⚠️ Auto-extracted by Claude on 2026-05-18 — PDF not yet downloaded locally (archive status: pending). Quantitative claims were extracted from PMC full text (PMC12490254) via WebFetch and are expected to be accurate, but verify exact HRs, CIs, and subgroup p-values against the primary PDF before relying on them for clinical or protocol decisions.
Berberine for preventing colorectal adenoma recurrence and neoplasm occurrence: 6-year follow-up of a randomized clinical trial
Tan YJ, Zou TH, Yu K et al. (Chen YX, Fang JY, corresponding) · Cell Rep Med 6(9):102293 · 2025 · DOI: 10.1016/j.xcrm.2025.102293 · PMID: 40795846 · PMC: PMC12490254
7 clinical centers, China. Published 2025. OA (gold). ClinicalTrials.gov: NCT02226185 (original CBAR trial).
TL;DR
This is the 6-year post-treatment observational follow-up to the chen-2020-berberine-adenoma-rct randomized trial. Patients who received berberine 300 mg twice daily (600 mg/day) for 2 years during the parent trial continued to show significantly lower adenoma recurrence rates versus the original placebo arm across a median 78 months of additional follow-up — despite no ongoing intervention. The primary outcome (any adenoma recurrence) showed an adjusted HR of 0.58 (95% CI: 0.45–0.74, p<0.001). This is the strongest human clinical evidence to date for a durable chemopreventive effect of berberine after treatment cessation. Causal interpretation is tempered by the observational, non-randomized design of the followup phase and the absence of mechanistic data on how the effect persists. no-mechanism
Background and relationship to Chen 2020
The parent Colorectal Adenoma and Berberine (CBAR) trial — chen-2020-berberine-adenoma-rct — enrolled patients with prior colorectal adenoma history across 7 Chinese clinical centers, randomizing 895 participants to berberine 300 mg BID (600 mg/day; n=432) or placebo (n=463) for 2 years. At the 2-year endpoint, berberine significantly reduced adenoma recurrence compared to placebo.
This extension study (CBAR-FE) followed up surviving, contactable participants for an additional ~6 years after berberine discontinuation, with colonoscopic surveillance according to standard Chinese guidelines (3-year intervals). The trial registration number is NCT02226185 (unchanged from original trial).
Key design shift: Once the 2-year intervention ended, no ongoing randomization or treatment was administered. Participants were followed as a cohort. This means the follow-up phase is observational — any confounders accumulated during the 6-year window (diet, lifestyle, competing medications, access to surveillance) are unadjusted by randomization. The authors addressed this with multivariate Cox regression adjusting for age, sex, clinical center, BMI, smoking, family colorectal cancer history, hypertension/diabetes history, and medication use (calcium, statin, aspirin), and a sensitivity analysis assuming all 133 patients who did not undergo colonoscopy were event-free.
Study design
| Feature | Detail |
|---|---|
| Design | Observational post-trial cohort follow-up (CBAR-FE) |
| Parent trial | CBAR RCT, NCT02226185 |
| Original CBAR completers | 895 (432 berberine, 463 placebo) |
| CBAR-FE enrolled | 781 (5 refusals/group; ~7–10 deaths/group; ~47–48 lost to contact/group) |
| Underwent colonoscopy (analyzed) | 648 (314 original berberine arm, 334 original placebo arm) |
| Follow-up period | December 2018 – October 2024 |
| Median follow-up | 78.0 months (IQR: 73.0–85.0) — total ~6.5 years post-randomization |
| Colonoscopy cadence | Per Chinese guidelines, 3-year intervals; range 1–7 colonoscopies/patient |
| Setting | 7 Chinese tertiary clinical centers |
| Population | Adults with prior colorectal adenoma history (elevated baseline risk) |
Baseline characteristics at CBAR-FE enrollment were well-matched between original berberine and placebo arms: median age 66–67, ~67% male, median BMI 24.1, ~14% smokers, ~4% family CRC history. No statistically significant differences in any baseline covariate (all p>0.05). Adenoma recurrence during the original CBAR trial was 41.1% (berberine) vs 48.5% (placebo), p=0.06 — a trend-level difference that did not reach significance during the active-treatment window alone, making the 6-year post-treatment persistence finding the more striking result.
