Every company publishes the studies that support it. This page publishes the ones that do not. These are peer-reviewed findings that contradict, complicate or narrow a claim made elsewhere on this site. We found every one of them ourselves, during our own research, usually while looking for something else — and we record them so they cannot be quietly lost. Each carries its identifier so you can read it without our summary in the way. Where a conflict has been resolved, the reasoning is shown. Where it has not, it says so.
The register
1. Lack of Effect of Oral Sulforaphane Administration on Nrf2 Expression in COPD: A Randomized, Double-Blind, Placebo Controlled Trial
PLOS One
Identifiers: PMID 27832073 · doi:10.1371/journal.pone.0163716
A HUMAN randomised, double-blind, placebo-controlled trial reporting NO EFFECT on Nrf2 expression, against an engine claim of 'NRF2 induction'. The trial measured gene expression in ALVEOLAR MACROPHAGES at 25 uM and 150 uM sulforaphane in COPD patients.
Found while: searching for orphan PMC4789124 (cited on this site as 'Sulforaphane-NRF2: Fahey/Talalay (PMC4789124); Frontiers Nutrition 2021; PMC4736808 - superior bioavailability & NRF2 induction')
Status — Open — not yet settled. Our claim is about NRF2 INDUCTION generally; this trial's endpoint is Nrf2 EXPRESSION IN ALVEOLAR MACROPHAGES IN COPD. Whether those are the same endpoint must be settled by retrieving BOTH sides in full — a null on one tissue in one disease does not automatically refute induction measured elsewhere.
2. Berberine Improves Glucose and Lipid Metabolism in HepG2 Cells Through AMPKalpha1 Activation
Frontiers in Pharmacology 2020
Identifiers: doi:10.3389/fphar.2020.00647
AMPKalpha1 KNOCKOUT in HepG2 cells ABOLISHED berberine's effect entirely — 'AMPKalpha1 is essential for BBR to promote glucose and lipid metabolism'. That is the OPPOSITE of the AMPK-independence claim the engine carries, in the SAME CELL LINE.
Found while: retrieving orphan PMC4114874 (engine: 'Berberine AMPK-independence: PLOS One 2014 (PMC4114874) — works even when AMPK is blocked')
Status — Open — not yet settled. Both are IN VITRO in HepG2. They differ in method: Xu 2014 blocked AMPK pharmacologically/by siRNA/by dominant-negative; the 2020 paper used a full knockout. Retrieve BOTH in full and compare the blockade completeness before either is called settled.
3. Efficacy and safety of berberine on the components of metabolic syndrome: a systematic review and meta-analysis of randomized placebo-controlled trials
Identifiers: PMID 40740996 · PMC12307485
SCOPE CORRECTION, not a refutation. Verbatim: berberine 'significantly reduces triglycerides ... fasting plasma glucose ... and waist circumference ... BUT HAS NO SIGNIFICANT EFFECT ON HIGH-DENSITY LIPOPROTEIN CHOLESTEROL (HDL-C), SYSTOLIC BLOOD PRESSURE (SBP), OR DIASTOLIC BLOOD PRESSURE (DBP).' The engine's summary reads '↓HbA1c/FPG/lipids'. TG, LDL-C and TC are supported; HDL-C is NOT, and 'lipids' unqualified overstates it.
Found while: retrieving the last orphan PMC8107691 (engine: 'Berberine meta-analyses & RCTs: Front Pharmacol 2025; PMC8107691; Atherosclerosis 2015 — AMPK; ↓HbA1c/FPG/lipids')
Status — Open — not yet settled. No grade moved. The claim needs narrowing to the lipid fractions the meta-analyses actually move, not overturning.
