Worst-case selection is question-specific: different variants may challenge different biological questions.
Define the device family and the evidence question, then compare patient contact, complete formulation, processing, sterilization, surface state, geometry, exposure, degradation and evidence gaps. The largest device is not automatically worst case, and more than one representative may be needed.
A family-level claim is only as strong as the relationship between the selected article and every variant it is meant to cover. “Worst case” should name a supported mechanism of challenge, not merely a convenient SKU.
Why there is rarely one universal worst-case SKU
One configuration may have the greatest patient-contacting surface, another the highest concentration of an additive, another a distinct coating, and another the most difficult geometry to clean. Those differences can make different variants challenging for chemical characterization, local tissue response, process residues or degradation. A single “largest device” claim can therefore obscure the question it is supposed to answer.
No ISO 10993-wide formula identifies one universal worst-case model. The defensible approach is to select a representative for a defined evidence question and state its coverage limits. If several questions drive the programme, the plan may need several representatives or a justified bracketing strategy.
Define the device-family boundary and the evidence question
Start with a controlled list of models and variants intended to share evidence. Confirm intended use, patient population, contact type, tissue site, duration, frequency, design architecture and manufacturing routes. A marketing family or catalogue grouping is not automatically a scientifically coherent biological-evaluation family.
Then write the question: which configuration most challenges an extractables profile, a sensitization rationale, a local tissue interaction, a degradation concern, a residue question or an evidence bridge? This prevents the team from changing the meaning of “worst case” after a candidate has already been chosen.
Build a controlled candidate matrix
| Dimension | Compare | Why it may matter |
|---|---|---|
| Contact and exposure | Tissue or fluid pathway, direct or indirect contact, duration, frequency and exposed area. | Defines the patient exposure the evidence must cover. |
| Materials | Complete formulations, concentration ranges, colourants, additives, adhesives, coatings and lubricants. | Same base polymer or trade name does not establish equivalent chemistry. |
| Processing | Moulding, machining, joining, cleaning, surface treatment, sterilization and reprocessing. | Finished-device residues and surfaces can differ even when materials match. |
| Geometry | Surface area, thickness, porosity, lumen, joints, edges and complex features. | Geometry can change exposure, cleaning or degradation rather than simply scale it. |
| Lifecycle state | Aging, storage, repeated cycles, wear, corrosion, degradation and particle generation. | The challenging state may differ by biological question. |
| Evidence | Prior tests, chemistry, supplier data, complaints, changes and unresolved gaps. | Evidence gaps may limit the proposed family bridge or require another representative; uncertainty alone does not prove that a variant is worst case. |
Select by question, not by a simplistic total score
A matrix helps organise facts; it should not turn unlike considerations into one arithmetic answer. A high formulation concentration may dominate a chemistry question while actual surface treatment dominates a local response. A smaller complex lumen may be more challenging for cleaning residues than the family member with the highest overall area.
- Name the property that creates the challenge and connect it to evidence.
- Separate known differences from assumptions and missing information.
- Confirm that the selected production lot and final-finished condition actually express that property.
- Record why non-selected variants are covered and what remains outside the claim.
Same supplier, base polymer, predicate or previous market history does not automatically transfer biological evidence. Explain both similarities and relevant differences.
When multiple representatives, bracketing or pooling may be needed
Multiple articles may be appropriate when no single variant preserves all relevant challenges. Bracketing can be useful when supported extremes genuinely cover the variants between them. Pooling may support a defined method when the combined set is traceable and does not mask a material-specific or configuration-specific concern. Each choice needs a question-specific rationale and laboratory feasibility review.
Do not assume one representative must cover every endpoint. An article chosen for chemistry can differ from the device needed for a surface-dependent or implantation question, provided the programme explains how the evidence reconnects in the biological evaluation.
Document the representative-device bridge
- List every model and variant within the proposed family boundary.
- State the specific biological or analytical question.
- Compare contact, formulation, processing, surface, geometry and lifecycle state.
- Identify the selected article by model, revision and lot.
- Explain the challenge mechanism rather than only naming the selection.
- Document evidence for similarity and the significance of each difference.
- Record excluded variants, unresolved information and additional representatives.
- Connect the decision to the protocol, report, risk file, BEP and BER.
- Obtain qualified scientific and regulatory approval before evidence is extended across the family.
When to reopen worst-case selection
Reassess the selection when material formulation, supplier, processing, sterilization, geometry, surface, packaging, intended use, contact or device-family boundaries change. Also reopen it when testing reveals an unexpected extract, residue, degradation product or biological response, or when new information shows that a previously covered variant is materially different.
Selection documentation supports review; it does not guarantee that FDA, a notified body or another authority will accept the chosen bridge. The final decision depends on the device, jurisdiction, method and complete evidence file.
Primary sources and standards records
- ISO 10993-1:2025 public record
- ISO 10993-12:2021 public record
- FDA guidance on use of ISO 10993-1
- FDA test-article and marketed-device comparison examples
- FDA test-report recommendations
- FDA recognition record for ISO 10993-1:2025 — review current recognition and transition details
Source status: checked August 22, 2026. Standards records describe scope; consult controlled standards and current jurisdiction-specific requirements for device decisions.
Why this perspective is practical
Arvind Rathore is the founder of MedDev Advisory, where his work focuses on ISO 10993 biological-evaluation strategy and documentation. Before establishing the practice, he was a Marie Skłodowska-Curie Early Stage Researcher at INSERM U1026 Biotis within the ImplantSens network. His research covered implantable electrochemical biosensors, cytotoxicity, oxidative stress, sterilization effects and biomaterial–cell interactions, with research placements in France, Germany and Sweden. Peer-reviewed work in Bioelectrochemistry and Advanced Sensor Research also informs his evidence-led approach. Read more about Arvind Rathore.
Need a controlled family comparison and representative-device justification?
The companion pack structures candidate evidence, question-specific selection and coverage boundaries. Qualified reviewers still authorise the selected article and protocol.