In plain words

Two of the four randomized trials and a before-after study found that berberine lowered insulin resistance. The other two of the four trials found no clear difference from their comparison groups.

The short answer

Two of four randomized trials found berberine improves insulin sensitivity: in 59 women with PCOS against placebo for 3 months (2011; the ledger has no dose), and in 184 people with fatty liver on 0.5 g three times a day for 16 weeks, against lifestyle advice alone (2015). Two found no clear difference: 116 people on 1.0 g a day for 3 months (2008) and 409 people on 0.6 g twice a day for 12 weeks (2020). A before-after study of 48 people on 500 mg three times a day for 3 months found lower insulin resistance. Certainty is low: the trials are few, split evenly and studied different groups.

Where the evidence sits

Insulin sensitivity improved

■ Randomized trial, Wei et al., 2011 · 59 people · 3 months
■ Randomized trial, Yan et al., 2015 · 184 people · 16 weeks
■ Before-after study, Yin, 2008 · 48 people · 3 months
◪ Pooled analysis, Dong, 2012 · may overlap studies counted here

No clear difference

■ Randomized trial, Zhang et al., 2020 · 409 people · 12 weeks
■ Randomized trial, Zhang et al., 2008 · 116 people · 3 months · rose from own baseline

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

Evidence weight7575 / 100?

Who says what

SourceFindingDesignPeopleDose and lengthFunding and independence
Zhang, 2008Not different from placeboRandomized, placebo-controlled1161.0 g a day, 3 monthsNot in ledger
Wei, 2011Lower HOMA-IR than placeboRandomized, placebo-controlled59 women, PCOSNot in ledger; 3 monthsNot in ledger
Yan, 2015Better HOMA-IR than lifestyle aloneRandomized, vs lifestyle184, NAFLD0.5 g thrice daily, 16 weeksNot in ledger
Zhang, 2020Lower HOMA-IR only with probioticsRandomized trial4090.6 g twice daily, 12 weeksNot in ledger
Yin, 2008Lower HOMA-IRBefore-after, no control48500 mg thrice daily, 3 monthsNot in ledger
Dong, 2012Insulin modestly lowerMeta-analysis271Not in ledger; 12 weeksNot in ledger
Zhang, 2008

Not different from placebo

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

Lower HOMA-IR than placebo

Design
Randomized, placebo-controlled
People
59 women, PCOS
Dose and length
Not in ledger; 3 months
Funding
Not in ledger
Yan, 2015

Better HOMA-IR than lifestyle alone

Design
Randomized, vs lifestyle
People
184, NAFLD
Dose and length
0.5 g thrice daily, 16 weeks
Funding
Not in ledger
Zhang, 2020

Lower HOMA-IR only with probiotics

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

Lower HOMA-IR

Design
Before-after, no control
People
48
Dose and length
500 mg thrice daily, 3 months
Funding
Not in ledger
Dong, 2012

Insulin modestly lower

Design
Meta-analysis
People
271
Dose and length
Not in ledger; 12 weeks
Funding
Not in ledger

Did berberine beat placebo?

Zhang and colleagues (2008) gave 116 people 1.0 g of berberine a day or a placebo for 3 months; by clamp test, glucose disposal did not differ significantly from placebo:

although no significant change was found between berberine and placebo groups (P = 0.063)

Against the group's own starting values, it rose:

The glucose disposal rate was increased after berberine treatment (P = 0.037)

Wei and colleagues (2011) found lower fasting insulin and HOMA-IR than placebo in 59 women with polycystic ovary syndrome over 3 months; the ledger has no dose for this trial.

HOMA-IR: a score from fasting glucose and insulin; lower means less insulin resistance.

What happened alongside other treatments?

Yan and colleagues (2015) gave 184 people with fatty liver disease lifestyle advice, with or without berberine at 0.5 g three times a day, for 16 weeks:

paralleled with better improvement in body weight, HOMA-IR, and serum lipid profiles (all p<0.05)

Zhang and colleagues (2020) gave 409 people berberine (BBR) at 0.6 g twice a day, alone or with probiotics (Prob), for 12 weeks:

except for homoeostasis model assessment index for insulin resistance (HOMA-IR), which was significantly lowered by Prob + BBR but not by BBR

Dong and colleagues' 2012 meta-analysis, a pooled estimate that may include the 2008 and 2011 studies above, found fasting insulin modestly lower with berberine plus lifestyle change in 271 people over 12 weeks; the ledger has no dose for it.

What did the before-after study find?

Yin and colleagues (2008) measured 48 people before and after 3 months of berberine at 500 mg three times a day, added to existing treatment, with no comparison group:

Fasting plasma insulin and HOMA-IR were reduced by 28.1% and 44.7% (P<0.001), respectively.

Why the sources disagree

The comparisons differ: placebo in the 2008 and 2011 trials, lifestyle advice alone in 2015, and each person's starting values in Yin's study. In the 2008 trial, glucose disposal rose from baseline yet did not differ significantly from placebo; both can be true at once. The people differ (women with PCOS, adults with fatty liver), as do the measures (a clamp, HOMA-IR, fasting insulin) and the doses in the ledger, from 1.0 g a day in 2008 to 0.5 g three times a day in 2015. The ledger holds no commentary on this question.

What would settle it

A placebo-controlled trial of berberine alone at 1.5 g a day for 6 months in adults with insulin resistance, with clamp-measured glucose disposal as the pre-registered main outcome, reporting the difference from placebo at 3 and 6 months. A matching result from a second, independent group would raise certainty from low to moderate.

How this page was made

The 6 sources are four randomized trials, a before-after study and a 2012 meta-analysis, a pooled estimate that may include the trials above, found through a PubMed literature search on berberine in people. The ledger holds 8 insulin sensitivity claims from 7 sources; a 2018 meta-analysis of blood glucose after meals was left out. Every claim was copied word for word 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: 2 Oct 2026, first version, 6 sources.