Retinol and bakuchiol in an anhydrous base: levels, limits and stability

The EU caps retinol at 0.3 percent for face and 0.05 for body. Retinyl palmitate potency, bakuchiol at 0.5 to 2 percent and the oxygen problem in a tin.

Retinol is the hardest cosmetic active to put into an anhydrous balm, and a wax balm in a wide tin is close to the worst vehicle anyone could pick for it. Bakuchiol is the opposite: oil soluble, heat tolerant, indifferent to light, and supported by a thinner evidence base than its marketing implies. This page sets out the EU limits that now apply, the stability arithmetic, and the point at which the honest answer is a different ingredient.

• EU: 0.3% RE, body lotion 0.05% RE
• Bakuchiol 0.5-2%
• Add retinol below 40-45 C
• Retinal and HPR not capped
• No retinoids on lips

The EU caps, and exactly what they cover

Commission Regulation (EU) 2024/996 added entry 376 to Annex III of Regulation (EC) 1223/2009, acting on Scientific Committee on Consumer Safety opinion SCCS/1639/21. Before that, retinol was unrestricted in EU cosmetics.

Two misreadings are common. The first is that 0.05 percent applies to anything used below the neck. It does not: the entry names body lotion, and the SCCS opinion behind it worked through hand cream and face cream at 0.3 percent, so a hand balm sits in the 0.3 percent bucket. The second is that the cap is about irritation. It is about total daily preformed vitamin A intake across diet and cosmetics, which is why the limit falls as the application area grows and why the label line talks about your daily intake rather than your skin.

Outside the EU the picture differs. The GB Cosmetics Regulation has not taken on entry 376, so at the time of writing there is no GB concentration limit and no mandatory vitamin A statement. That is a gap in the rulebook, not a licence: the general safety obligation still applies and your assessor may hold you to the SCCS numbers anyway. Check the current GB Annex III before you print labels, and read it alongside selling balms in the UK and EU. The United States sets no concentration cap at all, although the claim rules in labelling cosmetics in the US still bite.

Retinol equivalent: converting the esters

Retinol equivalent is a mass conversion by molecular weight, so an ester counts for less than its own weight. Do the arithmetic once: it changes what you may weigh out by nearly a factor of two.

Use more than one and their retinol equivalents add against the same ceiling. A second conversion catches people out: you do not buy pure retinol, it is sold diluted, usually at 10 or 50 percent in a carrier. To reach 0.3 percent retinol from a 10 percent solution you weigh 3.0 percent of the batch; from a 50 percent solution, 0.6 percent. Check the carrier before you buy, because the widely stocked 50 percent grade is dispersed in polysorbate 20, which bleeds and clouds in a fat base. For a balm you want retinol in caprylic/capric triglyceride, for the reasons in solubility in anhydrous formulas.

Four retinoids, ranked by what they cost you in stability

Every conversion step between what goes in the jar and retinoic acid at the receptor costs potency, and every step you remove costs stability. That trade is the whole story.

Retinaldehyde and hydroxypinacolone retinoate sit outside entry 376 only because the entry names three substances and they are not among them. Treat that as a drafting boundary rather than a safety verdict: an assessor writing your report under safety assessment and the CPSR may reason by analogy to the vitamin A opinion. Retinaldehyde has a further problem: it is bright orange, so it tints a pale balm, and it is the least stable of the lot. Hydroxypinacolone retinoate usually arrives dissolved in dimethyl isosorbide, a polar solvent not fully at home in a wax and butter base, so check miscibility on a small batch first.

The temperature problem a wax balm cannot solve

This is the section that decides most of the formulating. Retinol suppliers specify addition below 40 to 45 C. A balm containing 8 to 12 percent beeswax begins to build structure in the mid fifties and is a stiff, unpourable paste by 45 C. The windows do not overlap: you cannot follow the supplier instruction and pour a wax balm. There are three ways out and only one is good.

• Works

  Drop the wax

  Build the base from butters and oils only, set by cooling rather than by a wax network. A shea and cocoa butter base is still fluid at 45 C, so retinol goes in at 40 C and pours immediately.

• Compromise

  Change the retinoid

  Retinyl palmitate takes 60 to 70 C and hydroxypinacolone retinoate 60 C, both inside a normal pouring window. You lose potency with one and leave the regulated category with the other.

• Do not

  Add it hot anyway

  Retinol into wax at 65 C degrades an unknown fraction on contact. You cannot measure how much, so you no longer know what your label says.

