PMCF Survey Design That Notified Bodies Actually Accept: Evidence Level, Sample Size, Endpoints, and Bias Controls
How to design a PMCF survey under EU MDR that Notified Bodies accept as Level 4 clinical evidence — sample frame construction, inclusion/exclusion criteria, endpoint mapping to CER claims, acceptance criteria, statistical power, bias controls, adverse event capture, and when surveys alone are insufficient.
What This Article Covers
This article focuses on one PMCF method: the healthcare professional (HCP) survey. It explains how to design, execute, and defend a PMCF survey so that Notified Bodies (NBs) accept it as Level 4 clinical evidence under MDCG 2020-6, rather than defaulting it to Level 8 "general feedback." It covers the survey plan, sample size, endpoint mapping, bias controls, adverse event capture, response-rate defense, and the limits of what surveys can demonstrate.
This article covers:
- The critical difference between Level 4 and Level 8 PMCF surveys
- How to map CER residual uncertainties and clinical claims to survey endpoints
- Sample frame construction: inclusion/exclusion criteria, site selection, and response-rate defense
- Acceptance criteria design and statistical power calculations
- Bias identification and mitigation strategies
- Adverse event and complaint capture within the survey instrument
- The survey plan document structure that NBs expect
- Questionnaire design principles for structured clinical data collection
- When surveys are sufficient and when they are not
- Common NB objections and how to preempt them
This article does NOT cover:
- PMCF planning in general (see the PMCF guide)
- PMCF clinical investigations, registries, or literature reviews
- PMCF plan template structure (see the PMCF plan template)
- PSUR preparation or CER writing
- Digital health or SaMD-specific evidence strategies
Why Most PMCF Surveys Get Rejected by Notified Bodies
Under EU MDR, a PMCF survey is not a customer satisfaction questionnaire. It is a clinical evidence instrument. NBs routinely reject surveys that fail to qualify as Level 4 evidence, defaulting them to Level 8 "general feedback" status. The consequences: deficiency letters, costly rework, and delayed certification timelines.
The distinction is defined in Appendix III of MDCG 2020-6, which establishes an 8-level hierarchy of clinical evidence:
| Level | Evidence Type | PMCF Survey Classification |
|---|---|---|
| 1 | Systematic review / meta-analysis | — |
| 2 | Randomized controlled trial | — |
| 3 | Non-randomized comparative study | — |
| 4 | Case-series from clinical data sources (retrospective/prospective) | High-quality PMCF survey with structured, case-level clinical data |
| 5 | Case-series from registry data | — |
| 6 | Non-clinical data (bench, animal) | — |
| 7 | Expert opinion / consensus | — |
| 8 | General feedback, anecdotal, unstructured | Low-quality survey: satisfaction scores, general impressions, unstructured responses |
A survey qualifies as Level 4 only when it collects structured, case-level clinical observations linked to individual patient encounters, with predefined endpoints, acceptance criteria, and a justified sample size. Anything less is Level 8.
The stakes are high. A 2025 peer-reviewed survey of 13 European Notified Bodies found that 327 applications (approximately 11% of estimated certificates) were not approved, with PMCF studies required as certificate conditions for novel technologies (Dobrzynska et al., Frontiers in Medical Technology, 2025). In one case, deficient PMCF led to certificate cancellation.
