Semaglutide vs Liraglutide: Study Results

A 2026 real-world comparison found that people prescribed semaglutide had a lower recorded incidence of diabetes than matched people prescribed liraglutide. The difference was not detected in the first six months of follow-up, and the study did not detect a difference in cardiovascular disease events. This was an observational analysis of insurance claims, not a randomized trial, so it cannot establish that one medicine caused the difference^[1].

The study has attracted a published comment, which is a useful prompt to read the original paper carefully. The comment itself is a letter without an abstract in PubMed, so it does not provide enough accessible detail to support a separate account of its arguments. This article therefore reports the underlying study and its stated limits rather than inferring what the letter concluded^[2].

What the study compared

The authors used MarketScan health-insurance claims data to emulate, or approximate, a target trial. A target-trial emulation is an observational design that tries to specify the comparison a randomized trial would have made, then applies that plan to existing records. The study included people without recorded diabetes or cardiovascular disease who began semaglutide or liraglutide between 2021 and 2023^[1].

The medicines are both GLP-1 receptor agonists, a drug class used in approved medicines for metabolic conditions. The paper compared new users rather than people who had already spent an unknown amount of time on either treatment. It then matched people one-to-one on recorded characteristics and used a statistical adjustment based on those recorded factors^[1].

The recorded outcomes

After matching, the analysis included 57,456 users. Their mean age was 45 years and 83% were female. Over an average of one year of follow-up, the database recorded 1,104 diabetes events and 57 cardiovascular disease events. The cardiovascular outcome included myocardial infarction, heart failure, or stroke^[1].

For incident diabetes, the adjusted hazard ratio for semaglutide versus liraglutide was 0.88, with a 95% confidence interval from 0.78 to 0.99. That is an association with a lower recorded rate in this dataset, not a personal prediction. The paper also reported that the proportional-hazards assumption did not hold for this outcome, so the authors split the result by time in follow-up^[1].

In the first six months, the analysis did not find an association between semaglutide use and lower diabetes incidence: the reported hazard ratio was 0.99, with a confidence interval from 0.82 to 1.18. After six months, the reported hazard ratio was 0.80, with a confidence interval from 0.68 to 0.94. The study did not detect a difference in cardiovascular events, reporting a hazard ratio of 1.25 with a confidence interval from 0.74 to 2.11^[1].

Why the design needs careful reading

Matching can make observed groups more comparable than an unadjusted review of routine records. It cannot make treatment assignment random. Claims data may not fully capture factors that influence prescribing, such as clinical reasoning, access, changes in body weight, lifestyle support, or a person's reasons for switching or stopping treatment. Those factors can still affect an observed association.

The number of cardiovascular events was also small relative to the cohort size. Fifty-seven events can be enough to estimate a result, but the wide confidence interval means the study does not settle whether either medicine differs on cardiovascular risk in this population. The authors similarly called for more research on comparative cardiovascular effectiveness^[1].

This paper should not be read as a reason to start, stop, or switch a medicine. It does not compare dose schedules, assess every safety outcome, or establish the best choice for an individual. A clinician weighing these medicines will consider approved indication, medical history, other treatments, tolerability, access, and the evidence relevant to the condition being treated.

What the published comment changes

The October 2026 letter confirms that the paper is receiving scholarly attention. However, a title alone does not reveal whether a comment challenges the data source, the statistical choices, the interpretation, or something else. Treating the existence of a comment as proof that the study is wrong would be as misleading as treating the original association as proof of superiority^[2].

The practical takeaway remains narrow. This claims-database analysis offers a hypothesis-generating comparison between semaglutide and liraglutide in a defined population. It adds evidence to a question that randomized trials have not directly answered here. It does not replace randomized evidence or individual medical care. Readers can review the research context for semaglutide.

What we do not know yet

The study cannot rule out unrecorded differences between the matched groups. It does not show why the diabetes-incidence association appeared after six months but not earlier. It does not establish a cardiovascular advantage for either medicine, and its findings may not transfer to people with different medical histories, care systems, or reasons for treatment. Direct comparative trials with longer follow-up would give stronger evidence about cause and effect^[1].

Medical disclaimer: This article is for educational and journalistic purposes only and does not constitute medical advice. Peptides discussed may be classified as prescription medicines or research chemicals depending on your jurisdiction. Always consult a qualified healthcare professional before using any peptide product. PeptideMethods.com does not sell, distribute, or facilitate the sale of any peptide product.