Choosing Genotoxicity Test Methods: An ISO 10993-3 Guide

1. Settle "is it needed" before arguing about "which test"

Genotoxicity is not an endpoint every device has to address, and it is not one you can wave away with "it does not contact blood, so it does not apply". Its position in the framework is actually quite clear: whether the endpoint is needed is decided by the biological evaluation process of ISO 10993-1; once it is needed, the methodology is handed over to ISO 10993-3.

Many companies run the sequence backwards - they phone the laboratory to ask what genotoxicity costs, then go back and write the evaluation report once the quote is in hand. Two outcomes are typical. Either too much gets done, and money is spent generating data the evaluation report never uses; or too little gets done, the gap is filled during the deficiency response, and the whole registration cycle stretches by another round.

A safer sequence looks like this:

1. Classify the product by the nature of body contact (surface contact, externally communicating, implant) and by contact duration category (limited, prolonged, long-term);
2. Perform a risk-based analysis in the biological evaluation report to decide whether the genotoxicity endpoint needs experimental data or can be closed out with existing information;
3. If data are needed, first consider whether chemical characterisation (ISO 10993-18) plus a toxicological rationale can replace or reduce the amount of testing;
4. Only what genuinely remains goes on to method selection under ISO 10993-3.

Step three is the one most often underrated. If the materials evidence for a product is strong enough, the leachables list is complete, and every substance on that list can be supported by usable toxicological data, there is a real chance of closing the genotoxicity endpoint by rationale. The barrier on that route is not willingness to write the rationale - it is the quality of the chemical characterisation itself. An incomplete leachables list leads reviewers to conclude that you have not even established what needs evaluating, and you end up paying for characterisation and still being asked for testing, doubling the cost.

For a more detailed breakdown of route selection, read this alongside our biocompatibility testing service page.

2. ISO 10993-3 gives you a set of endpoints, not a list of tests

This is the key to understanding the whole subject. ISO 10993-3 is not organised as a table of mandatory genotoxicity tests. It first defines the genetic damage endpoints that have to be covered, then leaves you to cover them with as few tests as possible.

The endpoints to be covered fall broadly into three groups:

• Gene mutation: changes at the level of the DNA sequence, including base substitution and frameshift types;
• Structural chromosome aberration: structural damage such as chromosome breakage and rearrangement, mechanistically a clastogenic effect;
• Numerical chromosome aberration: aneuploidy arising from abnormal chromosome distribution during division, mechanistically an aneugenic effect.

The corresponding test types therefore fall into two families. The bacterial reverse mutation assay (commonly called the Ames test) uses reverse mutation in auxotrophic strains to address the gene mutation endpoint; it is inexpensive, short and information-dense, and nearly every combination puts it first. Mammalian cell in vitro assays are used to cover chromosome-level damage, and micronucleus-type endpoints have the advantage that a single observation reflects both clastogenic and aneugenic mechanisms, which makes them economical in combination design.

A caution here: exactly which endpoint categories exist, which test methods are recognised for each, and whether a given combination is considered adequate coverage are all governed by the current valid version of the ISO 10993-3 text. Checking this point with the laboratory before the protocol is finalised is far cheaper than adding tests afterwards on the basis of a remembered list.

There is one more design element that gets overlooked: metabolic activation. A substantial proportion of substances do not react with DNA directly and only produce a reactive intermediate after metabolism in the body. In vitro tests therefore normally run in parallel with and without metabolic activation. If only one condition was run at submission, coverage of that endpoint in the evaluation report will not hold up - this is one of the more frequent triggers for deficiency letters in real projects.

3. Where the fork sits: in vitro first, in vivo as a conditional follow-up

The main logic of method selection is in vitro first. The reasons are partly ethical, in terms of animal welfare, and partly practical engineering: an in vitro system lets you control the concentration gradient of the test article directly, whereas in an in vivo study the amount of test article reaching the target organ depends on absorption, distribution and metabolism, so it is often hard to confirm that the target tissue was genuinely exposed.