Step 1: Map CER Residual Uncertainties to Survey Objectives
Identifying What the Survey Must Address
Before designing any question, identify the specific clinical data gaps that justify the survey. These come from three sources:
- CER residual uncertainties — clinical claims in the CER that lack sufficient post-market confirmation
- Risk management residual risks — risks identified in ISO 14971 risk analysis that require ongoing monitoring
- NB conditions from prior certification — specific deficiencies or conditions that requested PMCF data
Table 1: CER-to-Survey Objective Mapping Matrix
| CER Gap ID | CER Residual Uncertainty | Affected Clinical Claim | Survey Objective | Primary Endpoint | Acceptance Criterion |
|---|---|---|---|---|---|
| GAP-001 | Limited data on hemostasis success in anticoagulated patients | "Device achieves hemostasis in external bleeding wounds" | Confirm hemostasis success rate in anticoagulated patients | Proportion of cases achieving hemostasis (visual assessment) | ≥ 75% success rate |
| GAP-002 | No post-market data on device use beyond 8 hours | "Device can remain in place for up to 24 hours" | Confirm duration-of-use safety profile | Proportion of cases with device in place > 8 hours without adverse event | ≤ 5% adverse event rate for extended use |
| GAP-003 | Off-label use in pediatric population observed in complaints | "Intended for adult use only" | Quantify extent and outcomes of pediatric off-label use | Number of pediatric cases identified; adverse event rate | Not applicable (surveillance objective — signal detection) |
Writing Survey Objectives
Each objective must be:
- Specific: tied to a measurable clinical parameter
- Justified: linked to a documented gap in the CER or risk management file
- Achievable: feasible within the survey methodology (retrospective chart review or prospective case capture)
Recommended Reading
PSUR Under EU MDR Article 86: Complete Periodic Safety Update Report Guide with 2026 EUDAMED Requirements
EU MDR / IVDR Post-Market Surveillance2026-04-17 · 19 min read
Step 2: Define Endpoints and Acceptance Criteria
Primary and Secondary Endpoints
Every PMCF survey must have at least one primary endpoint. This is the endpoint used for the sample size calculation.
Table 2: Endpoint Design Framework
| Element | Primary Endpoint | Secondary Endpoints |
|---|---|---|
| Definition | Single most important clinical outcome the survey measures | Additional outcomes that support or qualify the primary result |
| Example | Hemostasis success rate (binary: yes/no, per visual assessment by HCP) | Time to hemostasis; adverse event rate; device in-place duration; ease of use rating |
| Acceptance criterion | Pre-specified threshold (e.g., ≥ 75%) | Descriptive; may have exploratory thresholds |
| Role in sample size | Drives the calculation | Does not drive sample size |
| Analysis | Formal statistical test against acceptance criterion | Descriptive statistics; may use confidence intervals |
Acceptance Criteria
Acceptance criteria must be:
- Pre-specified before data collection begins (not retrofitted to results)
- Clinically meaningful (not just statistically convenient)
- Justified by pre-market clinical data and literature
Example acceptance criteria derivation:
| Endpoint | Pre-Market Data | Literature Benchmark | Acceptance Criterion | Justification |
|---|---|---|---|---|
| Hemostasis success | 85% in pivotal study (n=120) | 78–92% in published literature for comparable devices | ≥ 75% | Set below pre-market rate to account for real-world variability; literature supports this threshold as clinically acceptable |
Step 3: Construct the Sample Frame
Sample Size Calculation
The sample size must be statistically justified. For a single-proportion endpoint (e.g., success rate), use a one-sample exact binomial test or normal approximation.
Required inputs:
- Expected performance (from pre-market data or literature): e.g., 85%
- Acceptance criterion (the minimum acceptable rate): e.g., 75%
- Significance level (alpha): typically 0.025 (one-sided) or 0.05 (two-sided)
- Power (1 – beta): typically 80% or 90%
- Drop-out or non-evaluable rate: typically 10–20%
Illustrative calculation:
- Expected proportion: 0.85
- Null hypothesis threshold: 0.75
- Alpha: 0.025 (one-sided)
- Power: 80%
- Using exact binomial test: minimum evaluable sample = approximately 85 cases
- Adjusting for 15% non-evaluable rate: 85 / 0.85 ≈ 100 survey responses needed
This calculation must be documented in the survey plan and performed by a qualified statistician or using validated statistical software.
