☰

Clebre

Scaling a sleep apnea diagnostic from clinic to home

Industry
Medical Devices (Sleep Diagnostics)
Headquarters
Warsaw, Poland
Public information as of
January 2026

A4BEE prepared this analysis from publicly available sources. It reflects our own reading of Clebre's published strategy and is not endorsed by, or produced in cooperation with, Clebre. Company website

Strategic priorities

Clebre is commercialising a wireless tracheal acoustic sensor that records breathing sounds and motion at the suprasternal notch and turns them into a clinical sleep study. In a study of 89 patients the sensor reached 96.9 percent supine and 97.0 percent non-supine body-position accuracy against NOX A1 polysomnography (PSG, the standard overnight sleep study used in sleep labs), and 0.92 to 0.94 Pearson correlation for the Apnea-Hypopnea Index (AHI, the events-per-hour metric used to grade sleep apnea severity). Diagnostic sensitivity at the AHI ≥ 15 threshold reached 96 percent with 76 percent specificity, with slight underestimation (about 1.5 to 2.0 events per hour) at AHI ≥ 30.

The immediate priority is taking the device from clinical validation to commercial deployment. Funding to date includes a PLN 5.2 million (about EUR 1.2 million) grant from Poland's National Centre for Research and Development under Priority Axis I of the Smart Growth Operational Programme (2019–2023) and a EUR 2 million seed round for a 25 percent stake, valuing the company at roughly EUR 8 million. The next milestones are EU-MDR 2017/745 certification under the European Medical Device Regulation and FDA market entry through the agency's updated Real-World Evidence (RWE) policies.

Three operational threads run in parallel: a continuous Bluetooth Low Energy (BLE) data pipeline that has to scale from clinical pilots to thousands of home devices while staying GDPR- and HIPAA-compliant; an algorithm lifecycle that needs to keep improving with new clinical data without breaking the existing EU-MDR validation; and a hardware transition from ten-unit prototype batches to ISO 13485 (the international quality-system standard for medical devices) volume production through a contract development and manufacturing organisation (CDMO) partner.

A diagnostic-as-a-service model is the stated strategic direction — paediatric sleep studies, a subcontractor clinic network, and post-market clinical follow-up under EU-MDR — which puts longitudinal evidence collection, multi-jurisdiction regulatory submission work and clean Electronic Health Record (EHR) integration at the centre of the roadmap rather than at the end of it.

Challenges we see

  • Digital Integration

    Scaling BLE data ingestion from clinical pilots to home deployment

    The Clebre sensor streams audio and accelerometer data over Bluetooth Low Energy to a cloud backend, replacing the original SD-card-on-device design that consumed more than 50 percent of battery power. Commercial deployment needs to support thousands of concurrent home sensors with GDPR and HIPAA controls.

    Where home sensors stream continuously and clinicians depend on the data being there, the cloud path needs to be designed for the steady state of thousands of devices from day one rather than retrofitted once the deployment reaches that size.

  • Digital Integration

    Embedding diagnostic reports in hospital EHR workflows

    Sleep study results need to land inside Epic and Cerner (the two dominant hospital Electronic Health Record platforms) so that referring physicians can read them in their existing workflow. The integration standard is HL7 FHIR R4 (Fast Healthcare Interoperability Resources, the current healthcare data-exchange specification) with SMART on FHIR (a security and launch framework for apps that run inside EHR systems) for embedded apps.

    Where diagnostic reports live outside the systems physicians already use, the path from sensor to clinician involves manual transfer; standard interfaces that put the report in front of the clinician shorten that path.

  • Compliance Regulatory

    Maintaining EU-MDR certification while the algorithm evolves

    The deep-learning models for respiratory event detection are intended to improve as new clinical recordings accumulate. EU-MDR 2017/745 treats the algorithm as part of the device and requires post-market clinical follow-up (PMCF) plus continuous safety monitoring. The current device shows slight AHI underestimation at AHI ≥ 30.

    Where a learning algorithm sits inside a certified device, every model change has a regulatory footprint; a version-controlled lifecycle with documented revalidation keeps the improvement and the certification on the same schedule.

