How a generic is actually made

Buying a finished medicine on the open market and working backwards from the tablet to a manufacturing process that produces the same thing. In the industry it is called deformulation.

The originator company spent twelve years and a great deal of money finding the molecule, proving it works and proving it is safe. None of that is what gets reverse engineered. The molecule's structure is public the day the patent publishes, and the clinical evidence is public in the label. What is not published is everything that turns a kilogram of powder into a tablet that releases the drug at the right rate, survives two years in a Dhaka summer and performs in a patient the same way the original does.

So deformulation starts with the one thing anybody can buy: a pack of the reference product. From that pack the team has to recover the qualitative and quantitative composition, the solid form of the drug substance, the impurity fingerprint, the dissolution behaviour, and enough about the manufacturing route to reproduce all of it on their own equipment. The output is not a copy of someone's batch record. It is a target product profile and a process of your own that hits it.

Because it is the whole business. Roughly 98% of the medicine used in Bangladesh is made in Bangladesh, and almost none of those molecules were discovered here.

The domestic market is around 3.5 billion US dollars, and local manufacturers supply about 98% of it. Exports reached 213.16 million dollars in FY25 across 166 countries, according to Export Promotion Bureau figures, more than double the 103.46 million of FY18. Every one of those products began as somebody else's molecule, read backwards.

About a fifth of the drugs made in Bangladesh are under patent somewhere else. That fifth is the high-value part: oncology, biologics, the newer diabetes and hepatitis agents. It exists on the shelves here only because of the waiver described above, and it is the part that becomes legally complicated the moment the waiver ends. The remaining four fifths are off-patent worldwide, where reverse engineering is simply what the generic industry does everywhere, in Dhaka exactly as in Hyderabad or New Jersey.

Because there is no patent to infringe. Bangladeshi law does not grant patents on pharmaceutical products at all, under a WTO waiver that runs for as long as the country stays a least developed country.

The chain is worth knowing precisely, because it is often reported loosely. Article 66.1 of the TRIPS Agreement lets least developed countries delay compliance with most of TRIPS. Paragraph 7 of the 2001 Doha Declaration on TRIPS and Public Health applied that specifically to pharmaceuticals. The TRIPS Council then extended the pharmaceutical exemption to 1 January 2033 by its decision of 6 November 2015 (IP/C/73), and the General Council waived the related mailbox and exclusive marketing rights obligations on the same day (WT/L/971). Bangladesh's own Patent Act 2023 carries this into domestic law: pharmaceutical and agrochemical products sit on the list of things that cannot be patented, and the Department of Patents, Designs and Trademarks has in practice stopped granting pharmaceutical product patents since 2008.

Two consequences follow that people routinely get backwards. First, the protection is not a Bangladeshi policy that Bangladesh can keep; it is a status that expires on graduation, not on a calendar date. Second, graduation, not 2033, is what ends it. Bangladesh was due to graduate on 24 November 2026. On 29 October 2026 the UN General Assembly adopted resolution A/81/L.8, extending the preparatory period by three years to 24 November 2029. So the window in the figure at the top of this page is open for three more years than it was a week ago, and it will still close more than three years before the waiver's own expiry date.

The four documents that make it lawful

Studying a product you bought is lawful almost everywhere. Selling the result into a market where the molecule is still patented is not. The line falls between learning and launching.

Trade secret law is explicit about this. In the United States, the Defend Trade Secrets Act states that reverse engineering and independent derivation are not improper means of acquiring a secret. The European Union's Trade Secrets Directive, 2016/943, says in Article 3(1)(b) that observing, studying, disassembling or testing a product you have lawfully acquired is a lawful way to obtain information about it. A formulation kept as a secret rather than patented is therefore fair game for the analyst, which is precisely why originators patent formulations as well as molecules.

Patent law goes one step further and protects the act of preparing a generic even while the patent is alive. In 1984 a United States court held in Roche v. Bolar that testing a patented drug to prepare a regulatory filing was infringement; Congress reversed that within months in the Hatch-Waxman Act, and the resulting provision, 35 U.S.C. 271(e)(1), is why a generic can be developed and filed before expiry and launched the day after. Most jurisdictions now have a version of this, usually called the Bolar exception. Bangladesh's Patent Act 2023 includes one, alongside exceptions for private and non-commercial acts, for experimental and research use, and for teaching.

