R-TF-015-006 Clinical investigation report

Research Title

Simulated-use multi-reader multi-case (MRMC) investigation of Legit.Health Plus (hereinafter "the device"): effect on the top-1 diagnostic accuracy and referral decision-making of primary care physicians on a curated set of anonymised dermatological images.

Nature and positioning of the evidence

This investigation is a simulated-use MRMC reader study performed entirely on retrospective, fully anonymised dermatological images sourced from public dermatological atlases and openly accessible public sources. No patients were recruited, no patient-identifiable data were processed, and no diagnostic or therapeutic intervention was performed on any patient as a consequence of the investigation.

Under MDCG 2020-6 Appendix III the resulting evidence is Rank 11 (simulated-use reader study on retrospective images); it is distinct from clinical data on real patients within the meaning of MDR Article 2(48). Per MDCG 2020-1 §4.4 it contributes Pillar 3 Clinical Performance supporting evidence — measuring the clinician's diagnostic decision-making when using the device's Top-5 prioritised differential view, together with the malignancy-risk indicator and referral recommendation — at a lower rank than the prospective real-patient investigations that carry the primary Pillar 3 weight. Pillar 2 (algorithm-level analytical performance across the 346 ICD-11 categories at API level) is evidenced independently through the device verification-and-validation records and the published severity-validation manuscripts; it is not the subject of this investigation. Pillar 1 (Valid Clinical Association literature anchoring diagnostic accuracy to patient outcomes) is documented in the State of the Art (R-TF-015-011).

Product Identification

	Information
Device name	Legit.Health Plus (hereinafter, the device)
Model and type	NA
Version	1.1.0.0
Basic UDI-DI	8437025550LegitCADx6X
Certificate number (if available)	MDR 000000 (Pending)
EMDN code(s)	Z12040192 (General medicine diagnosis and monitoring instruments - Medical device software)
GMDN code	65975
EU MDR 2017/745	Class IIb
EU MDR Classification rule	Rule 11
Novel product (True/False)	TRUE
Novel related clinical procedure (True/False)	TRUE
SRN	ES-MF-000025345

Throughout this document, references to "the device" refer to the investigational product identified above.

Device version under investigation and bridging to the CE-marked release

The investigation was conducted using device version v1.1.0.0, which is the only version placed on the market and the version to which the present technical documentation applies. No intermediate development build was used during the conduct of the investigation. The bridging between the investigation-version and the CE-marked release is therefore an identity bridge (no algorithm, model, UI or training-data changes); no clinical-relevance assessment is required. The device-version statement has been reviewed and signed off by the PRRC.

Promoter Identification and Contact

	Manufacturer data
Legal manufacturer name	AI Labs Group S.L.
Address	Street Gran Vía 1, BAT Tower, 48001, Bilbao, Bizkaia (Spain)
SRN	ES-MF-000025345
Person responsible for regulatory compliance	Alfonso Medela, Saray Ugidos
E-mail	office@legit.health
Phone	+34 638127476
Trademark	Legit.Health
Authorized Representative	Not applicable (manufacturer is based in EU)

Identification of sponsors

Sponsor institution: Instituto de Investigación Sanitaria Puerta de Hierro — Fundación de Investigación Biomédica del Hospital Universitario Puerta del Hierro de Majadahonda, C/ Joaquín Rodrigo 2, 28222 Majadahonda, Madrid, Spain. The sponsor institution hosted the investigation and provided governance oversight; reader-participants were primary care physicians affiliated with the Las Rozas Healthcare Centre, which operates within the same regional health district as the sponsor institution.

Identification of the Clinical Investigation Plan (CIP)

	CIP
Title of the clinical investigation	Multi-Reader Multi-Case Study Assessing the Impact of Legit.Health Plus on the Diagnostic Accuracy and Referral Decision-Making of Primary Care Physicians for Skin Lesions.
Device under investigation	Legit.Health Plus
Protocol version	Version 1.0
Date	2024-07-04
Protocol code	LEGIT.HEALTH_PH_2024
Sponsor	Instituto de Investigación Sanitaria Puerta de Hierro
Coordinating Investigator	Dr. Gastón Roustán Gullon
Principal Investigator(s)	Dr. Gastón Roustán Gullon
Investigational site(s)	This study was conducted remotely by sending the images to the participating professionals.
Ethics Committee	This study does not require an Ethics Committee approval due to the nature of the images used. The images employed in this study are completely anonymized medical images sourced from public dermatological atlases and freely available public sources. These images are not derived from identifiable patients, and their anonymization makes patient recognition impossible. As these images constitute non-personal data, they do not fall under the scope of regulations requiring ethics committee approval for studies based on fully anonymized, publicly available data.

Trial Registrations

• ClinicalTrials.gov (NCT): NCT07428941
• EMA RWD Catalogue (EUPAS): EUPAS1000000644

Public Access Database

Please note that the database used in this study is not publicly accessible due to privacy and confidentiality considerations.

Research Team

Principal investigator

• Dr. Gastón Roustán Gullón

Collaborators — Clinical Staff (Primary Care Physicians)

• Dr. Esther Minguela
• Dr. Fernando León
• Dr. Ángel Rossell Díaz
• Reader-participants 202407-LR-004 through 202407-LR-009 (pseudonymised; un-pseudonymised mapping retained in the investigator identification-code list under the custody of the Principal Investigator, available on audit request)

Technical Support (Manufacturer)

• Mr. Alfonso Medela
• Mr. Taig Mac Carthy
• Mrs. Alba Rodríguez

Centre

The investigation was conducted remotely through a centralised web-based platform where primary care physicians reviewed anonymised dermatological images. All reader-participants were affiliated with Las Rozas Healthcare Centre, within the regional health district of the sponsor institution (Instituto de Investigación Sanitaria Puerta de Hierro, Majadahonda); image review and diagnostic annotation activities were performed online through the access-controlled platform.

Compliance Statement

The clinical investigation will be conducted according to the Clinical Investigation Plan (CIP) and other applicable guidances and regulations. This includes compliance with:

• The ethical principles originating from the World Medical Association's Declaration of Helsinki
• Harmonized standard UNE-EN ISO 14155:2020
• Regulation (EU) 2017/745 on medical devices (MDR), including the applicable General Safety and Performance Requirements (GSPR) as outlined in Annex I, and the requirements of Annex XV (Chapter I and Chapter II, Section 3)
• Harmonized standard UNE-EN ISO 13485:2016
• MDCG 2024-3 for its structural and content expectations, MDCG 2021-8 concerning application requirements, and MDCG 2020-10/1 Rev 1 for safety reporting timelines and definitions
• Regulation (EU) 2016/679 (GDPR)
• Spanish Organic Law 3/2018 on the Protection of Personal Data and guarantee of digital rights.

All data processing within the device is carried out in accordance with the highest standards of data protection and privacy. Patient information is managed in an encrypted manner to ensure confidentiality and security.

