Industries / Health / Medical devices / Medical sterilization
Radiation
Radiation sterilization sends gamma rays, electron beams or X-rays through sealed boxes of finished products, killing microbes without heat.
6
Milestones achieved, from 1956 to 2026.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1956
Sutures sterilized by electrons
Ethicon, Johnson & Johnson's suture maker, began passing packaged sutures under a beam of electrons from an accelerator, in what it believed was the first large commercial use of radiation. Sutures could now be sterilized already sealed in their final packets, and without heat they stayed 10 to 15 percent stronger, by the company history's account. By the end of 1958 almost all of Ethicon's sutures went through the beam.
Credit
Charles Artandi and his team at Ethicon, with High Voltage Engineering Corporation of Cambridge, Massachusetts
1959
A dose of 25 kGy
A kilogray (kGy) is the unit of absorbed radiation dose, one joule of energy per kilogram. After studying more than 150 kinds of microbes, Charles Artandi and Walton Van Winkle of Ethicon concluded that 25 kGy was 40% above the dose needed to kill the most resistant ones, and 25 kGy became the accepted dose almost everywhere for decades. Scandinavian health authorities instead tied the dose to the bioburden, the number of microbes on each item before treatment: 35 kGy below 50 microbes per item, rising to 50 kGy for 500 to 5,000.
LessonOne company's study set the world's number, and the Scandinavian approach of setting the dose from the bioburden is the one the international standard later adopted.
Credit
Charles Artandi and Walton Van Winkle; Scandinavian public health authorities
1964
Cobalt replaces the beam
Ethicon's early accelerators were unreliable, and once a division of Atomic Energy of Canada Limited (now Nordion) could make cobalt-60 in quantity, Ethicon switched to gamma rays from cobalt-60 and in 1964 built irradiators in Somerville, New Jersey, San Angelo, Texas, and Peterborough, Ontario. By the IAEA's figures, cobalt-60's gamma rays pass through about 300 mm of product, against about 38 mm for a 10 MeV electron beam, so whole cartons could be treated.
LessonReliability and penetration settled the first contest between machines and radioactive sources; in 2008 about 90% of radiation sterilization in the European Union still used gamma rays, by the IAEA's account.
Credit
Ethicon; Atomic Energy of Canada Limited
2006
Lower doses enter the standard
ISO 11137, the international standard for radiation sterilization, first published in 1995, was split into three parts in 2006. Part 2 kept two methods that set a product's dose from its bioburden, often below 25 kGy, and added VDmax, a method for confirming a dose of 25 kGy, or 15 kGy for products carrying fewer than 10 microbes. Fermilab's 2017 study shows the doses used at one Argentine sterilizer moving to the new lower options within a few years, and says later revisions added 17.5 and 20 kGy.
LessonWhen lower doses were allowed, makers took them up quickly, because a lower dose costs less.
Credit
International Organization for Standardization (ISO); the Association for the Advancement of Medical Instrumentation (AAMI), whose report TIR27 first set out the 25 kGy method
2010
An X-ray plant, little used
X-rays made by firing an electron beam at a metal plate pass through products as gamma rays do, but need no radioactive source. Leoni Studer built the first commercial plant designed for X-ray sterilization of medical devices at Däniken, Switzerland, in 2010, for 18 million euros. When Synergy Health bought it in 2012 the X-ray line ran at 22% of its capacity, against 75% for the gamma plant beside it, by evidence from a 2015 US court case quoted in Fermilab's study, which in 2017 still called it the only X-ray plant sterilizing medical devices.
LessonBuilding the plant was the easy part; device makers worried about how X-rays would change their products and were slow to take the risk of moving them.
Credit
Leoni Studer AG; Synergy Health
2026
X-ray lines multiply
Sterigenics opened an X-ray plant beside its gamma plant at Haw River, North Carolina, on 24 March 2026, as the trade journal Cleanroom Technology reported. STERIS says it completed X-ray processing at Chester, New York, in June 2026 and at Höchstadt, Germany, in July, and a second X-ray line at Venlo in the Netherlands in October. Moving products was made easier by AAMI's 2022 guidance TIR104 on transferring products between radiation sources, and by a 2023 FDA pilot that lets makers of high-risk (Class III) devices refer to their sterilizer's master file, a file the sterilizer gives the FDA, when they change radiation source; STERIS joined it in October 2023.
