Mānuka honey on skin: what the trials actually measured

Published 6 August 2026Zane Hitchcox, editor

There is a specific reasoning error underneath most writing about honey and skin, worth naming before any evidence arrives. A material that helps a wound close is described as though that tells you what it does to skin that is not wounded. Those are different organs in the only sense that matters here: one has a barrier and one does not.

What mānuka honey is, and what makes it different

Mānuka honey is honey made by Apis mellifera from the nectar of Leptospermum scoparium, a shrub native to New Zealand and south-east Australia. Cochrane's own background description of honey is the useful one for skincare purposes: it is "a viscous, supersaturated sugar solution derived from nectar gathered and modified by the honeybee."¹ The word supersaturated is doing work. Honey holds more sugar than water at that temperature can normally dissolve, which gives it very low water activity — a strong osmotic pull and an environment in which most bacteria cannot grow. That property belongs to honey in general, not to mānuka in particular, and it is a physical effect on the material in contact with the honey rather than something the honey does to skin cells.

Most honeys also generate hydrogen peroxide. The bee adds the enzyme glucose oxidase, which oxidises glucose to gluconolactone and reduces oxygen to hydrogen peroxide once the honey is diluted; a 2020 review in Food Chemistry describes hydrogen peroxide as playing a key role in the antibacterial activity of honey generally, while noting that the amount produced varies widely between honeys and that a non-enzymatic route contributes as well.¹¹ Hydrogen peroxide is heat-sensitive and is degraded by catalase, which is present in tissue and in blood.

Mānuka's distinction is that a large part of its antibacterial activity survives catalase. In 2008 Mavric and colleagues at TU Dresden measured 1,2-dicarbonyl compounds in commercial honeys by HPLC and found methylglyoxal in six New Zealand mānuka samples at 38 to 761 mg/kg, up to a hundredfold higher than conventional honeys. Most of the other honeys they tested showed no inhibition of E. coli or S. aureus at 80% dilution or below; the mānuka samples inhibited growth diluted to 15–30%, which corresponded to methylglyoxal concentrations of 1.1–1.8 mM — matching the minimum inhibitory concentration of 1.1 mM they measured for pure methylglyoxal against both organisms.³ That is a clean piece of chemistry: the activity tracks the compound.

The methylglyoxal starts out as the molecule in fake tan

Methylglyoxal is not put there by the bee, and it is not present in fresh mānuka honey in any quantity. In 2009 Adams, Manley-Harris and Molan at the University of Waikato showed that fresh mānuka honey contains low methylglyoxal and high dihydroxyacetone, that dihydroxyacetone falls and methylglyoxal rises correspondingly when the honey is stored at 37 °C, and that adding dihydroxyacetone to clover honey produces comparable methylglyoxal. Nectar taken directly from mānuka flowers contained substantial dihydroxyacetone and no detectable methylglyoxal.⁴ The marker compound is a storage product of a nectar sugar.

Dihydroxyacetone is the same molecule that browns skin in every sunless tanning product on the market, where it reacts with amino acids in the outermost dead layer of the epidermis. In the United States it is a listed colour additive under 21 CFR 73.2150, permitted "in externally applied cosmetics intended solely or in part to impart a color to the human body". The listing sets no other boundary; two general sections do. 21 CFR 70.3(v) defines an externally applied cosmetic as one kept off the lips and off any surface covered by mucous membrane, and 21 CFR 70.5(a) withholds the eye area from any additive whose listing does not name it — which this one does not.²¹ This is a coincidence of chemistry rather than a warning: the concentrations differ by orders of magnitude, and in mature honey most of the dihydroxyacetone has already converted. It is worth knowing because it shows how far the "just a natural sugar" framing sits from the chemistry.

UMF and MGO grade the honey, not the skin claim

The two numbers on a jar come from the trade. MGO is a direct statement of methylglyoxal content in mg/kg. UMF — Unique Mānuka Factor — is a certification scheme run by the Unique Mānuka Factor Honey Association, which tests four compounds: methylglyoxal for potency, leptosperin as an authenticity marker that cannot be added, dihydroxyacetone as an indicator that the methylglyoxal level will hold over shelf life, and hydroxymethylfurfural as a check on overheating. The association publishes the correspondence between grades and methylglyoxal: UMF 10+ at 261 mg/kg, UMF 15+ at 512 mg/kg, UMF 20+ at 826 mg/kg.⁷ It is an industry scheme about honey composition, not a government standard, and it makes no claim about skin.

