Three out of four people who swallowed the tablet in this trial reported a gastrointestinal adverse event, and almost none of them quit over it. Permanent discontinuation for an adverse reaction ran 6.9% on oral semaglutide 25 mg against 5.9% on placebo — a gap of one percentage point, recorded in the drug’s own FDA label.[3] That pairing is the most useful thing OASIS 4 produced, and it is the part no marketing page quotes in either direction. Whether a pill or a syringe is the better purchase is argued in the oral-versus-injectable article; what follows is the trial that put the 25 mg tablet on the market.
A small trial for a large indication
OASIS 4 ran for 71 weeks at 22 sites in four countries, double-blind and placebo-controlled. Adults without diabetes were eligible at a body-mass index of 30 or higher, or 27 or higher with at least one obesity-related complication, and were assigned 2:1 to a daily 25 mg tablet or placebo alongside lifestyle intervention. Randomization placed 205 participants on the drug and 102 on placebo. The coprimary endpoints were measured at week 64: percent change in body weight, and the proportion losing 5% or more.[1]
Three hundred and seven people is a small evidence base for a chronic medication, and the trial knows it. Every confirmatory secondary endpoint — losses of 10%, 15% and 20% or more, plus the physical-function score — was a comparison against placebo rather than against any other treatment.[1] Nothing in OASIS 4 tested this tablet against the injection, against the 50 mg tablet studied in the earlier OASIS trials, or against any competing molecule.
The headline, and the second number attached to it
Estimated mean body weight changed −13.6% on oral semaglutide against −2.2% on placebo, an estimated difference of −11.4 percentage points (95% CI, −13.9 to −9.0; P < 0.001).[1] The label’s intention-to-treat table gives the same drug figure against a placebo column of −2.4%, a difference of −11.2 points (95% CI, −13.6 to −8.8), and records that weight was missing at week 64 for 6.3% of the tablet group and 11.8% of the placebo group.[3]
What makes this endpoint worth reading twice is that the registry posts two analyses of it, each with its own name. Under the treatment policy estimand — everyone counted as randomized, whatever they went on to do — the difference is −11.43 points (95% CI, −13.88 to −8.98). Under the hypothetical estimand, which uses only responses recorded before a participant stopped the drug or started something else, it is −13.87 points (95% CI, −16.53 to −11.21).[2] The two answer different questions: the first is what the prescription achieves, the second is what the drug does in someone still taking it.
The gap widens sharply on the threshold endpoint. Reaching a 5% loss carried an odds ratio of 7.34 (95% CI, 4.22 to 12.76) under the treatment-policy estimand and 25.23 (95% CI, 13.24 to 48.07) under the hypothetical one.[2] A seller quoting the larger figure is not lying, and is describing an idealized adherence nobody buying by subscription should assume.
How many reached each threshold
On the full intention-to-treat set, the label records losses of 5% or more in 76.3% of tablet-treated patients against 31.3% on placebo; 10% or more in 59.8% against 14.4%; 15% or more in 47.0% against 5.5%; and 20% or more in 27.9% against 3.1%.[3] Just over a quarter of the drug arm crossed the twenty-percent line, which is the band a buyer usually has in mind when comparing a pill against the injectable semaglutide figures. The placebo arm losing 31.3% of its members past 5% is the other half of that comparison and is almost never shown.
Who stopped, and the arm they are measured against
Gastrointestinal adverse events affected 74.0% of the oral semaglutide group and 42.2% of the placebo group.[1] Across the whole trial the registry counts 1,239 treatment-emergent adverse events on the tablet against 432 on placebo.[2] On volume alone the tablet looks punishing.
The stopping column does not follow. Permanent discontinuation for an adverse reaction was 6.9% against 5.9%, with gastrointestinal reactions the most common cause at 3.4% against 2%.[3] For comparison, the same label puts permanent discontinuation on weekly injected semaglutide 2.4 mg at 6.8% against 3.2% on its own placebo across three trials.[3] The absolute quit rate is effectively identical between the two formulations. The excess over placebo is not: one point for the tablet against roughly three and a half for the injection, in trials that were never run against each other.
