GLP-1 Drugs and Alcohol Hospitalisation

A 2026 electronic-health-record study found that adults with alcohol use disorder who started semaglutide or tirzepatide had fewer alcohol-related emergency visits or hospitalisations than several active-comparator groups. The association appeared across four target-trial emulations. This was still an observational study. It cannot prove that the medicines prevented hospitalisation or support prescribing them as alcohol use disorder treatment.^[1]

The study included 40,703 adults in US healthcare systems who had alcohol use disorder plus type 2 diabetes or obesity. Researchers compared new users of a newer GLP-1-based medicine with people starting diabetes medicines, obesity medicines, or approved medications for alcohol use disorder. Follow-up lasted up to one year while patients remained on treatment.^[1]

How the target trials were built

Target-trial emulation starts by defining the randomized trial researchers would like to run, then approximates it with existing records. Here, the investigators specified eligibility, treatment initiation, comparators and follow-up for four clinical settings. They used propensity-score weighting or matching to balance recorded patient characteristics and Cox models to estimate the time to a first alcohol-related acute-care encounter.^[1]

The newer-medicine group combined semaglutide and tirzepatide. That improves sample size but prevents the headline estimate from establishing that both medicines have the same association. Tirzepatide activates GIP and GLP-1 receptors, while semaglutide targets the GLP-1 receptor. The study was designed around the combined newer-treatment category.^[1]

The outcome included emergency-department or inpatient encounters with an alcohol-related diagnosis or alcohol testing. Some visits may therefore have involved alcohol without alcohol being the principal reason for admission. A sensitivity analysis restricted the outcome to diagnostic codes, but routine records can still miss alcohol use disorder or record it inconsistently.^[1]

What the four comparisons found

Among people with type 2 diabetes starting metabolic treatment, newer GLP-1-based medicines were associated with a lower hazard of alcohol-related hospitalisation than sulfonylureas, with a hazard ratio of 0.74, and than other diabetes medicines, with a hazard ratio of 0.78. The analysis did not detect a difference against older GLP-1 receptor agonists, where the hazard ratio was 1.09 and the confidence interval crossed 1.^[1]

In the obesity-medication comparison, the hazard ratio versus other anti-obesity medicines was 0.68, with a 95% confidence interval from 0.54 to 0.85. Again, the comparison with older GLP-1 receptor agonists did not show a detected difference. These active-comparator results are more informative than comparing treated patients with untreated people, but unmeasured differences can remain.^[1]

The largest associations appeared when newer GLP-1-based medicines were compared with approved alcohol use disorder medications. Hazard ratios were 0.37 in the type 2 diabetes cohort and 0.35 in the obesity cohort. Those figures should be read with extra caution. The paper's negative-control analysis also found fewer non-alcohol-related hospitalisations in these comparisons, a result that points to residual confounding rather than an alcohol-specific effect alone.^[1]

Why the negative control changes the reading

A negative control is an outcome that should not plausibly change because of the treatment under study. The researchers used non-alcohol-related hospitalisation. There was no detected difference on that outcome in the diabetes- and obesity-medication trials, which supports greater specificity in those comparisons. In the alcohol-medication trials, however, the newer GLP-1 group also had fewer non-alcohol-related admissions.^[1]

That pattern suggests the groups may have differed in ways the records did not fully capture. Access to higher-cost medicines, clinical stability, healthcare engagement and alcohol use disorder severity are possible examples identified by the authors. More than 80% of comparator patients in the alcohol-medication trials were also censored during follow-up, adding another route for bias despite statistical adjustment.^[1]

How the result fits randomized evidence

A separate 2025 phase 2 trial randomly assigned 48 non-treatment-seeking adults with alcohol use disorder to semaglutide or placebo for nine weeks. Semaglutide reduced alcohol consumed in a laboratory self-administration task and lowered drinks per drinking day and weekly craving. It did not reduce average drinks per calendar day or the number of drinking days.^[2]

That small randomized study provides prospective evidence for some drinking outcomes, but it was not powered to test hospitalisation. The 2026 record study is much larger and examines acute-care use, yet treatment was not randomized. The two designs answer different questions and neither establishes a routine alcohol use disorder indication for semaglutide or tirzepatide.^[1][2]

What we do not know yet

The record study did not have a quantitative measure of alcohol use disorder severity and may have misclassified some hospital encounters. It could not remove all differences in income, access, treatment expectations or care-seeking. Its on-treatment design censored people after discontinuation or switching, which can narrow generalisability and introduce bias.^[1]

The association is strong enough to justify larger randomized trials with clinically meaningful outcomes. Those studies would need to separate semaglutide from tirzepatide, define how participants' existing alcohol use disorder care is handled, and track safety alongside drinking and hospitalisation outcomes. Until then, this evidence should not be used to replace approved alcohol use disorder treatment or individualized clinical care.

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.