A registry is only as valuable as the data it holds. We define what good data look like, monitor them as they arrive, and catch the problems while there is still time to fix them, so the evidence holds up in front of a notified body, a payer or a clinical audience.
The people who build your evidence have sat on the other side of the table.
Jimmy Andrew HayekHead of Quality and ComplianceISO 13485 certified quality system
Sebastien Meier PiantanidaChief Data OfficerData management and EDC
Charline PetitdemangeProject Delivery Lead, France and United KingdomStudy start-up and close-out
A registry is only as valuable as the data it holds. We define what good data look like, monitor them as they arrive, and catch the problems while there is still time to fix them, so the evidence holds up in front of a notified body, a payer or a clinical audience.
Leading manufacturers rely on Eclevar for registry-based PMCF and real-world evidence. Read all the client success stories.
Monitoring decisions are only as good as the understanding of how the evidence will eventually be read. Our teams combine clinical operations, data management, biostatistics and regulatory expertise, including former notified body reviewers, so the quality thresholds we set are the ones that matter when the dataset is reviewed.
We run our own clinical research associates across Europe rather than subcontracting them, which is what makes site initiation, central review and triggered on-site visits work to a single standard across a multi-year registry.
Is the device performing as expected in routine practice? Are there safety signals that only appear after several years? Do outcomes hold up across different patient populations, different centers and different surgical techniques? The answers are only as reliable as the data behind them. A medical device registry feeds post-market surveillance, post-market clinical follow-up, long-term safety and performance evaluation, health technology assessment submissions and, increasingly, reimbursement discussions.
Registries also carry risks that traditional clinical investigations do not. Many hospitals rather than a controlled set of investigational sites. Large populations, light protocols by design, and years of follow-up. Left unmanaged, those realities produce missing data, inconsistent coding, late entry and site-to-site variability that is impossible to separate from genuine clinical variation once the analysis starts. Good data quality does not remove them. It makes them visible early, keeps them within known limits, and documents them, so the analysis can account for what happened instead of guessing.
Each of them reads the registry for something different, and each of them fails differently when the data are weak.
They expect the clinical evidence in a PMCF evaluation report or a periodic safety update report to be traceable and defensible. A dataset with large gaps in follow-up, or with fields that different sites clearly interpreted in different ways, invites questions that are slow and expensive to answer.
They use registry outputs to understand real-world performance and to shape indications, training and technique. Data that do not reflect what actually happens in the operating room, or that quietly miss the patients who never came back, will mislead them.
They build reimbursement dossiers and commercial claims on real-world evidence. HTA bodies are explicit about the quality of evidence they will accept, and a registry that cannot demonstrate completeness or consistency will not carry the argument.
Data entry is often done by staff with other priorities. Patients move, change hospitals or stop attending. A registry is not a controlled investigational setting, and treating it as one is how monitoring budgets get spent on paperwork rather than on quality. The distinction, and what it changes in practice, is set out in registry versus clinical investigation.
The decisions taken during registry design determine how much rework will be needed later.
Agree what the registry has to deliver, and how good the data have to be to deliver it. That means naming the critical data the registry cannot afford to lose: the primary endpoint, device identification and UDI, the index procedure, serious adverse events, device deficiencies, and consent or the applicable legal basis. It also means naming the critical processes: enrollment, follow-up scheduling, event reporting and adjudication.
For each critical variable, define what acceptable looks like as a number rather than an intention. A minimum completeness level. A maximum acceptable loss to follow-up at each time point. A maximum delay between visit and data entry. A query resolution target. Setting these before the registry opens means later discussions are about whether a threshold was met, not about whether the quality being seen is good enough.
The document that holds it all together: critical data and processes, thresholds, the checks built into the electronic data capture and data management system, the metrics reviewed and how often, who reviews them, how issues escalate and what happens when a threshold is missed. It is a living document, versioned, and one of the first things an inspector will ask to see.
Registries fail on ownership more often than on method. The plan names the sponsor, the steering or scientific committee, the coordinating center or CRO, the monitoring lead, the data manager, the statistician, and at each site the principal investigator and the person who actually enters the data. Where an independent adjudication committee is used, its charter and its independence belong here too.
