How classification works under the EU MDR — and where this tool stops

Regulation (EU) 2017/745 on medical devices — the MDR — has applied since 26 May 2021, replacing the earlier Medical Devices Directive. Article 51 requires devices to be divided into Class I, IIa, IIb and III according to their intended purpose and inherent risks, in accordance with the rules set out in Annex VIII.

Classification is not a formality: it determines the conformity assessment route, whether a Notified Body must be involved, the depth of clinical evidence expected, and the post-market surveillance regime. Getting it wrong at the start of a project is expensive to correct at the end.

Annex VIII has three parts, and the middle one is the part most often overlooked.

• —Chapter I — Definitions of the terms the rules turn on: duration of use (transient, short term, long term), invasiveness, active device, and so on.
• —Chapter II — Implementing rules, which govern how the classification rules are applied. Crucially: classification follows the intended purpose, not the technology; accessories are classified in their own right; and where several rules apply, the strictest one wins.
• —Chapter III — the 22 classification rules, grouped as non-invasive devices (Rules 1–4), invasive devices (Rules 5–8), active devices (Rules 9–13) and special rules (Rules 14–22).

Two consequences of the implementing rules matter more than any individual rule. First, the intended purpose is the input — the same physical object can be Class I or Class III depending on what the manufacturer claims it is for. Second, because the strictest applicable rule prevails, a device must be tested against all rules, not just the most obvious one.

MDClassify asks a structured set of questions about intended purpose, invasiveness, duration of contact, whether the device is active, and a set of special-category flags. It then evaluates the applicable rules, applies the “strictest rule wins” principle, and returns the resulting class together with every rule that matched and the reasoning behind it — not just a letter. That transparency is deliberate: a class you cannot justify is of no use in technical documentation.

All 22 rules of Annex VIII are evaluated against every classification. 13 are determined from the answers alone; 5 are determined but rest on a regulatory judgement you declare, which the report names so it can be checked; 2 cannot be settled by an engine at all and are reported as open; and 2 have a branch that is not yet implemented, listed by name. The table below is the whole of it. It is generated from the same file the questionnaire, the report and the terms read, so this page cannot say one thing while another says something else — a discrepancy a reviewer found in an earlier version of this site, and the reason the table exists.

Grouping the special rules together and calling them all “interpretive” would be easier and would be false. Rule 17 has a single outcome and no judgement in it. Rule 19 has three outcomes and cannot be settled from a questionnaire at all. Stating the limit per rule is the only version of this claim that can be checked.

Every automated classification carries the same structural weakness: it can only be as good as the intended purpose it is given. In practice, most classification disputes are not arguments about Annex VIII at all — they are arguments about how the intended purpose was worded. This tool cannot draft your intended purpose, challenge it, or notice that the claims in your marketing material contradict it. A regulatory professional can, and that is the main thing you are buying when you engage one.

• —determine whether your product is a medical device at all — borderline and combination products need a qualification assessment first;
• —settle the rules whose outcome turns on an assessment rather than on a fact — the nanomaterial exposure assessment under Rule 19, and the qualification question under Rule 14 of whether the substance's action is ancillary or principal. These are named individually in the coverage table above rather than grouped, because most of the special rules are applied in full;
• —cover in vitro diagnostic devices, which fall under Regulation (EU) 2017/746 with its own class A–D framework;
• —produce the classification rationale in the form required for technical documentation, or defend it before a Notified Body;
• —account for national implementation details, guidance updates or Notified Body practice, which shift over time.

The MDR class does not travel, which is why the same answers are run through each market's own rules rather than translated. Six frameworks are implemented today: the EU MDR, Great Britain, Australia, Switzerland, Canada and Brazil. The United States is present too, but differently — it classifies by matching a device to a generic type in a catalogue rather than by running a cascade, so it appears as a search over the FDA product classification database and the choice of code is left to you.

The divergences are not cosmetic. Surgical mesh and total joint replacements are Class III under MDR Rule 8 and Class IIb in Great Britain and Australia, whose rules descend from the directives. A closed-loop therapeutic device is Class III in Europe and Class IV in Canada. Brazil transposes Annex VIII rule for rule but relabels the scale, so its Class III is the European Class IIb.

The frameworks also move. The MHRA published a draft amendment to the UK Medical Devices Regulations in May 2026 that would introduce a more risk-proportionate, IMDRF-aligned classification framework, with entry into force expected in 2027 — meaning a UK class determined today may not be the UK class of tomorrow.

See how each market classifies, and where they diverge

The manufacturer. Under Regulation (EU) 2017/745 the manufacturer determines the class of its device and must document the rationale in the technical documentation. Neither a software tool nor a consultant transfers that responsibility away from the manufacturer.

Four: Class I, Class IIa, Class IIb and Class III, in increasing order of risk. Class I is further sub-categorised as Is (supplied sterile), Im (with a measuring function) and Ir (reusable surgical instruments), which require limited Notified Body involvement for those specific aspects.

Annex VIII contains the classification framework: definitions, implementing rules that govern how the rules are applied, and 22 classification rules covering non-invasive devices, invasive devices, active devices and a set of special cases such as software, devices incorporating a medicinal substance and devices made of nanomaterials.

Largely, yes. Class I devices that are not sterile, have no measuring function and are not reusable surgical instruments can be self-declared. Every other class requires a Notified Body, with the applicable conformity assessment route set out in Annexes IX, X and XI.

Yes, and this is a frequent source of error. The implementing rules in Annex VIII state that if several rules apply to the same device, the strictest rule resulting in the higher classification applies. MDClassify evaluates all applicable rules and reports the strictest outcome together with every rule that matched.

No. It is an indicative, informational result intended for orientation. It is not a conformity assessment, not regulatory advice, and not a determination by a Notified Body or Competent Authority.