ACHIEVE-1 asked one question and answered a second one by accident. It was designed to find out how far a daily pill lowers glycated hemoglobin in people whose type 2 diabetes is not yet being treated with anything. It found that out, and it also produced two dose-response curves from the same 559 people that do not agree with each other: the glycemic effect flattens between 12 mg and 36 mg while the weight effect keeps climbing. The molecule itself is covered in the orforglipron article and the route question in the oral-versus-injectable article; this page is the trial.
The easiest population in the program
559 adults were randomly assigned 1:1:1:1, double-blind, to once-daily oral orforglipron at 3 mg, 12 mg or 36 mg, or to placebo, for 40 weeks — 143, 137, 141 and 138 people respectively. Entry required type 2 diabetes treated with diet and exercise only, a glycated hemoglobin between 7.0% and 9.5%, and a body-mass index of at least 23.0. Participants had to be insulin-naive and to have taken no oral or injectable antihyperglycemic medication in the 90 days before screening, with stable weight for the same period.[1][2]
That is the gentlest starting line available in diabetes: nobody failing metformin, nobody on insulin, nobody with a decade of escalating therapy behind them. Hold onto it, because it explains the most-quoted number in the trial and it is the reason that number does not travel.
What was preregistered, and what was not
The single primary endpoint was the change from baseline to week 40 in glycated hemoglobin. Weight was a key secondary endpoint, not a primary one.[1][2] The distinction matters, because almost every summary of this trial leads with the weight figure. A trial sized and analyzed to settle a glycemic question produces its weight result as a secondary reading, and the ordering of the endpoints is the author’s statement about which claim the trial is built to support.
The result, and the place the curve stops
Mean baseline glycated hemoglobin was 8.0%. At week 40 the estimated mean change was −1.24 percentage points at 3 mg, −1.47 at 12 mg and −1.48 at 36 mg, against −0.41 on placebo. All three doses beat placebo: the differences were −0.83 (95% CI, −1.10 to −0.56), −1.06 (95% CI, −1.33 to −0.79) and −1.07 (95% CI, −1.33 to −0.81), with P < 0.001 for all. Mean glycated hemoglobin at week 40 was 6.5% to 6.7% on orforglipron.[1]
Look at the top two intervals. They are not merely overlapping; they are nearly the same interval. Tripling the dose from 12 mg to 36 mg bought 0.01 percentage points of glycated hemoglobin, a difference indistinguishable from nothing at this sample size. The glycemic dose-response had already finished by 12 mg.
The weight curve had not. Percent change in body weight was −4.5%, −5.8% and −7.6% against −1.7% on placebo — a clean, continuing separation across the same three doses.[1] The registry’s threshold analysis says the same thing: a reduction of 5% or more was reached by 47.1%, 58.6% and 66.4% against 14.6% on placebo.[2] Whatever someone is taking the top dose for, it is not the last increment of glycemic control.
Two estimands, two sets of numbers
The figures above are the ones the publication reports. The registry posts a different primary analysis, restricted to data collected before a participant stopped study drug or started additional glucose-lowering medication: −1.26, −1.59 and −1.45 against −0.15 on placebo, with differences of −1.12 (95% CI, −1.43 to −0.80), −1.44 (95% CI, −1.72 to −1.16) and −1.30 (95% CI, −1.61 to −1.00).[2]
Both sets are honest and they describe different questions. The placebo arm is the tell: −0.41 when everyone is counted regardless of what they went on to take, −0.15 when rescue medication censors the record. The drug effect looks larger in the second, and under it the 12 mg dose actually finishes ahead of 36 mg. Anyone comparing a figure they have seen elsewhere against a figure here should check which analysis produced it before deciding one of them is wrong.
Targets reached, and the outcomes that did not move
On the registry’s analysis, a glycated hemoglobin below 7.0% was reached by 78.2%, 81.3% and 79.0% of the orforglipron groups against 37.1% on placebo, and 6.5% or lower by 66.4%, 68.8% and 68.9% against 18.6%. Fasting serum glucose fell 30.6, 37.4 and 37.8 mg/dL against 1.1, and systolic blood pressure 3.2, 6.1 and 6.0 mm Hg against 0.5.[2]
Two things did not move. Lipids responded only at the higher doses — non-HDL cholesterol was not significantly different from placebo at 3 mg (p = 0.100), nor were triglycerides (p = 0.072). And the Short Form-36 health survey produced nothing at all: the physical component difference against placebo was −0.19 (95% CI, −1.63 to 1.26), −0.04 (95% CI, −1.61 to 1.54) and −0.66 (95% CI, −2.25 to 0.92), with the mental component equally flat.[2] Every point estimate on the physical component ran numerically below placebo. Forty weeks of substantially better glucose control did not register as people feeling better, which is a useful corrective to any pitch built on how a treated month is supposed to feel.
