Medical Device Registries · Long-Term Evidence · International Delivery
Design and operate prospective, retrospective and hybrid medical device registries across protocol, sites, longitudinal data, biometrics and reporting, with the final evidence decision defined before the first record is collected.
The Starting Point
Registry design begins with a decision, not with a data model. Before eligibility, endpoints, sites or follow-up horizon can be fixed, the question the dataset must answer has to be written down and agreed.
What needs to remain observable beyond the original clinical investigation, once the device is used outside a controlled protocol?
Are revision, reintervention, recurrence or other long-term device outcomes the main issue the evidence has to address?
How does the device perform across broader populations, centers and clinical pathways than those represented in the premarket study?
Which decision will the evidence support: clinical evaluation, PMCF, reimbursement and market access, scientific publication, or another defined post-market obligation?
A registry should be designed around the decision it must support, not around the fact that a registry sounds like the obvious post-market solution.
Method Selection
A registry earns its cost when the evidence has to come from routine clinical practice, over a horizon that a controlled study cannot reasonably cover. Whether the required variables already exist in records or have to be predefined and collected going forward decides the model, not whether a registry is the right instrument.
Where these conditions apply, a clinical investigation, a focused observational study or a structured literature and data review is usually the more defensible route.
Do not choose a registry simply because it appears faster or less expensive than a clinical investigation. It should be selected because it is the model capable of answering the intended question at acceptable quality.
Registry Models
The term registry covers several distinct models. The right one depends on whether the required data already exists, how consistently it was recorded, and how much control the sponsor needs over the variables.
MODEL 01
Best when
The required endpoints and data do not yet exist consistently in routine records.
Main advantage
Control over eligibility, follow-up schedule and data definitions from the first patient.
Principal risk
Recruitment and loss to follow-up over a long horizon.
Decision before launch
Can sites collect this data consistently over the required number of years?
MODEL 02
Best when
Relevant historical records already exist for the device and the population.
Main advantage
Faster access to existing cohorts and to follow-up that has already accrued.
Principal risk
Missing data, device traceability and selection bias.
Decision before launch
Are the records genuinely fit for the purpose, tested on real charts rather than assumed?
MODEL 03
Best when
Historical baseline data exist but future follow-up needs standardization.
Main advantage
Existing exposure history is preserved while forward data is collected to a single definition.
Principal risk
Harmonization between historical and prospective variables.
Decision before launch
Can the two periods be analyzed together without the difference becoming the result?
MODEL 04
Best when
A national, specialty or healthcare dataset already contains relevant longitudinal data.
Main advantage
Access to scale and to follow-up duration that a sponsor cohort cannot reproduce.
Principal risk
Limited sponsor control over variable definitions, completeness and data release.
Decision before launch
Do the available variables identify the device and the outcome at the required precision?
Retrospective Feasibility Gate
Before a retrospective or hybrid registry is sized, a pilot review of real medical records tests whether the data supports the design. It is the least expensive point at which the model can still be changed.
Candidate sites, device period and expected chart volume identified.
A defined sample of charts is abstracted against the draft variable list.
Device traceability, endpoint availability, follow-up, missing data, safety documentation, selection method.
Deciding this before the protocol is finalized is what prevents a registry from being committed on assumptions that the first data extraction then contradicts.
Do not choose the registry model before you know the evidence question. The model is a consequence of the question, the available data and the follow-up horizon.
These four models describe how data is obtained over time. They are not a classification of study design: an observational post-market study can use any of them, and a program may combine more than one within a single evidence architecture.
Signature Method
One registry can serve several obligations, but only if each intended use is specified during design. Adding a use after the dataset is closed usually means the variable required by that use was never collected.
The decision the data must support, written before any variable is chosen.
Model, population, endpoints, follow-up horizon, sites and data sources.
Consistent capture across centers and years, with device identification maintained.
Prespecified methods, handling of missing data, and the limits stated openly.
The output enters the decision it was designed for, in the format that decision needs.
Longitudinal data structured so that it can be appraised in the clinical evaluation and referenced in the PMCF evaluation report, rather than summarized after the fact.
Complication, revision and reintervention data captured with enough traceability to be attributed to a device version.
Outcome, resource-use and comparator variables identified during design, since assessment bodies define their requirements independently.
Clinically meaningful research questions, authorship and reporting standards agreed in advance so publication is not a residual activity.
PMCF, reimbursement and publication are uses of a registry dataset. They are outputs of the design decision described above, and each has a dedicated route: read more about PMCF strategy and studies and about the clinical evaluation report.
Data Fitness
Registry value is decided by structure, not by volume. The tests below are applied during design, because most of them cannot be repaired once several years of data have been collected to the wrong definition.
Every variable exists because a defined question needs it. Endpoints are described so that two centers record the same event the same way, in the same time window.
