A prospective PMCF study is a structured clinical investigation conducted on a CE-marked device after it is placed on the market, designed in advance with a written protocol, pre-specified endpoints, a defined patient population, a planned sample size, and a fixed follow-up schedule, so that the data it produces can be analysed against the questions the study was built to answer. To design one that satisfies MDR Annex XIV Part B, fix the research question to a specific open clinical question from the clinical evaluation report, define endpoints that map to Annex XIV Part B objectives, choose the study type proportionate to the risk class and residual uncertainty, calculate a sample size that can actually detect the effect of interest, and conduct the study under EN ISO 14155:2020+A11:2024 for every aspect that applies. Where the activity qualifies as a clinical investigation under MDR Articles 62 to 82, the full regulatory submission regime applies.

By Tibor Zechmeister and Felix Lenhard. Last updated 10 April 2026.

TL;DR

  • A prospective PMCF study is a pre-planned clinical investigation on a CE-marked device, run against a written protocol, with pre-specified endpoints and a fixed follow-up schedule. It is the strongest PMCF evidence format and also the most expensive.
  • Annex XIV Part B of Regulation (EU) 2017/745 requires PMCF studies, where used, to be described in the PMCF plan with a rationale for the method, a link to the clinical evaluation report and the risk file, and a justified time schedule.
  • Where a PMCF study meets the definition of a clinical investigation, MDR Articles 62 to 82 apply, including protocol submission, competent authority and ethics committee review, informed consent, monitoring, and adverse event reporting, alongside the general requirements of Annex XV.
  • EN ISO 14155:2020+A11:2024 is the harmonised standard for good clinical practice in device clinical investigations and governs protocol content, investigator responsibilities, monitoring, data quality, and reporting.
  • The study design sequence is research question, endpoints, design type, sample size, recruitment, ethics, GCP conduct, reporting. Skipping steps produces a study that runs but does not answer the question it was supposed to answer.
  • The output of the study feeds the PMCF evaluation report and, through Article 61(11), the next clinical evaluation update.

Where this post sits in the PMCF cluster

Post 181 is the pillar on post-market clinical follow-up under MDR. Post 182 walks through how to write the PMCF plan under Annex XIV Part B. This post narrows further: it covers the design of one specific PMCF method — the prospective post-market clinical study — from the first decision about the research question through to the reporting of results. If your PMCF plan names a post-market study as one of its methods, this is the post that describes how to actually build it. If you are still choosing between methods, read PMCF methods for startups first.

Step 1 — Fix the research question

A prospective PMCF study that does not have a single, pre-specified research question is a study that cannot be analysed. The research question is the sentence that determines every downstream design decision, and it has to come from a specific gap in the clinical evaluation report.

The exercise is concrete. Open the CER. Find the sections that state conclusions and the sections that state residual clinical uncertainty. For each residual uncertainty, ask whether the uncertainty is something a prospective study could realistically resolve. If yes, and if it maps to one of the Annex XIV Part B objectives — confirming safety and performance over the device lifetime, identifying previously unknown side-effects, monitoring identified risks, keeping the benefit-risk ratio current, or identifying systematic misuse — then the uncertainty is a candidate research question.

A research question phrased well is narrow, falsifiable, and linked to an action. "Does the device continue to deliver the pre-specified clinical performance at 12 months in routine use?" is a research question. "Collect real-world data" is not. The rationale required by Annex XIV Part B for the choice of the PMCF study method is, at its core, the argument that this research question cannot be answered by lighter methods — literature, similar-device monitoring, user surveys, registry participation — and therefore a prospective study is necessary.

Step 2 — Define the endpoints

Endpoints are the measurable quantities the study will collect to answer the research question. A PMCF study needs at least one primary endpoint tied directly to the research question, and usually a small set of secondary endpoints that address supporting questions.

Primary endpoints are the quantities the sample size calculation is built on and the conclusions are drawn from. In a PMCF study, the primary endpoint is usually one of: a clinical performance measure (success rate, time to effect, measured physiological outcome), a safety measure (rate of a specific adverse event, rate of device-related serious incidents), or a benefit-risk measure that combines performance and safety. The primary endpoint must be specific enough to be measured consistently across sites and patients, and it must map to an Annex XIV Part B objective.

