⚠️ Auto-extracted by Claude on 2026-05-18 — full PDF not available (closed access; no PMC version; → not_oa). Abstract and PubMed record confirmed. Specific pooled effect sizes for most outcomes and I² heterogeneity values drawn from a partial secondary record of the PubMed landing page; they are potentially incomplete and require primary-PDF confirmation. The IMT subgroup section explicitly flags a methodological concern that also cannot be fully resolved without full text. no-fulltext-access
Efficacy and safety of berberine for several cardiovascular diseases: a systematic review and meta-analysis of randomized controlled trials
Yang L, Zhu W, Zhang X, Zhou X, Wu W, Shen T · Phytomedicine 2023;112:154716 · DOI: 10.1016/j.phymed.2023.154716 · PMID: 36805484
Citation impact (as of 2026-05-18): 25–29 citations (OpenAlex/Semantic Scholar); FWCI = 7.76; citation_percentile = 100
TL;DR
A 2023 systematic review and meta-analysis pooling 44 RCTs (n=4,606 patients) found that berberine alone did not significantly outperform routine therapy or statins for standard lipid endpoints (TC, TG, LDL-C, HDL-C, Crouse score), but did significantly reduce NIHSS score (stroke severity), hsCRP, IL-6, TNF-α, and IMT versus routine therapy alone. Berberine plus statins significantly outperformed statins alone on a broader set of endpoints including TC, TG, LDL-C, NIHSS, hsCRP, TNF-α, IMT, Crouse score, and number of unstable plaques, but not HDL-C or IL-6. No significant safety signal was identified.
Critical caveat: The IMT (plaque) endpoint — the finding with the most clinical significance — was derived from a 13-trial subgroup. A post-publication review of these 13 primary studies found that most could not be located in PubMed, ClinicalTrials.gov, or major academic search databases, raising concerns about source traceability. See § Source traceability concerns below. This concern is methodological, not a fraud allegation; it reflects a known problem in the Chinese-language TCM literature where small single-center trials are published in lower-tier regional journals not indexed in international databases.
Study design
Search strategy and inclusion criteria
- Databases searched: Ten electronic databases from inception to December 23, 2022. Databases listed in abstract include major Chinese-language databases (CNKI, Wanfang, VIP, CBM) as well as PubMed, Embase, Cochrane Library, Web of Science, ClinicalTrials.gov, and a WHO registry. The inclusion of Chinese-language databases is the primary reason the trial pool (44 RCTs) is substantially larger than what PubMed alone would return. no-fulltext-access — exact databases, search terms, and PRISMA flow diagram require primary PDF.
- Comparison groups: Berberine alone vs. routine therapy; berberine alone vs. statins; berberine + statins vs. statins alone. These three comparison arms were analyzed separately.
- Outcome domains: Lipid profile (TC, TG, LDL-C, HDL-C), cardiovascular plaque metrics (carotid IMT, Crouse score, number of unstable plaques), inflammatory markers (hsCRP, IL-6, TNF-α), and NIHSS score (stroke severity), plus adverse events.
- Quality assessment: Described as per Cochrane Handbook methodology; specific risk-of-bias scores for individual trials not accessible without full text. no-fulltext-access
Population
Patients with cardiovascular diseases broadly defined — the study title specifies “several CVDs,” which from the outcomes appears to include hyperlipidemia, atherosclerosis, and acute ischemic stroke (the NIHSS inclusion implies stroke patients). Exact disease-specific subgroup breakdown not accessible from abstract.
| Parameter | Value |
|---|---|
| Total RCTs included | 44 |
| Total pooled patients | 4,606 |
| IMT-subgroup RCTs | 13 (stated by reviewer; not independently confirmed from abstract) |
| Study period | Inception to December 23, 2022 |
Key results by outcome
Berberine alone vs. routine therapy / statins
The meta-analysis found no significant difference for standard lipid endpoints in this comparison arm. Reported effect sizes from a secondary summary of the PubMed landing page (verify against primary PDF):
| Outcome | SMD | 95% CI | Significant? |
|---|---|---|---|
| Total cholesterol (TC) | 0.43 | −0.39 to 1.24 | No |
| Triglycerides (TG) | −0.14 | −0.49 to 0.21 | No |
| LDL-C | 0.69 | −0.23 to 1.60 | No |
| HDL-C | 0.55 | −0.48 to 1.57 | No |
| NIHSS score | — | — | Yes (favors berberine) |
| hsCRP | — | — | Yes (favors berberine) |
| IL-6 | — | — | Yes (favors berberine) |
| TNF-α | — | — | Yes (favors berberine) |
| IMT | — | — | Yes (favors berberine) |
no-fulltext-access — specific SMD/WMD with 95% CI for inflammatory and IMT outcomes in the berberine-alone arm require primary PDF.
Berberine plus statins vs. statins alone
This is the clinically most salient comparison given the Kong 2004 mechanistic hypothesis (berberine stabilizes LDLR mRNA via ERK independently of SREBP-2/HMGCR, so the two agents should be additive) 1.
