Semaglutide: How the First Oral GLP-1 Changed Metabolic Medicine

The story of semaglutide is the story of a peptide that crossed the last barrier in drug delivery, and the trials that followed it, from STEP 1 to OASIS 4 to SELECT, rewrote what the field expects from a weight-loss molecule.

The verified record runs through a handful of landmark studies and one regulatory discrepancy, and each one changed the questions the next one asked.

I · The molecule that crossed the needle barrier

Peptide drugs spent their first half century tethered to the injection needle, and semaglutide broke that tether with a pill that carries a GLP-1 receptor agonist through the digestive tract intact.

Semaglutide belongs to the GLP-1 receptor agonist class, the drug family behind the STEP, OASIS, and SELECT programs, and it holds a specific place in that family because it became the first member available as a pill, a distinction that the FDA press record itself announced. The regulatory timeline for that oral version, sold as Rybelsus, carries a conflict that the verification record flags rather than resolves, since FDA files show the original new drug application action dated 01/16/2020 while press coverage of the first oral GLP-1 approval circulated in September 2019¹³. The injectable version of the same molecule, marketed as Wegovy, has a cleaner record, with NDA 215256 approved on 06/04/2021 in the FDA’s own database, and that verified date anchors the weight-loss story that follows¹³.

The phase 3 program that established the oral formulation in type 2 diabetes was PIONEER, which Vanita Aroda helped lead, and its trials put the absorption story on firm footing before the weight-loss studies began. The move from diabetes to obesity then shaped the semaglutide era, since the same molecule went on to anchor the STEP program with an injectable dose and the OASIS program with oral doses, and each program produced its own headline numbers.

II · STEP 1 and the numbers that reset expectations

Weight-loss trials before semaglutide rarely produced double-digit mean reductions, which is why the STEP 1 results looked like a typo when they appeared in the New England Journal of Medicine.

John Wilding and Rachel Batterham reported STEP 1 in 2021, and the phase 3 trial randomized adults with overweight or obesity, without diabetes, to once-weekly subcutaneous semaglutide 2.4 mg or placebo for 68 weeks, with mean weight change of -14.9% in the treatment group versus -2.4% in the placebo group¹. The categorical breakdown carried the more striking message, since 86.4% of treated participants lost at least 5% of body weight, 69.1% lost at least 10%, and 50.5% lost at least 15%, with a mean absolute loss of 15.3 kg across the trial¹.

“In STEP 1, mean weight fell 14.9% over 68 weeks versus 2.4% with placebo, and roughly half of participants lost at least 15% of their body weight.”

John Wilding and Rachel Batterham, New England Journal of Medicine, 2021

Those categorical figures mattered because they told clinicians that substantial weight loss was becoming the typical response rather than a rare one, and they set the standard that every later trial in the class had to answer to. The STEP program’s place in this article is therefore the baseline, since its numbers became the reference points against which the oral trials, the cardiovascular trials, and the body-composition meta-analyses all positioned themselves.

III · The OASIS program carries the oral route forward

The injectable had answered the efficacy question, so the OASIS program asked whether a pill could carry the same molecule to the same result, and the trials answered across two doses and two continents.

Filip Knop led OASIS 1, published in Lancet in 2023, which tested oral semaglutide 50 mg once daily and reported mean weight change of -15.1% versus -2.4% at week 68². The East Asian extension, OASIS 2, was reported by T. Kadowaki in JAMA Internal Medicine in 2025 and found -14.3% versus -1.3% at week 68 in the same 50 mg design, which extended the oral finding to a population with different body composition norms³.

“Oral semaglutide 50 mg produced 15.1% mean weight loss over 68 weeks in the OASIS 1 trial led by Filip Knop, against 2.4% in the placebo group.”

Filip Knop, Lancet, 2023

Sean Wharton reported OASIS 4 in the New England Journal of Medicine in 2025, and that trial tested the 25 mg dose rather than the 50 mg dose, producing -13.6% versus -2.2% at week 64 under the ClinicalTrials.gov identifier NCT05564117⁴. The three trials form a descriptive comparison rather than a head-to-head test, since the doses differ across them, and the trial means sit in a tight band between 13.6% and 15.1%, which told the field that oral dosing had matched the injectable in trial averages²³⁴.

IV · SELECT and the cardiovascular question

Weight trials answer the scale question, whereas cardiovascular outcome trials answer the question beneath it, and SELECT was built to settle that second one in a population without diabetes.

A. Michael Lincoff reported SELECT in the New England Journal of Medicine in 2023, and the trial enrolled adults with overweight or obesity and established cardiovascular disease while excluding anyone with diabetes, so the readout could be interpreted in a population without diabetes⁵. The primary results showed fewer major adverse cardiovascular events in the semaglutide group than in the placebo group, which extended the class’s cardiovascular evidence to a population defined by weight status⁵.

The SELECT population matters because the earlier cardiovascular outcome trials in the GLP-1 class had all enrolled people with type 2 diabetes, and the class-level mortality signal quantified in 2019 came from those diabetic cohorts⁵⁶. That lineage connects the weight-loss programs to the cardiovascular evidence, and it explains why the oral molecule’s story reaches beyond the scale, since SELECT pointed at the same cardiovascular direction in people without diabetes⁵⁶.

