Actives
The comedogenic numbers that circulate against this family are not numbers for these ingredients. Fulton's 1989 table, the source of nearly every score in circulation, has no row for any of the seven. His earlier 1984 paper did reach this territory, and it reached it as finished products: a column headed "Therapeutic agents" in which Retin-A Cream scored 4 of 5, Desquam-X 4, and Hytone, a hydrocortisone cream, 5 - nine numbers we read off a standard-resolution scan and have not re-verified at 400dpi, unlike that paper's ingredient rows. One member has an ingredient row of its own in the numeric documents we hold, and it is colloidal sulfur.
What they are
These are the ingredients a formula contains in order to DO something measurable to skin, rather than to make a product spread, feel or keep. The family holds seven: benzoyl peroxide, salicylic acid, azelaic acid and colloidal sulfur, all four of which are recognised acne treatments; retinol, which the skin oxidises by way of retinaldehyde to retinoic acid, the drug sold as tretinoin, though a retinol on a label is not interchangeable with it; niacinamide, a form of vitamin B3; and bakuchiol, a plant phenol sold as a retinol alternative. Two things follow from that composition and they pull against each other. Four of the seven are drugs used to TREAT the condition a pore-clogging checker is warning about, which is why a page like this one reads oddly at first. And the group is a functional category, not a chemical one: an organic peroxide, a hydroxy acid, a dicarboxylic acid, an element, a vitamin A alcohol, a vitamin B3 amide and a meroterpene phenol share no structure at all. Grouping them is a filing decision about what they are for, and it should not be mistaken for a claim that they behave alike. "Active" is OUR category. It is not one of Fulton's sections, and where a reader expects an overlap there usually is not one. Ascorbic acid, ascorbyl palmitate and tocopherol are the obvious near neighbours, and they are NOT in this family, and will not appear in the table below. We file all three as ANTIOXIDANTS. Fulton's own arrangement is different again, and we can only describe the part of it we hold: our transcription of his 1989 table records one heading of that kind, "X. Vitamins and herbs", and does not record which rows sit under it, while his sunscreen esters simply run together in the printed order. His 1984 table sorts by lanolins, surfactants, pigments and a therapeutics column that mixes drug classes freely. His sections and our families are different sets, and this page says which one it means every time.
What the evidence actually says
The family-level fact is about where the numbers came from, and it is checkable in two printed tables. FIRST, THE ABSENCE, AND IT IS BOUNDED TO A DOCUMENT WE HOLD. Fulton's 1989 survey is the 220-row rabbit-ear table from which nearly every comedogenic score now in circulation descends. We hold it transcribed verbatim from a 400dpi render of the journal scan, and we have read every row. It contains no row for benzoyl peroxide, salicylic acid, azelaic acid, retinol, niacinamide, bakuchiol or colloidal sulfur. Not one of the seven ingredients on this page is in it. Nor does the one later rabbit-ear paper in our corpus close the gap: Nguyen and Maibach 2007 names fourteen test materials in the abstract we hold, which is all of that paper we have read, and no member of this family is among them. What IS in it, sitting close enough to be mistaken for them, is three near misses. "Octyl salicylate" at 0 of 5 is a sunscreen ester, octisalate, printed in the run of rows that also carries octyl dimethyl PABA, oxybenzone and octyl methoxycinnamate; it shares a word with salicylic acid and nothing else. "Vitamin A palmitate" at 1-3* of 5 is retinyl palmitate, a larger molecule than retinol, and the asterisk is Fulton's footnote meaning results depend on source of raw material; the lists that quote the row print a bare 1-3. "Precipitated sulfur" at 0 of 5 and "Water-soluble sulfur" at 3 of 5 are two other sulfur preparations, and neither of them is the colloidal grade. Read a number off any of those three rows and you are reading a different material's result. SECOND, THE THING THAT WAS ACTUALLY TESTED. Fulton's earlier paper with Pay and Fulton, published in 1984, is titled "Comedogenicity of current therapeutic products, cosmetics, and ingredients in the rabbit ear", and it did reach this family. It reached it as FINISHED PRODUCTS. Its Table I carries a column headed "Therapeutic agents", and the rows read: Hytone 5, Retin-A Cream 4, Staticin 4, Sulfoxyl 4, Desquam-X 4, Van Oxide 4, Persadox HP Cream 3, Liquimat 