DATA JOURNALISM

Patents, prices and court files: How ICIJ used data to investigate an industry that thrives on secrecy

For the Cancer Calculus, ICIJ’s data team created datasets using records from patent offices and courts, and analyzed the opaque pricing of a lifesaving drug with a sky-high price tag.

By Denise Ajiri, Agustin Armendariz, Jelena Cosic, Jesús Escudero, Miguel Fiandor Gutiérrez, Karrie Kehoe and Delphine Reuter Image: Chelsea Conrad / ICIJ May 29, 2026

Drug patents are meant to help pharmaceutical companies recoup high development costs by preventing competitors from using the intellectual property for a defined period of time, typically 20 years in the U.S.

But the global patent system — a patchwork of national laws loosely connected by international treaties — is vulnerable to manipulation. In the case of Keytruda, a blockbuster cancer drug, companies exploited the patent system to try to extend market exclusivity well beyond the expiration of the drug’s initial patents, keeping competitors at bay and prices artificially high for years. Prolonged patent monopolies can delay cheaper alternatives entering the marketplace, prioritizing profit over patient access, straining governments’ healthcare budgets and putting patients’ health — sometimes even their lives — at risk.

For its Cancer Calculus project, the International Consortium of Investigative Journalists tracked Keytruda-related patents to show how Merck & Co. and other pharmaceutical companies created a dense web of patent applications that can make it harder for more affordable versions of the drug, known as biosimilars, to enter markets around the world. Merck, known as MSD outside the U.S. and Canada, did this by applying for patents for changes to formulation and dosing regimens, altering the drug’s use in combination with other agents, or for switching patients to a similar, newer version of the same drug — known as a “product hop.” Each change can potentially reset the patent clock and add years of exclusivity.

Merck’s scramble to fortify its dominance has included filing for patents that are combinations of Keytruda and another medication that aren’t necessarily new or innovative, according to experts interviewed by ICIJ.

Even if a patent isn’t ultimately approved by a patent office, the application itself can increase the complexity of the competitive landscape, creating legal and commercial uncertainty that can delay or deter competitors, patent experts said.

Patents were only part of the data that explains Keytruda’s price dominance and patients’ struggles to cope with it. ICIJ also reviewed the prices of Keytruda (known generically as pembrolizumab) across dozens of countries. Those prices can vary wildly depending on location and medical context — the result of opaque negotiations between governments and Merck. We also reviewed lawsuits and other court documents filed in Latin America to track the rising number of patients fighting in court, regulatory bodies and elsewhere to gain access to Keytruda, a trend due, in part, to its high prices. Researchers in the region see the phenomenon as part of an increasing judicialization of healthcare.

The patent thicket

ICIJ anchored its analysis on patent applications in the U.S., which accounts for 60% of Keytruda sales globally. The Initiative for Medicines, Access, and Knowledge (I-MAK), a U.S.-based, not-for-profit organization that advocates for affordable access to medicines, provided our starting dataset of 184 U.S. patent applications related to Keytruda. After speaking with patent lawyers and pharmaceutical industry experts to refine our methodology, we reviewed each patent application to confirm key details, including legal status, patent owners, known as assignees, and relevant dates using Google Patents, a free public search platform that aggregates patent information from major patent offices worldwide.

We limited the final set to 180 U.S. patent applications (166 filed or co-filed by Merck & Co. and 14 by Ono Pharmaceutical Co., Ltd. a Japanese company we included because its PD-1 patents underpin Keytruda’s core mechanism. (The patents involve the use of what are known as PD-1-blocking antibodies that restore the immune system’s ability to recognize and attack tumor cells; Keytruda was developed using PD-1-blocking antibody technology.) After a legal dispute, Merck bought licenses to Ono patents as part of an interlocking patent structure. Our final set excluded four patent applications not assigned to either Merck or Ono Pharmaceutical.

For each U.S. application, ICIJ then tracked its so-called patent family — a group of patent applications from around the world that cover the same or closely related content, which can include patents filed or co-filed by Merck or MSD and other cancer research businesses. For companies like Merck, the interconnectedness of patents in such families allows them to extend protection around a single drug by filing new applications related to the original patent over time and across markets that can complicate competitors’ decisions about whether to enter a market.

