Berberine's metformin comparison ran once, in 36 people

Thirty-six people split between two drugs is small enough that an ordinary run of luck moves the number about as much as the drug does. With no placebo group, the fall also contains whatever three months of clinic visits and daily pill-taking do on their own. About a third of patients in the add-on arm had short-lived stomach upset.

Whether berberine lowers blood sugar is not the open question. The head-to-head with metformin everyone quotes came from a pilot with no placebo group, never built to size an effect, and no larger trial has repeated it.

The bigger effects reported in women, in people already diagnosed, and in Asian populations each rest on a handful of trials, few enough that ordinary chance could produce them. Doses and trial lengths varied widely, so 0.4 points is an average across regimens rather than the result of any one of them. Berberine caused more stomach and bowel upset than placebo in the large trial.