+ Part V · Chemistry first
From what comes out to what is safe.
An evaluation today starts with chemistry and ends with a ratio. A laboratory first draws out what a device's materials can release, then names each compound and counts it (Chapters 12 and 13). Toxicology sets the daily amount of each compound that the body tolerates (Chapter 14). It then estimates how much reaches the patient and sets one against the other as a margin (Chapter 15). A compound that nobody has studied gets a threshold of its own (Chapter 16). Chapter 17 turns that threshold into the lowest concentration the laboratory must report, so toxicology sets what the chemistry has to see, and Chapter 18 asks what a biological test still adds. This part answers the second half of the central question: what a laboratory must measure to see harm before patients meet it.
Making a material talk.
+ The questionHow does a laboratory get a material to give up what it holds?
An extraction exaggerates use on purpose
To learn what a mask cushion can release, a laboratory soaks it in liquids the body never contains, such as hexane, and keeps it warmer than the body for days. Chapter 4 showed that the liquid, the temperature and the time set how much leaves a plastic, and an extraction pushes each of them past use. What a material gives up in use therefore generally comes out in the laboratory too, in larger amounts. The laboratory aims at a worst case, the choice of sample and conditions expected to release the most. FDA's draft guidance on chemical analysis asks that conditions be chosen "to obtain worst-case estimates of amounts of analytes", the substances measured, and it recommends conditions that exceed both the time and the temperature of clinical use. FDA issued the draft in September 2024, and it was still a draft in November 2026.
A laboratory exaggerates use because nobody has the full list of what a material holds before the analysis. A bill of materials that names only polymer families lists none of the material's residues or breakdown products (Chapters 3 and 5). For each substance present before use, an extract made under harsher conditions than use gives an upper estimate of how much the patient can receive (Chapter 4). An amount found tolerable in the extract is therefore tolerable in use (Chapters 14 and 15). An amount that is not tolerable shows no harm by itself, because it is larger than what the patient receives.
Food-contact testing shows how far apart a harsh test and real use can be. In the 2023 bench study of Chapter 6, Diamantidou and colleagues at Aristotle University of Thessaloniki tested half-liter bottles of new and of recycled PET. They used ethanol as a food simulant, a liquid that stands in for food in testing. Under that chapter's harsh conditions, 95 % ethanol at 60 °C for 10 days, the cyclic trimer reached about 600 to 3,000 µg/L. In virgin olive oil kept for 12 months at about 20 °C, the group detected no migration, by a method able to measure down to 10 to 50 µg/L. The liquid, the temperature and the time changed together, so the study shows the gap between the harsh test and the food, not which change produced it.
The liquid is chosen by its polarity, the uneven spread of electric charge across its molecules, which decides what it dissolves. Water is polar and draws out polar and charged substances, hexane is non-polar and draws out fat-loving ones such as plasticizers, and an alcohol such as isopropanol or ethanol is semi-polar, between the two. A molecule ends up mostly where it dissolves better, so each solvent is chosen to draw its kind of molecule more strongly than the watery or fatty body fluid it stands in for (Chapter 4). FDA's draft names water, isopropanol and hexane as the typical three, and it prefers undiluted alcohol to a mixture of alcohol and water. Unless otherwise justified, the draft recommends a polar and a non-polar solvent for contact up to 30 days, and adds a semi-polar one for long-term contact (Chapter 2).
Temperature, time and the volume of liquid
Each condition of an extraction works through a mechanism of Chapter 4. ICH's draft guideline for drug products lists conditions that set how much leaches, among them the ratio of surface area to liquid volume, the temperature and the duration of contact (Chapter 4). SCHEER, a scientific committee of the European Commission, adds three more in its 2024 guidelines on phthalates in devices: the polarity of the liquid, its flow rate and how often contact recurs. Flow matters because moving liquid carries the additive away before the liquid at the surface reaches its share, and the continuous agitation that FDA's draft describes renews the liquid at the surface in the same way.
