Semaglutide

Identity, and the one registry figure that disagrees

Semaglutide is a 31-residue peptide built on the human GLP-1(7-37) backbone. PubChem indexes it as CID 56843331 with the molecular formula C187H291N45O59 and an average molecular weight of 4114. The FDA prescribing information for the subcutaneous injection gives the same formula and a molecular weight of 4113.58 g/mol. CAS registry number 910463-68-2 appears as the primary CAS code in the FDA Global Substance Registration System record under UNII 53AXN4NNHX, and a PubChem name query on that registry number resolves back to the same CID. Three independent sources agreeing on formula, weight and registry number is not the usual position for compounds catalogued here.

The GSRS record prints the sequence as HAEGTFTSDVSSYLEGQAAKEFIAWLVRGRG and lists two site-specific structural modifications on it: 2-aminoisobutyric acid substituted at residue 2, and a modified lysine at residue 20. Under the conventional GLP-1 numbering those are positions 8 and 26. Lau and colleagues, describing the design in 2015, stated two amino acid substitutions relative to human GLP-1 - Aib8 and Arg34 - with derivatisation at lysine 26. The acyl group is an eighteen-carbon diacid attached through a gamma-glutamyl spacer and two AEEA units. The label's mechanism section states 94 percent sequence homology to human GLP-1, which is the arithmetic of two substitutions across thirty-one residues rather than a separate measurement.

One registry figure does not match. GSRS displays its own molecular formula for the substance, C187H294O60N45 at approximately 4100 Da, explicitly typed ESTIMATED and machine-calculated. It differs from the label and PubChem value by three hydrogens and one oxygen. The label figure is carried above because it is the one attached to a described structure and a regulatory filing, but the discrepancy is recorded rather than averaged away. Salt forms are a separate identity question. PubChem returns no compound at all for the name semaglutide sodium; semaglutide acetate resolves to a different record, CID 171360195, formula C189H295N45O61, CAS 1997361-85-9. Neither salt carries a GSRS substance record.

What the discovery paper measured, and in which species

Lau and colleagues published the design work in the Journal of Medicinal Chemistry in 2015. The stated objective was a once-weekly analogue obtained by raising albumin affinity and securing stability against metabolic degradation. Receptor affinity at the GLP-1 receptor was reported as 0.38 plus or minus 0.06 nM, described as three-fold lower than liraglutide, with albumin affinity increased. Pharmacokinetics were characterised in mini-pigs: a plasma half-life of 46.1 hours after intravenous administration, and a mean residence time of 63.6 hours after subcutaneous dosing. Those are pig figures from an in-house programme, and they are the only species-attributed half-life numbers in the discovery report.

Human pharmacokinetics come from the approved labelling rather than from that paper. Several distinct semaglutide injection applications exist with different sections; the figures carried here are from the type 2 diabetes presentation, NDA 209637. Absolute bioavailability after subcutaneous administration is stated as 89 percent, plasma albumin binding as greater than 99 percent, apparent volume of distribution as approximately 12.5 L, and elimination half-life as approximately one week, with steady state reached after four to five weeks of once-weekly administration. Clearance runs through proteolytic cleavage of the peptide backbone and beta-oxidation of the fatty acid side chain; roughly 3 percent is excreted unchanged in urine. The labelling attributes the long half-life principally to albumin binding, with DPP-4 stabilisation as a separate contribution.

Pharmacodynamic measurements in that labelling were taken at steady state after twelve weeks at 1 mg in patients with type 2 diabetes, and the document gives no sample size for them at all. Reductions relative to placebo were 29 mg/dL for fasting glucose, 74 mg/dL for two-hour postprandial glucose, and 30 mg/dL for mean 24-hour glucose. Gastric emptying is described twice and differently in the same document: the mechanism section calls it a minor delay in the early postprandial phase, while the warnings section states that the drug delays gastric emptying, may affect absorption of concomitant oral medicines, and has attracted rare postmarketing reports of pulmonary aspiration in the GLP-1 receptor agonist class.

