Oral semaglutide 25 mg vs the injection

Novo Nordisk now sells the same molecule two ways for weight management: a once-weekly injection at 2.4 mg and a once-daily 25 mg tablet. No trial has run the two against each other. In June 2026, a company-funded analysis in Diabetes, Obesity and Metabolism compared them indirectly and concluded that oral and injected semaglutide produce weight loss close enough to call equivalent^[1]. The estimated gap was about one percentage point of body weight, favouring the injection, with a confidence interval that crossed zero.

What the analysis compared

The authors used a naive Bucher indirect treatment comparison. It takes two placebo-controlled trials, treats placebo as the shared reference point, and subtracts one drug-versus-placebo result from the other^[1].

The inputs were OASIS 4, which tested oral semaglutide 25 mg against placebo in 307 adults^[5], and STEP 1, which tested subcutaneous semaglutide 2.4 mg against placebo in 1,961 adults^[6]. Both were double-blind phase 3 trials with the same entry criteria, both excluded people with diabetes, and both added the drug to one lifestyle programme: a 500 kcal daily deficit and 150 minutes of activity a week. Baseline weight was 105.9 kg and 105.3 kg^[1].

There is a pharmacological reason to expect a match. A separate modelling paper found that 82.2% of the oral 25 mg participants in OASIS 4 reached blood levels inside the range seen in STEP 1^[2]. Same molecule, comparable exposure, so comparable effect is the expected result rather than a surprising one.

The weight-loss numbers

Under the efficacy estimand, which models what happens if everyone stays on treatment as randomised, the injection led by 0.55 percentage points of body weight change (95% CI -2.25 to 3.35). Under the treatment regimen estimand, which counts everyone regardless of whether they stopped or added another drug, it led by 1.01 points (95% CI -1.61 to 3.63)^[1].

Both intervals span zero, and both estimates sit far under the 5% weight-loss threshold the FDA treats as the minimum clinically relevant effect, which is the yardstick the authors used^[1].

Responder analyses pointed the same way. For the proportion reaching at least 5% weight loss the estimated odds ratio was 0.68 (95% CI 0.34 to 1.38), an interval far too wide to conclude anything. The 10%, 15% and 20% thresholds showed no clear difference either^[1].

Everything else it measured

Waist circumference leaned the other way, favouring the tablet by 1.78 cm. Systolic blood pressure favoured the injection by 1.92 mmHg. HbA1c differed by 0.01 percentage points. Lipids, fasting glucose and hsCRP landed in the same place: small gaps inside intervals wide enough to be noise^[1].

Quality of life is the one endpoint that reached statistical significance. The tablet scored 7.25 points higher on the IWQoL-Lite-CT physical function scale (p = 0.006). The authors then noted that the minimal clinically important difference on that scale is 12 points, so the gap is not one patients would feel^[1].

Safety was compared descriptively, with no statistical test. Gastrointestinal disorders hit 74.0% of the oral group and 74.2% of the injectable group. Serious adverse events ran at 3.9% on the tablet against 9.8% on the injection, and discontinuation for an adverse event at 6.9% against 7.0%^[1].

Who paid for it

Novo Nordisk funded the work, and Novo Nordisk sells both products compared. Three of the seven authors were Novo Nordisk employees at the time, two of them holding company stock. A fourth was a health economic advisor to the company. Two more worked for Petauri Evidence, the consultancy contracted to run the study. The last author sat on a Novo Nordisk advisory board^[1].

That does not make the arithmetic wrong. It does mean the commercial interest and the conclusion point the same way, and readers can weigh where a result came from.

What the method cannot settle

OASIS 4 enrolled 307 people against 1,961 in STEP 1, and that imbalance drives most of the width in the intervals^[1]. A range of -2.25 to 3.35 percentage points fits equivalence. It also fits a three-point advantage either way.

No population adjustment was applied, on the grounds that recognised effect modifiers in obesity, including age, sex, BMI and weight, were aligned. Geography was not aligned between them. STEP 1 recruited across 16 countries on four continents, OASIS 4 only in the United States, Canada, Germany and Poland, leaving the oral arm 91.5% White against 75.1%^[1].

One design detail deserves particular attention here. The placebo in OASIS 4 was a tablet and the placebo in STEP 1 was an injection. The Bucher method rests on those two being interchangeable, and the authors state plainly that they assumed it and that the effect is unknown^[1]. Endpoints were also read at week 64 and week 68, because escalation took 12 weeks in one trial and 16 in the other.

What this means for a tablet-versus-pen decision

Both forms are prescription-only everywhere we cover, and our semaglutide regulation overview sets out the position country by country. In the European Union both forms sit under one Wegovy marketing authorisation, and the EMA product page states that the tablets are for adults only while the pens extend to adolescents from 12 years^[4].

Practical differences sit on the label rather than in the efficacy data. US prescribing information specifies a tablet escalation from 1.5 mg through 4 mg and 9 mg to a 25 mg maintenance dose over 90 days, against a four-step weekly escalation to 2.4 mg for the injection^[3]. Tablets have to be taken after eight hours of fasting, with a 30-minute wait before eating^[4]. The pen has no such restriction but needs cold storage and a needle.

The label anticipates people moving between the two, with switching instructions in both directions and a note to consider the 1.7 mg injection for anyone who cannot tolerate the tablet^[3]. None of that is a recommendation from us. This article is educational, not medical advice, and any decision belongs with your prescribing clinician. Our semaglutide reference page covers the mechanism and the safety signals.

What we don't yet know

An indirect comparison is a bridge, not a measurement. What exists here is two trials run six years apart, in different countries, against differently delivered placebos, joined by arithmetic. The company argues that head-to-head trials of two formulations are rarely run^[1]. The cost of that convention is an equivalence claim resting on modelled exposure^[2] and a wide-interval estimate.

Adherence is the question the analysis cannot reach. A daily fasted tablet and a weekly injection make different demands over a year, and neither trial was built to compare them. Real-world persistence data on the tablet will say more than any reworking of these datasets. Until then, the evidence points to similar efficacy without proving it. Our semaglutide page tracks the record as it grows.