Primary outcome: adenoma recurrence
Any adenoma recurrence (primary endpoint):
| Berberine arm (n=314) | Placebo arm (n=334) | Statistic | |
|---|---|---|---|
| Cumulative recurrence rate | 34.7% (109/314) | 52.1% (174/334) | — |
| Unadjusted HR | — | reference | 0.61 (95% CI: 0.45–0.78, log-rank p<0.001) |
| Adjusted HR | — | reference | 0.58 (95% CI: 0.45–0.74, p<0.001) |
| Sensitivity analysis HR | — | reference | 0.62 (95% CI: 0.49–0.79, p<0.001) |
At 6.9 years follow-up, 43.0% of the berberine arm remained event-free vs 18.5% of the placebo arm by Kaplan-Meier estimate. The divergence between arms persisted throughout the follow-up period rather than converging toward zero over time, which is the key observation driving the durability interpretation.
Year-by-year odds ratios (multivariate logistic regression):
| Follow-up period | OR (95% CI) | p-value |
|---|---|---|
| Within 1 year | Not significant (point estimate not extracted) | NS |
| 1–3 years | Not significant | NS |
| 3–5 years | 0.63 (95% CI: 0.41–0.97) | 0.04 |
| >5 years | 0.50 (95% CI: 0.35–0.70) | <0.001 |
The effect size strengthened over time — ORs became smaller (more protective) at years 3–5 and >5 compared to years 1–3. If anything, the inter-arm separation increased over the observational window. This pattern is unusual for a simple “residual drug” hypothesis and is more consistent with a durable reprogramming event (microbiome, immune, or epigenetic), though the design cannot establish this mechanistically. no-mechanism
Multiple event analysis (Andersen-Gill model):
- Adjusted HR for multi-recurrence of adenoma: 0.67 (95% CI: 0.54–0.83, p<0.001)
Secondary outcomes
| Outcome | Berberine | Placebo | Adjusted HR (95% CI) | p |
|---|---|---|---|---|
| Any colorectal neoplasm | 63.4% (199/314) | 71.0% (237/334) | 0.75 (0.62–0.91) | 0.004 |
| Non-advanced adenoma | — | — | 0.56 (0.43–0.71) | <0.001 |
| Advanced adenoma | 4.1% (13/314) | 5.1% (17/334) | 0.99 (0.96–1.02) | 0.57 (NS) |
| Colorectal cancer | 0.3% (1/314) | 0.3% (1/334) | — | 0.97 (NS) |
| Serrated lesions (overall) | — | — | 0.72 (0.52–0.99) | 0.05 |
| Multi-occurrence neoplasm | — | — | 0.88 (0.76–1.02) | 0.10 (NS) |
At 5.2 years, 35.5% of the berberine arm remained neoplasm-free vs 24.0% of placebo arm by Kaplan-Meier.
The non-significant advanced adenoma result (HR 0.99, p=0.57) is an important limitation: advanced adenomas are the highest-risk precursor to colorectal cancer. The study was likely underpowered for this outcome (13 vs 17 events). The authors note this in the limitations section. The failure to show a significant benefit on advanced adenoma weakens the case for clinical impact on cancer prevention, even if non-advanced adenoma recurrence is reliably reduced.
Subgroup analyses across all prespecified baseline characteristics (age, sex, BMI, smoking, family history, etc.) showed no significant interaction terms (all interaction p>0.05), suggesting the primary result is not driven by a specific subpopulation.
Proposed mechanistic interpretations
The paper does not present new mechanistic data — all hypotheses are speculative and drawn from prior literature. Three candidate mechanisms are discussed:
-
Durable gut microbiome reprogramming. Berberine is known to modulate gut microbiota composition, including suppression of taxa associated with colorectal tumorigenesis (cited: Fusobacterium nucleatum suppression; Veillonella parvula-mediated B-cell regulation). The authors speculate that a 2-year exposure may induce stable microbiome states that persist after cessation. The study did not measure microbiota at any timepoint. no-mechanism
-
Epigenetic or immune reprogramming of colonic mucosa. A 2-year berberine exposure could in principle alter CpG methylation patterns, mucosal immune tone, or stem cell compartment behavior in the colonic epithelium. No epigenetic or immunologic data are presented. no-mechanism
-
Prevention of first-wave high-risk lesions. Preventing large or advanced adenomas during the active-treatment period may reduce subsequent risk through a “path dependency” mechanism — fewer high-risk precursors in years 1–2 leads to fewer recurrences in years 3–8 even without ongoing treatment. The authors suggest “intermittent dosing might maximize chemopreventive effects while minimizing side effects.” This interpretation is consistent with the data but speculative.
