+ Part VII · The system

The evaluation as a system.

A biological evaluation belongs in the design file from the first choice of material. The method that puts it there took about four decades of guidance and six editions of ISO 10993-1 to form, moving from a list of tests chosen by contact to an estimate of risk made within the device's risk management. This part treats the evaluation as one system. It follows how the method changed and what the 2025 edition changed (Chapter 21), where the evaluation sits in risk management and design controls (Chapter 22), how a supplier's or a maker's change reopens it (Chapter 23), and how to read one (Chapter 24). With the quality system that keeps the file current, which Part IX treats, these chapters are this book's systems view.

21 — Checklist to estimate

From a checklist to a risk estimate.

+ The questionIf the 2025 edition still sets out tables of effects, what changed when the method moved from a checklist to a risk estimate?

A table chose the tests

For years the evaluation of a device began with a table. In 1986 FDA and two foreign agencies issued the Tripartite Biocompatibility Guidance for Medical Devices, as FDA's own history of its guidance records. A 1997 trade-press review by Bollen and Svendsen names the other two as the authorities of the United Kingdom and Canada. FDA then issued its memorandum G87-1 on 24 April 1987 and Blue Book Memorandum G95-1 on 1 May 1995. G95-1 took up ISO 10993-1, the international standard first published in 1992, and accepted it with a modified matrix and added requirements in several device categories. Its largest change from the Tripartite was "FDA's increased emphasis on a case-by-case testing strategy", in Bollen and Svendsen's reading.

The standard sorted a device before it chose any test. In Bollen and Svendsen's description, its first edition placed each device in one of three categories: surface devices, externally communicating devices and implant devices, each divided by the type of contact. Three bands counted duration: limited contact under 24 hours, prolonged contact from 24 hours to 30 days, and "permanent contact" beyond 30 days. The standard and G95-1 both identified 12 categories of potential biological concern, as Wallin and Upman of NAMSA, a test laboratory, reported in 1998. In their account, part 1 was then titled "Guidance on Selection of Tests". The matrix made it "relatively easy to select the categories of biological effects that are of concern", Wallin, NAMSA's president, wrote in 1996. Under that scheme the running device's cushion was a surface device on skin in permanent contact (Chapter 2).

Why a list of tests missed the material

The table era asked which tests to run. G95-1 "addresses only the selection of tests", Wallin wrote in 1996. In FDA's 2023 guidance, the revisions of 2009 and 2018 moved the focus "from how to determine which biocompatibility tests to conduct" to an approach that considers existing information first. A cell of the matrix names effects, not compounds, so two cushions of different silicones in the same contact drew the same tests, and no cell asked what either material released. Bollen and Svendsen warned that the matrix was not a checklist (Chapter 2), but a method built to select tests made that reading easy.

The tests were indirect, and some were borrowed. The two-rabbit muscle implant method "has its origins in the United States Pharmacopeia implant test for plastic pharmaceutical containers", Wallin noted. He also noted that FDA had in the past considered the Ames bacterial reverse mutation test sufficient for genotoxicity. Williams, a biomaterials scientist, argued in a 2019 review that most tests of the ISO 10993 series rest on the effect of a material's extracts on cells in culture (Chapter 18). In his view "it does not make sense to assess biocompatibility in a stress-free in vitro environment", and "there is still no valid test for assessing the risk of biomaterials-induced autoimmunity". He cites the Silzone heart valve, which was approved "after undergoing standard preclinical tests, including those for biological safety", in arguing that passing those tests did not show biocompatibility.

Existing information before a test

The editions that followed moved the question in steps. The 1997 and 2003 editions kept the title "Evaluation and testing", and the 2003 edition still covers the categorization of devices by "the nature and duration of their contact with the body" and "the selection of appropriate tests". The edition of 13 October 2009 added "within a risk management process" to the title, and the edition of 17 August 2018 kept that title. The 2009 edition covers "the evaluation of existing relevant data from all sources" and "the identification of gaps in the available data set on the basis of a risk analysis". A test became the answer to a gap in the data.

