MDR Article 88 requires every manufacturer to report any statistically significant increase in the frequency or severity of incidents that are not serious incidents, or of expected undesirable side-effects, where that increase could have a significant impact on the benefit-risk analysis and has led or may lead to unacceptable risks to patients, users, or other persons. The baseline, the statistical method, and the trigger criteria must be defined in advance in the PMS plan under Article 84 and in the technical documentation under Annexes II and III. Reports flow through Eudamed once the vigilance module is fully operational; in the interim, the applicable national reporting channels apply. The authoritative interpretive reference is MDCG 2023-3 Rev.2, revised January 2025.

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

TL;DR

  • Trend reporting MDR Article 88 is the obligation to report statistically significant increases in non-serious incidents or expected side-effects that could shift the benefit-risk balance into unacceptable territory.
  • Article 88 is distinct from Article 87. Article 87 reports individual serious incidents and field safety corrective actions. Article 88 reports patterns in events that do not individually meet the Article 2(65) serious incident threshold.
  • The manufacturer must predefine, in the PMS plan under Article 84, the baseline frequencies and severities and the method for determining what counts as a statistically significant increase. The Regulation does not prescribe the method, but it does require one.
  • Trend reports are submitted through the Eudamed electronic system under Article 92 once that module is fully operational, with national reporting channels used in the interim.
  • MDCG 2023-3 Rev.2 (January 2025) is the primary Q&A reference for interpreting trend reporting decisions and should be open whenever a borderline call is made.

The arm strap that made the trend obligation concrete

A small MedTech company we worked with shipped a sleep-monitoring device worn on the upper arm. The hardware passed every biocompatibility test before market. Weeks into real-world use, skin irritation reports started arriving. Most were mild. None, individually, met the serious incident threshold under Article 2(65). Classified one at a time, the Article 87 clock never started on most of them.

Then the PMS review compared the observed rate against the baseline the team had written into the PMS plan before launch. The observed frequency was meaningfully above the foreseeable rate. That was the Article 88 call. The individual events were not serious. The aggregated pattern was — and the benefit-risk balance was shifting. A trend report was filed, a material investigation was opened, and the specification was updated. The obligation that caught this was not Article 87. It was Article 88.

That story is the reason Article 88 exists as a distinct provision. Without it, slow-moving patterns made of individually-non-serious events would never reach the regulator until somebody got hurt badly enough to cross the serious incident threshold on their own.

What Article 88 actually requires

Article 88(1) of Regulation (EU) 2017/745 requires manufacturers to report any statistically significant increase in the frequency or severity of incidents that are not serious incidents, or that are expected undesirable side-effects, that could have a significant impact on the benefit-risk analysis referred to in Sections 1 and 5 of Annex I and which have led or may lead to risks to the health or safety of patients, users, or other persons that are unacceptable when weighed against the intended benefits. The significant increase is established in comparison to the foreseeable frequency or severity of such incidents in respect of the device or category or group of devices in question.

Paragraph 2 requires the manufacturer to specify how to manage the events concerned in the PMS plan referred to in Article 84 and in the technical documentation referred to in Annexes II and III, and to determine what constitutes a statistically significant increase. Paragraph 3 establishes the reporting path through the electronic system under Article 92.

Three operative phrases carry the obligation. "Statistically significant increase" means the manufacturer needs a predefined method. "Foreseeable frequency or severity" means the baseline has to be documented before the events start arriving. "Could have a significant impact on the benefit-risk analysis" means the threshold is not raw statistics alone — it is statistics interpreted against the clinical benefit and risk posture of the device.

Defining the trend — what counts as "foreseeable"

The foreseeable baseline is the anchor of the whole Article 88 framework, and it has to exist before the product ships. The baseline comes from several sources in combination.

Expected undesirable side-effects described in the clinical evaluation and the technical documentation — quantified where possible, with ranges rather than single numbers when the literature supports a range. Incidents known from similar devices on the market, captured through the literature and similar-device monitoring streams of the PMS system under Annex III. The risk management file under EN ISO 14971:2019 + A11:2021, which states the estimated occurrence rates for each identified hazard and feeds the baseline directly. Any pre-market clinical investigation data that quantifies the observed rate of specific events in the study population.

