Once-weekly semaglutide versus placebo in patients with alcohol use disorder and

# Once-weekly semaglutide versus placebo in patients with alcohol use disorder and comorbid obesity: a randomised, double-blind, placebo-controlled trial *The Lancet 2026* Articles Once-weekly semaglutide versus placebo in patients with alcohol use disorder and comorbid obesity: a randomised, double-blind, placebo-controlled trial Mette Kruse Klausen, Signe Keller Justesen, Julie Niemann Pedersen, Line Rasmussen, Andreas Jensen, Mathias Ebbesen Jensen, Ulla B Knorr, Marianne Lerbæk Bergmann, Jens Juul Holst, Bolette Hartmann, George F Koob, Helene Benveniste, Nora D Volkow, Claus Thorn Ekstrøm, Gitte Moos Knudsen, Tina Vilsbøll, Anders Fink-Jensen Summary Background Alcohol use disorder accounts for 5% of deaths worldwide annually, and there is an urgent need for new Lancet 2026; 407: 1687–98 therapeutic interventions. Preclinical and initial human studies indicate that the GLP-1 receptor agonist semaglutide See Comment page 1658 might reduce alcohol drinking. This study evaluated the ecacy of semaglutide once-weekly in treatment-seeking Mental Health Centre patients with alcohol use disorder and comorbid obesity. Copenhagen, Copenhagen University Hospital–Bispebjerg and Frederiksberg, Methods In a 26-week, single-centre, randomised, double-blinded, placebo-controlled trial, treatment-seeking Copenhagen, Denmark participants with moderate to severe alcohol use disorder and comorbid obesity were assigned (1:1) to receive once- (M K Klausen MD, weekly semaglutide (2·4 mg subcutaneously) or placebo (saline subcutaneously), in addition to standard cognitive S K Justesen MB, J N Pedersen RN, behavioural therapy. The primary endpoint was a reduction in the number of heavy drinking days assessed after L Rasmussen MB, M E Jensen MD, U B Knorr MD, 26 weeks of intervention, analysed with an ANCOVA model. Analysis adhered to the intention-to-treat principle, and Prof A Fink-Jensen DMSc); missing outcome data were addressed using multiple imputations. Safety was assessed in all treated patients. The Department of Pediatrics and trial is registered at ClinicalTrials.gov NCT05895643, and is complete. Adolescent Medicine, Rigshospitalet, Denmark (A Jensen PhD); Department of Findings From June 10, 2023, to Feb 4, 2025, 108 participants (53 women and 55 men) were enrolled, with Clinical Medicine, Faculty of 54 participants in each of the semaglutide and placebo treatment groups, and all were included in the data analysis. Health and Medical Sciences, Overall, 88 participants (81%) completed the full intervention. Semaglutide was associated with a reduction in heavy University of Copenhagen, drinking days (–41·1 percentage points from baseline, 95% CI –48·7 to –33·5) compared with placebo (–26·4, Copenhagen, Denmark (U B Knorr, –34·1 to –18·6; estimated treatment dierence –13·7 percentage points, –22·0 to –5·4; p=0·0015), and had substantial Prof G M Knudsen DMSc, eects on multiple secondary alcohol-related and somatic outcomes. Adverse events were transient, generally mild to Prof T Vilsbøll DMSc, moderate gastrointestinal eects, and occurred more frequently in the semaglutide group. Prof A Fink-Jensen); Department of Biochemistry and Immunology, University Interpretation Semaglutide showed robust therapeutic eects in treatment-seeking participants with obesity and Hospital of Southern Denmark, alcohol use disorder and this trial supports previous preclinical and clinical findings suggesting GLP-1 receptor Vejle, Denmark agonists as a potential novel treatment target for alcohol use disorder. (M L Bergmann MSc Pharm); Department of Biomedical Sciences, Faculty of Health and Funding The Research Foundation, Mental Health Services (Capital Region of Denmark), the Novo Nordisk Medical Sciences, University of Foundation, the Novavi Foundation, the Hartmann Foundation, and the Augustinus Foundation. Copenhagen, Copenhagen, Denmark (Prof J J Holst DMSc, Copyright © 2026 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 B Hartmann PhD); Novo Nordisk Foundation Center for license. Basic Metabolic Research and Faculty of Health and Medical Introduction consumption.4 Importantly, GLP-1 receptor agonists are Sciences, University of Copenhagen, Copenhagen, Alcohol use disorder is a chronic, relapsing brain generally well tolerated and have a favourable safety Denmark (Prof J J Holst); The disorder characterised by loss of control of alcohol profile, with a low risk of hypo glycaemia due to their National Institute on Drug consumption and compulsive use.1 Several behavioural glucose-dependent mechanism.5 The endogenous GLP-1- Abuse, National Institutes of and psychological treatments are available,1 and cognitive peptide is secreted from L-cells in the small intestine and Health, Bethesda, MD, USA (Prof G F Koob PhD); behavioural therapy (CBT) is among the treatments with also synthesised in brain areas implicated in reward and Department of Anesthesiology, the highest empirical support.2 However, despite decades addiction.6 Yale University, New Haven, of research, the US Food and Drug Administration Several GLP-1 receptor agonists have shown significant CT, USA (FDA) has approved only three medications—disulfiram, reductions in alcohol consumption, reward-processing, (Prof H Benveniste DMSc); National Institute on Alcohol acamprosate, and naltrexone1—highlighting the urgent and relapse-like behaviours, demonstrating robust and Abuse and Alcoholism, need for more eective treatments. GLP-1 receptor promising eects in preclinical models of alcohol National Institutes of Health, agonists, approved for the treatment of diabetes and addiction.7,8 In humans, register-based studies have Bethesda, MD, USA obesity,3 have gained wide attention for their eects on reported a lower risk of alcohol-related events or alcohol (Prof N D Volkow MD); Department of Public Health, brain pathways involved with appetite regulation and use disorder diagnosis among individuals treated with a Section of Biostatistics, reward, suggesting potential use for mitigating alcohol GLP-1 receptor agonist.9 Recently, a randomised University of Copenhagen, Articles Copenhagen, Denmark (Prof C T Ekstrøm PhD); Research in context Neurobiology Research Unit, Evidence before this study alcohol cues in reward-related brain regions, suggesting Copenhagen University Hospital Rigshospitalet, No formal literature review was done before commencing this reduced incentive salience. More recently, an RCT investigating Copenhagen, Denmark study; however, we previously published a Review on GLP-1 in the efficacy of the GLP-1 receptor agonist semaglutide in a low (Prof G M Knudsen); Steno addictive disorders in 2022. Alcohol use disorder is a chronic dose in 48 non-treatment-seeking individuals with alcohol use Diabetes Center Copenhagen, brain disorder marked by loss of control over drinking and disorder showed a significant reduction in alcohol intake in a Herlev, Denmark (Prof T Vilsbøll) compulsive use. Despite decades of research, only laboratory self-administration task. To our knowledge to date, Correspondence to: Prof Anders Fink-Jensen, Mental three medications are approved for alcohol use disorder, no other RCTs in treatment-seeking individuals have been Health Centre Copenhagen, underscoring the need for novel treatments. GLP-1 receptor published. Copenhagen University Hospital– agonists, established therapies for diabetes and obesity, have Bispebjerg and Frederiksberg, Added value of this study gained wide attention for their effects on reward pathways and Copenhagen 2000, Denmark This randomised, double-blind, placebo-controlled clinical trial anders.fink-jensen@ appetite regulation, suggesting potential use in mitigating shows, for the first time, to our knowledge, that the GLP-1 regionh.dk alcohol consumption. GLP-1 receptor agonists have shown receptor agonist semaglutide at 2·4 mg once-weekly reduces promising results in preclinical models of alcohol addiction. In alcohol consumption in treatment-seeking