Skip to main content
QMSQMS
QMS
  • Welcome to your QMS
  • Quality Manual
  • Procedures
  • Records
  • Legit.Health Plus Version 1.1.0.0
  • Legit.Health Plus Version 1.1.0.1
  • Legit.Health Utilities
  • Licenses and accreditations
  • Applicable Standards and Regulations
  • BSI Non-Conformities
    • Technical Review
    • Clinical Review
      • Round 1
        • Item 0: Background & Action Plan
          • X-3: Disease categorisation decision
          • Issue 6 — PMCF activities for X-3 gaps A & B
          • Item 0 — AI Context
          • Clinical Benefits Consolidation Options
        • Item 1: CER Update Frequency
        • Item 2: Device Description & Claims
        • Item 3: Clinical Data
        • Item 4: Usability
        • Item 5: PMS Plan
        • Item 6: PMCF Plan
        • Item 7: Risk
    • BSI Non-Conformities
  • Pricing
  • Public tenders
  • BSI Non-Conformities
  • Clinical Review
  • Round 1
  • Item 0: Background & Action Plan
  • Clinical Benefits Consolidation Options

Clinical Benefits Consolidation Options

Options for reducing from 7 to fewer clinical benefits.


Decision: Option C — 3 benefits. Implemented 2026-04-01. See meeting notes: _resources/meeting-transcript-and-notes-internal-2026-04-01-at-16-30.md

Rationale for choosing Option C over Option B​

The key question was whether the malignancy detection benefit (1QF) should be kept as a separate top-level benefit (Option B, 4 benefits) or merged into the general diagnostic accuracy benefit (Option C, 3 benefits).

Option C is the correct choice for the following reasons:

1. Architectural honesty: 1QF is not a separate algorithm​

The malignancy "detection" benefit is not produced by a distinct device capability. It is a derived calculation: the device outputs a probability distribution over all validated ICD-11 categories, and the malignancy index is the sum of probabilities assigned to ICD codes classified as malignant — Σ P(malignant ICD classes). The same classification algorithm that supports 7GH (general diagnosis) and 9VW (rare diseases) also produces the values that underpin 1QF.

Claiming 1QF as a separate top-level clinical benefit implies a separate device mechanism. The IFU already explicitly states: "The device does not diagnose specific conditions; it provides an interpretive distribution representation." Keeping 1QF separate would be inconsistent with that statement: the device does not independently detect malignancy — it classifies, and the clinical interpretation of the probability distribution for malignant ICD classes is a downstream use pattern.

2. One benefit, two metrics — not a problem​

The merger introduces heterogeneous metrics (Top-1 accuracy for general/rare, AUC for malignancy) within a single benefit. This is not a problem because they measure different facets of the same underlying capability:

  • Top-1 accuracy answers: does the highest-probability ICD code match the correct diagnosis?
  • AUC answers: does the probability distribution discriminate between malignant and non-malignant presentations?

Both are appropriate for their respective clinical questions. The CER explains the two sub-criteria clearly: they are complementary, not competing. Having two metrics for one benefit requires more explanation in the CER, but it requires less explanation than maintaining a benefit boundary that has no algorithmic basis.

3. The safety signal is preserved​

The main argument for keeping 1QF separate was that burying malignancy detection would weaken the safety story for BSI reviewers. This risk is mitigated by ensuring that the malignancy sub-criterion within Benefit 7GH has its own named row in the acceptance criteria table with its AUC threshold clearly displayed and prominently labelled. What BSI cares about is traceability: evidence for malignancy detection exists, is linked to specific performance claims, and is clearly visible in the CER and IFU. Hierarchy depth (top-level benefit vs. named sub-criterion) does not affect traceability.

4. Option C is more responsive to Erin's simplification request​

BSI's Erin Preiss asked for simplification because two benefits were too similar. Moving from 7 to 3 is more decisively responsive than 7 to 4. A smaller number of top-level benefit codes means fewer traceability threads, fewer places for a BSI reviewer to find a gap, and a cleaner CER argument.

5. FDA alignment is not relevant here​

The argument for keeping malignancy separate partly relied on FDA conventions, where malignancy detection tends to be called out separately. This submission is to BSI under MDR. FDA conventions do not govern the structure of the CER or the benefit listing under MDR Article 2(53).



Option A — 5 benefits​

Merge only: 3KX + 8PL + 0ZC → Care Pathway Optimisation. All other benefits unchanged.