What counts as "foreseeable" is never a single number pulled from a vacuum. It is a documented, traceable estimate built from the evidence the manufacturer has in hand at the time the PMS plan is written. When the observed real-world rate diverges meaningfully from that estimate, the Article 88 question activates. When there is no documented baseline, the manufacturer has no defensible way to answer the question at all — which is itself a non-conformity against Article 88(2).

Statistical methods — what the Regulation does and does not prescribe

Article 88 does not name a statistical method. Manufacturers frequently ask which test the Regulation requires. The answer is that the Regulation requires a method to exist, to be defensible, and to be written down in the PMS plan. The choice of method is left to the manufacturer, proportionate to the device, the data volumes, and the type of event.

In practice, the methods used across MedTech fall into a small number of families. Control-chart approaches based on Shewhart, CUSUM, or EWMA charts set upper control limits against the baseline rate and flag excursions. Poisson or binomial hypothesis tests compare observed event counts against the expected count over a defined observation window. Rolling-window frequency comparisons look at moving averages against a historical baseline. Bayesian updating approaches recompute a posterior event rate as new data arrives and flag when the posterior diverges from the prior beyond a pre-agreed threshold.

None of these is correct in every situation and none is wrong in every situation. The defensibility test is the test a Notified Body auditor will apply: is the method documented in the PMS plan, is it appropriate for the data volume and event type, does it produce reproducible results, and is the threshold for triggering a trend report written down before the data is analysed rather than after. Retrofitting a method after a trend has already appeared is the worst failure mode. It invariably looks like the manufacturer chose the method to match the conclusion.

MDCG 2023-3 Rev.2 does not prescribe methods but does describe the expectation that the method is proportionate, predefined, and documented. For low-volume devices where classical significance testing is underpowered, the Q&A endorses qualitative threshold definitions as long as they are defensible and written down.

Detection in the PMS system — where the signal comes from

Trend detection is a downstream use of the PMS data streams described in Annex III. The same streams that feed complaint handling, vigilance classification, and the post-market surveillance system feed the Article 88 trend analysis. There is no separate "trend reporting data collection" activity. There is the PMS data — complaints, service reports, literature, similar-device monitoring, user feedback, PMCF findings — filtered for events that are not serious incidents but are in scope for the baselines defined in the PMS plan.

The operational loop runs on a defined cadence. Every review cycle, the team compares observed rates against predefined baselines using the predefined method. If the method flags a trigger, the event goes into a documented decision log. A classification is made: is this a statistically significant increase, and does it have the potential to shift the benefit-risk analysis toward unacceptable. If yes, a trend report is prepared. If no, the reasoning is written down and referenced against MDCG 2023-3 Rev.2.

The cadence is not open to improvisation. Monthly is defensible for most devices. Quarterly can be defensible for very low-volume stable devices. Waiting until the annual PSUR cycle to run the Article 88 analysis is not defensible — the Article 88 obligation is continuous, not annual, and a trend that emerged in month three and was only noticed in month eleven is a trend the manufacturer did not detect in time.

When to report, when not to report

Four conditions must all be true before Article 88 reporting is triggered. The increase is in non-serious incidents or expected undesirable side-effects, not individual serious incidents (serious incidents go under Article 87 and the Article 2(64)–(65) definitions). The increase is statistically significant against the predefined baseline using the predefined method. The increase could have a significant impact on the benefit-risk analysis under Annex I Sections 1 and 5. And the resulting risk to patients, users, or other persons is unacceptable when weighed against the intended benefits.

Situations that do not meet all four conditions are still handled inside the PMS system and feed the PMS Report or PSUR under Articles 85 or 86, but they are not Article 88 trend reports. A frequency increase that is statistically significant but does not affect the benefit-risk balance is monitored and documented, not reported. A benefit-risk-relevant event pattern that fails the statistical significance test under the predefined method is also documented but not reported, with the reasoning written down against the specific method and threshold in the PMS plan. Every non-reporting decision has a documented rationale referencing the MDR text and the relevant MDCG 2023-3 Rev.2 Q.