FDA's guidance followed the standard. Its first guidance on ISO 10993-1 was announced on 16 June 2016 and replaced G87-1 and G95-1. FDA had revised it from its 2013 draft "to emphasize use of risk assessment and leveraging of prior information within a submission to potentially reduce the need for new biocompatibility testing". The version of 8 September 2023 superseded one of 2020 and added Attachment G, for certain devices on intact skin (Chapter 18). The 2023 guidance agrees with the 2018 edition's focus on minimizing the "number and exposure of test animals by giving preference to in vitro models". It calls its own matrix "only a framework for the selection of endpoints for consideration and not a comprehensive checklist".

The 2025 edition puts the evaluation in risk terms

The sixth edition, published in November 2025, builds risk management into its structure. The edition was completely reorganized, and its title "has been aligned with the risk management framework described in ISO 14971", as its foreword states. ISO 14971 is the standard for device risk management that Chapter 22 follows. The title now reads "Requirements and general principles for the evaluation of biological safety within a risk management process". Clause 6 is a biological risk analysis, clause 6.9 a biological risk estimation and clause 7 a biological risk evaluation. Clause 7's evaluation "compares the estimated biological risk against given risk criteria to determine the acceptability of the biological risk", in the words of the standard's introduction.

The edition also changes what the evaluation examines. The endpoints of Chapter 2 are now called biological effects, and the single matrix of 2018 became four tables, one for each type of contact, as three laboratories report (Chapter 2). The term "externally communicating" gave way to the specific tissue each component meets, and "effects after implantation" became "local effects after tissue contact", "as some non-implanted devices also will need this type of assessment". The edition adds guidance on calculating exposure duration, which Chapter 2 applies to the running device, and confines Annex A to materials characterization, its foreword records. Reasonably foreseeable misuse, such as "use for longer than the period intended by the manufacturer", must now enter the evaluation, in NAMSA's reading.

The tables stay, and the change lies in how they are read. For the running device's cushion, the 2025 tables still list effects for long-term contact with intact skin. The estimate, built from the contact and from the chemistry of Chapters 12 to 17, decides which of them need new data. Read as a checklist, the same table would send each effect straight to a test.

Solved example: the cushion read as a checklist and through the estimate
  • The contact. The cushion touches intact skin every night for years: in the first edition, a surface device in "permanent contact" beyond 30 days; in the 2025 edition, intact skin in long-term contact (Chapter 2).
  • The effects. Four apply in both readings: cytotoxicity, sensitization and irritation, which FDA names as the endpoints of concern for intact skin, and genotoxicity, which two laboratories read the 2025 tables as adding for contact of more than a day.
  • As a checklist. One test per effect gives a cytotoxicity test, an irritation test in rabbits, a sensitization test in guinea pigs and a genotoxicity battery of at least two tests: 1 + 1 + 1 + 2 = at least 5 tests, 2 of them in animals.
  • Through the estimate. Chapter 18's plan screens cytotoxicity on cells, reads irritation on reconstructed human epidermis and runs the local lymph node assay for sensitization: 3 tests, 1 in animals. The battery's 2 tests are added only if the chemistry leaves a compound above its threshold.
  • The result. The same four effects are considered, the animal tests fall from 2 to 1, and the chemistry, not the table, decides whether genotoxicity is tested at all.
  • Assumptions. The checklist applies the 2025 effects with the animal tests of the table era, and the plan is this book's (Chapter 18).
51 — One table, two readings
CUSHION, SILICONE intact skin · long-term (Chapter 2) TABLE FOR THIS CONTACT (ISO 10993-1:2025) biological effects for consideration, among them cytotoxicity sensitization irritation genotoxicity READ AS A CHECKLIST a test a test a test a test each effect straight to a test READ THROUGH THE ESTIMATE all four effects existing information and chemistry (Chapters 12 to 17) biological risk estimate: is the risk acceptable without new data? yes: no new test for that effect no: a test without animals where one answers it; otherwise the most refined animal test, justified (ISO 10993-2) Schematic. The 2025 edition keeps its tables; the estimate decides which effects need new data. FDA does not recognize clause 6.9, and a US file estimates under ISO 14971:2019. Effects: FDA for intact skin; genotoxicity as two laboratories read the 2025 tables.
Device materialReleased substanceTissue and cellsFocal detail
Plate 51 — The 2025 edition keeps the table. Read as a checklist, it sends every effect to a test; read through the estimate, existing information and chemistry come first, and a test answers only the effects whose risk they leave open.

