Inside the OASIS-1 Methodology: What Most Summaries Skip

At a glance

| Detail | Value | |---|---| | Trial name | OASIS-1 (NCT05035095) | | N randomized | 667 | | Intervention | Oral semaglutide 50 mg once daily | | Comparator | Matched oral placebo | | Population | Adults with BMI ≥30, or ≥27 with at least one weight-related comorbidity; no diabetes | | Duration | 68 weeks (16-week escalation + 52-week maintenance) | | Primary endpoint | Percentage change in body weight from baseline to week 68 | | Key result | −15.1% (semaglutide) vs −2.4% (placebo) | | Publication | The Lancet, 2023 |

Why the Design Matters More Than the Headline

A 15.1% weight reduction sounds definitive. But that single number conceals decisions about who was enrolled, how the drug was titrated, which statistical estimand was reported in the headline, and what happened to the roughly 1 in 4 participants who discontinued treatment early. Each choice is defensible on its own. Taken together, they form a frame that can make the same data look different depending on which lens you apply. This page unpacks that frame, section by section, using the full OASIS-1 publication and related regulatory and clinical sources.

Randomization and Blinding

OASIS-1 used a 1:1 randomization ratio, assigning 334 participants to oral semaglutide and 333 to placebo. Randomization was stratified by sex (male vs female) and by the presence or absence of prediabetes at screening. Stratification by prediabetes status was a smart inclusion: GLP-1 receptor agonists can affect glycemic parameters differently in people with impaired glucose tolerance, and unbalanced prediabetes rates between arms could confound both weight and metabolic secondary endpoints.

The trial was double-blind with matched placebo tablets. This is worth noting because earlier oral semaglutide trials in the PIONEER program used a dosing-condition design (fasting state, limited water, 30-minute post-dose food restriction) that made true blinding difficult in practice. OASIS-1 retained those dosing conditions for both arms, meaning every participant experienced the same morning routine. The practical blinding quality was therefore higher than it might appear on paper, though GI side effects (nausea, vomiting) could still unblind participants or investigators in some cases.

Inclusion and Exclusion Criteria: Who Got In

The enrolled population had a mean BMI of approximately 37.5 kg/m² and a mean body weight near 105 kg. Participants needed a BMI ≥30, or ≥27 with at least one weight-related comorbidity such as hypertension, dyslipidemia, or obstructive sleep apnea. People with type 1 or type 2 diabetes were excluded. So were those with a history of pancreatitis, personal or family history of medullary thyroid carcinoma or MEN2 syndrome (consistent with semaglutide prescribing information), or prior bariatric surgery.

Two exclusion criteria deserve attention from a generalizability standpoint. First, the diabetes exclusion. While it makes pharmacologic sense to isolate the weight-loss signal from glucose-lowering confounders, it means the OASIS-1 data cannot be directly applied to the large population of patients with obesity and coexisting type 2 diabetes. Separate OASIS trials (OASIS-2 and beyond) address that gap. Second, the trial excluded people using other anti-obesity medications within 90 days. This is standard practice, but it means we cannot extrapolate OASIS-1 results to patients switching from, say, phentermine-topiramate.

The 16-Week Dose Escalation

Oral semaglutide 50 mg was not the starting dose. Participants began at 3 mg daily for 4 weeks, then 7 mg for 4 weeks, then 14 mg for 4 weeks, and finally 25 mg for 4 weeks before reaching the target 50 mg dose at week 16. This 16-week ramp is longer than the escalation used in injectable semaglutide 2.4 mg trials like STEP 1, which reached full dose by week 16 as well but used fewer intermediate steps.

The escalation matters for two reasons. Clinically, it reduces GI adverse events during the titration period, when nausea and vomiting are most common with GLP-1 receptor agonists. Analytically, it means only 52 of the 68 trial weeks were spent at the full 50 mg dose. The weight-loss curve was still declining at week 68 in the published data, raising the question of whether a longer maintenance period would have produced even greater separation from placebo.

Primary Endpoint Definition

The primary endpoint was the percentage change in body weight from baseline (randomization) to week 68. This is a continuous outcome, not a responder threshold. The co-primary confirmatory endpoint was achievement of ≥5% body-weight reduction at week 68 (a binary outcome). Having both continuous and categorical co-primaries is now standard in obesity trials following FDA guidance on obesity drug development, which expects both types of evidence.

Weight was measured in a fasted state at clinic visits. The trial did not report home-scale weights, which limits our understanding of day-to-day weight variability. Body composition (fat mass vs lean mass) was not a pre-specified endpoint, a gap that matters because GLP-1-mediated weight loss can include meaningful lean-mass reduction, as documented in sub-studies of the STEP program.

The Estimand Framework: Two Numbers, One Trial

This is where OASIS-1 differs from older obesity RCTs in a way most summaries ignore entirely.

The trial pre-specified two estimands under the ICH E9(R1) addendum framework:

| Estimand | What it answers | Headline result | |---|---|---| | Treatment policy | What is the effect of being assigned to oral semaglutide, regardless of whether participants stayed on drug? | −12.7% vs −1.8% | | Trial product | What is the effect if participants stay on drug for the full 68 weeks? | −15.1% vs −2.4% |

The −15.1% figure that appears in most media coverage and conference slides comes from the trial-product estimand. This estimand effectively asks: "Among people who tolerated and continued the drug, how much weight did they lose?" It uses a mixed model for repeated measures (MMRM) that, for participants who discontinued treatment, censors their data at the point of discontinuation and imputes subsequent values from those who remained on treatment.