The government definition is a separate thing entirely, and it is more interesting than the marketing. Since 5 February 2018, all mānuka honey exported from New Zealand must be tested by a laboratory recognised by the Ministry for Primary Industries against a science definition with five attributes: four chemical markers and one DNA marker from mānuka pollen.⁵⁶ For a honey to be labelled monofloral mānuka it must show 3-phenyllactic acid at ≥400 mg/kg, 2'-methoxyacetophenone at ≥5 mg/kg, 2-methoxybenzoic acid at ≥1 mg/kg and 4-hydroxyphenyllactic acid at ≥1 mg/kg, plus a pollen DNA level below Cq 36. Multifloral mānuka is the same test with 3-phenyllactic acid between 20 and 400 mg/kg.⁵

Two things follow. First, methylglyoxal is not one of the five attributes. The compound the industry grades on and the compounds the state authenticates on are different sets; the government definition is a botanical-origin test, designed to stop non-mānuka honey being sold as mānuka, and it says nothing about potency. Second, the definition binds exports. In 2018 MPI consulted on whether to extend the same definition to mānuka honey sold on the domestic New Zealand market, stating plainly that it applied to exported honey and asking whether it should apply at home.⁶ A jar bought in Auckland and a jar shipped to London have not necessarily passed the same test.

The wound evidence, at the certainty Cochrane assigned it

The reference point is the Cochrane review by Jull and colleagues at the University of Auckland, updated in March 2015 as CD005083. It found 26 eligible randomised or quasi-randomised trials with 3,011 participants in total: three in minor acute wounds, eleven in burns, ten across various chronic wounds, and two in mixed populations.¹ This is honey in general; only some of the included trials used mānuka.

The finding usually quoted is the one for partial-thickness burns, and it is genuinely the strongest thing in the review: high-quality evidence from two trials with 992 participants that honey dressings heal partial-thickness burns more quickly than conventional dressings, by a weighted mean difference of 4.68 days (95% CI 5.09 to 4.28 days faster).¹ The comparator matters: the review spells out what "conventional treatment" meant in those trials — polyurethane film, paraffin gauze, soframycin-impregnated gauze, sterile linen, and leaving the burn exposed.

Everything else in the review is weaker, and the review says so. Burns treated with honey versus silver sulfadiazine: very low quality for time to healing, but high-quality evidence of no difference in the proportion healed within six weeks. Early excision and grafting beat honey in partial and full-thickness burns by 13.6 days, on low-quality evidence from a single 50-patient trial. Infected post-operative wounds: moderate quality, one trial, 50 patients. Pressure ulcers and Fournier's gangrene: very low quality, single small trials each. Venous leg ulcers, diabetic foot ulcers, minor acute wounds, cutaneous leishmaniasis and mixed chronic wounds: effects unclear.¹ The heterogeneity there is not a hedge in the write-up, it is the finding.

One trial in that literature deserves its own entry, because it is a New Zealand trial, it is properly sized, and it did not work. It randomised 368 people with venous leg ulcers to calcium alginate dressings impregnated with mānuka honey or to usual care, everyone under compression bandaging. At twelve weeks 55.6% of honey-treated ulcers and 49.7% of usual-care ulcers had healed — a difference of 5.9 percentage points with a confidence interval running from −4.3 to +15.7, and p = 0.258. Honey was probably more expensive and was associated with more adverse events, relative risk 1.3.²

Why a burn dressing is not a face mask

Every trial above studied skin whose barrier had been destroyed or breached. That is not a technicality. On a burn or an ulcer, honey sits in direct contact with living tissue, wound exudate and a bacterial population, and the mechanisms that make it useful — osmotic drawing of fluid, a moist wound environment, antibacterial activity in the wound bed, support of autolytic debridement — all operate on material the honey can physically reach.

On intact skin it cannot. The stratum corneum is a functioning physicochemical barrier, and the best-known heuristic for what crosses it is the 500-dalton rule proposed by Bos and Meinardi in Experimental Dermatology in 2000: compounds much above 500 Da are essentially excluded, which is why virtually all common contact allergens and all topical drugs used in transdermal delivery sit under that weight.¹⁷ Methylglyoxal, at about 72 Da, is small enough that the rule does not exclude it, and nor does it exclude glucose and fructose at about 180 Da. But "not excluded by molecular weight" is a very long way from "delivered at the concentration a trial measured in a wound bed", and nobody has published the measurement for honey on intact human skin. This constraint applies to every topical product, not just honey, and is set out at more length in what topical skincare can and cannot do.

There is also a regulatory tell. Products that carry the wound claims are not cosmetics anywhere. In the United States, mānuka honey wound dressings are cleared as medical devices through the FDA's 510(k) premarket notification route under product code FRO, "Dressing, Wound, Drug" — the Derma Sciences Medihoney dressings with active mānuka honey were cleared as K080315 on 23 April 2008, and the clearance rests on substantial equivalence to two predicate dressings rather than on a new efficacy trial.¹⁸ In New Zealand, anything supplied for a therapeutic purpose is a medical device under the Medicines Act 1981 and must be notified to Medsafe's WAND database — a notification, Medsafe is careful to say, that carries no assessment of quality, safety or efficacy.¹⁹ A honey face mask is regulated as a cosmetic precisely because it does not make those claims.

