ISO 10993-1:2025 frames biocompatibility as an evaluation, not a testing exercise — yet in roughly nine of every ten programs we see, FDA or a Notified Body still wants test data. A well-built Biological Evaluation Plan establishes how close a device sits to the conditions under which testing can be avoided or minimized, and where the data will still be needed and provides a wealth of information supporting the safety of a device.
ISO 10993-1 is now in its sixth edition, and with it a familiar line is back on many development programs: biocompatibility “isn’t a testing exercise; it’s an evaluation.” Strictly speaking, that’s true and correct. The standard has framed biological safety as an evaluation within a risk management process for several editions now, and the 2025 revision states that more plainly than ever before.
But in roughly nine of every ten programs we see at Groenakker, the FDA or a Notified Body still wants test data of some variety. So, the better question isn’t whether biocompatibility is an evaluation or a testing exercise, it’s how close a particular device actually sits to the conditions under which testing can be avoided or minimized.
FDA has gone some way toward spelling those conditions out. At a September 2026 town hall on the new edition of ISO 10993-1, the agency walked through hypothetical cases where a risk assessment may be acceptable instead of testing: a bare-metal stent of consensus-standard alloy, a PEEK implant made to its Master File without manufacturing chemicals, a processing change followed by a validated cleaning step, and a change of injection molding equipment. Add a coating, a drug, or a novel alloy to that same stent, and by the FDA’s own account, chemistry testing would likely be needed again.
The common thread is a well-characterized material, or a controlled change to an established device with proof that there is no manufacturing impact. Most of the Biological Evaluation Plan (BEP) work is establishing how far a given device sits from that narrow class, or if it’s well within it.
That isn’t how the plan usually gets approached. Programs often arrive with a test list already in hand — or the full expectation that there will be no test list. But the tests are only the visible and most costly/time consuming end of biocompatibility. The discipline is the evaluation framework around them, set down in the BEP, and any test matrix should follow from that, not the other way around.
The 2025 Edition and FDA’s Recognition
What Changed
The 2025 edition, published in November 2025, drops the word “testing” from its title and frames evaluation as a risk-management activity aligned with ISO 14971. Foreseeable misuse and the device life cycle are now in scope, and genotoxicity now extends to prolonged contact with all tissues other than intact skin. The most practical change, though, is how exposure gets counted:
Any contact within a day now counts as a contact day.
Intermittent days are summed.
Systematic use beyond the labeled duration is treated as foreseeable misuse.
A constituent known to bioaccumulate puts a device in the long-term category unless otherwise justified.
How One Extra Day Changes the Program
Follow that rule through and there’s a consequence chain few plans spell out. A device intended for one day of use that clinicians routinely carry into a second, or that a reviewer judges will be used past its labeled duration, now counts two contact days and moves from limited to prolonged contact.
Prolonged contact with anything other than intact skin now brings genotoxicity into the evaluation, either as a test battery or as chemistry and a TRA sensitive enough to address it, and both routes add cost and time that a limited-contact plan never budgeted for.
One extra day on paper can reshape the whole program. In other words, categorization has become one of the bigger calls in any plan and has significant impact on total effort.
How FDA Has Recognized It
For US submissions, all of this applies under FDA’s partial recognition (No. 2-313, May 2026). In short:
Excluded: clause 6.9 (biological risk estimation), as conflicting with ISO 14971:2019, along with the phrase “consumer products” in clause 6.5.11.3. The upshot is that biological risk in a US file is still estimated within the ISO 14971 framework.
Notes: the recognition carries two — one on concepts still under development in ISO/TC 194 and one on the added genotoxicity requirements.
Transition: declarations of conformity to the 2018 edition remain acceptable until July 1, 2029.
FDA has also said the 2025 edition doesn’t raise its testing expectations. That matches what we see from reviewers. The shift, which predates the new edition, is in how reviewers receive written rationales offered in place of data. They’ve steadily gotten less traction wherever the existing information can’t be tied back to the final finished device.
What a Good Plan Answers on Top of the Basics
Whichever edition you cite, the plan has to answer the same question. For each biological effect relevant to the device, what information already addresses it, and how well? Whatever’s left becomes the test plan.
