Porecelain

Flawless skin starts with the fine print.

Benzoyl Peroxide

Benzoyl peroxide is rated Clear, and there is no comedogenic number for the ingredient to correct, because it does not appear in the comedogenicity corpus we searched. The scores that do exist in that corpus are for a handful of 1980s finished benzoyl peroxide creams, at undisclosed strength and in bases we cannot see. The rating rests instead on clinical consensus: a dermatology guideline that recommends applying benzoyl peroxide to acne-prone skin, which is a treatment recommendation, not a measurement of pores.

Clear Moderate confidence No legacy score. It does not appear in the assay corpus we searched.

AAD 2024 STRONGLY recommends applying it to acne-prone skin. Not a comedogenicity assay.

What it is

Benzoyl peroxide is an organic peroxide — two benzoyl groups joined by a single oxygen–oxygen bond — and, unlike almost everything else rated on this site, it is a drug active rather than an emollient, a wax or a texture agent. A formulator does not reach for it to improve slip or feel. It is there to treat acne. It is used at 2.5% to 10%, in leave-on gels and creams and in wash-off cleansers, and it works by decomposing on the skin to release oxygen radicals that are bactericidal against Cutibacterium acnes (the bacterium implicated in inflammatory acne) and that have a mild drying, peeling effect. On a label it is Benzoyl Peroxide, declared as an active at its concentration.

Why we rate it Clear

We rate it Clear at Moderate confidence, and the most useful thing to say first is that the rating is not a corrected number — it is a rating built on top of a hole where a number is usually found. There is no comedogenicity score for benzoyl peroxide the ingredient. We looked in the primary documents, not on the charts. Fulton's 1989 rabbit-ear table — the 220-row survey nearly every circulating comedogenic score descends from, which we hold transcribed from a 400dpi render of the journal scan — has no benzoyl peroxide row anywhere in it. Fulton, Pay and Fulton's earlier 1984 paper has no benzoyl peroxide ingredient row either. Nguyen and Maibach's 2007 rabbit-ear study, on a 0–4 scale, quantified fourteen cosmetic materials; benzoyl peroxide is not among them. No study in the corpus we searched applied benzoyl peroxide to a follicle and measured what happened. What the 1984 paper does contain is a small group of finished, branded acne products, scored under the heading "Therapeutic agents": Desquam-X 4, Van Oxide 4, Sulfoxyl 4 and Persadox HP Cream 3, on the 0–5 scale. Those four are benzoyl peroxide products — Sulfoxyl is benzoyl peroxide combined with sulfur — and on Fulton's 1984 key a grade of 3 or above is "considered positive." It would be easy to read those numbers as benzoyl peroxide's. They are not. A score for a formulated cream, at a benzoyl peroxide concentration Fulton never states, in a base we cannot see, is a score for that product and not for the molecule — the same formula-for-ingredient substitution that once attached coconut butter's number to coconut oil. And the same "Therapeutic agents" group scored Retin-A Cream 4 and Hytone 5 — a tretinoin cream and a hydrocortisone cream, whose actives are not themselves comedogenic. Hytone's score demonstrates that a finished therapeutic product cannot be reduced to an ingredient score; it does not identify whether the active, the vehicle, another excipient or an interaction between them caused the response. Fulton reports no formula for any of these products, so a reader of his table cannot say which component the grade belongs to. So the Clear rating does not come from a comedogenicity measurement, because we identified none. It is worth setting out the order the evidence falls in, because the rating is built from the second rank down. First rank, exact-material induction on clean skin: nothing, in the sources and searches reviewed for this page. Second rank, finished-product rabbit assay: Fulton's 1984 cream scores, which belong to those creams. Third rank, vehicle-controlled human monotherapy trials. Fourth rank, guideline consensus. The third rank is where the practical reassurance sits. Kawashima and colleagues ran two large vehicle-controlled Japanese trials of benzoyl peroxide as monotherapy: in 2014, 360 patients on 3% benzoyl peroxide gel or vehicle for twelve weeks, with a difference in total lesion count of −21.0 in favour of the drug (P < 0.001); and in 2017, 609 patients randomised to 2.5% or 5% benzoyl peroxide gel or placebo, with median inflammatory-lesion reductions of 72.7% and 75.0% against 41.7% on placebo. Because the comparison is against the same gel base without the active, the difference is attributable to the benzoyl peroxide added to it. Those trials counted acne lesions on acne patients, not induced comedones on clear skin — but a material that reduces lesion counts against its own vehicle over twelve weeks on acne-prone faces is not behaving like a comedogen. The fourth rank is the American Academy of Dermatology's 2024 acne guideline, which gives benzoyl peroxide a strong recommendation, on moderate-certainty evidence, and recommends applying it to acne-prone skin — on its own and alongside topical antibiotics, where it also limits the development of resistance. A guideline that recommends putting a substance onto acne is a strong checkable proxy that the substance does not cause acne. It is a proxy, not a proof: it is a graded recommendation about treating a disease, not a count of comedones, and we are not going to launder one into the other.

