The drug industry published the grammar for classical names

Every naming guide recommends Latin and Greek roots. None of them mention that one industry got there first, worked out the rules, and published them.

The World Health Organization's stem book runs to 220 pages of classical morphemes, each assigned to a class of drug. -tinib is a tyrosine kinase inhibitor. -prazole is an antiulcer benzimidazole. -sartan is for your blood pressure. Some 9,824 substances have been given one of these names, and the number grows by 200 to 240 a year.

That book is why your Latin name sounds like medicine. It is also the only naming standard anybody has published, and every rule in it runs opposite to yours.

Two rules, both backwards

The stem book states the central requirement plainly:

"Whenever possible, an INN should include the 'common stem' expressing the pharmacologically-related group to which the substance belongs. Names that are likely to convey an anatomical, physiological, pathological or therapeutic suggestion are avoided."

The first sentence requires a drug to share a morpheme with its direct competitors — the naming equivalent of a mandate that every project tracker be called something-ly. The second forbids the name from hinting at what the drug does for you.

Your brief asks for the reverse on both counts: sound like nobody else, and say what you are for. An entire global industry has spent seventy years naming against that instinct, and it did so for a reason worth borrowing even if the rule is not. A prescriber needs to know instantly that two unfamiliar words belong to the same family, and needs never to be told by a name that a drug will work. Category legibility beats differentiation when the cost of confusion is measured in patients.

What the book does want, it wants in the letters it has left over. Names "should be distinctive in sound and spelling," and "should not be inconveniently long and not be liable to confusion with names in common use." That is your brief exactly, applied to the two syllables at the front.

The slot marked Random

The scheme for monoclonal antibodies sets out a drug name as three parts in a table. An infix encodes the target: -ta- for tumour, -ci- for cardiovascular, -ne- for neural, -vi- for viral, -os- for bone. A suffix encodes the class. The heading on the first column, the part where the naming actually happens, is a single word.

Random.

That is WHO's own label, printed in its own table. In the most rigorous naming system anybody has built, the one creative decision has been reduced to a slot and formally designated as arbitrary. Raxibacumab is the book's example of an antibacterial: raxi is the part somebody chose, and everything after it is grammar.

You can read that as a warning or as permission. It is worth noticing that the system works, and that the arbitrary syllable at the front is doing exactly the job an invented name does anywhere else, which is telling one thing apart from another, while the meaning is carried by parts the reader has already learned.

They ran out

The 2018 edition lists 505 substances whose names end in -mab. Three letters, and a population the size of a small language's vocabulary. In October 2021, the committee gave up:

"At the 73rd INN Consultation in October 2021, the INN Expert Group decided to discontinue the use of the stem -mab and to divide this group of substances into four different groups, with four brand-new stems, to avoid any confusion between the monoclonal antibodies named according to the old nomenclature and those named according to the new nomenclature."

The replacements are -tug for unmodified immunoglobulins, -bart for artificial ones, -ment for fragments, and -mig for the multi-specific. The most recognizable ending in pharmaceutical naming was retired because too many things had been named with it to tell them apart.

Any morpheme you pick has a population, and the useful ones are crowded for the same reason they are useful. The difference is that pharma counted.

The reservation is official

This is where the collision stops being an aesthetic problem. From the stem book's introduction:

"In 1993, the World Health Assembly endorsed resolution WHA46.19 which states that trade marks should not be derived from INNs and INN stems should not be used in trade marks. The Assembly reasoned that such practice could frustrate the rational selection of INNs and ultimately compromise the safety of patients by promoting confusion in drug nomenclature. Above all, INNs are protected for use in the public domain."

Be clear about what that is and is not. It is a resolution addressed to member states, not a statute, and no trademark office is going to refuse your application because your logistics company ended up with a stem in it. What it tells you is that these particular morphemes were taken deliberately, that the taking was ratified by health ministries, and that the reason given was safety. Borrowing from the same shelf is not forbidden. It is documented.

Check it against the stem book

The list is public, so a name can be checked against all of it.

The stem book · WHO 2018

Every morpheme the World Health Organization has reserved for naming drugs — 472 spellings, of which 311 occur in no English word — checked against a name of your own.

Candidate name

1 reserved stem

• -vastatin

  antihyperlipidaemic substances, HMG CoA reductase inhibitors

  Occurs in none of the 68,510 English words checked · the book files it under -stat

1 also ordinary English

• -stat-

  enzyme inhibitors

  Also in 91 English words, so this is probably the language rather than the stem book

What this cannot tell you is whether anyone hears it. That depends on which drugs your reader has met, and no list holds that.

Two readouts, because a match on its own proves nothing. Plenty of stems are also ordinary English endings: -al is the WHO's stem for aldehydes and it ends 2,140 English words, so a name finishing in those two letters has borrowed precisely nothing. Of the 472 spellings the book publishes, 311 occur in no English word at all — and those are the ones where a hit means your name really did reach into the pharmacopoeia.

One trap in that measurement is worth naming, because you will meet it in any tool that claims to check a name against a list. Count stems against an ordinary dictionary and -prazole and -vastatin come back as ordinary English — on the strength of omeprazole and atorvastatin, which are in the dictionary. A measure built that way accepts a stem's own children as proof that the stem was never borrowed. Take out the 6,487 drug names the same book lists, and the 68,510 words left over give a count that means what it says.

What it will not tell you is whether anyone hears it. That depends on which drugs your reader has met, and no list holds that.

What a dead language actually buys you

Classical roots buy one genuine legal advantage, and it comes from Latin being dead.

The general rule is unforgiving. The examiners' manual states that "the foreign equivalent of merely descriptive or generic English wording is no more registrable than the English wording itself," and the Federal Circuit applied that in 2025 to hold VETEMENTS generic for clothing. Call your bakery Le Pain and the office translates it before deciding. The doctrine reaches words from common, modern languages, and it applies only where the ordinary American purchaser would "stop and translate." The Federal Circuit found that American buyers would not translate veuve into widow, and it was talking about French.

Latin sits outside it. The manual says the doctrine "will generally not be applied to obscure languages," that whether a language is dead or obscure "is made on a case-by-case basis, based upon the meaning the word or term would have to the ordinary American purchaser," and then, directly: "Latin is generally considered a dead language." A Latin word for what you sell can be registrable where its English twin would be refused.

Now the sentence that removes most of the benefit, from the same paragraph: "if there is evidence that a Latin term is still in use by the ordinary American purchaser (such as the term appearing in news articles), then that Latin term would not be considered dead." The roots founders reach for first are exactly the ones still in circulation: vita, nova, terra, lux, veritas. The carve-out is real, and it is narrowest at the end you were heading for. The deeper into the language you go, the better the protection and the fewer people who know what you meant.

Pick the root for its sound

Treat the meaning as a bonus. Nobody translates your name. What survives contact with a customer is how it sounds and what English words it reminds them of, which is the subject of the page on how names sound, and neither of those is the Latin.

Check the ending before you fall in love with it. Two minutes, and the alternative is finding out from a customer.

Take the far root over the near one. Lux is in the dictionary; the obscure synonym is not. That is worth something legally, and it is worth more in a search box.

Do not build the whole name out of parts. Root plus root plus classical suffix is the construction the stem book formalized, and it produces the sound it was designed to produce. One borrowed element carried by one arbitrary one is the shape that reads as a company.

Notice which rule you are actually breaking. If your name shares a morpheme with six competitors, you have not made a classical name. You have made a generic one with better manners.