patients with humans, register-based studies have reported a lower risk of alcohol use disorder and comorbid obesity (BMI ≥30 kg/m²). alcohol-related events or alcohol use disorder diagnosis among individuals with diabetes or obesity treated with a GLP-1 Implications of all the available evidence receptor agonist. At the time of study initiation, only These data, when added to the growing evidence, demonstrate one randomised controlled trial (RCT) of GLP-1 receptor the potential of GLP-1 receptor agonists as a novel treatment agonists in treatment-seeking patients with alcohol use for alcohol use disorder. However, corroboration with larger disorder showed no overall effect on heavy drinking days. RCTs in patients without obesity is needed to address its However, reductions were observed among participants with a generalisability. BMI greater than 30 kg/m², along with decreased activation to See Online for appendix 1 controlled trial (RCT) that included 48 non-treatment- protocol (appendix 1) was approved by the Ethics seeking participants with alcohol use Committee of the Capital Region of Denmark and the disorder10—characterised by lower alcohol use disorder Danish National Board of Health (EU CT NUMBER: severity, fewer alcohol-related consequences, and less 2023-503371-25-00), and the Danish Data Protection motivation to reduce alcohol consumption11—received Agency (P-2023–187). Data were collected and managed low-dose semaglutide for 2 months. The study showed a using the REDCap.13 The trial was conducted in accordance significant reduction in alcohol consumption in a with the Declaration of Helsinki. Independent good laboratory-based alcohol self-administration task, and clinical practice (GCP) monitoring was provided by the decreases in drinks per drinking day and alcohol craving, regional GCP unit, ensuring compliance with compared with placebo.10 To our knowledge, only International Council for Harmonisation (ICH)-GCP one RCT has evaluated the eects of a GLP-1 guidelines, integrity of trial data, safety evaluations, and receptor agonist (once-weekly exenatide) in treatment- applicable regulatory requirements No patients were seeking patients with alcohol use disorder.12 No overall involved in the design, conduct, or reporting of the trial. See Online for appendix 2 dierence in the number of heavy drinking days was Appendix 1 contains the full protocol, and appendix 2 (p 4) observed, but an exploratory analysis showed that contains minor protocol amendments and minor protocol exenatide reduced alcohol consumption in participants deviations. The trial is registered at ClinicalTrials.gov with a BMI above 30 kg/m². In a subset of participants (NCT05895643), and is complete. undergoing functional MRI, treatment with exenatide compared with placebo decreased reactivity to alcohol Participants cues in reward-related brain areas.12 Consequently, we Participants were recruited via advertisements on a trial For more on this research conducted a 26-week RCT in treatment-seeking patients webpage and a patient recruitment service advertising on service platform see http:// with obesity and alcohol use disorder, aiming to social media, which were approved by the authorities. www.trialtree.com investigate the eect and safety of the GLP-1 Before the screening session, a 20-min telephone receptor agonist semaglutide, 2·4 mg subcutaneously interview was conducted, including an assessment of once-weekly, on alcohol consumption. treatment-seeking status (eg, seeking help to obtain professional help to reduce, control, or stop alcohol use). Methods Eligible participants were alcohol use disorder treatment- Study design seeking and age 18–70 years. At the screening session, a This randomised, single-centre, double-blinded, placebo- diagnostic interview was performed by a medical doctor, controlled, clinical trial was carried out at the Mental and an alcohol use disorder diagnosis was given according Health Center Copenhagen, Copenhagen, Denmark. The to the Diagnostic and Statistical Manual of Mental 1688 Articles Disorders, 5th edition (DSM-5), and alcohol dependence dose escalation every 4 weeks in accordance with the according to the ICD-10. All participants had a BMI of manufacturer’s recommendations, until the maximum 30 kg/m² or higher, an alcohol use disorders identification tolerated dose or the target dose of 2·4 mg was reached, test (AUDIT) score above 15, and a minimum of 6 heavy or placebo in equivalent dose. All participants were drinking days, (ie, ≥60 g for men or ≥48 g for women of oered up to ten standardised CBT sessions delivered by alcohol per day) during 30 days as estimated using the a trained nurse, lasting 45 min, and focused on validated, gold-standard tool, the Timeline Followback motivation, craving, strategies, permissive cognitions, (TLFB) method.14 Key exclusion criteria were severe and relapse prevention regarding alcohol use. No mental disorder, other substance use disorders (other nutritional counselling was provided, and monitoring of than tobacco), a history of diabetes, pancreatitis, or nutritional deficiencies was restricted to assessments of alcohol withdrawal seizures, and current use of vitamin D and vitamin B12 at weeks 0 and 26, with medications targeting alcohol use disorder. A full list of supplementation recommended for low levels. Regular inclusion and exclusion criteria is provided in appendix 2 assessments, including TLFB evaluations and blood (p 5). All participants were informed of their rights both sampling were performed in person and by personnel verbally and in writing by a medical doctor and were masked to treatment assignment at the research facility (a required to have a breath alcohol concentration detailed schedule is shown in appendix 2 pp 6–7). Women below 0·5‰, before providing written informed consent. of childbearing potential were required to use eective Data on gender (male, female, or other) and ethnicity contraception for the duration of the trial and underwent were obtained through participant self-reporting. In pregnancy testing at baseline and if pregnancy was accordance with Danish law, no participants received suspected. A follow-up telephone call to each participant financial incentives to remain in the trial. was conducted by a medical doctor 5 weeks after the end of trial participation to assess any safety concerns. All Randomisation and masking eligible participants were invited to a brain imaging Randomisation used a block design, with block sizes substudy, with MRI scans conducted before the first of two and four, stratified by sex assigned at birth, age injection and again at 26 weeks. Imaging data will be (two strata with a cuto at 40 years), and number of heavy reported elsewhere. The full protocol has been published.15 drinking days at baseline (four strata). Before trial initiation, personnel not involved in other trial activities Outcomes generated the allocation sequence using Sealed Envelope The primary endpoint was the change in heavy drinking For more on the randomisation and uploaded it to the REDCap13 randomisation module, days from baseline to week 26, estimated by the TLFB service see http://www. which implemented the computer-generated block method.14 Previous experience recruiting participants with sealedenvelope.com randomisation sequence to ensure balanced allocation alcohol use disorder has shown that an initial telephone across groups. Upon enrolment, participants were screening can lead to reductions in alcohol consumption.12,16 automatically assigned 1:1 in the REDCap system, to To mitigate this potential bias, baseline alcohol use was receive either once-weekly, subcutaneous semaglutide assessed based on a consecutive 30-day period with the (2·4 mg) or placebo (saline). The automatic assignment highest alcohol consumption—defined as the period with was performed by unmasked personnel who were not the largest number of heavy drinking days and involved in other trial activities. subsequently the greatest total alcohol intake—selected Participants