The research team assumes the role of Data Controller, responsible for the collection and management of study data. Legit.Health acts as the Data Processor and is not involved in the processing of patient data.

The storage and transfer of data comply with European data protection regulations. At the conclusion of the study, all information stored in the device will be permanently and securely deleted.

The device employs robust technical and organizational security measures to safeguard personal data against unauthorized access, alteration, loss, or processing.

Report Date

October 13th, 2024

Report author(s)

The full name, the ID and the signature for the authorship, as well as the approval process of this document, can be found in the verified commits at the repository. This information is saved alongside the digital signature, to ensure the integrity of the document.

Table of contents

Table of contents

• Research Title
  • Nature and positioning of the evidence
• Product Identification
  • Device version under investigation and bridging to the CE-marked release
• Promoter Identification and Contact
• Identification of sponsors
• Identification of the Clinical Investigation Plan (CIP)
  • Trial Registrations
• Public Access Database
• Research Team
  • Principal investigator
  • Collaborators — Clinical Staff (Primary Care Physicians)
  • Technical Support (Manufacturer)
  • Centre
• Compliance Statement
• Report Date
• Report author(s)
• Table of contents
• Abbreviations and Definitions
• Summary
  • Title
  • Introduction
  • Objectives
    • Primary objective
    • Secondary objectives
    • Acceptance criteria
  • Population
  • Sample size
  • Design and Methods
    • Design
    • Number of reader-participants
    • Initiation and completion dates
    • Duration
    • Methods
  • Results
  • Conclusions
• Introduction
• Material and methods
  • Product Description
  • Clinical Investigation Plan
    • Objectives
    • Design (type of research, assessment criteria, methods, active group, and control group)
    • Ethical considerations
    • Data Quality Assurance
    • Reader Population
    • Statistical Analysis
    • Performance claims cross-reference
• Results
  • Initiation and Completion Dates
  • Reader and Device Management
  • Image-Case Demographics (n = 30)
  • Reader-Participant Characteristics (n = 9)
  • Clinical Investigation Plan (CIP) Compliance
  • Analysis
    • Primary Analysis — confirmatory
    • Exploratory decomposition of device-attributable decision changes
    • Exploratory per-pathology analysis
    • Referral
    • Remote consultations
    • Adverse Events and Adverse Reactions to the Product
    • Product Deficiencies
    • Paediatric age-band subgroup — exploratory, non-evaluable for confirmatory claims
• Discussion and Overall Conclusions
  • Regulatory positioning of this evidence
  • Clinical Performance, Efficacy, and Safety
    • Acceptance-criteria pass/fail summary
  • Conclusions
  • Potential clinical-system implications (inferential, anchored to external literature)
• References
  • Implications for Future Research
  • Limitations of Clinical Research
  • Ethical Aspects of Clinical Research
• Investigators and Administrative Structure of Clinical Research
  • Brief Description
  • Investigators
    • Principal investigator
    • Collaborating Investigators (Clinical Staff — Primary Care Physicians)
  • Technical Support (Manufacturer)
  • External Organisation
  • Sponsor and Monitor
• Report Annexes

Abbreviations and Definitions

• AE: Adverse Event
• AEMPS: Spanish Agency of Medicines and Medical Devices
• AEP: Adverse Reaction to Product
• AUC: Area Under the ROC Curve
• CAD: Computer-Aided Diagnosis
• CMD: Data Monitoring Committee
• CIP: Clinical Investigation Plan
• CUS: Clinical Utility Questionnaire
• DLQI: Dermatology Quality of Life Index
• GCP: Standards of Good Clinical Practice
• ICH: International Conference of Harmonization
• IFU: Instructions For Use
• IRB: Institutional Review Board
• N/A: Not Applicable
• NCA: National Competent Authority
• PI: Principal Investigator
• PPV: Positive Predictive Value
• NPV: Negative Predictive Value
• SAE: Serious Adverse Events
• SAEP: Serious Adverse Event to Product
• SUAEP: Serious and Unexpected Adverse Event to the Product
• SUS: System Usability Scale

Summary

Title

Simulated-use multi-reader multi-case (MRMC) investigation of the device: effect on the top-1 diagnostic accuracy and referral decision-making of primary care physicians on a curated set of anonymised dermatological images.

Introduction

Dermatological conditions represent a significant portion of primary care consultations, constituting approximately 5% of all visits. Discrepancies between diagnoses made by primary care physicians and dermatologists remain substantial, with concordance rates between 57% and 65.52%; these discrepancies can contribute to misdiagnoses and incorrect referrals that affect the quality of life of patients with skin pathologies (the anchoring literature for this Valid Clinical Association is summarised in R-TF-015-011 State of the Art).

This investigation assessed, under simulated-use conditions on a curated set of anonymised dermatological images sourced from public dermatological atlases and openly accessible public sources, whether use of the device's Top-5 prioritised differential view (together with the malignancy-risk indicator and the referral recommendation) improves the top-1 diagnostic accuracy of primary care physicians relative to their unaided reads on the same image set. Nine primary care physicians each reviewed 30 anonymised dermatological images, first unaided and then with the device. The investigation was conducted following applicable data-protection and research-ethics principles; the sponsor's determination of non-applicability of biomedical-research law and ethics-committee review is documented in Annex E.

Objectives

Primary objective

• To demonstrate that the device, used in its Top-5 prioritised differential-view configuration, increases the top-1 diagnostic accuracy of primary care physicians on a curated, anonymised image set representing multiple dermatological conditions.

Secondary objectives

• To characterise the proportion of cases that the practitioner considers does not require referral to dermatology when aided by the device.
• To characterise the proportion of cases that the practitioner considers can be managed through remote consultation when aided by the device.

Acceptance criteria

• top-1 accuracy equal to or greater than 7.00%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 63.70%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 46.12%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 14.30%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 72.93%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 68.55%.(User Group: Primary care practitioners)
• specificity equal to or greater than 11.88%.(User Group: Primary care practitioners)
• specificity equal to or greater than 77.11%.(User Group: Primary care practitioners)
• specificity equal to or greater than 78.01%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 6.93%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 5.56%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 14.30%.(User Group: Primary care practitioners)
• sensitivity greater than 22.22%.(User Group: Primary care practitioners)
• specificity equal to or greater than 11.88%.(User Group: Primary care practitioners)
• specificity greater than 22.22%.(User Group: Primary care practitioners)
• reduction in the number of days equal to or greater than 40.00%.(User Group: Primary care practitioners)
• increase in patients that can be managed remotely equal to or greater than 40.00%.(User Group: Primary care practitioners)

Population

The reader population for this investigation consists of board-certified primary care physicians without prior regular exposure to the device. Nine reader-participants were enrolled and completed the investigation.

