ISO 10993 biocompatibility — the -17:2023 split is the real story nobody's talking about

You can ignore most of what you'll read about ISO 10993 right now. The test methods haven't changed in years. The contact categorisation matrix in 10993-1 is stable. What changed, and what's making every dual-submission biocompatibility report harder to write, is the ISO 10993-17:2023 revision and FDA's partial non-recognition of it.

Short version: ISO-17:2023 overhauled the toxicological risk assessment methodology. New Dose-Based Exposure (DBE) frameworks, revised TTC thresholds, tighter integration with chemical characterisation data. EU notified bodies have been expecting it since mid-2024. FDA's September 2023 recognition kept only portions of the standard. If your submission targets both markets, you're maintaining two toxicological assessment tracks — or you're writing a single evaluation report that explicitly reconciles them, which is what the stronger submissions do.

The rest of ISO 10993 is mechanics. But the -17 bifurcation is the thing sponsors are actually tripping over in 2026, and it affects how every other part of the evaluation gets framed.

The AET is the pivot

Analytical Evaluation Threshold per 10993-18 Annex E. This is the number the entire characterisation study hinges on. Above the AET, every identified extractable needs individual identification and toxicological assessment. Below it, compounds get dismissed as unlikely to matter.

Get the AET right and you have a tractable characterisation study. Too high and you miss compounds of concern — reviewers catch this. Too low and you're chasing trace impurities that cost weeks of analytical work for no risk reduction.

The AET calculation uses Dose-Based Threshold, device exposure duration, daily exposure amount, and an uncertainty factor. The DBT is where 10993-17:2023 changes the math. Historical practice used 1.5 µg/day per Cramer Class III / TTC methodology, and that still works for most adult devices. For paediatric exposure, FDA has been asking for 0.15 µg/day or lower. For short-term exposure, the Staged TTC framework in -17:2023 allows higher thresholds, but requires more evidence about duration.

In practice: if your BEP doesn't show the AET calculation with specific inputs and defensible uncertainty factors, you'll get an AI request. The calculation should appear in the BEP, not as a derived output in the final report. That's what distinguishes "we picked an AET" from "we justified an AET."

Extraction conditions — four decisions that actually matter

10993-12:2021 specifies extraction conditions. Solvent polarity, extraction ratios, temperature/duration combinations. The standard has flex built in, and the decisions you make within that flex drive characterisation cost and submission acceptability.

Exaggerated vs simulated. Exaggerated extractions use aggressive solvents at elevated temperatures. Worst-case. Simulated extractions mimic clinical exposure. For long-term implants, FDA typically expects both. EU notified bodies vary. A BEP proposing only exaggerated extraction is cheaper and can pass — but when the data flags solvent-chemistry artefacts that have no clinical relevance, you spend the savings (and more) explaining them away.

Solvent selection within categories. Dichloromethane vs ethyl acetate for the non-polar extraction produces different extractable profiles. Justify the choice against device chemistry and expected clinical exposure.

Extraction duration. 10993-18 Annex D requires exhaustive extraction — iterative until each extraction yields less than 10% of the first. For porous materials, that can be five or six cycles. Truncated three-point extractions presented without exhaustion demonstration get AI'd for additional cycles.

Analytical quantitation limit versus AET. The detection limit has to be below the AET. If your AET is low (paediatric populations, long-term implants) and your analytical method wasn't designed for that LOD, you redo the analysis. Catch this in method development, not after the data is collected.

Why one analytical technique isn't enough

GC-MS catches volatile and semi-volatile organics up to roughly 500 Da. Anything higher-MW — non-volatile polymers, oligomers, stabilisers — is invisible. LC-MS catches those. ICP-MS quantifies metals with sub-ppb sensitivity but tells you nothing about speciation. Silver ionic versus silver nanoparticulate versus silver complexed — same ICP signal, very different toxicological stories.

Submissions built on a single technique get AI'd for the coverage gap. Strong characterisation studies run orthogonal techniques — GC-MS headspace plus direct injection, LC-MS positive and negative ion, LC-HRMS non-target screening, ICP-MS with speciation where warranted, ion chromatography for ionic leachables from ionomer-based materials.

This is analytical chemistry, not compliance theatre. The technique stack has to actually address the chemistry you could reasonably expect to be there.

Bridging data — the powerful tool most submissions under-use

The most efficient biocompatibility submissions are the ones that don't generate new test data. FDA's 2023 guidance recognises three bridging paths: same device, same materials with process equivalence, or chemical equivalence established through characterisation.