Follow-up completeness is monitored as a live indicator, not discovered at analysis. The expected pattern of missing data is anticipated and its handling prespecified.
The specific device, version and where applicable the identifier are recorded, so an outcome can be attributed to what was actually implanted or used.
How patients enter the dataset is defined and documented. Selective inclusion is the most common reason a large registry cannot support the conclusion drawn from it.
Complications, reinterventions and device deficiencies are captured under definitions that hold across countries and reporting cultures.
Known sources of bias are named in the protocol with the analytical response. Data ownership, access, retention and publication rights are agreed before collection starts.
A large registry is not valuable because it is large. It is valuable when the dataset is fit for the intended evidence decision.
Technology Layer
Registry technology exists to lower the recording burden on sites and patients without weakening the definitions agreed in the protocol. The layer used on a registry typically includes:
Eclevar implements these on Milo Studio, its own clinical data platform, which allows the same team to change a definition, a validation rule or a follow-up window without a third-party release cycle. The platform is an implementation choice; the evidence design decides what it is configured to do. Related capability: clinical data management and EDC.
Long-Term Follow-Up
Registry follow-up may be strengthened by linking the registry cohort to data that already exists elsewhere in the health system, where this is legally and operationally feasible.
Linkage does not automatically solve loss to follow-up, and it is not available in every country or for every variable. Each of the following is assessed before linkage is written into a registry design: lawful basis and data governance approval; the availability of identifiers permitting a reliable match; the permissions required from data custodians and, where applicable, from patients; whether the linked variables are defined precisely enough to answer the registry question; and the release timelines that determine when the data can actually be analyzed.
Where those conditions cannot be met, the honest design response is to change the follow-up model rather than to assume linkage will close the gap later.
Delivery
Eclevar works only on medical devices and diagnostics. Registry strategy, operations, data management and biostatistics sit in one accountable team, so a definition agreed in design is the definition the database enforces.
Turn the evidence obligation into a defined question and a defensible model.
Write the document that every center, year and analysis will be held to.
Establish where the data can genuinely be collected to the agreed definition.
Open centers and keep them recording consistently over years, not months.
Hold one definition across centers, languages, versions and time.
Analyze what was prespecified and state clearly what the data cannot support.
Deliver the output in the form the receiving decision actually requires.
Engagement does not have to start at the beginning.
Delivery Experience
The program below demonstrates the operating discipline required for longitudinal registry delivery: one evidence model held consistently across several device families and several studies at portfolio level.
Client
Three families of implantable vascular devices, vascular patches, vascular prostheses and implantable ports, generated separate post-market evidence obligations. Handled study by study, each would have produced its own eligibility rules, its own variable definitions and its own reporting, leaving the manufacturer with datasets that could not be compared across the portfolio.
The program was instead built on one harmonized observational evidence model, applied across six registered observational studies covering more than 1,150 subjects, with follow-up extending to five years after implantation. Eclevar structured the program architecture and supports delivery, holding a single set of definitions, one data structure and one reporting approach across the portfolio.
Implantable vascular devices followed over several years, which is the scale and the horizon a registry has to sustain.
This program comprised observational PMCF studies rather than a sponsor-owned registry. It is presented as relevant long-term evidence delivery experience, not as a registry deliverable. Public disclosure on this page is limited to the device families, program scale and follow-up duration shown above. No study titles, registration numbers, site names, investigator names, endpoint results or regulatory outcomes are disclosed.
Registry Capability
Where sponsor-owned registry work cannot be disclosed publicly, relevant delivery experience can be discussed under confidentiality during qualification. The capability below is stated as capability, and is exercised on the same team, platform and processes used in the program above.
Further published work is collected in the client success stories.
Specialty Experience
A registry template does not transfer between therapeutic areas. What changes is the outcome that matters, the follow-up horizon and the way the device is identified in routine records.
Accountability
Registry decisions are taken by named people with a defined remit. The roles below are the ones a sponsor deals with directly when the design, the dataset or the analysis has to change.
Clinical and Regulatory Evidence
Whether the evidence question is the right one, and whether the design can defend it.
Dr Mark Da Costa
Chief Operating Officer and Head of Cardiovascular
Dr. Nikhil Khadabadi
Chief Medical Officer, Orthopedics and Spine
Registry Operations
Whether centers can be opened and kept recording to one definition over years.
Susanne Höfer
Head of Clinical Operations, DACH region
Owns country and site feasibility, activation and the retention model that decides follow-up completeness.
Data and Biometrics
Whether the dataset is structured to be analyzed, and analyzed as prespecified.
Sébastien Meier Piantanida
Chief Data Officer
Owns the variable dictionary, the database build, data quality controls and the statistical methodology.
Medical Writing and Evidence Integration
Whether the output reads as evidence in the document that receives it.