Temperature has a floor and a ceiling, each set by a mechanism. Below the temperature of use an additive diffuses more slowly than it does in use, so an extract made below it could hold less than the patient receives (Chapter 4). FDA's draft therefore does not recommend extracting below the temperature of clinical use, and its example condition is 50 °C with 72-hour cycles. The ceiling follows from Chapter 5, because heat that cuts a polymer's chains makes molecules that use may never form, and ISO 10993-12, the standard on preparing samples, notes that an exaggerated extraction must not change the material chemically. In its fifth edition, of 2021 and amended in 2025, the standard lists five pairs of temperature and time from the floor upward: 37 °C for 24 or 72 hours, 50 °C for 72 hours, 70 °C for 24 hours and 121 °C for 1 hour. Ma of Nelson Labs, a testing laboratory, cites them in a 2025 article and reserves 121 °C for stable materials such as metals.
The volume of liquid is tied to the area that releases, within two limits. Too little liquid fills toward its partition share and stops release early (Chapter 4). Too much liquid dilutes the extract below what the instruments can see. ISO 10993-12's first edition, of 1996, set the ratio by the thickness of the part, as Jansen and Wallin of NAMSA, a testing laboratory, described it in a 1998 trade-journal account. In that edition, films, sheets and tubing walls up to 0.5 mm thick got 6 cm² of surface for each mL of liquid, and thicker parts 3 cm²/mL. FDA's draft fixes no ratio. It asks that solvent cover the sample completely, and it says that ISO 10993-12's ratios could be considered, provided the method can still measure below the reporting threshold of Chapter 17.
Three intensities, three questions
How hard to extract depends on what the answer is for. ISO 10993-18, the standard on chemical characterization, defines three intensities in its 2020 edition, which was amended in 2022. A simulated-use extraction uses "a method that simulates clinical use" and estimates what use itself releases. An exaggerated extraction is "intended to result in a greater number or amount of chemical constituents being released" than use, and gives an upper estimate. An exhaustive extraction is repeated until its yield has fallen to a small fraction of the first cycle's, and estimates most of what the material will give up to that solvent under those conditions. FDA's draft assigns exaggerated extraction, or a clinically relevant worst case, to contact under 24 hours. For longer contact it assigns exhaustive or exaggerated extraction, and it recommends drying each extract and weighing what it leaves. Simulated use lies outside the draft's scope.
The contact decides which question matters. A contact of less than a day empties only a thin layer under the surface, so one extraction harsher than use already exceeds what the patient receives (Chapter 4). Over months or years the depleted layer reaches deeper, and the patient can receive a large share of what the material holds, so the bound that matters is the total. Heise, a consultant at MED Institute, summarized ISO 10993-18 in 2020 slides for an event of Nelson Labs. The slides state that exhaustive extractions are recommended for long-term devices "to assure that total exposure is estimated conservatively".
A simulated-use extraction happens in a laboratory under conditions chosen to resemble use, so what it releases is an extractable in Chapter 4's sense. A note to the definition in ISO 10993-12 describes a simulated-use extraction as one made under conditions that represent the intended use appropriately. FDA staff asked in 2026 for such conditions to be justified on the best available evidence. Heise's slides note that these extractions are often wrongly called leachables studies. What such an extraction releases stands for leachables only as far as the resemblance to use has been shown.
An exhaustive series ends by a rule. The laboratory extracts the sample, replaces the solvent with fresh solvent of the same kind and extracts again for the same fixed time. Each cycle draws on a deeper, emptier layer of the material, so the yield falls (Chapter 4). The series stops at the first cycle that yields less than 10 % of the first, the fraction that FDA's draft, ISO 10993-18 and ISO 10993-12 all set. FDA's draft measures the yield by drying the whole extract and weighing what remains, the non-volatile residue, in each solvent separately. ISO 10993-12 also allows other means. At a town hall in September 2026 FDA staff added that the stop must be measurable and that the balance must therefore be able to measure one-tenth of the first residue: 1 mg when the first cycle leaves 10 mg.