Materials, manufacturing, existing data, and clinical or literature history all contribute, and each one can be read as a measure of distance from FDA’s low-uncertainty examples that fully mitigate testing. A silicone with an FDA Master File and an established history in the same contact category sits close to the exception. A new blend from a supplier unwilling to disclose its formulation or specific manufacturing processes doesn’t. Neither does an undocumented process aid.
When uncertainty remains, it tends to concentrate in three places:
Similarity — claims of similarity to existing devices & their manufacturing processes.
Material behavior — the material’s behavior in the test system itself.
Life cycle — the device’s exposure over its life cycle (along with the impact of that life cycle on biocompatibility as a whole).
The standard asks every plan to cover a lot more than this, and it should. These three are what a good plan, or a great one, works through on top of all of that.
Similarity: Similar Is Not Same
Of the three, similarity is the one most often overestimated by manufacturers when they are considering the steps needed to complete a biocompatibility program for a project. Smaller manufacturers often assume that a device made of the same materials as one already on the market inherits its biological safety. It doesn’t. A 510(k) predicate establishes substantial equivalence for clearance; on its own, it isn’t evidence of biological equivalence. FDA’s 2023 guidance frames that evidence, in Attachment F, as a statement that the finished device is identical to the legally marketed device “in formulation, processing, sterilization, and geometry,” with no other chemicals added. The same guidance notes that prior experience is more informative when it’s the sponsor’s own, because for another manufacturer’s device “the manufacturing and processing of the device material may be unknown.”
In practice, reviewers go to manufacturing early, often asking for processing details that a sponsor relying on another company’s device can’t obtain. And when the same processes can’t be demonstrated, reviewers don’t treat similar processes as the same.
In Europe, under the MDR, it’s tougher still. To claim equivalence there (Article 61; Annex XIV, Part A, Section 3), you have to show technical, biological, and clinical equivalence all at once, not just one or two of them. On the biological side, that means the same materials or substances touching the same tissues or body fluids, a similar kind and duration of contact, and similar release characteristics, which covers degradation products and leachables too. MDCG 2020-5 confirms that the exceptions once allowed for similar materials no longer apply, and that the manufacturer must have sufficient access to the equivalent device’s data. For implantable and Class III devices made by another manufacturer, that means a contract granting full and ongoing access to its technical documentation.
So, the answer comes out the same in both jurisdictions: without specific data on the identical combination of material, process, and sterilization, the evaluation will need data.
Material Behavior: When the Material Meets the Test System
The second place uncertainty collects is one that teams see coming less often. For a component machined from 316L stainless steel, the evaluation mostly confirms processing and cleanliness. Coatings, nanomaterials and novel absorbable formulations are a different story. They interact less predictably with the body and, often just as importantly, with the defined test systems used to evaluate them.
Even common materials like flexible plastics used in tube sets, or the ubiquitously used silicone, can spark significant challenges in biocompatibility test systems, especially those related to chemical analysis. Most of those systems were developed around conventional, stable materials and not the types of materials modern devices are utilizing to promote healing, fixation, or absorption. Put in a material that behaves differently than what ISO 10993-12 was developed around, and you can get a result that describes the test rather than the device.
Nanomaterials are the best-documented case. Nanoparticles can adsorb assay reagents, directly reduce tetrazolium dyes, or absorb at the measurement wavelength, causing a colorimetric cytotoxicity assay to misstate viability. That’s why ISO/TR 10993-22 addresses assay interference, and why FDA’s guidance calls for extraction conditions that avoid testing artifacts — easier said than done.
Absorbables interfere differently. A magnesium alloy that’s degrading pushes an extract’s pH and osmolality up. PLGA and other absorbable polyesters go the other way, dropping the pH as lactic and glycolic acid build up. And since FDA wants extracts used as-is, with no pH adjustment unless you can justify one, a static extract can read as cytotoxic or hemolytic even though the body would buffer and clear those byproducts in vivo. For coatings and other heat-sensitive constituents, the guidance also cautions that extraction above 37 °C can generate chemicals that wouldn’t occur in clinical use.
In a regulatory framework that is bent toward “check box” passing without deviation — all of these challenges create significant impact to a biocompatibility program, especially if they are identified in testing and not in planning.
