15 — Reaching the patient

From extract to exposure.

+ The questionHow much of a compound reaches the patient, and what margin is enough?

An extract's amount is not a patient's dose

An extract's list gives amounts in a flask: so many micrograms of a compound in a measured volume of solvent, drawn from a part held for days in a liquid the body never contains (Chapter 12). That amount is not what the patient receives. Between the flask and the patient stand the size of the part, the length and the frequency of the contact, how much the part gives up in use rather than in solvent, and the patient's own weight. ISO 10993-17:2023 accounts for each of them in one computed number, which it then sets against the tolerable intake of Chapter 14.

That number is the worst-case estimated exposure dose, which the standard defines as an "exposure dose that is a maximum value for a specified intended clinical-use scenario". It is a dose in Chapter 14's sense, counted in µg/kg/d like the tolerable intake, so the two can be compared directly. The definition carries two qualifiers. The value is a maximum, built to stand above what the patient receives, and it holds for a scenario of use that has to be specified. A second way of using the same device, or a second patient group, therefore needs its own estimate.

The estimate divides the amount by the patient and the days

One of two starting points for the estimate is the total quantity of the compound in the device. Slides from a 2024 Society of Toxicology webinar, given by presenters from FDA, a device maker and a toxicology consultancy, give the equation. It multiplies the total by a scaling factor, then divides it by the patient's body mass and by the number of days over which the release is spread. The total comes from the chemistry of Chapters 12 and 13, where conditions built to exaggerate give an upper estimate of what the part held before use. The body mass is the patient's: 30 kg for the running device's lightest user (Chapter 1). The other inputs reach the estimate through the total: the device's size and materials, how it is used and how often, the patient group and the route, as Gruber of the Johner Institute, a consultancy, lists the edition's inputs.

The days matter because a compound released over a month gives a smaller daily dose than the same amount released on the first night. Where no release study exists, the total is spread over fixed periods, and for a long-term contact the slides list four: 1 day, 2 days, 31 days and 366 days. Each period yields its own dose, and so its own margin against the tolerable intake, and the lowest margin is the one used, in the account of Efor, a consultancy. Where a release study does exist, the second starting point is the highest quantity released in a single day, which needs no division by days.

37 — From an amount in one device to a dose per kilogram each day
WHAT THE PART HOLDS OR RELEASES total quantity in the device (µg), from the extract (Chapters 12 and 13) highest quantity released in a day (µg/d), from a release study × a scaling factor for what the patient's contact actually reaches the standard’s SF; the slides do not spell the term out the amount reaching the patient each day (µg/d) ÷ body mass 60 kg for an adult by default; 30 kg for the running device's lightest user (Chapter 1) ÷ the release days 1 d · 2 d · 31 d · 366 d the same total divided over 1 day or 366 days differs by about 366 times worst-case estimated exposure dose, EEDmax (µg/kg/d) a maximum for one stated use of the device ÷ the body-contacting surface area instead, for an irritant dose per area (µg/cm²), against a tolerable contact level (Chapter 11) the release-study starting point the release days divide only the total quantity: the release study’s own daily quantity needs no division by days Equations as the 2024 Society of Toxicology webinar slides give them for ISO 10993-17:2023 (E.1, E.3 and E.5; presenters from FDA, a device maker and a consultancy), with the 2023 account of Efor, another consultancy, agreeing on the four periods. Schematic.
Released substanceLimit or thresholdFocal detail
Plate 37 — The estimated exposure dose is built from the amount in the device, a scaling factor, the patient’s body mass and the days over which the release is spread. Only the division by release days differs between the two starting points, and that choice alone moves the dose by up to about 366 times for a long-term contact.

For an irritant the comparison moves to the surface, since what matters is the amount held against a patch of skin rather than the amount entering the body. The dose is then the released quantity divided by the area in contact, in µg/cm². It is set against a tolerable contact level, which the standard defines as an "estimate of the surface-contact exposure to an identified constituent ... that is without appreciable irritation". The running device's cushion seals against the skin of the face, so both comparisons apply to it (Chapter 11).

A model can take the place of a measurement

A total quantity is not always to hand, and a release can be modeled instead of measured. Chapter 4's release model is the simplest kind: the amount that diffusion brings out of a plastic grows with the contact area, the starting concentration and the square roots of the contact time and the diffusion coefficient. With those inputs a model returns an amount for each day of contact, which is what the estimate needs, and it returns that amount before any part exists to be tested.