  • Operations Manufacturing

    Moving from ten-unit prototypes to volume production

    Phase III produced ten prototype units that successfully recorded 8-hour overnight sessions. Commercial volume requires a contract development and manufacturing organisation partner with ISO 13485 quality management, validated processes for acoustic membrane assembly, BLE module integration and medical-grade patch application, and a qualified supplier base for the acoustic and electronic components.

    Where production batches move from tens to thousands of units, the supply chain, incoming inspection and release testing that worked for prototypes become the constraint; designing them with the partner up front keeps the scale-up on the clinical timeline.

  • Digital Regulatory

    Turning device-generated data into FDA-ready Real-World Evidence

    The FDA's late-2025 policy change accepts appropriately designed Real-World Evidence from de-identified databases of millions of patient records for pre-market submissions, label expansions and post-approval safety evaluations. Clebre's home sensors generate exactly that kind of data if the de-identification, aggregation and analysis pipeline is in place.

    Where device-generated data is treated as a regulatory asset rather than a by-product, the same data set that supports post-market follow-up under EU-MDR also feeds U.S. submissions; designing that data path once serves both.

Opportunities, by urgency and business impact

Each bubble is one opportunity, numbered to match the list below. Further right means it bites sooner; higher means a bigger effect on the business. A bigger bubble means a bigger implementation effort.

Source: A4BEE analysis of public sources

  1. A multi-region cloud platform for continuous sensor data

    The BLE ingestion path was sized for clinical pilots and does not yet have the multi-region redundancy, automated scaling and compliance controls needed for home deployment at scale.

    An explicit cloud architecture with multi-region failover, containerised services that scale with sensor count, and GDPR- and HIPAA-aligned controls lets the diagnostic-as-a-service model operate continuously from clinical pilots through commercial scale without re-platforming.

    • Clebre Sensor product page, accessed 16 January 2026
    • Cloud migration best practices, A4BEE article

  2. FHIR-based EHR integration so reports reach the referring physician

    Sleep study results currently require manual transfer into hospital systems, which adds a step between the sensor and the clinical decision and creates an adoption barrier for physicians who already work inside Epic or Cerner.

    HL7 FHIR R4 interfaces plus a SMART on FHIR application embed the AHI score, body position data and sleep study summary directly into the hospital record, so the report appears in the existing physician workflow.

    • The Only Thing to FHIR is FHIR Itself?, Medium, accessed 16 January 2026
    • Caristix HL7 message viewing and editing documentation

  3. Versioned algorithm lifecycle with documented revalidation

    Deep-learning models for respiratory event scoring must improve with new clinical data while keeping the EU-MDR-validated device classification intact.

    A model registry with training-data lineage, hyperparameter tracking, automated validation against clinical benchmarks and a deployment gate that requires documented revalidation lets the algorithm keep improving on the same regulatory cadence as the device.

    • EU-MDR 2017/745 framework, Medoc, accessed 16 January 2026
    • Mantra Systems webinar: From USA to Europe - Accelerating Your Path to the Medical Device Market

  4. Validated manufacturing bridge through a medical-device CDMO partner

    Commercial volume requires ISO 13485 processes, validated assembly for the acoustic membrane and BLE module, and a qualified supplier network — none of which exists in-house.

    Working with a contract development and manufacturing organisation that operates under ISO 13485, with IQ/OQ/PQ (Installation, Operational and Performance Qualification, the three stages of equipment and process validation under ISO 13485) protocols, incoming inspection and batch release, gives Clebre a manufacturing path that meets EU-MDR submission expectations from first commercial batch.

    • Consonance case study: Clebre — Device for Home Diagnosis of Respiratory Disorders
    • Clebre one-pager (NCBR project documentation), December 2023

  5. A Real-World Evidence pipeline that serves FDA submissions and post-market follow-up

    The FDA's updated RWE policy opens a path to U.S. market entry and label expansion, but only for sponsors whose device-generated data is de-identified, aggregated and analysable at population scale.

    An RWE pipeline that de-identifies home-sensor data at ingestion, aggregates it into a queryable dataset and produces pre-formatted analysis reports aligned with FDA RWE guidance lets Clebre run pre-market submissions, label expansions and EU-MDR post-market clinical follow-up from the same evidence base.