Secondary endpoints address questions the primary endpoint does not cover but the CER still needs updated information on. Common secondaries include durability measures, user-reported outcomes, specific subgroup performance, and descriptive safety endpoints that are not powered but are worth collecting. Secondary endpoints do not drive sample size; they ride along on the sample the primary endpoint requires.

Every endpoint, primary or secondary, must be definable in the protocol with a clear operational definition — what is measured, when it is measured, who measures it, which instrument or scale is used, and how missing data is handled. Endpoints without operational definitions produce data that cannot be analysed and arguments at audit about what the numbers actually mean.

Step 3 — Choose the design type

Prospective PMCF studies come in several design types, and the choice shapes the rest of the study.

Single-arm observational cohort. The most common PMCF design. All enrolled patients receive the CE-marked device in routine clinical practice, and outcomes are measured prospectively against pre-specified endpoints. No randomisation, no comparator arm. Appropriate when the research question is about the performance of the device in real-world use and a comparator arm is neither clinically justified nor feasible.

Comparative cohort. Two or more groups are followed prospectively, one receiving the device, one or more receiving a comparator. The comparator can be a different device, a standard-of-care treatment, or a historical control if justified. Appropriate when the research question is about relative performance or when the CER claims comparative benefit that needs real-world confirmation.

Registry-embedded study. A study built inside an existing disease registry or device registry, using the registry infrastructure for recruitment and follow-up but adding protocol-specified data collection for the study population. Often the most efficient design when a suitable registry exists.

Real-world data study with prospective protocol. A study that draws on routine clinical data — electronic health records, device telemetry, claims data — but does so against a protocol written in advance that pre-specifies the population, the endpoints, the analysis, and the time windows. Appropriate for connected software devices and for research questions where routine data is rich enough to answer the question.

The design type is documented in the PMCF plan and defended in the protocol. The choice depends on the research question, the feasibility of each design for the device and the target population, the resources available, and whether the activity qualifies as a clinical investigation under MDR Articles 62 to 82.

Step 4 — Calculate the sample size

A sample size that is too small produces a study that runs without being able to detect anything. A sample size that is too large wastes money and time and, for higher-risk devices, exposes more patients than necessary. The calculation is a design responsibility, not a reporting afterthought.

Sample size calculation needs four inputs: the primary endpoint, an assumed effect size or rate the study is designed to detect, a statistical power (conventionally 80 percent or higher for PMCF studies), and a significance level. The calculation produces the number of evaluable patients required and, by working backward through expected dropout and screen-failure rates, the number of patients to enrol.

For single-arm observational PMCF studies, the calculation is often built around a target precision for the primary endpoint estimate rather than a hypothesis test — for example, the sample needed so that the 95 percent confidence interval around the observed rate is narrower than a pre-specified width. For comparative designs, the calculation is the standard hypothesis-test sample size for the comparison of interest.

The sample size calculation, the assumptions behind it, and the reasoning for each assumption all belong in the protocol. Assumptions that cannot be defended from the CER, the risk file, or published literature weaken the protocol and the study.

Step 5 — Plan recruitment and follow-up

A study that cannot enrol the patients it needs, in the time it has, at the sites it has access to, is a study that will fail — slowly and expensively. Recruitment planning is where optimism and reality meet.

The recruitment plan names the investigator sites, the expected enrolment rate per site per month, the screening strategy, the inclusion and exclusion criteria, and the contingency for under-enrolment. Inclusion and exclusion criteria have to be tight enough to produce an interpretable cohort and loose enough that real patients actually qualify. Criteria that are too tight produce a study that never recruits; criteria that are too loose produce a cohort that cannot be interpreted.

Follow-up is the schedule of visits, measurements, and contacts. The follow-up duration has to be long enough to capture the primary endpoint and any clinically meaningful secondary endpoints — for implantable devices, this often means years, not months. The follow-up schedule has to be realistic for both sites and patients; over-ambitious schedules produce missing data, and missing data is the most common reason a PMCF study fails to answer its research question.

Step 6 — Secure ethics and regulatory approvals

Any PMCF study on human subjects requires ethics committee approval. Where the PMCF activity qualifies as a clinical investigation under MDR Articles 62 to 82, the full competent authority submission regime applies alongside ethics committee review, and the study must comply with the general requirements for clinical investigations set out in Annex XV.

