Research and planning

Internal working document

This document is for internal use only. It contains analysis, gap identification, and response strategy for Item 6b of the BSI Clinical Review Round 1. It will not be included in the final response to BSI.

1. What BSI is asking

BSI requests:

"Please provide a justification that the planned activities will collect sufficient data on safety and performance of the device throughout its lifetime."

The index-level observation adds:

"Overall justification that the PMCF plan is expected to collect sufficient data in quality and quantity to confirm the benefit, safety, and performance of the device over its lifetime based on all activities is not found."

This is a deficiency finding. BSI wants a consolidated argument — not just a list of activities, but an overarching justification that the totality of PMCF activities (general + specific) will produce enough data, in sufficient quality, to cover the device's entire expected lifetime.

2. What "data sufficiency throughout its lifetime" means

Annex XIV 5 requires PMCF to be a "continuous process" that confirms safety and performance "throughout the expected lifetime of the device." For software, the device's lifetime is typically defined by the period during which the manufacturer supports it — this should be stated in the technical documentation.

The data sufficiency argument must demonstrate:

1. Quality: Activities use appropriate methodologies (validated study designs, relevant endpoints, recognised standards)
2. Quantity: Sample sizes are statistically adequate; the total evidence base covers all claims
3. Coverage: Activities cover all intended uses, all user populations, all identified risks, and all performance claims
4. Continuity: The plan does not just collect data once — it describes how data collection continues throughout the device's lifetime (annual cycles, periodic monitoring)
5. Integration: PMCF data feeds back into the clinical evaluation (CER updates) and risk management

3. Gap analysis: why the current plan fails this requirement

The PMCF Plan describes 9 specific activities and 4 general methods. However:

Sufficiency criterion	Status in current PMCF Plan
Quality of methodology	Partial — study types are stated but methodology is not detailed enough to assess quality
Quantity (sample sizes)	Insufficient — only A.1 states a sample size (30,000 images); 8 of 9 activities have no sample size
Coverage of claims	Partial — CER gaps 1–3 are addressed but not all performance claims are explicitly mapped to activities
Coverage of risks	Not demonstrated — no explicit mapping of identified risks to PMCF activities that will monitor them
Continuity	Partial — the plan says PMCF is "continuous" and reports annually, but specific activities are one-off studies with start dates in 2026; no description of what happens after initial studies complete
Integration with CER	Partial — the plan states results will be in the PSUR, but the mechanism for CER update is not explicit
Overall sufficiency statement	Missing — no consolidated argument that collectively, these activities produce sufficient evidence

The "lifetime" problem

All 9 specific activities have start dates in 2026 and no end dates. Once these studies complete (presumably 2026–2027), the PMCF Plan does not describe what happens next:

• Are the studies repeated?
• Are new studies designed based on results?
• How is ongoing performance monitoring maintained beyond the initial studies?

The general methods (routine PMS data collection) provide continuity, but the PMCF Plan does not explicitly state that these general methods, combined with periodic specific studies, constitute the lifetime monitoring strategy.

The coverage problem

The PMCF Plan maps activities to three CER gaps, but does not map to:

• All identified risks in the risk management record (R-TF-013-002)
• All clinical performance claims in the IFU/CER
• All user populations (HCPs and ITPs — all specific activities focus on HCPs)
• Safety specifically (activities focus on performance endpoints; safety monitoring relies on general PMS methods, but this is not stated)

4. Equivalent/similar device evaluation

BSI notes: "An evaluation of the clinical data relating to the equivalent or similar devices is not found."

Annex XIV 6.2 requires the PMCF Plan to include an evaluation of clinical data relating to equivalent or similar devices. The current PMCF Plan has no section addressing this.

This section needs:

1. Identification of equivalent or similar devices (AI-based dermatology diagnostic support tools)
2. Summary of available clinical data for those devices (published studies, vigilance data)
3. Assessment of how that data informs the PMCF strategy (e.g., what risks have been identified in similar devices that our PMCF should monitor?)

For a novel AI medical device, equivalent device data may be limited. But the section must exist and explain the search methodology and findings.

5. Cross-NC connections

Clinical Review Item 6a — Activity descriptions

Item 6a research addresses the detail of each individual activity. Item 6b builds on top of 6a — even if each activity is described in detail, the sufficiency argument requires showing how the activities work collectively. The response to 6b should reference the expanded activity descriptions from 6a.