Credit
Sterigenics; STERIS; FDA; AAMI
Watch for
X-ray takes gamma's work
STERIS and Sterigenics present their new X-ray lines as added capacity and a backup for gamma and ethylene oxide. Nordion, the largest cobalt-60 supplier and, like Sterigenics, part of Sotera Health, said in January 2026 that accelerator-based irradiation "is facing significant challenges in deployment and adoption". In May 2024 STERIS wrote off an X-ray accelerator it had developed itself, as part of a restructuring, without giving a reason.
Achieved whensomeone outside the sterilizing companies publishes, from records rather than estimates, the share of devices sterilized by X-ray.
Claimed by
STERIS and Sterigenics, March to October 2026; disputed by Nordion, January 2026
Open
01
Dose-setting methods, rerun in public
An open program that carries out the published methods for setting a radiation sterilization dose (from a product's bioburden, from sterility tests at small incremental doses, and VDmax) and runs them on many thousands of made-up microbial populations whose resistance is known. It reports how often each method gives a dose too low to reach the target sterility assurance level and how often far too high, and lets anyone rerun the test for the lower doses of 15 to 20 kGy and for new resistance data. This is computer work only.
The briefCredit · unclaimed02
An open dose-map benchmark
A shared test set of published dose-mapping studies, in which a plant's dose was measured with dosimeters (small strips or vials that record the dose they absorb) and also computed with a Monte Carlo model, a simulation that follows individual photons. Each case would hold, in machine-readable form, the plant's geometry, source strength, product density and every dosimeter position with its reading, so any open simulation code can be scored against all of them. This is computer work only: no source, plant or dosimeter is touched.
The briefCredit · unclaimed
Open 01 · the brief
Dose-setting methods, rerun in public
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
An open program that carries out the published methods for setting a radiation sterilization dose (from a product's bioburden, from sterility tests at small incremental doses, and VDmax) and runs them on many thousands of made-up microbial populations whose resistance is known. It reports how often each method gives a dose too low to reach the target sterility assurance level and how often far too high, and lets anyone rerun the test for the lower doses of 15 to 20 kGy and for new resistance data. This is computer work only.
Why it is still open
The methods rest on models published from 1971 to 2000: Tallentire, Dwyer and Ley's model relating the share of contaminated items to rising dose (1971); Davis, Strawderman and Whitby's incremental-dose method (1984), which they checked by computer simulation and found to overdose and underdose some populations whose kill curves start with a "shoulder", a stretch of dose over which few microbes die; and Kowalski and Tallentire's VDmax (2000). The dose tables built from these models are printed in ISO 11137-2, a paid standard. Searches of GitHub and the literature found no open code that reproduces the methods or repeats the 1984 simulation with new data.
eutils.ncbi.nlm.nih.goveutils.ncbi.nlm.nih.goveutils.ncbi.nlm.nih.gov
Built with what exists
Python with NumPy and SciPy; the published papers for each model's formulas; published survival curves of device bioburden under radiation for realistic populations; an AI coding agent to turn each paper's formulas into code with tests, each checked by a person against the paper.
First step
Code Davis, Strawderman and Whitby's formula for the dose at a sterility level of one in a hundred, and rerun one of the populations they simulated.
Done looks like
The program reproduces Davis and colleagues' 1984 finding of which simulated populations were underdosed, and matches the worked examples in ISO 11137-2 to their stated rounding, as checked by someone who holds the standard. The share of 10,000 simulated populations underdosed at a sterility assurance level of one in a million is published for each method, with the code.
Open 02 · the brief
An open dose-map benchmark
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A shared test set of published dose-mapping studies, in which a plant's dose was measured with dosimeters (small strips or vials that record the dose they absorb) and also computed with a Monte Carlo model, a simulation that follows individual photons. Each case would hold, in machine-readable form, the plant's geometry, source strength, product density and every dosimeter position with its reading, so any open simulation code can be scored against all of them. This is computer work only: no source, plant or dosimeter is touched.
Why it is still open
Each study checks one code against one plant. Majer, Pasariček and Knežević modeled Croatia's cobalt-60 facility with the PHITS code in 2024 and matched chemical dosimeters within 6% inside the source rack; similar studies exist for plants in Tunisia, Bulgaria and Brazil. A search found no public set of the measured readings and no shared benchmark across plants.
Built with what exists
OpenMC and Geant4, open-source simulation codes that track photons; the published papers and their accepted versions in repositories such as FULIR; an AI model to pull geometry tables and dosimeter readings from PDFs, each checked by a person.
First step
Enter the Croatian paper's geometry and readings into a simple file format and try to reproduce its in-rack results with OpenMC.
Done looks like
Ten published cases, each run in two open codes, with the difference from every measured reading published. A new user can run any case with one command.