Article 74 specifically addresses clinical investigations of CE-marked devices used within their intended purpose — the Article 74(1) regime for PMCF studies that do not involve additional invasive or burdensome procedures differs from the Article 62 regime for pre-market investigations, and the distinction matters operationally. The protocol, the investigator's brochure, the informed consent documents, the insurance coverage, the serious adverse event reporting procedures, and the monitoring plan all have to be in place before the first patient is enrolled.

Where Article 74(1) applies, competent authority notification may be required in some Member States and not others, and national implementing rules vary. The rule of thumb is to assume the full clinical investigation regime applies and to work with regulatory counsel in each Member State where the study will run to confirm the exact obligations. Under-submission is a regulatory risk; over-submission is a scheduling cost.

Informed consent is non-negotiable. Even in PMCF studies where the device is used within its intended purpose and the procedures are those of routine care, patients must consent to participation in the study and to the specific data collection the protocol requires. EN ISO 14155:2020+A11:2024 specifies the content and process.

Step 7 — Conduct the study under EN ISO 14155:2020+A11:2024

EN ISO 14155:2020+A11:2024 is the harmonised standard for good clinical practice in medical device clinical investigations. Where a PMCF study qualifies as a clinical investigation, the standard applies in full for the aspects that apply. Where the activity is lighter — a real-world data study that does not meet the clinical investigation definition — the standard still provides the benchmark for good design and conduct, and auditors will read the protocol against it regardless.

The standard governs protocol content, investigator responsibilities, site qualification, monitoring plans, source data verification, data management, deviation handling, serious adverse event reporting, investigator training, and the clinical investigation report. For PMCF studies, the practical implications are that the protocol needs every element EN ISO 14155:2020+A11:2024 requires, the sites need qualified investigators with documented training, the monitoring plan has to be risk-based and written down, and the data management system has to produce a clean, auditable dataset at the end.

Monitoring is often where startup PMCF studies go wrong. A monitoring plan that is too thin produces data quality problems at lock; a monitoring plan that is too thick burns through budget before enrolment is complete. The standard supports a risk-based approach — monitoring intensity scaled to the risks in the specific protocol — and the plan documents the reasoning.

Step 8 — Report the results

The output of the study is the clinical investigation report written under EN ISO 14155:2020+A11:2024, and the findings of that report feed directly into the PMCF evaluation report required by Annex XIV Part B. The PMCF evaluation report in turn feeds the clinical evaluation update cycle under Article 61(11).

The clinical investigation report documents the study as it was conducted, the data as it was collected, the analysis as it was performed, any deviations from the protocol and their impact, the conclusions, and the limitations. For publication purposes, the study may also produce peer-reviewed manuscripts, but the CIR is the regulatory artefact and the one the notified body will read.

The findings then travel. The PMCF evaluation report summarises the CIR findings against the Annex XIV Part B objectives and the pre-specified acceptance criteria from the PMCF plan. The CER is updated with the new clinical data. The risk file is updated if the findings change risk estimates. The labelling is updated if the findings change what clinicians or patients need to know. The PSUR under Article 86 cites the findings. For Class III and implantable devices, the Summary of Safety and Clinical Performance is updated.

A study that is well-designed but poorly reported does not close the loop. Reporting discipline is part of the design.

The Subtract to Ship angle — the lightest prospective study that answers the question

The Subtract to Ship framework for MDR applied to PMCF study design produces a clear rule. Start from the research question. Build the lightest design that can credibly answer it — fewest endpoints, smallest defensible sample size, shortest follow-up that captures the primary endpoint, leanest monitoring plan that EN ISO 14155:2020+A11:2024 supports. Add complexity only where the research question cannot be answered without it.

The opposite pattern — building a study around the methodology textbook and then looking for a research question it can answer — produces bloated protocols that burn budget without closing the clinical loop. A lean prospective PMCF study that answers one well-chosen research question is worth more, regulatorily and clinically, than an elaborate study that collects everything and concludes nothing.

Reality Check — where do you stand?

