Scientific deep-dive
GLP-1 Drugs and Addiction Hospitalizations: What the Largest Cohorts Show
Large cohorts link GLP-1 drugs to fewer substance-related admissions and overdoses, but results shift with the comparison drug and fade after stopping. A review of the observational evidence and what it cannot show.
In the largest cohorts of people with alcohol or other substance use disorders, GLP-1 drugs are linked to fewer hospital admissions: a 26% lower hazard of substance-related admission among 81,617 US veterans with type 2 diabetes[1], and a 36% lower hazard of alcohol-related admission during semaglutide use in a Swedish national cohort[2]. These are associations from health records, and not every analysis finds them: a separate veterans study comparing GLP-1 drugs with DPP-4 inhibitors found no significant difference in alcohol-related admissions[5].
About this article
This review stays on hard endpoints: admissions, emergency visits, overdoses and deaths. For drinking and craving outcomes, see the trial evidence in alcohol use disorder. Every cohort below is observational, so none can show that a GLP-1 drug caused the difference it reports.
Hospital admissions: what each cohort found
| Study | Population | Compared with | Result (95% CI) |
|---|---|---|---|
| Cai, BMJ 2026 | 81,617 US veterans with type 2 diabetes and a prior substance use disorder | SGLT-2 inhibitor starters, up to 3 years | Substance-related admission: HR 0.74 (0.65 to 0.85) |
| Lähteenvuo, JAMA Psychiatry 2025 | 227,866 Swedes aged 16 to 64 with alcohol use disorder; 4,321 used semaglutide, 2,509 liraglutide | The same person's periods off the drug | Alcohol admission: semaglutide aHR 0.64 (0.50 to 0.83), liraglutide aHR 0.72 (0.57 to 0.92) |
| Bach, Lancet Psychiatry 2026 | 167,026 Stockholm County residents with type 2 diabetes; 41,109 received a GLP-1 drug | The same person's unexposed periods | Alcohol admission: RR 0.55 (0.43 to 0.70). Any substance: RR 0.61 (0.50 to 0.74) |
| Rodriguez, BMJ Open 2026 | 40,703 US adults with alcohol use disorder and type 2 diabetes or obesity, starting semaglutide, tirzepatide or a comparator drug | Four sets of active comparators, 1 year | Alcohol-related emergency visit or admission: HR 0.74 (0.62 to 0.89) vs sulfonylureas; 0.78 (0.65 to 0.92) vs other diabetes drugs; 0.68 (0.54 to 0.85) vs other obesity drugs |
| Al-Moussally, Drugs 2025 | 7,302 matched pairs of US veterans with alcohol use disorder and type 2 diabetes | DPP-4 inhibitor starters | Alcohol admission: 1.4% vs 1.7%, OR 0.85 (0.65 to 1.11), not significant |
| Dai, Obesity 2026 | US Medicare adults aged 65 and over with type 2 diabetes and a substance use disorder | DPP-4 inhibitors (n=4,920); SGLT-2 inhibitors (n=4,620) | Any substance admission vs DPP-4: HR 0.76 (0.67 to 0.86). No significant difference vs SGLT-2 |
The largest active-comparator analysis here comes from the US Department of Veterans Affairs. From 606,434 veterans with type 2 diabetes, it followed 81,617 who already had a substance use disorder and started either a GLP-1 drug (16,768) or an SGLT-2 inhibitor (64,849)[1]. GLP-1 starters had lower hazards of substance-related emergency visits (HR 0.69, 95% CI 0.61 to 0.78), admissions (HR 0.74, 0.65 to 0.85) and substance-related death (HR 0.50, 0.32 to 0.79)[1]. The absolute gaps were small: the net three-year risk difference was 6.23 per 1,000 people for admissions and 1.52 per 1,000 for death[1].
The two Swedish studies ask a different question. Each person serves as their own control, so traits that stay fixed, such as genetics or lifelong drinking history, cannot explain the gap. In the national cohort, use of any alcohol use disorder medication barely moved admissions (aHR 0.98, 95% CI 0.96 to 1.00)[2]. Combination questions about those medicines are covered in the review of GLP-1 drugs with naltrexone. The same cohort also saw fewer admissions for physical illness on semaglutide (aHR 0.78, 0.68 to 0.90) and liraglutide (aHR 0.79, 0.69 to 0.91)[2]. That could be a benefit in its own right, or a sign that periods on the drug were healthier periods overall. The design cannot separate those readings.
