A new analysis from Washington University and Ziyad Al-Aly's group, published this week in BMJ Medicine, followed more than 333,000 adults with type 2 diabetes over three years. It is, by any measure, an impressive piece of work — 16 simulated treatment scenarios, target trial emulation methodology, a real-world dataset large enough to detect signals that randomized trials often miss.

The headline finding: GLP-1 receptor agonists significantly reduce the risk of heart attack, stroke, and cardiovascular death — but only in patients who stay on them. Stop the medication, and that protection erodes systematically over the following 24 months, eventually disappearing entirely.

This is important clinical information. It is also — depending on how you frame it — either an argument for better treatment infrastructure or a cautionary note about the nature of these drugs. The data support one of those interpretations far more than the other.

The cardiovascular signal from continuous GLP-1 use is real and clinically meaningful. Three years of uninterrupted therapy produced an approximately 18% relative reduction in major adverse cardiovascular events compared to sulfonylurea use — a 3.7% absolute risk reduction and a number needed to treat of roughly 27. For context, that NNT compares favorably to many interventions we consider standard of care in preventive cardiology.

What happens when patients stop is where the data become both striking and, I'd argue, clarifying:

One detail in these data deserves careful attention: the +22% at 24 months represents excess risk compared to patients who stayed on treatment — not compared to where these patients started. The paper's own data indicate that cardiovascular risk after stopping returns toward, not above, pre-treatment baseline. That distinction matters enormously for how we interpret the clinical implications.

The framing that has attached itself to this paper — that GLP-1s function as "chronic suppressive therapy" rather than legitimate disease treatment — rests on a standard we apply to no other medication in medicine.

Consider what that logic looks like applied consistently:

The pattern is identical because the underlying biology is identical: these are all chronic diseases in which the pathophysiology is suppressed, modulated, or compensated by pharmacotherapy — and reasserts itself when the pharmacotherapy is withdrawn. That is not a drug failure. That is what chronic disease looks like.

Obesity is a chronic neuroendocrine condition involving dysregulation of appetite-regulating circuits, adipokine signaling, and energy homeostasis set points. The STEP 1 Extension (Wilding et al., Diabetes, Obesity and Metabolism, 2022) demonstrated this precisely: two-thirds of weight lost during semaglutide treatment was regained within one year of stopping, accompanied by deterioration in blood pressure, lipid profiles, and C-reactive protein. The authors' stated conclusion was that "ongoing treatment is required to maintain improvements in weight and cardiometabolic risk factors." Not drug failure. Not suppression. Chronic disease requiring chronic treatment.

The most substantive critique in this conversation deserves direct engagement. The concern runs as follows: GLP-1-induced weight loss is not pure fat loss. DXA studies suggest that roughly 25–40% of weight lost may be lean mass — and if patients subsequently stop medication and regain weight primarily as fat, they may end up with a worse body composition phenotype than where they started: more fat mass, less muscle, poorer insulin sensitivity, and greater metabolic fragility.

This is a biologically coherent hypothesis. It is also one that the best available imaging data — not DXA, but gold-standard MRI — substantially complicates.

A post-hoc analysis of the SURPASS-3 trial, published in The Lancet Diabetes & Endocrinology in 2025 (Sattar et al.), assessed muscle composition in 246 participants using MRI at baseline and week 52. This matters methodologically: DXA cannot reliably distinguish true muscle atrophy from reductions in intramuscular glycogen and water — changes that are adaptive, not pathological. MRI measures actual muscle tissue and, critically, intramuscular fat infiltration, which is an independent determinant of muscle function.

The findings are more reassuring than the DXA-derived narrative suggests:

Two findings here deserve emphasis. First, the reduction in muscle volume was proportionate to the degree of weight loss — exactly what you would expect in a healthy physiological response, since lighter bodies require less muscle mass to support their weight. Sattar noted that the muscle-to-fat ratio actually improves with such therapies. This is the opposite of pathological muscle wasting.

Second — and this is the more striking finding — tirzepatide removed intramuscular fat at a rate significantly exceeding what weight loss alone would predict. Excess fat within muscle tissue impairs glucose uptake, force generation, and metabolic function. Clearing it beyond the expected amount is, if anything, a signal pointing toward improved muscle quality, not compromised muscle health.