When the four conditions are met, the report is filed through Eudamed once the vigilance module under Article 92 is fully operational, and through the applicable national reporting channel in the interim. The content of the trend report includes the baseline, the observed data, the method applied, the statistical result, the benefit-risk assessment, and the corrective or preventive action the manufacturer is initiating.

How MDCG 2023-3 Rev.2 interprets the edge cases

MDCG 2023-3 Rev.2 (January 2025) is the interpretive reference every trend reporting decision should be checked against. The Q&A document covers, among the twenty-four questions, the distinction between Article 87 and Article 88 reporting, the treatment of expected side-effects that were quantified in the technical documentation versus unexpected side-effects that were not, the handling of use-error patterns that cluster without any single event meeting the serious incident threshold, the interaction between trend reporting and periodic summary reporting, and the expectation that the trend-reporting method is predefined and proportionate.

Two interpretive points from MDCG 2023-3 Rev.2 come up repeatedly in practice. First, an "expected undesirable side-effect" is only expected in the Article 88 sense if it was quantified in the technical documentation — a vague mention in the IFU is not sufficient to treat the side-effect as expected. Second, a cluster of non-serious events that reveals a hazard not previously identified in the risk management file is a trend signal even if the individual events are mild, because the benefit-risk balance has shifted on account of the new hazard rather than the frequency alone.

The general principle: when in doubt, open MDCG 2023-3 Rev.2, find the closest Q, and document the decision against it. Unexplained divergence from the Q&A is a risk that a startup does not need to carry.

Common mistakes

  • No predefined baseline. The PMS plan names "trend monitoring" as an activity but does not define the foreseeable frequencies or severities for any event category. Article 88(2) is failed before any data arrives.
  • No predefined method. The plan says the manufacturer will "monitor for statistically significant increases" without naming the method, the threshold, or the cadence. The first trend analysis performed under this setup is always retroactive.
  • Treating Article 88 as a subset of Article 87. Teams assume that if no individual event is a serious incident, there is nothing to report. The Article 88 obligation is exactly the gap Article 87 does not cover.
  • Waiting for the PSUR cycle. Trend monitoring is continuous. Discovering a trend in the annual PSUR preparation window and back-dating the detection is not defensible and leaves a paper trail an auditor will find.
  • Confusing "expected" with "mentioned." A side-effect is expected in the Article 88 sense only if it was quantified in the technical documentation. A side-effect mentioned in the IFU without a quantification is not an expected side-effect for trend-reporting purposes.
  • Ignoring severity trends. Article 88 covers increases in frequency or severity. A stable frequency with a drift toward higher severity of the same event category is a trend signal under the Regulation.
  • Running the method only when a complaint feels alarming. The method runs on the cadence, every cycle, regardless of subjective alarm. Selective execution defeats the purpose and fails the audit.

The Subtract to Ship angle

The Article 88 obligation tempts startups into two failure modes. The first is building an elaborate statistical framework nobody can actually run — multivariate models, dashboards, significance tests configured for data volumes the device will never produce. The second is doing nothing and hoping trends do not occur.

The Subtract to Ship framework for MDR rejects both. The lean Article 88 system has five components and no more. A documented baseline for each event category that matters, built from the risk file, the clinical evaluation, and any pre-market data. A predefined method proportionate to the data volume — for most Class I and IIa startups, a simple control-chart or Poisson-test approach is defensible. A cadence — monthly for most devices, with the review documented and signed. A decision log capturing every threshold excursion and the reasoning for reporting or not reporting, cross-referenced to MDCG 2023-3 Rev.2. A path into the CAPA process under EN ISO 13485:2016 + A11:2021 and the risk file under EN ISO 14971:2019 + A11:2021. That is the whole system. Every component traces to Article 88, Article 84, Annex III, or the Q&A. Anything beyond it is ceremony.