Inclusion and Exclusion Criteria
Table 3: Inclusion/Exclusion Criteria Template
| Criterion Type | Parameter | Rationale |
|---|---|---|
| Inclusion | Device used per IFU (on-label use) | Ensures data reflects intended use population |
| Inclusion | HCP has used device at least once prior | Ensures familiarity with device; reduces learning-curve bias |
| Inclusion | Complete patient chart available for case | Enables case-level data extraction; no missing outcomes |
| Inclusion | Case occurred within defined reporting period | Ensures contemporary practice patterns |
| Exclusion | Device used off-label (unless GAP-03 specifically targets off-label use) | Prevents confounding of primary endpoint |
| Exclusion | Insufficient documentation to determine outcome | Maintains data quality |
| Exclusion | HCP involved in device development or clinical investigation | Reduces investigator bias |
Site and Respondent Selection
- Select sites that represent the intended use environment (hospital types, geographies, experience levels)
- Avoid convenience sampling (only sites from one country or one hospital network)
- Target a minimum of 3–5 sites to capture practice variation
- Document the selection rationale and site characteristics
Step 4: Design the Questionnaire
Principles for Level 4-Qualifying Questionnaires
- One questionnaire per case (per patient encounter), not per HCP impression of overall experience
- Structured, closed-ended responses linked to pre-defined categories, not free-text opinions
- Clinical outcome questions first, satisfaction questions last (if included at all)
- Adverse event capture built in — every case must include an AE/SAE determination
- Source data reference — each question should map to a field extractable from the patient chart
Table 4: Sample Questionnaire Structure (Illustrative)
| Section | Questions | Response Type | Maps to |
|---|---|---|---|
| A. Case identification | Case ID, date of procedure, site ID, HCP specialty | Structured fields | Data management |
| B. Patient demographics | Age, sex, relevant comorbidities (anticoagulation status) | Structured fields | Subgroup analysis |
| C. Device use | Device model/size, anatomical site, duration of placement | Structured fields | Device description |
| D. Primary endpoint | Was hemostasis achieved? (Yes/No) | Binary | GAP-001 acceptance criterion |
| E. Secondary endpoints | Time to hemostasis (minutes), adverse events (list), device removal method | Structured fields | Exploratory analysis |
| F. Adverse events | Did any adverse event occur? If yes: type, severity, relatedness to device, outcome | Structured AE form | Safety monitoring, PMS |
| G. Device performance rating | Ease of application (5-point Likert), would you use again? (Yes/No) | Structured | Context only (not primary evidence) |
Questionnaire Validation
Before deployment:
- Pilot test with 3–5 HCPs not involved in the main study
- Assess completion time (target: ≤ 10 minutes per case)
- Verify all response options are mutually exclusive and collectively exhaustive
- Confirm data extraction is feasible from typical patient charts
- Document validation results in the survey plan
Recommended Reading
GB PMSR/PSUR Dual-Report Architecture: How to Structure Post-Market Surveillance Reports for Devices Sold in Both EU and Great Britain
Post-Market Surveillance EU MDR / IVDR2026-05-05 · 18 min read
Step 5: Bias Identification and Mitigation
Table 5: Bias Matrix for PMCF Surveys
| Bias Type | Description | Mitigation Strategy |
|---|---|---|
| Selection bias | Only "successful" cases are reported by HCPs | Request consecutive case reporting; provide case log template; audit compliance |
| Recall bias | HCPs rely on memory rather than chart data | Require case-level data extracted from patient records at time of encounter |
| Response bias | HCPs give socially desirable responses | Use objective outcome measures (hemostasis yes/no) over subjective ratings; anonymize data |
| Non-response bias | HCPs who had complications may be less likely to respond | Track response rates by site; follow up with non-responders; report response rate in results |
| Investigator bias | HCPs with financial relationships report favorable outcomes | Exclude HCPs with conflicts of interest; disclose relationships in survey plan |
| Sampling bias | Sites/HCPs are not representative of intended use population | Multi-site design; document site characteristics; compare against market demographics |
| Lead-time bias | Early adopters may have different outcomes than later users | Define reporting window; stratify analysis by experience level if sample allows |
| Survivorship bias | Only patients who returned for follow-up are captured | Define minimum follow-up period in inclusion criteria; document loss to follow-up |
Step 6: Adverse Event Capture
Every PMCF survey must include a mechanism for adverse event detection and reporting. This is not optional — it serves both the PMCF objectives and the manufacturer's vigilance obligations.