The meta-analysis found significant improvement for berberine + statins vs. statins alone on: TC, TG, LDL-C, NIHSS score, hsCRP, TNF-α, IMT, Crouse score, and number of unstable plaques. No significant difference for HDL-C and IL-6.
no-fulltext-access — exact pooled effect sizes for this arm require primary PDF.
Mechanistic consistency note: The finding that berberine + statins outperforms statins alone on lipid endpoints is consistent with Kong 2004’s mechanistic prediction — LDLR mRNA stabilization is additive to SREBP-2-mediated LDLR transcription induction. However, whether the magnitude of benefit is explained by the Kong mechanism alone vs. berberine’s AMPK-activation / gut-microbiome effects is not resolved here. no-mechanism
Heterogeneity
Substantial to extreme heterogeneity (I²) was observed across outcomes, particularly for lipid endpoints in the berberine-alone arm. Reported from secondary records:
| Outcome | Approximate I² (berberine alone arm) |
|---|---|
| TC | ~high (exact value: no-fulltext-access) |
| TG | 76% |
| LDL-C | 96% |
| HDL-C | 96% |
I² of 76–96% indicates that the pooled estimates are dominated by heterogeneity between trials rather than a consistent underlying effect. At these levels, the pooled point estimate is of limited clinical interpretive value without subgroup analysis explaining the source of heterogeneity. no-fulltext-access — subgroup analyses (by disease, dose, duration, quality) require primary PDF.
Adverse events
No significant differences between groups in incidence of adverse reactions were reported. Specific adverse event types and rates not available from abstract. Safety profile appeared acceptable across the 44 RCTs. no-fulltext-access
Source traceability concerns (IMT subgroup)
This section documents a methodological concern identified by a post-publication reviewer (via YouTube, referenced in this wiki via user communication, 2026-05-18). The concern is presented here for transparency. It does not constitute a fraud allegation; it reflects a recognized structural problem in the Chinese-language TCM trial literature.
The reviewer’s claim
A reviewer examined the 44 RCTs pooled in this meta-analysis, focusing on the 13-trial IMT subgroup. The reviewer reported:
- Approximately 30% of the 13 IMT-subgroup trials could be traced to verifiable publication records in standard international databases.
- Approximately 60% had “no record of existence” in Google Scholar, PubMed, ClinicalTrials.gov, or major academic AI search tools.
- One trial from the subgroup was located in a Chinese-language source; hand-translation reportedly showed it did not actually measure IMT, raising concerns about outcome misclassification in the original screening.
The reviewer did not allege fabrication; the concern is about indexing opacity making independent verification of the primary inputs impossible.
What PubMed indexing shows
A targeted search of PubMed (performed 2026-05-18) for berberine + IMT/carotid + RCT design returns at most 1 indexed RCT (PMID 28929697, Li Y et al., Zhongguo Zhong Yao Za Zhi 2016, n=120, Chinese-language journal). An additional search specifically combining berberine + carotid + IMT + atherosclerosis + RCT returns 0 results.
If Yang 2023 pooled 13 IMT-measuring RCTs and PubMed indexes at most 1, the remaining 12 are either:
- Published in Chinese-language regional journals not indexed in PubMed (the most likely explanation given the database scope of the search strategy), or
- Not identifiable by standard academic-database searching.
Explanation (1) is methodologically problematic but not fraudulent — it is a structural feature of how the Chinese TCM trial literature is published and aggregated. However, it creates a serious verification gap: the primary inputs to the pooled IMT estimate are effectively unreviewed by the international biomedical community.
Why this matters for the pooled effect estimate
The headline result — that berberine (alone or + statins) reduces IMT — rests entirely on this 13-trial subgroup. IMT reduction is the most clinically consequential endpoint in this paper: LDL-C and TG improvements are modest and not robust (I² 76–96% in the berberine-alone arm), but IMT reduction would directly imply plaque regression — an antiatherosclerotic structural effect. If the IMT subgroup trials are unverifiable or if any subset miscoded the outcome (as the reviewer found in at least one case), the pooled IMT effect may be unreliable.
| Risk dimension | Assessment |
|---|---|
| Outcome misclassification in individual trials | Possible: 1 case identified by reviewer |
| Publication bias from non-indexed journals | Cannot be assessed without access to the underlying trial pool |
| Heterogeneity within IMT subgroup | Not accessible without full text |
| Funnel plot / Egger test | Not confirmed; likely performed per Cochrane Handbook but details inaccessible |
contradictory-evidence — The IMT pooled effect should be treated as low-confidence until the underlying trials can be independently verified. This is not contradictory-evidence in the traditional sense (no RCT contradicts the finding), but reflects a confidence problem in the evidence base itself.
