17 — How low to look

From a daily dose to a line on the chromatogram.

+ The questionA threshold is set in micrograms a day, and a laboratory measures micrograms in a vial: where does the laboratory draw the line?

A daily threshold is not a line an instrument can see

Chapter 16 left the evaluation with thresholds in micrograms a day, but an instrument reports concentrations. A chromatogram of the cushion's extract shows peaks whose areas become concentrations in the vial, in micrograms per milliliter, and a peak is only as reliable as the method that found it. The extraction of Chapter 12 and the patient's use stand between the two units: how many cushions went into the extract, how much liquid they were extracted into, and how many the patient meets in a day. The reporting line has to be built from all three.

That line is the analytical evaluation threshold, or AET, the reporting threshold Chapter 16 pointed forward to. ISO 10993-18:2020 defines it as the threshold below which the analyst need not identify or quantify extractables or leachables. PMDA, Japan's regulator, calls it the concentration below which identification is not required, in its information page on the same edition. The definition sets the laboratory's work. Above the line every peak is named, counted and sent to a toxicological risk assessment, and below it a peak may be left alone. Where the line sits therefore decides how much of the extract is ever looked at.

The chain of divisions

FDA's 2024 draft guidance on chemical analysis prints the calculation as its Eq. 1, in Appendix D. The calculation starts from a dose-based threshold in micrograms a day and multiplies it by the number of devices extracted. It then divides the result by the extract volume, by the devices a user meets in a day and by any dilution, and divides it again by the analytical uncertainty factor of Chapter 13. Each term accounts for one step of the extraction, the analysis or the patient's use.

The first division changes the unit. A threshold of 1.5 µg a day belongs to a patient, and dividing it by the volume the devices were extracted into turns it into a concentration that a detector can be held to. Multiplying by the number of devices extracted keeps the arithmetic per device. The second division shares the allowance out. FDA's draft calls this term C, the clinical exposure, and defines it as the number of devices a user would be exposed to in a day under expected clinical practice. A user who meets two devices in a day may take half as much from each, so C divides the line again.

Processing the extract moves the line as well. The draft's factor D is the dilution or concentration factor, the final volume over the initial volume. Diluting an extract gives a lower AET than the unprocessed extract, the draft states, and concentrating one gives a higher AET. For an unprocessed extract D is 1 and may be left out. The rule follows from what dilution does: it spreads the same mass through more liquid, so the same mass per device shows up as a smaller concentration. The line must fall by the same factor, or compounds would slip below it unexamined.

41 — One threshold, three reporting lines
one cushion (A = 1), extracted into 50 mL (B = 50 mL), analyzed unprocessed (D = 1) dose-based threshold: 1.5 µg a day contact of 30 days or more, FDA's 2024 draft ÷ 50 mL of extract ÷ 1 device a day 0.03 µg/mL the concentration at which one cushion would deliver the whole threshold ÷ uncertainty factor UF 2: 0.015 µg/mL noted by ISO as proposed for GC screening UF 4: 0.0075 µg/mL FDA's draft default for GC-FID and GC-MS UF 10: 0.003 µg/mL Nelson Labs' LC-MS (electrospray) value the same extract, three reporting lines: fivefold between them the uncertainty factor belongs to the analytical method, not to the device Arithmetic as FDA's draft guidance on chemical analysis prints it (Appendix D, Eq. 1, September 2024, draft). Factors: ISO 10993-18:2020/Amd 1:2022; FDA's draft; Nelson Labs' slides at FDA's ASCA meetings, 2024 (the laboratory's own figure). Extract volume from a published worked example (Nelson Labs, 2023); the single cushion a day is this book's choice.
Device materialReleased substanceGas or fluid pathLimit or thresholdFocal detail
Plate 41 — A daily threshold becomes a concentration by the extraction's own numbers, and the uncertainty factor then lowers it. Because no value of that factor is agreed, the same extract carries reporting lines a fivefold apart, and each one decides how many peaks are named.