The intermediate trick of cooling a wax balm to 45 C and pouring fast is worse than it sounds. The mix is already thickening, traps air on the way into the pot and sets with a mottled, dimpled surface, the faults described in air bubbles in balm. The pouring windows are in pour temperatures and the structural reason a wax-free base behaves differently is in anhydrous formulation. Bakuchiol tolerates the whole normal process, which is the real reason it appears in balms far more often than retinol.

What actually destroys a retinoid

Retinol is a conjugated polyene with an alcohol on the end. The chain of alternating double bonds is what makes it biologically interesting and what makes it fragile. Three routes matter.

Oxygen. Autoxidation at the allylic positions is the dominant loss route in an open pack, producing epoxides and fragments with no retinoid activity. Light. UV and short-wavelength visible light drive cis and trans isomerisation and photodegradation to anhydroretinol. Heat. Reaction rate roughly doubles per 10 C rise, the rule of thumb behind accelerated ageing and the reason a balm left in a car window is a different product by September.

Trace metals accelerate all of it. Iron and copper catalyse the same peroxidation chemistry described in rancidity and oxidation, which is a direct argument against combining a retinoid with the iron oxides and mineral pigments a tinted face balm would need.

Why an open tin is the worst possible pack

Packaging is not a finishing decision for a retinoid, it is a formulation decision, and the standard balm pack is the problem. A tin is opened, scooped with a finger, and closed over a headspace that has just been refreshed with air. Every gram removed is replaced by air, and the exposed surface stays constant as the product depletes.

An aluminium tin wins outright on light and loses outright on oxygen, so the ranking flips with whichever route dominates, and for a retinoid it is oxygen. A squeeze tube exposes about one eightieth of the surface of a 60 mm tin and never fully refreshes its headspace. An airless pump is better again, which is why commercial retinol products use one. Weigh those against filling difficulty and cost in the packaging guide, and check the closure and liner against your oil phase using packaging compatibility.

Bakuchiol: the ingredient, and what the trial actually showed

Bakuchiol (INCI Bakuchiol) is a meroterpene phenol from the seed of Psoralea corylifolia. It is not a retinoid: no vitamin A structure, no conversion to retinoic acid. The interest came from gene expression profiling published by Chaudhuri and Bojanowski in the International Journal of Cosmetic Science in 2014, which found it upregulated collagen and matrix genes in a pattern resembling retinol's.

As a raw material it is everything retinol is not: a viscous pale yellow to amber oil, fully oil soluble, stable across the pH range, tolerant of processing to around 80 C and far less light sensitive. Use it at 0.5 to 2 percent, with most published clinical work at 0.5 to 1 percent. It carries a herbal smell that becomes noticeable above about 1 percent.

The study everybody cites is Dhaliwal and colleagues in the British Journal of Dermatology in 2019: 44 patients, randomised and double blind, 12 weeks, bakuchiol 0.5 percent cream twice daily against retinol 0.5 percent cream once daily. Both arms significantly reduced wrinkle surface area and hyperpigmentation with no statistically significant difference between them, and the retinol arm reported more scaling and stinging.

One sourcing point matters. Psoralea corylifolia seed also contains furocoumarins including psoralen and isopsoralen, which are photosensitisers. Bakuchiol itself is not one, but crude babchi seed oil is a different material from purified bakuchiol and is not a substitute for it. Buy a grade specified at 98 percent or better with a stated furocoumarin limit on the certificate, and treat the photosensitivity question as the class of problem described in essential oil allergy and photosensitivity.

Why retinoids do not belong on lips

The vermilion has a stratum corneum a few cell layers thick against ten to twenty elsewhere on the face, almost no sebaceous glands across most of its surface, and a barrier that is repeatedly wetted and abraded. A retinoid concentration a cheek tolerates produces retinoid cheilitis there: scaling, fissuring, burning and cracked corners. That looks exactly like the chapping the balm was bought to fix, which drives reapplication, which makes it worse. That feedback loop is the mechanism described in does lip balm dry your lips, and the specific fault pattern is in cracked corners of the mouth.

