Evidence grading
How we rate each outcome
Each herb outcome is graded independently. An herb can have strong evidence for one outcome (immune function) and preliminary evidence for another (anti-cancer). We never aggregate across outcomes — that's how you end up calling something "proven" when one benefit is real and five others are speculative.
Our grading adapts the Cochrane risk-of-bias framework for the specific challenges of herbal medicine research: small sample sizes, poor standardisation of extracts between trials, and the predominance of Chinese-language literature in databases that Western meta-analyses frequently miss.
A
Strong — ≥5 high-quality RCTs, consistent effect, ≥1 meta-analysis
Multiple independent double-blind placebo-controlled trials. Adequate statistical power (n>50 per arm). Low Cochrane RoB2 risk. Effect direction consistent across trials. Meta-analysis confirms pooled effect. Examples: Ashwagandha for cortisol reduction, Ginkgo EGb 761 for claudication.
B
Moderate — 2–4 RCTs or consistent cohort data with plausible mechanism
Some RCT evidence but limited by small sample sizes, industry funding, or effect size inconsistency across trials. Mechanism is characterised and supported by mechanistic studies. Clinical relevance is plausible. Examples: Lion's Mane for MCI, Schisandra for hepatoprotection.
C
Preliminary — ≤1 RCT, or primarily animal/in vitro evidence
Mechanism is biologically plausible. Animal data exists and is internally consistent. Human trial data is absent or limited to a single small pilot. Cannot be recommended with confidence for specific outcomes. We stock these herbs only when they have a strong traditional use record alongside the preliminary data.
D
Insufficient or contradictory — we document and do not stock
Studies are contradictory without explanation, methodological quality is consistently poor across all available trials, or the preclinical safety signal is concerning enough that human trials are not yet appropriate. We document these decisions in full.
Process
Our review protocol
Every herb dossier follows the same structured review process, adapted from systematic review methodology used in clinical pharmacology.
Step 01
Systematic literature search
PubMed, Cochrane Library, Web of Science, and CNKI (Chinese National Knowledge Infrastructure). Date range: all years. Search terms: binomial Latin name + MeSH conditions. No language restriction for RCTs — Chinese-language trials are included and translated.
Step 02
Risk-of-bias assessment
Each RCT assessed using Cochrane RoB 2.0 tool: randomisation, allocation concealment, blinding, outcome reporting, and other biases. Industry-funded trials are flagged. Open-label trials are downgraded by one evidence tier. Animal model data is explicitly labelled as preclinical.
Step 03
Effect size extraction
Primary endpoints extracted with confidence intervals. Cohen's d or standardised mean difference calculated where possible. Clinically meaningful threshold defined per outcome — statistical significance alone is insufficient. We distinguish between a statistically significant effect at p=0.04 and a clinically relevant one.
Step 04
Standardisation assessment
We record which extract form, manufacturer, and standardisation specification was used in each trial. This is critical: a trial using Ashwagandha KSM-66 at 600mg does not validate an unstandardised ashwagandha powder product at the same dose. We only stock products whose standardisation matches the trial evidence.
Step 05
Safety data extraction
Adverse events, drug interactions, and contraindications are extracted from all trials and supplemented with pharmacovigilance case reports, FDA/TGA/MHRA databases, and published toxicology reviews. Safety data is always published regardless of efficacy grade.
Step 06
Grading and quarterly update
Grade is assigned, published, and reviewed quarterly. When a new meta-analysis changes the picture — even if it weakens the case for an herb we sell — the rating is updated. Negative findings receive equal prominence to positive findings. We have downgraded three herbs in the last 12 months.
Highest-evidence herbs
Our strongest dossiers
These herbs have the most robust clinical evidence bases. Each has been studied in multiple independent RCTs with consistent findings and a characterised mechanism of action.
A−
Ashwagandha (Withania somnifera, 印度人参)
74 studies · 28 RCTs · 10 meta-analyses
Strongest cortisol-reduction evidence of any adaptogen. 2019 Chandrasekhar et al.: serum cortisol −27.9% vs placebo (p<0.001). Consistent across 8 independent RCTs. KSM-66 and Sensoril are the validated extract forms. Also demonstrated: +14.7% testosterone in healthy men (Wankhede 2015), significant sleep quality improvement (Langade 2019).
A−
Panax Ginseng (人参)
112 studies · 38 RCTs · 14 meta-analyses
Most-studied adaptogen globally. 2024 meta-analysis (n=671, 15 RCTs): significant memory improvement SMD=0.19, stronger at high doses. Strong evidence for glucose metabolism (57 RCTs systematic review). Well-characterised ginsenoside mechanisms. G115 extract is the reference preparation.
A
Turmeric / Curcumin (姜黄)
96 studies · 34 RCTs · 18 meta-analyses
Strongest anti-inflammatory evidence base of any natural compound. 2018 meta-analysis (n=1,438): significant OA pain reduction comparable to NSAIDs. 2019 meta-analysis (n=531): significant antidepressant effect. Key limitation: bioavailability — piperine or phytosomal formulations required for clinical equivalence to trial doses.
