The Licences That Left: Withdrawal Rates Across the Korean Register

Every ingredient query against the Korean register returns two numbers, and most readings use only one of them. Polydeoxyribonucleotide returns 51 licences, of which 47 are valid. Retinol returns 246, of which 157 are valid. The second number is the one people quote and the gap between them is the one they skip, which is a shame, because that gap runs from zero to 36 percent depending on what you query and it is a far more interesting figure than it looks. It is also the figure most likely to be misread as a safety signal, which it is not.

Withdrawal rates across eleven queries, from zero to thirty-six percent

We swept eleven ingredient and category queries on 27 August 2026, recording total records and currently valid records for each, and derived the withdrawal rate as the difference. The spread is wide enough that any single figure quoted as typical would be wrong for most of the set.

At the high end, retinol returns 246 records of which only 157 remain valid, a withdrawal rate of 36.2 percent. Ascorbic acid sits at 27.2 percent, the tissue restoration device category at 24.8 percent, and nicotinamide at 24.8 percent. In the middle, lidocaine withdraws at 16.7 percent, tranexamic acid at 14.3 percent, thioctic acid at 13.3 percent. At the low end, glutathione withdraws at 11.8 percent and polydeoxyribonucleotide at 7.8 percent. Two queries return no withdrawals at all: hyaluronic acid in the drug register, and human placenta extract.

Before that spread gets interpreted, one thing has to be said plainly, because everything downstream depends on it. The register records that a licence is no longer valid. It does not record why. A licence holder who stopped renewing a product they no longer make, a company that voluntarily surrendered a licence during a portfolio consolidation, and a licence ended by administrative action all produce the same result in the same column. Nothing in the public record separates them, and no amount of careful reading will extract a reason that was never written down.

That single limitation disposes of the most common inference immediately. A withdrawn licence is not evidence that a product was found wanting, because the field that would tell you whether it was found wanting does not exist. Anyone presenting a withdrawal count as a safety observation has read a fact about administration as a fact about substances.

It is worth being concrete about how easily that inference forms, because it rarely arrives as an explicit claim. It arrives as an arrangement: a withdrawal count placed next to a competitor's product, or a valid count presented alone as though it were the whole record. Neither states anything false. Both invite a conclusion the data does not carry. The defence is procedural rather than interpretive, which is to write down what the field says and refuse to write down what it implies.

The spread tracks how old the category is

Sort that table by withdrawal rate and a pattern surfaces that has nothing to do with the substances themselves. The three highest rates belong to retinol, ascorbic acid and nicotinamide, which are among the oldest and most widely filed ingredients in the register, each carrying hundreds of records accumulated over decades of pharmaceutical practice. The lowest rates belong to polydeoxyribonucleotide, hyaluronic acid in the drug register, and human placenta extract, all of which are either recent arrivals or small specialised entries.

The mechanism is straightforward once stated. Withdrawal is cumulative and it takes time. A category that has been open for forty years has had forty years for licence holders to exit, reformulate, merge, or simply stop renewing paperwork on a product line that stopped selling. A category whose oldest injectable licence dates from 2008, as polydeoxyribonucleotide's does, has had a fraction of that exposure. The rate is measuring elapsed opportunity to leave, not quality of what stayed.

Human placenta extract makes the point in its cleanest form. Thirty-eight licences, not one withdrawn, and it would be absurd to read that as evidence of exceptional product quality. What it reflects is a small, stable field where the same holders have kept the same registrations current. The zero is a fact about turnover, and turnover is a fact about commercial behaviour.

Ascorbic acid deserves a footnote here, because the figures in this sweep and in an earlier one do not match exactly. We recorded 842 total records on 27 August 2026, while a piece published three days later worked from 840. The difference is two records out of eight hundred and is almost certainly a matter of query timing rather than a substantive change. We note it rather than quietly harmonising the numbers, because a register is a live dataset and pretending two snapshots are identical is how small discrepancies become invisible ones.

The middle of the table is where the age explanation runs out

Category age accounts for the two ends of the spread cleanly. It accounts for the middle less well, and saying so is more useful than stretching one explanation across everything.

Lidocaine sits at 16.7 percent with 263 records, and it is an old entry: its earliest licences date from 1979. On an age-only account it should sit near retinol and ascorbic acid, and it sits well below both. Thioctic acid at 13.3 percent and tranexamic acid at 14.3 percent are close to each other while differing enormously in size, 181 records against 35. Glutathione at 11.8 percent has entries going back to 1971 and a lower rate than categories half its age.

Several things could produce that and the register does not let us choose between them. A category dominated by a few large holders may simply renew more consistently than one with many small ones. Products that remain in routine clinical use generate renewal regardless of fashion, while consumer-facing formulations turn over with the market. Some entries may have absorbed reformulations under the same licence rather than filing new ones. All plausible, none visible in the data we have.

The disciplined version of the finding is therefore narrower than the pattern first suggests: age explains why the extremes sit where they do, and it does not explain the ordering within the middle. That is a real limit rather than a gap to be filled with a better sounding story. A reader who takes away only "older categories withdraw more" has taken away something true at the edges and unreliable in between, which is where most queries land.

The register keeps its withdrawals, and that is the useful part

A register could simply delete records when a licence ends, presenting only what is currently valid. The Korean register does not. Withdrawn entries stay in place and stay queryable, which is why every ingredient sweep returns two numbers instead of one. That design choice is easy to overlook and it is the reason most of this article is possible.

What it buys a buyer is history. A supplier says a product was licensed in Korea, the product is not in the current valid set, and the question becomes whether the claim was ever true. With withdrawals preserved, that question has an answer: the record is either there with an ended status, or it is not there at all. Those are very different findings, and a register that deleted its dead entries would collapse them into one indistinguishable absence.