4. Acute exposure to resveratrol inhibits AMPK activity in human skeletal muscle cells
Diabetologia
Identifiers: doi:10.1007/s00125-012-2691-1
Directly opposes a live grade: resveratrol holds B+ on the AMPK pathway in Holy Grail, and it is the driver that opened that pathway. Verbatim from the paper: primary myotubes from six healthy men, 4 h at 100 micromol/l resveratrol +/- palmitate -- 'RSV did NOT reverse palmitate-induced impairment of glucose metabolism. Surprisingly, RSV DECREASED glucose uptake and glycogen synthesis in human skeletal muscle cells.'
Found while: Sweeping RESVERATROL for the CERAMIDE BLOCKER build -- i.e. found while working on a DIFFERENT blend. Nobody was looking for it. That is the whole reason this register exists.
Status — Resolved. No grade moved, and the reason is that the endpoints genuinely differ. OURS: Eur J Nutr 2025 (doi:10.1007/s00394-025-03850-3), 2-MONTH randomised trial, HUMAN MUSCLE BIOPSY IN VIVO, chronic oral dosing, AMPK phosphorylation UP alongside ACC phosphorylation UP. THEIRS: 4 HOURS, 100 micromol/l, PRIMARY MYOTUBES IN VITRO, acute. Acute-in-vitro vs chronic-in-vivo is a different exposure, a different system and a different question. AND IT CORROBORATES RATHER THAN REFUTES: the lane's own customer note already reads 'appears to need weeks, not a dose' -- an acute high-concentration exposure doing the opposite is exactly the shape of a duration-conditional effect. SECOND SIGNAL FROM THE SAME SWEEP, ALSO REGISTERED: PMC5372570 / doi:10.3390/ijms18030554, a 4-month randomised placebo-controlled trial in metabolic-syndrome men, reports resveratrol 'INCREASED FRUCTOSAMINE levels after four months of treatment, indicating POORER insulin sensitivity'. That is a HUMAN, CHRONIC, RANDOMISED result pointing the wrong way on GLYCAEMIC CONTROL -- a DIFFERENT endpoint from AMPK phosphorylation, so it does not move the `AMPK` grade either, but it is the stronger of the two and any future resveratrol work must read it.
5. Deletion of serine palmitoyl transferase 2 in hepatocytes impairs ceramide/sphingomyelin balance, prevents obesity and leads to liver damage in mice
BBA - Molecular and Cell Biology of Lipids (peer-reviewed; the bioRxiv preprint is superseded)
Identifiers: PMID 37224999 · doi:10.1016/j.bbalip.2023.159333 (ScienceDirect S1388198123000574) · biorxiv 10.1101/2021.09.16.460588 (preprint, now superseded)
Directly complicates the SPT-to-ceramide edge in the published causal map. Verbatim: 'Surprisingly, in our genetic model, Sptlc2 deficiency in hepatocytes does NOT induce ceramide (Cer) reduction in liver of RC mice and even ELICITS HEPATIC Cer ACCUMULATION in mice fed HFD. Moreover, Sptlc2 deletion modifies hepatic Cer species distribution. Thus, we showed C16:0-Cer, C18:0-Cer and C24:1-Cer ACCUMULATION and C22:0-Cer reduction.' ⇒ DELETING THE RATE-LIMITING ENZYME RAISED THE THREE CVD-ASSOCIATED SPECIES.
Found while: Sweeping SULFORAPHANE for the CERAMIDE BLOCKER roster. Found in the same result set as the paper that supports the driver.