Clinical Review Item 3b — Data sufficiency justification

Item 3b research raises the same "data sufficiency" concern at the CER level. The PMCF Plan's data sufficiency justification should be consistent with and complementary to the CER's data sufficiency argument. Where 3b identifies evidence gaps, the PMCF Plan should show how those gaps will be filled by PMCF activities.

Clinical Review Item 3a — Clinical data analysis

Item 3a research identified that the CER does not integrate legacy PMS data. The PMCF data sufficiency argument should reference the existing PMS evidence base (7 non-serious incidents over 4+ years of market experience with the legacy device) as part of the baseline safety profile that PMCF will continue to monitor.

Clinical Review Item 5 — PMS Plan

The general PMCF methods draw data from the PMS system. The PMS Plan (R-TF-007-001) describes the data collection infrastructure. The data sufficiency argument should reference the PMS Plan to show how routine data collection supplements specific PMCF studies.

6. Response strategy

The response should:

1. Provide the overall data sufficiency statement that BSI could not find — a consolidated argument that collectively, the general methods plus 9 specific activities produce sufficient evidence in quality and quantity
2. Map PMCF activities to the complete evidence needs — not just CER gaps, but all performance claims, identified risks, and user populations
3. Address continuity — explain the lifecycle of PMCF: initial targeted studies → results integration → updated gap analysis → next cycle of studies
4. Address quality — reference methodological standards (e.g., STARD for diagnostic accuracy, ISO 14155 for clinical investigations)
5. Provide the equivalent/similar device evaluation
6. Confirm that PMCF results feed into CER and risk management updates

Fixes required (in R-TF-007-002)

Fix 1: Add a "Data Sufficiency Justification" section

Add a section early in the PMCF Plan (after Rationale and Objectives) that consolidates:

• How the general + specific methods collectively cover all evidence needs
• How quality is ensured (methodological standards)
• How quantity is ensured (sample size adequacy across the portfolio)
• How coverage is achieved (mapping to claims, risks, populations)
• How continuity is maintained throughout the device lifetime

Fix 2: Add a PMCF lifecycle / continuity section

Explain that PMCF operates in annual cycles:

1. Execute planned activities
2. Analyse results and integrate into CER/PSUR
3. Reassess CER gaps and risks based on new data
4. Design next cycle of activities if gaps remain
5. Repeat

This demonstrates that the plan covers the device's lifetime, not just the initial post-market period.

Fix 3: Add an "Equivalent/Similar Device Evaluation" section

Per Annex XIV 6.2, include:

• Identification of comparable AI dermatology devices
• Summary of their clinical data
• How this data informs the PMCF strategy

Fix 4: Add a coverage mapping

Create an explicit mapping showing:

• Each performance claim → which PMCF activity monitors it
• Each identified risk → which PMCF activity monitors it
• Each user population → which PMCF activity covers it

This directly addresses BSI's concern that there's no overall justification the plan is comprehensive.

7. Risk assessment

Risk	Impact	Mitigation
BSI may find the sufficiency argument unconvincing without sample size data	High — sample sizes are critical for the "quantity" argument	Obtain sample size calculations for all 9 activities (depends on 6a open items)
BSI may question "lifetime" coverage since all activities are one-off studies in 2026	Medium — BSI may ask what happens after initial studies complete	Add lifecycle/continuity section describing annual PMCF cycles
BSI may note that safety monitoring is under-represented in specific activities	Medium — most activities focus on performance endpoints, not safety	Emphasise that general PMCF methods (vigilance, PMS data) provide continuous safety monitoring, and that specific activities include safety endpoints where relevant
Equivalent device evaluation may be thin for a novel AI device	Low — BSI understands novel devices may have limited comparators	Conduct the search, document methodology, and explain the limited findings
ITP user population not covered by any specific PMCF activity	Low — ITP users interact via API only; clinical risk is indirect	Explain in coverage mapping that ITP usability is monitored via PMS (complaints, feedback) and C.1/C.2 activities evaluate API output quality

8. Open items

#	Item	Owner	Status
1	All open items from 6a (sample sizes, acceptance criteria, completion dates) are prerequisites for the data sufficiency argument	Jordi	Complete
2	Define the device's expected lifetime (years) — needed for the "throughout its lifetime" argument	Jordi	Complete
3	Map all identified risks from R-TF-013-002 to PMCF activities that monitor them	Jordi	Complete
4	Map all performance claims from IFU/CER to PMCF activities that confirm them	Jordi	Complete
5	Search for equivalent/similar AI dermatology devices and their published clinical data	Jordi	Complete
6	Confirm the PMCF annual review cycle is documented in the QMS (e.g., in GP-007 or the PMS Plan)	Jordi	Complete