The uncertainty factor has no agreed value

The last division covers what the method does not know. A screening method cannot tell how strongly an unknown compound answers its detector, so it counts that compound against a surrogate. Chapter 13 showed how far response factors differ between compounds, and the analytical uncertainty factor lowers the reporting line to cover that spread. FDA's draft computes it as 1 divided by 1 minus the relative standard deviation of the response factors in a reference database. Once the spread is as wide as the mean, the result is infinite or negative, and ISO's 2022 amendment states that its formula then gives no usable value (Chapter 13).

The value itself is contested. FDA's draft allows a default as low as 4 for GC-FID and GC-MS without further justification. It establishes none for LC-MS, whose response factors, the draft says, can span more than a thousandfold, a wider range than on GC-MS. ISO 10993-18's Amendment 1 of 2022 notes instead that a factor of 2 has been proposed for GC screening and that a factor of 1 can be justified where response variation is known to be low. In a 2021 poster, Oktem and colleagues, scientists at FDA, found factors of 6 to 10 in initial measurements of polymer additives on headspace GC.

Laboratories also report factors of their own. Nelson Labs, a testing laboratory, presented its figures in late 2024 at FDA's meetings on ASCA, the accreditation scheme of Chapter 25. Its slides give a factor of 2.0 on GC-MS and a factor of 10 on LC-MS with electrospray, and Chapter 13 shows that the formula gives the first but not the second. A larger set, the laboratory's in-house GC-MS database of 2,724 compounds, gave a spread of 56 % and a factor of 2.27. The slides mark these values as specific to the laboratory that measured them. A factor of 2 and a factor of 10 move the reporting line fivefold, so two laboratories can report different sets of compounds from the same extract.

Solved example: the reporting line for the cushion's extract
  • The threshold. The cushion is in long-term contact. For contact of 30 days or more, FDA's draft takes 1.5 µg a day as conservatively protective (Chapter 16).
  • The extraction. This book's plan extracts one cushion in 50 mL of solvent, so A = 1 and B = 50 mL. The volume is that of a worked example in Nelson Labs' 2023 workshop slides. The extract is analyzed as it comes, so D = 1.
  • The use. A user meets one mask in a night, so C = 1.
  • Without the uncertainty factor. A threshold of 1.5 µg a day, divided by 50 mL and by one device a day, gives 0.03 µg/mL: the concentration at which one cushion would deliver the whole threshold.
  • With it. Divided by a factor of 4, FDA's GC default, the line falls to 0.0075 µg/mL. A compound sitting exactly on it delivers 0.375 µg a day, a quarter of the threshold. The factor of 2 that ISO's amendment notes as proposed for GC screening would put the line at 0.015 µg/mL, and Nelson Labs' LC-MS factor of 10 at 0.003 µg/mL.
  • The check. The worked example in Nelson Labs' slides reaches the same 0.0075 µg/mL for a long-term device from the same threshold and volume, with two devices a day and a factor of 2.
  • Assumptions. The extraction plan and the single cushion a day are this book's choices, not a maker's record, and 50 mL is the worked example's volume, not a measured one. Every value holds for one analytical method, since the factor belongs to the method, not to the device.

A line set too high hides what it never reports

Each term in the chain can be argued toward a higher line, and a higher line reports fewer peaks. The calculation still runs correctly on such a line, so nothing in it signals the loss. Yun and colleagues, the FDA chemists of Chapter 13, built a reference set of 106 polymer additives in a 2025 bench study. They warn that badly chosen reference standards understate the spread of response factors. In their account the uncertainty factor then comes out too small and the AET too high, and extractables are under-reported. The report cannot show compounds it never listed, and none of them is ever held to a threshold of Chapter 16.

42 — The same chromatogram at two reporting lines
0.003 0.0075 0.015 0.03 CONCENTRATION IN THE EXTRACT (µg/mL) reporting line at UF 4: 0.0075 µg/mL at UF 10: 0.003 µg/mL no match in the library: unknown tentative name named, counted, and sent to a toxicological risk assessment never identified, never assessed reported only at the lower line Schematic; peak heights illustrative; peaks of the cushion's separate extracts drawn on one trace. Threshold values computed in this chapter. Below the line a compound need not be identified or quantified (ISO 10993-18:2020); a cohort-of-concern compound known or suspected to be present is investigated even below it (FDA's 2024 draft, Chapter 16).
Released substanceGas or fluid pathLimit or thresholdFocal detail
Plate 42 — The reporting line decides what the analysis has to name. Four peaks clear the line set with FDA's default factor, two more clear a line set 2.5 times lower, and the rest are never identified, so a report shows nothing of what its own threshold excluded.