The comparator changes the answer
Results shift with the comparison drug, and they do not shift in one tidy direction. Against SGLT-2 inhibitors, the veterans study found fewer admissions[1], while the Medicare study of adults 65 and over found no significant difference[6]. Against DPP-4 inhibitors, the Medicare study found fewer admissions for any substance (HR 0.76, 95% CI 0.67 to 0.86), but its alcohol-specific result was not significant (HR 0.76, 0.53 to 1.08)[6], in line with the veterans study of alcohol-related admissions[5]. Follow-up, age, drug era and outcome definitions all differ, so these are not direct contradictions, but they argue against reading any one hazard ratio as the effect size.
The steepest numbers deserve the most caution. Against medications for alcohol use disorder, the US multi-system study reported HRs of 0.37 (95% CI 0.29 to 0.46) in people with type 2 diabetes and 0.35 (0.26 to 0.47) in people with obesity[4]. People prescribed naltrexone or similar drugs may differ in severity or stage of treatment from people prescribed a weight-loss or diabetes drug, and matching can only address what the records capture. The study’s own check supports that worry. Non-alcohol hospitalizations, where no real effect would be expected, were also lower in these two comparisons (HR 0.70, 95% CI 0.53 to 0.93, and HR 0.54, 0.41 to 0.73), and more than 80% of comparator patients were censored during follow-up. The authors call this evidence of residual confounding and urge greater caution with these results. The same check showed no significant difference in the comparisons against diabetes and obesity drugs[4]. Its data came from Truveta, some authors are current or former Truveta employees, and the first author now works at Eli Lilly, with the disclosure stating that Lilly had no role[4].
What happens after stopping
The Stockholm study is the one that looked past the end of treatment. For alcohol-related admissions, the lower rate held during the first 182 days after stopping (RR 0.70, 95% CI 0.50 to 0.98) and was gone by days 183 to 364 (RR 1.07, 0.71 to 1.63)[3]. For any substance use disorder, neither window reached significance (days 1 to 182: RR 0.79, 0.61 to 1.02; days 183 to 364: RR 0.89, 0.63 to 1.26)[3]. The later estimates sit near 1 and do not show a rebound above the unexposed rate, though their intervals are wide. The authors concluded that the findings support monitoring substance use when treatment stops[3].
Overdose and death in opioid use disorder
| Study | Population | Compared with | Overdose result (95% CI where the source gives one) |
|---|---|---|---|
| Qeadan, Addiction 2025 | 503,747 US patients with opioid use disorder | No GIP or GLP-1 prescription | Opioid overdose: aIRR 0.60 (0.43 to 0.83) |
| Wang, JAMA Netw Open 2024 | 33,006 US patients with type 2 diabetes and opioid use disorder; 3,034 on semaglutide | Each other diabetes drug class, 1 year | Opioid overdose: HRs from 0.32 (0.12 to 0.89) to 0.58 (0.38 to 0.87) |
| Lakhani, J Addict Dis 2026 | 8,052 matched pairs with opioid use disorder | No GLP-1 prescription, up to 5 years | Overdose: 3.0% vs 4.8%, HR 0.81 |
| Kennedy, J Clin Psychiatry 2026 | US veterans with type 2 diabetes and opioid use disorder starting semaglutide or tirzepatide | Five diabetes drug classes, 12 months | All-cause overdose: HR 0.32 vs insulin, 0.25 vs SGLT-2; not significant vs metformin, sulfonylureas or DPP-4 |
| Cai, BMJ 2026 | 81,617 US veterans with a prior substance use disorder of any type | SGLT-2 inhibitors, up to 3 years | Drug overdose: HR 0.61 (0.42 to 0.88) |
The veterans overdose study is the most useful replication because it is mixed. Matched samples were small, from 432 to 1,016 people. The drugs were tied to fewer overdoses against insulin (HR 0.32, 95% CI 0.14 to 0.71) and SGLT-2 inhibitors (HR 0.25, 0.09 to 0.68), and showed no significant difference against metformin (HR 0.75, 0.38 to 1.45), sulfonylureas (HR 0.82, 0.33 to 1.96) or DPP-4 inhibitors (HR 1.20, 0.52 to 2.78)[10]. A review of the field calls evidence beyond alcohol and tobacco, and tirzepatide-specific evidence, preliminary[14].
The largest number in the table is also the loosest comparison. The 503,747-patient study compared people with at least one prescription against people with none[7], rather than against people starting a similar drug. A separate study of 473,906 people with opioid use disorder reported lower five-year mortality with a GLP-1 prescription (4.9% vs 10.9%, HR 0.63), but only 8,737 of them received one[9], and the abstract gives no interval. In 2,314 methadone-treated adults with diabetes, GLP-1 use was associated with more opioid use disorder remission (HR 1.75, 95% CI 1.24 to 2.47)[11]. How these drugs fit with standard treatment is covered in GLP-1 drugs alongside buprenorphine and methadone.
