This is a trade, not a win. Nothing about glucose control got better.
What changed is everything around it. [1]
The outcome that did not move
HbA1c was the primary endpoint. The difference was −0.08 percentage points, with an interval running from −0.28 to 0.12.
That interval sits across zero. It is a null, and here a null is the objective: the point of the strategy is to take mealtime insulin away without losing control of blood sugar.
A trial designed to show something is better and a trial designed to show something is no worse are different objects, and the second gets read as the first constantly.
What the swap bought
Weight fell 4.94 kg, interval 7.37 to 2.51. Total daily insulin fell 33.32 units, interval 52.23 to 14.41.
Thirty-three units a day is a large reduction. So is five kilograms. For somebody on four injections a day, fewer of them is its own benefit that no endpoint captures — and weight removed this way is still weight that comes back when the drug stops.
What it costs
A GLP-1 costs many times what mealtime insulin costs. This analysis does not price anything.
So the trade in money is the reverse of the trade in convenience: fewer injections, less weight, and a much larger bill. Whether that is worth it depends on what coverage does with it, which is the question that decides most of these decisions before any seller quotes a figure.
Nothing on this page is a reason to change an insulin regimen. Reducing insulin without a prescriber directing it is dangerous, and this is a clinical decision rather than a comparison-shopping one.
How long it holds
Unknown. The trials are short and the authors say so.
Insulin requirements move over years, and a strategy that holds HbA1c for months is not proof of one that holds it for a decade — the same gap that leaves most of the value question resting on modeled time rather than measured time.