ATTAIN-1 has reported, which matters because a great deal was written about this trial before it did. It is the phase 3 obesity trial of orforglipron, the first GLP-1 receptor agonist that is a small molecule rather than a peptide, and its results landed in a journal in late 2025 with a full results posting on the trial registry behind them.[1][2] What the molecule is and how it is taken is covered in the orforglipron article. This page is about the trial: how big the effect was, which number you are looking at when someone quotes it, and the two columns that refuse to line up with dose.
What was run
A multinational, double-blind, phase 3 trial randomized 3,127 adults with obesity and no diabetes in a 3:3:3:4 ratio to once-daily orforglipron at 6 mg, 12 mg or 36 mg, or to placebo, alongside diet and activity advice, for 72 weeks.[1] The posted results give the arm sizes as 723, 725 and 730 against 949 on placebo. Mean age was 45.1 years (SD 12.1), mean starting body weight was 103.16 kg (SD 22.11), and 2,009 of the 3,127 participants — 64.2% — were women.[2]
This was a placebo-controlled trial, not a comparison. No arm received semaglutide, tirzepatide or any other active drug, so nothing in it answers the question most buyers arrive with, which is how a daily tablet stacks up against a weekly injection.
Two estimands, two headline numbers
The published abstract states that the primary endpoint was percent change in body weight at week 72 assessed under the treatment-regimen estimand — the analysis that counts everyone as randomized, whatever they later did. Under it, mean weight change was −7.5% (95% CI, −8.2 to −6.8) at 6 mg, −8.4% (−9.1 to −7.7) at 12 mg and −11.2% (−12.0 to −10.4) at 36 mg, against −2.1% (−2.8 to −1.4) on placebo, with P < 0.001 for every comparison.[1] Those are per-arm intervals. The abstract does not publish the between-group difference or its confidence interval at all.
The registry does, on a different analysis. Its posted primary outcome is a mixed-model estimate described as the unconditional average treatment effect, and it reports −12.35% at 36 mg against −0.87% on placebo, a difference of −11.48 percentage points (95% CI, −12.35 to −10.61). At 12 mg the difference is −8.40 (−9.22 to −7.58) and at 6 mg −6.95 (−7.71 to −6.18).[2]
Both are real and they are not interchangeable. The placebo arm moves most: −2.1% in one account and −0.87% in the other, which is most of why the gap widens from roughly nine points to eleven and a half. Anyone quoting “eleven and a half points” is quoting the registry; anyone quoting “11.2% weight loss” is quoting a single arm, not a difference. The thresholds come from the abstract: at 36 mg, 54.6% lost 10% or more, 36.0% lost 15% or more and 18.4% lost 20% or more, against 12.9%, 5.9% and 2.8% on placebo.[1]
The dose on the box is not the dose in the paper
The trial arms are 6, 12 and 36 mg in both the paper and the registry. A later subgroup analysis of these same ATTAIN-1 arms names them 5.5, 9 and 17.2 mg, the strengths the marketed tablet is sold in.[3] Nothing was changed about who received what: one set of arms carries two sets of milligram labels, because the trial used an investigational formulation and the approved product is counted differently. A reader holding a product listing next to the published trial will meet both, and neither is a typographical error. The approved ladder and where it comes from belong to the orforglipron article.
Where tolerability stops tracking dose
The weight column climbs at every step. Two of the tolerability columns do not. Nausea was reported by 209 of 723 (28.9%) at 6 mg, 260 of 724 (35.9%) at 12 mg and 245 of 728 (33.7%) at 36 mg, against 99 of 948 (10.4%) on placebo. Constipation followed the same shape: 21.6%, then 29.8%, then 25.4%, against 9.3%.[2] The middle dose was the least comfortable one in the trial.
Vomiting and diarrhea did rise monotonically — vomiting 13.0%, 21.4% and 24.0% against 3.5%, diarrhea 21.0%, 22.8% and 23.1% against 9.6% — so this is not a general artifact.[2] The plausible reading is escalation schedule rather than steady-state exposure: participants climbing toward 36 mg spend longer stepping through intermediate strengths, and symptom reporting is concentrated in that climb. The trial does not test that explanation, and nothing here should be read as a reason to expect a higher dose to feel better.