attended weekly sessions to receive the from the 40 days preceding enrolment. This approach was assigned treatment, administered by an unmasked a pragmatic way to mitigate potential bias in the baseline nurse. Participants were wearing blindfolds and assessment and to obtain a baseline measure that more headphones, listening to music, to mask a potential closely reflects pre-enrolment drinking behaviour. audible click sound from the semaglutide subcutaneous Importantly, the same procedure was applied at all injection pen. Placebo injections were administered subsequent assessments (weeks 6, 12, 20, and 26), subcutaneously and matched semaglutide in volume and ensuring consistency across all timepoints, and ensuring needle size. CBT and assessments were performed by that results will still be comparable to other studies. masked personnel, and analyses were conducted by a Secondary endpoints were—from baseline to week 26— masked external statistician. changes in total alcohol consumption (g), number of days without alcohol consumption, number of drinks Procedures per drinking day, alcohol craving (Penn alcohol craving Semaglutide was supplied by the Pharmacy of the Capital scale [PACS] score), screen for harmful alcohol use Region of Copenhagen as prefilled FlexTouch pen (AUDIT and AUDIT for consumption[-C] scores), time to injectors (Wegovy [Novo Nordisk]) in doses of 0·25 mg, relapse, WHO risk drinking levels, screen for harmful 0·5 mg, 1·0 mg, 1·7 mg, and 2·4 mg. Placebo consisted drug use (drug use disorders identification test [DUDIT] of prefilled saline syringes (BD PosiFlush [Becton score), Fagerström Test for Nicotine Dependence (FTND) Dickinson]). Participants initiated treatment with score, plasma liver enzymes (γ-glutamyl transferase and subcutaneous semaglutide 0·25 mg once-weekly, with alanine aminotransferase), plasma mean cell volume, Articles plasma phosphatidyl ethanol, Fibrosis-4 Index for Liver per drinking day. Spearman’s rank correlation test was Fibrosis score, glycated haemoglobin (HbA₁), bodyweight used to assess the relationship between weight loss and c (kg), blood pressure (systolic and diastolic, mm Hg), pulse the endpoints of heavy drinking days and total alcohol (beats per min), waist circumference (cm), and measures consumption. All tests were two-sided with a significance of health and life quality (WHO quality of life brief [QOL- level of 0·05, and 95% CIs were reported for estimated BREF] questionnaire score). A reference list of assessment treatment dierences. No multiplicity adjustments were instruments is available in appendix 2 (p 8). Adverse performed, and no additional covariate adjustments were events were evaluated non-systematically at the scheduled applied, except for the endpoint’s baseline value. However, visits and by self-reporting. For each type of event, the for the primary endpoint, an additional analysis adjusted number and percentage of participants who had that event for sex assigned at birth, age (two levels, age 40 years as a at least once were presented by treatment group. cuto), and number of heavy drinking days at baseline (four levels) was provided. The model fit was evaluated Statistical analysis visually using Q–Q plots and residual plots based on a The sample size was based on an RCT investigating randomly selected imputed dataset. No substantial model psilocybin-assisted therapy for alcohol use disorder, which deviations were observed. However, as a precaution, a showed a 47 percentage point reduction in heavy drinking sensitivity analysis using the Huber–White sandwich days in the intervention group versus 25 percentage points estimator to make inferences more robust to potential in the placebo group.17 Assuming 90% power, a two-sided model misspecification was done. No interim analyses alpha of 0·05, and an estimated standard deviation of were done. Analyses were performed in R (version 4.3.2). 26·4 percentage points, we calculated that 64 participants No data monitoring committee was established; this was a would be required. Because dropout rates of 10–35% are small, single-centre study, without pharmaceutical common in clinical alcohol use disorder trials,18 we industry involvement, and with investigators who saw anticipated a 40% attrition rate and set the target sample participants on a regular basis. The statistical analysis size at 108 participants (54 per group). Analysis adhered to plan (appendix 1) was uploaded to ClinicalTrials.gov a the intention-to-treat principle, including all randomised priori, and the dataset was locked before analysis. participants who received their first injection. The primary endpoint—change in the percentage of heavy drinking Role of the funding source days from baseline to week 26, adjusted for baseline—was The funding sources and the manufacturer of analysed using baseline-adjusted ANCOVA. Secondary semaglutide once-weekly (Wegovy) had no role in study endpoints were analysed using ANCOVA models for design, data collection, data analysis, data interpretation, continuous outcomes and Cox proportional hazards or writing of the report. models for time-to-event outcomes. Subgroup analyses were prespecified for baseline heavy drinking days Results (days 6–11, 12–17, 18–23, and 24–30) and DSM-5 severity From June 10, 2023, to Feb 4, 2025, 302 participants were (mild, moderate, and severe). Missing outcome data were pre-screened for eligibility; of the 135 participants addressed using multiple imputation by predictive mean screened, 108 were enrolled and received at least matching, based on all available repeated measures for one injection, and were included in the final analysis each endpoint. This approach assumed that outcome data (54 per group; table 1). In total, 88 participants (81%) were missing at random in the sense that the probability completed the trial (figure 1). There was no significant of missingness depended on observed data, but not the dierence in time to trial discontinuation between the missing values themselves. In the context of this trial, two groups (and the p value was not adjusted), with missingness could be due to loss to follow-up or 14 patients in the placebo group discontinuing by week 26 withdrawal, factors expected to be related to baseline compared with six in the semaglutide group (hazard drinking patterns, but possibly also to the actual endpoint ratio [HR] 0·39, 95% CI 0·15–1·01; appendix 2 p 12). values. Thus, since the missing at random assumption Baseline analysis showed a gender distribution of cannot be formally tested, pre-planned sensitivity analyses 53 (49%) women and 55 (51%) men overall, and a mean were conducted under more conservative scenarios: age of 52·3 years (SD 9·8). The 108 participants had a (1) complete-case analysis including only individuals with mean of 17·2 heavy drinking days (SD 7·8), an overall values observed at both baseline and endpoint, alcohol intake of 2200·9 g (SD 1128·0) of pure alcohol over (2) imputing missing outcome values at week 26 as a the last 30 consecutive days, an overall AUDIT score return to baseline levels, and (3) imputing missing of 22·8 (SD 4·4), and 92 (85%) fulfilled the criteria for outcome values at week 26 as 50% reduction from severe alcohol use disorder according to DSM-5. baseline. In addition, Cohen’s d eect size, which For the primary endpoint, participants treated with quantifies the magnitude of the dierence between semaglutide had greater reductions in the number of two means relative to the total variability in data, was heavy drinking days compared with the placebo group. calculated for the endpoints of number of heavy drinking The mean change was –41·1 percentage points (95% CI days, total alcohol consumption, and number of drinks –48·7 to –33·5) for the semaglutide group versus 1690 Articles Placebo Semaglutide Placebo Semaglutide (n=54) (n=54) (n=54) (n=54) Sex assigned at birth* (Continued from previous column) Male 28 (52%) 27 (50·0) Heavy drinking days§ 17·2 (7·4) 17·1 (8·2) Female 26 (48%) 27 (50·0) Heavy drinking days