OASIS 1 produced the largest weight reduction ever published for a GLP-1 taken by mouth: an estimated 15.1% of body weight over 68 weeks.[1] It did it at a strength nobody can buy. The tablet in this trial was escalated to 50 mg once daily; the oral product that reached the market is the 25 mg tablet, whose own trial is covered in the OASIS 4 article. Anyone shopping for the fifteen-percent pill is shopping for a strength that exists only in this dataset.
What was run
A randomized, double-blind, placebo-controlled phase 3 superiority trial at 50 outpatient clinics across nine countries in Asia, Europe and North America enrolled adults with a body-mass index of at least 30, or at least 27 with a weight-related complication, and without type 2 diabetes. Of 709 people screened, 667 were randomly assigned 1:1 — 334 to oral semaglutide escalated to 50 mg once daily and 333 to a visually matching placebo — for 68 weeks plus lifestyle intervention.[1] Mean age was 50 (SD 13) and 485 of the 667 participants, 72.7%, were women.[2]
There was no active comparator. Nothing in this trial compares the tablet with the injection; every such comparison made from it is a comparison across trials, with everything that implies.
The coprimary pair, and the estimand that quadruples an odds ratio
Two endpoints shared primary billing: percentage change in body weight, and whether a participant reached a reduction of at least 5%, both at week 68 and both assessed regardless of whether the person kept taking the tablet or started something else. On that basis, estimated mean weight change was −15.1% (SE 0.5) against −2.4% (SE 0.5) — an estimated treatment difference of −12.7 percentage points (95% CI, −14.2 to −11.3; p < 0.0001). Reaching 5% or more: 269 of 317 (85%) against 76 of 295 (26%), odds ratio 12.6 (95% CI, 8.5 to 18.7).[1]
The registry posts that analysis and a second one beside it, and the second is where the numbers change shape. Restricted to responses recorded before a participant first stopped the tablet or started another weight-lowering treatment, the weight difference widens to −15.63 percentage points (95% CI, −17.07 to −14.18), and the odds of reaching 5% rise from 12.62 (8.50 to 18.74) to 55.21 (95% CI, 32.98 to 92.41).[2]
A fourfold move in an odds ratio on one endpoint in one trial is a useful thing to have seen. Neither figure is wrong. The first answers what happens to people prescribed the drug, including everyone who abandoned it; the second answers what happens to the drug in the people still taking it. Marketing has an obvious preference between the two, and a reader who does not know which one is on the page cannot tell a forty-point difference in odds from a rounding choice. The rest of the threshold ladder, on the first analysis, ran 220 (69%) against 35 (12%) at 10%, 170 (54%) against 17 (6%) at 15%, and 107 (34%) against 8 (3%) at 20%.[1] Measured in kilograms, the tablet arm lost 16.1 kg (SD 10.9) against 2.4 kg (SD 7.9).[2]
One detail in those threshold figures is easy to skim past. The denominators are 317 and 295, not the 334 and 333 who were randomized: the percentages are shares of participants with a week-68 weight on record, so 17 people on the tablet and 38 on placebo are absent from them.[1] More than twice as many are missing from the placebo column as from the tablet column, and the registry’s own withdrawal reasons point the same way: 14 placebo participants were lost to follow-up against 4 on the tablet.[2] Every analytical choice about who counts moves these numbers, and in this trial they all move in one direction.
Doubling the milligrams buys about a point
The 25 mg tablet that reached the market was tested against placebo in 205 and 102 participants over 64 weeks and produced an estimated difference of −11.4 percentage points (95% CI, −13.9 to −9.0).[4] Against the −12.7 here, doubling the daily milligrams bought roughly one and a third percentage points of separation from placebo — across two different trials, two different populations and two endpoints assessed four weeks apart, so the comparison is indicative rather than measured.
The other comparison worth making runs the opposite direction and is more surprising. Weekly injected semaglutide at 2.4 mg over 68 weeks produced an estimated treatment difference of −12.4 percentage points (95% CI, −13.4 to −11.5).[5] This tablet delivered −12.7. Seven daily 50 mg tablets come to 350 mg of drug a week against 2.4 mg injected, and they arrive at approximately the same place. The milligram figures on the two routes are not on one scale and cannot be compared as though they were, which is the subject of the oral-versus-injectable article, and the semaglutide dose article sets out what each labeled strength is for.
Four in five reported a gut symptom. Six in a hundred stopped.
