| Trial results
An Italian study of 80,302 adults with type 2 diabetes found that people who stopped a GLP-1 medicine had about twice the rate of heart attack, stroke or death of matched people who stayed on it, and about 4.8 times the rate of death from any cause. Those are differences in rates in medical records. They do not show that stopping caused them. The study did not record why people stopped or what they died of, and it covers only people with type 2 diabetes, few of them on semaglutide (Ozempic, Wegovy) and none on tirzepatide (Mounjaro, Zepbound).
The headline numbers
| What was counted | Stayed on (events per 100 person-years) | Stopped (events per 100 person-years) | Hazard ratio (95% CI) |
|---|---|---|---|
| Non-fatal heart attack, non-fatal stroke or death from any cause | 1.96 | 3.94 | 2.03 (1.92 to 2.14) |
| Death from any cause | 0.48 | 2.23 | 4.81 (4.39 to 5.26) |
Each group had 28,713 people after matching. A person-year is one person followed for one year. The ranges in brackets are 95% confidence intervals. Observational study: it shows a link, not a cause. Source: Ceriello et al., The Lancet Regional Health Europe, online first November 9, 2026.
What the researchers did
Ceriello and colleagues used electronic records from 301 diabetes clinics across Italy, covering 2004 to 2024. They looked at adults with type 2 diabetes who had never had a heart attack or stroke and who started a GLP-1 receptor agonist, the class of medicine that includes semaglutide (Ozempic, Wegovy, Rybelsus). In all, 80,302 people qualified.
A person counted as having stopped if no GLP-1 prescription was recorded at two or more visits in a row. By the last visit on record, 31,687 people (39.5%) were still on the medicine and 48,615 (60.5%) had stopped. To compare like with like, the researchers matched 28,713 people who stayed on with 28,713 who stopped. The matching used age, sex, body mass index, blood sugar control (HbA1c), how long they had had diabetes, and use of a diabetes pill called a sulfonylurea.
Each person's time on the medicine counted toward the "stayed on" group, and their time after stopping counted toward the "stopped" group. The study then compared how often events happened per year of follow-up, which averaged roughly four years.
The main measure combined three things: a heart attack that was not fatal, a stroke that was not fatal, and death from any cause. Death from any cause was also counted on its own.
What it found
People who stopped had a higher rate of the combined outcome: 3.94 events per 100 person-years against 1.96 for people who stayed on, a hazard ratio of 2.03. For death from any cause the rates were 2.23 and 0.48 per 100 person-years, a hazard ratio of 4.81. A stricter definition of stopping gave similar results.
In everyday terms, if 1,000 people were followed for a year, the study's rates work out to about 22 deaths among people who had stopped and about 5 among people who stayed on. The average age was about 65.
The gap was larger in people with obesity (a BMI of 30 or more) than in people without. The hazard ratio for the combined outcome was 2.27 in the first group and 1.61 in the second.
The study also tracked weight, blood sugar, blood pressure and cholesterol after stopping. Those changes were small. Among people who stopped, the ones who went on to have a heart attack, stroke or death differed only slightly on those measures from the ones who did not. The authors say any explanation that does not run through those risk factors is exploratory.
What a hazard ratio means
You may have seen this study summarized as "five times higher risk of death." A hazard ratio compares how often something happened per year of follow-up in two groups. In this observational comparison, a ratio of 4.81 for death from any cause means deaths were recorded at about 4.8 times the rate in the stopped group. It does not give any one person's chance of dying, and it does not say how many people died over a lifetime.
The rates are the better guide: about half a death per 100 people per year in the stayed-on group, and about two per 100 per year in the stopped group. The ratio that also counts non-fatal heart attacks and strokes is 2.03. Neither ratio shows that stopping caused the difference.
Why stopping and getting sicker can be the same story
People rarely stop a medicine at random. Someone who develops cancer, is hospitalized or becomes frail may stop picking up a prescription or be taken off it, and is also more likely to die. In that case the illness explains both the stopping and the death, and the missing medicine is a sign of trouble rather than the cause. Researchers call this reverse causation.
This study could not rule it out. It had no record of why anyone stopped and no cause of death. The authors raise the problem themselves. They note that the gap in deaths is larger than the heart benefit seen in randomized trials of GLP-1 medicines, that such a large gap is unlikely to come from stopping alone, and that unmeasured factors may explain part of it. They write that stopping may be a marker of higher underlying risk rather than a direct cause, and they describe their findings as hypothesis generating, meaning the idea needs testing in studies built for the question.
The paper does speak to one version of the worry. At the moment people stopped, their blood sugar, weight, blood pressure and cholesterol were similar to or better than when they started, so the stopped group was not obviously sicker on those measures. The authors also point out that Italy's health system pays for GLP-1 medicines used for diabetes, which makes it less likely that people stopped because they could not afford them. The records did not say whether an illness such as cancer or frailty led anyone to stop.
What the study cannot say
- It cannot say that stopping caused the extra events or deaths.