randomisation strata Age 6–11 days* 15 (28%) 17 (32%) Mean age, years 51·7 (9·8) 52·8 (9·8) 12–17 days* 14 (26%) 14 (26%) 40 years and younger* 6 (11%) 5 (9%) 18–23 days* 10 (19%) 7 (13%) Older than 40 years* 48 (89%) 49 (91%) 24–30 days* 15 (28%) 16 (30%) Cohabiting or married 29 (54%) 22 (41%) Days without alcohol consumption 8·5 (7·3) 9·1 (7·5) Race or ethnicity Total alcohol consumption, g of alcohol 2246·6 2155·3 Asian 0 4 (7%) per 30 days§ (1091·9) (1171·5) Black or African American 1 (2%) 1 (2%) Phosphatidyl ethanol, µmol/L 0·7 (0·6) 0·5 (0·5) White 53 (98%) 49 (91%) WHO risk drinking level¶ Employed 40 (74%) 39 (72%) Low 1 (1·9%) 1 (1·9%) Education Medium 12 (22·6%) 16 (29·6%) Lower secondary school 3 (6%) 3 (6%) High 23 (42·6%) 19 (35·2%) Upper secondary school 2 (4%) 2 (4%) Very high 18 (33·3%) 18 (33·3%) Vocational education or short-cycle 21 (39%) 14 (26%) Fagerströms test for Nicotine 4·2 (2·8) 3·5 (2·5) higher education Dependence score|| Medium-cycle higher education or 28 (52%) 35 (65%) Cigarettes per day 14·9 (10·1) 13·2 (9·5) higher education Bodyweight, kg 105·2 (15·5) 100·3 (12·0) Previous treatment with a glucagon-like 8 (15%) 5 (9%) BMI, kg/m² 35·2 (4·4) 33·7 (3·3) peptide-1 Glycated haemoglobin** 5·4 (0·4) 5·5 (0·3) Previous pharmacological treatment for alcohol use disorder Data are n (%) or mean (SD). *Randomisation strata. †Total score ranges Disulfiram 9 (17%) 10 (19%) from 0 to 40, with higher scores reflecting higher alcohol dependence. ‡Subscale Acamprosate 5 (9%) 5 (9%) with total score ranging from 0 to 12 focusing solely on alcohol consumption Naltrexone 1 (2%) 1 (2%) patterns. §The 30 consecutive days with highest alcohol use (most heavy drinking days and greatest total intake) within the 40 days before evaluation, measured by Nalmefene 0 0 the Timeline Followback method. ¶Low risk: 1–40 g/day for men, 1–20 g/day for Alcohol use disorders identification test 22·3 (4·5) 23·2 (4·3) women; moderate risk: 41–60 g/day for men, 21–40 g/day for women; high risk: score† 61–100 g/day for men, 41–60 g/day for women; very high risk: >100 g/day for Alcohol use disorders identification test 9·8 (1·3) 9·7 (1·2) men, >60 g/day for women. ||Only individuals who reported current smoking at consumption score‡ baseline (n=31; placebo n=15 and semaglutide n=16) were included; the number of daily cigarettes were self-reported at baseline. **To convert glycated ICD-10: alcohol dependence haemoglobin from percentage to mmol/mol, subtract 2·15 and multiply 3 symptoms 5 (9%) 2 (4%) by 10·929. 4 symptoms 32 (59%) 26 (48%) Table 1: Baseline characteristics 5 symptoms 13 (24%) 21 (39%) 6 symptoms 4 (7%) 5 (9%) vs –2·1 units; estimated dierence –1·5 units Diagnostic and Statistical Manual of Mental Disorders, 5th edition: alcohol use disorder [–2·6 to –0·5]). Improvements were also seen in mean Mild (2–3 symptoms) 0 0 PACS score (–9·2 vs –6·1; estimated dierence –3·1 Moderate (4–5 symptoms) 11 (21%) 5 (9%) [–5·1 to –1·2]), mean AUDIT score (–9·9 vs –6·3; Severe (>5 symptoms) 43 (80%) 49 (91%) estimated dierence –3·3 [–5·5 to –1·1]), mean AUDIT-C (Table 1 continues in next column) score (–4·2 vs –2·7; estimated dierence –1·5 [–2·6 to –0·4]), and overall WHO risk drinking levels (–1·75 vs –1·24; estimated dierence –0·52 –26·4 percentage points (–34·1 to –18·6) for the placebo [–0·89 to –0·16]; table 2). The semaglutide group also group, corresponding to a mean dierence of had a significant 2-level reduction in the WHO risk –13·7 percentage points (95% CI –22·0 to –5·4; p=0·0015; drinking level compared with the placebo group table 2, figure 2A). (appendix 2 p 9). For the secondary alcohol consumption outcomes, Analysis of biomarkers also showed favourable changes semaglutide led to greater improvements than in those with semaglutide, including plasma phosphatidyl receiving placebo across multiple measures. Mean total ethanol (–0·24 µmol/L with semaglutide vs 0·00 µmol/L alcohol consumption decreased (–1550·2 g/30 days with with placebo; estimated dierence –0·28 [–0·41 to –0·15]; semaglutide vs –1025·9 g/30 days with placebo; estimated appendix 2 p 12), γ-glutamyl transferase (–36·0 U/L dierence –467·5 g/30 days [95% CI –739·5 to –195·4]; vs –10·2 U/L; estimated dierence –24·2 [–33·4 to –15·1]), figure 2B) and mean drinks per drinking day (–3·5 units and plasma mean cell volume (–1·3 fL vs 0·4 fL; estimated Articles above upper limit, with the highest value recorded as 302 patients pre-screened for eligibility 123 U/L (normal range 10–65 U/L; table 3). Other biomarkers did not significantly dier between groups (table 2). 167 ineligible 51 medical reasons Analysis of metabolic measures showed favourable 24 BMI <30 kg/m changes of semaglutide, with regard to bodyweight 9 severe psychiatric comorbidity (–11·2 kg with semaglutide vs –2·2 kg with placebo; 7 GLP-1RA exposure within 3 months 4 previous delirium tremens or alcohol- estimated dierence –9·0 [–11·2 to –6·7]; appendix 2 withdrawal seizures p 13), mean percent change in bodyweight (–11·36% 4 severe somatic disease 1 diabetes vs –2·0%; estimated dierence –9·0 [–11·3 to –6·8]), waist 1 age >70 years circumference (–12·1 cm vs –3·8 cm; estimated 1 history of acute or chronic pancreatitis dierence –8·3 [–11·2 to –5·4]), BMI (–3·8 vs –0·7; 46 logistical challenges or not interested 27 logistical challenges estimated dierence –3·1 [–3·8 to –2·3]), and HbA₁ 14 declined participation c (–0·3% vs 0·0%; estimated dierence –0·3 [–0·4 to –0·2]). 4 did not attend screening 1 inability to speak or understand Danish Other metabolic measures did not significantly dier 70 AUD related between groups (table 2). 29 DUDIT ≥2 21 heavy drinking days ≤5 Analysis of patient-reported outcomes from the 19 AUDIT ≤15 WHOQOL-BREF questionnaire showed that 1 concomitant pharmacotherapy for AUD semaglutide improved self-evaluated general health (0·98 with semaglutide vs 0·26 with placebo; estimated 135 enrolled and screened dierence 0·48 [0·13–0·83]) and psychological health (3·86 vs 1·81; estimated dierence 1·51 [0·15–2·88] appendix 2 p 9), but there was no improvement in overall 27 ineligible quality of life or physical health. A sensitivity analysis 13 medical reasons 5 BMI ≤30 kg/m using the Huber–White sandwich estimator was done: 3 glycated haemoglobin ≥48 across the 100 imputations, the use of the sandwich 2 intention to conceive 1 impaired liver performance estimator resulted in an average increase in confidence 1 severe psychiatric comorbidity interval widths of 0·29, which did not change the overall 1 GLP-1RA exposure within 3 months conclusion. 14 AUD related 7 DUDIT ≥2 In our prespecified subgroup analyses, a significant 5 AUDIT ≤15 eect of semaglutide on the reduction in heavy drinking 2 heavy drinking days ≤5 days was detected in the subgroup who reported 12–17 heavy drinking days at baseline compared with the 108 randomly assigned placebo group (–38·7 with semaglutide vs –16·7 with placebo; estimated dierence –22·7 [–33·3 to –12·2]) and in the subgroup having severe alcohol use disorder (–43·2 vs –26·9; estimated dierence –12·0 [–19·4 to 54 assigned placebo 54 assigned semaglutide –4·6]; appendix 2 p 10). The interpretation of these subgroup results is addressed in the Discussion. A 14 lost to follow-up 6 lost to follow-up post-hoc sensitivity analysis of the primary endpoint, 4 logistical challenges 4 side-effects adjusted for sex, age, and number of heavy drinking days 1 side-effects 1 no effect 3 no effect 1 no contact at baseline, showed the robustness of the results 5 semaglutide prescription (appendix 2 p 8). 1 reported to authorities The most common adverse events reported were gastrointestinal symptoms, with a higher incidence in 40 assessed at week 26 48 assessed at week 26 the semaglutide group compared with the placebo group (nausea in 31 [57%] of 54 patients vs four [7%] of 54; loss of appetite in 19 [35%] vs eight [15%]; food 54 included in intention-to-treat analysis 54 included in intention-to-treat analysis aversion in 13 [24%] vs one [2%]; vomiting in eight [15%] vs one [2%]; abdominal pain in 11 [20%] vs four [7%]; Figure 1: Trial profile constipation in 19 [35%] vs nine [17%]; and reflux in Pre-screenings conducted via telephone. AUD=alcohol use disorder. AUDIT=alcohol use disorders identification test. DUDIT=drug use disorders identification test. GLP-1RA=glucagon-like peptide-1 receptor agonist. 