Reality Check — Where do you stand?

  1. Is the foreseeable baseline for each monitored event category documented in your PMS plan today, with the evidence it was built from?
  2. Is the statistical method for determining a significant increase named in the PMS plan, with the threshold written down before the data arrives?
  3. Is the trend review cadence defined and actually run on schedule, with each review documented and signed?
  4. Can you name the difference between an Article 87 report and an Article 88 report for your device, with a concrete example of each?
  5. If a non-serious-event cluster appeared this month, who would classify it and against which MDCG 2023-3 Rev.2 Q would the decision be checked?
  6. Does your decision log contain non-reporting decisions as well as reporting decisions, with written reasoning referencing the Regulation and the Q&A?
  7. Is your trend monitoring connected to the risk file and the CAPA system, or does it live in a spreadsheet nobody else sees?

Frequently Asked Questions

What is trend reporting under MDR Article 88? Trend reporting under MDR Article 88 is the manufacturer obligation to report any statistically significant increase in the frequency or severity of incidents that are not serious incidents, or of expected undesirable side-effects, where that increase could significantly impact the benefit-risk analysis and the resulting risks to patients, users, or other persons are unacceptable when weighed against the intended benefits. It is distinct from the Article 87 obligation to report individual serious incidents.

How is Article 88 different from Article 87? Article 87 reports individual serious incidents meeting the Article 2(65) threshold and field safety corrective actions. Article 88 reports patterns in events that do not individually meet the serious incident threshold — non-serious incidents and expected side-effects whose aggregated frequency or severity increase shifts the benefit-risk balance toward unacceptable. The two articles are complementary and both are required.

Does the MDR prescribe a specific statistical method for trend reporting? No. The Regulation requires the manufacturer to determine what constitutes a statistically significant increase under Article 88(2) and to document the method in the PMS plan under Article 84 and in the technical documentation under Annexes II and III. The choice of method is the manufacturer's responsibility and must be proportionate, defensible, and predefined — not retrofitted after the data arrives.

How is "foreseeable frequency or severity" determined? From the evidence available when the PMS plan is written: the risk management file under EN ISO 14971:2019 + A11:2021, the clinical evaluation, the expected undesirable side-effects quantified in the technical documentation, the literature on similar devices, and any pre-market clinical investigation data. The baseline must be traceable to documented sources, not invented.

Where is the trend report submitted? Under Article 88(3), trend reports are submitted through the Eudamed electronic system established by Article 92 once the vigilance module is fully operational. In the interim, reports flow through the applicable national competent authority channel as the competent authorities direct. MDCG 2023-3 Rev.2 describes the interim path.

Does Article 88 apply to Class I devices? Yes. Article 88 applies to every manufacturer of every class of device placed on the Union market. The method and baseline are proportionate to the device and the data volume, but the obligation itself does not scale with team size or device class.

Sources

  1. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, Article 2(64) (definition of incident), Article 2(65) (definition of serious incident), Article 84 (post-market surveillance plan), Article 87 (reporting of serious incidents and field safety corrective actions), Article 88 (trend reporting), Article 92 (electronic system on vigilance and on post-market surveillance), Annex I Sections 1 and 5 (general safety and performance requirements, benefit-risk analysis), Annex II (technical documentation), Annex III (technical documentation on post-market surveillance). Official Journal L 117, 5.5.2017.
  2. MDCG 2023-3 Rev.2 — Questions and Answers on vigilance terms and concepts as outlined in Regulation (EU) 2017/745 and Regulation (EU) 2017/746, first publication February 2023, Revision 2 January 2025.
  3. EN ISO 13485:2016 + A11:2021 — Medical devices — Quality management systems — Requirements for regulatory purposes.
  4. EN ISO 14971:2019 + A11:2021 — Medical devices — Application of risk management to medical devices.

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. Article 88 is the article most startups forget exists — and the one that catches the slow-moving patterns the rest of the vigilance framework is not built to see.