3, and, filed in the same column by the table itself, D & C Red No. 40* at 4. All on a 0-5 scale, and all nine transcribed from a standard-resolution scan rather than from the 400dpi render we used for that paper's ingredient rows, so read them as our reading of a scan pending a second look. Every row in the column is 3 or above, and that paper states its own key: "a grading score of 3 or above is uniformly reproducible and considered positive." Those brands are, between them, most of this family's drugs. Retin-A is tretinoin, named as such in a 1983 paper by Mills and Kligman that used it. Desquam-X, Van Oxide, Persadox and Sulfoxyl are benzoyl peroxide products, Sulfoxyl being benzoyl peroxide with sulfur. Staticin is topical erythromycin. Liquimat is a sulfur acne lotion. Hytone is a hydrocortisone cream. AND HYTONE IS WHY THE COLUMN CANNOT BE READ AS A SET OF INGREDIENT SCORES. The cream graded 5, the highest score in the column. Fulton prints no formula and no active concentration for any of these products, so his table cannot tell a reader whether the grade belonged to the drug, the base, an excipient or the interaction. This is the same substitution that once attached coconut butter's number to coconut oil, and it cuts in both directions: the column is not evidence that tretinoin is comedogenic, and it is not evidence that it is not. It is also worth saying what the assay did NOT do, because the tidy version of this page would claim it. The 1984 table did not grade everything positive. In its other columns, Revlon's Natural Wonder Oil-Free Base, Almay's Pure Beauty Foundation for Oily Skin, Helena Rubinstein's Bio Clear and Clinique's Pore Minimizer all scored 0, as did Oil-of-Olay, Albolene, mineral oil and petrolatum among the moisturisers. The therapeutic column stands out from the rest of its own table. THIRD, THE ONE INGREDIENT ROW. Colloidal sulfur is the exception in this family and has its own page. Fulton 1984 prints the row unasterisked and scores it 3 of 5, which is positive on that paper's key. Read the key with the number: a 3 counts as positive in 1984 and would count as borderline under the 1989 paper's different key, so a 3 from one of Fulton's papers is not the same claim as a 3 from the other. The same 1984 table separately prints "Precipitated sulfur*" at 0-2 of 5, the asterisk there meaning material diluted in mineral oil or made up as a 10 per cent alcohol gel. One element, two grades, two different answers, in one table. FOURTH, AND THIS IS THE PART THAT SURPRISED US. The same laboratory that built the rabbit-ear model ran the measurement in the opposite direction, on human faces, and published a correction of its own rabbit reading. Mills and Kligman's "Assay of Comedolytic Activity in Acne Patients" (Acta Dermatovener (Stockholm) 63:68-71, 1983; received April 1982, and not the same paper as Mills and Kligman 1982 in Archives of Dermatology 118:903-905) recruited college students with moderately severe acne, required at least 15 microcomedones per square centimetre at baseline, treated cheeks and foreheads twice daily, and counted microcomedones by cyanoacrylate follicular biopsy across five one-square-centimetre fields, in panels of ten to twelve, at 8 and 12 weeks. Of seven anti-acne preparations, they judged only two comedolytic: tretinoin 0.05 per cent, as Retin-A liquid, at a 75 per cent reduction by 8 weeks, and salicylic acid 5 per cent, at 40 per cent by 8 weeks and 48 per cent by 12. Benzoyl peroxide 10 per cent as Persa-Gel (10 and 19 per cent), clindamycin, lotio alba, sulfur 5 per cent in a shake lotion (7 and 9 per cent) and resorcinol were judged inactive. Their discussion contains the sentence this whole page turns on: "In a previous study, using the rabbit ear model, we reported that benzoyl peroxide was moderately comedolytic. Assays with various commercial formulations since that time have led us to revise this assessment." They give the reason in the next lines: "The softer comedones of the rabbit are much more easily exfoliated; hence, comedolytic activity tends to be overrated. The vehicle itself also becomes a factor." So the shape of the evidence for this family is: one rabbit-ear ingredient row, one rabbit-ear product column, and one human follicular-biopsy measurement of a different endpoint. Only the first of the three is a comedogenicity score for an ingredient a reader will find on a label, and it belongs to colloidal sulfur. That is why six of the seven member ratings on this site rest on a treatment guideline or on an absence rather than on a number, why the seventh rests on a printed row, and why saying which is which is more useful than printing a 0.