Table 6: AE Capture Requirements Within PMCF Surveys
| Requirement | How to Implement in the Survey |
|---|---|
| AE screening question | "Did the patient experience any adverse event during or after device use?" (Yes/No) |
| AE characterization | If yes: type (dropdown from pre-defined list aligned with CER risk categories), severity (mild/moderate/serious), device relatedness (related/unlikely related/unrelated) |
| SAE escalation | Protocol must specify: if SAE identified, HCP must report to manufacturer's vigilance system within defined timeline (per MDR Article 87 requirements) |
| Complaint linkage | AE data must be cross-referenceable with the complaint management system for trending |
| PMCF-CER feedback | AE summary from survey must feed into CER update and PSUR |
Step 7: Survey Plan Document Structure
The survey plan is the master document the NB reviews. It must contain all of the following:
Table 7: PMCF Survey Plan Structure
| Section | Content | NB Review Focus |
|---|---|---|
| 1. Device description | Device name, model, Basic UDI-DI, intended use | Confirms scope |
| 2. Survey objectives | Mapped to CER gaps (Table 1) | Justification for survey vs. other PMCF methods |
| 3. Justification for survey design | Why Level 4 survey is appropriate; why not clinical investigation or registry | Methodology choice |
| 4. Endpoints and acceptance criteria | Primary and secondary endpoints with pre-specified criteria | Statistical rigor |
| 5. Inclusion/exclusion criteria | Patient and HCP criteria (Table 3) | Population definition |
| 6. Sample size calculation | Statistical method, inputs, output, adjustment for non-evaluable cases | Power adequacy |
| 7. Site and respondent selection | Number of sites, selection rationale, HCP qualification criteria | Representativeness |
| 8. Questionnaire | Final validated questionnaire (appendix) | Data quality |
| 9. Data collection method | Electronic vs. paper; data capture system; GDPR compliance | Data integrity |
| 10. Data analysis plan | Statistical methods for primary endpoint; handling of missing data; subgroup analyses | Analytical rigor |
| 11. Bias controls | Bias matrix and mitigation strategies (Table 5) | Evidence quality |
| 12. AE/Serious AE procedures | Escalation pathway; vigilance reporting; complaint linkage | Safety monitoring |
| 13. Timeline | Recruitment period, data collection period, analysis period, report delivery | Planning |
| 14. Premature termination criteria | Conditions under which the survey is stopped early | Ethical/safety |
| 15. Ethical considerations | IRB/Ethics committee notification or approval status; informed consent if applicable; GDPR compliance | Regulatory compliance |
Recommended Reading
EU MDR Summary of Safety and Clinical Performance (SSCP): Complete Guide for Class III and Implantable Devices
EU MDR / IVDR CE Marking2026-04-17 · 16 min read
When Surveys Are Insufficient
PMCF surveys are one tool in a portfolio. They are insufficient as the sole PMCF method for:
- High-risk implantable devices where long-term performance data is needed (use registries or clinical investigations)
- Devices with significant residual risks that require controlled follow-up (use targeted clinical follow-up)
- Devices with emerging safety signals identified through vigilance (use formal clinical investigation)
- New indications or significant design changes (use clinical investigation)
- Devices where equivalence was claimed but post-market confirmation of the equivalence link is needed (use comparative studies)
Table 8: PMCF Method Selection Decision Guide
| Scenario | Is Survey Alone Sufficient? | Recommended Complement |
|---|---|---|
| Mature device, established safety profile, minor CER gaps | Yes, if Level 4 criteria met | Literature review (routine) |
| Device with off-label use signals | Partially (for quantification) | Add focused complaint trend analysis |
| Device with new adverse event trend in PSUR | No | Formal clinical investigation or registry |