A screening model can stand in for a measurement because it errs high at every choice. FDA's screening tool for color additives in devices shows how, and in November 2026 it was still under development. It treats a part as a flat sheet, because that shape releases the most, takes the additive as spread evenly through the polymer, and holds the diffusion coefficient fixed as release goes on. For a blend of polymers it takes as the worst case the softest component, the one with the lowest glass-transition temperature. That is the temperature at which a polymer turns from glassy to rubbery and its chains gain the freedom that lets an additive move.

Calling a model conservative describes its assumptions and proves nothing about its result. The European Commission's 2010 migration report describes its upper-bound model as meant to predict conservative values for compliance, and it warns that the model "might not always overestimate in all cases". The warning matters because such models may replace a test or add to it. Saylor, an FDA scientist, offered them to device makers as estimates of the most a patient could receive, "in lieu of or supplementary to extraction testing" in the words of Saylor's slides. An estimate is therefore defended by naming its assumptions one by one and showing which way each of them errs.

The route decides how much of the release gets in

What leaves a part and what enters the body are two amounts, and the route decides the difference. A compound swallowed meets the gut and the liver first, the same compound in the blood arrives whole, and on the skin it must cross the outer layer. Differences between routes already enter the tolerable intake through Chapter 14's uncertainty factors, in a 2023 reading of the edition by MCRA, a regulatory consultancy. A value derived for one route is therefore not moved to another without that factor.

How much of the released amount gets in is the other half of the question. None of the accounts of the 2023 edition that this chapter cites states that the edition assumes the whole released amount is absorbed. This book therefore takes the whole released amount as reaching the patient, as its own choice, and says so wherever an estimate rests on it. The choice errs high, as a screening estimate should, and only data on how much the route lets through can relax it.

The running device's cushion rests on intact skin for hours a night, so its estimate takes the skin route. EPA's draft assessment of D4 siloxane gives two human doses for repeated skin contact: 326 mg/kg/d where the skin is uncovered and 179 mg/kg/d where it is covered. Under the headgear, the skin beneath a mask's cushion is nearer covered skin than open skin, so the worked example takes the lower value. That reading is this book's, and the values remain a draft's (Chapter 14). What the same device adds to the air the patient breathes is estimated differently, under ISO 18562 (Chapter 19).

The margin of safety is a ratio of two estimates

The margin of safety completes the calculation. The standard defines it as the ratio of a constituent's tolerable contact level or tolerable intake to the worst-case estimated exposure dose, so the division gives a plain number: how many times the estimated dose fits inside the tolerable one. A margin above 1 "indicates a low toxicological risk", in Gruber's reading of the edition, and Efor writes that it must be greater than 1 for the risk to be acceptable. FDA's slides for its town hall of 7 October 2026 pair a margin above 1 with a second outcome, a margin below 1 "with root cause analysis and adequate risk management". None of the three says what a margin of exactly 1 means.

Solved example: the margin of safety for D4 siloxane from the running device's cushion
  • The tolerable intake. EPA's draft for D4 gives 179 mg/kg/d for repeated contact with covered skin. Divided by the draft's own factor of 30 (Chapter 14), that dose gives about 5.97 mg/kg/d, or about 5,970 µg/kg/d.
  • The amount in the silicone. In a 2012 study, Zhang and colleagues at FDA measured the siloxanes D4, D5 and D6 in silicone products made for food contact, and found all three in every nipple sample, at about 0.5 to 269 µg of siloxane per gram. The study gives that range for the three rings together, not for D4 alone, so taking its top, 269 µg per gram, as the D4 that a gram of cushion silicone could hold is a choice that errs high.
  • The estimated dose. At worst, one gram of the cushion gives up all its D4 in one night and all of it reaches the patient: 269 µg divided by 30 kg and by one day is about 8.97 µg/kg/d.
  • The margin. 5,970 divided by 8.97 is about 670 for each gram of cushion silicone. Put the other way, the cushion would have to give up the D4 of about 670 g of silicone in a single night to bring the margin to 1.
  • The uncovered reading. On the draft's value for uncovered skin, 326 mg/kg/d, the same arithmetic gives a tolerable intake of about 10.9 mg/kg/d and a margin of about 1,200.
  • Assumptions. EPA's values come from its draft risk evaluation of D4. EPA's science advisory committee reviewed the draft at a public meeting in December 2025, and EPA has not made the evaluation final. EPA sets no tolerable intake for D4, so the division by 30 is this book's application of the standard's method to draft numbers. The content was measured in food-contact silicone, not in a mask cushion, and stands in for it by analogy. The release of a whole gram's content in one night, and the uptake of all of it, are this book's worst case rather than a measurement. Chapter 14 gives a further factor of 3 for two groups of young patients. Applied to the 30 kg user, that factor would bring the margin to about 220 per gram, still far above 1.