What the evidence doesn’t tell you
The absence is bounded, and it has to be. What we can say is that we identified no comedogenicity assay of benzoyl peroxide, salicylic acid, azelaic acid, retinol, niacinamide or bakuchiol as isolated ingredients, in the numeric documents we hold or in the searches run for their pages. That is a statement about a corpus, not about the world. Our own provenance file says to treat the corpus as OPEN, and it says so because it grew: a ninth numeric paper, Fulton 1976 in Cutis, turned up after we had written down that there were eight. An absent score is not a zero, and unproven is not disproven. Nothing on this page establishes that any of these ingredients is safe for pores. It establishes that the reassuring 0 printed against them in consumer checkers has no assay behind it that we could trace, which is a different and smaller claim. The therapeutic column is the weakest transcription on this page, which is why its disclosure is printed beside the numbers themselves and again in the summary rather than kept down here: our 1984 file records that the branded left-hand columns were transcribed from a standard-resolution scan and NOT re-verified at 400dpi, unlike the ingredient rows, so the nine numbers above are our reading of a scan pending a second look. The same caution applies to the comedolysis figures: we read that paper from the publisher's PDF text layer rather than from a visual render, and its prose and its Table I disagree about tretinoin's 12-week result, the prose giving 87 per cent where the table appears to print 20. We print only the 8-week 75 per cent, which both carry, and flag the discrepancy rather than choose between them. Comedolysis is not the inverse of comedogenicity. A preparation that removes existing microcomedones from a face that already has acne has not been shown incapable of provoking one on skin that does not. Different endpoint, different population, different starting condition. And the comedolysis preparations were formulations too, which cuts against the reassuring reading exactly as hard as the Hytone problem cuts against the alarming one: salicylic acid was tested in 85 per cent ethanol with 15 per cent propylene glycol, sulfur in a shake lotion, tretinoin as a branded liquid. This family is not evidence-free in the alarming direction either, and the counter-example is inside our own site. Mills and Kligman reported in 1972, in the British Journal of Dermatology, that suspensions and creams containing elemental sulphur "were shown to be capable of inducing comedones in the rabbit ear and on human backs", while adding that "sulphur was not strongly comedogenic". We have read the abstract of that paper and no more. It is a human comedogenicity observation about a member of this family, it points the way the checkers point, and any sentence on this page claiming the family has only reassuring evidence would be false. Three of the seven ratings lean on the American Academy of Dermatology's 2024 acne guideline. That is a graded treatment recommendation, not a comedone count, and it recommends benzoyl peroxide strongly while recommending salicylic acid only conditionally. A guideline that tells a clinician to put something on acne is a checkable proxy for the thing not causing acne. It is a proxy. Two members are untouched by everything above. Niacinamide and bakuchiol appear in none of the three documents this page rests on, and neither has a comedogenicity measurement we could find in any direction. Their pages are built on treatment trials and on an honest blank respectively, and this hub adds nothing to either. Finally, the generalisation itself. Seven ingredients filed together by function is a small and arbitrary set, and a family-level claim is exactly the kind of claim this site exists to distrust, including when we are the ones making it. The useful sentence here is narrow: for most of this family, the number you will meet is not a measurement of the material, and for one of them, colloidal sulfur, it is.
All 7 actives we rate
The legacy score is printed exactly as it appears in the paper, including the ranges, the asterisks and the parentheses. Those marks are not decoration. They are the paper telling you what its own number is worth.