Serious adverse events went the same unexpected way and further. They affected 28 of 728 participants (3.8%) at 36 mg — fewer, in proportion, than the 46 of 948 (4.9%) on placebo, and fewer than either lower dose at 5.5% and 5.4%. One death occurred in each of the placebo, 6 mg and 12 mg arms and none at 36 mg.[2] In a trial of this size those are small counts, and a difference of eighteen events across two arms is well inside what chance produces. The correct conclusion is that the highest dose showed no excess of serious harm, not that it prevented any.
Who stopped, and who left
The discontinuation figure that matters is in the abstract, and it does rise with dose: adverse events led to treatment discontinuation in 5.3% to 10.3% of participants across the three orforglipron arms, against 2.7% on placebo.[1] One in ten leaving the top dose over side effects sits oddly next to the serious-event column, and both are true: the events that made people stop were overwhelmingly the ordinary gastrointestinal ones, which are miserable without being serious in the regulatory sense.
Leaving the study is a separate count and it runs the other way. Of those randomized, 181 of 949 on placebo did not complete, against 98, 95 and 91 in the three drug arms — roughly 19% against 13%. The single largest contributor to that gap is the withdrawal reason recorded as lack of efficacy: 26 on placebo, against 2, 1 and none on orforglipron.[2] People on a tablet that was not working stopped showing up. That is a familiar pattern in obesity trials and it is the reason the estimand question above is not academic.
What a pill buys against an injection
Against the injectable benchmarks, the honest summary is that orforglipron at its top dose lands between a placebo and the better injections rather than alongside them. Semaglutide 2.4 mg over 68 weeks in 1,961 adults produced −14.9% against −2.4%, an estimated treatment difference of −12.4 percentage points (95% CI, −13.4 to −11.5), detailed in the STEP 1 article.[5] Tirzepatide 15 mg over the same 72 weeks used here produced −20.9% (95% CI, −21.8 to −19.9) against −3.1%, covered in the SURMOUNT-1 article.[4]
The comparison that complicates the pitch is the tolerability one. Tirzepatide discontinued for adverse events in 4.3%, 7.1% and 6.2% of participants across its three doses against 2.6% on placebo.[4] Orforglipron’s top dose reached 10.3% while delivering roughly half the weight change. A tablet removes the needle; on these two trials it does not remove the reason people quit. What each route costs and demands is set out in the oral-versus-injectable article, and the expected weight-loss tool puts trial averages next to a starting weight.
The older subgroup did not do worse
A post hoc analysis pooled this trial with its diabetes counterpart and looked at the 616 randomized participants aged 65 or older, 613 of whom were treated. Within this trial, week-72 weight change at the top dose was −13.0% (95% CI, −15.7 to −10.4) against −1.6% (−3.2 to 0.1) on placebo, with −11.3% at the middle dose and −7.9% at the lowest.[3] That top-dose figure is larger than the whole-trial figure, not smaller, and the authors report the findings in those under 65 as comparable. It is a subgroup analysis run after the results were known, on a few hundred people, and it is descriptive rather than a test — but it is the opposite of what the usual caution about older patients would predict.
What ATTAIN-1 did not test
It did not test orforglipron against another drug. It did not test it in people with type 2 diabetes — that was a separate trial. It did not measure cardiovascular outcomes, and it was never designed to: no events endpoint was adjudicated and the population was a 45-year-old obesity cohort rather than a cardiac one. Its glycemic column is small by design, with glycated hemoglobin falling 0.38 points at 36 mg against 0.03 on placebo, a difference of 0.35 points.[2]
It also has not finished. Two secondary endpoints — weight change to week 176 and time to onset of type 2 diabetes in participants who entered with prediabetes — are listed on the registry with no results posted against them.[2] Nothing about durability past 72 weeks, or about whether the drug prevents diabetes, can be read out of this trial yet.
What was in the tablet
Every figure above belongs to a manufacturer-supplied investigational formulation of orforglipron, taken once daily under protocol for 72 weeks with diet and activity support, in adults with obesity and no diabetes. Any product sold online that is described as orforglipron but is not the approved branded tablet at a labeled strength is a compounded preparation, and compounded drugs are not FDA-approved and receive no FDA review of safety, efficacy or quality before they are dispensed. The formats currently listed on the oral GLP-1 board are worth reading against that distinction, because a small molecule with a marketed tablet has an approved counterpart to be compared with, which is not true of every oral preparation on sale.