Consistency across sites comes from the tools, not from goodwill. A data dictionary with an unambiguous definition for every field. An eCRF with edit checks, range checks and cross-field logic that stops obvious errors at the point of entry. Written procedures, so two monitors visiting two sites apply the same standard. Training at initiation, and refresher training when staff change, which over a multi-year registry they will.
Oversight concentrates on the data and processes most critical to participant safety and to the reliability of the evidence. A low-risk observational registry of a well-established device does not need the monitoring intensity of a registry supporting a novel implant with a new mechanism of action. Our monitoring services are built and priced on that principle.
The risk assessment carried out during design is what produces the monitoring plan. Risks are ranked on likelihood, impact and how easily they would be detected. A site that enrolls heavily but has never run a registry is a different risk from a site enrolling two patients a year. A field that requires clinical judgment carries a different risk from a date field. The ranking tells the plan where to look first.
Most quality problems are visible in the data before anyone visits a site: an implausibly low event rate, a variable recorded in different units, entry that stops for six weeks, values distributed too neatly to be real. On-site and remote monitoring then works best when visits are triggered by a central signal, high enrollment, staff turnover, a query backlog or a site's first participants, with a purpose defined in advance.
Verifying every field against source documents is neither realistic nor useful in a large multi-year registry. Targeted SDV concentrates on the variables that carry the evidence: consent or legal basis, eligibility, device identification, the primary endpoint, serious adverse events and device deficiencies. Everything else is covered by edit checks, central review and source data review on a sample. The percentage and the selection method are stated in the plan, so the approach can be defended later.
Data quality is not something a registry achieves once. It is a state that has to be held for the life of the study, which for an implantable device may be ten years or more.
Critical data, critical processes and the numerical thresholds that go with them, written into the data quality plan before the first site opens.
Metrics reviewed against those thresholds on a defined cadence, so problems get looked at before somebody happens to notice them.
Root cause analysis on what the metrics show, because retraining a site that was never the problem fixes nothing.
A correction for the data already affected, a preventive action on the system rather than on a person, each with an owner, a due date and an effectiveness check.
What is learned goes back into the data quality plan, the monitoring plan, the training material and the eCRF. Over several years, that loop drives more quality than any individual monitoring activity.
Reviewed typically monthly at operational level, quarterly with the sponsor, and at every steering committee meeting for the scientific view. One metric moving is information. Several metrics moving at the same site is a pattern, and patterns are what oversight exists for.
Overall and site by site, against the minimum level agreed in the data quality plan.
Visits completed within the protocol window, patients retained at each time point, and loss to follow-up by site and by cohort.
Days from visit to data entry, and days from awareness of an event to its reporting.
Query rates, query types, time to resolution, and edit check failure rates by site and by field.
Enrollment against plan, deviation rates, outstanding queries and monitoring findings, combined so it is clear which sites need support. Trend analysis is designed with our biostatistics team.
Reporting timelines met, and the completeness of event documentation. A site whose entry delay is growing, whose query backlog is rising and whose follow-up compliance is slipping is usually a site that has lost its study coordinator. That is fixable in month two, much less so in month nine.
Reliable evidence depends on the data, and on the ability to show how those data came to be.
Registries that start preparing when an inspection is announced prepare badly. Registries that keep their documentation current as they go can respond in days. In practice that means filing as you go, keeping the registry master file complete, closing out monitoring findings instead of letting them accumulate, and periodically running an internal check against the documentation an inspector would request. Teams that want the underlying standard in depth can follow our ISO 14155 training.
Shaped around what the sponsor already has in place, not around a fixed package.
Defining critical data and critical processes, setting monitoring thresholds and writing the registry data quality plan, before the first site opens.
Building the risk assessment and a proportionate monitoring plan, then running central, remote and on-site monitoring against it with our own CRA teams across Europe.
Standing quality metric reporting, site performance review, query management, and the escalation process that goes with them.
An issue log that records what was found, how, where and with what impact, then root cause analysis, CAPA definition and effectiveness checks.
