Proven in whom

Adding FDA-approved products to this register was supposed to give us a reference class — products where a regulator saw the dossier, so the concentration, the endpoint and the effect size are all checkable. It did. It also produced a finding we did not go looking for, which is that the demographic tables in those dossiers describe someone other than the person reading this.

Every approved label prints who was in the trials. Almost nobody reads that section, because it looks like administrative furniture. Read seventeen of them together and it stops looking administrative.

The age table

Three programmes in that list break the pattern — Finacea Foam at a mean of 50.6, Epsolay at 51, and Klisyri at an average of 70 — and none of them broke it on purpose. Rosacea presents in midlife and actinic keratosis in old age, so the enrolment followed the disease. Nobody set out to study the people this site is written for; two diseases happened to belong to them.

The retinoid a forty-two-year-old is prescribed for lines was measured in a population where nearly half the subjects were children.

This is not a scandal and it is not fraud. Acne is commonest in adolescence, so acne trials enrol adolescents. Retinoids work through the same receptors at every age. Prescribing an acne-indicated retinoid to an adult for photoageing is routine, defensible and often the right thing to do. But it is extrapolation, twice over — a different age and a different indication — and it is worth naming out loud rather than letting the word 'proven' do work it has not earned.

The second axis is skin type, and it is worse

Age at least gets printed. Ethnic composition gets printed too, and then nothing is done with it.

• ▸Soolantra: 96% of 1,371 subjects were Caucasian, in a condition that is under-recognised rather than absent in skin of colour.
• ▸Finacea Foam: 95.7% of 1,362 subjects were Caucasian.
• ▸Winlevi: 91% of 1,421 subjects were Caucasian.
• ▸Mirvaso: 99% of 553 subjects were Caucasian — the narrowest population in the register.
• ▸Zoryve foam is the exception that shows it is possible: 79% White, 11% Black, 5% Asian, 21% Hispanic or Latino, ages 9 to 87.
• ▸Seysara is the other: 2,002 subjects, 51% adults, 15% Black or African American.
• ▸Klisyri runs the other way — 99% White and 87% male — which reflects who presents with actinic keratosis and still narrows what the evidence covers.
• ▸Tri-Luma: Fitzpatrick V and VI were not studied at all, and the label concedes that excessive bleaching in darker skin cannot be excluded — for a drug treating a condition that disproportionately affects pigmented skin.
• ▸Renova: in the one trial run in Fitzpatrick IV–VI subjects, 43% improved on vehicle and 29% on the drug.

That last one deserves to be stated twice. The only FDA-approved topical for facial wrinkles has a single trial in darkly pigmented skin, and in it the cream without the drug outperformed the cream with it. The label reports the number, notes correctly that the study was small, and leaves it there. It is the most honest disclosure in this register and also the largest unanswered question in it.

What the compounders do with this

Nothing, because they cannot. A compounded formula has no registration programme, so it has no demographic table to be inadequate. When a telehealth page says its actives are clinically proven, the proof it is pointing at is the proof described above — generated in someone else's trial, at someone else's concentration, in a different vehicle, usually in a much younger population.

That is not an argument that compounded products do not work. It is an argument about what the word proven is doing on those pages. The evidence exists; it belongs to the molecule, at strengths that were tested, in bodies that were enrolled. Every step away from those conditions is an inference, and inferences compound.

The question to ask

When something is described as clinically proven, the useful follow-up is not whether the trial existed. It is: proven in whom, at what concentration, in what vehicle, against what comparator, and for how long. Four of those five are usually answerable from an approved label in about ninety seconds. On a compounded product page, most of the time, none of them is.

We will keep recording the demographic tables in every review of an approved product, and keep noting their absence everywhere else. It is the cheapest form of scepticism available and it costs the reader nothing.