    • FDA eliminates major barrier to using Real-World Evidence in drug and device application reviews, FDA press release
    • FDA removes a major barrier to Real-World Evidence, Pure Clinical, accessed 16 January 2026

What we'd propose

  • Enterprise AI

    Multi-region cloud platform for home sensor data

    A cloud architecture for continuous BLE sensor data ingestion with multi-region redundancy, containerised scaling and built-in GDPR and HIPAA controls, designed for the diagnostic-as-a-service operating model from clinical pilots through commercial deployment.

    • Ingestion and stream processing

      Sensor data captured at source

      Build the BLE-to-cloud ingestion path as a managed stream so audio, accelerometer and metadata arrive with sensor identity, firmware version and timestamp attached, and can be replayed for algorithm development without losing clinical context.

    • Multi-region reliability

      Failover for clinical continuity

      Design active-passive failover across geographic regions so that a regional outage does not interrupt diagnostic studies in progress or lose recorded sensor data.

    • Compliance and cost controls

      GDPR, HIPAA and unit economics together

      Apply the Well-Architected Framework (a structured review of operational excellence, security, reliability, performance efficiency and cost optimisation, originally published by AWS and adopted by Azure and Google Cloud) across operational excellence, security, reliability and cost optimisation so the platform meets regulatory expectations and stays within the unit economics of a home diagnostic.

    • Continuous diagnostic availability from clinical pilots through home deployment.
    • Compliance controls designed in from day one rather than bolted on later.
    • A platform whose cost scales with deployment, not with the engineering team.

    We've done this before

    See how we do Enterprise AI
  • Digital Lab

    HL7 FHIR integration into Epic and Cerner workflows

    An interoperability layer that exposes Clebre sleep study results as HL7 FHIR R4 resources and delivers them inside Epic and Cerner through a SMART on FHIR embedded application, so referring physicians read the report in their existing clinical workflow.

    • FHIR R4 resource model

      Sleep study as a standard resource

      Define AHI score, body position time series, sensor metadata and clinician summary as HL7 FHIR R4 resources so they map onto the structures Epic and Cerner already store and display.

    • SMART on FHIR embedded app

      Report inside the EHR

      Build a SMART on FHIR application that launches from the patient's record and renders the Clebre sleep study with the same identity, audit trail and user permissions as native EHR content.

    • Clinical decision support hooks

      Actionable signals at the point of care

      Implement CDS Hooks (a standard for triggering clinical decision support from within an EHR workflow) so that significant findings, such as AHI thresholds consistent with moderate-to-severe sleep apnea, surface as recommendations to the clinician without requiring a separate report view.

    • Reports appear in the workflow the physician already uses.
    • Manual data transfer between systems is removed from the clinical path.
    • Decision support surfaces inside the EHR rather than in a separate application.

    We've done this before

    See how we do Digital Lab
  • Enterprise AI

    Versioned model lifecycle with documented revalidation

    An MLOps platform for the deep-learning models behind respiratory event scoring, with training-data lineage, version control, automated revalidation against clinical benchmarks and a deployment gate tied to EU-MDR change control.

    • Model registry with lineage

      Every model version traceable

      Store every model version with the training dataset identifier, hyperparameter set and performance metrics, so that any deployed algorithm can be reproduced and audited from its lineage.

    • Automated revalidation

      Clinical benchmarks run on every change

      Run the validation test suite against clinical recordings on every proposed model update, and gate deployment on documented revalidation evidence aligned with EU-MDR change control.

    • Field performance monitoring

      Drift detected early

      Compare algorithm predictions against confirmed clinical outcomes on an ongoing basis to detect model drift and trigger retraining before the deployed model falls outside its validated operating envelope.

    • Algorithm improvements advance on the same schedule as regulatory submissions.
    • Model drift is detected before it becomes a clinical question.
    • Every deployed version is reproducible from its lineage.

    We've done this before

    See how we do Enterprise AI
  • Digital CDMO

    Validated manufacturing bridge through an ISO 13485 CDMO partner

    A manufacturing partnership framework that takes Clebre from ten-unit prototype batches to commercial ISO 13485 production, with documented process validation, qualified suppliers and batch release aligned with EU-MDR expectations.

