PMCF Clinical Trials: How to Design an Effective eCRF

Post-Market Clinical Follow-up (PMCF) enables medical device manufacturers to collect and evaluate clinical data after a device has been placed on the market, helping document its safety, performance, and clinical benefit over time.

To structure this data collection, manufacturers can use EDC/eCRF software (Electronic Data Capture/electronic Case Report Form), allowing investigators to enter protocol-defined data in a standardized environment throughout the clinical trial.

The way an eCRF is designed has a direct impact on data quality. An overly complex form increases investigator workload and the risk of missing data. In contrast, an eCRF built around the scientific objectives of the clinical trial streamlines data collection and prepares the data for analysis from the outset.

The design of the clinical trial should therefore begin well before the eCRF is configured in the EDC software.

Why Shouldn't a PMCF eCRF Be Modeled on a Pharmaceutical CRF?

Medical device clinical trials have characteristics that differ from pharmaceutical trials.

Three differences should be considered in particular: the economic and operational constraints differ from those of large pharmaceutical clinical trials; certain methodologies, such as double-blinding, may not be feasible when an intervention necessarily requires the healthcare professional to know which device is being used; and clinical outcomes may depend not only on the device itself but also on how it is used by the operator.

Follow-up periods can also differ considerably.

It is therefore rarely appropriate to simply reuse a CRF designed for a pharmaceutical clinical trial. The eCRF should instead be built around the medical device, how it is used, the objectives of the protocol, and the data that will actually be analyzed.

Adapt Your eCRF to Your PMCF Strategy Over Time

Short term: prioritize agility without compromising methodology

Some PMCF clinical trials need to be deployed quickly to address an immediate clinical or regulatory question.

They may, for example, be retrospective and rely on information already available in medical records. In this situation, the eCRF should be designed around the data that are actually available, rather than all the data the clinical team would ideally like to collect.

A more comprehensive eCRF cannot recover information that was never documented in the source records.

These clinical trials may also involve a limited number of investigators and patients. Statistical power may therefore be limited, while a small number of events can have a significant impact on observed rates. In a cohort of 50 patients, for example, two complications already represent 4%, whereas their relative impact would be very different in a much larger population.

The CRF should therefore remain proportionate to the analyses that can realistically be performed.

Speed should not come at the expense of investigator experience either. A clinical trial that is deployed quickly but designed without sufficient input from clinical teams may make data entry unnecessarily burdensome and negatively affect investigator engagement in future studies.

Medium term: maintain data consistency

An individual PMCF study may be part of a broader clinical strategy.



If several successive clinical trials involve the same medical device or product range, it is useful to maintain consistency across databases, including variable naming, definitions, units, visit structures, and data collection formats.

Previously validated variables can also be reused to avoid rebuilding the same components for every new clinical trial.

This does not mean blindly copying an existing CRF. Every variable should remain relevant to the objectives of the new protocol.

Long term: think beyond individual clinical trials

When follow-up extends over 5, 10, or 15 years, the approach changes.

Rather than designing an eCRF for a single clinical trial, the objective may be to establish a long-term clinical registry. This database can support a primary study and, depending on the protocol, contribute to ancillary studies or additional analyses.

Over several years, the clinical environment will evolve. New investigators may join, others may leave, and scientific objectives may change.

The eCRF therefore needs to be stable enough to maintain data consistency while remaining flexible enough to evolve over time.

How to Design an Effective eCRF for a PMCF Clinical Trial

1. Start With the Objectives, Not the Variables

The first question should not be: "What data can we collect?" but rather: "What questions do we want to answer?"

The objectives and hypotheses of the clinical trial should be defined before the CRF is designed. They determine the endpoints, influence sample size and the planned analyses, and ultimately define which variables need to be collected.

The logic should therefore be:

For every variable, ask:

Will this data point actually be used to address an objective or perform a planned analysis?

2. Resist the Temptation to Collect Data "Just in Case"

A common mistake is to make a CRF as comprehensive as possible by adding scores, measurements, and other information that might potentially be useful later.

The more items an eCRF contains, the greater the burden on investigators and the higher the risk of missing data.

A theoretically interesting variable also loses much of its value when it is only available for a minority of patients. In a clinical trial involving 50 patients, for example, a variable completed for only 20 patients substantially reduces the population available for analysis. Dividing those patients into two groups can reduce the usable sample even further.

An effective eCRF does not collect as much data as possible. It collects the data that matter, with the highest possible level of completeness.

3. Align the eCRF With the Clinical Trial Protocol

The eCRF should not make it easy to enter responses that are incompatible with the protocol.

For example, if only three diagnoses meet the inclusion criteria, adding an “Other” option to the diagnosis field may not make sense. Any other diagnosis would mean that the patient does not meet the inclusion criteria.

The CRF should reflect the logic of the clinical trial protocol without introducing unnecessary pathways or opportunities for inconsistent data entry.

4. Use Structured Variables Instead of Free Text

Free-text fields are easy to create but much harder to analyze.

Consider "Surgical history." One investigator may provide a detailed description, while another may simply enter “multiple surgeries.” The resulting information is neither standardized nor directly comparable.

When the information of interest is known in advance, it is generally preferable to break it down into structured variables, such as whether previous surgery occurred, the type of procedure, the number of procedures, or the presence of an infection.

Free text remains useful for exceptional comments or information that cannot reasonably be anticipated, but it should not replace structured variables when the expected information can be defined in advance.

5. Define Exactly What You Are Measuring

A subjective or ambiguous variable reduces data comparability across investigators and clinical sites.

For example, simply asking investigators to assess "bone quality" without providing a standardized method can lead each investigator to apply their own interpretation.