Whether this reflects a direct pharmacological effect of tirzepatide on muscle tissue — beyond the consequences of caloric restriction and weight loss — remains an open question. Sattar's group notes that it is difficult to disentangle the two. But the implication that these drugs are silently degrading muscle health while patients and clinicians focus on the scale is not supported by the best available data.

Does this close the conversation? No. Functional outcomes — strength testing, chair-rise performance, step counts, gait speed — have not been systematically assessed at scale, and Sattar's own conclusion is that trials measuring these endpoints are needed. The question of what happens to body composition after discontinuation, specifically, remains poorly characterized in imaging studies. And the sarcopenia concern is most acute in older adults, a population underrepresented in most GLP-1 trial populations to date.

What the SURPASS-3 MRI data do is shift the burden of proof. The claim that GLP-1s cause clinically meaningful muscle degradation is now a hypothesis that requires evidence — not a default assumption that opponents of treatment must rebut. The best available gold-standard imaging says the opposite.

And regardless: the correct clinical response to any residual uncertainty here is not to withhold treatment. It is to add resistance training, ensure protein-adequate nutrition, monitor body composition longitudinally, and develop a comprehensive obesity care infrastructure that optimizes outcomes. "GLP-1s alone may be insufficient for ideal muscle preservation" is an argument for better care — not for less of it.

"Are we treating disease — or pharmacologically compensating for a toxic food environment?"

This is a genuinely important question. And the honest answer is that both things are simultaneously true — and that framing them as competing alternatives is where the argument goes wrong.

Yes: we exist in an obesogenic environment that drives chronic energy dysregulation at a population level. Highly processed foods engineered for palatability and overconsumption, sedentary infrastructure, food insecurity, agricultural policy — these are real structural drivers of the obesity epidemic, and medication cannot substitute for addressing them. The argument for comprehensive food policy, urban design, and equitable access to whole foods is correct and urgent.

Also yes: patients are suffering today, with a disease that has hard cardiovascular endpoints, in bodies with neurobiological dysregulation that no amount of individual willpower was ever going to correct. We have medications with robust randomized trial evidence, confirmed cardiovascular outcome benefits, and a mechanism of action that addresses the actual biology of the disease. The answer to systemic environmental problems is never to withhold individual treatment while we wait for the system to be reformed.

Fix the food environment and treat the disease. These are not competing priorities. They are complementary obligations operating at different scales and timelines.

There is also a subtler nutritional concern embedded in this debate that deserves more attention than it receives: reduced appetite does not guarantee adequate nutrition. Patients on GLP-1 therapy often eat substantially less total food — but not necessarily better-quality food. Fiber, micronutrient, and protein intake can all become insufficient on a smaller dietary volume, particularly in patients who are not receiving structured nutritional support. The gut microbiome implications of sustained reductions in dietary diversity and quantity are poorly characterized. These are real clinical considerations — and they are arguments for better comprehensive care in obesity medicine, not for less treatment.

Up to 50% of patients prescribed GLP-1 medications stop within the first year. The WashU data quantify exactly what that means: real cardiovascular events. Not theoretical risk. Not risk factors. Hard clinical outcomes — heart attacks, strokes, deaths — that are preventable when treatment is maintained.

Why do patients stop? The data are consistent: cost is the primary driver. Insurance coverage for obesity pharmacotherapy remains patchy, exclusionary, and structurally discriminatory in ways that would be considered unconscionable if applied to diabetes or cardiovascular disease management. Side effect burden — nausea, vomiting, gastrointestinal discomfort — is a secondary driver that is often underaddressed through dose titration and supportive counseling. The healthcare infrastructure was not built to support long-term chronic disease treatment for obesity, the way it was built to support hypertension management or heart failure care.

These are systems failures. Attributing them to characteristics of the pharmacology — framing discontinuation as evidence that these are somehow "different" from other chronic disease medications — is a diagnostic error that points us toward the wrong solutions.

The right question is not "should we be concerned that GLP-1 benefits are contingent on continued use?" Every effective medication for chronic disease is contingent on continued use. That is not a feature of the drug; it is a feature of the disease.

