Multiple meta-analyses of randomised studies show that berberine produces statistically significant reductions in blood sugar, HbA1c, and insulin resistance in type 2 diabetes. The evidence is reasonably consistent but is limited by the low methodological quality of the underlying studies, their almost exclusively Chinese origin, and uncertainty about effects with use beyond three months. The comparison with metformin is based on one small pilot study and has not yet been established as a clinical fact.
Berberine lowers blood sugar in people with type 2 diabetes, and this has now been investigated in dozens of clinical studies. A meta-analysis of 28 randomised studies (2,313 patients,1 shows that berberine reduces fasting blood sugar by an average of 0.54 mmol/L, post-meal blood sugar by 0.94 mmol/L, and HbA1c by 0.54 percentage points compared with a control group. A second meta-analysis of 46 studies2 confirms this and adds that insulin resistance and LDL cholesterol also improve. The authors of that latter analysis describe the evidence as 'strong', but almost all of the underlying studies were conducted in China, which raises the question of whether the results also apply to Western patients.
The comparison with metformin comes from a small randomised pilot study involving just 36 newly diagnosed diabetic patients3. In that study, berberine and metformin (both 0.5 g three times daily for three months) achieved comparable results: HbA1c fell from 9.5% to 7.5% and fasting blood sugar from 10.6 to 6.9 mmol/L. These are impressive figures, but the very small sample size means this is not yet a firm conclusion. The nickname 'natural metformin' is therefore based on one small study and on shared mechanisms of action, not on large comparative trials.
How berberine works has been reasonably well described in animal and cell studies4,5,6. It activates the energy-regulating enzyme AMPK, similarly to metformin, and inhibits the production of new sugar in the liver. It also stimulates the release of the hormone GLP-1 via gut bacteria -- the same hormone mimicked by popular weight-loss drugs such as semaglutide. This provides a biological explanation for the observed effects, but direct evidence in humans for these mechanisms is more limited than the animal study research.
A special salt form of berberine, HTD1801 (combined with ursodeoxycholic acid), was tested in a phase-2 RCT in 2025 (n=113,7. That study showed dose-dependent HbA1c reduction: 0.4% at the low dose and 0.7% at the high dose compared with placebo. However, this is a small study, conducted in China and partly funded by the manufacturer Shenzhen HighTide Biopharmaceutical. Phase-3 confirmation is still ongoing, so caution is warranted when interpreting these results.
An umbrella review of 11 meta-analyses8 raises a critical caveat: the methodological quality of the underlying studies falls short. Larger, independently funded, and longer-duration studies are needed. The meta-analysis of 28 RCTs also indicates that the blood-sugar-lowering effect weakened with treatment lasting longer than 90 days, with daily doses above 2 grams, and in patients older than 60 years. These are relevant limitations that matter for long-term use.
The most common side effect of berberine is gastrointestinal discomfort: diarrhoea, constipation, and abdominal pain. In the pilot RCT, more than a third of participants experienced this temporarily3. Liver or kidney damage was not observed in that study, and berberine is generally considered safe on the basis of the short-term trials. However, data on long-term use are scarce, and long-term safety has therefore not been well established. Anyone considering berberine as a supplement to or substitute for diabetes medication would be wise to discuss this with a doctor.
Based on 1 umbrella review (11 meta-analyses, PMID 36999891), 2 meta-analyses (PMID 30393248 and 34956436), 1 small pilot RCT (n=36, PMID 18442638), 1 phase-2 RCT (n=113, PMID 40029660), and mechanistic studies (PMID 32353823, 35209140, 37921026). Total RCT population approximately 2,400+ patients, but quality and geographical variation are limiting factors.