The sample is the finished device
What is extracted matters as much as how. Each step from pellet to part can add residues the resin never held, and ethylene oxide leaves residuals of its own (Chapters 6 and 7). FDA's draft calls the sample the test article, and it therefore defines the test article as the device or component in its final, finished form, with its sterilization and packaging where they apply. It asks for whole devices, typically the model with the largest surface area and the most sterilization or reprocessing cycles, and it states that extracting a smaller model and scaling the result by surface area is not a worst case.
Age and use change the device further. A polymer that water, heat or ozone breaks down forms molecules the new part does not hold, so an extract of the new part cannot contain them (Chapter 5). FDA's slides for the 2026 town hall list factors that can change what a device releases, among them shelf life and the sterilization, cleaning or rinsing a user performs before first use. The slides state that "for some devices, testing a device at shelf life may be an important consideration". The aging has to resemble use, and an oven's heat alone does not (Chapter 7).
The running device: the cushion is extracted, the foam is judged by its gas
The running device shows how the contact decides the extraction. Its silicone mask cushion rests on intact skin for hours each night, so its contact is long-term after about a month of nightly use (Chapter 2). The cushion is therefore extracted under ISO 10993-12 and ISO 10993-18, once as made and again as aged and cleaned in use. The plan is this book's illustration, not a maker's record. The sound-damping foam is a polyether polyurethane, by this book's design choice (Chapter 5). It sits in the gas pathway, so what it gives the patient arrives through the breathing gas. ISO 18562 evaluates the foam by what the device adds to the air the patient breathes and, where water condenses in the path, to that condensate (Chapter 19).
- Contact and solvents. In FDA's draft, long-term contact, more than 30 days, calls for three solvents: here water, isopropanol and hexane, the draft's typical three.
- Intensity. The draft accepts exhaustive or exaggerated extraction for this contact. The plan takes exhaustive extraction, which Heise's slides report as recommended for long-term devices.
- Extracts per cycle. The draft asks for triplicate extractions in each solvent. The plan reads that as three cushions per solvent, each carried through its own series, so each cycle makes 3 times 3, or 9, extracts.
- Cycles. The rule compares a later cycle with the first, so a series needs at least two. At the draft's example of 50 °C with 72-hour cycles, two cycles take 144 hours, or 6 days, and make 18 extracts.
- The stop. If a first cycle leaves 10 mg, as in FDA staff's 2026 example, that solvent's series ends at the first cycle that leaves less than 1 mg.
- The result. The plan takes at least 6 days and 18 extracts, and each further cycle adds 3 days and 3 extracts for the solvent that needs it.
- Assumptions. The guidance is a draft, and neither it nor its example condition is a requirement. The 10 mg is FDA staff's example, not a measurement on a cushion. The three solvents run side by side, and the days leave out drying and weighing. The silicone is assumed to tolerate the solvents; the draft names alternatives, such as heptane, for materials that do not.
Write each part's extraction plan from its contact before choosing a laboratory: the contact category (Chapter 2), the solvents and intensity it calls for, the temperature with its margin to use and to breakdown, and the sample, finished, sterilized, packed and aged. Ask the laboratory how it will show exhaustion in each solvent and what its balance can weigh. Send gas-pathway parts, such as the running device's foam, to the breathing-gas tests of Chapter 19.
- An extraction pushes the liquid, the temperature and the time past use on purpose, so for the substances present before use its amounts give an upper estimate of what use releases.
- Solvents of different polarity draw out different molecules, and FDA's draft asks for two for contact up to 30 days and three for long-term contact.
- Simulated-use, exaggerated and exhaustive extractions answer different questions, and an exhaustive series repeats with fresh solvent until a cycle yields less than 10 % of the first.
- The sample is the finished device in its worst case, aged where it degrades in use, and a gas-pathway part such as the running device's foam goes to Chapter 19's tests of the breathing gas instead.
Let's Talk