Where reverse engineering does become unlawful is when the information does not come from the product. A former employee bringing a batch record, a contractor sharing a validation report, a stolen specification: none of that is reverse engineering, and all of it is actionable everywhere, waiver or no waiver. The legal safety of deformulation rests entirely on the information having been derived from a lawfully purchased pack.

Because the originator published it. A patent needs novelty and an inventive step, and a faithful copy of a disclosed molecule has neither, by design.

This is the bargain at the heart of the patent system and it is easy to miss. A patent is not a reward for secrecy; it is a trade. The inventor discloses the invention in enough detail that a skilled person could repeat it, and in exchange gets twenty years of exclusivity. Once that specification publishes, usually eighteen months after filing and long before the drug reaches a pharmacy, the molecule is permanently part of the prior art. Nobody can ever patent it again, including the originator. The exclusivity comes from the patent's twenty-year term, not from anyone being unable to work out what the compound is.

A successful deformulation is, in patent terms, an exercise in landing exactly where the prior art already sits. If your tablet is equivalent to the reference product, it is by definition not new. If your analytical work is good, it is still routine work with known instruments, which fails the inventive step. There is nothing to patent, and that is the intended outcome rather than a loophole.

What generic companies can and do patent is everything they had to invent along the way:

Almost never the chemistry of the molecule. The failures cluster in the solid form, the impurities, and the gap between a tablet that dissolves in a beaker and one that works in a patient.

What the regulators actually published

The two nitrosamine episodes are not industry folklore. Both regulators documented them as they went, and both files are worth reading before anyone writes an impurity specification.

Yes, and more than help: for the drug substance they have usually done it already. An Indian or Chinese API house sells the molecule, the solid form, the impurity standards and the paperwork as one package.

This is the part of the story that rarely gets told, and it explains why the formulation side of a generic project can look deceptively easy. By the time a Bangladeshi formulator starts work, somebody else has already reverse engineered the synthesis, chosen a polymorph that avoids the originator's form patent, characterised the impurities and filed a drug master file. What the formulator buys is not just powder.

• A drug master file or CEPA US Type II DMF, or a Certificate of Suitability from the EDQM, filed with the regulator by the supplier. It has an open part you receive and a closed part you never see, which is how the supplier shares regulatory standing without sharing its process.
• A named polymorphic formSpecified, controlled and usually chosen to sit outside whichever form the originator patented. This single choice decides much of your dissolution behaviour before you weigh anything.
• Impurity reference standardsIsolated, characterised samples of each known impurity, without which your HPLC method cannot prove it is seeing what the reference product contains.
• Physical specificationParticle size distribution, bulk and tapped density, flow, residual solvents. Change the particle size of the same molecule and you have changed the product.
• Methods and a batch historyValidated analytical methods, certificates of analysis, and stability data on the substance itself, which saves repeating work but does not substitute for your own.

Six or seven analytical techniques, each answering a different question, and a dossier that is mostly proof rather than discovery.

No single instrument deformulates anything. The craft is in knowing which question each one answers and where they disagree.

The documents are the other half, and in a regulated submission they are the larger half. The Directorate General of Drug Administration works from the same Common Technical Document structure as everyone else, so a file built properly for Dhaka is a file that can travel.

• Reference product, with provenancePurchased through a documented chain with the pack, batch number and expiry recorded. A deformulation against an unidentifiable sample is worthless.
• A patent and freedom-to-operate searchCompound, polymorph, formulation, process and use patents, for every market you intend to sell in, not only this one. For the United States the patents and exclusivities on a reference product are listed in the FDA Orange Book.
• Pharmacopoeial monographBP, USP or Ph. Eur. for the substance and, where it exists, the product. It sets the limits you are arguing against.
• The supplier's DMF or CEPWith the open part in your file and the closed part filed directly with the regulator, as FDA and EDQM each set out.
• Development reportWhat you tried, what failed, and why the final formula is the final formula. This is where reverse engineering becomes defensible science rather than guesswork.
• Stability data, zone IVbAccelerated and long-term, at 30 °C and 75% relative humidity for the local market, under ICH Q1A(R2).
• A bioequivalence studyAgainst the named reference product, or a documented biowaiver justification where the dissolution and BCS data support one.
• ValidationAnalytical methods under ICH Q2, cleaning, and three consecutive process validation batches at commercial scale.