How we evaluate evidence
Natural health is a field where strong evidence, emerging research, traditional wisdom, and outright misinformation often sit side by side. Here is exactly how Coeheal distinguishes between them.
The five-tier evidence hierarchy
Every article on Coeheal is assigned an evidence tier. The tier appears as a coloured banner at the top of the article and as a full evidence card that you can expand to see study types, populations studied, date range, limitations, and source links.
Multiple large randomised controlled trials (RCTs) and/or high-quality meta-analyses in humans. Results are consistent across independent research groups and populations.
Examples on Coeheal: Mediterranean diet and cardiovascular disease; exercise and all-cause mortality; vitamin D deficiency and bone health.
At least one well-designed RCT or several consistent cohort studies in humans. Some limitations in study size, duration, or population diversity.
Examples on Coeheal: Omega-3 and triglyceride reduction; intermittent fasting and metabolic markers; lion's mane and cognitive function in older adults.
Human studies exist but results are inconsistent, populations are small, or methodological quality is limited. Promising but not yet conclusive.
Examples on Coeheal: Ketogenic diet as cancer adjunct; mistletoe therapy in integrative oncology; resveratrol and longevity markers.
Primarily animal or in vitro studies, with limited or no human trial data. Mechanistic plausibility exists but clinical translation is unproven.
Examples on Coeheal: Fenbendazole and cancer (preclinical only); most parasite-cleanse protocols; many high-dose antioxidant claims.
No credible human evidence, or evidence of harm outweighs any plausible benefit. May include traditional use with documented safety risks.
Examples on Coeheal: Coffee enemas for detoxification; Gerson therapy as cancer treatment; high-dose supplements with documented toxicity.
Why is there no grade E?
Our grading system uses A, B, C, D, and F — deliberately skipping E. This is an intentional design choice, not an oversight.
The scale is adapted from academic grading conventions, where F has a clear, universally understood meaning: failure — in our context, no credible evidence or documented harm. Inserting an E between D and F would create ambiguity. Readers familiar with graded scales tend to read E as "nearly failing but not quite" — a softer signal than we intend for content that carries genuine safety concerns.
By jumping directly from D to F, we preserve a sharp, unambiguous boundary. D means weak or indirect evidence — the claim may have biological plausibility but lacks human trial support. F means the evidence is absent or the risk of harm is documented. There is no meaningful middle ground between those two positions that warrants its own tier.
This also keeps the system consistent with how Coeheal's Claim Decoder grades individual health claims — the same A–F scale appears there, and users who move between articles and the decoder see the same framework without needing to relearn it.
Our editorial process
Six principles govern how every article is researched, written, reviewed, and maintained.
Source hierarchy
We prioritise Cochrane reviews, IARC monographs, NICE guidelines, and peer-reviewed journals indexed in PubMed. We cite primary sources, not secondary health websites. Where a claim relies on a single study, we say so.
Separating mechanism from outcome
A plausible biological mechanism (e.g. "curcumin inhibits NF-κB in cell cultures") does not mean a clinical outcome in humans. We explicitly distinguish lab findings from human trial results in every article.
Cancer claim protocol
Any article discussing natural approaches to cancer carries a mandatory red-flagged panel separating preclinical findings (cell/animal) from human clinical evidence. We never present unproven treatments as alternatives to conventional care.
Contributor accountability
Every article lists its author's credentials and role. High-risk articles (cancer, serious conditions) are reviewed by a named clinician with disclosed qualifications and conflict-of-interest statement. See our Contributors page.
Correction policy
When we make a material error, we correct it transparently. The correction is logged with a date, a summary of what changed, and why. Corrections are visible on the article itself — we do not silently edit content.
CoeAI safety rules
Our AI assistant follows five explicit safety rules: it routes urgent symptoms to emergency services, never diagnoses, flags supplement-drug interactions, never advises stopping prescribed treatment, and applies extra caution with vulnerable populations.
Test any claim yourself
The Claim Decoder applies this same evidence framework to any health claim you paste in. It returns a verdict, an A–F evidence grade, what researchers actually found, safety concerns, and links to original sources.
Open the Claim DecoderWhat Coeheal does not do
Diagnose, treat, or prescribe. Coeheal is an educational resource, not a medical service.
Present preclinical findings as human evidence. Cell studies and animal models are labelled as such.
Recommend stopping or replacing prescribed treatment with natural alternatives.
Publish affiliate-driven supplement recommendations without disclosing the relationship.
Silently correct errors. All material corrections are logged with a date and explanation.
Allow CoeAI to fabricate source URLs. The AI is instructed to cite only real, verifiable sources.