  1. Can you state the research question of your prospective PMCF study in one sentence that ties to a specific open clinical question in the CER?
  2. Does every endpoint in the protocol map to an Annex XIV Part B objective, and does the primary endpoint have a defensible operational definition?
  3. Is the study design type — single-arm cohort, comparative, registry-embedded, real-world data — justified in writing against the research question?
  4. Does the sample size calculation use assumptions that can be defended from the CER, the risk file, or published literature?
  5. Is the recruitment plan realistic against actual site enrolment rates, or is it a wish list?
  6. Do you know whether the study qualifies as a clinical investigation under MDR Articles 62 to 82, and have you mapped the obligations in each Member State where the study will run?
  7. Does the monitoring plan match a risk-based reading of EN ISO 14155:2020+A11:2024, or is it copied from a pre-market investigation protocol?
  8. Is the path from the clinical investigation report to the PMCF evaluation report to the CER update written down and owned, or will it be improvised when the data arrives?

Frequently Asked Questions

Does every PMCF study qualify as a clinical investigation under MDR Articles 62 to 82?

No. A PMCF study qualifies as a clinical investigation when it meets the definition of a clinical investigation under the MDR — a systematic investigation involving one or more human subjects to assess the safety or performance of a device. Studies on CE-marked devices used within their intended purpose that do not involve additional invasive or burdensome procedures fall under Article 74(1), which is a lighter regime than the pre-market Article 62 route. Studies that involve additional procedures, or that fall outside the intended purpose, fall under the full clinical investigation regime. The boundary is a legal judgement that belongs with regulatory counsel.

Does EN ISO 14155:2020+A11:2024 apply to every PMCF study?

EN ISO 14155:2020+A11:2024 applies in full where the PMCF activity is a clinical investigation. Where the activity is lighter — real-world data collection against a prospective protocol without the procedural features of a clinical investigation — the standard may not apply in full, but the design and conduct principles it codifies are still the benchmark auditors will measure against, and the plan should name which parts of the standard it follows and which it does not.

Can a single-arm prospective study satisfy PMCF obligations for a Class III device?

It can, if the research question is such that a single-arm design is the appropriate way to answer it and the rationale is defensible in the PMCF plan. For Class III devices the bar is higher because Article 61(11) requires annual CER updates with PMCF data and Article 86 names PMCF findings in the PSUR, and notified bodies scrutinise the design accordingly. Many Class III PMCF programmes combine a prospective study with registry participation and literature surveillance to cover the full range of Annex XIV Part B objectives.

How long does a prospective PMCF study usually take?

From protocol draft to final clinical investigation report, prospective PMCF studies typically take between 18 months and several years, depending on the follow-up duration, the enrolment rate, and the regulatory submission timelines. Implantable device studies with multi-year follow-up can run for longer. The time schedule required by Annex XIV Part B for the PMCF plan has to account for this realistically.

What is the link between the clinical investigation report and the PMCF evaluation report?

The clinical investigation report documents the study. The PMCF evaluation report, required by Annex XIV Part B, summarises PMCF findings across all methods — including the study — against the pre-specified objectives and acceptance criteria from the PMCF plan, and feeds the clinical evaluation update cycle under Article 61(11). The CIR is an input to the PMCF evaluation report, not a substitute for it.

Can real-world data replace a prospective PMCF study?

For some research questions, yes. A connected software device generating rich clinical data through routine use can answer PMCF research questions that would have required a prospective study for a non-connected device. The test is whether the real-world data can genuinely answer the research question with the rigour the clinical claim requires. Where it can, the lighter design is preferable. Where it cannot, the prospective study is the appropriate method.

Sources

  1. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, Article 61(11) (PMCF update of clinical evaluation), Articles 62 to 82 (clinical investigations regime), Annex XIV Part B (post-market clinical follow-up), and Annex XV (clinical investigations — general requirements). Official Journal L 117, 5.5.2017.
  2. MDCG 2025-10 — Guidance on post-market surveillance of medical devices and in vitro diagnostic medical devices. Medical Device Coordination Group, December 2025.
  3. EN ISO 14155:2020+A11:2024 — Clinical investigation of medical devices for human subjects — Good clinical practice.

This post is part of the Post-Market Surveillance & Vigilance series in the Subtract to Ship: MDR blog. Authored by Felix Lenhard and Tibor Zechmeister. A prospective PMCF study is a commitment to answer one clinical question rigorously, and the discipline of the design is what determines whether the answer is worth the investment.