The second reason is the one the regulation cares about: lip products are eaten. The cap exists to control total daily preformed vitamin A intake, and a lip balm bypasses the skin entirely for part of the dose. Take 50 mg a day as a working figure for lip product use. At 0.3 percent retinol that is 150 micrograms of retinol equivalent swallowed daily, roughly 5 percent of the 3,000 microgram per day tolerable upper intake level EFSA sets for preformed vitamin A in adults, from lip balm alone and before anything you eat. Nothing a lip balm does justifies spending that. Use the materials in ingredients for chapped lips instead. Bakuchiol is not dangerous there, but it tastes unpleasant and has no evidence for the site.

Irritation, pregnancy and what has to go on the label

Retinoid dermatitis is dose dependent, peaks in the first two to four weeks and largely settles with continued use, but a balm changes the exposure. An occlusive film holds the active against the skin and raises delivery, so 0.3 percent retinol in a heavy balm is not the same exposure as 0.3 percent in a light lotion, for the reasons in lipid penetration into skin. Start lower and less often than a cream would need, and when a customer reports stinging work through balm stings or burns first, because fragrance and menthol are commoner culprits.

Pregnancy is a place to be precise rather than reassuring. Oral high-dose vitamin A and oral isotretinoin are teratogenic. Absorption from topical cosmetic retinol is low and no causal link to birth defects has been established for it, but clinical guidance routinely advises avoiding topical retinoids in pregnancy as a precaution, and the SCCS reasoning behind the cap is about systemic vitamin A burden. A voluntary precautionary line on the pack is a reasonable commercial decision. Phrasing it as medical advice is not, and the boundary is in balms in pregnancy.

In the EU the label line is not optional: "Contains Vitamin A. Consider your daily intake before use", in the language of each market you sell into, alongside everything else required by labelling cosmetics in the EU. Claims are a separate discipline again. Appearance claims are cosmetic; treating acne, reversing photodamage or repairing sun damage are not, in any of these jurisdictions, and the line is drawn claim by claim in cosmetic vs drug claims.

Antioxidants, chelators and dating the batch honestly

Tocopherol at 0.1 to 0.5 percent is the usual vendor recommendation alongside a retinoid, above the dose that protects the base oils. Be clear what it buys: it slows peroxidation in the carrier, sparing the retinoid indirectly, rather than protecting the retinoid molecule itself. The measured dose response for the oils, including the point at which tocopherol turns pro-oxidant, is in vitamin E and antioxidants. BHT is more effective per unit weight and has been restricted in EU and GB Annex III since 2022 at a leave-on limit well above anything a balm needs; check the current entry and the comparison in synthetic antioxidants.

Chelators are where anhydrous formulating diverges from everything written for creams. Disodium EDTA needs a water phase to dissolve and to shuttle metal ions, and in a genuinely anhydrous balm it does neither: it sits there as undissolved powder. Metal exclusion is mechanical instead. Stainless steel or glass vessels, never copper or bare mild steel, no iron oxide pigments, and raw materials whose certificates report metal content.

Then date it honestly. You cannot assay retinol on a bench, so the number on your label is an assertion about what you weighed in, not about what the customer applies. If you sell on a stated percentage, the defensible route is an HPLC assay at time zero and again at the end of the intended shelf life, on product in its real pack: the framework is in shelf life testing and the traceability that makes a retest meaningful is in batch records. One asymmetry runs in your favour. Because the Annex III limit applies to the ready for use preparation, degradation only ever moves you further under the cap. It is the marketing claim that becomes untrue, not the legal limit that is breached.

The decision rule

Work down this list and stop at the first line that matches. If your base contains wax and pours above 55 C, use bakuchiol at 0.5 to 1 percent and do not attempt retinol. If the product goes in a tin the customer opens and scoops with a finger, use no retinoid at all, because the oxygen exposure makes the label figure meaningless within weeks. If it goes on lips, use no retinoid, for both reasons above.

If none of those apply, retinol in an anhydrous product is achievable and looks like this: a wax-free butter and oil base fluid at 40 C, a stabilised or encapsulated retinol in a triglyceride carrier, 0.3 percent retinol equivalent or less, tocopherol at 0.2 to 0.5 percent, a small opaque airless pack, the vitamin A statement, and a short stated period after opening. The base is the one used in barrier repair face balm, built without wax for unrelated reasons and suited to this one.

The honest limit on the other side is that bakuchiol's evidence base is one 44-patient trial without a vehicle arm plus supplier gene expression work. It is well tolerated, it survives your process, and it is probably doing something. It has not been shown to match retinol, and any product page saying otherwise is reading that trial wrong.

Reviewed and updated 6 September 2026. Spotted an error? Tell us and we will fix and log it.