ATTAIN-2 is the diabetes half of orforglipron’s phase 3 obesity program, and the more interesting half, because a drug that lowers blood sugar and body weight at the same time is being sold on both and tested against neither. It randomized adults who already had type 2 diabetes to a once-daily tablet or placebo for 72 weeks, on top of whatever they were already taking.[1] What the molecule is and how the approved tablet is dosed belongs to the orforglipron article; the sibling trial in people without diabetes is covered in the ATTAIN-1 article. This page is about which of ATTAIN-2’s columns moved and which did not.
What was run
A 72-week, phase 3, double-blind, placebo-controlled trial at 136 sites in ten countries screened 2,859 people and randomized 1,613 in a 1:1:1:2 ratio after a dose-escalation phase: 329 to orforglipron 6 mg, 332 to 12 mg, 322 to 36 mg and 630 to placebo, as an adjunct to lifestyle modification. Eligibility required a body-mass index of 27 or above and a glycated hemoglobin of 7 to 10%. Baseline body weight was 101.4 kg (SD 22.5), body-mass index 35.6 (SD 6.6) and glycated hemoglobin 8.05% (SD 0.75). Of those randomized, 757 (46.9%) were female, and 1,444 (89.5%) completed the study.[1]
The control arm took a placebo tablet while continuing background glucose-lowering therapy. No arm received semaglutide, tirzepatide, metformin as a comparator or any other active drug, so the trial contains no answer to the question a buyer holding two price lists is actually asking. What each route costs and demands is set out in the oral-versus-injectable article.
The weight result, and which version of it is being quoted
The primary endpoint was mean percent change in body weight at week 72 under the treatment-regimen estimand, which counts every randomized participant regardless of what happened afterward. Under it, weight changed −5.1% (95% CI, −6.0 to −4.2) at 6 mg, −7.0% (−7.8 to −6.2) at 12 mg and −9.6% (−10.5 to −8.7) at 36 mg, against −2.5% (−3.0 to −1.9) on placebo. The estimated treatment differences were −2.7 (95% CI, −3.7 to −1.6), −4.5 (−5.5 to −3.6) and −7.1 percentage points (−8.2 to −6.1), all at p < 0.0001.[1]
The registry posts the same endpoint on a different analysis and returns larger separations — −10.54% at 36 mg against −2.21%, a difference of −8.33 (95% CI, −9.34 to −7.32).[2] Both are real, neither is wrong, and the eight-point figure and the seven-point figure describe the same trial under two pre-declared rules about who counts.
The threshold columns follow the dose cleanly. A reduction of 5% or more was reached by 49.8%, 60.2% and 72.8% against 24.4% on placebo; 10% or more by 23.9%, 35.5% and 50.1% against 7.0%; and 15% or more by 7.3%, 17.7% and 28.4% against 1.9%. Waist circumference fell 9.16 cm at the top dose against 2.67 cm.[2]
The glucose column is the larger one
Glycated hemoglobin fell 1.29, 1.60 and 1.79 points across the three doses against 0.14 on placebo, a difference at the top dose of −1.65 points (95% CI, −1.82 to −1.49). Fasting serum glucose fell 45.8 mg/dL against a rise of 1.2. A target below 7.0% was reached by 85.1% against 23.0%, and below 6.5% by 75.0% against 10.6% — a risk difference of 64.42 points (95% CI, 58.67 to 70.18).[2]
Set against the weight result, that is the trial’s real asymmetry. Seven points of weight is a moderate obesity result; three quarters of a diabetes population dropping below the diagnostic threshold on glycated hemoglobin is a large diabetes result. A tablet marketed on the first of those is carrying the weaker of its two findings.
Three columns that did not move
The abstract summarizes the secondary results in a single sentence: all prespecified weight and cardiometabolic measures, including glycated hemoglobin, improved significantly.[1] The posted results list is longer than that sentence, and three of its entries did not separate from placebo.
Pooled across the three doses, diastolic blood pressure fell 1.32 mm Hg against 1.39 on placebo — a difference of 0.07 mm Hg (95% CI, −0.74 to 0.88; p = 0.869), which is no difference at all. Systolic pressure did separate, at −3.43 (95% CI, −4.75 to −2.10).[2] A drug that lowers one half of a blood-pressure reading and not the other is not a blood-pressure drug, and what the class does to pressure generally is set out in the blood pressure article.
Fasting insulin moved only at the top dose. The geometric mean difference against placebo was +2.6% at 6 mg (p = 0.530), +0.6% at 12 mg (p = 0.898) and −10.0% at 36 mg (95% CI, −17.7 to −1.7; p = 0.019).[2] Two of the three doses lowered glycated hemoglobin by more than a point while leaving circulating insulin where it was.