- It cannot say what people died of. There was no cause-of-death information, and hospital admissions for heart failure were not available.
- It cannot say why people stopped. Side effects, a new illness, cost and other reasons were not recorded.
- It cannot say how stopping compares with never starting. The comparison is between people who stayed on and people who stopped, and the authors say they could not compare either group with people who never took a GLP-1 medicine.
- It does not describe what treatment people moved to after stopping in the main text of the paper.
Who the study does not cover
Everyone in the study had type 2 diabetes. The authors say the findings may not apply to people who use these medicines for weight loss without diabetes, so this page does not apply them there. Only a minor share of people took semaglutide (Ozempic, Wegovy, Rybelsus). Tirzepatide (Mounjaro, Zepbound) was not yet available during the study, so the results say nothing about it.
The records come from diabetes clinics in Italy. Of 127,975 people who started a GLP-1 medicine in those clinics, 47,673 (37%) were left out of the analysis: 41,205 had no follow-up or a single prescription that was never renewed, and 6,468 had already had a heart attack or stroke.
A second study, from US veterans
A separate analysis of US Veterans Affairs records, by Xie, Choi and Al-Aly, was published in BMJ Medicine in March 2026. It followed 132,551 veterans with type 2 diabetes who started a GLP-1 medicine and compared different patterns of stopping over three years. Compared with continuing the medicine for the whole period, one year of discontinuation was linked to a 14% higher risk of heart attack, stroke or death from any cause (risk ratio 1.14, 95% CI 1.09 to 1.18), and two years of discontinuation to a 22% higher risk (1.22, 1.16 to 1.27).
That study is also observational. It used a different method and measured risk over a three-year window, so its numbers cannot be lined up against the Italian ones. The ratios are much smaller than the Italian hazard ratios, which is one more reason not to treat the two papers as one finding.
Where this fits on the Atlas
The randomized trials summarized on the Atlas page on weight regain after stopping Wegovy or Zepbound followed weight, blood sugar and blood pressure for up to about a year after stopping. They were not designed to count heart attacks or deaths. This study asks a different question, using medical records and a longer follow-up. For the eating side of stopping, see eating after you stop. For how to read any headline figure, see the beginner guide and trial results, with context. PeptideBlog.net has a separate analysis of the same paper, Stopping GLP-1 drugs and death risk: what the 4.81 does and does not mean. If you are thinking about stopping a medicine you were prescribed, talk with the clinician who prescribes it first, and the page on questions to ask your prescriber can help you prepare.
What is still unknown
- Whether stopping itself changes the risk of heart attack, stroke or death, or whether stopping marks people who were already at higher risk. The authors call for studies designed to answer that.
- Whether the pattern holds in people who use these medicines for weight loss without diabetes, or in people taking tirzepatide.
- What people died of, and in what order events happened.
- How heart attacks and strokes behave on their own. The main text of the paper reports the combined figure and death, but not the two non-fatal events separately.
Quick answers
- Does stopping a GLP-1 medicine raise the risk of death?
- In this Italian study of people with type 2 diabetes, people who stopped had a higher death rate than matched people who stayed on. The study cannot show that stopping caused it.
- Is the risk of death really 5 times higher after stopping a GLP-1 medicine?
- The hazard ratio for death from any cause was 4.81, which compares rates per year of follow-up and not each person's chance of dying. The figure for heart attack, stroke or death combined was 2.03. Both come from an observational study that cannot show cause.
- Does this apply to Wegovy or Zepbound used for weight loss?
- The study included only people with type 2 diabetes, few of them on semaglutide, and tirzepatide was not included. The authors say it may not apply to people using these medicines for weight loss alone.
- Could people who stop be sicker to begin with?
- That is possible, and the authors say it may explain part of the gap. People who become seriously ill may stop treatment and are also more likely to die. The study had no information on why people stopped.
- What did the study not measure?
- Cause of death, reasons for stopping, hospital admissions for heart failure, and what people took after stopping.
- Has another study looked at stopping a GLP-1 medicine and heart events?
- A US Veterans Affairs analysis in BMJ Medicine linked one year of discontinuation to a 14% higher risk of heart attack, stroke or death, and two years to 22%. It is also observational.
This page is general information, not medical advice.
Sources: Ceriello A and colleagues. GLP-1 receptor agonist discontinuation and cardiovascular events and all-cause mortality in people with type 2 diabetes: a retrospective cohort study. The Lancet Regional Health Europe, online first November 9, 2026 Peer-reviewed paper, observational cohort study, authors report fees from GLP-1 makers. The study was funded by Italian Ministry of Health research funds to IRCCS MultiMedica, and the funder had no role in the analysis. Xie Y, Choi T, Al-Aly Z. Glucagon-like peptide 1 receptor agonist discontinuation and risks of major adverse cardiovascular events in adults with type 2 diabetes: target trial emulation. BMJ Medicine 2026;5(1):e002150, published March 18, 2026 Peer-reviewed paper, observational target trial emulation.