In vivo testing generally appears in two situations: first, when an in vitro result is positive and confirmation is needed at whole-animal level as to whether that positive has biological significance; second, when the nature of the test article makes the in vitro system unreliable - for example the extract precipitates in the culture system and seriously interferes with readout, or the substance can only show activity after a specific metabolic pathway in vivo.

It is worth stating plainly that "positive in vitro" does not equal "the product cannot be used". Concentrations in an in vitro system are usually far above real clinical exposure, and cytotoxicity, pH shift and osmolality change can all produce non-specific positive signals. So the first move on receiving a positive result is not to change the formulation but to go back and check whether the test conditions sat within the valid ranges the standard requires.

4. Preparing the test article: the step where projects come off the rails

Like cytotoxicity testing, genotoxicity testing in most cases does not examine the device itself but an extract of the device. If sample preparation goes wrong, everything done downstream is wasted no matter how correctly it is executed.

Polar and non-polar extraction media are a pair. The evaluation framework requires extraction with both polar and non-polar media so that leachables with different solubility characteristics are drawn out as far as possible. Using only one medium amounts to assuming the other class of substance does not exist in your product, and that assumption rarely holds. Non-polar extracts also raise dispersion and vehicle questions when they enter a cell culture system, so the handling approach needs to be discussed with the laboratory in advance.

Extraction ratio and conditions. The ratio of surface area to extraction medium volume, the extraction temperature and the duration all follow defined categories. Note that these conditions are not set by the methodology standard itself but by the part of the ISO 10993 series that deals with sample preparation - which part governs and which edition applies has to be confirmed against the current valid version of the standard text and against your evaluation route for this particular project. Do not copy a protocol written for a different product. The real difficulty lies with products of unusual form: porous materials, hydrogels, powders and coated devices, where how to calculate "surface area" is itself something that has to be argued. For these, write the area calculation method or the mass-based alternative into a short statement before submission and hand it to the laboratory with the samples, so the representativeness of the test article is not challenged after the report is issued.

False-negative risk, as an engineering analysis rather than a statistical conclusion. From a mass transfer perspective, extraction is a diffusion-controlled process: a leachable has to migrate from the bulk of the material to the surface before entering the medium. For dense polymer substrates, short-range diffusion resistance is significant, and the concentration of the target substance in the extract may fall below what the test system can resolve, giving a "negative" result. That negative reflects insufficient extraction efficiency, not the absence of genotoxicity in the material. The most direct way to check whether you have fallen into this trap is to read the chemical characterisation data against the extraction conditions - if characterisation shows a class of leachable is present but it is not detectable in the extract, the extraction plan needs rethinking. This is a mechanistic inference and implies no statistical conclusion.

Interference factors. pH shift, osmolality change, colour or turbidity in the extract all affect readout, and excessive cytotoxicity masks the genotoxicity signal outright, because chromosome behaviour cannot be observed once the cells are dead. Concentration design for in vitro genotoxicity therefore usually refers to cytotoxicity data from the same extract batch. Submitting the two projects together, sharing one extract batch, is generally simpler than running them separately and makes the evaluation report easier to keep internally consistent.

5. Method route comparison

The table below organises the common decision scenarios so you can locate quickly which route your product should take.

Note that this table reflects common engineering decision routes and does not replace a risk assessment carried out on your specific product. Similar products can reach completely different conclusions because of differences in material, process and contact site.

6. What to do when the result comes back positive

For an in vitro positive in a real project, the effective order of handling is as follows.

Re-check first; do not rush to change the design. Confirm that the negative and positive controls behaved as expected, that the cytotoxicity level sat within the range the standard allows, and that the extract showed no precipitation or marked pH shift. A fair proportion of positives can be characterised as a conditions problem at this stage.

Then trace the source and find the suspect substance. This step relies on chemical characterisation. ISO 10993-18 data tell you what is in the extract; picking out the substances that carry structural alerts often narrows the suspects quickly. Common origins include residual monomer, oligomers, antioxidant degradation products and processing aid residues.

Next comes the process-side response. Enhanced cleaning, post-curing, replacing an additive - what these have in common is that they lower the amount of leachables rather than changing the material itself. Reformulating usually costs far more and is normally the last option considered.