Sample size

The investigation was powered to detect a pre-specified absolute improvement of at least 10 percentage points (pp) in pooled top-1 diagnostic accuracy between unaided and device-aided reads on the same image set, using McNemar's test for paired proportions at a two-sided significance level of 0.05. The target sample of 9 reader-participants each reviewing 30 anonymised dermatological images yields 270 paired per-observation reads; for a paired-binary design with an expected baseline accuracy of approximately 65% this provides power ≥ 80% to detect the pre-specified 10 pp absolute improvement.

The reader cohort size was chosen to contain inter-reader variability while remaining practically feasible under simulated-use conditions; the image set size per reader was chosen to represent the dermatological conditions of interest within a single reader-session window. Per-pathology, per-age-band and per-specialty analyses were pre-specified as exploratory / hypothesis-generating; the investigation is not powered for confirmatory per-stratum conclusions.

Design and Methods

Design

The investigation proceeded as follows.

Reader recruitment and image presentation

A centralised, access-controlled web-based platform served as the electronic Case Report Form (eCRF) for image presentation and data capture. The UI surface under evaluation — the Top-5 prioritised differential view together with the malignancy-risk indicator and the referral recommendation — is the integration surface the device mandates on any integrator; the Pillar 3 Clinical Performance claim supported by this investigation applies to that UI configuration. Each reader-participant logged in with individual credentials and was presented, for each case, with the following pre-specified question set:

1. Based on the provided image, what diagnosis do you consider most appropriate? This question was accompanied by anamnesis inquiries regarding allergies, ongoing treatments, and other relevant medical history.
2. Considering both the image and the information provided by the device, what diagnosis do you now consider most appropriate? At this step, the same information from question 1 was supplemented with the device's Top-5 prioritised differential diagnoses and their respective confidence levels.
3. Based on the information provided by the device, does this patient require a dermatology referral? At this step, the malignancy-risk indicator and the referral recommendation were additionally provided.
4. Based on the information provided by the device, can diagnosis and treatment be managed remotely? No additional information was provided at this stage.

Each participant was presented with 30 cases or images to review. These images had been previously confirmed by dermatologists and by anatomical pathology for cases of skin cancer. The conditions were distributed as follows:

Condition	ICD-11 code	Number of images
Atypical melanocytic nevus	2F20.1	2
Melanocytic nevus	2F20	3
Melanoma	2C30	5
Basal cell carcinoma	2C32	3
Urticaria	EB05	5
Pustular psoriasis	EA90.4	2
Actinic keratosis	EK90.0	2
Plaque psoriasis	EA90.0	3
Hidradenitis suppurativa	ED92.0	5

All reader responses were recorded in the secure central database and, upon database lock, exported in a controlled format for analysis. The pre-specified statistical analysis was implemented in a deterministic, version-controlled analytics environment maintained by the manufacturer; statistical tests (including McNemar's test for the primary endpoint) and confidence intervals were computed as specified in the CIP §Statistical analysis. Atypical melanocytic nevus (ICD-11 2F20.1) and melanocytic nevus (ICD-11 2F20) were evaluated together under a single pooled nevus category (ICD-11 extension XH4L78 Pigmented Nevus) and are referred to in the following sections as "nevus".

Number of reader-participants

A total of 9 primary care physicians were recruited and completed the investigation.

Initiation and completion dates

• Initiation: 4 June 2024
• Completion: 13 September 2024

Duration

The investigation spanned approximately three months, from the first reader session on 4 June 2024 to database lock on 13 September 2024. The planned duration per the CIP was approximately four months; the investigation completed approximately one month earlier than planned. This variance is documented as CIP deviation in §Clinical Investigation Plan (CIP) Compliance below.

Methods

The investigation applied a prospective, simulated-use, multi-reader multi-case (MRMC) self-controlled design to evaluate whether the use of the device improved the top-1 diagnostic accuracy of primary care physicians on a curated set of 30 anonymised dermatological images. Each reader-participant served as their own comparator, first providing diagnoses unaided and subsequently providing diagnoses with the device's Top-5 prioritised differential view, malignancy-risk indicator and referral recommendation on the same image set. The investigation included 9 reader-participants. Data collection captured the per-read diagnosis together with the referral decision and remote-consultation feasibility assessment.

Results

Nine reader-participants reviewed the full 30-image set, yielding 270 paired per-observation reads.

On the pre-specified primary endpoint (pooled per-observation paired analysis), the device produced an absolute improvement of 18.15 percentage points in top-1 diagnostic accuracy. All 95% confidence intervals below are reported using the Wilson score method; Newcombe 95% CIs for the paired difference are reported in the Analysis section.

• Top-1 diagnostic accuracy: 63.70% (95% CI 57.9–69.2%) unaided → 81.85% (95% CI 76.8–85.9%) aided; absolute improvement +18.15 pp (McNemar p < 0.001).
• Sensitivity: 68.55% (95% CI 62.7–73.9%) unaided → 83.15% (95% CI 78.1–87.2%) aided; absolute improvement +14.60 pp.
• Specificity: 78.01% (95% CI 72.7–82.7%) unaided → 89.91% (95% CI 85.8–92.9%) aided; absolute improvement +11.90 pp.

Per-pathology analyses (2–5 images per pathology) are pre-specified as exploratory / hypothesis-generating and are reported in the Analysis section; no per-pathology confirmatory benefit claim is made on the basis of this investigation.

On the secondary endpoints: 48.89% of cases were judged by the aided reader not to require specialist referral, and 60.74% of cases were judged amenable to remote consultation.

Conclusions

The pre-specified primary endpoint was met: use of the device's Top-5 prioritised differential view increased pooled top-1 diagnostic accuracy among primary care physicians from 63.70% to 81.85% (absolute improvement +18.15 pp; McNemar p < 0.001) on the curated anonymised image set. Under MDCG 2020-6 Appendix III this is Rank 11 supporting evidence; per MDCG 2020-1 §4.4 it contributes Pillar 3 Clinical Performance supporting evidence — that the clinician, using the device's Top-5 prioritised differential view, makes measurably more accurate top-1 diagnostic decisions than without the device on the curated image set.

Diagnostic accuracy is a surrogate endpoint. The patient-benefit chain (earlier correct treatment, reduced disease progression, more efficient referral pathways) is indirect and is anchored by published Valid Clinical Association literature summarised in R-TF-015-011 State of the Art; it is not demonstrated by this investigation in isolation and is assessed at Clinical Evaluation level in R-TF-015-003.

Per-pathology analyses were exploratory and under-powered; no per-pathology confirmatory benefit claim is made on the basis of this investigation. Claims concerning malignancy-specific performance, Fitzpatrick V/VI performance, paediatric performance, or real-world primary-care performance are not made on the basis of this investigation alone and are addressed by dedicated evidence within the Clinical Evaluation and by the PMCF Plan (R-TF-007-002).