The skin evidence that does exist, with its numbers said out loud

For atopic dermatitis, the study everyone cites is Alangari and colleagues, published in Immunity, Inflammation and Disease in 2017. Adult volunteers with bilateral eczema lesions applied mānuka honey to one site overnight for seven consecutive nights and left the matching lesion on the other side untreated. Lesions improved significantly after treatment compared with the untreated control sites. The number of patients was fourteen. One result inside it is rarely quoted and cuts against its own premise: the study was built on mānuka honey's anti-staphylococcal activity, and skin swabs found no significant change in staphylococci by day 7, treated site or untreated.⁸ The authors themselves write that the finding needs to be confirmed by randomised, controlled clinical trials.

A 2019 review of alternative treatments for atopic dermatitis in the American Journal of Clinical Dermatology placed mānuka honey among several interventions with "positive clinical effects" in preliminary data and concluded there is not enough evidence to recommend any of them for atopic dermatitis, calling for replication at larger sample sizes in populations more representative of the general patient.⁹ That is the state of it: one fourteen-person pilot and a reviewer saying it is not enough.

The one properly powered trial of a New Zealand honey on a facial skin condition used a different species. Braithwaite and colleagues at the Medical Research Institute of New Zealand randomised 138 adults with rosacea across five primary-care sites to 90% medical-grade kānuka honey with 10% glycerine, or to a Cetomacrogol control cream, twice daily for eight weeks with blinded assessment of the primary outcome. A ≥2-point improvement on the seven-point Investigator Global Assessment was reached by 24/68 (34.3%) on the honey and 12/69 (17.4%) on the control, relative risk 2.03 (95% CI 1.11–3.72), p=0.020.¹⁰ That trial, its limits, and the less impressive companion work are read line by line in kānuka honey and rosacea. It is a different plant, Kunzea ericoides, and the result does not transfer to mānuka.

Safety: allergy, sterility, and the methylglyoxal question

Honey is not a common contact allergen, but it is a real one, and the best-documented sensitiser among bee products is not the honey itself. Propolis — the resin bees use to seal the hive — is a recognised sensitiser; DermNet's review notes European patch-test positivity of 1.2% to 6.6% and much higher rates in some paediatric eczema series.¹² A 2025 study in Contact Dermatitis patch-tested consecutive dermatitis patients at two European clinics and found 16 of 257 (6.2%) reacting to propolis in Genoa, though only 3 of 329 (0.9%) Swedish patients reacted to the Chinese propolis used there — the source of the propolis changed the answer.¹³ Anyone with a known bee-product allergy has an obvious reason to avoid honey-containing cosmetics.

Kitchen honey is not sterile, and the difference between it and a dressing is not marketing. A 2014 study in The Veterinary Journal, screening honeys for equine wound care, cultured 29 honey products — irradiated and non-irradiated medical honeys, supermarket honeys and beekeepers' honeys — and recovered aerobic bacteria or fungi from 18 of them, concluding that non-sterile honeys may not be suitable for wound treatment.¹⁴ Honey is a recognised vehicle for Clostridium botulinum spores; a 2020 case report in the International Journal of Infectious Diseases describes an infant who developed progressive flaccid paralysis days after topical medical-grade honey was applied to a debrided umbilical wound.²⁰ The practical point is narrow: broken skin is a doctor's territory, not a jar's.

Methylglyoxal is easy to overclaim in both directions. It is a reactive dicarbonyl and a precursor of advanced glycation end-products. In a 2005 study in Clinical Science, rats given methylglyoxal systemically at 50–75 mg/kg developed degenerative changes in cutaneous microvessels, impaired granulation tissue and slowed wound healing in the absence of high blood glucose.¹⁵ In cell culture, methylglyoxal is cytotoxic to human keratinocytes and causes measurable DNA strand breaks — which is precisely why it is used as the insult in protection experiments, as in a 2020 study in Dermatologic Therapy.¹⁶

None of that is evidence that mānuka honey harms skin. The rat study used systemic administration at doses unrelated to topical exposure, the cell work applies methylglyoxal directly to cells with no barrier in the way, and the wound trials — where honey is in direct contact with tissue for weeks — did not surface a glycation signal. The fair claim is narrow: the compound mānuka is graded on is a reactive glycating agent, nobody has measured how much of it crosses intact stratum corneum from a cosmetic, and nobody has run the long-term topical safety study that would settle it. The same uncertainty runs through most of the ingredients covered in New Zealand botanicals in skincare.

One label note. Honey appears in an ingredient list as Mel, its INCI name; mānuka essential oil — a steam-distilled leaf oil, a different material from the honey and not graded on methylglyoxal at all — appears as Leptospermum Scoparium Branch/Leaf Oil. Marketing copy confuses them constantly. Where either sits relative to the preservative tells you roughly how much is present, the trick set out in how to read an INCI list.