The subcutaneous outcome trials

SUSTAIN-6 was the cardiovascular safety trial. Marso and colleagues randomised 3,297 adults with type 2 diabetes to once-weekly semaglutide at 0.5 mg or 1.0 mg, or placebo, for 104 weeks; 83.0 percent had established cardiovascular disease, chronic kidney disease, or both at baseline. The primary composite of cardiovascular death, non-fatal myocardial infarction and non-fatal stroke occurred in 108 of 1,648 (6.6 percent) against 146 of 1,649 (8.9 percent), hazard ratio 0.74. The same trial recorded a significantly higher rate of retinopathy complications in the treated group, hazard ratio 1.76 with a 95 percent confidence interval of 1.11 to 2.78. Both findings come from one trial and are logged together.

Weight was the endpoint in STEP 1. Wilding and colleagues randomised 1,961 adults without diabetes, 2:1, to 68 weeks of once-weekly subcutaneous semaglutide at 2.4 mg or placebo, both with a lifestyle intervention. Mean change in body weight was minus 14.9 percent against minus 2.4 percent. Reductions of 15 percent or more were reached by 612 participants (50.5 percent) against 28 (4.9 percent). Nausea and diarrhoea were the most frequently reported adverse events, and discontinuation for gastrointestinal events ran at 4.5 percent against 0.8 percent.

SELECT is the largest published trial of the molecule. Lincoff and colleagues enrolled 17,604 patients aged 45 or older with pre-existing cardiovascular disease and a body-mass index of 27 or greater, without diabetes, and followed them for a mean of 39.8 months. A primary cardiovascular endpoint event occurred in 569 of 8,803 (6.5 percent) against 701 of 8,801 (8.0 percent), hazard ratio 0.80. Permanent discontinuation for adverse events occurred in 16.6 percent against 8.2 percent. In FLOW, Perkovic and colleagues randomised 3,533 patients with type 2 diabetes and chronic kidney disease and stopped early at a prespecified interim analysis, with 331 against 410 first primary-outcome events, hazard ratio 0.76, at a median follow-up of 3.4 years.

Two further trials measured organ-level endpoints. STEP-HFpEF randomised 529 patients with heart failure with preserved ejection fraction and a body-mass index of 30 or higher to 52 weeks of 2.4 mg weekly; the Kansas City Cardiomyopathy Questionnaire clinical summary score changed by 16.6 points against 8.7, an estimated difference of 7.8 points, alongside a body weight change of minus 13.3 percent against minus 2.6 percent (Kosiborod 2023, PMID 37622681). ESSENCE, reported by Sanyal and colleagues, is a 1,197-patient trial in biopsy-defined steatohepatitis with fibrosis stage 2 or 3; the planned week-72 interim analysis of the first 800 patients found resolution of steatohepatitis without worsening fibrosis in 62.9 percent against 34.3 percent. That trial runs to 240 weeks and has not finished.

The oral formulation is a separate evidence base

Oral semaglutide is the same molecule co-formulated with an absorption enhancer, and its outcome data are its own. PIONEER 6 was a non-inferiority trial: Husain and colleagues randomised 3,183 patients at high cardiovascular risk, median time in trial 15.9 months, and recorded major adverse cardiovascular events in 61 of 1,591 (3.8 percent) against 76 of 1,592 (4.8 percent), hazard ratio 0.79 with a confidence interval spanning 0.57 to 1.11. The trial was designed to rule out an 80 percent excess risk, not to demonstrate benefit, and its conclusion was stated in those terms.

SOUL was powered for superiority. McGuire and colleagues randomised 9,650 participants aged 50 or older with type 2 diabetes and atherosclerotic cardiovascular disease, chronic kidney disease or both, to once-daily oral semaglutide at a maximal dose of 14 mg or placebo, with mean follow-up of 47.5 months and median follow-up of 49.5 months. A primary-outcome event occurred in 579 of 4,825 (12.0 percent) against 668 of 4,825 (13.8 percent), hazard ratio 0.86. Results for the confirmatory secondary outcomes, which included a five-point composite of major kidney disease events, did not differ significantly between groups. Anyone treating oral and subcutaneous figures as interchangeable is combining different exposures, different trials and different endpoints.