Secondary observations include a 48.89% proportion of aided-read cases judged not to require specialist referral and a 60.74% proportion judged amenable to remote consultation; these are descriptive reader-judgement outcomes under simulated-use conditions and do not, by themselves, demonstrate a change in real-world referral pathways or teleconsultation workflows.

Introduction

Dermatological conditions represent a significant portion of primary care consultations, constituting approximately 5% of all visits. Discrepancies between diagnoses made by primary care physicians and dermatologists remain substantial, with concordance rates between 57% and 65.52%, and the limited availability of dermatologists (particularly in rural settings) further complicates patient access. This clinical gap motivates interest in diagnostic-decision-support tools that can operate within the primary-care consultation.

Under MDCG 2020-6 Appendix III the resulting evidence from this investigation is Rank 11 (simulated-use reader study on retrospective images); per MDCG 2020-1 §4.4 it contributes Pillar 3 Clinical Performance supporting evidence — measuring the clinician's diagnostic decision-making when using the device's Top-5 prioritised differential view on a curated anonymised image set. The investigation does not, and is not intended to, evidence Pillar 2 (algorithm-level analytical performance across the 346 ICD-11 categories at API level, which is covered by the device's verification-and-validation records and the published severity-validation manuscripts) or Pillar 1 (Valid Clinical Association literature anchoring diagnostic accuracy to patient outcomes, which is documented in R-TF-015-011 State of the Art). It measures the change in top-1 diagnostic accuracy of primary care physicians attributable to use of the device's Top-5 prioritised differential view on the curated image set, and nothing more; real-world patient-outcome, triage, teledermatology and healthcare-economics claims are explicitly out of scope.

Material and methods

Product Description

This section contains a short summary of the device. A complete description of the intended purpose, including device description, can be found in the record Legit.Health Plus description and specifications.

Product description

The device is a computational software-only medical device leveraging computer vision algorithms to process images of the epidermis, the dermis and its appendages, among other skin structures. Its principal function is to provide a wide range of clinical data from the analyzed images to assist healthcare practitioners in their clinical evaluations and allow healthcare provider organisations to gather data and improve their workflows.

The generated data is intended to aid healthcare practitioners and organizations in their clinical decision-making process, thus enhancing the efficiency and accuracy of care delivery.

The device should never be used to confirm a clinical diagnosis. On the contrary, its result is one element of the overall clinical assessment. Indeed, the device is designed to be used when a healthcare practitioner chooses to obtain additional information to consider a decision.

Intended purpose

The device is a computational software-only medical device intended to support health care providers in the assessment of skin structures, enhancing efficiency and accuracy of care delivery, by providing:

• quantification of intensity, count, extent of visible clinical signs
• interpretative distribution representation of possible International Classification of Diseases (ICD) categories.

Intended previous uses

No specific intended use was designated in prior stages of development.

Product changes during clinical research

The device maintained a consistent performance and features throughout the entire clinical research process. No alterations or modifications were made during this period.

Clinical Investigation Plan

Objectives

To demonstrate that the device, used in its Top-5 prioritised differential-view configuration, increases the top-1 diagnostic accuracy of primary care physicians on a curated, anonymised image set representing multiple dermatological conditions.

Acceptance criteria

• top-1 accuracy equal to or greater than 7.00%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 63.70%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 46.12%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 14.30%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 72.93%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 68.55%.(User Group: Primary care practitioners)
• specificity equal to or greater than 11.88%.(User Group: Primary care practitioners)
• specificity equal to or greater than 77.11%.(User Group: Primary care practitioners)
• specificity equal to or greater than 78.01%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 6.93%.(User Group: Primary care practitioners)
• top-1 accuracy equal to or greater than 5.56%.(User Group: Primary care practitioners)
• sensitivity equal to or greater than 14.30%.(User Group: Primary care practitioners)
• sensitivity greater than 22.22%.(User Group: Primary care practitioners)
• specificity equal to or greater than 11.88%.(User Group: Primary care practitioners)
• specificity greater than 22.22%.(User Group: Primary care practitioners)
• reduction in the number of days equal to or greater than 40.00%.(User Group: Primary care practitioners)
• increase in patients that can be managed remotely equal to or greater than 40.00%.(User Group: Primary care practitioners)

The primary confirmatory endpoint is the paired absolute improvement of ≥ 10 pp in pooled top-1 diagnostic accuracy across the full image set (pre-specified per CIP §Justification of acceptance thresholds; thresholds derived from R-TF-015-011 State of the Art). All other thresholds are pre-specified as secondary / supporting; stratum-level comparisons are pre-specified as exploratory / hypothesis-generating.

Design (type of research, assessment criteria, methods, active group, and control group)

The investigation is a prospective, simulated-use, multi-reader multi-case (MRMC) self-controlled reader study. There is no separate active or control group: each reader-participant serves as their own comparator, first diagnosing each image unaided and then re-diagnosing the same image with the device's Top-5 prioritised differential view, malignancy-risk indicator and referral recommendation. The assessment criteria are the paired change in top-1 diagnostic accuracy, the aided-read referral-decision distribution, and the aided-read remote-consultation-feasibility distribution, all captured through a secure, access-controlled web-based eCRF platform.

Ethical considerations

This study adhered to international Good Clinical Practice (GCP) guidelines, the Declaration of Helsinki in its latest amendment, and applicable international and national regulations. As applicable, approval from the relevant Ethics Committee was obtained prior to the initiation of the study. When applicable, modifications to the protocol were reviewed and approved by the Principal Investigator (PI) and subsequently evaluated by the Ethics Committee before subjects were enrolled under a modified protocol.

This study was conducted in compliance with European Regulation 2016/679, of 27 April, concerning the protection of natural persons with regard to the processing of personal data and the free movement of such data (General Data Protection Regulation, GDPR), and Organic Law 3/2018, of 5 December, on the Protection of Personal Data and the guarantee of digital rights. In accordance with these regulations, no data enabling the personal identification of participants was collected, and all information was managed securely in an encrypted format.

Participants were informed both orally and in writing about all relevant aspects of the study, with the information being tailored to their level of understanding. They were provided with a copy of the informed consent form and the accompanying patient information sheet. Adequate time was given to patients to ask questions and fully comprehend the details of the study before providing their consent.

The PI was responsible for the preparation of the informed consent form, ensuring it included all elements required by the International Conference on Harmonisation (ICH), adhered to current regulatory guidelines, and complied with the ethical principles of GCP and the Declaration of Helsinki.

The original signed informed consent forms were securely stored in a restricted access area under the custody of the PI. These documents remained at the research site at all times. Participants were provided with a copy of their signed consent form for their records.

Data Quality Assurance

The Principal Investigator is responsible for reviewing and approving the protocol, signing the Principal Investigator commitment, guaranteeing that the persons involved in the centre will respect the confidentiality of patient information and protect personal data, and reviewing and approving the final study report together with the sponsor. All the clinical members of the research team assess the eligibility of the patients in the study, inform and request written informed consent, collect the source data of the study in the clinical record and transfer them to the Data Collection Notebook (DCN) or Data Collection Forms (CRF).

