The longest randomised evidence for GLP-1 medication runs to about 3.3 years, and it comes from trials that deliberately enrolled narrow, high-risk populations — people who had already had a heart attack, or who already had both diabetes and kidney disease. Those trials found real benefits in those groups. What they cannot tell you is what happens over a decade, or what happens to someone whose only reason for taking the medicine is weight. This article covers what the big trials show, who was in them, and where long-term data does not exist yet.
Outcome trial: a trial that measures events — heart attacks, kidney failure, death — rather than a number on a scale. They are large, slow and expensive, which is why there are so few.
The largest and longest trial: cardiovascular outcomes
SELECT randomised 17,604 adults and followed them for a mean of 39.8 months.
Everyone enrolled was aged 45 or over, had a BMI of 27 or above, and had established cardiovascular disease — a prior heart attack, a prior stroke, or symptomatic peripheral artery disease. Anyone with type 1 or type 2 diabetes was excluded. Over that follow-up, 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 (Lincoff 2023).
Two things about that result deserve equal weight. It is a genuine 20% relative reduction in hard cardiovascular events over more than three years. And it is secondary prevention — every participant already had cardiovascular disease before starting, so it says nothing about someone with no cardiovascular history taking the same medicine for weight.
The same trial also reported that more participants discontinued because of an adverse event on semaglutide than on placebo on placebo. Long-term tolerability is part of the long-term picture.
The kidney trial, and who was in it
FLOW randomised 3,533 adults over a median 3.4 years and was stopped early because the benefit was clear.
The enrolment criteria were narrow and specific: adults who had both type 2 diabetes and chronic kidney disease, defined by particular combinations of eGFR and urinary albumin-to-creatinine ratio. In that group, the composite of kidney failure, a 50% or greater fall in eGFR, or death from kidney or cardiovascular causes occurred less often on semaglutide, a hazard ratio of 0.76 (Perkovic 2024).
This is the strongest kidney evidence in the class, and it belongs entirely to people who already had diabetic kidney disease within defined thresholds — not to someone taking the medicine for weight with normal kidney function.
The sleep apnoea trials
Two trials reported together randomised 469 adults with obesity and moderate-to-severe obstructive sleep apnoea over 52 weeks.
In participants not using positive airway pressure therapy, tirzepatide reduced the apnoea-hypopnoea index by 20.0 more events per hour than placebo; in those also using PAP therapy, by 23.8 more (Malhotra 2024). Everyone enrolled had a BMI of 30 or above and an AHI of 15 or more, so mild sleep apnoea and sleep apnoea without obesity were both outside the trial.
Note the length: one of the marquee non-weight results in this field ran for a single year. Individual results vary, and these figures describe the populations enrolled.
Why the enrolled population is the finding, not a footnote
Each of these trials selected for the exact condition it then improved.
A trial that enrols only people with established cardiovascular disease can measure whether cardiovascular events fall in people with established cardiovascular disease. It cannot measure what happens to anyone else, because nobody else was in it. The same holds for the kidney trial, where diabetes and kidney disease were both entry requirements, and for the sleep-apnoea trials, where obesity and an apnoea-hypopnoea index of 15 or more were both entry requirements.
The distinction is substantive rather than pedantic. "This medicine reduced repeat cardiovascular events in people who had already had one" and "this medicine protects your heart" are different claims, and only the first was tested.
The question the trials are still too short to answer
Thyroid cancer is the long-latency worry, and the evidence genuinely disagrees.
A French nested case-control study of 2,562 thyroid cancer cases found that one to three years of GLP-1 receptor agonist use in people with type 2 diabetes was associated with a higher relative risk of thyroid cancer overall (adjusted HR 1.58) and of medullary thyroid cancer (HR 1.78) (Bezin 2023). A Scandinavian cohort of 145,410 GLP-1 initiators compared against 291,667 users of a different drug class — a design that reduces the risk of simply detecting more cancers because patients were scanned more — found no significant association, HR 0.93 over a mean 3.9 years (Pasternak 2024). The most comprehensive analysis available, combining 93 trials, more than 18 million patient-years of post-marketing exposure and a US real-world database, concludes that the totality of the data does not suggest an association (Vilsbøll 2026).