Who stopped, and why that number does not travel
Permanent discontinuation of study drug for an adverse event occurred in 4.4% to 7.8% of orforglipron participants against 1.4% on placebo, with no episodes of severe hypoglycemia.[1] Completion was high: 134 of 143, 125 of 137 and 136 of 141 against 131 of 138.[2] That is the friendliest tolerability figure in the entire orforglipron program, and it is quoted constantly.
It does not survive leaving this population. In ACHIEVE-2, the same three doses in adults already on metformin produced study-drug discontinuations in 15%, 18% and 20% against 6% on dapagliflozin, with gastrointestinal events in 46% to 54% against 12%.[3] In ACHIEVE-3, against oral semaglutide, discontinuation ran 9% and 10% against 4% and 5%, gastrointestinal events 58% to 59% against 37% to 45%, and mean pulse rate rose 3.7 and 4.7 beats per minute against 1.0 and 1.5.[4] Same molecule, same milligrams, three and four times the quit rate.
ACHIEVE-1’s own event table shows where the pressure comes from. Nausea ran 3 of 138 on placebo against 18, 26 and 24; vomiting 2 against 7, 9 and 20; diarrhea 13 against 27, 29 and 36; constipation 5 against 12, 23 and 20; decreased appetite 3 against 5, 14 and 17.[2] Vomiting is the one that scales hard with dose, and it does so at the dose that adds no further glycemic benefit. How the class behaves generally when it is swallowed rather than injected is in the oral semaglutide article.
What the trial does not describe
It ran 40 weeks with a placebo comparator and no active one. It was not powered for, and did not assess, cardiovascular or renal outcomes; five serious adverse events occurred on placebo against 8, 7 and 4 on the orforglipron doses, and four deaths occurred in total across 559 people.[2] The active comparisons live in the sibling trials — against dapagliflozin,[3] against oral semaglutide,[4] and added to titrated insulin glargine, where 546 participants with a median 14.6 years of diabetes and a mean glycated hemoglobin of 8.50% saw changes of −1.58, −1.88 and −1.82 against −0.79 on placebo.[5] The same flattening between 12 mg and 36 mg appears there too.
Forty weeks is also a short horizon for a chronic disease. There was no randomized withdrawal period, so the trial says nothing about what happens to a glycated hemoglobin of 6.5% when the tablets stop, and no maintenance phase, so it says nothing about year two. Its participants entered having taken nothing for their diabetes for at least 90 days; the question a prescriber usually faces — what to add, and when, to someone already on two agents — is a question this design was built to avoid rather than answer.
It is also not a trial of what orforglipron is approved to do. The single orforglipron label on file with the FDA — FOUNDAYO, from Eli Lilly, under NDA220934 — is indicated for weight management, in combination with a reduced-calorie diet and increased physical activity, in adults with obesity or with overweight plus a weight-related comorbid condition. Type 2 diabetes is not among its indications.[6] The weight program behind that approval is a different trial, covered in the ATTAIN-1 article, and why a diabetes result and a weight result are not interchangeable is in the diabetes-versus-weight article.
One further gap is worth naming rather than papering over. The approved product’s labeled ladder runs 0.8, 2.5, 5.5, 9 and 14.5 mg to a maximum of 17.2 mg once daily. ACHIEVE-1 tested 3, 12 and 36 mg. The label states those tablet strengths are orforglipron, giving the calcium-salt equivalents as only about 2% higher, so a salt conversion does not explain the difference, and the label supplies no bridge between the two scales.[6] Until one exists, a trial milligram and a prescription milligram should not be treated as the same quantity.
Finally, the people. 165 of 559 participants enrolled in the United States, 159 in Mexico, 101 in Japan, 83 in India and 51 in China; 245 were recorded as Asian, 145 as White, 143 as American Indian or Alaska Native and 24 as Black or African American, with a mean age of 53.4.[2] No compounded orforglipron exists to buy, and the oral products sellers do supply are a different molecule at a different dose — see the oral board for what is actually on offer.