Reader Population

The investigation enrolled primary care physicians to review anonymised dermatological images and record their diagnostic decisions. Nine reader-participants were enrolled and completed the investigation; each reviewed the full set of 30 images, first unaided and subsequently with the device.

Inclusion criteria

• Board-certified primary care physicians, regardless of years of professional experience, without prior regular exposure to the device.
• High-quality dermatological images sourced from public dermatological atlases and openly accessible public sources, with confirmed reference diagnosis.

Exclusion criteria

• Images of insufficient technical quality (blurred, poorly framed, or with insufficient lesion coverage) that cannot be properly analysed.

Statistical Analysis

The pre-specified primary analysis tested the paired change in pooled top-1 diagnostic accuracy across the full image set using McNemar's test for paired proportions at a two-sided significance level of 0.05, with Wilson 95% confidence intervals for each proportion and Newcombe 95% confidence intervals for the paired difference. Secondary analyses of the referral-decision and remote-consultation-feasibility distributions used descriptive proportions with Wilson 95% CIs and tested the association between the two decisions with Pearson's chi-squared test. Per-pathology, per-age-band and per-specialty analyses were pre-specified as exploratory / hypothesis-generating; no multiplicity correction was applied across exploratory analyses, consistent with their hypothesis-generating purpose. Missing reader responses were excluded from the paired analysis for that reader-image pair (complete-case handling); zero-cell strata used exact confidence intervals and were flagged as non-evaluable for confirmatory purposes. Inter-reader agreement on the primary endpoint was characterised descriptively. All analyses were executed in a deterministic, version-controlled analytics environment maintained by the manufacturer.

Performance claims cross-reference

Error: Could not determine study code. Provide studyCode or folderSlug prop, or ensure this component is used within an Investigation document with a registered folder slug.

Results

Initiation and Completion Dates

• Initiation: 4 June 2024
• Completion: 13 September 2024

Reader and Device Management

A total of 9 primary care physicians (reader-participants) took part in the investigation. Each reader-participant was presented with the full 30-image set and, for each image, recorded an unaided diagnosis followed by a device-aided diagnosis together with a referral decision and a remote-consultation feasibility judgement. The device — a software-only, web-accessible medical device — was provisioned to reader-participants through individual access-controlled credentials on the secure eCRF platform; version identification (v1.1.0.0), session logging, and complete audit-trail records were maintained for the duration of the investigation and retained under the custody of the Principal Investigator.

Image-Case Demographics (n = 30)

The characteristics reported in this section describe the 30 anonymised image cases presented to the reader-participants (i.e., the patient cases depicted in the images sourced from public atlases), not the 9 reader-participants. Age and phototype metadata were extracted from the source atlas records; where a source did not report a given attribute, the image was still included in the pooled primary analysis as pre-specified.

Gender (image-case level): Men 14 (46.7%), Women 16 (53.3%).

Age Group	Count	Percentage
Newborn (birth to 1 month)	1	3.3%
Infant (1 month to 2 years)	1	3.3%
Child (2 to 12 years)	0	0.0%
Adolescent (12 to 21 years)	0	0.0%
Adult (22 to 64 years)	21	70.0%
Elderly (≥ 65 years)	7	23.3%

Fitzpatrick Phototype	Count	Percentage
Phototype I	10	33.3%
Phototype II	12	40.0%
Phototype III	7	23.3%
Phototype IV	1	3.3%
Phototype V	0	0.0%
Phototype VI	0	0.0%

The image set under-represents darker Fitzpatrick phototypes (V and VI absent, IV present in a single case) and the paediatric age bands (2–12 years and 12–21 years absent; newborn and infant bands each represented by a single case). Accordingly, this investigation does not, on its own, support performance claims on Fitzpatrick V/VI, on the 2–12 year or 12–21 year age bands, or on pooled paediatric use; these coverage gaps are addressed by dedicated evidence within the Clinical Evaluation (R-TF-015-003) and by the PMCF Plan (R-TF-007-002).

Reader-Participant Characteristics (n = 9)

Nine primary care physicians affiliated with Las Rozas Healthcare Centre acted as reader-participants. They are identified under pseudonymised codes 202407-LR-001 to 202407-LR-009 in the exported analysis dataset; the un-pseudonymised mapping is retained in the investigator identification-code list under the custody of the Principal Investigator and is available on audit request. Reader-participants were board-certified primary care physicians without prior regular exposure to the device.

Clinical Investigation Plan (CIP) Compliance

The investigation was conducted in accordance with the CIP. The following CIP deviations are recorded:

Ref	Description	Impact assessment	Action
D-01	Investigation duration: approximately three months (4 June — 13 September 2024) vs approximately four months planned per CIP §Duration and §Calendar.	No impact on data integrity, analysis population, endpoints, or conclusions; reader workload and session pacing unchanged.	Documented as a minor deviation; no CAPA required.
D-02	Reader cohort size: 9 reader-participants enrolled and retained vs "minimum of 8" stated in early CIP drafts; 9 is the pre-specified target used for powering.	Matches the pre-specified powering basis for the primary endpoint; no impact on pre-specified analyses.	Documented for traceability; no CAPA required.

No other CIP deviations occurred during the conduct of the investigation. No deviations impacted reader or image-case safety, data integrity, analysis population, or pre-specified endpoints.

Analysis

Primary Analysis — confirmatory

The pre-specified primary endpoint was the paired change in pooled top-1 diagnostic accuracy across the full image set (270 paired per-observation reads). Unaided top-1 accuracy was 63.70% (95% Wilson CI 57.9–69.2%); aided top-1 accuracy was 81.85% (95% Wilson CI 76.8–85.9%). The absolute paired improvement was +18.15 pp (95% Newcombe CI for the paired difference 11.9–24.0 pp); McNemar's test for paired proportions gave p < 0.001. The primary endpoint was met (pre-specified threshold ≥ +10 pp absolute improvement).

Exploratory decomposition of device-attributable decision changes

The following decomposition of the 270 paired per-observation reads is reported as an exploratory / hypothesis-generating description of how the device-aided decision related to the unaided decision and the reference diagnosis. It is not a pre-specified endpoint and is not the basis of any confirmatory claim:

• Device reinforced a correct unaided decision (both reads correct) — 69.26% of reads.
• Device improved an incorrect unaided decision (unaided incorrect, aided correct) — 12.96% of reads.
• No change between reads (unaided and aided decisions identical and incorrect) — 14.07% of reads.
• Device-aided decision switched a correct unaided decision to an incorrect one (automation-bias direction) — 3.70% of reads.