Why large numbers do not settle it
- Who gets prescribed. Someone in an active crisis, or recently hospitalized for one, may be less likely to start a weekly injection for weight or diabetes. That can make GLP-1 use look protective, and within-person designs do not fully remove it because it changes over time.
- The cohorts are not independent. Three of the studies above draw on TriNetX[8][9][11] and three on veterans' records[1][5][10], and the two Swedish studies use national registers over overlapping years, so they may include some of the same people[2][3]. Agreement among them counts for less than the study count suggests.
- Who was studied. Most cohorts required type 2 diabetes or obesity, and the Stockholm group had a mean age near 64[3]. The results say little about people with a substance use disorder and neither condition.
- Records miss events. An admission or overdose appears only if it reaches a participating health system and is coded as such. Overdose deaths outside a hospital can be a particular gap in electronic records.
What randomized evidence adds
The strongest randomized test here did not measure admissions. In a 26-week, single-center Danish trial, 108 treatment-seeking adults with moderate to severe alcohol use disorder and obesity received weekly semaglutide 2.4 mg or placebo on top of cognitive behavioral therapy. In an intention-to-treat analysis with multiple imputation, heavy drinking days fell 41.1 percentage points on semaglutide and 26.4 on placebo, an estimated treatment difference of 13.7 points favoring semaglutide (95% CI 5.4 to 22.0)[12]. The Novo Nordisk Foundation was among the funders[12]. A trial that size cannot speak to rare events such as overdose.
A ClinicalTrials.gov review through July 2025 found 33 registered trials: 15 in alcohol use disorder, 9 in nicotine or tobacco, 4 in cocaine, 4 in opioids, 1 in methamphetamine and none in cannabis[13]. Those trials were selected for substance-use measures such as urine toxicology and craving[13], so they should not be expected to settle whether admissions or overdoses fall.
What to raise with a prescriber
Substance use disorder is not among the indications in the Wegovy label (revised June 2026) or the Zepbound label (revised August 2026)[15][16]. If you have a substance use disorder and qualify for a GLP-1 drug for diabetes or weight, ask whether your addiction and prescribing clinicians are coordinating, how it fits with any medication you already take for addiction, and what monitoring is planned if the GLP-1 stops. These cohorts offer no basis for replacing an established addiction treatment with a GLP-1 drug.
Frequently Asked Questions
References
- 1.Cai M, Choi T, Xie Y, Al-Aly Z. Glucagon-like peptide-1 receptor agonists and risk of substance use disorders among US veterans with type 2 diabetes: cohort study. BMJ. 2026. PMID: 41781010.
- 2.Lähteenvuo M, Tiihonen J, Solismaa A, Tanskanen A, Mittendorfer-Rutz E, et al. Repurposing Semaglutide and Liraglutide for Alcohol Use Disorder. JAMA Psychiatry. 2025. PMID: 39535805.
- 3.Bach P, Franck J, Hällgren J, Widing H, et al. Association of GLP-1 receptor agonists with alcohol use disorder-related and substance use disorder-related hospital admissions during treatment and after discontinuation: a Swedish register-based within-individual observational study. Lancet Psychiatry. 2026. PMID: 42456704.
- 4.Rodriguez PJ, Lusk JB, Mehta HB, Levy JF, et al. Association between GLP-1 receptor agonists and alcohol-related hospitalisations among adults with alcohol use disorder: multi-target trial emulation study. BMJ Open. 2026. PMID: 42481079.
- 5.Al-Moussally F, Khan S, Katukuri V, Kinaan M, Mansi IA, et al. Association of Glucagon-Like Peptide-1 Receptor Agonist with Progression to Liver Cirrhosis and Alcohol-Related Admissions in Patients with Alcohol Use Disorder and Diabetes: A Retrospective Cohort Study. Drugs. 2025. PMID: 40223043.
- 6.Dai H, Radwan RM, Scheiffele GD, Tang H, Sheer A, et al. GLP-1 Receptor Agonists and Substance Use Disorders in Older Adults With Type 2 Diabetes: A Target Trial Emulation. Obesity (Silver Spring). 2026. PMID: 40954547.
- 7.Qeadan F, McCunn A, Tingey B. The association between glucose-dependent insulinotropic polypeptide and/or glucagon-like peptide-1 receptor agonist prescriptions and substance-related outcomes in patients with opioid and alcohol use disorders: A real-world data analysis. Addiction. 2025. PMID: 39415416.