Life Cycle: Exposure Over Repeated Use
The third, exposure over the life cycle, is now in scope under the 2025 edition, and it shows up most clearly in reusable devices. Run a silicone part through enough cleaning and disinfection cycles and it can soak up aldehyde disinfectants, glutaraldehyde and ortho-phthalaldehyde included, then give them back later. Now those residues are part of what the patient is exposed to. Because the absorbed fraction leaves the bulk of the elastomer only slowly, a rinse that clears the surface can still leave a device that releases disinfectant into tissue or into an extract.
We’ve worked with clients on this failure mode directly at Groenakker, the unexpected impact of a useful life on the biocompatibility of an otherwise biocompatible device. It’s why the rinse steps, which FDA’s 2015 reprocessing guidance expects to be validated, matter as much to the biological evaluation as they do to the cleaning validation.
Predict Before You Test
In each of these cases, the plan earns its keep by providing opportunity to predict how a material will respond, in the body and in the test system, before a study is commissioned. An invalid result or an unexplained failure generally costs more than the study itself. That’s why predictive analytics and simulations are part of Groenakker’s BEP process. How will a coating behave under extraction? How will degradation shift an extract? How does a disinfectant partition into an elastomer? We work those questions through before testing begins, not after a result has to be explained.
Chemistry and the Burden of Proof
With those sources of uncertainty handled, what’s left usually starts with chemistry: characterizing the finished device per ISO 10993-18, then running a toxicological risk assessment (TRA) per ISO 10993-17. Done right, that pairing can often cover systemic toxicity, genotoxicity, and carcinogenicity without putting a single animal on study.
Don’t mistake that for a shortcut, though; it still runs on data. Quantified constituents are compared against tolerable intakes or, where compound data are lacking, a threshold of toxicological concern. The chemistry substitutes for the animal work. It doesn’t substitute for evidence.
That moves the burden of proof onto the chemistry, which has become a principal source of deficiencies in its own right. FDA put numbers on this at a second town hall in September 2026: of the fiscal year 2025 letters it audited, 19 had chemistry or TRA deficiencies, 32 of them in total. And it was the same handful of problems over and over:
how the test article was defined
extraction conditions
analytical thresholds
chemicals nobody identified
The Endpoints That Still Need Testing
Even a strong chemistry package rarely closes out cytotoxicity, sensitization, and irritation. These are the endpoints the evaluation should expect to test, so the BEP’s job shifts from deciding whether to test to defining the test precisely enough that it runs once and only once. That comes down to three decisions made on paper: what the test article is, which configuration represents the worst case, and which extraction conditions are justified for the device and its materials.
Those decisions carry the schedule. Once samples land at the test facility, a sensitization study typically takes 8 to 10 weeks, so a test article that has to be redefined after the fact can set a program back two months or more. Study design works the same way. For absorbable devices, sutures included, we’ve consistently seen reviewers ask for implantation at three time points: 2 weeks, 13 weeks, and full absorption. A plan that sets those time points before anything goes on study avoids extending or repeating it later.
This is where the evaluation becomes a test matrix — the one that should follow from the plan, not the other way around.
One Plan, Two Reviews
For devices intended for both markets, the same evidence gets read against two references. FDA reviews against its 2023 guidance and its recognition of the standard. Notified Bodies, on the other hand, already treat the 2025 edition as the state of the art. In our experience EU reviewers also dig further into the reasoning behind an evaluation, so the quality of the BEP and the Biological Evaluation Report (BER) carries more weight there. A single BEP written to the 2025 structure and mapped explicitly to the FDA position generally serves both reviews better than parallel plans.
The Bottom Line
ISO 10993-1:2025 is an evaluation standard by design, and FDA’s own examples show that evaluation can stand in for testing where uncertainty is genuinely low. For most devices, that condition is only partly met, so the evaluation will still include some testing. A well-built BEP gives you a clear record, set down early, of which parts of the safety case rest on existing knowledge, which need data, and how the materials can be expected to behave once that data is generated.
So what’s the point of all this? Do the BEP, do it well, and do it as early as you reasonably can. Every expensive surprise in this post, from a device that slips into prolonged contact, to a similarity claim that won’t hold, to an extract that fails for reasons that have nothing to do with the patient, is cheapest to find on paper, while materials, suppliers, labeling and reprocessing instructions can still change. Found in the lab, the same problem costs a repeat study and months of schedule. Found in a deficiency letter, it costs more than that.
Started at design input rather than after design freeze, the BEP stops being a record of decisions already made and becomes the tool that makes them.