In plain words

In three of four randomized trials, fasting blood sugar ended lower with berberine than without it. The fourth found no clear difference. Two studies with no comparison group also saw blood sugar fall.

The short answer

Three of four randomized trials found lower fasting glucose with berberine: 116 people with diabetes and high lipids on 1.0 g a day for 3 months (2008), 59 women with polycystic ovary syndrome over 3 months (2011; the ledger has no dose for it) and 409 people on 0.6 g twice a day for 12 weeks (2020). The fourth, 155 people on 0.5 g three times a day for 16 weeks, found no significant difference from lifestyle change alone (2015). Two before-after studies also saw fasting glucose fall. Certainty is low, because the four trials differ in people and comparison.

Where the evidence sits

Fasting glucose fell

■ Randomized trial, Zhang et al., 2008 · 116 people · 3 months
■ Randomized trial, Wei et al., 2011 · 59 people · 3 months
■ Randomized trial, Zhang et al., 2020 · 409 people · 12 weeks
■ Before-after study, Yin, 2008 · 36 people · 3 months
■ Before-after study, Zhang et al., 2009 · ledger has no n
◪ Pooled analysis, Liang et al., 2018 · may overlap studies counted here · faded after 90 days

No clear difference

■ Randomized trial, Yan et al., 2015 · 155 people · 16 weeks

■ measured by the source's own study · ◪ pooled analysis, may overlap studies counted here · □ reported, reviewed or argued, no new data

Evidence weight7878 / 100?

Who says what

SourceFindingDesignPeopleDose and lengthFunding and independence
Zhang Y, 2008LowerRandomized trial1161.0 g a day, 3 monthsNot in ledger
Wei, 2011LowerRandomized trial59Not in ledger; 3 monthsNot in ledger
Yan, 2015No clear differenceRandomized trial1550.5 g three times a day, 16 weeksNot in ledger
Zhang Y, 2020LowerRandomized trial4090.6 g twice a day, 12 weeksNot in ledger
Yin, 2008LowerBefore-after study36 and 48500 mg three times a day, 3 monthsNot in ledger
Zhang H, 2009LowerBefore-after studyNot in ledgerNot in ledgerNot in ledger
Liang, 2018Associated with lower glucoseMeta-analysis of 28 trials2,313Not in ledgerNot in ledger
Zhang Y, 2008

Lower

Design
Randomized trial
People
116
Dose and length
1.0 g a day, 3 months
Funding
Not in ledger
Wei, 2011

Lower

Design
Randomized trial
People
59
Dose and length
Not in ledger; 3 months
Funding
Not in ledger
Yan, 2015

No clear difference

Design
Randomized trial
People
155
Dose and length
0.5 g three times a day, 16 weeks
Funding
Not in ledger
Zhang Y, 2020

Lower

Design
Randomized trial
People
409
Dose and length
0.6 g twice a day, 12 weeks
Funding
Not in ledger
Yin, 2008

Lower

Design
Before-after study
People
36 and 48
Dose and length
500 mg three times a day, 3 months
Funding
Not in ledger
Zhang H, 2009

Lower

Design
Before-after study
People
Not in ledger
Dose and length
Not in ledger
Funding
Not in ledger
Liang, 2018

Associated with lower glucose

Design
Meta-analysis of 28 trials
People
2,313
Dose and length
Not in ledger
Funding
Not in ledger

What did the placebo-controlled trials find?

Zhang and colleagues (2008) gave 116 people with type 2 diabetes and high blood lipids 1.0 g of berberine a day or placebo for 3 months:

In the berberine group, fasting and postload plasma glucose decreased from 7.0 +/- 0.8 to 5.6 +/- 0.9

Zhang and colleagues (2020) gave 409 people 0.6 g twice a day before meals for 12 weeks; the arms that included berberine had lower fasting glucose than placebo. Wei and colleagues (2011) saw fasting glucose fall more with berberine than placebo in 59 women with polycystic ovary syndrome over 3 months, both groups also taking cyproterone acetate; the ledger has no dose for it.

Did any trial find no difference?

Yan and colleagues (2015) compared 0.5 g three times a day plus a lifestyle intervention with the lifestyle intervention alone; across 155 people over 16 weeks, the change in fasting glucose did not differ significantly (p = 0.096). It is the only trial here without a placebo group.

What did studies without a comparison group find?

Yin and colleagues (2008) measured 36 people with newly diagnosed type 2 diabetes before and after 3 months of 500 mg three times a day:

fasting blood glucose (FBG; from 10.6 ± 0.9 mmol/L to 6.9 ± 0.5 mmol/L, P<0.01)

In 48 people who added the same dose to existing treatment for 3 months, fasting glucose was lower from the first week. Zhang and colleagues (2009) also report lower fasting glucose in type 2 diabetes; the ledger has no n, dose or duration for that study.

Does the effect last past three months?

Liang and colleagues (2018) pooled 28 randomized trials in 2,313 people with type 2 diabetes, a pooled estimate that may include the trials above; berberine was associated with a 0.54 mmol/L larger fall in fasting glucose than control groups, and the ledger has no dose or duration for this estimate. Their subgroup analyses found:

effects of Berberine on blood glucose became unremarkable as the treatment lasted more than 90 days

Why the sources disagree

Only Yan and colleagues found no clear difference. Their comparison group had a lifestyle intervention instead of placebo, and their trial in 155 people lasted 16 weeks; the other three, in 59 to 409 people, lasted 12 weeks to 3 months. Liang's subgroups, from a pooled estimate in 2,313 people that may include these trials, saw the effect fade after 90 days. The people differ too: type 2 diabetes in one trial, polycystic ovary syndrome in another, and the ledger has no condition for the other two. The ledger has no funding details for any study.

What would settle it

An independent placebo-controlled trial in adults with type 2 diabetes, giving 0.5 g of berberine three times a day for 12 months, measuring fasting glucose at 3, 6 and 12 months, with a pre-registered analysis. A difference from placebo that held at 12 months would test the fading seen in Liang's pooled subgroups, which may include the trials above.

How this page was made

The 7 sources are six studies and one pooled analysis that may include them, found by a PubMed literature search; papers on cells, animals or unrelated diseases were left out. Every claim was copied word for word from its source into the Longevity Circle ledger and checked against the original text, and a person approved each one. Certainty follows the GRADE levels used in medical reviews. Drafting was assisted by AI.

Changelog: 30 Sep 2026, first version, 7 sources.