- 8.Wang W, Volkow ND, Wang Q, Berger NA, Davis PB, et al. Semaglutide and Opioid Overdose Risk in Patients With Type 2 Diabetes and Opioid Use Disorder. JAMA Netw Open. 2024. PMID: 39320894.
- 9.Lakhani DA, Zheng W. Association of GLP-1 receptor agonist use with long-term outcomes in patients with opioid use disorder. J Addict Dis. 2026. PMID: 42116696.
- 10.Kennedy KP, Bremseth-Vining R, Lynch KG, Reville N, et al. GLP-1 Receptor Agonists and All-Cause Overdose Risk in Veterans With Type 2 Diabetes and Opioid Use Disorder. J Clin Psychiatry. 2026. PMID: 42474323.
- 11.Pan JY, Elman I, Panchal K, Ramsey KM, Albanese M, et al. GLP-1 Receptor Agonist Treatment and Health Outcomes in Methadone-Treated Patients with Opioid Use Disorder and Diabetes. J Gen Intern Med. 2026. PMID: 42151525.
- 12.Klausen MK, Justesen SK, Pedersen JN, Rasmussen L, Jensen A, et al. Once-weekly semaglutide versus placebo in patients with alcohol use disorder and comorbid obesity: a randomised, double-blind, placebo-controlled trial. Lancet. 2026. PMID: 42070571.
- 13.Patil S, Jha N, Jha MK. Glucagon-like peptide 1 receptor agonists in substance use disorders: A systematic review of ClinicalTrials.Gov. Addict Behav Rep. 2026. PMID: 41696398.
- 14.De Filippo O, Bruno F, Giacobbe F, Beccuti G, Braia V, et al. Incretin-based therapies, alcohol and tobacco use, and acute substance-related events: emerging implications for cardiovascular medicine. J Cardiovasc Pharmacol. 2026. PMID: 42520399.
- 15.Novo Nordisk. WEGOVY (semaglutide) prescribing information, revised June 2026. DailyMed, US National Library of Medicine. 2026. https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=ee06186f-2aa3-4990-a760-757579d8f77b
- 16.Eli Lilly and Company. ZEPBOUND (tirzepatide) prescribing information, revised August 2026. DailyMed, US National Library of Medicine. 2026. https://dailymed.nlm.nih.gov/dailymed/lookup.cfm?setid=487cd7e7-434c-4925-99fa-aa80b1cc776b
Related research
GLP-1 Drugs and Alcohol Use Disorder: What the Trial Evidence Actually Shows in 2026
Semaglutide's effect on alcohol use disorder has thin trial support so far: three small randomized trials (n=48, 108, and 50). Hendershot 2025 reduced drinks per drinking day but not drinking days. Naltrexone, acamprosate, and disulfiram remain first-line.
13 min read
GLP-1 and Addiction Circuitry: Alcohol, Nicotine, More (2026)
GLP-1 drugs act on the brain reward circuit behind cravings. Honest 2026 evidence for alcohol and nicotine: strong in animals, early in humans, not approved.
12 min read
NAD+ IV for Addiction Recovery: Does It Actually Work?
IV NAD+ is marketed for opioid, alcohol, and benzodiazepine withdrawal — but there are zero published RCTs. The only clinical data is an uncontrolled 50-patient pilot with a conflict of interest. An honest review.
12 min read
7-Day GLP-1 Meal Plan: A Structured Framework (Not a Recipe Card)
A structured 7-day GLP-1 meal-plan framework — daily protein, fiber, and hydration targets and a cumulative habit-by-habit structure grounded in 2026 GLP-1 nutrition research, not a fabricated recipe card.
8 min read
Air Bubbles in Your GLP-1 Pen or Syringe: Do They Matter?
A small air bubble in your Ozempic, Wegovy, Mounjaro, or Zepbound pen is normal and safe. Here's why, plus when bubbles matter for compounded syringes.
6 min read
Best Apps for Tracking Protein, Water & Weight on a GLP-1 (2026)
On a GLP-1 the things worth tracking shift to protein, fiber, water, your weekly dose, and weight trend. The features to look for in a tracking app — and the categories that fit GLP-1 users best.
6 min read
Where to get tirzepatide (Mounjaro / Zepbound) safely: vetted online providers
Vetted telehealth providers that prescribe online. We compare pricing, form, and states served.
No insurance needed · vetted by our editors
WeightLossRankings.org is reader-supported. When you buy through links on our site, we may earn an affiliate commission. How we make money
Oak
Compounded semaglutide and tirzepatide at a low starting price, with a full price list before you sign up
Read full review →From $250/mo (starting dose)
Get started →From $149/mo
Get started →From $209/mo (2-month plan rate)
Get started →