9. Final Auditor Review (Simulated)

A final review of the updated documentation, performed from the perspective of the BSI auditor assessing against MDR Annex XIV Part B 6.1 & 6.2, confirmed:

1. Overall justification for data sufficiency: Section 15.1.3 ("Data Sufficiency Justification") now explicitly combines general and specific methods to cover the 5-year expected lifetime.
2. Quality and quantity: The section correctly points to standard methodologies (ISO 14155, STARD) and references the robust, justified sample sizes established in our response to Item 6a.
3. Comprehensive coverage: The "Comprehensive Coverage Mapping" matrix successfully and explicitly links specific activities to all IFU clinical performance claims, residual risks from the RMF, and all intended user populations (HCPs and ITPs).
4. Lifetime continuity: The "Continuous Evaluation Lifecycle" sub-section clearly outlines an iterative, annual loop, showing that 2026 activities feed into continuous CER/PSUR updates, gap reassessment, and subsequent monitoring phases.
5. Equivalent/similar devices: Section 15.3 directly addresses Annex XIV 6.2(a) by explicitly naming comparable devices (SkinVision, DERM, etc.) and describing how their data will proactively inform risk management.

Conclusion: The PMCF plan is no longer just a list of studies; it is a well-justified, continuous strategy explicitly covering all claims, risks, and the entire device lifetime. The deficiency is resolved.

---

Regulatory framework: what the BSI meeting revealed

Severity warning

Nick stated during the BSI meeting (2026-03-25) that PMCF sufficiency is a condition for CE mark approval. His specific language was that the PMCF plan must be "robust" for BSI to certify. Item 6b is the overall sufficiency argument for the PMCF system — if BSI cannot assess whether the plan will produce sufficient evidence throughout the device's lifetime, Section F of the CEAR fails regardless of how strong the individual activity descriptions are.

The four applicable guidance documents

Document	Role for Item 6b
MDR Annex XIV Part B	The binding requirement. Annex XIV Part B requires PMCF to be a "continuous process" that "confirms the safety and performance of the device throughout its expected lifetime." This is not fulfilled by listing nine studies scheduled for 2026. "Throughout the expected lifetime" requires a documented mechanism for ongoing monitoring beyond the initial study cycle — including what happens after 2026 studies complete, how results feed into updated gap analyses, and what triggers new study designs. The device's expected lifetime must be explicitly stated (the PMCF Plan currently does not define it). Without an explicit lifetime and a continuity mechanism, the "throughout its lifetime" requirement cannot be assessed.
MDCG 2020-6, § 6.4	PMCF activities must confirm pre-market evidence, not fill pre-market gaps. This applies to the overall sufficiency argument, not just individual activity framing. The consolidated sufficiency statement (Fix 1) must be structured as: "The following evidence base is sufficient for CE marking, and the PMCF activities collectively confirm this sufficiency under real-world deployment conditions throughout the device's 5-year expected lifetime." If the sufficiency statement instead implies that PMCF will fill missing evidence, BSI will interpret this as an admission that the pre-market evidence is insufficient — triggering MEDDEV A7.4 non-conformity. The § 6.4 distinction between "confirming" and "filling" must be present explicitly in the sufficiency section.
MDCG 2020-7, §§ 7.1 and 7.4	Section 7.1 requires the PMCF Plan to include an overall data sufficiency justification — not just descriptions of individual activities. Section 7.4 requires the plan to describe how PMCF data will be evaluated and used to update the clinical evaluation (CER) and risk management record. These are distinct requirements: (1) justification that the plan is sufficient and (2) mechanism for integration. The current PMCF Plan (R-TF-007-002) addresses neither. The fixes (data sufficiency section, lifecycle continuity section, coverage mapping, CER integration mechanism) map directly to §§ 7.1 and 7.4 obligations.
MDCG 2020-1, three-pillar continuous monitoring	For MDSW, MDCG 2020-1 requires that PMCF addresses all three pillars continuously: Valid Clinical Association (VCA), Technical Performance, and Clinical Performance. The sufficiency justification must demonstrate that collectively, the general PMCF methods and specific activities monitor all three pillars across the device's lifetime — not just Clinical Performance. VCA is monitored through general surveillance (adverse events, complaint analysis, literature updates confirming the clinical basis of the algorithm output). Technical Performance is monitored through routine PMS data and performance metric tracking. Clinical Performance is monitored through Activities A.1–C.2 and D.1–D.2. This three-pillar structure is the correct framework for the coverage mapping required in Fix 4.