    • Process validation for sensor assembly

      IQ/OQ/PQ for every step

      Define and document Installation, Operational and Performance Qualification protocols for the acoustic membrane assembly, BLE module integration and medical-grade patch application steps, so the manufacturing process is auditable from the first commercial batch.

    • Supplier qualification and incoming inspection

      Component quality at intake

      Establish a qualified supplier network for the acoustic membrane, BLE module and patch components with incoming inspection protocols and lot traceability from raw materials through finished devices.

    • Quality system and batch release

      EU-MDR-ready documentation

      Implement document control, CAPA (Corrective and Preventive Action) management and batch release procedures aligned with EU-MDR for Class IIa medical devices, so the first regulatory submission already references an operational quality system.

    • Manufacturing readiness matches the certification timeline rather than trailing it.
    • Every device carries the documentation a notified body (an EU-authorised organisation that audits medical devices for MDR compliance) will ask to see.
    • Supplier risk is identified and managed before it becomes a release issue.

    We've done this before

    See how we do Digital CDMO
  • Agents

    AI agents for EU-MDR and FDA regulatory document work

    Narrow, reviewable agents that draft Clinical Evaluation Reports, post-market clinical follow-up summaries, RWE analysis reports and FDA submission sections from the underlying study data and device records. A named person approves every output.

    • Drafting Clinical Evaluation and PMCF documents

      First drafts from study data

      Generate the first draft of Clinical Evaluation Reports and post-market clinical follow-up summaries directly from the source clinical study records and the literature review database, so the regulatory author edits and judges rather than assembles.

    • RWE analysis report drafting

      Pre-formatted FDA RWE outputs

      Produce analysis reports aligned with FDA Real-World Evidence guidance — including study design documentation, statistical analysis and confidence intervals — from the device-generated RWE dataset, ready for regulatory review.

    • Template completeness and change impact

      Gaps found before review

      Check a draft regulatory document against the EU-MDR or FDA template and the company's own checklist before it enters the human review queue, and identify every controlled document that a standards or labelling change touches.

    • Regulatory queues move faster because documents arrive complete.
    • Submission work happens on the certification timeline, not behind it.
    • Every output is traceable to the source records it came from and signed off by a named reviewer.

    We've done this before

    See how we do Agents

Digital maturity: today and target

Scored out of 100 across six dimensions. The target is what Clebre's own published ambition implies — not a perfect score.

Source: A4BEE analysis of public sources
Cloud infrastructure 45 → 85
The BLE ingestion path has been validated through clinical trials but does not yet carry the multi-region redundancy, automated scaling and compliance controls needed for thousands of concurrent home sensors.
Data integration 32 → 78
Sleep study results are produced inside the device pipeline but not yet exposed as HL7 FHIR R4 resources for Epic or Cerner, so the report still reaches the physician through a manual transfer.
AI/ML operations 50 → 85
Deep-learning models for respiratory event scoring have reached 0.92 to 0.94 AHI correlation against PSG but operate without a version-controlled lifecycle, training-data lineage or automated revalidation pipeline.
Manufacturing digitalisation 40 → 75
Phase III prototypes were produced in collaboration with a specialised firm but the validated ISO 13485 processes, qualified supplier network and batch-release procedures for commercial volume are not yet in place.
Regulatory data management 55 → 88
Clinical evidence for EU-MDR submission is being assembled and the FDA Real-World Evidence policy creates a clear submission path, but the de-identification pipeline, population-scale aggregation and analysis reporting infrastructure for RWE are still ahead of the company.
Cybersecurity 50 → 85
GDPR and HIPAA awareness is established but the production deployment needs an explicit Zero Trust architecture (a security model in which every access request is verified regardless of network location), encryption standards and vulnerability management suited to medical-grade continuous data streams.

Check this yourself

Our Service Portal has free self-assessments and market comparisons. These are the ones that line up with what we've read above — no sales call required.

Think we've read this right?

Talk to us

Related reading

This is an independent analysis prepared by A4BEE from publicly available information as of January 2026. It reflects A4BEE's own interpretation and opinion, is not affiliated with, endorsed by, or verified with Clebre, and may be incomplete or inaccurate. All company names and trademarks are the property of their respective owners. To request a correction or removal, contact [email protected].