Chemical equivalence bridging is the one most sponsors under-use. If your new device uses the same materials and processes as a cleared predicate, characterisation of both can establish extractables-level equivalence. Profiles match within defined tolerances — same compounds, quantities within roughly 2-fold — and the biological evaluation leverages predicate biological data. Profiles differ — the specific differences drive targeted testing, not full re-test.

The arguments that fail in review are the ones that claim equivalence without the characterisation. "Material X has been in cleared devices for 20 years" is a supporting point, not a bridging argument. The evidence has to be device-specific data showing specific chemistry equivalence to a specific predicate with specific biological data addressing specific endpoints. Without that, the bridge doesn't hold.

Long-contact timelines — why biocompatibility is the critical path

Surface and external-communicating devices with limited or prolonged contact run 4–6 months from BEP approval to final report. Permanent implants are a different problem. Subchronic toxicity under 10993-11 is 90 days rodent or 13 weeks non-rodent with multi-dose levels and full histopathology. Chronic extends to 6–12 months for permanent implants. Carcinogenicity where required goes 18–24 months with tumour latency considerations. Genotoxicity battery runs 8–12 weeks in parallel but complicates scheduling if anything comes back positive.

Practical consequence: biocompatibility for Class III implants needs to start at least 18 months before the target PMA submission or it becomes the critical path. Chemical characterisation and the BEP need to be earlier still because they determine what the test battery even is. Programs that start biocompat work alongside design verification find out, too late, that the implantation studies are the timing constraint.

Manufacturing residuals — the avoidable deficiency

Biological data often reflects the raw material. What ships includes mould release agents, machining lubricants, cleaning agent residuals, sterilisation residuals (ethylene oxide per 10993-7, gamma-induced free radicals, gas plasma byproducts), and packaging migrants. All of these can affect finished-device biocompatibility.

FDA guidance is explicit about evaluating residuals. Submissions that present raw-material data without demonstrating that manufacturing residuals were characterised or controlled get AI'd for the missing analysis. EO sterilisation residuals per 10993-7 are a particular recurring issue — testing done but thresholds not justified against the specific clinical exposure scenario.

The rule I use: whatever the patient sees in the finished device should be what gets characterised, not what the raw material supplier sold you.

How MANKAIND handles biocompatibility

Materials, processes, residuals, extractables data, and biological test results live as connected objects in the engineering record. When a material substitution gets proposed, the platform surfaces which characterisation remains applicable, which processing-specific data needs regenerating, which endpoints need re-evaluation, which clearance data can bridge. The BEP and BER aren't documents assembled before submission — they're live views of the biocompatibility record as it accumulates.

For long-term implants, that integration is the difference between a program that ships with a defensible biocompat file and one that spends two years reconciling undocumented changes.

Frequently asked questions about ISO 10993

What is ISO 10993?

ISO 10993 is a multi-part international standard series for the biological evaluation of medical devices. It provides a structured framework for determining whether a medical device — or its constituent materials — may produce adverse biological responses in the body. Part 1 (ISO 10993-1:2018) governs the evaluation framework; subsequent parts cover specific test methods and evaluation approaches.

Which ISO 10993 parts apply to my device?

Applicable parts depend on device contact type (surface, external communicating, implant) and contact duration (limited, prolonged, permanent). ISO 10993-1 provides the matrix that maps these categories to relevant biological endpoints. Most devices require a subset: cytotoxicity (10993-5), sensitization (10993-10), and irritation (10993-23) are near-universal; implants add systemic toxicity, genotoxicity, and implantation testing.

Is ISO 10993 testing always required?

No. ISO 10993-1 explicitly favours evaluation over testing. If existing data — chemical characterization, toxicological assessment, published literature, prior device history — adequately addresses a biological endpoint, additional testing is not required. The biological evaluation plan (BEP) documents this rationale. FDA and notified bodies accept evaluation-based arguments when they are rigorous and well-documented.

What is ISO 10993-18 chemical characterization?

ISO 10993-18:2020 specifies how to identify and quantify the chemical constituents of a medical device — including extractables (substances that can be released under exaggerated conditions) and leachables (substances released under actual use conditions). Chemical characterization is the foundation of modern biological evaluation; in many cases it substantially reduces or eliminates the need for animal testing.

How long does ISO 10993 testing take?

Typical timelines run 3–6 months for a standard test battery on a surface or short-contact device. Long-term implant testing with subchronic toxicity, genotoxicity, and chronic toxicity can extend to 9–18 months depending on implant duration requirements. Chemical characterization is often the longest lead time — careful extraction planning and analytical method development can take 8–12 weeks before the first data is available.