Skip to content
This GLP
← Research
Evidence

A trial built to show a tie, and the tablet won instead

Orforglipron only had to prove it was no worse than dapagliflozin. All three doses beat it, by up to 0.75 percentage points of HbA1c.

Dana Sullivan6 min read
HbA1c fall at week 40, percentage pointsorforglipron 36 mg1.56orforglipron 12 mg1.50orforglipron 3 mg1.23dapagliflozin 10 mg0.81The trial only had to prove it was no worse.

Orforglipron was approved for weight management in April 2026, covered in the oral one is approved now. This trial tested it where most diabetes drugs are actually decided — as the thing added when metformin is no longer enough [1].

The design deserves a moment because the phrase appears constantly and is rarely explained. A non-inferiority trial does not set out to show a drug is better; it sets out to show it is not worse by more than an agreed margin, here 0.3 percentage points of HbA1c. That is a lower bar, chosen when a new drug offers some other advantage — in this case a tablet rather than an injection — and only needs to hold its own on effect.

It did considerably more than hold its own. At week 40, HbA1c fell 1.23, 1.50 and 1.56 percentage points on the 3, 12 and 36 mg doses against 0.81 with dapagliflozin, giving treatment differences of -0.42 (95% CI -0.62 to -0.23), -0.70 (-0.90 to -0.49) and -0.75 (-0.96 and below). A trial built to demonstrate equivalence producing superiority is a stronger signal than a superiority trial succeeding, because nobody designed it to find that.

One design choice makes the result more useful than it might have been. The primary analysis used a treatment-regimen estimand, counting each participant’s data whether or not they stopped the study drug or added another glucose-lowering agent. That measures what happens when someone is put on a drug rather than what happens if they stay on it perfectly, which is the more honest question given how many people discontinue — the figures in the network meta-analysis of nineteen drugs put orforglipron among the highest for stopping because of adverse events.

The trial was open-label, so everyone knew what they were taking. HbA1c is a laboratory measure and difficult to influence directly, but knowing you are on the new drug can change adherence and effort in ways the estimand absorbs rather than removes. That is a smaller concern here than in a trial with subjective endpoints, and it is worth noting alongside the pipeline picture in the ecnoglutide trial.

Frequently asked

What is a non-inferiority trial?
One designed to show a drug is not worse than a comparator by more than an agreed margin — here 0.3 percentage points of HbA1c. It is a lower bar than proving superiority, used when the new drug offers another advantage such as being a tablet.
Did orforglipron beat dapagliflozin?
Yes, on blood sugar. All three doses cleared non-inferiority and then showed significantly greater HbA1c reductions, by 0.42 to 0.75 percentage points.
Does that make it the better drug?
Not necessarily. Dapagliflozin belongs to a class with benefits for heart failure and kidney disease that an HbA1c endpoint does not measure, so this answers a narrow question.

Sources

  1. [1] Welch M, Forst T, Jia W, Del Pino PO, et al. (2026). Orforglipron compared with dapagliflozin in adults with type 2 diabetes and inadequate glycaemic control with metformin (ACHIEVE-2): a multicentre, randomised, non-inferiority, open-label, phase 3 trial The Lancet. PMID 42259339

Where to get it

Best GLP-1 injections

Every injectable seller we can verify, with the price each one publishes and an honest read of what the trials measured.

Compare providers →

More in Evidence