How a patent works for a drug

The right to stop other people, for twenty years from the filing date, inside one country. It is a right to exclude. It is not permission to sell your own drug.

That distinction is the one that costs companies money, and it is worth sitting with for a moment. A patent is a negative right. It lets you go to court and stop somebody else making, using, selling or importing what your claims cover. It says nothing whatsoever about whether you are allowed to do those things yourself. You can hold a perfectly valid patent on a molecule and still be unable to sell it, because a competitor's earlier patent covers the class of molecules yours sits inside, or the formulation you need, or the process that is the only economic way to make it. Question four is about exactly that gap.

Three other properties follow from the same idea and explain most of what the system looks like in practice.

Utility, design and plant. A drug is a utility patent, and the category it falls into inside that is composition of matter.

The USPTO's own wording for a utility patent is worth knowing by heart, because every argument about what is patentable starts from it: anyone who invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof. Four categories. A drug substance is the fourth. The office defines composition of matter as covering chemical compositions, mixtures of ingredients and new chemical compounds, and gives a pharmaceutical drug as its example.

A real drug is not one patent. It is a family built in layers, filed over a decade or more, and the layers expire at different times. The compound patent is the valuable one and the first to go. What is filed afterwards, on a crystal form, a formulation or a new indication, is how an originator tries to extend the commercial life of the molecule, and it is why a generic company has to clear far more than one patent before it can launch.

Four gates, not three: eligible subject matter, novelty, non-obviousness, and utility. A fifth requirement sits in the document itself, and it fails more chemistry applications than people expect.

Novelty and non-obviousness are often spoken of together, but they fail in completely different ways, and knowing which one you are arguing about is half of patent practice.

Whether you can actually sell your product in a given country without infringing somebody else's live patent. It is a different question from whether your own invention is patentable, and your own patent is no answer to it.

Picture two patents. A competitor holds a granted claim to a broad class: benzimidazoles bearing a sulfinyl group. You invent one specific member of that class, and it turns out to be dramatically better than anything in the class. Your compound is new, it is non-obvious, it is useful. You will get your patent.

And you still cannot sell it. Making your compound means making a benzimidazole bearing a sulfinyl group, which is what their claim covers. Their patent blocks you; yours blocks them from using your specific compound. Neither of you can practise without the other. This is a blocking pair, it is extremely common in pharmaceutical chemistry, and it is ordinarily resolved by a cross-licence or by somebody paying somebody.

In the United States, almost always a provisional first, then a non-provisional within twelve months, then a PCT application within the same twelve months if you want other countries, then each country separately at about thirty months.

The provisional application is the most useful thing in the system and the least understood. It is filed under 35 U.S.C. 111(b). It needs a written description and drawings where necessary; it does not need claims, an oath, or a prior art statement. It is never examined. It is automatically abandoned twelve months after filing, and that twelve months cannot be extended. What it does in exchange is fix your priority date, let you say "patent pending", and start the Paris Convention year. Critically, those twelve months are not counted against the twenty-year term, so filing provisionally first can push the end of your patent out by up to a year.

It is also the point at which most inventors go wrong, in one specific way: a provisional only protects what it actually describes. If your non-provisional claims something the provisional never disclosed, that claim gets the later date, and anything published in between can be used against it. A thin provisional filed in a hurry gives a date for almost nothing.

Yourself, or an attorney?

You are allowed to file your own application, and the USPTO runs a programme specifically for inventors who do. For a provisional, written carefully and with real technical detail, doing it yourself is reasonable and many academics do. For the non-provisional it is a different task, because the value of a patent is almost entirely in the claims, and claim drafting is a specialist legal craft rather than a technical one. A claim drafted too narrowly is worthless because competitors step around it; too broadly and the prior art kills it. If the invention matters commercially, this is not the place to save money.

And then the rest of the world

The Patent Cooperation Treaty, concluded in 1970 and in force since 1978, does not grant anything. What it does is buy time and simplify paperwork. One international application, filed within twelve months of your priority date, holds your place in all of its contracting states, which now number 159. You get an international search report and a written opinion on patentability, which is a genuinely useful early read on whether the thing is worth pursuing. The application publishes at eighteen months from priority. Then, at around thirty months, you must enter the national phase in each country you actually want, pay each office, and translate where required. That is where the cost arrives, and it is why the PCT is best understood as an eighteen-month deferral of a very large bill rather than as a cheap global patent.

Most of the decision is eight yes-or-no questions, and the first one eliminates more applications than all the others together.