Adverse events of any kind were reported by 307 of 334 (92%) on the tablet and 285 of 333 (86%) on placebo, with gastrointestinal events — mostly mild to moderate — in 268 (80%) against 154 (46%).[1] By term, the registry records nausea in 173 of 334 (51.8%) against 51 of 333 (15.3%), vomiting in 80 (24.0%) against 12 (3.6%), constipation in 92 (27.5%) against 50 (15.0%) and diarrhea in 89 (26.6%) against 56 (16.8%).[2]
Eighty percent is the number that gets quoted, and on its own it describes a drug people cannot stay on. They stayed on it. Neither the abstract nor the registry publishes a discontinuation rate, but a published review of the OASIS program reproduces this trial’s safety table, and in it adverse events led to permanent discontinuation of the trial product in 19 participants (6%) against 12 (4%) on placebo, with gastrointestinal events specifically accounting for 12 (4%) against 5 (2%).[3] A gap of two percentage points over placebo is what the eighty-percent figure actually costs in abandonment. For scale, the injectable trial at 2.4 mg discontinued for gastrointestinal events in 59 participants (4.5%) against 5 (0.8%).[5]
The arm people left was the placebo arm
Trial withdrawal is a different count again, and it runs against expectation. 14 of 334 on the tablet did not complete, against 26 of 333 on placebo; 4 against 14 were lost to follow-up.[2] Twice as many people walked away from the inert assignment as from the one causing nausea in half its recipients.
The serious-event columns need both of their numbers for the same reason. Participants with at least one serious adverse event were 32 (9.6%) on the tablet against 29 (8.7%) on placebo — slightly more. Serious adverse events, counted individually, were 44 against 48 — slightly fewer. No deaths occurred in either arm. The one serious term clearly concentrated on the drug was gallstones, at 4 against 0.[2] Four events is not a rate, and a trial of 667 people over 68 weeks cannot establish one — but a zero in the comparison column is the reason the imbalance is worth printing rather than smoothing away.
The columns that are not about weight
Waist circumference fell 13.4 cm (SD 10.0) against 2.8 (SD 7.3), and body-mass index 5.9 (SD 4.0) against 0.9 (SD 2.8); 129 participants on the tablet who entered with a body-mass index of 30 or above finished below it, against 19 on placebo. Systolic blood pressure fell 7 mmHg (SD 14) against 1 (SD 14), and high-sensitivity C-reactive protein ended at 0.42 times baseline against 0.85.[2]
Glycemic movement was small in absolute terms and large in category terms. Glycated hemoglobin fell 0.2 points (SD 0.3) on the tablet and rose 0.1 (SD 0.3) on placebo — this was a trial that excluded diabetes, so there was little to lower. But of participants classified at baseline, those ending in the normoglycemic band went from 200 to 274 on the tablet and from 200 to 139 on placebo, while the prediabetes band fell from 132 to 36 against 130 to 143.[2] Placebo participants moved the wrong way over sixteen months.
One column moved in a direction nobody markets. Mean pulse rose 4 beats per minute (SD 9) on the tablet against no change on placebo.[2] That is a class effect rather than a surprise, and what is known about it is set out in the heart rate article.
What OASIS 1 did not show
It did not test the tablet against an injection, or against any other drug. It did not enroll anyone with type 2 diabetes, so the glycemic columns above describe people who mostly had normal or near-normal blood sugar. It measured no cardiovascular events and adjudicated none. It ran 68 weeks, so it says nothing about the second year, and nothing at all about what happens when the tablet stops.
And it did not describe a product. The 50 mg strength was not brought to market; the trial that supported the marketed oral tablet used half that dose and reported a smaller difference. Quoting 15.1% as what an oral semaglutide product delivers attaches this trial’s result to a different tablet. Where the published semaglutide weight figures actually sit, by dose and by route, is collected in the semaglutide weight-loss article.
What was in the tablet
Every figure above belongs to a manufacturer-supplied, coformulated oral semaglutide tablet, taken under the strict fasting and water conditions that formulation requires, escalated by protocol to a 50 mg strength that was never marketed, with lifestyle support attached, in a population selected for obesity without diabetes. Products listed on the oral semaglutide board are generally compounded oral or sublingual preparations. Compounded drugs are not FDA-approved, and the FDA does not review them for safety, efficacy or quality before they are dispensed.
A compounded preparation that prints 50 mg on the label is not thereby this trial’s tablet, because the number on an oral peptide is meaningless without the absorption system that carries it across the gut wall — which is the coformulation, not the milligrams. That is the gap to hold onto: this trial measured a delivery system as much as a molecule, and it is the delivery system that a compounded copy has no obligation to reproduce.