15 [28%] vs one [2%]), and fatigue (17 [32%] of 54 patients vs ten [19%] of 54; table 3). The gastrointestinal side- dierence –1·7 [–2·6 to –0·7]; table 2). Elevated amylase eects reported were mainly mild to moderate and was found in the semaglutide group (4·7 U/L vs –0·9 U/L transient; however, five participants (four in the in the placebo group; 5·1 [2·5–7·7]; table 2), and semaglutide group) discontinued the trial due to four participants had asymptomatic elevations of amylase side-eects. 1692 Articles Placebo (n=54) Semaglutide (n=54) Estimated treatment p value difference, placebo vs semaglutide (95% Cl) Self-reported drinking and alcohol scales Heavy drinking days, percentage points –26·4 (–34·1 to –18·6) –41·1 (–48·7 to –33·5) –13·7 (–22·0 to –5·4) 0·0015 (primary endpoint)* Total alcohol consumption, g/30 days* –1025·9 (–1260·0 to –791·1) –1550·2 (–1868·2 to –1232·1) –467·5 (–739·5 to –195·4) <0·0009 Days without alcohol consumption, 27·6 (19·8–35·5) 38·9 (30·6–47·3) 10·1 (–0·0 to 20·2) 0·051 percentage points* Change in drinks per drinking day* –2·1 (–3·1 to –1·1) –3·5 (–4·5 to –2·6) –1·5 (–2·6 to –0·5) 0·0051 Reduction in WHO risk drinking levels† –1·24 (–1·54 to –0·94) –1·75 (–2·04 to –1·45) –0·52 (–0·89 to –0·16) 0·0055 Alcohol craving (Penn alcohol craving scale score)‡ –6·1 (–7·6 to –4·6) –9·2 (–10·8 to –7·6) –3·1 (–5·1 to –1·2) 0·0020 Alcohol use disorders identification test score§ –6·3 (–7·8 to –4·7) –9·9 (–11·6 to –8·2) –3·3 (–5·5 to –1·1) 0·0044 Alcohol use disorders identification test –2·7 (–3·4 to –1·9) –4·2 (–5·0 to –3·5) –1·5 (–2·6 to –0·4) 0·0071 consumption score¶ Alcohol biomarkers Phosphatidyl ethanol, µmol/L 0·00 (–0·10 to 0·11) –0·24 (–0·34 to –0·13) –0·28 (–0·41 to –0·15) <0·0001 Liver and pancreas parameters Alanine aminotransferase, U/L –6·5 (–11·0 to –2·0) –12·5 (–21·2 to –3·9) –3·3 (–11·0 to 4·4) 0·40 γ-glutamyl transferase, U/L –10·2 (–17·5 to –2·9) –36·0 (–52·2 to –19·7) –24·2 (–33·4 to –15·1) <0·0001 Mean corpuscular volume, fL 0·4 (–0·6 to 1·1) –1·3 (–1·9 to –0·6) –1·7 (–2·6 to –0·7) 0·0007 Amylase, U/L –0·9 (–2·7 to 1·0) 4·7 (2·8 to 6·6) 5·1 (2·5 to 7·7) 0·0002 FIB-4|| 0·0 (–0·1 to 0·1) –1·1 (–3·2 to 1·0) 0·0 (–0·1 to 0·2) 0·67 Clinical measures Bodyweight, kg –2·2 (–3·7 to –0·6) –11·2 (–12·9 to –9·6) –9·0 (–11·2 to –6·7) <0·0001 Bodyweight, % –2·0 (–3·4 to –0·5) –11·36 (–13·0 to –9·7) –9·0 (–11·3 to –6·8) <0·0001 Systolic blood pressure, mm Hg –10·6 (–16·2 to –5·0) –14·5 (–19·8 to –9·2) –6·8 (–13·5 to –0·1) 0·047 Diastolic blood pressure, mm Hg –4·5 (–7·7 to –1·3) –5·6 (–8·8 to –2·4) –2·1 (–5·7 to 1·6) 0·26 Pulse, beat per min 2·7 (–0·8 to 6·2) 1·8 (–1·3 to 5·0) –0·7 (–4·8 to 3·4) 0·74 Waist circumference, cm –3·8 (–6·0 to –1·5) –12·1 (–14·2 to –10·0) –8·3 (–11·2 to –5·4) <0·0001 BMI, kg/m² –0·7 (–1·2 to –0·2) –3·8 (–4·4 to –3·2) –3·1 (–3·8 to –2·3) <0·0001 Glucose metabolism Glycated haemoglobin**, % 0·0 (–0·0 to 0·0) –0·3 (–0·4 to –0·2) –0·3 (–0·4 to –0·2) <0·0001 Other drugs Drug use disorders identification test score†† 0·2 (–0·1 to 0·5) 0·1 (–0·1 to –0·2) –0·1 (–0·4 to 0·3) 0·77 FTND score‡‡ –0·7 (–1·6 to 0·3) –0·1 (–0·6 to 0·4) 0·3 (–0·9 to 1·5) 0·60 Number of cigarettes per day‡‡ –1·8 (–3·5 to –0·1) –2·5 (–5·6 to 0·7) –0·5 (–3·7 to 2·7) 0·75 Data are mean (95% CI) unless otherwise stated. FIB-4=fibrosis-4 index. FTND=Fagerströms test for nicotine dependence. *The 30 consecutive days with highest alcohol use (most heavy drinking days and greatest total intake) within the 40 days before evaluation, measured by the Timeline Followback method. †Low risk: 1–40 g/day for men, 1–20 g/day for women; moderate risk: 41–60 g/day for men, 21–40 g/day for women; high risk: 61–100 g/day for men, 41–60 g/day for women; very high risk: >100 g/day for men, >60 g/day for women. ‡Total scores ranges from 0 to 30, with higher scores reflecting higher alcohol craving. §Total scores range from 0 to 40, with higher scores reflecting greater likelihood of alcohol-related problems. ¶Subscale with total score ranging from 0 to 12 focusing solely on alcohol consumption patterns. ||FIB-4=(age × aspartate aminotransferase) / (platelets × √alanine aminotransferase); non-invasive liver fibrosis estimate, higher scores indicate greater risk. ** To convert glycated haemoglobin from percentage to mmol/mol, subtract 2·15 and multiply by 10·929. ††Total score ranges from 0 to 44, with higher scores indicating greater likelihood of drug-related problems. ‡‡Only individuals who reported current smoking at baseline (n=31; placebo n=15 and semaglutide n=16) were included; a lower FTND score indicates lower nicotine dependence; the number of daily cigarettes were self-reported at baseline and at week 26. Table 2: Change in endpoints from baseline to week 26 The only serious adverse event recorded in the symptoms. By the 5-week post-trial safety follow-up call, semaglutide group was one hospital admission due to all serious adverse events, adverse events, and adverse abdominal pain (semaglutide 0·25 mg), without trial reactions had resolved. discontinuation. In the placebo group, one participant was The mean number of injections given was distributed admitted for alcohol withdrawal symptoms, one participant equally, with 24·7 (SD 1·4) in the semaglutide group, was treated at hospital for an eye issue, and one patient versus 24·2 (1·8) in the placebo group (appendix 2 p 10). was admitted for acute coronary syndrome observation, In the semaglutide group, three participants reached a without trial discontinuation (table 3). No other maximum tolerable once-weekly dose at 0·5 mg, participants had treatment-requiring alcohol withdrawal two participants at 1·0 mg, five participants at 1·7 mg, and Articles 80 40 0 Figure 2: Effects of once-weekly semaglutide compared with placebo on reduction in heavy drinking days and total alcohol consumption Changes from baseline (week 0) to follow-up (week 26) are shown for mean percent heavy drinking days in the last 30 days (A), and mean alcohol (g) consumed during the past 30 days (B), in participants randomly assigned to semaglutide 2·4 mg (n=54) or placebo (n=54). Participant level changes by treatment group are in appendix 2 (p 15). SEM=standard error of the mean. the remaining participants at 2·4 mg. The placebo group completed the titration regimen corresponding to the 2·4 mg semaglutide dose. Analysis of the plasma semaglutide concentrations in patients without detectable plasma semaglutide at baseline showed 59·2 pmol/mL in the semaglutide group versus 0·00 pmol/mL in the placebo group (59·3 [95% CI 52·5 to 66·1; appendix 2 p 10), which is comparable to plasma semaglutide levels in other clinical trials.19 The number of therapy sessions received was similar in the two groups, with a mean of 6·8 (SD 2·3) sessions in the semaglutide group versus 6·6 (2·2) sessions in the Discussion placebo group (appendix 2 p 11). To our knowledge, this is the first RCT to investigate the Post-hoc analysis showed an eect size in the moderate eects of the GLP-1 receptor agonist semaglutide in range on heavy drinking days (Cohen’s d=0·57 [95% CI treatment-seeking patients with alcohol use disorder, and 0·14–0·99]), total alcohol consumption (0·54 [0·11–0·96]), it showed significant eects on multiple alcohol-related and number of drinks per drinking days (0·45 [0·02–0·88]). and somatic outcomes, aligning with preclinical evidence The number needed to treat for semaglutide according to and registry studies.7 the 2-level decrease in WHO risk drinking level was 4·3 Recent advances in pharmacotherapy have shifted focus (95% CI 2·33–30·3). Spearman’s rank correlations from strict abstinence to harm reduction via reduced indicated a significant association between heavy drinking alcohol