There's Vitamin K3 in a Toner We Scored 9.5. We Flagged It Anyway.

Menadione sits twentieth on a list of thirty ingredients. The product is a 400 ml hydrating toner from eple:me, it scores 9.5 out of 10 in our screen, and by most measures it is one of the gentlest things we sell. Fragrance-free. Alcohol-free. Twenty-eight of its thirty ingredients screen low risk.

It is also the only face product in our catalog carrying an ingredient we graded high.

We listed that grade last week, in the piece on the eight worst-graded ingredients in our own catalog. What we didn't do there was explain the chemistry, or answer the obvious follow-up: if you flagged it, why is the product still 9.5?

This is the answer. It takes a detour through a 1958 poultry-feed sanction.

The tail is the vitamin

Vitamin K is not one molecule. It is a shared naphthoquinone ring with different things bolted onto it, and the bolted-on part decides almost everything.

Phylloquinone (K1) is what plants make, and it is most of what you eat. The menaquinones (K2) are a family, built by gut bacteria and by fermentation, and the Linus Pauling Institute's micronutrient monograph describes them as differing by 1 to 13 repeats of a five-carbon unit in the side chain. MK-4 through MK-13. Same head, longer and longer tails.

Menadione has no tail at all. It is synthetic, and it is not a vitamer in the way the other two are; it is a provitamin that has to be alkylated before it can act as a cofactor for anything.

Which sounds like a knock on it, and mostly is, except that the alkylation genuinely happens. Hirota and colleagues showed in The Journal of Biological Chemistry in 2013 that menadione is not merely an industrial shortcut — it is an actual metabolic intermediate. Roughly 2% of absorbed phylloquinone gets its side chain cleaved off in the intestine, releasing menadione, which travels through the lymph and gets rebuilt in tissue by an enzyme called UBIAD1 into menaquinone-4. The work was done in rats, and the authors reported human time-concentration profiles that looked similar.

So menadione is a real waypoint in human vitamin K metabolism. That is the strongest thing anyone can say for it, and it is worth saying before the rest.

The one form small enough to get in

In 2000, Jan Bos and Marc Meinardi published a short argument in Experimental Dermatology that has held up better than most rules of thumb in this field. Compounds above roughly 500 daltons do not cross intact human skin. Their evidence was three-sided: virtually every common contact allergen is under 500 Da, every widely used topical dermatological drug is under 500 Da, and every compound anyone has managed to put into a transdermal patch is under 500 Da.

Now put the vitamin K family against that bar. K1 and the menaquinones carry twenty carbons or more of isoprenoid tail, and that tail is most of their mass. Menadione is the bare ring.

Strip the tail and you get a small molecule. Keep it and you get a large one. The only member of the vitamin K family light enough to plausibly clear the 500-dalton bar is the member that is not yet functional vitamin K.

We want to be careful here, because this is an inference from structure, not a measurement. We went looking for a percutaneous absorption study of menadione in human skin, the kind that would give an actual flux number, and we did not find one. If it exists, we missed it. What we have is a size argument with no data stacked on top of it, and we would rather say that than round it up.

The other name on the label

Ascorbic acid is the twenty-seventh ingredient in the same bottle.

Menadione and ascorbate have a relationship. Beck and colleagues laid it out in International Journal of Toxicology in 2009: ascorbate reduces the quinone to its semiquinone radical, molecular oxygen pulls the electron back off, and you are left with superoxide and then hydrogen peroxide, over and over, from catalytic amounts of the quinone. In their breast cancer cells the consequences were fast. Lactate formation fell 76% in three hours. Intracellular ATP dropped about 82% in one. The cells died by necrosis rather than apoptosis, with DNA damage and calcium dysregulation along the way.

And we cannot close the gap, because we do not know either concentration. Korean labels do not print percentages for non-functional ingredients, this one does not either, and position on an INCI list only tells you so much once you are past the 1% line. Menadione is twentieth of thirty. Ascorbic acid is twenty-seventh. Both are almost certainly present in trace amounts. "Almost certainly" is the honest limit of what we can say.

Vitamin K3 is legal in your dog's food

Here the regulatory picture goes sideways, and the clearest way to show it is to compare a cosmetic shelf with a feed mill.

The FDA maintains a page on vitamin K substances in animal feed, last updated 24 October 2024. Menadione and menadione sodium bisulfite complex are prior-sanctioned for poultry, with a history of safe use predating 1958. Menadione dimethylpyrimidinol bisulfite is a permitted food additive for chicken and turkey up to 2 grams per ton and for swine up to 10 grams per ton. The agency's own summary says menadione "can be added up to levels as high as 1,000 times the dietary requirement without adverse effects" — with an exception carved out for horses.

For humans the story runs the other way. The Linus Pauling Institute's monograph, last revised in May 2022, notes that injected menadione caused liver toxicity, jaundice and hemolytic anemia in infants, and that menadione is no longer used to treat vitamin K deficiency at all. It does not appear in supplements. K1 and K2 do.

Cosmetics fall into the gap. The EU banned phytonadione (vitamin K1, the safe one) from cosmetic products outright, which we covered in the worst-grades piece. Nobody has banned K3. As far as we can find, no regulator anywhere restricts menadione in a leave-on cosmetic.

That is not a clearance. Under MoCRA the FDA does not pre-approve cosmetic ingredients at all, a point we leaned on in our benchmark write-up. An absent rule usually means nobody filed the paperwork, not that somebody checked. We also still have not read Korea's MFDS list on this ingredient. Two runs at the primary text, two failures. So we are not going to tell you what Korea thinks.

So why is it still 9.5?

Because our screen counts ingredients, and thirty ingredients with one high and one mid is still a mostly clean list. Disodium EDTA is the mid, at position nine. Everything else screens low: eight vitamins, eight forms of hyaluronic acid, niacinamide, panthenol, allantoin. The arithmetic is not wrong. It just does not know that one entry deserves more weight than the twenty-eight around it.

Then we went to check our own copy, and found something worse.

The product write-up for this toner says, in our own words, that it screens 9.7 out of 10. The score field on the same product says 9.5. A third field, our curator score, says 9. Three numbers, one product, all written by us.

We flagged this pattern for a single product, a hair dye, in the benchmark piece. It is not a single product.

We opened nine product write-ups. Five of them state a score in the sentence. Four of those five disagree with the field. The only one that matches is the hair dye, which is the lowest-scoring product we stock and the one where the copy was written defensively enough that somebody checked it.

We have not audited all fifty-four. We will, and we will publish the count.

What our screening found

The practical takeaway

Read the whole vitamin list, not the front of the bottle. A label that says "vitamin complex" can mean seven B vitamins and a naphthoquinone that got quietly retired from human medicine in the twentieth century, and both of those things are true of this toner at once.

Check for K3 specifically. It appears as Menadione, and it is rare — one product in fifty-four, in our catalog. If you see it beside ascorbic acid, that is not a reason to panic, but it is a reason to patch test rather than to layer it under an acid the same night.

And treat a missing regulation as missing information. We give scores to everything we stock, and the scores are useful, and they are also a tally that can hand a 9.5 to a product with a high-risk entry in it. That is why we publish the flag next to the number instead of only the number. If we ever stop doing that, stop trusting the number.