Documentation review, registry master file support and readiness checks ahead of notified body or authority scrutiny.
Assessment and recovery of registries where data quality has drifted, including gap analysis and a prioritized remediation plan.
This page is the data quality and monitoring layer of our medical device registry practice under the EU MDR. It applies the same way to a cardiovascular PMCF registry and to an arthroplasty PMCF registry in the DACH region, and it sits inside the wider medical device CRO offer.
Whitepapers, client voices and publications produced by our teams and our partners (BSI, TÜV SÜD, RegenLab).
It is the document that describes the registry's critical data and critical processes, the numerical thresholds attached to each of them, the checks built into the EDC system, the metrics that will be reviewed and how often, who reviews them, how issues escalate and what happens when a threshold is missed. It is versioned and kept current. Auditors and inspectors ask for it early, and so do notified bodies when the registry supports a PMCF evaluation report.
No, and in a large multi-year registry it is neither realistic nor useful. Targeted source data verification concentrates on the variables that carry the evidence: consent or the applicable legal basis, eligibility, device identification, the primary endpoint, serious adverse events and device deficiencies. Everything else is covered by edit checks, central review and source data review on a sample. What matters is that the percentage and the selection method are written into the monitoring plan in advance, so the approach can be defended later.
A common cadence is monthly at operational level, quarterly with the sponsor, and at every steering committee meeting for the scientific view. The exact frequency belongs in the data quality plan. The point of having a cadence at all is that problems get looked at on a schedule rather than when somebody happens to notice them.
It should. A monitoring plan written at registry start is a hypothesis about where the risks are, and twelve months of data will show which parts of it were right. The plan is reviewed on a defined cadence and after significant events such as a protocol amendment, a new country joining, a change of EDC system or a safety signal, and intensity moves toward the sites and variables where the evidence says it belongs.
Usually, yes, and the earlier the assessment the less it costs. We run a gap analysis against what the evidence has to support, quantify what is recoverable and what is not, and produce a prioritized remediation plan covering data cleaning, targeted re-monitoring, documentation reconstruction and the process changes that stop the drift recurring. Talk to our registry team about where your registry stands.
It is the document that describes the registry's critical data and critical processes, the numerical thresholds attached to each of them, the checks built into the EDC system, the metrics that will be reviewed and how often, who reviews them, how issues escalate and what happens when a threshold is missed. It is versioned and kept current. Auditors and inspectors ask for it early, and so do notified bodies when the registry supports a PMCF evaluation report.
No, and in a large multi-year registry it is neither realistic nor useful. Targeted source data verification concentrates on the variables that carry the evidence: consent or the applicable legal basis, eligibility, device identification, the primary endpoint, serious adverse events and device deficiencies. Everything else is covered by edit checks, central review and source data review on a sample. What matters is that the percentage and the selection method are written into the monitoring plan in advance, so the approach can be defended later.
A common cadence is monthly at operational level, quarterly with the sponsor, and at every steering committee meeting for the scientific view. The exact frequency belongs in the data quality plan. The point of having a cadence at all is that problems get looked at on a schedule rather than when somebody happens to notice them.
It should. A monitoring plan written at registry start is a hypothesis about where the risks are, and twelve months of data will show which parts of it were right. The plan is reviewed on a defined cadence and after significant events such as a protocol amendment, a new country joining, a change of EDC system or a safety signal, and intensity moves toward the sites and variables where the evidence says it belongs.
Usually, yes, and the earlier the assessment the less it costs. We run a gap analysis against what the evidence has to support, quantify what is recoverable and what is not, and produce a prioritized remediation plan covering data cleaning, targeted re-monitoring, documentation reconstruction and the process changes that stop the drift recurring. Talk to our registry team about where your registry stands.
Whitepapers and publications produced by our teams with our notified body partners.
Written with Notified Body BSI: a practical reading of the clinical evidence expectations under EU MDR 2017/745, the same evidence your file has to support.
The services and topics connected to this page.
Send us the device, the claim and the deadline. You get a written answer within 24 hours.
Your documents are reviewed confidentially. An NDA can be signed before we receive any technical or clinical information.