The 3.70% device-aided "switch to incorrect" proportion is traced to the risk management record R-TF-013-002 as a known automation-bias direction, is covered by the existing risk controls (IFU warnings on the non-binding nature of the device output, requirement that the clinician remains the decision-maker, and the Top-5 prioritised differential view rather than a single binding answer), and is monitored as part of the PMCF Plan (R-TF-007-002).

Exploratory per-pathology analysis

Per-pathology analyses are pre-specified as exploratory / hypothesis-generating and are under-powered (2–5 images per category). No per-pathology confirmatory benefit claim is made on the basis of this investigation. P-values below are labelled exploratory and were not corrected for multiplicity; they are reported to characterise directional trends only.

Condition	Images	Unaided accuracy (%)	Aided accuracy (%)	Absolute difference (pp)	p-value (exploratory)
Actinic keratosis	2	55.56	83.33	+27.77	0.125
Pustular psoriasis	2	5.56	22.22	+16.66	0.25
Plaque psoriasis	3	96.30	96.30	0.00	1.00
Nevus	5	75.56	77.78	+2.22	1.00
Melanoma	5	86.67	91.11	+4.44	0.69
Urticaria	5	73.33	91.11	+17.78	0.02
Hidradenitis suppurativa	5	64.44	80.00	+15.56	0.04
Basal cell carcinoma	3	91.67	88.89	−2.78	1.00

The directional trend was toward improvement in six of eight pathology categories. Per-pathology p-values did not reach statistical significance for melanoma, basal cell carcinoma, plaque psoriasis, nevus, actinic keratosis or pustular psoriasis; the investigation is not powered for, and does not support, per-pathology confirmatory claims for any of these conditions, including any malignancy-specific benefit claim. Claims relating to these pathologies are addressed at Clinical Evaluation level (R-TF-015-003) using the broader body of clinical evidence, including the published severity-validation manuscripts and the prospective real-patient investigations at Ranks 2–4.

Referral

Aided reads judged 48.89% of cases not to require specialist referral.

Condition	Do not require referral (%)
Nevus	60.00
Melanoma	2.22
Basal cell carcinoma	7.41
Urticaria	88.89
Pustular psoriasis	11.11
Actinic keratosis	33.33
Plaque psoriasis	81.48
Hidradenitis suppurativa	71.11

Remote consultations

Aided reads judged 60.74% of cases to be amenable to remote consultation.

A Pearson's chi-squared test identified a statistically significant association between the aided-read referral decision and the aided-read remote-consultation feasibility judgement at the 5% level; the association pattern is described below:

• 36.67% of the cases do not require a referral and can have follow-up remotely
• 12.22% of the cases do not require a referral but require an in-person appointment
• 24.07% of the cases require a referral and remote consultation
• 27.04% of the cases require a referral in addition to an in-person appointment

Pathology	Can be handled remotely (%)
Nevus	55.56
Melanoma	42.22
Basal cell carcinoma	44.44
Urticaria	75.56
Pustular psoriasis	38.89
Actinic keratosis	61.11
Plaque psoriasis	81.48
Hidradenitis suppurativa	75.56

Adverse Events and Adverse Reactions to the Product

Because the investigation recruited no patients and performed no intervention on patients, patient-level adverse events are not applicable. At the reader-participant level, no adverse events related to the investigational product were observed during the conduct of the investigation. The detection mechanism combined: (i) platform-side error logging and session monitoring, and (ii) a per-session free-text feedback field through which reader-participants could report any issue. This reporting path feeds the post-market surveillance procedure (GP-009 Post-Market Surveillance).

Product Deficiencies

No product deficiencies were logged during the conduct of the investigation. The detection mechanism combined platform-side error logging, session monitoring, and the reader-participant feedback field described above. This reporting path feeds the non-conforming product control procedure and, as applicable, the post-market surveillance procedure (GP-009).

Paediatric age-band subgroup — exploratory, non-evaluable for confirmatory claims

A paediatric age-band subgroup description is reported below for completeness. The paediatric representation in the image set is minimal — one image in the newborn band (birth to 1 month), one image in the infant band (1 month to 2 years), zero images in the child band (2 to 12 years) and zero images in the adolescent band (12 to 21 years) — and both analysable paediatric cases are urticaria. For a single-positive-case subgroup, per-class sensitivity and specificity do not carry independent statistical meaning; accordingly, the subgroup is reported as exploratory and is flagged as non-evaluable for confirmatory purposes. The investigation does not support confirmatory paediatric performance claims; paediatric coverage is addressed by dedicated evidence at Clinical Evaluation level (R-TF-015-003) and by the PMCF Plan (R-TF-007-002).

Infant band (1 month to 2 years) — n = 1 image, exploratory

Per-reader correct identifications: 4 of 9 unaided, 7 of 9 device-aided (absolute directional change +3 readers, exact 95% CI wide given the very small sample). The subgroup is flagged as non-evaluable for confirmatory purposes.

Newborn band (birth to 1 month) — n = 1 image, exploratory

Per-reader correct identifications for this band are reported in the annex to the locked analysis dataset; the subgroup is flagged as non-evaluable for confirmatory purposes on the same grounds.

Discussion and Overall Conclusions

Regulatory positioning of this evidence

Per MDCG 2020-6 Appendix III this investigation produces Rank 11 evidence (simulated-use reader study on retrospective images); per MDCG 2020-1 §4.4 it contributes Pillar 3 Clinical Performance supporting evidence — measuring the clinician's diagnostic decision-making when using the device's Top-5 prioritised differential view, together with the malignancy-risk indicator and the referral recommendation. Pillar 2 (algorithm-level analytical performance across the 346 ICD-11 categories at API level) is evidenced independently through the device verification-and-validation records and the published severity-validation manuscripts; it is not the subject of this investigation. Pillar 1 (Valid Clinical Association literature anchoring diagnostic accuracy to patient outcomes) is documented in R-TF-015-011 State of the Art.

Clinical Performance, Efficacy, and Safety

Acceptance-criteria pass/fail summary

The table below renders each pre-specified acceptance criterion from R-TF-015-011 (as configured in the performance-claims data source for PH_2024) against the observed value, with pass/fail called explicitly:

Error: Could not determine study code. Provide studyCode or folderSlug prop, or ensure this component is used within an Investigation document with a registered folder slug.

Conclusions

On the pre-specified primary endpoint, use of the device's Top-5 prioritised differential view increased pooled top-1 diagnostic accuracy among primary care physicians from 63.70% (95% CI 57.9–69.2%) to 81.85% (95% CI 76.8–85.9%) on the curated anonymised image set; absolute paired improvement +18.15 pp, McNemar p < 0.001. The primary endpoint was met.