A line set too low fails in another way. It can fall below the lowest concentration the instrument can reliably quantify, the method's limit of quantitation, and a laboratory cannot report what it cannot measure. For that reason FDA's slides for its town hall of 23 September 2026 ask that the AET stay above the limit of quantitation, in line with Annex E of ISO 10993-18. The same slides list unclear AET and uncertainty-factor calculations among the common deficiencies FDA sees. The line is therefore bounded from both sides: the toxicology sets its ceiling and the instrument its floor.

Two kinds of compound are not governed by the line. A member of the cohort of concern that is known or suspected to be present should be investigated even below the reporting threshold, FDA's draft states, because the threshold of Chapter 16 was never meant to cover it. A compound with a tolerable intake of its own is held to that intake instead of to a threshold. The reporting line is therefore a screen for unknown compounds, not a verdict on known ones.

The reporting line is a threshold divided, not a limit of safety

A compound below the AET is one the analyst need not name, in ISO 10993-18's words. It is not a compound shown to be safe. Nor is the line a dose: it is a daily threshold spread over an extract and then lowered for the method's own uncertainty. The dose a patient receives comes from the amount measured, the use and the body mass of Chapters 14 and 15, and Chapter 18 adds what chemistry cannot see at all.

Where the standards put it, and what is still draft

Two documents carry the calculation, with different standing. ISO 10993-18:2020 defines the threshold and sets out the arithmetic in its Annex E, and the standard was reviewed and confirmed in 2025. Its Amendment 1 of 2022, titled for the determination of the uncertainty factor, replaced one clause of that annex in full. Heise, a consultant at MED Institute, had noted in slides on the standard in 2020 that its printed formula for the factor should read 1 divided by 1 minus the relative standard deviation.

As of November 2026, FDA's guidance is still the September 2024 draft: every page carries "Draft — Not for Implementation", FDA itself still called it a draft at its town hall of 23 September 2026, and it does not appear on the agency's proposed guidance lists for fiscal year 2027. The draft bases its equation on the edition of ISO 10993-18 that FDA recognizes, the second edition of 2020. It names one difference from that edition, a factor D for processing the extract, which it says the standard itself notes.

Chemistry, from the extraction on, fixes how much of a compound a method can see, and toxicology fixes how much of it a patient may receive. The line is the one number where the two meet: a threshold from Chapter 16 divided by the laboratory's own conditions. The compounds above it form the list that Chapter 18's tests and Chapter 21's risk estimate work from, and Chapter 24 reads an evaluation back through the same chain.

In practice

Derive the AET before the analysis, not after it. Record each value in the calculation: the threshold and the contact it comes from, the devices extracted, the extract volume, the devices a user meets in a day, any dilution, and the uncertainty factor with the method and the database behind it. State the factor for each analytical method separately, as FDA's draft asks, and check that the line sits above the method's limit of quantitation. Where the factor rests on a laboratory's own response-factor database, name that database in the report, since the same extract reported against a different one yields a different set of compounds.

+ What this chapter established
  • The analytical evaluation threshold is the concentration below which an analyst need not identify or quantify a compound, and it converts a daily threshold into a line on a chromatogram.
  • The conversion divides the threshold by the extract volume, the devices a user meets in a day and any dilution, and multiplies it by the devices extracted, in the arithmetic FDA's draft prints as Eq. 1.
  • The analytical uncertainty factor covers the spread of response factors between compounds, and published values run from 1 and 2 to 4 and from 6 to 10, so the same extract can yield different reported sets. A spread as wide as the mean leaves the formula with no value.
  • A line set high under-reports compounds that then reach no risk assessment, and a line below the method's limit of quantitation cannot be met, so the AET is bounded by toxicology above and by the instrument below.
Created with Claude (Anthropic).
Educational reference material, not engineering, regulatory or clinical advice. Product and company names referenced are the property of their respective owners.