Sources
- 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 survey nearly every circulating comedogenic score descends from - ingredient at roughly 10 per cent in propylene glycol, applied to the inner ear of three New Zealand albino rabbits, five days a week for two weeks, follicular keratosis graded 0-5. Held here transcribed verbatim from a 400dpi render of the journal scan, because OCR drops the numeric columns. Cited on this page overwhelmingly for what it does NOT contain: no row for any of the seven ingredients in this family, and three near-miss rows that look as though they might be, all on his 0-5 scale - octyl salicylate 0, vitamin A palmitate 1-3*, precipitated sulfur 0 and water-soluble sulfur 3. Its grade key, which almost no republisher prints, reads 0-1 not significant, 2-3 borderline, 4-5 positive.
- 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). PMID 6229554 The earlier and far less quoted rabbit-ear paper - two New Zealand albino rabbits per test, one ear dosed daily for two weeks with the other as control, materials used as they came unless asterisked for dilution. It is the source of everything specific on this page about products. Its Table I sets a therapeutic column beside cosmetics and moisturisers, and its key is its own: 1-2 not significant, 3 or above positive, which is NOT the 1989 key. Two disclosures. Its ingredient rows are transcribed here at 400dpi; the branded left-hand columns, which include the therapeutic agents quoted above, were transcribed from a standard-resolution scan and have not been re-verified. And it prints no formula or active concentration for any branded product, so none of its product grades can be assigned to an ingredient.
- Mills OH Jr, Kligman AM. "Assay of Comedolytic Activity in Acne Patients." Acta Dermatovener (Stockholm) 63:68-71 (1983). PMID 6191495 A short report, read in full from the journal's own PDF, and a HUMAN measurement - but of the opposite endpoint, and not the only human reading cited on this page. It measures comedoLYSIS, the removal of existing microcomedones, not comedone induction, so it cannot answer whether these materials can cause a comedone. Design: acne patients with at least 15 microcomedones per square centimetre, cyanoacrylate follicular biopsy, counts across five one-square-centimetre fields, panels of ten to twelve, cheeks and foreheads treated twice daily, read at 8 and 12 weeks. It tested finished preparations rather than raw ingredients. Cited for three things: that only tretinoin and salicylic acid of seven anti-acne preparations were judged comedolytic; that the authors used it to REVISE their own earlier rabbit-ear reading of benzoyl peroxide, writing that in the rabbit "comedolytic activity tends to be overrated" and that "the vehicle itself also becomes a factor"; and for the internal discrepancy noted in the limits above. Its volume year is 1983 and it was received in April 1982. It is a DIFFERENT paper from Mills and Kligman 1982 in Archives of Dermatology 118:903-905, which is also in this corpus, runs a 0-3 scale and carries human back data.
- Mills OH, Kligman AM. "Is sulphur helpful or harmful in acne vulgaris?" Br J Dermatol 86(6):620-627 (1972) Held at arm's length: we have read the publisher's abstract and no more, so methods, numbers and grading are not in hand, and it reports no score on any scale. It is cited on this page for one reason - it is the counter-example to a reassuring family-wide sentence we might otherwise have written. Its abstract states that suspensions and creams containing elemental sulphur "were shown to be capable of inducing comedones in the rabbit ear and on human backs", while also saying sulphur "was not strongly comedogenic". Note that this is a DIFFERENT paper from the Kligman and Mills 1972 study cited elsewhere on this site, which is in Archives of Dermatology 106:843-850 and runs a 0-3 scale.
- 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 that three of this family's seven ratings lean on. A work group grading published evidence: benzoyl peroxide carries a strong recommendation on moderate-certainty evidence, salicylic acid and azelaic acid conditional ones. It counts no comedones, measures no follicle and assays no ingredient. Cited as exactly what it is - a treatment recommendation standing in for a measurement that does not exist, which is the best available proxy and still a proxy.
- "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 rather than a new measurement, and independent of Fulton only in authorship, since the corpus it surveys includes him. Cited for two field-wide points that bear directly on this family: 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, which is the precise reason the 1984 therapeutic column cannot be read onto the drugs in it.