Broader context: Chinese TCM meta-analyses and indexing gaps
The structural issue is well-documented in the meta-research literature: systematic reviews drawing on Chinese-language databases (CNKI, Wanfang, VIP, CBM) consistently include trials that are not indexed in international databases, reducing their audibility. This is not unique to berberine or to this paper. The appropriate response is not to reject the finding but to treat it as hypothesis-generating pending verification of the primary trial pool against CONSORT-compliant reporting standards.
Heterogeneity and publication bias
- Heterogeneity: Substantial to extreme across lipid endpoints (I² 76–96% reported for berberine-alone arm). The high heterogeneity means the pooled effects for TC, LDL-C, HDL-C, and TG in the berberine-alone arm are not reliable summary estimates; the confidence intervals span null. This is partly expected given the diverse patient populations (hyperlipidemia, atherosclerosis, stroke) and dose heterogeneity across the 44 trials.
- Publication bias: Cochrane Handbook methods required funnel plots and/or Egger test. Specific results not accessible from abstract. Given that 10 databases were searched including Chinese-language sources, the authors likely had better publication bias control than a PubMed-only search; however, the non-indexed nature of many included trials means the funnel plot may not capture the true trial universe. no-fulltext-access
Mechanistic interpretation
The finding that berberine + statins outperforms statins alone for lipid and plaque endpoints is mechanistically consistent with the Kong 2004 model 1: berberine stabilizes LDLR mRNA via ERK-dependent post-transcriptional mechanisms, while statins increase LDLR transcription via HMGCR inhibition → cholesterol depletion → SREBP-2 cleavage. These two mechanisms are non-overlapping in principle, predicting additivity. The meta-analysis pooled result is consistent with this prediction but does not constitute mechanistic confirmation — it is observational pooling across heterogeneous populations.
Additionally, berberine’s AMPK-activating effects (independent of the LDLR mechanism) and gut-microbiome modification effects may contribute to the inflammatory marker improvements (hsCRP, IL-6, TNF-α), which are not easily explained by the LDLR pathway alone. The multi-mechanism nature of berberine makes it difficult to attribute specific outcomes in the pooled result to specific pathways. no-mechanism
Extrapolation table
| Dimension | Status | Notes |
|---|---|---|
| Pathway conserved in humans? | yes | Direct human RCT data (though with traceability caveats) |
| Phenotype conserved in humans? | partial | Lipid-lowering confirmed; plaque/IMT endpoint low-confidence (see § Source traceability concerns) |
| Replicated independently? | partial | Some lipid-lowering RCTs are PubMed-indexed and independently verifiable; IMT subgroup is not |
Limitations and gaps
- Full text inaccessible — effect sizes, I² values, subgroup analyses, funnel plots, GRADE assessments, and PRISMA flow diagram cannot be confirmed without primary PDF access. no-fulltext-access
- Source traceability of IMT subgroup — the clinically most impactful outcome rests on 13 trials, at most 1 of which is PubMed-indexed. The remainder cannot be independently audited without institutional access to Chinese-language databases. contradictory-evidence
- Extreme heterogeneity — I² 76–96% for lipid outcomes in the berberine-alone arm renders pooled point estimates unreliable as guides to expected individual-level effects.
- Patient-level disease heterogeneity — pooling atherosclerosis, hyperlipidemia, and acute ischemic stroke patients under “cardiovascular diseases” may obscure disease-specific effects.
- No hard clinical endpoints — IMT is a surrogate; no MACE (myocardial infarction, stroke recurrence, cardiovascular death) endpoints were assessed.
- Risk of bias in primary trials — Chinese-language single-center trials in TCM literature have well-documented high rates of unclear or high risk of bias on allocation concealment and blinding. Specific ROB scores not confirmed. no-fulltext-access
- Dose and duration heterogeneity — berberine dose and treatment duration likely varied substantially across the 44 trials. Dose-response analysis not confirmable from abstract. dose-response-unclear
Cross-references
| Page | Relationship |
|---|---|
| berberine | Primary compound; this study is a major human-evidence aggregator for berberine’s CV effects |
| kong-2004-berberine-ldlr-mechanism | Mechanistic foundation for berberine + statin additivity hypothesis; the pooled result in Yang 2023 is consistent with Kong 2004’s model |
| ldlr | Direct mechanistic target of berberine (LDLR mRNA stabilization) |
| deregulated-nutrient-sensing | AMPK-mediated berberine effects intersect with nutrient-sensing hallmark |
| chronic-inflammation | hsCRP, IL-6, TNF-α improvements implicate this hallmark |
| ampk-activators | Berberine is classified here as a pharmacological AMPK activator |
Footnotes
Footnotes
-
kong-2004-berberine-ldlr-mechanism · n=32 (human arm) · open-label pilot · model: hypercholesterolemic adults + HepG2 cells + hyperlipidemic hamster · doi:10.1038/nm1135 · PMID: 15531889 · established LDLR mRNA stabilization via ERK as berberine’s primary cholesterol-lowering mechanism, distinct from HMGCR/SREBP-2 statin pathway — the mechanistic basis for predicted additivity of berberine + statin combination ↩ ↩2