Work down this honestly, answering as an examiner would rather than as you would like. It is not a substitute for an attorney, and it cannot tell you whether an invention is commercially worth protecting. What it does do is tell you whether there is anything left to protect at all, which is the question people most often skip.

Questions people actually ask

Can I file a patent on an idea only?

No. This is the question that comes up more than any other and the answer does not bend. A patent is granted on an invention described well enough that a skilled person could make and use it without inventing anything further, which is what 35 U.S.C. 112 demands. An idea, a concept, a research plan or a hypothesis is none of those. What you can do is file as soon as you have something concrete, which is usually far earlier than people think: one worked example and a credible use is often enough for a provisional.

Does a provisional protect everything I claim later?

Only what it actually describes. Anything in your non-provisional that the provisional did not disclose gets the later date, and anything published in between can then be used against it. This is why a two-page provisional filed in a panic is close to worthless, and why adding data during the twelve months usually means filing a second provisional.

Is a departmental seminar really a public disclosure?

Usually yes. The test is whether a person interested in the subject could have attended and learned the invention, not whether anybody did. An open seminar, a poster session, a thesis placed in a library, a slide deck left on a public drive: all have been held to be disclosures. A talk given under a signed confidentiality agreement is not.

Do I even own my invention?

Often not. If it was made under a contract of employment, or using the employer's equipment, data or facilities, the rights usually vest in the employer; the Bangladesh Patent Act 2023 says so expressly, and most universities have the same rule. You remain the inventor and are named as such. The applicant, who owns the patent, may be somebody else entirely. Find out which you are before you file anything.

How much does it cost to find out if it is worth filing?

Nothing but time. The databases are free, the USPTO's tutorial is free, and the questions above are free. The money starts at the drafting, and by then you should already know the answer to everything in this flow.

Because publication is automatic and grant is not. An application publishes at eighteen months whether it is brilliant or hopeless, and you cannot sue anybody on it.

This is the part worth being blunt about, because it produces a lot of misplaced celebration. A laboratory files an application. Eighteen months later it appears on Google Patents and in the WIPO database with a number, a title, a figure and the inventors' names. It looks exactly like a patent. Press releases get written. CVs get updated. And nothing has been granted, nothing has been examined, and nobody has agreed that any of it is new.

There is one narrow exception in the United States, and it proves the rule rather than softening it. Under 35 U.S.C. 154(d), a patent owner can claim a reasonable royalty for the period between publication and grant, but only if the published claims and the granted claims are substantially identical, only if the infringer had actual notice of the publication, and only once the patent has actually granted. Everything still waits on the grant.

From a constitutional clause and a 1790 statute signed by George Washington. The American patent office is one of the oldest continuously operating in the world, and almost every rule above is a scar from some specific argument.

The United States wrote patents into its constitution before it had an office to issue them. The first Patent Act followed in 1790. The office that examines applications today handles more than half a million a year with over thirteen thousand employees, from Alexandria, Virginia, with regional offices in Detroit, Dallas, Denver and San Jose. A few dates are worth carrying around, because they explain why the modern system behaves as it does.

Behind, and in a specific way that has nothing to do with the quality of its scientists. Bangladesh is not a member of the Patent Cooperation Treaty, so a Bangladeshi inventor cannot use the route the rest of the world uses.

Bangladesh joined the Paris Convention on 3 March 1991, which means a Bangladeshi applicant can claim the twelve-month priority year. It has never joined the PCT. Practically, that means the whole middle row of the timeline at the top of this page is unavailable: no single international application, no international search report, no thirty-month breathing space. An inventor in Dhaka who wants protection in ten countries must file in ten countries, directly, within twelve months, paying ten sets of fees and ten sets of attorneys at once. For a university group or a small company that is not a hard decision, it is an impossible one, and the usual outcome is that nothing gets filed abroad at all.

The domestic system has in fairness just been rebuilt. The Bangladesh Patent Act 2023, Act No. 53 of 2023, came into force on 27 February 2025, replacing the Patents and Designs Act of 1911 and bringing a twenty-year term from filing, publication at eighteen months, and a requirement to request examination within thirty-six months or lose the application. A provisional specification is allowed, with the complete specification due within twelve months, which will look familiar from question five.

The numbers are the uncomfortable part.

None of this is about talent. It is about three things that can each be fixed: a treaty Bangladesh has not joined, an examination capacity that is thin enough that applications wait years, and a research culture in which almost nobody files before publishing. The third is the cheapest to change and the most within reach of a university department. A provisional specification filed at the DPDT before a thesis is defended costs very little and preserves an option that is otherwise destroyed the day the thesis goes on a shelf.