consumption, and WHO drinking risk levels are days versus weight loss (ρ=–0·40, p=0·0038) in the now accepted by the FDA as the primary endpoint in semaglutide group (figure 3A), but not in the placebo alcohol RCTs.20 Our finding of a reduction in WHO risk group (ρ=0·03, p=0·85; figure 3B). drinking levels, and the 2-level reduction with semaglutide 1694 syad gniknird yvaeh tnecrep naeM % ,)MES( syad tsal eht ni A 2000 1000 0 0 6 12 20 26 )MES( syad tsal eht ni )g( lohocla naeM Placebo Semaglutide Placebo (n=54) (n=54) Semaglutide Serious adverse events Admitted with abdominal pain* 0 1 (2%) Hospitalised due to withdrawal symptoms 1 (2%) 0 Posterior vitreous detachment 1 (2%) 0 Hospitalised for acute coronary syndrome 1 (2%) 0 observation Adverse events and adverse reactions Nausea 4 (7%) 31 (57%) Loss of appetite 8 (15%) 19 (35%) Food aversion 1 (2%) 13 (24%) B Vomiting 1 (2%) 8 (15%) Abdominal pain 4 (7%) 11 (20%) Diarrhoea 11 (20%) 15 (28%) Constipation 9 (17%) 19 (35%) Reflux 1 (2%) 15 (28%) Abdominal bloating 3 (6%) 2 (4%) Increased belching 0 2 (4%) Fatigue 10 (19%) 17 (32%) Generalised itching 1 (2%) 3 (6%) Headache 10 (19%) 11 (21%) Dizziness 4 (7%) 7 (13%) Injection site reaction 2 (4%) 1 (2%) Week Outpatient detoxification 1 (2%) 0 Elevated plasma amylase† 0 4 (7%) Elevated liver parameters‡ (AST, ALT, 21 (39%) 15 (28%) and GGT) Sleep disturbances 2 (4%) 4 (7%) Depression diagnosed in primary care 1 (2%) 0 Worsening of anxiety level 0 1 (2%) Suicidal thoughts 1 (2%) 0 Fall (no hospitalisation) 2 (4%) 3 (6%) Miscellaneous 21 (39%) 30 (56%) Data are n (%). ALT=alanine aminotransferase. AST=aspartate aminotransferase. GGT=γ-glutamyl transferase. *The only serious adverse event recorded in the semaglutide group was one hospital admission due to unspecific abdominal pain (semaglutide 0·25 mg), but this did not lead to trial discontinuation. †Reference range <105 U/L; highest observed level 123 U/L. ‡Total participants n=24; AST (n=8), highest observed level 125 U/L; ALT (n=12), highest observed level 121 U/L; GGT (n=16), highest observed level 209 U/L. Table 3: Serious adverse events, adverse events, and adverse reactions Articles compared with placebo (appendix 2 p 9), is aligned with this reductionist approach and with the drinking reduction goals of many patients.20 This reduction in WHO risk drinking levels is associated with long-term reduction in alcohol-related negative consequences, improvement in mental health,21 and aligns with the improve ment of the self-evaluated general and psychol ogical health in the present semaglutide group, although these improvements were modest (appendix 2 p 9). –25 Analysis of medication Cohen’s d eect sizes indicated noteworthy findings, with estimates falling within the medium range for number of heavy drinking days, total –50 alcohol consumption, and number of drinks per drinking day. The eect size estimates are clinically meaningful in the context of already established FDA-approved pharma- –75 cotherapies for alcohol use disorder, for which RCTs indicate that FDA-approved agents generally confer small to moderate eects in reducing heavy drinking episodes.22,23 For acamprosate, a meta-analysis reported –100 0 10 20 30 no significant reduction in heavy drinking days compared with placebo,24 and for extended-release naltrexone, a meta-analysis showed a moderate eect, corresponding to approximately 1·2 fewer heavy drinking days per month relative to placebo.25 In a recent review, the ecacy of several therapeutics for alcohol use disorder was evaluated with respect to the number needed to treat for a 2-level decrease in the WHO risk level endpoint.20 On the basis of the present data (appendix 2 p 9), the number needed to treat for semaglutide is 4·3, indicating that semaglutide could be more ecacious than approved medications for alcohol use disorder, for which the number needed to treat is 7 or higher.20 The precise mechanisms of action of GLP-1 receptor agonists remain incompletely understood. Preclinical studies have reported that stimulation of brain GLP-1 receptors is a prerequisite for GLP-1 RA-induced reduction in alcohol consumption,26 and in an earlier paper we have shown that treatment with the GLP-1 RA exenatide reduced alcohol cue-induced activation in reward-related brain areas in patients with alcohol use disorder.12 These findings point towards a central mechanism of action. Figure 3: Change in heavy drinking days correlated with weight loss However, the precise molecular mechanism of action has (A) Changes from baseline (week 0) to follow-up (week 26) are shown for heavy drinking days versus weight loss not been elucidated; the prevailing hypothesis suggests in the semaglutide group. (B) Changes from baseline (week 0) to follow-up (week 26) are shown for heavy that the therapeutic eects arise from overlapping drinking days versus weight loss in the placebo group. neurobiological and peripheral mechanisms that modulate both metabolic regulation and reward-related pathways attributable to the caloric content of alcohol. When shared between obesity and alcohol use disorder.27 In line evaluating changes in beverage preference (appendix 2 with this, we observed a significant association between p 7), we observed a similar shift in self-reported weight loss and reduced alcohol consumption.27 preference patterns across the two groups, indicating no In this cohort, the mean weight loss was comparable to semaglutide-induced change in consumption patterns. that of individuals with obesity, treated with semaglutide.28 As only three individuals in the semaglutide group had The Spearman’s rank correlation indicates an association weight loss of less than 2·5 kg (appendix 2 p 13), it is not between heavy drinking days and weight loss in the possible to conclude whether the potential eects are semaglutide group, but not in the placebo group independent of weight loss. Semaglutide-induced (figure 3). As caloric intake was not assessed, it is not gastrointestinal symptoms are mostly transient. possible to establish whether the observed eect of Consequently, it is unlikely that adverse eects semaglutide on alcohol consumption was primarily (eg, nausea) should account for the reduction in alcohol )%( syad gniknird yvaeh ni egnahC A Change in heavy drinking days (%) vs weight loss (semaglutide group) 0 –25 –50 –75 –100 –10 0 10 20 )%( syad gniknird yvaeh ni egnahC Spearman's rank correlation=–0·40 p=0·0038 B Change in heavy drinking days (%) vs weight loss (placebo group) Spearman's rank correlation=0·03 p=0·85 Weight loss (kg) Articles consumption, as the eects of semaglutide continued conditions and underscore the need for further studies throughout the study (figure 2). in routine clinical practice. Semaglutide reduced alcohol consumption on average, This trial has several limitations. The inclusion but a subset of participants showed little or no response, criterion of BMI of 30 kg/m² or higher limits the even at the highest dose of 2·4 mg once-weekly generalisability of the findings to the entire population of (appendix 2 p 15), paralleling the non-responder pheno- patients with alcohol use disorder. Approximately me non observed in obesity studies28 and highlighting the 890 million adults worldwide (~16%) have obesity influence of individual phenotypes on treatment ecacy.27 (BMI ≥30).30 The findings from this trial provide support We found significant reductions within the subgroup of for broadening the indication for semaglutide to alcohol participants with severe alcohol use disorder and within use disorder in patients with a BMI of 30 kg/m² or the subset with 12–17 heavy drinking days at baseline higher, which could potentially benefit millions of (appendix 2 p 10). However, the number of participants individuals, given that an estimated 8 million adults in diered considerably between the two DSM-5 alcohol use the USA alone have both obesity and high consumption disorder symptom subgroups with 16 in the moderate of alcohol.31 Such targeted approaches are consistent with subgroup