| Device claiming equivalence with limited post-market data | No | Comparative registry or clinical study |
| Device with design change affecting clinical performance | No | Clinical investigation |
| Class III implantable device with limited long-term data | No | Registry participation + literature review |
Common NB Objections and Preemptive Responses
Table 9: NB Objection and Preemptive Response Table
| NB Objection | Root Cause | How to Preempt in the Survey Plan |
|---|---|---|
| "Survey generates only Level 8 evidence" | Unstructured questionnaire; no case-level data; no acceptance criteria | Use Level 4 design: case-level, structured endpoints, pre-specified acceptance criteria, justified sample size |
| "Sample size is not justified" | No power calculation; arbitrary target (e.g., "50 responses") | Include formal sample size calculation by statistician with all inputs documented |
| "Endpoints do not map to CER claims" | Survey asks about satisfaction, not clinical outcomes | Map every question to a CER claim or risk management residual risk (Table 1) |
| "No bias controls documented" | Survey plan omits bias discussion | Include bias matrix (Table 5) with mitigation strategies |
| "Consecutive case reporting not verified" | HCPs cherry-pick successful cases | Build case-log requirement into protocol; audit a sample of sites |
| "AE capture is inadequate" | AE questions are optional or absent | Mandatory AE screening for every case; SAE escalation pathway documented |
| "Survey is the only PMCF activity for a high-risk device" | Over-reliance on surveys for Class III | Include survey as part of a multi-method PMCF strategy; justify why clinical investigation is not needed |
| "Response rate too low to draw conclusions" | No response-rate target; no follow-up strategy | Set target response rate ≥ 70%; document follow-up procedures for non-responders |
| "HCP selection is biased" | Only manufacturer's key opinion leaders invited | Use multi-site design with documented, representative selection criteria |
| "No statistical analysis plan" | Plan says "data will be summarized" without specifying methods | Include detailed statistical analysis plan with methods for each endpoint |
Source-to-Evidence Traceability
Table 10: PMCF Survey Evidence Traceability Chain
| Source Document | Evidence Element | Survey Component | PMCF Evaluation Report Section | CER Update Section |
|---|---|---|---|---|
| CER v8.2, Gap Table | GAP-001: Limited hemostasis data in anticoagulated patients | Primary endpoint: hemostasis success rate | Section 4.1: Primary endpoint results | Updated CER Section 8.3: PMCF data incorporated |
| Risk Management File v6.1 | Residual risk: infection at application site | AE capture: infection type, rate | Section 4.3: Safety results | Updated CER Section 7.2: Post-market safety data |
| IFU Section 2.3 | Contraindication: active systemic infection | Exclusion criterion | Section 3: Population description | Cross-check: no change to IFU indicated |
| PMCF Plan v3.1 | Activity 2.3: HCP survey for real-world performance | Entire survey plan | Section 5: PMCF plan completion status | N/A (PMCF plan tracking) |
Recommended Reading
IVDR Companion Diagnostic Consultation File: How to Prepare the NB–EMA/National Authority Interface Package
Companion Diagnostics EU MDR / IVDR2026-05-05 · 17 min read
Key Regulatory References
- MDR Annex XIV Part B — PMCF requirements
- MDR Article 61 — Clinical evaluation requirements
- MDR Article 86 — Periodic safety update reports
- MDCG 2020-6 — Guidance on sufficient clinical evidence (Appendix III: evidence hierarchy)
- MDCG 2020-7 — PMCF plan template
- MDCG 2020-8 — PMCF evaluation report template
- MDCG 2020-5 — Guidance on clinical evaluation — equivalence
- ISO 14971:2019 — Risk management, source of residual risk identification for PMCF objectives
- MedTech Europe (2025) — Risk-based approach to PMCF: proportionality for lower-risk devices
- IMDRF N91 (2026 draft) — Updated clinical evidence framework for IVD medical devices