Ingredient Beauty BenefitsOther · Caffeine
Caffeine: it is in more eye products than almost anything — and we found no dark-circle trial
In a 12-week double-blind, placebo-controlled trial on thigh cellulite, a 2% caffeine cream brought scores from 3.96 to 2.50. But the placebo also fell, from 3.88 to 2.83. And for the place this ingredient is most used — the eye area — we found no human trial with dark circles as an endpoint.
Last checked Sep 19, 2026
- In pregnancy
- Not established
- Irritation
- Low
- Daytime use
- ✓ Fine
Not a measurement — reported irritation grouped into five steps. It varies between people.
Caffeine is a xanthine alkaloid — the same substance you drink. In cosmetics it turns up mainly in two places: eye products and body slimming products.
Two mechanisms are usually offered: that it constricts blood vessels and reduces puffiness, and that it inhibits phosphodiesterase, an enzyme that blocks fat breakdown inside fat cells.
But writing this page, what we ran into was a mismatch.
The place caffeine is most widely used is the eye area. And on PubMed we found no human trial of caffeine with dark circles or under-eye puffiness as an endpoint. What we found was one thigh cellulite trial and one hair-loss trial in which caffeine was mixed with something else.
This ingredient shows well that being widely used and being well evidenced are different things. The dark circles page explains why the evidence for that whole concern is thin.
Contents
What is established
The same ingredient at a different strength, in a different formula, is a different product. These are the conditions the studies used.
Cellulite scores fell — and so did the placebo's
A 2026 study. Most of the paper is formulation work, with the clinical part attached at the end.
Briefly on the formulation side. The researchers built nano-creams in several combinations to get caffeine further into skin, and the one containing lanolin permeated best — 8.829 µg/cm² per hour, more than three times an aqueous solution (2.533) or a commercial caffeine cellulite product (2.827). But those are numbers from a skin permeation test (IVPT), not from people.
That formulation then went into a clinical trial.
- Double-blind, randomised, placebo-controlled paired design
- Each volunteer applied the active cream to one upper thigh and placebo to the other
- Twice daily for 12 weeks
Comparing two sides of one person controls automatically for age, weight and habits. That is an unusually good condition for this layer.
The results.
- Active: cellulite score 3.96 (95% CI 3.16–4.76) → 2.50 (1.70–3.30)
- Placebo: 3.88 (3.08–4.67) → 2.83 (2.03–3.63)
- Effect sizes: active 0.475 (moderate), placebo 0.286 (small)
Two things have to be read together here.
The active cream did fall further. But the placebo fell too. Applying a cream with nothing in it to a thigh twice a day for 12 weeks took the score from 3.88 to 2.83.
So caffeine's contribution is not the whole of 3.96 → 2.50, but the amount beyond placebo. In effect-size terms, that is 0.475 against 0.286.
And the abstract does not report the number of participants. We left that field blank rather than fill it. The value of a paired design leans heavily on its size, and we are reading it without knowing that size.
Finally, this result belongs to one specific lanolin nano-formulation. The same paper showed that at the same 2%, formulation changed how much got into skin by a factor of three.
- Randomised controlled trial in people · sample size not reported · 12 weeks · Compared with vehicle (same formula minus the active) · Funding not declared · Endpoint scores from blinded assessors PMID 41754894
The hair trial was not caffeine on its own
A 2025 24-week, double-blind, randomised, placebo-controlled trial in male pattern hair loss.
What it tested was a topical product containing dimethylglycine sodium salt (DMG-Na) together with caffeine.
So the question it answers is not 'does caffeine work'. It is whether that combination beats placebo. What happens if you take the caffeine out is not in this trial.
The ferulic acid page on this site records a problem of the same shape — there a vitamin C, E and ferulic serum made it impossible to apportion the result to ferulic acid. Separating them needs a design that adds or removes one ingredient, and the kojic acid page has an example of exactly that.
And this is about scalp. This site covers skin, not hair loss. It is recorded here for one reason: it was the only other human caffeine trial we found. We write it down rather than hide it, and we also write down that it is not a question this site can answer.
This trial too gives no participant count in its abstract.
- Randomised controlled trial in people · sample size not reported · 24 weeks · Compared with vehicle (same formula minus the active) · Funding not declared · Endpoint scores from blinded assessors PMID 40820949
We found no trial looking at dark circles
This is the most important item on the page.
The place caffeine is most widely used in cosmetics is the eye area. You will have seen the name on eye-cream ingredient lists more than once.
We found no human trial comparing caffeine against placebo with dark circles or under-eye puffiness as an endpoint.
To be precise: this is not 'it does not work'. A trial that ran and came out negative and a trial we could not find are different things, and this site writes them with different words.
Why would there be no trial here? There is a guessable reason. As the dark circles page sets out, under-eye darkness comes by three routes — pigment, vessels and shadow — and more than one is often present in the same person. Recruiting a clean group is hard, and there is no standard way to measure it.
The commonly cited rationale is that caffeine constricts vessels and so reduces puffiness and show-through. We are not saying that is wrong. We are saying we could not find it confirmed in people.
The practical implication: we have no grounds to tell you not to use a caffeine eye cream either. But if a product says 'clinically proven', it is fair to ask proven for what — it might be cellulite, or it might be a formulation permeation test.
How to use it
- The cellulite trial used 2% cream twice daily for 12 weeks on the thighs.
- Formulation matters a lot. In the same paper, a lanolin nano-formulation permeated more than three times better than an aqueous solution.
- The trials above say nothing about use around the eyes. No concentration, frequency or duration is established for that.
- Body products go on large areas. If irritation appears, it appears over that same area.
Cautions
- We have not verified the range and frequency of adverse events reported in the trials above.
- Neither trial gives a participant count in its abstract.
- We found no established data on use during pregnancy.
- Cellulite moves with body weight. What happened to participants' weight over the 12 weeks is not in the abstract.
With a new ingredient, apply a small amount somewhere inconspicuous, such as the inner arm, for two days first. Stop if it reddens or stings.
Related skin concerns
What this article relies on
- PubMed — 미국 국립의학도서관 문헌 데이터베이스National Library of Medicine · Checked on 2026-09-16