and 92 in the severe subgroup (appendix 2 the recognition of alcohol use disorder heterogeneity and p 10). Low statistical power in the moderate symptom the move towards precision medicine.32 subgroup could partly explain why a significant eect of In our trial, the administration pens required weekly semaglutide was detected only in the severe symptom visits for 26 weeks, for all participants, to receive the group. Moreover, the size of the four subgroups defined assigned treatment. To optimise retention, all participants by the baseline number of heavy drinking days varied were oered CBT, consistent with current recommen- from 17 to 32 participants per subgroup, which might dations for alcohol RCTs, in which investigational explain why the overall significant eect could not be treatments are evaluated alongside eective interventions detected in all subgroups despite consistent trends such as psychotherapy.33 Although this design of CBT towards benefit (appendix 2 p 10). It should be stressed oering and weekly visits among all participants likely that dierences in significance in each subgroup in itself increased the placebo response, it also underscores that do not reflect that the estimated treatment dierences our findings show even more robust ecacy of between two groups are necessarily dierent. semaglutide. We found no eect of semaglutide on the FTND score The overall dropout rate was comparable to that of other or daily cigarette consumption, despite preclinical alcohol RCTs;18 more participants in the placebo group reports suggesting a potential eect.29 The absence of discontinued than in the semaglutide group, and this eect might be attributable to the small cohort of pattern aligns with observations from other clinical smokers or the relatively low baseline severity of nicotine obesity trials of semaglutide.28 The primary reasons for dependence. These findings are consistent with data dropout in the placebo group were the initiation of from the exenatide alcohol study,12 and results from RCTs semaglutide treatment from external sources and a in smokers, reporting mixed eects.29 perceived lack of treatment eect. These findings GLP-1 receptor agonists are considered safe; however, underscore the risk of functional unblinding in incretin safety data in individuals with alcohol use disorder therapy trials as, in general, the absence of expected side- remain scarce. On the one hand, weekly visits provided eects or weight loss in placebo recipients and their substantial oversight of safety, but CBT might have presence in active treatment recipients could compromise mitigated some adverse eects. On the other hand, the blinding and lead CBT providers to reinforce treatment focus on and systematic inquiry into adverse eects eects inadvertently. However, in our trial, substantial might have contributed to a higher number of reported heterogeneity in weight loss was observed, with some events. The gastrointestinal side-eects reported were participants in the placebo group losing up to 25 kg not measured systematically (eg, with a visual analogue (appendix 2 p 13), and some participants receiving scale score), so no correlation with alcohol consumption semaglutide had little or no weight loss. Additionally, the can be done. The reported gastrointestinal side-eects high frequency of reported adverse events in the placebo were mostly transient and mild to moderate. When group in this trial (table 3) further hampered the potential compared with other semaglutide RCTs, symptoms unblinding for both participants and study personnel. occurred slightly more frequently for all participants in A key limitation of this study is that no alcohol this trial than in individuals with a comparable BMI but follow-up data were collected after week 26 and, therefore, without alcohol use disorder.28 These adverse events were alcohol consumption after trial termination could not be also more frequent in the placebo group, suggesting that assessed. Future studies should include longer-term this population has an increased gastrointestinal follow-up to evaluate the potential sustained eects of the vulnerability in general. In real-world settings without treatment. Lastly, a limitation was the study population such intensive monitoring, the safety profile might dier. being predominantly White, limiting the generalisability Nonetheless, our findings provide important insight into to a broader population, so greater eort should be made potential adverse eects under closely monitored for diverse inclusion in future trials. 1696 Articles In summary, the eects of semaglutide suggest that (DOI 10.17632/fr522rsvyp.1). The criteria for access to data are a the treatment eect was suciently large to be detected methodologically sound proposal with an approved aim directed to the despite the modest sample size, and self-reported alcohol corresponding author via email. Data will be available for 5 years from publication. consumption was validated with the gold-standard Acknowledgments phosphatidyl ethanol biomarker, further enhancing the We thank all the participants in this trial for their valuable commitment. validity of our findings. We also thank Tugba Kuzey for providing therapy sessions and clinical To our knowledge, this RCT is the first to show that investigations; Sofie Hoppe Soe, Natasja Aaby Randa Nielsen, once-weekly semaglutide reduces heavy drinking days Eva Gunnarsdóttir Justinussen, and Mads-Gustav Nebel Hvidt for assistance with injections; Linnea Rosalinde Koeber for help with clinical and WHO drinking-risk levels in treatment-seeking investigations; Joan Marie Andersen for randomisation of treatment to patients with alcohol use disorder and comorbid obesity. participants; Lene Brus Albaek for performing the semaglutide plasma This finding adds to the growing evidence for use of analysis; Benjamin Graungaard and www.trialtree.com for assistance GLP-1 receptor agonists in alcohol use disorder, with recruitment of participants; and the Department of Clinical Biochemistry, Copenhagen University Hospital–Bispebjerg and supporting an expanded indication for semaglutide, Frederiksberg, Copenhagen, Denmark, where part of the laboratory work potentially aecting millions of individuals, given the was performed. Finally, we thank the funding sources: the Novo Nordisk global burden of alcohol use disorder and comorbid Foundation; the Research Foundation, Mental Health Services, Capital obesity. Importantly, key limitations and safety Region of Denmark; the Novavi Foundation; the Hartmann Foundation; and the Augustinus Foundation. During the preparation of this work the uncertainties do persist, and additional research is authors used ChatGPT for language optimisation. After using this tool/ needed before o-label use can be endorsed. service, the authors reviewed and edited the content as needed and take full responsibility for the content of the published Article. Contributors Conceptualisation: AF-J and MKK. Data curation: AJ, MKK. Statistical References power analysis: AF-J, MKK, and CTE. Statistical analysis plan: MKK, AJ, 1 Rehm J, Assanangkornchai S, Hendershot CS, et al. Alcohol use AF-J, and CTE. Funding acquisition: MKK and AF-J. Clinical disorders. Lancet 2025; 406: 2269–81. investigations: MKK, JNP, LR, and SKJ. Plasma phosphatidyl ethanol 2 Carroll KM, Kiluk BD. Cognitive behavioral interventions for concentration analysis: MLB. Plasma semaglutide concentration alcohol and drug use disorders: through the stage model and back measurements: JJH and BH. Methodology: MKK, AFJ, TV, HB, NDV, again. Psychol Addict Behav 2017; 31: 847–61. GFK, and GMK. MKK was a study physician and subinvestigator, AF-J 3 Drucker DJ. GLP-1 physiology informs the pharmacotherapy of was the principal investigator, and UBK was the institutional head of the obesity. Mol Metab 2022; 57: 101351. department with responsibility for initiating and leading the clinical 4 Leggio L, Hendershot CS, Farokhnia M, et al. GLP-1 receptor trial. Validation of clinical data: AF-J, AJ, MKK, and CTE. Preparation of agonists are promising but unproven treatments for alcohol and figures: AJ, MKK, MEJ, and AF-J. Original manuscript drafting: MKK. substance use disorders. Nat Med 2023; 29: 2993–95. All authors had access to the data, contributed to the review of the 5 Lyseng-Williamson KA. Glucagon-like peptide-1 receptor analogues manuscript, approved the final version, and had final responsibility for in type 2 diabetes: their use and dierential features. Clin Drug Investig 2019; 39: 805–19. the decision to submit for publication. 