term budget impact for employers is negative.
| Parameter | Hwang et al. (JAMA Health Forum, Mar 2025) | ICER (Final report, Dec 2025) |
|---|---|---|
| Semaglutide ICER | $467,676/QALY | $61,400/QALY |
| Tirzepatide ICER | $197,023/QALY | $53,400/QALY |
| Cost-effective at $100K/QALY? | No (0% probability) | Yes (well below threshold) |
| SELECT CV data used? | No (weight-mediated benefits only) | Yes (full CV outcomes) |
| Net price used | ~$8,412/yr (sema); ~$6,236/yr (tirze) | ~$6,829/yr (sema); ~$7,973/yr (tirze) |
The critical distinction: ICER's 2025 update incorporated SELECT cardiovascular outcomes, MASH/CKD/HF benefits, and lower net prices. Hwang's model focused on weight-mediated disease pathways. The ICER model is more optimistic about lifetime value — but lifetime value and employer ROI are different questions. Even ICER's model shows cost-effectiveness, not cost-savings: the drug still costs more than it saves even over a lifetime. It just produces enough health benefit to justify the excess cost at standard willingness-to-pay thresholds.
Semaglutide would need to reach ~$1,522/year (~$127/month) — an additional 81.9% discount beyond current net prices — to meet the $100,000/QALY threshold.
Tirzepatide would need to reach ~$4,334/year (~$361/month) — an additional 30.5% discount.
For context, naltrexone-bupropion was found to be cost-saving at current prices. Full cost-effectiveness analysis on page 2.
The other side
The counterarguments — presented honestly
The case for GLP-1 coverage extends beyond direct medical cost offsets. These are real arguments, even if harder to quantify — and the ethical argument deserves serious consideration regardless of ROI.
Indirect cost burden of obesity
The total burden of obesity — including presenteeism, absenteeism, disability, and workers' comp — is estimated at $6,472 per employee with obesity per year. This is roughly comparable to annual net GLP-1 drug costs, which is the core of the manufacturer argument: the drug cost replaces existing burden rather than adding to it.
The gap: This argument depends on GLP-1s actually eliminating these costs, which requires sustained treatment. No study has yet directly measured GLP-1-specific absenteeism or productivity gains.
Source: GlobalData/PMC 2024; JOEM 2021 (instrumental variable analysis).
Recruitment and retention value
GLP-1 coverage was valued more than unlimited PTO or hybrid work in one survey (9amHealth). These surveys are from organizations with commercial interests (9amHealth sells GLP-1 care; NFP is a benefits consultancy), but the consistency across sources lends credibility.
Sources: NFP 2026; 9amHealth 2026; WTW 2025.
The coverage parity and equity argument
The American Medical Association (November 2023) passed a resolution urging all payers to ensure coverage parity for evidence-based obesity treatment. The New England Journal of Medicine published arguments that excluding obesity medications undermines opportunities for addressing inequities, given disproportionate obesity prevalence among Black (49.9%) and Hispanic (45.6%) adults.
The philosophical core: No employer demands ROI proof before covering statins, insulin, or antihypertensives. Applying this standard uniquely to obesity treatment may reflect stigma rather than sound benefits policy.
The Aon study as strongest positive evidence
Aon's analysis — 192,000 users across 50 million commercial lives — found MACE hospitalizations down 44%, medical cost growth slowing 7–13% by year two, and striking cancer-related findings in women. Aon launched its own subsidized GLP-1 program based on these findings.
Caveats: Aon is a benefits consulting firm with commercial interest; the study measures cost growth reduction, not absolute savings; 70% of the cohort has T2D (easier to demonstrate offsets); and results are strongest only for the ≥80% adherence subgroup.
Break-even analysis
What would need to be true for positive ROI
For an employer to achieve positive ROI, three conditions must hold simultaneously: drug price must drop substantially, persistence must be high, and the time horizon must be long enough. Currently, none of these conditions is met for obesity-only populations in commercial plans.
Novo Nordisk announced $675/month list price for all semaglutide products starting January 2027. Eli Lilly launched Employer Connect offering Zepbound at $449/month bypassing PBMs. NovoCare sells Wegovy at $349/month cash pay. The Trump administration secured $245/month for Medicare.
These represent 50–75% reductions from list prices, moving toward but not yet reaching break-even thresholds for most employer analyses. Full pricing data on page 2.
How GLP-1 ROI compares to alternatives
| Intervention | Annual cost | Break-even | Published ROI | ICER ($/QALY) |
|---|---|---|---|---|
| DPP (lifestyle) | $400–$500/yr | ~3 years | $4,250 net savings over 10 yrs | $5,494–$12,878 |
| Bariatric surgery | $15K–$30K (one-time) | 2–4 years | 22.6% cost reduction at 2 yrs | Cost-effective |
| Omada Health (digital DPP) | Outcomes-based | ~3 years | $1,169/yr savings (Dow Chemical) | ~$5,500–$13,000 |
| Virta Health | $900–$2,808/yr | Year 1 (claimed) | $425 PPPM savings; 2:1 ROI claimed | Not independently published |
| GLP-1s | $6,830–$16,000/yr (ongoing) | Not achieved | Net -$6,540/yr (AssuredPartners) | $53K–$468K |
Sources: DPP: ICER 2016; University of Michigan 2025. Bariatric: Crémieux et al. Omada: peer-reviewed claims analysis, Dow Chemical. Virta: actuary-validated internal data. GLP-1s: AssuredPartners; Hwang et al.; ICER 2025.
Model it
Honest ROI calculator
Adjust the inputs below to model GLP-1 ROI under different persistence and offset assumptions. The key feature: toggle between clinical trial and real-world persistence to see how dramatically the answer changes.
GLP-1 ROI estimator
Model total pharmacy spend vs. medical cost offsets under different assumptions. All outputs update in real time.
This calculator uses simplified models for illustrative purposes. Medical cost offsets are applied only to members who persist on treatment. Persistence rates decay across years based on published data. Year 2–3 rates for 2024 cohorts are extrapolated — no confirming data exists. The "clinical trial" toggle uses STEP/SURMOUNT persistence rates. Real-world data from Prime Therapeutics. This is not actuarial guidance — consult Milliman, Mercer, or your PBM's analytics team for plan-specific modeling.
Methodology
How this page was built
All data on this page is drawn from publicly available sources: peer-reviewed studies, government publications, consulting firm reports, actuarial analyses, and industry surveys. Where sources conflict — as with the Aon and Prime Therapeutics analyses on medical cost trends — both are presented with the methodological differences that may explain the divergence.
The ROI calculator uses simplified models for illustration. It does not incorporate PBM administrative fees, member cost-sharing, indirect cost savings, or the impact of weight regain after discontinuation. Persistence rates for 2024 cohorts at years 2–3 are extrapolated from 2021 cohort trajectories — no confirming data yet exists. The "clinical trial" comparison uses reported on-treatment persistence from STEP and SURMOUNT trials.
No data points have been inferred or extrapolated beyond what the source material supports; gaps are noted as gaps. The manufacturer-funded studies are clearly labeled as such. Corrections and updates can be submitted via the contact page.
References
Sources
Questions about this data? Corrections or updates?
Get in touch