Pierre-Marie Boutanquoi
Head of Medical Writing
Owns registry reporting and its integration into clinical evaluation, PMCF and market-access documentation.
International and U.S. Strategy
Whether a European registry design also serves the sponsor's U.S. ambitions.
Dawn Heimer, PhD
Strategic Clinical Advisor, United States
Provides strategic input on U.S. medical-device clinical operations, regulatory considerations and real-world evidence programs.
Full biographies are on the Eclevar executive team page.
Start Here
A structured working session that converts an evidence obligation into a registry decision. It is the point at which the model can still be changed at low cost, and it is deliberately scheduled before any operational budget is committed.
If the workshop concludes that a registry is not the right instrument, that conclusion is delivered with the reasoning and the alternative. That outcome is cheaper for the sponsor than a registry that runs for three years and cannot answer the question.
Frequently Asked
A medical device registry is an organized system for collecting clinical and device-related information from a defined patient population over time. It typically relies primarily on routine-care data, with predefined registry-specific data collection or follow-up where the evidence question requires it. Its protocol, variables and follow-up horizon should be defined in advance.
The distinction from a controlled clinical investigation is one of control rather than of documentation: a registry typically does not impose the study-specific procedures, comparator or measurement conditions that a controlled design requires. Its purpose is to keep long-term safety, performance and outcome data available after the premarket evidence closes.
A registry is appropriate when the evidence question concerns long-term outcomes, implant survivorship, recurrence or reintervention, or performance across broader real-world populations than those studied before market entry. It also fits post-listing and reimbursement requirements that call for continued data collection. It is not appropriate when controlled procedures, a defined comparator or highly standardized measurement are required, because routine care will not reliably produce them.
A prospective registry defines the variables first and then collects them as patients are treated, which gives control over eligibility, follow-up and definitions but carries recruitment and retention risk. A retrospective registry works from records that already exist, which is faster and gives immediate access to accrued follow-up, but the data was recorded for clinical rather than research purposes, so missing data, device traceability and selection bias have to be assessed before the design is fixed.
They can, provided their suitability is demonstrated rather than assumed. The reliable way to establish this is a pilot review of a defined sample of real charts, testing whether the device and its version can be identified, whether the intended endpoints are documented, how much follow-up actually exists, how much data is missing, how complications were recorded, and how patients would be selected. That review should be completed before sample size, sites, timeline and budget are committed.
Under EU MDR, post-market clinical follow-up has to generate data that feeds the clinical evaluation on a continuing basis. A registry can serve that purpose when its endpoints are aligned with the questions the clinical evaluation leaves open, when its follow-up horizon matches the residual risks identified, and when its outputs are produced in a form the PMCF evaluation report can appraise. That alignment is a design decision, not a reporting exercise. See the PMCF pillar for the wider PMCF route.
A registry suits questions about what happens over years in ordinary practice. A clinical investigation suits questions that need controlled conditions, defined comparison or measurements that routine care does not produce. Where a specific hypothesis must be tested under control, a registry will not substitute for a clinical investigation, and choosing it on cost or speed grounds usually results in a dataset that cannot support the intended conclusion.
Cost is driven by the model rather than by patient numbers alone. The main drivers are the number of centers and countries, the length of follow-up, whether data is collected prospectively or abstracted from records, the volume of variables per visit, whether imaging or central adjudication is required, whether linkage is involved, and how frequently evidence has to be reported. A feasibility assessment produces these drivers explicitly, so that a budget can be built on the design rather than on an assumed unit price.
The determining factors are country selection, the ethics and data protection pathway in each country, contracting with each center, and the time required to build and validate the database against the agreed variable list. Retrospective and hybrid designs additionally require the pilot record review to conclude before the protocol can be finalized. Rather than quote a standard duration, Eclevar builds a country-by-country activation plan during feasibility so that the timeline reflects the actual pathway.
Yes. Common engagements include reviewing or amending the design of a running registry, taking over data management where completeness or data quality has become a problem, extending an existing registry into additional countries, and performing analysis and reporting on a longitudinal dataset that another party collected. The starting point in each case is an assessment of whether the existing dataset can still support the intended evidence use.
Next Step
A medical device registry should do more than accumulate patient records. Define the evidence question, data source and follow-up model before committing the operational budget.
Related Capabilities
PMCF strategy and studies
The wider post-market clinical follow-up route, of which a registry is one instrument.
Medical device clinical investigations
Controlled study designs for questions a registry cannot answer.
Clinical data management
Database build, validation and data quality across long-running datasets.
Clinical evaluation reporting
Where registry outputs are appraised and where the next evidence gap is defined.
Feasibility and site selection
Country and center assessment against the target population.
Client success stories
Published examples of Eclevar evidence programs across device specialties.