Where the published record is negative or contested

The largest negative result is recent. Cummings and colleagues reported evoke and evoke+ in The Lancet in 2026: two phase 3 trials across 566 sites in 40 countries, 3,808 participants aged 55 to 85 with amyloid-confirmed early Alzheimer's disease, randomised to once-daily oral semaglutide up to 14 mg or placebo for up to 156 weeks. Change in the Clinical Dementia Rating-Sum of Boxes score at week 104 gave an estimated difference of minus 0.08 in evoke and 0.10 in evoke+, neither approaching significance. Both trials were discontinued for negative clinical outcome. The dementia-protection claim that circulated before this readout rested on animal work, observational cohorts and pharmacoepidemiology, and it did not survive a direct test.

Optic neuropathy is where the sources disagree in magnitude while agreeing in direction. Hathaway and colleagues reported a single-institution matched cohort study in 2024 covering two populations. Among patients with type 2 diabetes, 17 events occurred in 194 patients prescribed semaglutide against 6 among 516 comparators, hazard ratio 4.28. Among 979 overweight or obese patients, 20 events occurred among the 361 prescribed semaglutide against 3 among 618 comparators, hazard ratio 7.64 with a 95 percent confidence interval of 2.21 to 26.36. Heberer and colleagues then emulated a target trial in 102,361 US veterans, comparing 11,478 semaglutide initiators with 90,883 SGLT2-inhibitor initiators; 173 incident events occurred in total, at 123 against 67 per 100,000 person-years, hazard ratio 2.33, with an absolute weighted incidence of 0.29 percent against 0.13 percent. That paper carries a published erratum correcting Figure 2.

The one synthesis of that literature does not collapse to a single number either. A 2026 systematic review and meta-analysis of eight retrospective cohorts totalling 14,255,247 participants pooled the hazard ratio at 3.36, with a 95 percent confidence interval of 1.44 to 7.84, at one-year follow-up, and reported heterogeneity of I squared 97 percent on that estimate. At two-year follow-up the pooled figure was 2.37 with I squared of zero, at three years 2.37 after sensitivity analysis with I squared of 28 percent, and at five years 2.37 again. The one-year figure is the highest of the four and by far the least homogeneous, and quoting it alone would misdescribe what the analysis reported.

Retinopathy has its own unfinished thread. Beyond the SUSTAIN-6 hazard ratio of 1.76, the labelling records diabetic retinopathy complications in 3.0 percent of semaglutide-treated patients against 1.8 percent on placebo, and attributes the pattern in part to a temporary worsening that accompanies rapid improvement in glucose control. FOCUS, registered as NCT03811561 with an estimated 1,500 participants, is the trial designed to settle the question; it is listed as active, not recruiting, with an estimated completion date of November 2027 and no published result.

Alcohol consumption produced a small positive signal that is frequently overstated. Hendershot and colleagues ran a phase 2, double-blind, nine-week trial in 48 non-treatment-seeking adults with alcohol use disorder at a single US academic centre. The primary laboratory measure was taken in the 0.5 mg arm. Grams of alcohol consumed in a laboratory self-administration task fell relative to placebo, and weekly craving and drinks per drinking day fell, while average drinks per calendar day and number of drinking days did not change. Forty-eight participants at one site over nine weeks is a signal worth registering and not a demonstration of anything durable.

Rodent carcinogenicity, and material outside the approved supply chain

The approved labelling carries a boxed warning for thyroid C-cell tumours. In a two-year carcinogenicity study in CD-1 mice, subcutaneous doses of 0.3, 1 and 3 mg/kg/day in males and 0.1, 0.3 and 1 mg/kg/day in females produced a statistically significant increase in thyroid C-cell adenomas and a numerical increase in carcinomas. In Sprague Dawley rats at 0.0025 to 0.1 mg/kg/day, C-cell adenomas increased significantly in males and females at all dose levels, and carcinomas in males at 0.01 mg/kg/day and above. The labelling states that human relevance of the rat C-cell finding is unknown and could not be determined by clinical or nonclinical studies; it makes no equivalent statement about the mouse study. Semaglutide was not mutagenic or clastogenic across the standard genotoxicity battery, and in rats an increase in oestrus cycle length was observed at all dose levels.