Status — Resolved. IT DOES NOT REFUTE THE SULFORAPHANE DRIVER — different intervention (pharmacological partial inhibition of SPTLC3 vs total genetic deletion of SPTLC2), different isoform, and compensatory salvage-pathway flux is the obvious explanation for why total ablation behaves unlike partial inhibition. BUT IT STRONGLY CORROBORATES THE SAFETY BOUND ALREADY WRITTEN INTO THE `spt|ceramide` EDGE: 'SPT is NOT a viable therapeutic target, owing to safety issues that result from the extreme diminution of all sphingolipids.' Here total deletion did not merely fail to lower ceramide — it RAISED the harmful species AND caused liver damage. This blend is never positioned as shutting down SPT. Selectivity and partiality are the whole point. The preprint has since been superseded by the peer-reviewed version: PMID 37224999. It supplies THE MECHANISM THAT RESOLVES THE PARADOX, VERBATIM: 'Despite lower expression of hepatic Sptlc2, we observed an INCREASED concentration of hepatic Cer, associated with a 10-FOLD INCREASE IN NEUTRAL SPHINGOMYELINASE 2 (nSMase2) EXPRESSION, and a decreased sphingomyelin content in the liver ... Our data suggest a COMPENSATORY MECHANISM to regulate hepatic ceramide content FROM SPHINGOMYELIN HYDROLYSIS, with deleterious impact on liver homeostasis.' ⇒ BLOCKING ONE VALVE OPENS ANOTHER. Shutting down de novo synthesis triggered a ten-fold upregulation of the hydrolysis route. AND THE DOSE DISTINCTION IS NOW MEASURED, NOT HYPOTHESISED: the same literature records that 'HETEROZYGOUS Sptlc2 knockout mice were PROTECTED from high-fat-diet-induced obesity, insulin resistance and atherosclerosis' (PMC12594915). PARTIAL inhibition protects; TOTAL deletion backfires. That is exactly the distinction the blend rests on.
6. Effect of Homocysteine-Lowering Nutrients on Blood Lipids: Results from Four Randomised, Placebo-Controlled Studies in Healthy Humans
PLoS Medicine
Identifiers: PMC1140947
BETAINE MOVES TWO LANES THIS BLEND CARRIES IN THE WRONG DIRECTION, IN HUMANS. Verbatim: 'Betaine supplementation (6 g/d) for 6 wk INCREASED blood LDL cholesterol concentrations by 0.36 mmol/l (95% CI: 0.25-0.46), and triacylglycerol concentrations by 0.14 mmol/l (0.04-0.23) relative to placebo. The ratio of total to HDL cholesterol increased by 0.23 (0.14-0.32).' FOUR randomised placebo-controlled studies, n=151 for the betaine arm. SECOND FINDING, same sweep: 'A randomized placebo-controlled trial in human patients reported that betaine DID NOT IMPROVE HEPATIC STEATOSIS' (cited in PMC9180950).
Found while: Sweeping BETAINE for the CERAMIDE BLOCKER roster. Betaine has NO ceramide measurement; this is what the sweep found instead.
Status — Resolved. MEASURED AND VERIFIED SAFE AS BUILT: Betaine sits ONLY on CERAMIDE/weight (B+). It is NOT on CERAMIDE/lipids and NOT on CERAMIDE/steatosis - the two lanes these findings touch. NO GRADE MOVES and nothing needs removing. This entry is prospective: Betaine is not added to the lipids or steatosis pathways on any blend. A randomised human trial says it RAISES LDL and triglycerides, and a separate randomised trial found no steatosis benefit. Anyone reading only betaine's methyl-donor and NAFLD-mechanism literature would find plenty of reasons to add it to exactly those lanes. The weight placement is a DIFFERENT ENDPOINT and is untouched by these results. AND THE DOSE MATTERS: 6 g/day is high; the LDL effect may not generalise to lower formulation doses - but that is a reason to state the bound, not to ignore it.
How far a search goes before this site will say no evidence exists is described at How We Search.
Why this page exists
A grade assigned by the seller is worth exactly as much as the seller’s willingness to argue against it. No independent body audits the grades on this site. This register is what we do instead: we retrieve the opposing evidence, we record it with its identifier, and we state plainly whether it changed a grade or did not, and why. Three of the conflicts above are resolved. Three are open. A register with no open items would mean we had stopped looking.
No grade on this site has ever been moved on a single finding. Comparing two studies requires that both be retrieved in full and their endpoints compared — an acute four-hour cell experiment and a two-month human trial can disagree without either being wrong. That comparison is the work, and the register is where it is shown.
These statements have not been evaluated by the Food and Drug Administration. This product is not intended to diagnose, treat, cure, or prevent any disease.