And on quality of life the split is between body and mind. The physical component of the SF-36 improved by 1.05 points more than placebo (95% CI, 0.36 to 1.73; p = 0.003); the mental component differed by 0.40 points (95% CI, −0.28 to 1.07; p = 0.249), with both arms scoring slightly lower at week 72 than at baseline.[2] Losing ten percent of body weight and reaching a glucose target did not register on the mental-health scale the trial itself chose.
Tolerability, and the counts that do not track dose
Treatment discontinuations due to adverse events, mainly gastrointestinal, ran 6.1% to 9.9% across the orforglipron arms against 4.1% on placebo. The most common adverse events were mild to moderate gastrointestinal ones, predominantly during dose escalation.[1]
The serious-event and death columns do not order themselves by dose. Serious adverse events affected 24 of 328 (7.3%) at 6 mg, 34 of 331 (10.3%) at 12 mg and 35 of 321 (10.9%) at 36 mg, against 55 of 628 (8.8%) on placebo — the lowest rate in the trial belongs to the lowest drug dose, and the placebo arm sits between the drug arms. Ten deaths were recorded: four on placebo, none at 6 mg, four at 12 mg and two at 36 mg.[2] Investigators deemed all of them unrelated to study treatment except one case in the placebo group and one in the 12 mg group, and for the orforglipron case the abstract records that no treatment-related association was reported.[1]
Ten deaths spread across four arms in a 72-week trial is a small-numbers column, and the arm carrying the most of them is the middle dose rather than the highest. Nothing in that pattern supports a dose-related harm reading, and nothing in it supports the opposite reading either.
Where a daily tablet lands against a weekly injection
No injectable was in this trial, so the only available comparison is across trials with separate placebo arms — a weaker instrument than randomization, and one that ignores differences in protocol, duration and background therapy. With that carried in full: in adults with obesity and type 2 diabetes, tirzepatide 15 mg over 72 weeks returned −14.7% against −3.2%, a difference of −11.6 percentage points (95% CI, −13.0 to −10.1),[4] and semaglutide 2.4 mg over 68 weeks returned −9.6% against −3.4%, a difference of −6.2 points (95% CI, −7.3 to −5.2).[5] Those trials are covered in the SURMOUNT-2 article and the STEP 2 article.
Orforglipron’s 7.1 points sits above the weekly semaglutide figure and well below the weekly tirzepatide one, in the same indication and on the same estimand. A tablet that beats one injection and loses to another is a genuine option rather than a replacement, and the discontinuation gap runs the wrong way for it: tirzepatide in that trial recorded fewer than 5% discontinuing for adverse events.[4]
Older participants, and a second set of milligram labels
A post hoc analysis pooled this trial with its sibling and examined the 616 randomized participants aged 65 or older, 613 of whom were treated. It names the arms 5.5, 9 and 17.2 mg — the strengths the marketed tablet carries — for the same doses this trial calls 6, 12 and 36 mg. Within ATTAIN-2, week-72 weight change among those 65 and older was −7.5% (95% CI, −9.1 to −5.9), −8.3% (−9.7 to −6.8) and −12.2% (−13.9 to −10.6) against −2.3% (−3.1 to −1.4) on placebo, all at p < 0.001, with findings in those under 65 reported as comparable.[3]
The top-dose figure in the older subgroup is larger than the whole-trial figure, not smaller. It is a post hoc subgroup on a few hundred people and it is descriptive rather than a test, but it runs against the usual expectation that the drug does less in older patients. What the class demands of an older body is covered in the older adults article.
What ATTAIN-2 did not establish
It did not compare orforglipron with any active drug. It adjudicated no cardiovascular outcome and was not designed to: the population was selected by body-mass index and glycated hemoglobin, not by cardiac risk, and a 72-week window accumulates too few events to test. It did not run past 72 weeks, so nothing in it speaks to durability, and it withdrew no arm, so nothing in it speaks to what happens on stopping. It did not enroll anyone without diabetes; that was the sibling trial.
It also did not measure what the tablet does for a person whose diabetes is already well controlled. Entry required a glycated hemoglobin of at least 7%, and the largest results in the trial are glycemic, so the population it describes is one with room to move on glucose. Someone buying this drug purely for weight at a normal glycated hemoglobin sits outside every column on this page.
What was in the tablet
Every figure above belongs to a manufacturer-supplied formulation of orforglipron, escalated under protocol and taken once daily for 72 weeks with lifestyle support, in adults whose diabetes was already under treatment and whose study drug cost them nothing. Any product sold online described as orforglipron that is not the approved branded tablet at a labeled strength is a compounded preparation, and compounded drugs are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed. The formats currently listed on the oral GLP-1 board are worth reading against that distinction.
The number to be careful with is the one from the other trial. Eleven and a half points is ATTAIN-1’s registry figure in people without diabetes. In diabetes the same tablet at the same top dose returned 7.1 points on the published estimand, and 2.5 of the percentage points it was measured against belonged to a placebo arm taking the same background drugs.