Approval attaches to specific manufactured presentations, not to the name. The FDA states that semaglutide salt forms, naming semaglutide sodium and semaglutide acetate, are different active ingredients from the one used in the approved drugs, that the agency has no information on whether those salts share the chemical and pharmacologic properties of the approved ingredient, and that it is not aware of a lawful basis for their use in compounding. The agency has separately warned companies selling unapproved semaglutide falsely labelled for research purposes or not for human consumption, noting that such products have been sold directly to consumers for human use with dosing instructions.

Analytical work on non-originator material exists and is unflattering. Kopp and colleagues profiled follow-on and compounded semaglutide by mass spectrometry alongside an MHC-II-associated peptide proteomics assay, photostability testing and a fibrillation assay, and reported distinct impurity profiles - amino acid deletions and additions plus unidentified impurities - against the originator products, greater disparity in strength and high-molecular-weight protein content on light exposure, and peptide impurities that presented on monocyte-derived dendritic cells. Adverse-event counts sit alongside that. As of 31 May 2026 the FDA reported receiving 990 adverse event reports associated with compounded semaglutide, while noting that state-licensed pharmacies that are not outsourcing facilities are not required to report, so the count is a floor rather than a measurement.

Material bought without a prescription has also been measured directly. Ashraf and colleagues screened 1,080 search-engine links, identified 59 unique illegal online pharmacy websites, and completed test purchases from six of them; three injection vials were delivered and three prefilled-pen orders were never shipped. Liquid chromatography coupled with mass spectrometry put the measured semaglutide purity of those three vials between 7.7 and 14.37 percent, against the 99 percent claimed on the product labels, while semaglutide content exceeded the labelled amount by 28.56 to 38.69 percent. No viable microorganisms were recovered at the time of testing, but endotoxin was detected in every sample at 2.1645 to 8.9511 EU/mg. Visual inspection found noncompliance on 59 to 63 percent of the criteria assessed.

What is not known

The failed Alzheimer's programme marks the boundary of what the trial record supports: a compound with large cardiovascular, kidney and hepatic results is not thereby established in every organ, and evoke and evoke+ demonstrated that directly across 3,808 participants. Retinopathy is the oldest unresolved question on the file - SUSTAIN-6 recorded a hazard ratio of 1.76 in 2016 and FOCUS, the 1,500-participant trial designed to characterise it, is listed as active, not recruiting, with an estimated completion date of November 2027. Optic neuropathy has published estimates ranging from 2.33 to 7.64 and no randomised evidence at all; every study located is observational, the highest figure comes from the overweight and obese cohort of the 2024 single-institution study (20 against 3 events, hazard ratio 7.64, 95% CI 2.21-26.36), and the one meta-analysis returns a different pooled figure at each follow-up horizon, from 3.36 at one year with I squared of 97 percent down to 2.37 at two, three and five years. Body composition is characterised only in small studies with wide confidence intervals, so the split of the recorded weight change between fat and lean tissue is not settled. Human relevance of the rat thyroid C-cell finding is stated as undetermined in the labelling itself, which makes no equivalent statement about the mouse study. Long-term data exist but are bounded: the longest published follow-up is the 47.5-month mean in SOUL (PMID 40162642), with FLOW at a median of 3.4 years and SELECT at a 39.8-month mean, and ESSENCE runs to 240 weeks without having reported past its week-72 interim. No published trial addresses use in pregnancy or lactation. Semaglutide is an approved medicine in the United States and elsewhere, which is unusual for this catalogue, but approval attaches to specific manufactured presentations; salt forms, compounded preparations and material labelled for research purposes are not that article, and the one published analysis of product bought without a prescription measured purity between 7.7 and 14.37 percent.

Questions