Semaglutide: What the Published Studies Report

Semaglutide is a glucagon-like peptide-1 (GLP-1) receptor agonist, and reviews of this drug class describe shared mechanisms observed in humans including glucose-dependent augmentation of insulin secretion, suppression of glucagon secretion, deceleration of gastric emptying, and reduced calorie intake [9]. That review characterises long-acting agents in the class, semaglutide among them, as having more pronounced effects on overnight and fasting plasma glucose and on HbA1c than short-acting agents, with the gastric-emptying effect diminishing over time [9]. Semaglutide is also described as the only agent of the class available in both subcutaneous and oral formulations, with phase 3 registration programmes reported for each [2]. In head-to-head human trials in type 2 diabetes, the open-label SUSTAIN 7 trial randomised 1201 patients on metformin monotherapy to once-weekly semaglutide or dulaglutide and reported HbA1c reductions of 1.5 and 1.8 percentage points with semaglutide at the two dose levels tested versus 1.1 and 1.4 percentage points with dulaglutide, alongside bodyweight reductions of 4.6 kg and 6.5 kg versus 2.3 kg and 3.0 kg [14]. Gastrointestinal disorders were the most frequently reported adverse event in that trial and the most common reason for treatment discontinuation in both groups [14]. In the placebo-controlled SUSTAIN 5 trial, semaglutide added to basal insulin in 397 patients with uncontrolled type 2 diabetes was associated with HbA1c reductions of 1.4 and 1.8 percentage points versus 0.1 percentage points with placebo, and with mean bodyweight changes of 3.7 kg and 6.4 kg versus 1.4 kg, with premature discontinuation due to adverse events higher than placebo and mainly gastrointestinal [16]. Several human studies have compared semaglutide with the dual agonist tirzepatide. In the 40-week phase 3 SURPASS-2 trial in 1879 patients with type 2 diabetes, the estimated mean HbA1c change was -1.86 percentage points with semaglutide, and tirzepatide was reported as non-inferior and superior on that endpoint, with greater bodyweight reductions and broadly similar rates of nausea, diarrhoea and vomiting between groups [1]. In the 72-week open-label SURMOUNT-5 trial in 751 adults with obesity but without type 2 diabetes, the least-squares mean weight change was -13.7% with semaglutide and -20.2% with tirzepatide, with waist-circumference changes of -13.0 cm and -18.4 cm respectively, and gastrointestinal events most common in both groups [5]. A systematic review and meta-analysis of four direct comparative human studies including 28,827 patients reported mean weight change of -7.3% with semaglutide and -11.4% with tirzepatide, with a pooled mean difference of -4.84 kg favouring tirzepatide and mostly mild to moderate gastrointestinal adverse events in both groups [11]. A propensity-matched electronic-health-record cohort of 18,386 adults with overweight or obesity reported higher hazards of reaching 5%, 10% and 15% weight loss with tirzepatide than semaglutide and similar rates of gastrointestinal adverse events, with treatment discontinuation ending follow-up for 52.5% of the semaglutide group [3]. Combination work has been reported in two phase 3a human trials. In REDEFINE 1, which randomised 3417 adults with overweight or obesity and without diabetes, the estimated mean bodyweight change at week 68 was -20.4% with coadministered cagrilintide and semaglutide versus -3.0% with placebo, with gastrointestinal adverse events in 79.6% and 39.9% of the two groups and described as mainly transient and mild to moderate [4]. In REDEFINE 2, in 1206 adults with overweight or obesity and type 2 diabetes, the estimated mean bodyweight change at week 68 was -13.7% with cagrilintide-semaglutide versus -3.4% with placebo, HbA1c of 6.5% or less was recorded in 73.5% versus 15.9% of patients, and gastrointestinal adverse events were reported by 72.5% versus 34.4% [10]. A systematic review of 26 placebo-controlled randomised trials in adults with overweight or obesity and without diabetes reported weight loss of up to 13.9% with once-weekly semaglutide after 68 weeks, with adverse events frequent but predominantly gastrointestinal and serious adverse events rare [17]. Body-composition and safety reporting has also been synthesised. A systematic review of six human studies in 1541 overweight or obese adults reported that weight reduction with semaglutide was driven primarily by fat mass, while lean-mass changes ranged from close to zero to about 40% of total weight reduction across trials, with the largest absolute lean-mass reductions noted in the larger trials [7]. A narrative safety review of the human registration programme concluded that semaglutide was associated mostly with mild-to-moderate and transient gastrointestinal disturbance and with an increased risk of cholelithiasis, that definitive conclusions on pancreatic and thyroid cancer could not be drawn given low event counts, and that patients with existing diabetic retinopathy warranted careful monitoring [2]. A retrospective matched cohort study at one academic neuro-ophthalmology registry reported higher hazards of non-arteritic anterior ischaemic optic neuropathy among patients prescribed semaglutide in both a type 2 diabetes population (hazard ratio 4.28) and an overweight or obese population (hazard ratio 7.64), while stating that the observational design does not establish causality [8]. A narrative review of six human studies encompassing 480,825 adults reported vitamin D deficiency in 7.5% of GLP-1 receptor agonist users at 6 months and 13.6% at 12 months and 26%-30% lower ferritin than SGLT2-inhibitor comparators, again noting that causality cannot be established from observational datasets [13]. Real-world human data are reported to diverge from trial data in several respects. A narrative review of real-world GLP-1 receptor agonist use found weight reduction generally lower than in randomised trials but approaching trial outcomes among highly adherent patients, discontinuation of 20%-50% within the first year, use of much lower doses than those tested in trials, and no clear increase in pancreatitis, pancreatic cancer, thyroid disorders, or depression and self-harm in observational populations [6]. Five claims-based cohort studies in patients with cardiometabolic heart failure with preserved ejection fraction reported a lower hazard of the composite of heart-failure hospitalisation or all-cause mortality for semaglutide initiators than for sitagliptin used as a placebo proxy (hazard ratio 0.58), with no meaningful difference between tirzepatide and semaglutide (hazard ratio 0.86) [12]. A review of therapeutics for metabolic dysfunction-associated steatohepatitis reported that human studies of semaglutide showed effects on steatohepatitis resolution with limited histological improvement in fibrosis, and flagged gastrointestinal tolerability as a concern [19]. Two further research directions appear in this set. A study combining machine-learning analysis of approximately 68,250 social-media posts with a remote study of 153 alcohol-drinking adults with obesity reported lower self-reported alcohol intake, fewer drinks per drinking episode, lower binge-drinking odds and lower AUDIT scores among people self-reporting semaglutide or tirzepatide use than before starting the medication and than an untreated control group [18]. In rodents, semaglutide given acutely and repeatedly reduced alcohol intake and relapse-like drinking in male and female rats, fluorescently labelled semaglutide was detected in the nucleus accumbens, and in male mice semaglutide attenuated alcohol-induced hyperlocomotion and accumbal dopamine release while raising dopamine metabolites and the expression of dopamine-metabolising enzymes [20]. These are animal findings and have not been shown to correspond to human outcomes [20]. Separately, a review of GLP-1 receptor agonist exposure in pregnancy reported adverse offspring outcomes including reduced fetal growth, skeletal and visceral anomalies and embryonic death in some small-animal studies, alongside a human population-based cohort of 938 pregnancies affected by type 2 diabetes that found no significantly increased risk of major congenital malformations, and concluded that available human evidence remains insufficient to predict effects of periconceptional exposure [15].