The grading scale
Seven levels, ordered by the strength of the best available human evidence. The grade describes the evidence, not the compound — it is not a rating of how good, safe or useful something is.
Licensed by at least one major regulator for a specific indication, on the strength of completed phase 3 trials. Approval is for that indication in that population — it is not a general endorsement of the compound.
Large randomised trials in humans are complete or under way, but the compound is not licensed. Efficacy and common adverse effects are reasonably well characterised; long-term safety may not be.
Tested in humans for efficacy at a modest scale. Enough to show a signal and characterise short-term tolerability, not enough to establish safety or durability. Most compounds that reach this stage still fail.
Studied in small numbers of humans, usually for safety and pharmacokinetics rather than whether it works. A phase 1 result says almost nothing about effectiveness.
Evidence comes from rodents or other animal models, with no published controlled human trials. Most compounds that look promising in animals do not replicate in people — the attrition rate from animal model to approved drug is well over 90%.
Evidence comes from cell culture or biochemical assays. This can establish that a compound does something to cells in a dish. It establishes nothing about what it does in a living body.
Taken into human trials and did not outperform placebo on its primary endpoint, or development was discontinued. This is stronger evidence than "untested" — it is evidence of absence, not absence of evidence.
How a grade is assigned
Three rules decide it, and they are applied consistently:
- Human evidence sets the ceiling. However extensive the animal literature, a compound with no published human trials is graded "animal studies only". Volume does not substitute for translation.
- The best evidence counts, not the average. A compound with one good phase 3 trial and a pile of weak observational work is graded on the trial.
- Negative results are graded as such. A compound taken into human trials that failed its primary endpoint is graded "failed", which sits below "animal only". That is deliberate: a failed trial is stronger evidence than no trial, it just points the other way.
Regulatory status is shown separately, because the two are not the same. A compound can be approved in one country and unlicensed in another without the underlying evidence changing at all.
Reading a study yourself
Every compound page links to the primary literature. That is deliberate — you should not have to take our word for any of this. A few things worth checking when you get there:
- Was it in humans? The single most important question, and the one most often lost when findings get summarised. "Shown to increase collagen synthesis" frequently means in cells in a dish.
- How many participants? Twelve people tells you about twelve people. Rare adverse effects need thousands to detect.
- Was there a control group, and was it blinded? Without a placebo arm you cannot separate the compound from everything else that changed. Unblinded studies of subjective outcomes are especially unreliable.
- What was the primary endpoint? Trials declare in advance what they are measuring. A trial that missed its primary endpoint but reports a positive secondary one has not shown what it appears to show.
- How long did it run? Twelve weeks says nothing about twelve months.
- Who funded it, and who ran it? Industry funding does not invalidate a study, but it is context. So is a body of work coming overwhelmingly from one research group — independent replication is what turns a finding into knowledge.
- Is it a review or primary research? Narrative reviews can repeat claims without re-checking them, and citation chains sometimes trace back to a single weak source.
Claims worth distrusting
- "Clinically proven" attached to a compound with no clinical trials. The phrase has no legal meaning in this context.
- "Studies show" with no study named. If it cannot be linked, it cannot be checked.
- Mechanism presented as outcome. "Activates AMPK" describes something happening in a cell. It is not a result.
- Animal findings stated without the animal. "Accelerates tendon healing" usually means in rats.
- A certificate of analysis offered as proof of safety. It attests to purity of a batch. It says nothing about whether the compound is safe, or whether that batch is the one you received.
- Absence of reported harm treated as evidence of safety. If nobody has studied it, nobody has found anything — which is not reassurance.
- Dosing protocols from anyone with something to sell. Where no trial has established a dose, a protocol is someone's guess presented with confidence.
Our editorial rules
- Describe what studies found. Never what anyone should do.
- No dosing, administration routes or protocols, on any page, and not on request. That is the line between an information resource and something that guides use.
- Every compound page carries a "what the evidence does not show" section, given the same prominence as the findings. A page with only positive results is marketing.
- Where a claim is widely repeated but poorly supported, say so plainly.
- Link the primary literature so the reader can check us.
- Grade on human evidence. A thousand mouse papers is still animal evidence.
- Report negative results. Compounds that were tested and failed are more informative than compounds nobody has looked at.
- Sell nothing. Genexis Health Ltd. holds no stock and arranges no supply, so no page here has a commercial reason to overstate a finding.
Corrections
We will get things wrong. Research moves, summaries simplify, and we are reading the same literature as everyone else.
If you believe a page misstates the evidence, email research@genexishealth.uk with the compound and, where you can, the study you think we have misread. We will check it against the source and correct it if you are right. Substantive corrections are noted on the page.
This applies especially if you are a researcher working on something described here. We would rather be corrected by someone who knows the field than be confidently wrong in public.