#CodeBenefit
17GHThe device improves the accuracy of healthcare professionals in the diagnosis of dermatological conditions. This has a positive impact on patient management and outcomes related to diagnosis and monitoring of patients.
21QFThe device improves the accuracy of healthcare professionals in the diagnosis of lesions suspicious for skin cancer. This has a positive impact on patient management and health outcomes related to diagnosis, including reducing the risk of delayed diagnosis of malignant conditions and the need for unnecessary invasive diagnostic procedures.
39VWThe device improves the accuracy of healthcare professionals in the diagnosis of rare dermatological diseases. This has a positive impact on patient management and outcomes related to diagnosis and monitoring of patients, especially those suffering from rare diseases.
45RBThe device measures the degree of dermatological disease involvement objectively, quantitatively, and reproducibly, increasing the precision of healthcare professionals in the clinical monitoring of patients. This has a positive impact on patient management and outcomes related to the monitoring of patients and the management of their treatment.
53KX + 8PL + 0ZCThe device improves the precision of healthcare professionals in managing dermatological care pathways, encompassing referral decisions, resource allocation, and clinical assessment in remote care settings. This has a positive impact on patient management and outcomes related to the diagnosis and monitoring of patients, resulting in reduced waiting times for specialist consultation, improved adequacy of referrals, and expanded access to dermatological assessment across in-person and remote care settings.

Option B — 4 benefits​

Merge: 7GH + 9VW → Diagnostic Accuracy and 3KX + 8PL + 0ZC → Care Pathway Optimisation.

#CodeBenefit
17GH + 9VWThe device improves the accuracy of healthcare professionals in the diagnosis of dermatological conditions across a broad spectrum of clinical presentations, including rare dermatological diseases. This has a positive impact on patient management and health outcomes related to diagnosis, enabling more appropriate clinical decision-making and more timely identification of conditions across all dermatological presentations.
21QFThe device improves the accuracy of healthcare professionals in the diagnosis of lesions suspicious for skin cancer. This has a positive impact on patient management and health outcomes related to diagnosis, including reducing the risk of delayed diagnosis of malignant conditions and the need for unnecessary invasive diagnostic procedures.
35RBThe device measures the degree of dermatological disease involvement objectively, quantitatively, and reproducibly, increasing the precision of healthcare professionals in the clinical monitoring of patients. This has a positive impact on patient management and outcomes related to the monitoring of patients and the management of their treatment.
43KX + 8PL + 0ZCThe device improves the precision of healthcare professionals in managing dermatological care pathways, encompassing referral decisions, resource allocation, and clinical assessment in remote care settings. This has a positive impact on patient management and outcomes related to the diagnosis and monitoring of patients, resulting in reduced waiting times for specialist consultation, improved adequacy of referrals, and expanded access to dermatological assessment across in-person and remote care settings.

Option C — 3 benefits​

Merge: 7GH + 9VW + 1QF → Diagnostic Accuracy and 3KX + 8PL + 0ZC → Care Pathway Optimisation.

#CodeBenefit
17GH + 9VW + 1QFThe device improves the accuracy of healthcare professionals in the diagnosis of dermatological conditions across a broad spectrum of clinical presentations, including rare diseases and lesions suspicious for skin cancer. This has a positive impact on patient management and health outcomes related to diagnosis, enabling more appropriate clinical decision-making, earlier identification of rare conditions, and, in cases of suspected malignancy, reducing the risk of delayed diagnosis and the need for unnecessary invasive procedures.
25RBThe device measures the degree of dermatological disease involvement objectively, quantitatively, and reproducibly, increasing the precision of healthcare professionals in the clinical monitoring of patients. This has a positive impact on patient management and outcomes related to the monitoring of patients and the management of their treatment.
33KX + 8PL + 0ZCThe device improves the precision of healthcare professionals in managing dermatological care pathways, encompassing referral decisions, resource allocation, and clinical assessment in remote care settings. This has a positive impact on patient management and outcomes related to the diagnosis and monitoring of patients, resulting in reduced waiting times for specialist consultation, improved adequacy of referrals, and expanded access to dermatological assessment across in-person and remote care settings.

Rationale for each merger​

7GH + 9VW (diagnostic accuracy, general + rare)​

These two are the easiest merge to justify. Both use the same metric (Top-1 accuracy), the same domain ("Diagnostic accuracy"), and share studies (BI_2024 and PH_2024 support both). The only difference is indication scope: "Multiple conditions" vs "Rare diseases." Rare diseases are a subpopulation within general dermatology — calling them out separately in the current structure was a choice of emphasis, not a clinical necessity. The merged benefit just extends the scope of what "general diagnostic accuracy" covers.

In the filter criteria in clinicalBenefits.ts, this merge is a one-liner: add "Rare diseases" to 7GH's indications array and delete the 9VW entry.

7GH + 9VW + 1QF (add malignancy)​

The argument extends naturally: the device improves HCP diagnostic accuracy, and malignant lesions are a subset of dermatological conditions. The underlying mechanism is identical — AI analysis supports the HCP's classification decision. The population differs in clinical stakes, not in how the benefit works.

The counter-argument is that 1QF uses a different metric (AUC vs Top-1 accuracy) because detection of rare events like cancer is better measured by AUC than by classification accuracy. Merging them means the combined benefit has heterogeneous sub-criteria, which requires more explanation in the CER.

