Retinol: the wrinkle evidence is real, but the numbers came from the arms of 87-year-olds

A 0.4% retinol lotion was applied up to three times a week for 24 weeks. The fine-wrinkle score changed by −1.64 (95% CI −2.06 to −1.22), against −0.08 for the vehicle: the same formulation minus the active. The difference is clear.

Three qualifications. Participants had a mean age of 87 and lived in senior residences; the product went on the arm, not the face; and the tissue findings (increased glycosaminoglycan and procollagen I) come from subgroups of six and four people.

There is a facial trial too: 44 people, double-blind, 0.5% retinol for 12 weeks, wrinkle area reduced. But that trial compared retinol against bakuchiol, not against vehicle. Both sides improving does not, on its own, establish that retinol beats applying nothing.

In the same 44-person, 12-week trial, both 0.5% retinol and 0.5% bakuchiol reduced hyperpigmentation, and the difference between them was not statistically significant. Irritation ran the other way: scaling and stinging were reported more often with retinol.

With no vehicle arm, the result supports 'retinol does about as much as bakuchiol' and no further. What retinol alone does for pigmentation needs a larger trial.

A 2026 systematic review and meta-analysis pooled eight studies of first-trimester topical retinoid exposure covering 3,632,118 pregnancies. The risk ratio for major congenital malformations was 0.83 (95% CI 0.64 to 1.07) and for spontaneous abortion 0.99 (0.68 to 1.43) — no increase was found.

These are pooled observational studies, not randomised ones, and the authors note the result is driven largely by a single large registry study.

And the teratogenicity of oral retinoids such as isotretinoin is established. Do not read the topical and the oral literature as one.

The row at the top of this page says 'do not use' because that is what guidance says, not because harm has been demonstrated. If you are pregnant or planning to be, this is a decision to make with a doctor.

Korea's Regulation on the Review of Functional Cosmetics, Annex 4, lists retinol at 2,500 IU/g for wrinkle improvement. One IU of vitamin A is defined as 0.3 µg of retinol, so 2,500 IU/g is 750 µg/g, that is 0.075%.

That figure does not mean 'this concentration works'. It means 'at this concentration a product can be filed as functional without submitting its own data'. The trials above used 0.4 to 0.5% — five or six times higher. A 'wrinkle-improvement functional cosmetic' label on the back of a bottle tells you nothing about how much retinol is in it.