How we read the evidence
This page stands in place of a byline. Rather than asking you to trust a name, it sets out which documents count here, in what order, and how a sentence signals the strength of the record behind it. Every article on this site links back to it.
The source ranking
Four kinds of document, in this order. A claim that can only be supported further down the list is written so that the sentence itself says so.
1. The regulatory record
FDA-approved prescribing information and device labelling, premarket approval and 510(k) clearance decisions, safety communications, recalls and warning letters. These sit at the top for a specific reason: a label is the only document in this field that is legally required to describe the limits of a product as carefully as its uses. It states the approved indication, the population studied, the warnings, and often the sentence a marketing page will not print.
Labels are read at their source — usually DailyMed for drug products and FDA’s own device pages — and dated in the citation, because they are revised.
2. Systematic reviews and meta-analyses
Read for two numbers before any conclusion: how many studies the review screened, and how many it could actually use. A review that screens a hundred articles and finds a dozen usable ones has told you something important about the field before it says a word about the treatment.
3. Primary trials
Randomised and controlled work first, then uncontrolled series. The sample size travels inline, in parentheses, in the same sentence as the finding, because a result in nineteen people and a result in two thousand are different kinds of fact and should not look the same on the page.
4. Professional-body guidance
Guidance from medical societies and national health bodies is used, and named as guidance rather than as evidence. It reflects a consensus of practitioners, which is useful and is not the same thing as a measurement.
What is not treated as a source
Manufacturer marketing material, including brochures, device-maker white papers and sponsored summaries. Single case reports presented as if typical. Clinic-published photography of any kind, which has no control, no scale and no independent scoring. Where a manufacturer’s own figure is the only figure that exists — and in this field it sometimes is — it is printed as a manufacturer figure, beside the independent numbers where independent numbers exist.
The six rungs
Every substantive claim on this site sits on one of six rungs. The rung appears as a small mark in the margin beside the sentence, and the sentence is written in that rung’s grammar, so the strength of the claim is legible from the wording alone even if the mark is missed.
The marks are typographic and monochrome. Colour is never used to encode how strong a piece of evidence is, because colour coding invites a reader to skim for green and there is very little green in this field.
| Rung | Mark | The grammar a sentence on this rung must use |
|---|---|---|
| Label | “FDA has approved X for Y in adults aged 22 and over.” The claim is a quotation of the regulatory record, and the document is named and dated. | |
| Consistent trials | “Randomised trials consistently find…” More than one controlled trial, pointing the same way, with the sample sizes given. | |
| Limited | “The evidence is limited to a small number of trials; they found…” Real evidence, too little of it to be settled. | |
| Contested | “Two reviews reach different conclusions: one found…, the other found…” Both sides are named. The disagreement is the finding. | |
| Uncertain | “It is not clear whether…” Evidence exists and does not answer the question asked. | |
| Not evaluated | “This has not been evaluated in clinical studies.” Nobody has measured it. This is a finding in its own right and it is printed in the same typeface as every other rung. |
The bottom two rungs are the reason this publication exists. A page that can print “not evaluated” in the same weight it uses for “FDA has approved” is doing something a page that is also selling the treatment has no way to do.
Rules the prose follows
- The source is the subject of the sentence. If the document cannot be named in the same sentence or the one before it, the claim drops a rung or comes out.
- Four phrases are barred. Studies show, research suggests, experts agree and clinically proven all assert a consensus while naming nobody. Each is replaced by the document that would have to exist for the sentence to be true.
- Sample sizes travel inline. A number without its denominator is decoration.
- Nothing is called safe. No treatment on this site is described as safe, as free of risk, or as gentle. Risks are reported by how common they are and by whether they are permanent, and those two things are stated together, because a rare complication that cannot be reversed reads very differently from a common one that resolves in a week.
- No doses, no technique. Where a study used a stated dose, that is reported as a fact about the study and attributed to it. Nothing on this site is written as a quantity or a method anyone could act on.
- No first-person experience. Nobody at this publication has had any of these treatments, so no page describes having one.
The two dates
Every page carries two. Reviewed is the date the page was last read through in full. Evidence current to is the date the sources behind it were last checked against their originals.
There is no third date. A stated date for a future review would be a claim about what happens next rather than a description of what has already been done, and a page that carries one and then misses it has published a broken promise on top of an old fact. Two dates, both about the past, are enough for a reader to judge how much weight the page can hold.
Corrections
A misread source is the most useful thing a reader can send. The contact form has a correction route: naming the page, the sentence and the document it should have followed makes it possible to check quickly.
Reviewed: 8 August 2026 · Evidence current to: 8 August 2026