3KX + 8PL + 0ZC (care pathway)​

These three share the same causal mechanism: the device improves clinical decision quality → HCPs route patients more appropriately → patients access the right level of care faster. Waiting times (3KX), referral adequacy (8PL), and remote access (0ZC) are all downstream expressions of the same underlying benefit.

Shared studies support the merge: DAO_Derivación_O_2022 covers both 3KX and 8PL; COVIDX_EVCDAO_2022 covers 3KX and 0ZC. The evidence base is compatible.

MDR Article 2(53) explicitly names "positive impact on patient management or public health" as a valid clinical benefit category — care pathway optimisation sits squarely within this definition.

Why 5RB cannot be merged​

5RB is categorically different from all other benefits. It is a measurement benefit (ICC), not a diagnostic accuracy or care pathway benefit. The clinical activity is monitoring, not diagnosis. The patient outcome is treatment adjustment, not correct identification of a condition. Merging it with anything would produce a benefit statement that is incoherent on its face.


Gotchas​

Compound acceptance criteria are not eliminated, only reorganised​

Merging benefits does not reduce the number of acceptance criteria — it groups them. All options require defending the same 7 underlying sub-criteria. The difference is that merged benefits require compound acceptance criterion displays (multiple sub-criteria per benefit row), which the current component system may need to support explicitly.

7GH and 9VW use different acceptance criterion forms​

7GH uses a relative improvement threshold (≥+15% Top-1 accuracy over baseline). 9VW uses an absolute threshold (≥54% Top-1 accuracy). They cannot be collapsed into a single criterion — the merged benefit requires two sub-criteria with different forms. This needs to be clear in both the component display and the CER text.

Burying 1QF (Options C) weakens the safety story​

Malignancy detection is the benefit with the highest clinical consequence if the device fails. BSI reviewers are likely to expect it called out at the top level of the benefit structure, because it maps directly to patient safety (delayed cancer diagnosis = serious harm). If it is merged into a broad diagnostic accuracy benefit, the cancer detection claim becomes a sub-item rather than a named clinical benefit — which could be interpreted as downplaying safety relevance. The mitigation is ensuring the malignancy sub-criterion is prominently displayed in the acceptance criteria table, but this is a risk to be aware of.

AUC and Top-1 accuracy are not comparable metrics​

If 1QF is merged with 7GH/9VW, the combined benefit contains both Top-1 accuracy claims and AUC claims. These measure fundamentally different things and cannot be pooled into a single aggregate value. The CER must explain why both metrics are appropriate within the same benefit and how each maps to the benefit statement. This is additional argumentative burden that Option B avoids.

The 0ZC filter logic is complex​

0ZC uses a special OR logic in the filter criteria: claims qualify if their metric contains "remote" OR their domain is "Expert consensus." Some claims (WL4, WOI) are intentionally dual-assigned to both 3KX and 0ZC. When merging 3KX+8PL+0ZC, this dual-assignment needs to be resolved — the merged benefit should absorb both without duplication. Verify the filter criteria logic in clinicalBenefits.ts before implementing.

Any merge requires updating all downstream references​

Clinical benefit IDs (7GH, 9VW, 3KX, 8PL, 0ZC) are referenced in the IFU, CER, performance claims component, and potentially other QMS documents. Removing a benefit ID means finding and updating every reference. The component system handles assignment programmatically, but any MDX file that hard-codes a benefit ID will break.


Trade-offs​

Option AOption BOption C
Benefits543
Malignancy visible at top levelYesYesBuried in Benefit 1
Benefits with compound criteria122
BSI scrutiny riskLowestLowMedium
Previous
Item 0 — AI Context
Next
Question
  • Rationale for choosing Option C over Option B
    • 1. Architectural honesty: 1QF is not a separate algorithm
    • 2. One benefit, two metrics — not a problem
    • 3. The safety signal is preserved
    • 4. Option C is more responsive to Erin's simplification request
    • 5. FDA alignment is not relevant here
  • Option A — 5 benefits
  • Option B — 4 benefits
  • Option C — 3 benefits
  • Rationale for each merger
    • 7GH + 9VW (diagnostic accuracy, general + rare)
    • 7GH + 9VW + 1QF (add malignancy)
    • 3KX + 8PL + 0ZC (care pathway)
    • Why 5RB cannot be merged
  • Gotchas
    • Compound acceptance criteria are not eliminated, only reorganised
    • 7GH and 9VW use different acceptance criterion forms
    • Burying 1QF (Options C) weakens the safety story
    • AUC and Top-1 accuracy are not comparable metrics
    • The 0ZC filter logic is complex
    • Any merge requires updating all downstream references
  • Trade-offs
All the information contained in this QMS is confidential. The recipient agrees not to transmit or reproduce the information, neither by himself nor by third parties, through whichever means, without obtaining the prior written permission of Legit.Health (AI Labs Group S.L.)