What FDA recognized, and what stayed

FDA recognized the 2025 edition in part, and the clause it left out is the one that names the new step. FDA does not recognize clause 6.9, "Biological risk estimation", because it finds the clause "in conflict with another FDA-recognized standard", and its record points to clauses 5.5, 6 and 7 of ISO 14971:2019, on risk estimation, evaluation and control. FDA also leaves out the phrase "consumer products" in clause 6.5.11.3, as "in conflict with an existing published final guidance". Complife, a testing laboratory, reads that guidance as Attachment G. A US file therefore estimates the biological risk under those clauses of ISO 14971:2019. FDA's 2023 guidance stays as it was: "No changes to FDA's Biocompatibility Guidance related to ISO 10993-1:2025 at this time", FDA's town hall slides stated on 9 September 2026. Chapter 26 gives the recognition's dates and the transition from the 2018 edition.

Read in sequence, each generation answered a sharper question. The table era asked which tests a device's contact called for. The editions of 2009 and 2018 asked what existing information already answered before a test was chosen, and the 2025 edition asks how large the biological risk is and whether it is acceptable. The sorting of Chapter 2 held constant: the three bands of duration run from the first edition to the sixth, as a 2025 note by Labcorp, a contract laboratory, records. The newer editions did not make the older tools obsolete: the tests of Chapter 18 are still the tests, and the tables still list the effects. What changed is the step that decides when a test is run.

In practice

State in the evaluation plan which edition of ISO 10993-1 it follows and, for a US file, that the biological risk is estimated under ISO 14971:2019 in place of clause 6.9. For each effect in the table for the device's contact, record the existing information and the chemistry first, then the reason for each test that remains.

52 — Four decades of the method: what each generation asked
WHICH TESTS DOES THE CONTACT CALL FOR? WHAT DOES EXISTING INFORMATION ALREADY ANSWER? HOW LARGE IS THE BIOLOGICAL RISK, AND IS IT ACCEPTABLE? 1986 1992 1995 1997 2003 2009 2016 2018 2023 2025 FDA ISO 10993-1 Tripartite guidance (FDA and two foreign agencies) G87-1 (24 April 1987) G95-1 (1 May): ISO 10993-1 with a modified matrix first guidance on ISO 10993-1 (notice 16 June); replaces G87-1 and G95-1 final guidance (8 September); Attachment G, intact skin first edition (year of publication) Evaluation and testing Evaluation and testing the title gains “within a risk management process” same title sixth edition: … the evaluation of biological safety within a risk management process held constant: sorting by tissue and duration; three duration bands, limited · prolonged · “permanent”, now long-term Dates: FDA's history page (Tripartite, G87-1, G95-1); Federal Register notices (2016, 2023); national standards catalogues (1997 to 2018); ISO (2025). 1992: year of publication. The Tripartite's other issuers are not named by FDA.
Focal detail
Plate 52 — Across four decades the sorting by tissue and duration held, and the question changed: from which tests a contact calls for, to what existing information already answers, to how large the biological risk is and whether it is acceptable.
+ What this chapter established
  • The table era, from the Tripartite guidance of 1986 to G95-1 and the first editions of ISO 10993-1, sorted a device by contact and duration and read from a matrix which effects to test.
  • A matrix that names effects and not compounds chose tests without asking what a material released, and Williams argued in 2019 that passing the standard tests did not show biocompatibility.
  • The 2009 and 2018 editions and FDA's guidance from 2016 put existing information before testing, and the 2025 edition makes the evaluation a biological risk estimate aligned with ISO 14971 while keeping its tables of biological effects.
  • FDA's recognition leaves out clause 6.9 and the phrase "consumer products", so a US file estimates the risk under ISO 14971:2019, and in September 2026 FDA reported no change to its 2023 guidance for the 2025 edition.
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.