In vivo confirmation comes last. Only when source tracing and process adjustment have failed do you use an in vivo test to judge whether the positive has biological significance at whole-animal level.

Walking the chain end to end shows that your ability to deal with genotoxicity depends on how solid the chemical characterisation is. Without characterisation data, a positive result becomes a black box that cannot be attributed, leaving only trial and error, with time and cost both out of control.

7. What to prepare before submission

From the laboratory's sample-receiving perspective, whether the following are complete decides directly whether the project runs through in one pass.

For specific sample submission requirements and formats, prepare against the testing requirements page; for how the milestones are scheduled, see the testing process.

8. How this maps onto the domestic registration route

In domestic registration submissions, genotoxicity requirements are usually cited from the GB/T 16886 series (here "series" means the biological evaluation framework as a whole, not one particular part), with the methodology part corresponding to GB/T 16886.3; on the international route the same roles are played by the ISO 10993 series, with methodology corresponding to ISO 10993-3. Conceptually the two routes are the same: the evaluation process decides whether the endpoint is needed, and the methodology standard says how to address it.

What companies genuinely have to handle themselves comes down to three things. All three sit exactly where writing from memory produces errors, and where those errors directly change which tests you choose.

First, do not write version and adoption relationships into a protocol from impression. The current valid versions of GB/T 16886.3 and ISO 10993-3, the adoption relationship between them, and how the series is divided into parts can all change as standards are revised. Project planning should follow the standard text and the current valid version published by the standards authority, and the specific edition to be applied should be confirmed with the laboratory before the order is placed. That one confirmation often saves a full round of rework, because protocol design, the reasoning behind acceptance decisions and the citation format in the report all hang on the version. A common failure is freezing a version number into a quality document and leaving it untouched for years, only to discover at submission time that the cited edition is no longer the current valid one and the report has to be reissued.

Second, organise the protocol by endpoint, not by a list of tests. Whichever edition applies, the methodology standard is organised by first defining the genetic damage endpoints to be covered and then covering them with as few tests as possible. The right question at protocol design is therefore not "how many tests do we run?" but "do the tests I have chosen cover both gene mutation and chromosome-level damage, and can the rationale for that coverage be written into the evaluation report?". The difference shows most clearly when a standard is revised: a protocol organised by endpoint usually needs only local adjustment of test selection, while a protocol organised as a fixed test list often has to be rewritten from scratch.

Third, the same data may be accepted differently on different submission routes. Domestic registration, overseas registration and different product categories do not necessarily align on report format requirements, laboratory qualification requirements, or whether a chemical characterisation rationale can substitute for testing. Plan by confirming these separately for each target market rather than assuming one report covers everything; the current published requirements of each regulator and the actual review opinion govern.

One further point deserves separate mention: the quality of the evaluation report usually affects review efficiency more than the number of tests performed. Making three statements clearly - why these tests were chosen, why they are sufficient to cover the endpoints, and what closes out the endpoints that were not tested - is worth more than adding two more tests. A test that has no supporting line of argument in the report is just data nobody reads.

More material on related topics is collected in the knowledge centre.

9. What SUNGO Lab can provide

SUNGO Lab (Shanghai Shage Medical Technology Co., Ltd.) operates laboratories in Shanghai and Hefei, accredited by CNAS, CMA and IAS (USA), and provides biocompatibility testing and supporting evaluation services for medical devices, including protocol design for genotoxicity-related work, rationale for test article preparation, coordination with chemical characterisation data, and support in drafting the evaluation report. To be clear: an accreditation mark only demonstrates that the laboratory has the technical competence within its accredited scope and does not constitute a commitment regarding market access outcomes; registration and market access conclusions rest with the review opinions of the relevant regulators.

If your product is stuck on whether genotoxicity applies, which endpoints to cover or how to fix the extraction plan, bring the material information and the contact profile to us and our engineers will work out the route first, then propose a matching test combination and schedule. Call +86 132 4819 8029, or request a quote directly.