The authors explicitly acknowledge: “The treatment efficacy of berberine might be mediated by long-term gut microbiota changes, but we did not further investigate the potential mechanisms.”
Limitations
Limitations acknowledged by the paper, plus additional structural concerns:
-
Observational followup — not randomized. The core limitation. After the 2-year CBAR intervention ended, participants were not re-randomized. The observed inter-arm difference could reflect unmeasured confounders accumulated over 6 years (dietary changes, self-initiation of berberine use, differences in surveillance behavior). The authors adjusted for known confounders via Cox regression, but residual confounding from unmeasured variables (dietary fiber, physical activity, self-medication) is not addressable with available data.
-
Incomplete colonoscopy coverage. 133 of 781 enrolled participants (17%) did not undergo any follow-up colonoscopy and were excluded from the primary analysis. The sensitivity analysis (assuming these 133 had no events) preserved the result (adjusted HR 0.62), but if the non-colonoscopy group had differential event rates by arm, bias is possible.
-
Adenoma morphology data incomplete. Only 336 of 648 participants with colonoscopy (51.9%) had accurate recordings of adenoma number, size, and location. Analyses using these characteristics were restricted to this subset.
-
Advanced adenoma and CRC underpowered. Only 13 (berberine) vs 17 (placebo) advanced adenoma events; 1 CRC per arm. These are the clinically most relevant outcomes; the study was not powered to demonstrate effects on them.
-
Entirely Chinese cohort. Results may not generalize to Western populations with different baseline microbiome composition, dietary patterns, and colorectal adenoma epidemiology.
-
Selected high-risk population. All participants had prior adenoma history. Effects in average-risk populations are unknown.
-
No mechanistic data collected. Gut microbiota, epigenetics, and immune profiling were not measured at any timepoint in CBAR-FE.
-
Six-year follow-up shorter than adenoma-to-carcinoma progression. The typical adenoma-to-carcinoma sequence takes ~10 years; this follow-up does not yet capture the endpoint most relevant to mortality benefit.
Assessment of durability claim
The observation that protective effects strengthened over time (OR 0.50 at >5 years vs OR 0.63 at 3–5 years) is striking and is not a trend confined to a small subgroup — the primary analysis involved 648 participants with a median of 3 colonoscopies each. The CIs are relatively tight for the primary outcome (adjusted HR 0.58, 95% CI: 0.45–0.74). This is not a marginal result.
However, several features temper confidence in the durability interpretation:
- The non-significant advanced adenoma result (HR 0.99) suggests that whatever is being prevented consists predominantly of lower-risk, non-advanced lesions. This could reflect detection bias (more colonoscopies in the berberine arm finding more low-grade lesions) or a genuine difference in the lesion spectrum. The paper does not fully resolve this.
- The strengthening effect over time is equally consistent with increasing non-randomized confounding accumulation as with genuine biologic durability. Diet and lifestyle in the berberine arm may have shifted post-trial in ways correlated with the original assignment.
- No microbiome data exists to confirm the proposed durable-reprogramming mechanism.
Bottom line: The durability claim is supported by the primary outcome data and is not obviously fragile statistically. It is the most provocative finding in the paper. The causal interpretation requires an RCT with true post-cessation randomization or a mechanistic follow-up study before it should anchor clinical recommendations.
Cross-references
| Page | Relationship |
|---|---|
| berberine | Subject compound; this is the strongest human clinical trial data for berberine as a chemopreventive agent |
| chen-2020-berberine-adenoma-rct | Parent RCT (CBAR); this page is the 6-year follow-up extension |
| cancer | Phenotype: colorectal adenoma recurrence and colorectal neoplasm incidence |
| gut-microbiome-aging-shifts | Proposed mechanism: durable gut microbiome reprogramming; relevant to whether berberine’s effects persist via the microbiome |
| sasp | Indirect relevance: adenoma microenvironment and inflammatory signaling; berberine has documented anti-inflammatory effects in the gut |