The right questions are: Why can't patients afford to stay on them? Why won't insurers cover them? Why are we not pairing pharmacotherapy with behavioral support, nutritional counseling, and resistance training to maximize benefits and mitigate legitimate risks? Why does our healthcare system treat obesity as a lifestyle choice requiring periodic intervention rather than a chronic disease requiring chronic management?

GLP-1 drugs are not a cure. No serious clinician claims they are. They are what we have spent decades in obesity medicine arguing we needed: effective, evidence-based pharmacotherapy for a chronic neurobiological disease that we have spent those same decades undertreating, stigmatizing, and pretending it could be addressed with better personal choices.

The WashU analysis is good science, and the discomfort it generates is productive discomfort — the kind that should sharpen clinical practice and accelerate the policy conversations we need to have. But that sharpening cuts in the right direction only if we diagnose correctly what the data are telling us.

The data are telling us that three years of continuous GLP-1 use builds substantial cardiovascular protection — protection that took three years to accumulate and can be lost in 12 to 18 months. That is a compelling argument for designing systems that help patients stay on these medications: better insurance coverage, better side-effect management protocols, better integration of behavioral and nutritional support, better long-term monitoring.

The question is no longer whether to prescribe GLP-1 medications. That question has been answered by LEADER, SUSTAIN-6, SELECT, SURPASS-CVOT, and now the WashU real-world outcomes data.

The question is how we build a system where the patients who need these medications can actually stay on them — and how we pair that pharmacotherapy with the comprehensive care that makes the treatment more than the sum of its molecular parts.

References

1. Xie Y, Choi T, Al-Aly Z. Cardiovascular consequences of discontinuing glucagon-like peptide-1 receptor agonists in adults with type 2 diabetes. BMJ Medicine. 2026. [Target trial emulation; N=333,000+; 3-year follow-up; 16 treatment scenarios.]

2. Wilding JPH, Batterham RL, Davies M, et al. Weight regain and cardiometabolic effects after withdrawal of semaglutide: The STEP 1 trial extension. Diabetes Obes Metab. 2022;24(8):1553–1564. doi:10.1111/dom.14725.

3. Marso SP, Daniels GH, Brown-Frandsen K, et al. Liraglutide and Cardiovascular Outcomes in Type 2 Diabetes (LEADER). N Engl J Med. 2016;375:311–322.

4. Marso SP, Bain SC, Consoli A, et al. Semaglutide and Cardiovascular Outcomes in Patients with Type 2 Diabetes (SUSTAIN-6). N Engl J Med. 2016;375:1834–1844.

5. Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and Cardiovascular Outcomes in Obesity without Diabetes (SELECT). N Engl J Med. 2023;389:2221–2232.

6. Nicholls SJ, Pavo I, Bhatt DL, et al. Cardiovascular outcomes with tirzepatide versus dulaglutide in type 2 diabetes (SURPASS-CVOT). N Engl J Med. 2025;393:2409–2420. [Tirzepatide noninferior to dulaglutide for 3-point MACE over median 4-year follow-up in patients with T2D and established ASCVD. Superiority not met on primary endpoint; expanded MACE endpoint significantly favored tirzepatide (HR 0.88).]

7. Lam CSP, Rodriguez A, Aminian A, et al. Tirzepatide for reduction of morbidity and mortality in adults with obesity: rationale and design of the SURMOUNT-MMO trial. Obesity (Silver Spring). 2025;33(9):1645–1656. doi:10.1002/oby.24332. [Trial design paper only. SURMOUNT-MMO outcomes results have not yet been published as of March 2026; trial is ongoing.]

8. Sattar N, Neeland IJ, Dahlqvist Leinhard O, et al. Tirzepatide and muscle composition changes in people with type 2 diabetes (SURPASS-3 MRI): a post-hoc analysis of a randomised, open-label, parallel-group, phase 3 trial. Lancet Diabetes Endocrinol. 2025;13(6):482–493. doi:10.1016/S2213-8587(25)00027-0. [MRI-based assessment of thigh muscle volume, muscle fat infiltration, and muscle volume Z-score at 52 weeks; N=246; observed changes benchmarked against UK Biobank population-based estimates (n≈2,942).]