Serious events ran the unexpected way too. The registry records serious treatment-emergent events in 8 of 204 tablet-treated participants and 9 of 102 on placebo, with no deaths in either arm.[2] Severe gastrointestinal reactions were reported in 2% of tablet-treated patients and 0% of placebo, against 4.1% and 0.9% in the injection trials.[3] Alopecia appeared in 13 of 204 on the tablet against 2 of 102 on placebo, a signal discussed for the class in the hair loss article.
The cardiometabolic columns are modest
The label’s table for this trial shows systolic blood pressure falling 7 mm Hg on the tablet and 5.4 on placebo — a difference of 1.6 mm Hg. Glycated hemoglobin fell 0.3 points against 0.1, a difference of 0.2. Heart rate rose 2.7 beats per minute against 2.4, a difference of 0.4. Low-density lipoprotein cholesterol was 3.1% lower than placebo and triglycerides 9.9% lower.[3]
Those are smaller than the corresponding columns in the injection trials, where the same label records a systolic difference of 5.1 mm Hg and a heart-rate difference of 4.3 beats per minute against placebo.[3] Separate trials, separate populations, no head-to-head — but a reader expecting the pill to carry the class’s pulse effect should note that in this trial it barely separated from placebo, in contrast to the pattern described in the heart rate article.
The glycemic movement is worth a line of its own. Of participants who entered with prediabetes, 64 on the tablet returned to normoglycemia by week 64 against 13 on placebo, while 25 and 26 respectively stayed prediabetic. One tablet-treated participant progressed to diabetes and none on placebo did.[2] A single case decides nothing; it is recorded here because a page that prints only the 64 is quoting a real number while describing a cleaner picture than the table contains.
The figures that were modeled rather than measured
OASIS 4 excluded type 2 diabetes, and no trial of 25 mg has been run in people who have both obesity and diabetes. The gap was filled by simulation. A population-pharmacokinetic and exposure-response analysis reports that the 25 mg dose was approved for obesity on the OASIS program together with historical semaglutide data, using a modeled phase 2 trial to establish efficacy across oral dose levels and a modeled phase 3 trial to estimate efficacy in a diabetes population. In those simulated studies, weight reduction against placebo was 9.4% and 7.8% with diabetes, and 16.3% and 14.7% without it, for the trial-product and treatment-policy estimands respectively.[4]
The same analysis reports that 82.2% of OASIS 4 participants on 25 mg reached drug exposures within the range seen on weekly injected semaglutide 2.4 mg in STEP 1.[4] That is the bridge the approval rests on, and it is a pharmacokinetic argument rather than a clinical one. The distinction between a measured endpoint and a modeled one is the distinction a buyer should hold on to, and it sits alongside the regulatory vocabulary set out in the approval-status article.
Who the 307 were
Mean age was 48. 242 of 307 participants were women and 281 of 307 were White, with 22 Black or African American and 2 Asian participants across both arms.[2] Mean baseline weight was 106.4 kg and mean body-mass index 37.5; 44% had hypertension, 31% dyslipidemia, and 1.5% coronary artery disease.[3] An East Asian population was studied separately in OASIS 2, which randomized participants with and without type 2 diabetes to the 50 mg dose rather than this one.[5]
What OASIS 4 did not establish
It did not measure a cardiovascular outcome, a kidney outcome or a death rate; its endpoints were weight and a quality-of-life score. It did not run past week 71, so nothing in it speaks to maintenance, and no arm was withdrawn, so it says nothing about what happens on stopping. It did not enroll anyone with type 2 diabetes. And with 205 people on the drug, a rare harm would not appear at all.
Every figure above belongs to branded, FDA-approved oral semaglutide at 25 mg daily, supplied free, taken under the fasting and water conditions the formulation requires, in a supervised trial. Sellers listed on the oral semaglutide board generally dispense compounded oral or sublingual preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed. No compounded oral preparation has been through a trial of any size, so the exposure argument that carried this approval — the one showing 82.2% of participants landing in the injection’s range — has no counterpart for a product bought that way.