Diagnostic accuracy is a surrogate endpoint. The patient-benefit chain (earlier correct treatment, reduced disease progression, more efficient referral pathways) is indirect and is anchored by the Valid Clinical Association literature summarised in R-TF-015-011 State of the Art (Pillar 1); it is not demonstrated by this investigation in isolation and is assessed at Clinical Evaluation level in R-TF-015-003. Published literature documents the relationship between timely and accurate primary-care dermatological diagnosis and downstream patient outcomes (Escalé-Besa et al. 2023; Giavina-Bianchi et al. 2020; Maul et al. 2024; Liu et al. 2016), and documents the burden of diagnostic delays in rare dermatological conditions such as hidradenitis suppurativa and pustular psoriasis (Strober et al. 2021; Costanzo et al. 2022; Kokolakis et al. 2019; Willmen et al. 2024); these anchors describe the downstream leg of the surrogate-to-outcome chain that this investigation supports, not outcomes demonstrated by this investigation itself.

Per-pathology analyses were exploratory and under-powered. Per-pathology p-values did not reach statistical significance for melanoma, basal cell carcinoma, plaque psoriasis, nevus, actinic keratosis or pustular psoriasis. No per-pathology confirmatory benefit claim is made on the basis of this investigation; in particular, malignancy-specific benefit claims are not made from this investigation alone and are addressed at Clinical Evaluation level using the broader body of clinical evidence (including the prospective real-patient investigations at Ranks 2–4).

The paediatric age-band subgroup is exploratory and, on sample-size grounds, non-evaluable for confirmatory purposes. Paediatric coverage is addressed by dedicated evidence at Clinical Evaluation level (R-TF-015-003) and by the PMCF Plan (R-TF-007-002).

On the secondary endpoints, 48.89% of aided reads judged the case not to require specialist referral and 60.74% judged the case amenable to remote consultation, with a statistically significant association between the two aided-read decisions. These are descriptive reader-judgement outcomes under simulated-use conditions and do not, by themselves, demonstrate a change in real-world referral pathways or teleconsultation workflows.

Potential clinical-system implications (inferential, anchored to external literature)

The two secondary reader-judgement outcomes — 48.89% of aided reads judged not to require specialist referral, 60.74% judged amenable to remote consultation — are reported here for their bearing on the Pillar 1 Valid Clinical Association chain documented in R-TF-015-011. They are not endpoints that this investigation demonstrates in real-world practice. The published literature characterises the baseline: approximately 66% of patients seen in primary care for dermatological complaints are referred to dermatology (González-López et al. 2019); a substantial proportion of urgent referrals from primary care to dermatology present as benign diagnoses at the specialist visit (Pagani et al. 2023); teledermatology has been associated with substantial reductions in waiting times and increased throughput per dermatologist-hour (Ludwick et al. 2010; Zakaria et al. 2019; Finnane et al. 2017; Giavina-Bianchi et al. 2020); and modelling studies have estimated cost savings associated with reductions in avoidable dermatology consultations (Liu et al. 2016). Within that external context, aided-reader judgement patterns at 48.89% "no referral" and 60.74% "remote consultation feasible" are consistent with, but do not themselves prove, the mechanism by which better primary-care diagnostic accuracy may translate into more efficient referral pathways at system level. Demonstration of real-world referral-pathway and teleconsultation outcomes is not a claim of this investigation; it is addressed at Clinical Evaluation level and by the PMCF Plan.

References

1. Escalé-Besa A, Yélamos O, Vidal-Alaball J, et al. Exploring the potential of artificial intelligence in improving skin lesion diagnosis in primary care. Scientific Reports. 2023 Mar 15;13(1): 4293.

2. Giavina-Bianchi M, Santos AP, Cordioli E. Teledermatology reduces dermatology referrals and improves access to specialists. EClinicalMedicine. 2020 Nov 21;29-30:100641.

3. Maul LV, Jamiolkowski D, Lapides RA, et al. Health Economic Consequences Associated With COVID-19-Related Delay in Melanoma Diagnosis in Europe. JAMA Netw Open. 2024 Feb 5;7(2):e2356479.

4. Liu KJ, Hartman RI, Joyce C, et al. Modeling the Effect of Shared Care to Optimize Acne Referrals From Primary Care Clinicians to Dermatologists. JAMA Dermatol. 2016 Jun 1;152(6):655-60.

5. Strober B, Kotowsky N, Medeiros R, et al. Unmet Medical Needs in the Treatment and Management of Generalized Pustular Psoriasis Flares: Evidence from a Survey of Corrona Registry Dermatologists. Dermatol Ther (Heidelb). 2021 Apr;11(2):529-541.

6. Costanzo A, Bardazzi F, DE Simone C, et al. Pustular psoriasis with a focus on generalized pustular psoriasis: classification and diagnostic criteria. An Italian expert consensus. Ital J Dermatol Venerol. 2022 Dec;157(6):489-496.

7. Kokolakis G, Wolk K, Schneider-Burrus S, et al. Delayed Diagnosis of Hidradenitis Suppurativa and Its Effect on Patients and Healthcare System. Dermatology. 2020;236(5):421-430.

8. Willmen L, Völkel L, Willmen T, et al. The economic burden of diagnostic uncertainty on rare disease patients. BMC Health Serv Res. 2024 Nov 12;24(1):1388.

9. González-López G, Descalzo-Gallego MÁ, Arias-Santiago S, et al. Derivación de pacientes en consulta de dermatología y de teledermatología en España. Estudio DIADERM. Actas Dermo-Sifiliográficas. 2019 Mar 1;110(2):146-52.

10. Pagani K, Lukac D, Olbricht SM, et al. Urgent referrals from primary care to dermatology for lesions suspicious for skin cancer: patterns, outcomes, and need for systems improvement. Archives of dermatological research. 2023 Jul;315(5):1397-400.

11. Ludwick DA, Lortie C, Doucette J, et al. Evaluation of a telehealth clinic as a means to facilitate dermatologic consultation: pilot project to assess the efficiency and experience of teledermatology used in a primary care network. J Cutan Med Surg. 2010 Jan-Feb;14(1):7-12.

12. Zakaria A, Maurer T, Su G, Amerson E. Impact of teledermatology on the accessibility and efficiency of dermatology care in an urban safety-net hospital: A pre-post analysis. J Am Acad Dermatol. 2019 Dec;81(6):1446-1452.

13. Finnane A, Dallest K, Janda M, Soyer HP. Teledermatology for the Diagnosis and Management of Skin Cancer: A Systematic Review. JAMA Dermatol. 2017 Mar 1;153(3):319-327.

14. Naka F, Lu J, Porto A, et al. Impact of dermatology eConsults on access to care and skin cancer screening in underserved populations: A model for teledermatology services in community health centers. J Am Acad Dermatol. 2018 Feb;78(2):293-302.

