The First Pill-Versus-Pill Trial

ACHIEVE-3 put two oral GLP-1s directly against each other in 1,698 adults with type 2 diabetes, and orforglipron beat oral semaglutide on blood sugar at every dose pairing tested — including the low orforglipron dose against the high semaglutide dose.^[1] It also produced more gastrointestinal side effects, roughly double the dropouts, and a larger rise in resting heart rate. Both halves are the result.

What the trial did

Adults with type 2 diabetes inadequately controlled on at least 1,500 mg of metformin daily, with an HbA1c between 7.0% and 10.5% and a BMI of 25 or above, were randomized in equal quarters to orforglipron 12 mg, orforglipron 36 mg, oral semaglutide 7 mg, or oral semaglutide 14 mg, and followed for 52 weeks across 131 centers in Argentina, China, Japan, Mexico and the United States. Baseline HbA1c averaged 8.3%.

The design asked non-inferiority first — could orforglipron match semaglutide within 0.3% — with superiority tested only if that was met. It was met, and superiority followed.

The row worth pausing on is the first one, third column: the low dose of orforglipron beat the high dose of semaglutide, by 0.24 percentage points. That is a modest difference and a meaningful one, because it is the comparison a formulary makes when deciding what to put where.

The design problem this particular pair creates

That creates a route to the result that is not pharmacology. Semaglutide’s blood level depends on a daily ritual being performed correctly; orforglipron’s does not. A gap of 0.24 to 0.68 percentage points could reflect the molecule, or the fasting window, or both, and a trial where participants know their assignment cannot separate them.

We are not claiming that is what happened. We are saying the trial cannot tell you, and that this ambiguity is worth more here than in an ordinary head-to-head. It also cuts in an interesting direction: if convenience is part of why orforglipron did better, that is still a real advantage to a real patient — it is just not the advantage the word “superior” implies.

The cost side of the result

Gastrointestinal events hit 59% of the orforglipron 12 mg arm and 58% of the 36 mg arm, against 37% on semaglutide 7 mg and 45% on 14 mg. Most were mild to moderate. The consequential number is what people did about them.

Roughly twice as many people quit orforglipron because of side effects. That is part of the finding, not a footnote to it.

The pulse difference deserves its own sentence. A resting heart rate rise of 4 to 5 beats per minute is a known class effect, but it was two to three times larger with orforglipron than with oral semaglutide here. In a population with type 2 diabetes and a mean BMI above 25, that is a variable a cardiologist would want to know about, and it has no bearing on the HbA1c headline whatsoever.

Four deaths occurred across the trial — one in each orforglipron arm and two on semaglutide 7 mg. With 1,698 participants over a year, that is not a signal in either direction.

What this does not tell you

• It is not a weight-loss trial. The endpoint was HbA1c in people with type 2 diabetes on metformin. Orforglipron’s obesity evidence is a separate trial with a separate design, and reading a diabetes HbA1c result as a weight result is the error this register calls out in sellers.
• The comparator was 7 and 14 mg oral semaglutide — not the higher 25 and 50 mg strengths. A different comparator arm could produce a different gap.
• Eli Lilly funded it, and Eli Lilly makes orforglipron. The comparator is a Novo Nordisk product. That does not invalidate a Lancet phase 3 trial; it is context every reader of an industry head-to-head is entitled to.
• Fifty-two weeks is not a long time for a chronic condition, and nothing here speaks to cardiovascular or kidney outcomes, which is where semaglutide has an evidence base orforglipron does not yet have.

For what orforglipron is and how it is labeled, see Foundayo. For the drugs still behind it, the oral pipeline. And for how side-effect burden compares across the class generally, which GLP-1 has the worst side effects.