Companion to the cornerstone article. Ten plants, the compound actually responsible, the mechanism, an evidence grade, and the specific caveat that goes with each. Printable, and intended for clinic use.
A — Human RCT evidence, replicated. Would use clinically and state confidently.
B — Human trials exist but are small, short, heterogeneous or industry-funded. Would use with the caveat stated.
C — Mechanism established, human outcome data thin or absent. Worth knowing, not worth claiming.
D — The popular claim does not survive the primary literature. Would actively correct.
Compound: ellagitannins → urolithins A and B (bacterial metabolites, not present in the fruit)
Mechanism: receptor modulation — binds ERα and ERβ, mixed agonist/antagonist
Systems: cardiovascular, prostate, gut, bone
Grade: C for the receptor claim; B for cardiovascular effects via a separate mechanism
Caveat: SERM-like behaviour shown in MCF-7 cells at micromolar concentrations only (Larrosa 2006). Urolithin production is microbiome-dependent — a substantial proportion of people are non-producers, so the same intake yields different metabolites in different people.
Compound: curcuminoids, principally curcumin
Mechanism: NF-κB inhibition upstream of the inflammatory transcription programme; also Nrf2, MMPs, eicosanoids
Systems: joints, gut, skin, vascular
Grade: A for knee OA symptom relief
Caveat: trials used standardised extracts at roughly 500–1,500 mg curcuminoids daily, often with bioavailability enhancement. Culinary turmeric is around 3% curcuminoids and poorly absorbed — not the same intervention. Heterogeneity across trials is high (I² >85%), all trials short, benefit falls as BMI rises, and no structural or imaging change has been shown.
Compound: allicin (unstable), plus S-allylcysteine in aged preparations
Mechanism: vascular — nitric oxide and hydrogen sulphide signalling; antimicrobial — non-specific thiol-group interference
Systems: cardiovascular, immune, metabolic, microbiome
Grade: A cardiovascular; C antimicrobial
Caveat: hypertensive subgroup shows around 8.7/6.1 mmHg reduction (Ried 2016), but heterogeneity is high and effect sizes shrink in higher-quality trials with blinded measurement. Antimicrobial activity is in vitro; allicin degrades within hours and is poorly represented in supplements. Interacts with anticoagulants.
Compound: artemisinin (A. annua); thujone and others (A. absinthium)
Mechanism: endoperoxide bridge generates reactive species on contact with haem iron
Systems: antiparasitic
Grade: A as a pharmaceutical antimalarial; D as a herbal preparation
Caveat: WHO advises against non-pharmaceutical A. annua preparations for malaria — sub-therapeutic dosing drives resistance. Antimalarial efficacy does not generalise to intestinal parasites or dysbiosis. Where a stool test identifies an actual parasite, treat it specifically and retest. Thujone toxicity constrains A. absinthium dosing.
Compound: polyphenols, proanthocyanidins
Mechanism: anti-adhesive — interferes with bacterial surface attachment
Systems: oropharyngeal, gut
Grade: C
Caveat: the weakest entry here. Anti-adhesion in vitro prevents biofilm formation in a dish; that is not the same as disrupting established biofilm in a human gut, and there is no routine way to measure whether it has happened. Safe and pleasant as a tea — that is the level of claim supported.
Compound: not definitively identified
Mechanism: proposed metal binding
Systems: hepatic, renal, neural
Grade: C in rodents; no human data in either direction
Caveat: all supporting work is animal (Aga 2001; Téllez-López 2017; Mustafa 2021) and all of it uses prevention models — co-administration during exposure, not mobilisation of stored burden. The counter-claim that cilantro redistributes metals without a binder is equally unevidenced. Established heavy metal toxicity requires medical chelation with monitoring, not food.
Compound: rosmarinic acid
Mechanism: GABA transaminase inhibition — slows GABA degradation
Systems: nervous, cardiovascular
Grade: B
Caveat: trials are small, short and often product-linked, though consistently in the same direction for anxiety and sleep quality. Cortisol outcomes are measured inconsistently — the claim that lemon balm regulates the HPA axis goes further than the trials support. Effect on subjective anxiety and autonomic arousal is the defensible version.
Compound: GOPO (a galactolipid) — not primarily vitamin C
Mechanism: anti-inflammatory, independent of ascorbate; separately, vitamin C as cofactor for prolyl/lysyl hydroxylase in collagen synthesis
Systems: joints, connective tissue
Grade: B
Caveat: Christensen 2008 pooled three RCTs (n=287), effect size 0.37, NNT 6 — but all three were manufacturer-supported and the authors called for independent replication that has not arrived. Vitamin C content is heat- and oxidation-labile and varies hugely with processing; dried powders may retain a fraction of fresh values.
Compound: polymethoxyflavones — nobiletin, tangeretin
Mechanism: hepatic lipid handling, ApoB secretion, insulin sensitivity; nobiletin acts on RORα/β circadian nuclear receptors
Systems: metabolic, hepatic, circadian
Grade: C
Caveat: results in rodent metabolic syndrome models are striking; human trials are sparse, small and short. Methoxylation makes these unusually bioavailable for flavonoids, which is what makes the mechanism worth watching. Included as promising, not as clinically actionable. Use the zest — the compounds are in the peel, not the flesh.
Compound: astragaloside IV; cycloastragenol as purified derivative (TA-65)
Mechanism claimed: telomerase activation
Systems: claimed as systemic / longevity
Grade: D for telomerase; C as immune-supportive adaptogen
Caveat: the compound studied is a purified derivative, not the root. Telomerase activation shown in cell culture only. Zhu 2010 found only 8.2% of cycloastragenol surviving thirty minutes in human liver microsomes, concluding oral bioavailability would be limited. The single human RCT (Salvador 2016) found significance at the low dose but not the high dose — a non-monotonic response inconsistent with a real drug effect — with all authors commercially connected to the product. An alternative explanation exists: immune cell redistribution toward naive T cells raises measured average leukocyte telomere length without any telomere lengthening.
Where claims fail, they usually fail by escalation rather than fabrication: a cell-culture mechanism becomes a property of the compound, then of the plant, then a clinical indication — and nobody records which step it stopped being true at.
In most cases a smaller, better-supported claim sits directly alongside the inflated one. Using the smaller claim costs almost nothing and keeps the credibility available for the things that do hold up.
Educational content, not medical advice. Several of these interact with prescribed medication — particularly garlic and turmeric with anticoagulants. Check before combining.
Which of these is relevant depends on what is actually going on.
Read the full article → Book a discovery call