V · The mortality signal in the meta-analysis

Individual trials are sized for their own primary endpoints, which leaves rarer outcomes such as death to the statisticians who pool them, and the Kristensen meta-analysis assembled the largest dataset of that kind in the class.

Søren Kristensen and colleagues published the meta-analysis in Lancet Diabetes & Endocrinology in 2019, and it pooled 7 cardiovascular outcome trials covering 56,004 participants with type 2 diabetes, reporting an all-cause mortality hazard ratio of 0.88 with a 95% confidence interval of 0.83 to 0.95⁶. That hazard ratio translates to a 12% relative reduction in all-cause mortality, and the verification record is explicit that the figure is relative, so the absolute difference depends on the baseline mortality rate in each trial population⁶.

“A 2019 meta-analysis led by Søren Kristensen reported a 12% relative reduction in all-cause mortality across GLP-1 receptor agonists, pooling 7 cardiovascular outcome trials with 56,004 participants.”

Søren Kristensen, Lancet Diabetes & Endocrinology, 2019

The type 2 diabetes scope of the pooled trials matters because the 12% figure describes that population specifically, and the obesity trials that followed were part of a later evidence wave⁶. The open question, then, is whether the same relative signal holds in the weight-defined populations that SELECT enrolled, since the meta-analysis predates those trials and the SELECT readout points in the same direction without being pooled into it⁶⁵.

VI · Lean mass and the body composition question

A scale reports total weight, so the composition of that weight became the next question, and the 2026 meta-analyses answered it by correcting a claim that had circulated for years.

The claim that roughly 40% of weight lost on semaglutide comes from lean mass is a circulating misattribution, since the verification record found no paper behind it, and the likely source is a 2026 JAMA Network Open body-composition analysis by Wang and colleagues that was read loosely⁷¹⁵. The corrected estimate comes from the Eisa 2026 meta-analysis of 15,782 participants, which places lean mass at 25% to 39% of weight lost, with a pooled figure of 35.2% for semaglutide (95% CI 31.5-38.9), and a parallel analysis by Batsis and colleagues reported a median of 28.3% (IQR 15.9-39.9) from a different statistical approach⁷⁸. A further analysis by Laverde and colleagues in 2026 found that lean mass as a share of total body weight rose by 1.81% during treatment, which means lean tissue holds or grows its share of total weight as the scale falls⁹.

“Meta-analytic estimates place lean mass at 25% to 39% of the weight lost on semaglutide, with a pooled figure near 35.2%, while lean mass as a share of total weight typically increases.”

Eisa and colleagues, 2026 meta-analysis, 2026

The composition picture therefore comes down to measurement, since different analyses bracket lean loss between 25% and 39% of weight lost while simultaneously showing that lean mass holds or grows its share of total weight⁷⁸⁹. Commentators including Dubin, Heymsfield, Ravussin, and Greenway wrote about the interpretation of weight-loss composition in Diabetes, Obesity and Metabolism in 2024, which signals an active methodological debate about how these fractions should be read and reported¹⁴. The practical takeaway is that lean loss on semaglutide is real, measurable, and smaller than the circulating claim, with the protein literature offering a reference point for the resistance-training context rather than a protocol for these trials⁷¹⁰.

VII · The population taking GLP-1 drugs

The trials describe what the molecule can do under protocol, whereas a polling firm asked what people actually do, and the answer put GLP-1 use squarely into the mainstream.

The KFF Health Tracking Poll of November 14, 2025 found that 12% of US adults, about 1 in 8, reported currently taking a GLP-1 drug for weight loss, diabetes, or another condition, and that 12% is the verified figure after earlier reports of 12.4% failed verification¹¹. The poll captures self-reported current use of the whole drug class, so it counts injectable and oral forms and every GLP-1 molecule, which makes it a population snapshot rather than a semaglutide-specific number¹¹.

The East Asian evidence rounds out the population picture, since T. Kadowaki reported STEP 6 in Lancet Diabetes & Endocrinology in 2022, testing semaglutide in East Asian adults with overweight or obesity and complementing the OASIS 2 readout from the same investigator group¹²³. The combination of the 12% population figure and the East Asian trial data shows a class spreading across geography and indication, which is the context every single-molecule trial result has to be read inside¹¹¹².

VIII · Open questions and the honest limits

Every dataset in this article carries a caveat, so the honest summary is a list of open questions rather than a verdict, and the verification record names them plainly.

The regulatory record still carries the Rybelsus date conflict, since FDA files show the ORIG-1 action on 01/16/2020 while press coverage dated the first oral GLP-1 approval to September 2019, and the verification record flags the discrepancy rather than resolving it¹³. The oral 25 mg formulation’s approval for obesity appeared in January 2026 news coverage but could not be confirmed against a primary source, and a circulating 17% figure attributed to the OASIS program appears in none of the three verified trials, so both remain unverified in the current record¹³²³⁴.

The search-trend claim that GLP-1 overtook Ozempic in search volume by early 2025 also failed verification, and every trial number quoted in this article is a trial average that describes the group rather than any individual outcome²¹. The verified record supports a coherent story: an oral GLP-1 receptor agonist with double-digit mean weight loss across STEP 1 and the OASIS trials, a cardiovascular readout in SELECT, a 12% relative mortality association in the Kristensen meta-analysis, and a body-composition picture corrected by the 2026 meta-analyses¹²⁵⁶⁷. Those open questions are the honest boundary of the evidence, and they mark where the next round of verification work begins.