Implications for Future Research

The findings of this investigation are hypothesis-generating with respect to the real-world Pillar 3 Clinical Performance claims that require Rank 2–4 prospective real-patient evidence. Specifically, questions that are explicitly outside the scope of this Rank-11 simulated-use investigation and are addressed by dedicated evidence at Clinical Evaluation level (R-TF-015-003) and by the PMCF Plan (R-TF-007-002) include:

• Real-world top-1 accuracy in consulting primary-care populations (outside the curated atlas image set).
• Performance on Fitzpatrick phototypes V and VI (not represented in this image set).
• Performance in the 2–12 year and 12–21 year paediatric age bands (not represented in this image set).
• Malignancy-specific performance (melanoma, basal cell carcinoma, actinic keratosis) at pre-specified power.
• Real-world referral-pathway and teleconsultation-workflow outcomes (this investigation measures reader judgements under simulated use, not pathway outcomes).
• Long-term post-deployment diagnostic performance, including drift and continued-use effects.

Continued refinement of the device's algorithms and UI integration remains an engineering question addressed under the change-control framework of the QMS; it is not an open question for this investigation.

Limitations of Clinical Research

The following structural limitations apply to this investigation and constrain the claims that it can support:

1. Rank 11 simulated-use evidence. Under MDCG 2020-6 Appendix III this investigation produces Rank 11 evidence (simulated-use reader study on retrospective images) and is distinct from clinical data on real patients within the meaning of MDR Article 2(48). It contributes Pillar 3 §4.4 supporting evidence only; primary Pillar 3 Clinical Performance evidence is supplied by the prospective real-patient investigations at Ranks 2–4.
2. Curated image set vs real-world primary-care imaging. The image set is sourced from public dermatological atlases with confirmed reference diagnoses; atlas-grade image quality over-represents optimal acquisition conditions relative to typical primary-care phone-camera captures, and this is expected to bias observed accuracy upward relative to real-world deployment.
3. Fitzpatrick phototype coverage. The image set contains no Fitzpatrick V or VI cases and only one Fitzpatrick IV case; this investigation does not support Fitzpatrick V/VI performance claims on its own, and phototype coverage is addressed by dedicated evidence at Clinical Evaluation level.
4. Paediatric subgroup scope. Paediatric representation is minimal (newborn n = 1, infant n = 1, child n = 0, adolescent n = 0); per-age-band subgroup analyses are non-evaluable for confirmatory purposes.
5. Recall bias in a within-subject design. The unaided and device-aided reads evaluate the same image set sequentially without a washout period; the observed difference reflects device impact plus any recall-assisted improvement across reads, with expected net bias upward on accuracy.
6. Per-pathology statistical power. With 2–5 images per pathology category, per-pathology analyses are exploratory and under-powered; no per-pathology confirmatory benefit claim is made.
7. Per-pathology non-significance for malignancies. Per-pathology p-values did not reach statistical significance for melanoma, basal cell carcinoma, plaque psoriasis, nevus, actinic keratosis or pustular psoriasis; no malignancy-specific benefit claim is made on the basis of this investigation alone.
8. Device-aided "switch to incorrect" proportion. A 3.70% proportion of device-aided reads switched a correct unaided decision to an incorrect one (automation-bias direction); this is traced to the risk management record R-TF-013-002, covered by existing risk controls (IFU warnings, Top-5 prioritised differential rather than single binding answer, the clinician remains the decision-maker), and monitored through the PMCF Plan.
9. No real-world validation within this investigation. Real-world patient-outcome, triage, teledermatology and healthcare-economics outcomes are not measured here; they are addressed at Clinical Evaluation level and by the PMCF Plan.
10. Hawthorne effect under simulated use. Readers are aware that they are being observed; this may modulate their behaviour relative to routine practice.
11. Reader-behaviour variability. Variation between readers in how readily they incorporate the device's output affects observed inter-reader variability; this is reported rather than suppressed.

Ethical Aspects of Clinical Research

The investigation is conducted in accordance with the ethical principles of the Declaration of Helsinki to the extent applicable to retrospective studies of anonymised material, and in compliance with Regulation (EU) 2016/679 (GDPR) and Spanish Ley Orgánica 3/2018 on the protection of personal data. No data allowing the personal identification of any individual is included, and all information is managed under appropriate technical and organisational security measures.

The sponsor's determination of non-applicability of biomedical-research law (Ley 14/2007) and of ethics-committee review for this investigation is documented in Annex E (R-TF-015-010) under "Ethics Committee Non-Applicability Determination". The investigation does not recruit patients, does not interfere with any clinical care, and uses only completely anonymised images from public dermatological atlases and openly accessible public sources; it therefore does not fall within the material scope of the biomedical-research framework that would require ethics-committee authorisation. Reader-participants receive comprehensive written and oral information about the investigation and sign a participation agreement with the sponsor; formal informed-consent forms are not required because the investigation is observational and non-interventional with respect to their clinical practice.

The Data Controller for the reader-participant data processed under this investigation is the sponsor institution. The manufacturer provides the validated reader platform as Data Processor under a data-processing agreement and does not process data beyond the scope of that agreement. Storage and handling of reader-participant data are aligned with GDPR and Ley Orgánica 3/2018; at the conclusion of the investigation, reader-session data retained beyond the analytic record are deleted according to the sponsor's data-retention schedule.

Investigators and Administrative Structure of Clinical Research

Brief Description

The investigation was conducted between the participating primary care physicians (reader-participants), the Instituto de Investigación Sanitaria Puerta de Hierro (sponsor institution), and the manufacturer.

Investigators

Principal investigator

• Dr. Gastón Roustán Gullón

Collaborating Investigators (Clinical Staff — Primary Care Physicians)

• Dr. Esther Minguela
• Dr. Fernando León
• Dr. Ángel Rossell Díaz
• Reader-participants 202407-LR-004 through 202407-LR-009 (pseudonymised; un-pseudonymised mapping retained in the investigator identification-code list under the custody of the Principal Investigator, available on audit request)

Technical Support (Manufacturer)

• Mr. Alfonso Medela
• Mr. Taig Mac Carthy
• Mrs. Alba Rodríguez

External Organisation

No additional organisations, beyond those listed above, contributed to the investigation. The investigation was conducted with the collaboration and resources of the specified entities.

Sponsor and Monitor

Sponsor institution: Instituto de Investigación Sanitaria Puerta de Hierro — Fundación de Investigación Biomédica del Hospital Universitario Puerta del Hierro de Majadahonda. Monitoring was performed by the sponsor as described in the CIP §Monitoring plan.

Report Annexes

• Instructions For Use (IFU) v1.1.0.0 was provided to all reader-participants at investigation onboarding and served as the investigational-device orientation document; the IFU is cross-referenced from the ISO 14155 applicability checklist in Annex E (R-TF-015-010).

Signature meaning

The signatures for the approval process of this document can be found in the verified commits at the repository for the QMS. As a reference, the team members who are expected to participate in this document and their roles in the approval process, as defined in Annex I Responsibility Matrix of the GP-001, are:

• Author: Team members involved
• Reviewer: JD-018 Clinical Research Coordinator
• Approver: JD-022 Medical Manager