Do we really need this task? — scope-narrowing memo
Written: 2026-04-21
Purpose: before anyone picks task-3b5 up and tries to execute all five work-streams in its CLAUDE.md / index.md, read this. The task was scoped on 2026-04-19 before the 2026-04-20 Celine-conclusions meeting tightened the plan. The live scope is narrower than what the task brief suggests.
Resolution (2026-04-21). WS5 delivered per this memo; task completed and moved to
completed-tasks/. See the task'sCLAUDE.mdheader block for the full list of audit-visible propagations (CEP R-TF-015-001, CER R-TF-015-003, PMCF Plan R-TF-007-002,WarningsDeviceOutputreusable + en/es/pt translations, EU IFU MDR Precautions in en/es/pt). WS2 (Rank-7 legacy-PMS slice) is committed for delivery in the firstR-TF-007-003update at approximately 6 months post-certification. WS4 (fast MRMC) is not executed. The §6.5(e) declaration count in the CEP and CER is now three (Fitzpatrick V–VI, Pillar 3 severity, paediatric); autoimmune dermatoses and genodermatoses are a §6.3 sufficient-evidence determination. The indication remains unchanged; the CLAIM for these sub-categories is qualified by the Device Output Warning.
Short answer
The task concept is necessary. The task as currently written is overscoped for BSI Round 1 (submission 2026-04-21). Only one of its five work-streams is mission-critical; one should be deleted per Saray's explicit direction; one should be deferred to PMCF.
Execute WS5 only. Delete WS4. Defer WS2. WS1 and WS3 are already done.
The underlying gap is real — so the task cannot simply be dropped
Driver: pre-submission finding A.2.C5 (../../../pre-submission-review-2026-04-19-cep-cer.md). CEP lines 900–908 currently declare autoimmune (Gap 4) and genodermatoses (Gap 5) as "addressed through passive surveillance through PMS/PMCF data collection." Two concrete defects:
- MDCG 2020-6 §6.4 violation. Passive surveillance cannot fill pre-certification evidence gaps. This is a textbook §6.4 hit.
- §6.5(e) over-use pattern. Five §6.5(e) declarations in the CEP weaken every individual declaration by association. Reviewers read the pattern as indication over-ambition.
Submitting Round 1 with that wording intact is a known own-goal. Erin/Nick already pushed back on MRMC-as-primary for dark phototypes (A.2.C6 precedent). The same reviewers will flag the same defect on autoimmune/genodermatoses if the "passive surveillance" framing stays in.
Implication: the gap is inside the CEP, not inside the evidence base. What must ship is a re-framing of the CEP/CER wording, not a new clinical study.
Saray's rule of the day (2026-04-20)
From ../resources/_meeting-notes/2026-04-20-celine-conclusions-review.md §8:
Saray: "If Taig spends today on recruiting, he will not have time to pass the broom (revisar) over everything else. And a Round-1 submission needs the broom more than a sixth study."
Saray (general framing for the day): "T, prefer revising what is already done to opening a new study the day before the submission."
This framing governs every scope decision below. The question to ask on each work-stream is: "is this the broom, or is it a sixth study?"
The explicit strategy decisions from the meeting:
- Strategy A (fast MRMC). Rejected. ← this is WS4 of task-3b5.
- Strategy B (triangulation: literature + Pillar 2 per-group + legacy PMS + PMCF + low-prevalence justification). Confirmed. ← this is WS3 + WS1 + WS2 + part of WS5.
- Strategy C (narrow the indication). Rejected.
Work-stream by work-stream
| WS | Ingredient | What it produces | Status | Necessary for 2026-04-21? | Verdict |
|---|---|---|---|---|---|
| WS1 | Ingredient 2 (Pillar 2 per-group V&V) | AUC 0.948 autoimmune / 0.905 genodermatoses on held-out V&V set | ✅ Done (via task-3b7) | Already in | Leave as-is |
| WS3 | Ingredient 1 (Pillar 1 literature) | 22 load-bearing VCA anchors, pillar-1-literature.md | ✅ Done | Already in | Leave as-is |
| WS5 | Ingredient 5 + CEP/CER/PMCF/IFU propagation | Four-test §6.5(e) rewrite, PMCF spec, narrowed-claim language | NOT STARTED | YES — essential | Execute. This is the broom. It rewrites what is already written to match evidence we already have. |
| WS2 | Ingredient 3 (Rank-7 legacy PMS filter) | Autoimmune/genodermatoses slice of the 250k+ legacy corpus | NOT STARTED | No — deferrable | Defer to PMCF. Commit to analysing the slice in the first PMS Update Report (R-TF-007-003). §6.3 permits this. |
| WS4 | Ingredient 4 (new MRMC study) | Rank-11 supporting evidence from a dedicated reader study | NOT STARTED | No — delete | Delete. Saray rejected on the spot; Erin/Nick will pick apart a rushed MRMC; Rank-11 cannot substitute for Pillar 3 on any named sub-indication. |
Why WS5 alone is enough to de-risk BSI Round 1
If WS5 ships:
- CEP lines 900–908 become a four-test §6.5(e) analysis backed by evidence we already have — Ingredient 1 (literature, WS3) and Ingredient 2 (Pillar 2 per-group AUCs, WS1) — plus a pre-specified PMCF activity framed as confirms / strengthens, never fills / closes.
- The §6.4 violation goes away: pre-certification evidence is now two independent anchors, not "passive surveillance."
- The §6.5(e) over-use pattern drops from five declarations to three (autoimmune and genodermatoses exit the §6.5(e) list by being resolved under §6.3 sufficient-evidence).
