Ingredient Beauty BenefitsRetinoids
Retinoids are sold as a ladder. Tretinoin strongest, then retinaldehyde, then retinol, with HPR as the 'next generation'.
The trials on this site do not build that ladder. The table shows where it comes apart.
- the endpoints differ. Retinol's moderate evidence hangs on a blinded rater's score; retinaldehyde's hangs on an instrument reading. The two numbers were never counting the same thing
- almost none were compared with each other. We found no trial pitting one rung of the ladder against another. The only head-to-head in this set is bakuchiol against retinol — the one the ladder does not even admit is the one with the head-to-head
- adapalene's evidence here is not about wrinkles. The moderate finding on the adapalene page is about atrophic acne scars. That is not a number you can put on the same ladder
- HPR is still empty. Both findings on the HPR page are 'not established', and one of them is lab-only. The other was measured only inside a serum with four actives in it
- the retinaldehyde trial answered a question about concentration. It compared 0.1% with 0.05% and found no difference on most measures
So what this table tells you is not 'which is stronger'. It is which rung was measured with what, and which rung is empty.
The oral side is not in the table. Isotretinoin is the same family but there is no axis on which to compare it with something you apply. And the tretinoin and retinol pages both carry a separate finding about pregnancy — that is the thing to read before picking a rung.
| Item | Grade | Endpoint | Randomised trials in humans | Comparator · Funding |
|---|---|---|---|---|
| TretinoinMedicine · 0.01–0.05%79% improved in a 251-person trial, against 48% on vehicle | Moderate evidence | scores from blinded assessors | 2 · 29–251 participants · 24–26 weeks | vehicle (same formula minus the active)not declared |
| AdapaleneMedicine · 0.3%On one face, one side lost scars and the other gained them | Moderate evidence | scores from blinded assessors | 1 · 67 participants · 24 weeks | vehicle (same formula minus the active)not declared |
| RetinaldehydeIngredient · 0.05–0.1%0.1% versus 0.05% over 13 weeks | Moderate evidence | instrument measurement | 1 · 40 participants · 13 weeks | another active ingredientnot declared |
| RetinolIngredient · 0.4–0.5%Fine wrinkles decreased — in 36 people, over 24 weeks, on the arm | Moderate evidence | scores from blinded assessors | 2 · 36–44 participants · 12–24 weeks | vehicle (same formula minus the active), another active ingredientnot declared |
| Hydroxypinacolone retinoateIngredientThe human material is only from a serum with four actives in it | Not established | instrument measurement | 0 | not stated in the abstractmanufacturer-funded |
| BakuchiolIngredient · 0.5%No difference from retinol, and less irritation | Moderate evidence | scores from blinded assessors | 1 · 44 participants · 12 weeks | another active ingredientnot declared |
This is not a ranking
This site does not rank or star-rate ingredients and procedures. The comparison tables are no different.
The row order is the layer and the seed order, and being higher up does not mean better. Where several items share a grade, adding any second sort key would make that key another ranking, so there is none.
And every row carries the title of the finding that grade hangs on. Without it, the row reads as 'this ingredient's score'. It is the same reason we do not average when a page has several findings — an average is a number nobody measured.
Which finding goes in the table is a rule too. If the comparison names a question, that finding (as with the three measured against hydroquinone in the brightening set); if it names none, the highest-graded finding. Entries marked as ingredient data — facts from an ingredient list or a regulatory document — are not efficacy claims, so they are never the representative.