A margin below 1 says that the estimated dose exceeds the tolerable intake. NAMSA, a contract research organization, reads that result as an exposure that "could be harmful" and as grounds for further risk analysis, risk control or risk evaluation. Three answers follow, and each works on a different number. The estimate can be refined, by measuring the release instead of spreading a whole quantity over one day. The toxicology can be improved, by data that replace a default factor (Chapter 14) or by a test of the effect itself (Chapter 18). Or the part can change: less of the compound, another grade, or another material (Chapter 23).

A margin of safety grades two estimates against each other

Both of its numbers are built to err toward safety. The dose stands above what the patient receives, and the tolerable intake lies below the point where harm begins by the whole product of Chapter 14's factors. A margin of 3 therefore does not mean a patient takes a third of a harmful dose. Yet the 2024 slides say that a dose can be harmful when the tolerable intake, the tolerable contact level or the estimated dose is not conservative.

38 — The margin of safety on a tenfold scale, and what each reading leads to
MARGIN OF SAFETY 1,000 100 10 1 0.1 0.01 1 — the tolerable intake equals the estimated dose an MoS above 1 indicates a low toxicological risk, in the Johner Institute's reading of the edition FDA's 2026 town hall slide states the outcome as an MoS above 1, or below 1 with root cause analysis and adequate risk management the running device's cushion, for each gram of its silicone: about 670 (EPA draft values for skin; this book's arithmetic) a cushion would have to give up the D4 of about 670 g of silicone in one day to reach 1 neither reading above says what an MoS of exactly 1 means the ratio judges the estimate, and both of its numbers are built to err toward safety below 1: the estimate, not yet the patient, says the dose could be harmful refine the estimate: a release study in place of the total quantity add data in place of a default factor, or a test (Chapter 18) change the part: less of the compound, or another material (Chapter 23) MoS = tolerable intake ÷ worst-case estimated exposure dose (as ISO 10993-17:2023 defines it and the 2024 webinar slides give the equation). Readings: the Johner Institute, a consultancy, undated; a contract research organization, undated; FDA town hall slides, 7 October 2026. Cushion values: EPA's draft hazard assessment for D4, September 2025, and an FDA measurement of siloxanes in food-contact silicone products, 2012. Schematic.
Device materialReleased substanceTissue and cellsLimit or thresholdFocal detail
Plate 38 — The margin of safety is the tolerable intake divided by the worst-case estimated dose, so it rises as either the estimate falls or the tolerable intake rises. A value below 1 sends the assessor back to the estimate, to the data or to the material; neither the Johner Institute's reading nor FDA's slides say what an exact 1 means.

Compounds that do the same harm are added

A device releases more than one compound, and two compounds that do the same harm are not judged apart. Where the same harm, the same target organ and the same tolerable intake apply, the individual estimated doses "shall be summed" before the comparison, in the words the 2024 slides quote. Margins are also combined for compounds met at once with the same harm, target organ and mode of action, although the slides show no formula for that combination. Exponent, a consultancy, reads the rule in a 2023 article as covering compounds that share a mechanism of action or a structural resemblance.

Where a compound's critical effect is unknown, its margin stands alone, because nothing says which other margin it belongs with. Grouping works the other way and makes the test stricter. EFSA's group limit for DEHP and three other phthalates, expressed as DEHP (Chapter 14), applies the same reasoning in food law: under it, a device releasing two of those phthalates spends one allowance rather than two. Chapter 16 takes up the compounds that have no tolerable intake at all. Chapter 24 reads a finished evaluation, in which these sums appear compound by compound.

In practice

Write the exposure estimate as a table with one row for each compound and one column for each assumption: the quantity and where it came from, the scaling factor, the body mass, the release period, the route and the share taken up. Compute a margin for every release period the contact allows, and carry the lowest. Mark the status of every value, draft or final, in the row that uses it. Where a margin falls below 1, record which of the three answers was taken and why.

+ What this chapter established
  • The estimated exposure dose turns an amount per device into a dose per kilogram each day, through a scaling factor, the patient's body mass and the days over which the release is spread.
  • Where no release study exists, the total quantity is spread over fixed periods, and the shortest period gives the largest dose and the smallest margin.
  • Release models, and the screening tools built on them, err high at every assumption, which is what lets them stand in for a test and what has to be stated with the result.
  • The margin of safety is the tolerable intake divided by the worst-case estimated dose, and a value below 1 sends the assessor back to the estimate, to the toxicology or to the material.
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.