The "throughout its lifetime" obligation: what annual cycling means in regulatory terms

Annex XIV Part B's "throughout its expected lifetime" requirement means the PMCF Plan must credibly describe monitoring at every point from CE marking to end-of-support. For a software device with a 5-year expected lifetime (or however long it is declared), this means:

1. Year 1 (post-CE mark): Execute planned 2026 activities (A.1–C.2, D.1–D.2); collect general PMS data
2. Year 1 PMCF Report: Analyse results, integrate into CER and PSUR, update gap assessment
3. Year 2 planning: Based on Year 1 results, design next cycle — are gaps confirmed closed? Are new risks emerging? Do acceptance criteria need revision?
4. Years 2–5: Repeat cycle; general surveillance continues continuously; specific studies launched where Year 1 or Year 2 data shows residual gaps or new concerns

The PMCF Plan must describe this mechanism explicitly. The current plan says PMCF is "continuous" but does not describe the review-and-planning loop that makes it continuous. Without this mechanism, BSI cannot assess how the plan addresses the device's full expected lifetime — only the initial study phase.

Activities D.1 and D.2: their role in the sufficiency argument

Activities D.1 (autoimmune monitoring) and D.2 (genodermatosis monitoring) from issue-6-pmcf.mdx are not just additional activity descriptions — they are the regulatory load-bearing elements of the declared-gap sufficiency strategy:

Element	D.1 contribution	D.2 contribution
Sufficiency for Gap A (autoimmune, 3%)	D.1 is the specific activity committed per MDCG 2020-6 § 6.5(e) — without it, Gap A cannot be declared acceptable	—
Sufficiency for Gap B (genodermatoses, 1%)	—	D.2 is the § 6.5(e) activity for Gap B — without it, Gap B cannot be declared acceptable
Coverage mapping	Maps to autoimmune conditions (ICD-11 EA) across HCP user population	Maps to genodermatosis conditions (ICD-11 EC) across HCP user population
Lifetime continuity	D.1 is designed as an annual monitoring cycle, not a one-off study	D.2 uses a retrospective/longitudinal design that accumulates data over successive years

The coverage mapping in Fix 4 must explicitly include D.1 and D.2 as the PMCF activities addressing Gaps A and B. If they are absent from the coverage matrix, BSI will conclude that no monitoring is planned for 4% of the claimed indication space.

The § 6.4 implication for the consolidated sufficiency statement

The consolidated sufficiency statement (Fix 1) must be written to satisfy § 6.4. This means it cannot begin with "the PMCF activities will provide sufficient evidence for..." — this framing implies evidence is not yet sufficient. The correct framing is:

"The pre-market clinical evaluation demonstrates sufficient evidence for CE marking across all claimed performance dimensions. The PMCF activities described herein confirm this sufficiency under real-world deployment conditions and monitor for changes over the 5-year expected device lifetime. Specifically: [activity-by-activity confirmation role]."

This distinction is not bureaucratic — Nick confirmed during the meeting that BSI reads the PMCF sufficiency statement to determine whether the manufacturer believes pre-market evidence is adequate. A statement implying inadequacy triggers an automatic re-examination of Item 3b.

What the equivalent/similar device evaluation must actually demonstrate

Annex XIV 6.2(a) requires evaluation of clinical data from equivalent or similar devices. For a novel AI medical device, this section cannot be empty — but it also need not show exact equivalent devices. The section must:

1. Document the search methodology (databases searched, terms used, date of search)
2. Identify the most comparable published AI dermatology systems (e.g., SkinVision, DERM, published dermatology AI literature)
3. Assess what risks, use errors, and performance characteristics have been identified in comparable systems
4. State explicitly how this assessment informs the PMCF monitoring strategy — which PMCF activities address the risks identified in comparable devices

The value of this section for Item 6b is that it demonstrates the PMCF Plan is designed with awareness of known risks in the field — not just gap-filling for this device's studies. BSI will read this as evidence that the sufficiency strategy is grounded in the broader clinical evidence landscape, which strengthens the overall plan's credibility.