To conduct our analysis, we scraped relevant records from two main websites: Espacenet, a patent search platform developed by the European Patent Office, and Google Patents. Espacenet constructs patent families using an automated system based on so-called shared priority claims, which links a later application to the filing date of an earlier one.

The addition of patent families brought the total to 1,212 global patent applications, including the original 180 U.S. filings. Not included were 129 Patent Cooperation Treaty applications we identified as part of the patent families. Submitted through the World Intellectual Property Organization, these are international applications that do not themselves result in granted patents but instead serve as a unified filing that allows applicants to pursue protection in multiple national or regional jurisdictions. We also excluded nine filings with the Eurasian Patent Organization — which represents Russia and seven former Soviet republics — as they duplicated European Patent Office applications already in our count. We did retain the 134 European Patent Office applications that we found represented substantive regional filings that, once granted and validated, could confer enforceable patent rights across multiple European jurisdictions.

We updated the details of each patent family, including the status, assignees and dates, using Google Patents.

ICIJ relied on both current and original patent assignees, depending on the context, when analyzing how many patents were filed and by whom. Particularly for non-U.S. patent applications, the current assignee or co-assignee is not always Merck & Co. but can also be other businesses involved in cancer research. The European Patent Office has identified these patents as part of Merck patent families. So while they’re not Merck patents, they are connected to Keytruda. For this reason, we refer to these as “Keytruda-related” patents rather than attributing them solely to Merck.

Of the total 1,212 identified applications, most were assigned to Merck as of early 2026 — sometimes with co-applicants: 590, including subsidiaries or companies later acquired by Merck; 44 assigned to Ono Pharmaceutical; 45 assigned to other entities not affiliated with Merck; and 533 listed with no identified current assignee. Of the 533, 455 had originally been filed or co-filed by Merck (or by subsidiaries or companies later acquired by Merck); 14 were filed by Ono; and 34 were not related to Merck. ICIJ couldn’t determine the assignee for 30 of the applications. All applications included in ICIJ’s dataset are part of patent families related to Keytruda.

ICIJ relied on the date that an application was filed rather than the publication date to reflect when inventions were first formally claimed, which was most relevant to our analysis.

ICIJ included patents across all relevant legal statuses, including 211 granted, 337 pending, 120 abandoned, 24 ceased, 41 expired, six revoked, 75 withdrawn, and 398 whose status we couldn’t determine, to capture the full global landscape of patents related to Keytruda that fall within patent families identified by the European Patent Office. Including all statuses allows ICIJ to capture not only enforceable rights, but also the broader ecosystem shaping access and competition. Pending applications may, if granted, translate into enforceable rights with defined expiration dates. Abandoned applications, while no longer pursued, can still be used as evidence to restrict what others can patent.

While this analysis focuses on Keytruda, similar patenting strategies are common across the pharmaceutical industry. As such, the dynamics highlighted in this dataset reflect broader structural features of the global patent system, which is administered through national and regional offices, such as the U.S. Patent and Trademark Office and the European Patent Office, and linked through international frameworks like the World Intellectual Property Organization. A published patent application or successful defense in major markets can deter competitors from entering other countries where there are patents for the same invention.

 

Patent data tools

ICIJ’s use of Espacenet and Google Patents as patent data sources presented different challenges. While Espacenet blocks automated compilation of patents data, which made it difficult to extract and use for analysis, Google Patents data can be compiled using Google Big Query service. But compiling bigger datasets from there can be costly, so we confirmed that Google Patents allowed us to retrieve information about our target list of patents in an automated and careful way. We also collected some data manually to populate our analysis spreadsheet before fact-checking.

Both sources returned hard-to-read webpage content that is tricky to transform into a structured format, a task made even more difficult by the many properties a patent can contain. At this stage, ICIJ used AI large language models to generate code in the easy-to-read Python language, which used popular Python libraries (pre-existing collections of code) to create parsers that transformed the content we extracted into a single, structured spreadsheet. Two of the Python libraries were Beautiful Soup, which selects the pieces from messy HTML, and pandas, which is used for data analysis.  We then used this dataset for the patent analysis.

Uncloaking secrecy around prices