The same principle applies to pain scores, definitions such as "chronic" use, or radiological measurements: the eCRF should specify what needs to be measured, under which conditions, and according to which definition.

The more interpretation a variable requires, the more precise the instructions should be.

6. Define Units and Appropriate Validation Ranges

Units of measurement should be clearly defined, particularly for clinical, biological, or radiological data.

Where appropriate, EDC/eCRF software can also apply validation ranges to identify values that are improbable or inconsistent with expected clinical values at the point of data entry.

These controls help detect potential errors earlier in the clinical trial and reduce unnecessary data cleaning later.

7. Use Data Consistency Checks to Identify Errors

Not every inconsistency can be identified by validating a single variable.

EDC software can apply consistency checks across multiple data points. A visit date that precedes the inclusion date, a response that conflicts with an eligibility criterion, or two contradictory pieces of information can be flagged during data entry or data review.

These checks can identify errors earlier and reduce the amount of verification and data cleaning required downstream.

However, they should remain relevant to the clinical trial protocol. Too many unnecessary alerts can make data entry more difficult and increase investigator burden.

8. Automate Derived Variables

When information can be calculated directly from data that have already been collected, investigators should not have to enter it again.

Certain indices, scores, or sub-scores can be automatically calculated from source variables within the EDC/eCRF software.

This reduces duplicate data entry and limits the risk of discrepancies between source variables and derived values.

9. Distinguish Between "Not Applicable," "Not Available," and Missing Data

Missing information does not always have the same meaning.

A data point may not apply to the patient, may not be available, or may simply have been omitted.

For certain imaging examinations, for example, it may be useful to first determine whether the CT scan or MRI is available. If it is not, there is little value in displaying all the variables that depend on that examination.

Similarly, a classification may include a "Not applicable" option rather than artificially generating missing data.

10. Use Visuals When Interpretation Depends on Imaging or Classification

Some clinical or radiological classifications involve several stages or categories.

Embedding an illustration directly in the eCRF can help investigators identify the appropriate classification and reduce differences in interpretation across clinical sites.

For imaging measurements, visual guidance can also demonstrate precisely how a particular measurement should be performed.

11. Simplify Complex Data Collection Structures

An electronic form should not simply reproduce a complex paper table.

Consider a table describing the characteristics of a medical device implant, with numerous cells and dependencies between responses. The same information can sometimes be broken down into three or four simple questions.

This can reduce data entry time, errors, and potentially the monitoring workload. If a data collection grid requires extensive instructions, consider whether it could be replaced by several simple, independent variables.

12. Structure Complication Data

A field such as: "Did the patient experience a complication? Yes/No. If yes, please describe."

may seem intuitive, but free-text descriptions make it considerably more difficult to calculate complication rates and compare events across patients or clinical sites.

When the main expected events are known, it is preferable to use a structured variable listing the relevant complications, while retaining an "Other" category when genuinely necessary.

The resulting clinical data are immediately easier to quantify, compare, and analyze.

13. Adapt Score Collection to Investigators' Clinical Practice

A clinical or functional score can be collected in different ways.

The eCRF can include every individual question required to automatically calculate the final score. However, if the investigator already records that score in their clinical software and the planned analysis only requires the total score or selected sub-scores, asking them to enter every individual item again may create unnecessary work.

Conversely, when the score is not already available, collecting the individual items directly in the EDC/eCRF software may be appropriate.

The decision should therefore be based on the data that are actually available in clinical practice and the data required for the analysis. Discussing existing data collection workflows with investigators is essential.

It is also important to avoid collecting multiple scores that measure essentially the same dimensions, as this can unnecessarily increase the burden on investigators or patients.

14. Minimize Unnecessary Identifiable Data

Designing an eCRF also requires careful consideration of which personal data are genuinely necessary for the clinical trial.

If a complete date of birth is not required, for example, collecting only the month and year may be sufficient where permitted by the protocol and applicable regulatory framework.

The principle is straightforward: do not collect identifiable information simply because it is available if it is not necessary for the clinical trial.

15. Plan for Analysis, Monitoring, and Data Cleaning From the Start

Every variable should be designed with the next stages of the clinical trial in mind.

Can it be filtered? Can it be compared between groups? Can it be used in a statistical analysis? Can its value be validated? Can missing data be distinguished from "Not applicable"?

This explains many of the previous recommendations: limit free-text fields, standardize categories, define units, implement consistency checks, automate derived variables, and avoid poorly defined subjective measures.

Structured and validated data at the point of entry can also reduce the amount of verification and cleaning required later.



Finally, the value of a PMCF database can extend beyond an immediate regulatory requirement. Structured clinical data collected in collaboration with investigators may support new scientific questions and generate value beyond the initial clinical trial.

Which eCRF Software Should You Use for a PMCF Clinical Trial?

eCRF design depends not only on the clinical trial protocol, but also on the capabilities of the software being used. Structured variables, data consistency checks, missing data management, and the ability to reuse variables across multiple clinical trials can all help streamline the implementation of a PMCF strategy.



To learn more, discover how to choose the right EDC/eCRF software for a PMCF clinical trial.

An Effective PMCF eCRF Collects Less Data, but Better Data

Optimizing an eCRF does not mean creating the most comprehensive form possible.



The objective is to collect the data that are genuinely required to address the clinical trial objectives, in a format that is easy for investigators to understand and directly usable for analysis.

For PMCF clinical trials, the eCRF design should also reflect the long-term clinical strategy: a rapid retrospective study and a clinical registry followed for 15 years should not be designed in the same way.

An eCRF is therefore more than a technical component of a clinical trial. Its design is an integral part of the strategy for collecting, controlling, and ultimately using clinical data.

Want to Go Further?

For more practical examples and guidance on designing an eCRF for PMCF clinical trials, watch our webinar.