What the evidence doesn’t tell you

The rating is an inference, and it should be read as one. We identified no comedogenicity assay of benzoyl peroxide as an isolated ingredient, in any primary document we hold or could find, and we identified no human comedone-induction study of it either. "Not measured" is not the same as "measured and found clear." The Kawashima trials tested finished 2.5%, 3% and 5% gels on people who already had acne, counting acne lesions rather than induced comedones, so they answer a neighbouring question rather than this one, and they yield no dose–response for the molecule. A clinical guideline recommending a drug for acne is strong evidence that the drug is safe and useful on acne-prone skin; it is not an experiment that isolated benzoyl peroxide and watched a follicle. The rating is built on the first standing in for the second, and that substitution, though it is the best available, is exactly the kind of move this site exists to name — so we are naming it in our own rating. The 1984 finished-product scores tell you about creams, not about the molecule: an undisclosed benzoyl peroxide concentration, an unseen base, the rabbit ear rather than a human face, and a grade key that differs from the 1989 one, so a "4" there does not mean a "4" elsewhere. They cannot be read onto the ingredient, and we have not. One distinction matters for how Clear is read here. Comedogenicity is not tolerability. Benzoyl peroxide is a well-documented irritant: dryness, peeling, stinging and contact dermatitis are common with it, and it bleaches fabric, hair and bedding. None of that is comedogenicity, and none of it is what this rating speaks to. Clear here means that no isolated-material comedogenic signal was identified, and that vehicle-controlled acne trials provide strong practical reassurance for the tested benzoyl peroxide gels. It says nothing about whether a given person will find it comfortable, and for many people it is not.

Where you’ll see it

On a label it is Benzoyl Peroxide, listed as an active at its strength — typically 2.5%, 4%, 5% or 10%. You meet it in dedicated acne treatments: leave-on spot gels and creams, benzoyl peroxide face and body washes, and fixed-dose prescription combinations with adapalene, clindamycin or erythromycin. It is the acne ingredient in those products, not a supporting one. Keep it separate from the four brand names in the evidence above. Desquam-X, Van Oxide, Sulfoxyl and Persadox HP are the finished 1980s products Fulton actually scored; the numbers belong to them, not to the ingredient on this page. We print "Van Oxide" as two words wherever we are reporting Fulton's row, because that is how his table prints it; the drug references that identify the brand set it as "Vanoxide," and we leave each source's spelling as we found it rather than silently reconciling them.

Sources

Each source says what it actually is. Several widely-cited “sources” for comedogenic ratings are republishing the same 1989 assay, and counting them as independent agreement is how a thin evidence base gets made to look thick.