- The CER §Representativeness + §Sufficiency sections inherit the same narrative. Note: CER §Sufficiency is gated by task-3b6 (surrogate-endpoint literature review) for the indirect-benefit causal chain — see
index.mdline 45.
Why WS2 can be deferred without losing the Round-1 argument
Two independently-scoring anchors (Pillar 1 literature coverage + Pillar 2 AUCs ≥ 0.80 on held-out V&V, both with tight CIs) already outgun "passive surveillance." Ingredient 3 (Rank-7 legacy PMS slice) is a fourth anchor that strengthens the package, but it is not load-bearing for the §6.4 argument. The defensible PMCF commitment language is:
"The autoimmune-dermatoses and genodermatoses slice of the 250,000+ legacy post-market report corpus will be analysed and reported as part of the first PMS Update Report (R-TF-007-003), consistent with the MDCG 2020-6 §6.3 provision that PMCF confirms and strengthens an adequately-evidenced pre-certification base."
That is §6.3-compliant (confirms/strengthens), not §6.4-violating (fills/closes). Do not blur the verbs.
Why WS4 is a negative de-risker and must be deleted
Three layered reasons a rushed pre-certification MRMC hurts rather than helps:
- Rank hierarchy. MRMC is Rank 11 and the CEP itself (line 801) already says MRMC "is not clinical data under the strict MDR Article 2(48) definition." Filling a Clinical Performance gap with evidence the document says is not clinical data is self-defeating.
- Erin/Nick precedent. BSI pushed back on MRMC-as-primary for dark phototypes (A.2.C6). The same logic applies here uniformly.
- Methodological floor. A rushed MRMC with <30 images per category or <5 readers is worse than no MRMC — reviewers pick it apart. There is no time between now and 2026-04-21 to meet the floor.
Saray's direct words: "If Taig spends today on recruiting, he will not have time to pass the broom over everything else." That is the ruling.
The four-test §6.5(e) rewrite can leave the Rank-11 slot either empty (three anchors are already sufficient) or described as a planned PMCF-linked MRMC — but not generated pre-certification.
What to actually execute in task-3b5
Rename the working scope to:
- Rewrite CEP lines 900–908 (
R-TF-015-001) as the four-test §6.5(e) analysis:- Test 1: narrow-and-bounded (~4 % combined)
- Test 2: core benefit-risk independence (the three benefits 7GH/5RB/3KX are evidenced on the remaining ~96 %)
- Test 3: adequate residual evidence — Pillar 1 literature (Ingredient 1, done), Pillar 2 per-group V&V (Ingredient 2, done), plus the deferred PMCF commitment standing in for Ingredient 3
- Test 4: PMCF addresses remaining uncertainty — with pre-specified enrolment targets, diagnostic-accuracy thresholds, user-concordance thresholds
- Update CER §Representativeness (
R-TF-015-003) with the triangulated-evidence narrative. Hold back on CER §Sufficiency until task-3b6 closes the indirect-benefit causal chain. - Pre-specify the PMCF activity in
R-TF-007-002 PMCF Plan:- Enrolment target per category group (N autoimmune, M genodermatoses within T months post-CE)
- Pre-specified diagnostic-accuracy thresholds
- Pre-specified user-concordance thresholds
- Trigger conditions for unscheduled CER update if thresholds breached
- Commit to the WS2 legacy-PMS slice analysis here
- Wording discipline: "confirms" / "strengthens" — never "fills" / "closes"
- Update the narrowed-claim language in
packages/reusable/snippets/intendedPurpose.*and the IFU (apps/eu-ifu-mdr/) for the autoimmune and genodermatoses sub-categories. Keep them in the intended use; narrow the CLAIM, not the INDICATION. - Cross-reference the Risk Management File (
R-TF-013-002) if any risks are affected by the narrowed-claim framing.
Produced in this folder:
four-test-rewrite.md— prose for CEP §900–908 and CER §Representativenesspmcf-activity-spec.md— PMCF activity specification with thresholds and the WS2 deferralnarrowed-claim-language.md— IFU + reusables wording
Skipped on purpose (record here so it is visible to the next person picking the task up):
evidence-package/rank-7-legacy-pms-filter.md— deferred to the first PMS Update Reportevidence-package/mrmc-protocol.mdandevidence-package/mrmc-results.md— deleted from scope per Saray 2026-04-20
If scope pressure disappears (hypothetical future Round-2 posture)
If, in a later round, time is no longer the constraint, WS2 and WS4 become genuinely useful:
- WS2 — a Rank-7 filter of the legacy 250k+ corpus gives a concrete numerator/denominator anchor (rule-of-three on zero-event categories), which strengthens the four-test §6.5(e) Test 3.
- WS4 — a properly-designed MRMC (≥30 images/category, ≥5 readers, pre-specified acceptance criteria, protocol locked before data collection) would push Ingredient 4 from "not applicable pre-certification" to "supporting Pillar 3 evidence." Do it in the PMCF window, not the week before a CE-mark deadline.
Cross-references
- Meeting that set the scope:
../resources/_meeting-notes/2026-04-20-celine-conclusions-review.md§8 - Pre-submission finding driving the task:
../../../pre-submission-review-2026-04-19-cep-cer.md§A.2.C5 - Related precedent (MRMC-as-primary rejection):
../../../pre-submission-review-2026-04-19-cep-cer.md§A.2.C6, and../../task-3b4-mrmc-dark-phototypes/CLAUDE.md - Task brief this memo narrows:
./CLAUDE.md,./index.md - Gated dependency for CER §Sufficiency:
../task-3b6-surrogate-endpoint-literature-review/ - Completed upstream work feeding Ingredients 1 and 2:
./evidence-package/pillar-1-literature.md,../completed-tasks/task-3b7-icd-per-epidemiological-group-vv/