A−
Ginkgo biloba (银杏叶)
88 studies · 30 RCTs · 15 meta-analyses
Cochrane 2009 (36 trials): consistent benefit for dementia and cognitive impairment. Gold standard: EGb 761 extract standardised to 24% flavonoid glycosides / 6% terpene lactones. Strong evidence for peripheral arterial disease and tinnitus. Antiplatelet effect is clinically significant — monitor with anticoagulants.
B+
Lion's Mane (Hericium erinaceus, 猴头菇)
54 studies · 14 RCTs · 6 meta-analyses
Only natural compound with strong RCT evidence for NGF synthesis in humans. Mori 2009 (n=30): significant improvement on MMSE and ADAS-cog at 16 weeks 3g/day. Effects reverse after cessation — confirming active mechanism. Three additional RCTs replicated findings. Also: anxiety reduction (Nagano 2010), sleep quality improvement (Saitsu 2019).
Transparency
What we don't sell — and why
Our review database includes herbs we've assessed and decided not to stock. These decisions are documented and public. This is not marketing — it is our obligation to users who might otherwise find these herbs sold without safety context.
Raw He Shou Wu (Reynoutria multiflora, 何首乌 — unprocessed)
Clear hepatotoxicity signal in published systematic reviews and case series. The hepatotoxic compounds (anthraquinones including emodin) are substantially reduced in properly processed (prepared/zhi) form. We stock only processed form with strict dose caps (≤6g/day) and explicit liver monitoring guidance. At least 23 WHO-documented cases of liver injury with raw form.
Aconite (Aconitum carmichaelii, 附子 — unprocessed)
Aconitine and related alkaloids are potent cardiotoxins. Unprocessed aconite root has a documented fatality record. Processed forms (pao fu zi) reduce alkaloid content significantly but require precise preparation protocols unavailable in standardised supplement manufacturing. We will not stock any aconite preparation until pharmaceutical-grade standardisation is available.
Aristolochia species (马兜铃 and related)
Aristolochic acid is a proven nephrotoxin and human carcinogen (IARC Group 1). Causes aristolochic acid nephropathy (AAN) — progressive renal fibrosis leading to dialysis or transplant. Banned by regulatory authorities in Australia, EU, USA, and most of Asia. We will never stock any Aristolochia species regardless of dose or preparation.
Ephedra (Ephedra sinica, 麻黄 — high-dose supplements)
Ephedrine and pseudoephedrine have a documented cardiovascular adverse event record at supplement doses (not traditional decoction doses). FDA banned ephedra dietary supplements in 2004 following deaths. We do not stock ephedra-containing products in supplement form. Traditional low-dose decoction use is a different risk context.
Quality control
Certificate of analysis explained
Every product batch is tested by an independent, ISO/IEC 17025-accredited laboratory before release. We require six core test categories.
What the COA certifies
Every batch COA is accessible via QR code on the product label. The document is hosted on our server and cannot be printed and reused — the URL is batch-specific and live.
IdentityHPTLC fingerprinting or DNA barcoding confirms correct species. Adulterations with related but different species is a documented problem in TCM supply chains.
PotencyActive compound percentage declared and verified. Beta-glucans by AOAC 995.16. Ginsenosides by HPLC. Withanolides by HPLC-UV. Curcuminoids by AOAC 2012.24.
Heavy metalsArsenic, cadmium, lead, mercury. All below USP <2232> limits. ICP-MS method. Particularly important for herbs grown in Chinese agricultural soil.
MicrobialsTotal aerobic plate count, yeast and mould, E. coli, Salmonella, Staphylococcus aureus. USP <2021> and <62> methods.
PesticidesMulti-residue screen covering 400+ pesticides by LC-MS/MS and GC-MS/MS. Not-detected required for release. EU MRL limits applied as standard.
AflatoxinsB1, B2, G1, G2 by HPLC. Critical for mushroom products. Total aflatoxin limit: <4 ppb (USP). Not detected is our internal standard.
Limitations
Honest caveats
We want to be direct about what the clinical evidence for TCM herbs can and cannot tell you.
Most trials are small. The majority of TCM RCTs have fewer than 100 participants. This limits statistical power and increases the chance of false positives. We note sample sizes explicitly for this reason.
Publication bias exists. Positive trials are more likely to be published than negative ones, particularly in Chinese-language journals. We search Chinese databases specifically but cannot guarantee we have found all negative results.
Extract standardisation is inconsistent. A trial using KSM-66 Ashwagandha does not validate a bulk ashwagandha powder product. We only recommend products whose standardisation matches the evidence we cite — but other vendors may not share this constraint.
Long-term human data is limited. Most trials run 4–16 weeks. Safety data beyond 12 months is sparse for most herbs. Our safety assessments note this limitation explicitly where relevant.
We are not physicians. The information on this site is educational and does not constitute medical advice. Discuss any supplement use with a qualified healthcare provider, especially if you are on prescription medications or have a chronic condition.