It also makes the size of a category legible in a way a valid-only view would not. Polydeoxyribonucleotide's 51 total records against 47 valid ones says something about a young field with almost no turnover. A view showing only the 47 would look identical to a field that had churned through two hundred registrations and settled at 47, and those are not the same market. Total and valid read together describe a shape; either one alone describes a snapshot.

The limit on that history is worth stating alongside its value. Preserved records carry the status and the identifying fields, not a narrative. You can establish that a registration existed and ended. You cannot establish when it ended relative to a shipment, why it ended, or whether anything was recalled, because those are not fields in this dataset. The archive is real and it is thin, and treating it as thicker than it is produces confident answers to questions it never addressed.

What a withdrawal does and does not change for a buyer

The practical question is narrower than the interpretive one. If a product you source appears in the register with a withdrawn licence, what follows?

What follows immediately is that the licence is no longer current, and that is a fact with real consequences for import documentation and for any claim that a product is registered in Korea. Whatever the reason, a withdrawn licence cannot be cited as evidence of current registration, and a supplier presenting one as though it were is either working from stale records or hoping you will not check the status field. Checking that field is a two second operation once you have the record in front of you.

What does not follow is any conclusion about the material itself. The same molecule almost always continues under other valid licences held by other companies, which is visible directly in the table: retinol has 89 withdrawn records and 157 valid ones running alongside them. A withdrawal is a statement about one company's registration of one product, not about a substance. Reading it as the latter would mean concluding something about retinol from the fact that a manufacturer stopped renewing a cream in 2014.

There is a third case that catches people out, and it is the one worth building a habit around. A product can be genuinely unavailable while its licence remains perfectly valid, and a product can remain in circulation through existing stock after a licence lapses. Licence status and commercial availability are separate axes that happen to correlate loosely. If the question is "can I still buy this", the register is the wrong instrument, and if the question is "is this currently registered", it is the right one and the only one.

The asymmetry between those two axes runs in a particular direction that is worth internalising. A valid licence tells you very little about availability, since holders routinely keep registrations current on products they have stopped pushing. A withdrawn licence tells you rather more, because a company that has let a registration lapse has usually made a decision about that product line first. Neither is a reliable inference, but if you are going to lean on one, lean on the second and lean lightly.

Why a product leaves a catalogue, and why the register will not tell you

Distributor catalogues turn over constantly, and the most common question that generates is why something that was listed last season is not listed now. The register is the natural place to look for an answer and it is a poor source for this particular question.

The reasons a product leaves a catalogue mostly live outside any register. Distribution agreements end. Minimum order quantities change. A manufacturer restructures a line and replaces three products with two. Shipping economics shift for a temperature sensitive item. Demand for a presentation falls below the point where holding inventory makes sense. None of those events touches licence status, and all of them remove a product from a catalogue.

Meanwhile licence withdrawal, when it does occur, often lags the commercial reality by a long interval. A licence holder who stopped making a product in 2019 may leave the registration current until it comes up for renewal, so the register can show a valid licence for something nobody has manufactured in years. The two timelines are not synchronised and there is no reason to expect them to be.

So the honest position is that the register answers one question well and several adjacent questions badly. It will tell you whether a specific licence is currently valid, who holds it, what route and classification it carries, and when it was granted. It will not tell you why a licence ended, whether a product is still made, whether it is still shippable, or what a distributor's reason was for delisting it. Knowing which of your questions belong in which column saves a great deal of time spent interrogating a dataset that was never going to answer.

How to record a withdrawal in a sourcing note

The habit worth adopting is to record the withdrawal as a status with a date and to stop there, resisting the pull toward an explanatory sentence. A note reading "licence withdrawn, status confirmed 2026-08-27" is complete and durable. A note reading "licence withdrawn, presumably discontinued after quality issues" has invented the second half, and six months later nobody will remember which half was observed.

The date on the note is doing more work than it appears to. A register status is a reading taken at a moment, and this dataset changes underneath a cached answer without announcing it, which is exactly how the two record discrepancy in the ascorbic acid figures above arose. A note that says "withdrawn" with no date is a claim with no expiry attached, and it will be quoted a year later by someone who assumes it was checked recently. Writing the retrieval date beside the status costs nothing and is the single difference between a record and a rumour.

Recording the valid alternatives at the same time is the other half of the habit, because it converts a dead end into a working answer. If a licence under a given ingredient has lapsed, the same query almost always shows other valid licences under other holders, and that is the information a buyer actually needs. The question "is this exact registration current" and the question "is this ingredient still licensed in Korea" have different answers surprisingly often, and only the second one determines whether there is anything to source.

Where a brand spans several products, the check has to run per item rather than per brand, since licence status attaches to a registration and not to a name. That granularity problem is set out in one brand name, five products that share almost nothing. The broader shape of how Korea regulates and records these products, including who holds what and how classification works, is covered in inside Korea's MFDS regulation, and the specific case of licences that carry an export-only condition in what the export-only suffix means.

Across the KSTATION range we run this check against current records rather than cached ones, because a status field is only as good as the date attached to it. Products move between the skin boosters and ampoules and dermal fillers collections for commercial reasons that have no register counterpart, which is exactly the gap this piece is about.

Disclaimer. This article is general educational information and is not medical, legal or regulatory advice. It describes how withdrawal is recorded in the Korean drug and medical device licence registers, what the withdrawal rate measures across a sample of eleven queries, and which questions those registers can and cannot answer. The registers do not record a reason for withdrawal, and nothing in this article should be read as a statement about the safety, quality or performance of any substance or product. Figures are from a single sweep dated 2026-08-27 on a live dataset. Regulatory status varies by country and changes over time.