  1. Fulton JE Jr, Pay SR, Fulton JE III. "Comedogenicity of current therapeutic products, cosmetics, and ingredients in the rabbit ear." J Am Acad Dermatol 10(1):96–105 (1984) A primary rabbit-ear comedogenicity study on the 0–5 scale — two New Zealand albino rabbits per test, one ear dosed daily for two weeks, the other the control — read in full off a 400dpi render of the journal scan and transcribed into this repository. It has NO benzoyl peroxide ingredient row. What it has, under "Therapeutic agents," is a group of finished branded acne products: Desquam-X 4, Van Oxide 4, Sulfoxyl 4 (benzoyl peroxide with sulfur) and Persadox HP Cream 3 — benzoyl peroxide products — alongside Retin-A Cream 4 and Hytone 5, whose actives (tretinoin, hydrocortisone) are not comedogenic. On the 1984 key, 3 and above is "positive." Cited here as a result about formulated creams at undisclosed active concentration, not as a score for benzoyl peroxide, which it is not.
  2. Fulton JE. "Comedogenicity and irritancy of commonly used ingredients in skin care products." J Soc Cosmet Chem 40:321–333 (1989) The 220-row rabbit-ear table (roughly 10% in propylene glycol, three New Zealand albino rabbits) that nearly every comedogenic score in circulation descends from, held here transcribed from a 400dpi render. Cited for what it does NOT contain: there is no benzoyl peroxide row anywhere in the paper, in any table. The single most-quoted source of comedogenic numbers never gave benzoyl peroxide one.
  3. Reynolds RV, Yeung H, Cheng CE, et al. "Guidelines of care for the management of acne vulgaris." J Am Acad Dermatol 90(5):1006.e1–1006.e30 (2024). PMID 38300170 The clinical practice guideline the rating rests on. Produced by a work group using graded evidence review, it gives benzoyl peroxide a STRONG recommendation on moderate-certainty evidence and recommends applying it to acne-prone skin, including in combination with topical antibiotics to limit resistance. It is a treatment recommendation, not a comedogenicity assay: it counts no comedones produced by benzoyl peroxide and measures no follicle. Cited as exactly what it is — the strongest checkable proxy that this ingredient does not cause acne, and a proxy is what it is.
  4. Nguyen SH, Dang TP, Maibach HI. "Comedogenicity in rabbit: some cosmetic ingredients/vehicles." Cutan Ocul Toxicol 26(4):287–292 (2007). PMID 18058303 An independent rabbit-ear comedogenicity study on a 0–4 scale (abstract read verbatim via EuropePMC; full text not read). Cited to bound the absence claim: it quantified fourteen cosmetic materials — isopropyl palmitate, isopropyl myristate, butyl stearate and the rest — and benzoyl peroxide is not among them. A second primary comedogenicity paper that never tested this ingredient.
  5. "Comedogenicity in cosmeceuticals: a review of clinical relevance, regulatory gaps, and future directions." JAAD Reviews (2025) A peer-reviewed review of the comedogenicity field — a synthesis of the legacy literature, not a new measurement, and independent of Fulton only in authorship, since the corpus it surveys includes his. Cited for two field-wide facts it establishes: that no standardised comedogenicity test exists and "non-comedogenic" is an unregulated label; and that testing a finished product is not equivalent to testing an ingredient, because the mixture and the base change the result — which is the exact reason Fulton's 1984 cream scores cannot be read onto benzoyl peroxide.
  6. Kawashima M, Nagare T, Katsuta Y, et al. "Efficacy and safety of benzoyl peroxide gel 3% in acne vulgaris: a randomized, double-blind, vehicle-controlled study." J Dermatol 41(9):795–801 (2014). PMID 25132461 A vehicle-controlled human monotherapy trial, abstract read verbatim via EuropePMC; full text not read. 360 Japanese patients randomised to 3% benzoyl peroxide gel or its vehicle for twelve weeks, with a difference in total lesion count of −21.0 in favour of the drug (P < 0.001). What it establishes: because the comparator is the same gel without the active, the effect is attributable to the benzoyl peroxide rather than to the base — lesion counts fell. What it does not: it counts acne lesions in acne patients, not comedones induced on clear skin, so it is not a comedogenicity assay, and it tests one strength in one finished gel.
  7. Kawashima M, Sato S, Furukawa F, et al. "Twelve-week, multicenter, placebo-controlled, randomized, double-blind, parallel-group comparative phase II/III study of benzoyl peroxide gel in patients with acne vulgaris." J Dermatol 44(7):774–782 (2017). PMID 28295516 The larger companion trial, abstract read verbatim via EuropePMC; full text not read. 609 patients randomised (607 in the full analysis set) to 2.5% or 5% benzoyl peroxide gel or placebo, with median inflammatory-lesion reductions of 72.7% and 75.0% against 41.7% on placebo. Cited for the same reason and with the same limit as the 2014 trial: a placebo-controlled reduction of existing acne lesions, not a comedone-induction measurement.
  8. Drug references identifying the 1980s brand names — Medscape drug monograph (benzoyl peroxide; Desquam-X) and RxMed monograph (Sulfoxyl, sulfur–benzoyl peroxide) Pharmacology references establishing that Desquam-X, Vanoxide, Sulfoxyl and Persadox are benzoyl peroxide products (Sulfoxyl being benzoyl peroxide combined with sulfur). Cited only to identify what the "Therapeutic agents" rows in Fulton 1984 actually were. Not comedogenicity evidence.

Others in the same family

Actives behave similarly enough that the evidence for one is often wrongly read across to the others. These are its structural relatives, not a guess. What the evidence says about actives as a family ›

Last reviewed 2026-08-05 · How we decide

‹ Check a full ingredient list