That last source is reassuring, and it should be read completely: its all-trials hazard ratio was 1.70 (95% CI 0.99–3.03), just under statistical significance rather than comfortably at null.
The structural limit applies to all of it. Mean follow-up in the largest cohort was under four years, and thyroid cancers can take far longer than that to appear. These studies can say that no signal has emerged in the first few years. None can say what a decade looks like.
What has no long-term data at all
Four gaps are worth naming, because they are the ones most often filled with confident guesses.
Body composition beyond a year. The longest lean-mass follow-up available is a single-arm cohort of 106 completers with a BMI of 40 or above, in which lean-mass loss plateaued after seven months while fat loss continued and grip strength improved (Alissou 2026). With no placebo arm, it cannot show the pattern belongs to the drug — and it stops at 12 months.
What regained weight is made of. The largest meta-analysis of stopping GLP-1 medication pooled body weight, waist circumference and BMI, and nothing else — no fat mass, no lean mass (Berg 2025). The common claim that weight returns as fat is an extrapolation from a different literature, not a measured GLP-1 finding.
Stopping after years rather than months. Every published withdrawal trial stops the medicine after 68 weeks or less. In the STEP 1 extension, participants regained about two-thirds of the weight lost within a year of stopping, and most cardiometabolic improvements reverted towards baseline (Wilding 2022); pooled across cessation trials, about 60% returns within a year, decelerating thereafter (Budini 2026). What happens to someone stopping after four years is unmeasured. We cover the decision in how long you should stay on GLP-1 medication.
Tapering. No randomised trial has tested tapering a GLP-1 dose. Every published withdrawal trial is an abrupt switch to placebo. Anyone presenting a taper as evidence-based is describing a clinical preference, not a trial result.
At GetLean, our philosophy is that the medication is the catalyst and what you keep is the result. That view is a response to the absence of long-term data, and to the one thing measured repeatedly: what happens when the medicine stops.
Individual results vary, and clinical-trial figures describe the populations studied. Whether any of this evidence applies to you is a question for a doctor who knows your history.
Common questions
How long have GLP-1 medications been studied?
The longest and largest randomised trial reported so far followed 17,604 adults for a mean of 39.8 months - about 3.3 years (Lincoff 2023). Beyond roughly four years, randomised evidence in this drug class does not yet exist.
Do GLP-1 medications protect the heart?
In one specific population, yes. In 17,604 adults aged 45 or over with overweight or obesity and established cardiovascular disease but without diabetes, semaglutide reduced cardiovascular death, heart attack or stroke by 20% versus placebo (Lincoff 2023). Everyone enrolled already had cardiovascular disease, so this does not describe a general weight-management population. Individual results vary.
Do GLP-1 medications cause thyroid cancer?
The evidence disagrees. A French case-control study found a raised relative risk (Bezin 2023); a 145,410-person Scandinavian cohort found no significant association (Pasternak 2024); and the most comprehensive integrated analysis concludes the totality of data does not suggest one, though its all-trials estimate sits just under significance (Vilsbøll 2026). Follow-up under four years cannot rule out a long-latency effect.
What happens to muscle over the long term on GLP-1?
The longest body-composition follow-up available is 12 months, in a single-arm cohort of 106 people with a BMI of 40 or above, where lean-mass loss plateaued after seven months while fat loss continued (Alissou 2026). With no placebo arm and only a year of data, multi-year lean-mass evidence does not yet exist.
What happens if you stop after several years?
Nobody has tested stopping after several years. The withdrawal evidence comes from trials of 68 weeks or less: adults who stopped semaglutide regained about two-thirds of the weight lost within a year, with most cardiometabolic gains reverting towards baseline (Wilding 2022).