The treatment-policy estimand, at −12.7%, includes all randomized participants and uses a pattern-mixture model that assumes discontinued participants regain weight at rates observed in similar populations. This number is closer to an intention-to-treat result and is the more conservative figure.

Both estimands are valid. But the gap between them (15.1% vs 12.7%) tells you something important: roughly 2.4 percentage points of the headline result reflect the removal of early discontinuers from the analytic frame. Clinicians prescribing oral semaglutide 50 mg in practice, where adherence and discontinuation rates may be higher than in a controlled trial, should keep the treatment-policy number in mind.

Discontinuation Rates and Their Impact

Approximately 26% of participants in the semaglutide arm and 16% in the placebo arm discontinued treatment before week 68. Among semaglutide discontinuers, the most common reasons were adverse events (predominantly GI), followed by withdrawal of consent. This differential discontinuation is not unusual for GLP-1 trials, but it does amplify the gap between the two estimands described above.

A useful comparison: in the STEP 1 trial of injectable semaglutide 2.4 mg, the discontinuation rate in the active arm was approximately 17%. The higher rate in OASIS-1 (26%) may reflect the GI burden of the high oral dose, the 50 mg tablet being roughly 3.5 times the previously approved maximum oral dose of 14 mg. This comparison is imperfect (different populations, different trial designs), but it does raise a practical question about real-world tolerability of the 50 mg oral formulation.

Statistical Analysis Plan

The primary analysis used an MMRM model with treatment, visit, and treatment-by-visit interaction as fixed effects, along with the stratification factors (sex, prediabetes status) and baseline body weight. The MMRM approach handles missing data under a missing-at-random assumption for the trial-product estimand. For the treatment-policy estimand, a multiple-imputation approach using a pattern-mixture model was applied, referencing the placebo arm's weight trajectory for participants who discontinued.

Multiplicity was controlled using a pre-specified hierarchical testing procedure. The co-primary endpoints (percentage weight change and proportion achieving ≥5% loss) were tested first, followed by key secondary endpoints in a fixed sequence. This means the p-values for secondary outcomes like waist circumference or systolic blood pressure are protected from type I error inflation, but only if all preceding tests in the hierarchy reached significance, which they did.

Sample size was calculated to provide greater than 90% power to detect a treatment difference in percentage weight change, assuming a 10-percentage-point difference between arms with a standard deviation of approximately 8%. The observed 12.7-percentage-point difference (treatment-policy estimand) exceeded the assumption, meaning the trial was well powered for its primary question.

What the Comparator Choice Tells You

OASIS-1 was placebo-controlled. There was no active comparator arm testing injectable semaglutide 2.4 mg or another anti-obesity medication. This means OASIS-1 can tell you that oral semaglutide 50 mg beats placebo, but it cannot tell you whether it matches injectable semaglutide 2.4 mg. Cross-trial comparisons with STEP 1 are tempting (both show approximately 15% weight loss at their trial-product estimand), but differences in population, baseline weight, concomitant lifestyle intervention intensity, and discontinuation handling make direct comparison unreliable.

A head-to-head oral vs injectable trial would be the only rigorous way to answer that question. Until such data exist, the claim that "oral semaglutide matches injectable" remains an inference, not evidence.

Limitations the Authors Acknowledged

The OASIS-1 investigators noted several limitations in the published paper:

• Short duration relative to a chronic disease. Obesity is a lifelong condition. 68 weeks tells us about medium-term efficacy but nothing about 2-year, 5-year, or lifetime weight maintenance.
• Limited diversity. The trial enrolled participants across multiple countries, but demographic breakdowns suggest a predominantly White population, limiting generalizability to other racial and ethnic groups.
• No cardiovascular outcomes. Weight loss is a surrogate endpoint. Whether the oral 50 mg dose reduces major adverse cardiovascular events (MACE) is being addressed in separate trials. Until then, the cardiovascular benefit seen with injectable semaglutide in the SELECT trial cannot be assumed to apply to the oral formulation.
• No body-composition data. As noted above, the proportion of weight lost as fat vs lean mass was not assessed.

What This Means for Clinical Practice

The OASIS-1 methodology is rigorous. The two-estimand design is transparent and gives clinicians both an optimistic and a conservative reading of the data. The 15.1% trial-product result is the upper bound of expected benefit for patients who tolerate and persist with the drug. The 12.7% treatment-policy result is the more realistic expectation for a clinic population where a quarter of patients may discontinue within the first year.

Clinicians should also consider the 16-week escalation period when counseling patients. Full-dose efficacy does not begin until month 4. Early discontinuation during the escalation phase, often driven by GI side effects, will reduce real-world effectiveness below even the treatment-policy estimate.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. Knop FK, Aroda VR, do Vale RD, et al. Oral semaglutide 50 mg taken once a day in adults with overweight or obesity (OASIS 1): a randomised, double-blind, placebo-controlled, phase 3 trial. The Lancet. 2023;402(10403):705-719. PubMed
2. Wilding JPH, Batterham RL, Calanna S, et al. Once-weekly semaglutide in adults with overweight or obesity (STEP 1). N Engl J Med. 2021;384(11):989-1002. PubMed
3. Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and cardiovascular outcomes in obesity without diabetes (SELECT). N Engl J Med. 2023;389(24):2221-2232. PubMed
4. FDA. Rybelsus (semaglutide) prescribing information. FDA Label
5. ICH E9(R1) Addendum on estimands and sensitivity analysis in clinical trials. PubMed
6. FDA Guidance: Developing Products for Weight Management (Revision 1). FDA