6 Heppner KM, Kirigiti M, Secher A, et al. Expression and Declaration of interests distribution of glucagon-like peptide-1 receptor mRNA, protein and MKK has served as a consultant for Phamacotherapies for Alcohol and binding in the male nonhuman primate (Macaca mulatta) brain. Substance Use Disorders Alliance and Guidepoint. AFJ has received an Endocrinology 2015; 156: 255–67. unrestricted grant from Novo Nordisk to investigate the eects of 7 Klausen MK, Knudsen GM, Vilsbøll T, Fink-Jensen A. Eects of semaglutide on metabolic disturbances in participants with GLP-1 receptor agonists in alcohol use disorder. schizophrenia treated with antipsychotics; serves on a RCT advisory Basic Clin Pharmacol Toxicol 2025; 136: e70004. panel for Novo Nordisk, receiving no honorarium; and has served as a 8 Klausen MK, Thomsen M, Wortwein G, Fink-Jensen A. The role of consultant for Phamacotherapies for Alcohol and Substance Use glucagon-like peptide 1 (GLP-1) in addictive disorders. Br J Disorders Alliance and Guidepoint. TV has served on scientific advisory Pharmacol 2022; 179: 625–41. panels and been part of speakers bureaus for Amgen, AstraZeneca, 9 Wang W, Volkow ND, Berger NA, Davis PB, Kaelber DC, Xu R. BMS, Boehringer Ingelheim, Eli Lilly, Gilead, GSK, Mundipharma, Novo Associations of semaglutide with incidence and recurrence of Nordisk, Carmot/Roche, Sanofi, Sun Pharmaceuticals, and Zealand alcohol use disorder in real-world population. Nat Commun 2024; 15: 4548. Pharma; and served as a consultant for Amgen, Astra-Zeneca, BMS, Boehringer Ingelheim, Eli Lilly, Gilead, GSK, Mundipharma, Novo 10 Hendershot CS, Bremmer MP, Paladino MB, et al. Once-weekly semaglutide in adults with alcohol use disorder: a randomized Nordisk, Carmot/Roche, Sanofi, Sun Pharmaceuticals, and Zealand clinical trial. JAMA Psychiatry 2025; 82: 395–405. Pharma. JJH has received a grant from the Novo Nordisk Foundation to 11 Rohn MCH, Lee MR, Kleuter SB, Schwandt ML, Falk DE, Leggio L. the Center for Basic Metabolic Research at Copenhagen University Dierences between treatment-seeking and nontreatment-seeking (NNF23SA0084103); has given lectures and received travel support from alcohol-dependent research participants: an exploratory analysis. AstraZeneca, Eli Lilly, and Decheng Capital Global Lifesciences Fund; Alcohol Clin Exp Res 2017; 41: 414–20. has served as a consultant for Third Bridge, Metaphor Biotechnologies, 12 Klausen MK, Jensen ME, Møller M, et al. Exenatide once weekly for Vial Health Technology, Crinetics Pharmaceuticals, Immunic alcohol use disorder investigated in a randomized, placebo- Therapeutics, Guidepoint, Fractyl Health, Alcimed, Sonova Partners, controlled clinical trial. JCI Insight 2022; 7: e159863. and Jeeries International; has received support for attending meetings 13 Harris PA, Taylor R, Minor BL, et al, and the REDCap Consortium. and/or travel from the Endocrine Society and Eli Lilly; participates on an The REDCap consortium: building an international community of advisory board for Novo Nordisk; is a cofounder and board member of software platform partners. J Biomed Inform 2019; 95: 103208. Antag Therapeutics and Bainan Biotech, and sits on the boards of both, 14 Sobell LC, Sobell MB. Timeline follow-back. In: Litten RZ, unpaid; and has stock options with Antag Therapeutics. All other Allen JP, eds. Measuring alcohol consumption. Humana Press, authors declare no competing interests. 1992: 41–72. 15 Klausen MK, Kuzey T, Pedersen JN, et al. Does semaglutide reduce Data sharing alcohol intake in Danish patients with alcohol use disorder and All de-identified individual participant data, except brain imaging data comorbid obesity? Trial protocol of a randomised, double-blinded, and alcohol diaries, and the protocol and statistical analysis plan placebo-controlled clinical trial (the SEMALCO trial). BMJ Open (appendix 1), are available in the Mendeley database from publication 2025; 15: e086454. Articles 16 Kaner EFS, Beyer FR, Muirhead C, et al. Eectiveness of brief 26 Sirohi S, Schurdak JD, Seeley RJ, Benoit SC, Davis JF. Central and alcohol interventions in primary care populations. peripheral glucagon-like peptide-1 receptor signaling dierentially Cochrane Database Syst Rev 2018; 2: CD004148. regulate addictive behaviors. Physiol Behav 2016; 161: 140–44. 17 Bogenschutz MP, Ross S, Bhatt S, et al. Percentage of heavy 27 Leggio L, Farokhnia M, Kenny PJ, Pepino MY, Simmons WK. drinking days following psilocybin-assisted psychotherapy vs Crosstalk between alcohol use disorder and obesity: two sides of the placebo in the treatment of adult patients with alcohol use disorder: same coin? Mol Psychiatry 2025; 30: 5938–52. a randomized clinical trial. JAMA Psychiatry 2022; 79: 953–62. 28 Wilding JPH, Batterham RL, Calanna S, et al, and the STEP 1 Study 18 Hallgren KA, Witkiewitz K. Missing data in alcohol clinical trials: Group. Once-weekly semaglutide in adults with overweight or a comparison of methods. Alcohol Clin Exp Res 2013; 37: 2152–60. obesity. N Engl J Med 2021; 384: 989–1002. 19 Strathe A, Horn DB, Larsen MS, et al. A model-based approach to 29 Herman RJ, Schmidt HD. Targeting GLP-1 receptors to reduce predict individual weight loss with semaglutide in people with nicotine use disorder: preclinical and clinical evidence. overweight or obesity. Diabetes Obes Metab 2023; 25: 3171–80. Physiol Behav 2024; 281: 114565. 20 Witkiewitz K, Anton RF, O’Malley SS, et al, and the Alcohol Clinical 30 WHO. Obesity and overweight. Dec 8, 2025. https://www.who.int/ Trials Initiative (ACTIVE) Workgroup. Reductions in World Health news-room/fact-sheets/detail/obesity-and-overweight (accessed Organization risk drinking levels as a primary ecacy end point for Dec 20, 2025). alcohol clinical trials: a review. JAMA Psychiatry 2025; 82: 1246–54. 31 Raza SA, Sokale IO, Thrift AP. Burden of high-risk phenotype of 21 Witkiewitz K, Kranzler HR, Hallgren KA, et al. Stability of drinking heavy alcohol consumption among obese US population: results reductions and long-term functioning among patients with alcohol from National Health and Nutrition Examination Survey, use disorder. J Gen Intern Med 2021; 36: 404–12. 1999–2020. Lancet Reg Health Am 2023; 23: 100525. 22 Kranzler HR, Hartwell EE. Medications for treating alcohol use 32 Litten RZ, Ryan ML, Falk DE, Reilly M, Fertig JB, Koob GF. disorder: a narrative review. Alcohol Clin Exp Res 2023; 47: 1224–37. Heterogeneity of alcohol use disorder: understanding mechanisms 23 McPheeters M, O’Connor EA, Riley S, et al. Pharmacotherapy for to advance personalized treatment. Alcohol Clin Exp Res 2015; alcohol use disorder: a systematic review and meta-analysis. JAMA 39: 579–84. 2023; 330: 1653–65. 33 Witkiewitz K, Finney JW, Harris AHS, Kivlahan DR, Kranzler HR. 24 Witkiewitz K, Saville K, Hamreus K. Acamprosate for treatment of Recommendations for the design and analysis of treatment trials alcohol dependence: mechanisms, ecacy, and clinical utility. for alcohol use disorders. Alcohol Clin Exp Res 2015; 39: 1557–70. Ther Clin Risk Manag 2012; 8: 45–53. 25 Murphy CE 4th, Wang RC, Montoy JC, Whittaker E, Raven M. Eect of extended-release naltrexone on alcohol consumption: a systematic review and meta-analysis. Addiction 2022; 117: 271–81. 1698 --- [PDF原文](https://sci-net.xyz/storage/7932541/8356c5b15ee58462f88e2544a000a4a48a406b21db4e6d9d89a95b3e8ece41bb/Once-weekly-semaglutide-versus-placebo-in-patients-with-alcohol-use-disorder-and-comorbid.pdf) DOI: 10.1016/S0140-6736(26)00305-3