The Top Dose Keeps Not Being Worth It

A phase 3 trial put oral orforglipron against dapagliflozin, an SGLT2 inhibitor, in 962 people with type 2 diabetes. Orforglipron won on blood sugar and lost badly on tolerability — one in five stopped taking it at the top dose, against one in seventeen on the comparator.^[1] And that top dose is the third in a row, across three separate trials, to buy almost nothing.

The trial's own result

Adults with type 2 diabetes inadequately controlled on metformin were randomized across 73 sites to orforglipron at 3, 12 or 36 mg, or dapagliflozin 10 mg, for 40 weeks. Average age was 56.1, average HbA1c 8.14%, and average diabetes duration eight years.

Every orforglipron dose beat dapagliflozin on blood sugar at P<0.0001, and no severe hypoglycemia occurred in any arm. The last two columns are the part that decides a real prescription.

Three quarters of a percentage point of HbA1c, for four times the gut trouble and three times the dropouts.

The pattern across three trials

Look at the 12 mg and 36 mg rows above. Tripling the dose produced 0.06 percentage points more HbA1c reduction and took gastrointestinal events from 46% to 54%.

That is not an oddity of this trial. It is the third time this register has found it, in three separately conducted phase 3 studies, and no single paper points it out because each reports only its own doses.

It is worth saying what this does not mean. Higher doses do more for weight in several of these trials even where they do nothing more for blood sugar — the two endpoints plateau at different points, which we set out in the dose that stopped helping. And an individual is not an average: some people genuinely need the top dose. The point is that the last step should be a decision rather than a default.

Why beating dapagliflozin does not settle anything

SGLT2 inhibitors are not a weak comparator chosen to be beaten. They have their own cardiovascular and kidney outcome evidence, they are taken by mouth, they are becoming inexpensive as generics arrive, and in this trial 12% had gastrointestinal side effects against 46–54%.

So a clinician choosing between them is not choosing the one with the better HbA1c. They are weighing three quarters of a percentage point against a threefold difference in the chance the patient abandons treatment — and a drug someone stops taking delivers nothing at all.

The combination question — whether to use both together — is separate and has its own disappointing evidence, in why combination therapy fails in practice.

Limits

• Open-label, though the orforglipron doses were blinded to participants.
• Forty weeks, in a condition managed for decades, with no cardiovascular or kidney outcomes reported.
• Funded by Eli Lilly, which develops orforglipron.
• The discontinuation figures are trial figures. Real-world dropout is generally higher, not lower.