Under MDR Article 27(4) and Annex VI Part C, the UDI must appear on the label of the device and on all higher levels of packaging in two forms at the same time: an AIDC (Automatic Identification and Data Capture) form that a machine can read, and an HRI (Human-Readable Interpretation) form that a person can read without a scanner. Accepted AIDC carriers are the ones defined by the issuing entity the manufacturer has chosen under Article 27(2) — in practice GS1-128 linear barcodes and GS1 DataMatrix 2D codes for GS1 users, the HIBC linear and 2D equivalents for HIBCC users, ISBT 128 carriers for ICCBBA users, and PZN for IFA users. RFID is permitted as an additional carrier alongside a compliant AIDC barcode, not as a replacement. For reusable devices that must be reprocessed, the UDI must also be applied directly on the device itself through direct part marking. Where the size or nature of the device makes one of the forms impossible, Annex VI Part C allows the carrier to be placed on the lowest packaging level instead, under defined conditions.

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

TL;DR

  • The UDI must appear on the label in two forms at the same time: an AIDC (machine-readable) form and an HRI (human-readable) form. Annex VI Part C of the MDR sets this as a joint requirement, not a choice between them.
  • The accepted AIDC carrier types are the carriers defined by the issuing entity under Article 27(2). For GS1 users this is GS1-128 for linear and GS1 DataMatrix for 2D. For HIBCC, ICCBBA, and IFA, the equivalent carriers defined by each entity apply.
  • The UDI is required at the unit-of-use level where applicable and at each higher packaging level up to, but not including, shipping containers. Each packaging level carries its own UDI-DI.
  • RFID is permitted as an additional carrier alongside a compliant barcode, not as a replacement for the AIDC barcode plus HRI requirement.
  • Reusable devices that require reprocessing between uses must also carry the UDI directly on the device itself through direct part marking.
  • Where a device is too small or its nature makes the label carrier impossible, Annex VI Part C allows the carrier to move to the lowest packaging level instead. The exception is defined, not open-ended.

A founder opens the label file and discovers the barcode is only half the job

The scenario is familiar. The manufacturer has chosen an issuing entity, allocated UDI-DIs, and handed the label design to the print vendor. The vendor returns a proof with a clean GS1 DataMatrix in the corner. The founder signs it off and moves on to the next task. Six months later, during the Notified Body review, the auditor points at the label and asks where the human-readable UDI is. It is not there. The DataMatrix is compliant as an AIDC carrier, but the label does not show the UDI in a form a hospital storekeeper can read without a scanner. Annex VI Part C requires both. The label goes back into revision, the printed stock is scrapped, and the launch slips.

The lesson is not that the founder was careless. The lesson is that the AIDC versus HRI distinction is easy to miss on a first reading of Article 27 and Annex VI Part C, because the phrase "UDI on the label" sounds like one thing when it is really two things applied together. This post walks through what the regulation actually requires on the carrier side, where the boundaries sit, and how a startup gets the label right the first time.

For the conceptual orientation to UDI itself, start with What is UDI?. For the article-by-article reading of the UDI provisions, see MDR Articles 27-29 UDI requirements decoded. For the identifier structure that sits behind the carrier, see UDI-DI vs UDI-PI in practice and How to assign a Basic UDI-DI.

AIDC and human-readable are a joint requirement, not a choice

Annex VI Part C of the MDR defines the general principles for the UDI carrier. The UDI carrier is the physical object on the label that holds the UDI in a form a scanner or a person can read. Annex VI Part C requires that the UDI carrier include the UDI in two forms at the same time. The first is AIDC — Automatic Identification and Data Capture — which is the machine-readable encoding the scanner parses. The second is HRI — Human-Readable Interpretation — which is the printed characters a person can read without equipment.

These are not two alternatives. They are two parts of one compliant carrier. A DataMatrix without an HRI block beside it is not a compliant UDI carrier. A row of printed characters without an AIDC barcode beside them is not a compliant UDI carrier either. Both must be present, and both must encode the same UDI-DI and the same applicable UDI-PI elements.

The reason both forms are required is operational, not bureaucratic. Hospital logistics, pharmacy dispensing, and traceability systems rely on scanners where they exist and on human reading where they do not. A broken scanner, a damaged barcode, a manual stock check, or a recall notice posted on paper all require the HRI. A normal receiving workflow relies on the AIDC. Requiring both is what makes the UDI system work at every point in the supply chain, not only at the point where the scanner works.

The HRI must be legible. Annex VI Part C does not specify a minimum font size in millimetres, but the practical test is whether a person with normal vision can read it at arm's length without a magnifier. Print vendors sometimes compress the HRI to the point where it fails this test. A label that squeezes the HRI into unreadable 3-point type is not compliant even if the characters are technically present.

Accepted carrier types and where the choice comes from

The accepted AIDC carrier types are not listed in the MDR text as a fixed catalogue. Annex VI Part C requires that the UDI carrier conform to the rules of the issuing entity the manufacturer has chosen under Article 27(2). Each designated issuing entity — GS1, HIBCC, ICCBBA, and IFA — publishes the carrier formats its system supports, and those formats are what the manufacturer implements. The ISO/IEC 15459 family on unique identification provides the general international framing for the structure of such identifiers, and the entity-specific carriers live within that framing.

For GS1 users. The accepted carriers are GS1-128 for linear barcodes and GS1 DataMatrix for 2D codes. In practice, most new medical device labels use GS1 DataMatrix because it holds more data in a smaller footprint and is what hospital scanners across Europe are already configured to read. A GS1 DataMatrix encodes the GTIN under Application Identifier (01) as the UDI-DI and appends the applicable UDI-PI elements — batch or lot under AI (10), serial number under AI (21), expiry date under AI (17), and so on — each introduced by its Application Identifier. The same identifier is printed underneath the symbol as the HRI, typically in the form "(01) 0xxxxxxxxxxxxx (17) YYMMDD (10) LOT".

For HIBCC users. HIBCC publishes its own linear and 2D code formats compatible with the HIBC (Health Industry Bar Code) standard. The carrier is structured under HIBCC data identifier conventions rather than GS1 Application Identifiers. HRI is required in the same way.

For ICCBBA users. ICCBBA publishes the ISBT 128 carrier formats used for products of human origin. For devices in the ICCBBA scope, the carrier follows ISBT 128 data structures.

For IFA users. IFA publishes the PZN (Pharmazentralnummer) carrier conventions used in the German pharmacy supply chain.

The choice of carrier flows from the choice of issuing entity. A manufacturer does not pick a barcode symbology independently of the entity. This is why the entity decision documented in how to obtain a UDI by choosing an issuing entity is the upstream decision that determines what the label will look like.

Packaging-level versus unit-level carriers

Annex VI Part C requires the UDI to appear on the label of the device and on all higher levels of packaging. "All higher levels of packaging" means every packaging layer up to, but not including, the shipping container. A device in a primary pouch inside a secondary box inside a case inside a pallet typically carries its own UDI-DI at the pouch level, a different UDI-DI at the box level, and a different UDI-DI at the case level. The pallet and the shipping container are excluded.

Each packaging level has its own UDI-DI because the quantity and configuration differ at each level. A box of ten pouches is not the same trade item as one pouch, and the GTIN — or its equivalent under another issuing entity — is different. The UDI-PI elements (batch, expiry, serial) generally track with the packaged content so the traceability chain up through the levels remains consistent.

The unit-of-use concept covers devices that are not individually packaged but still need to be traceable down to the individual unit for patient-level documentation. Where a unit-of-use UDI-DI applies, it exists in the database as an additional identifier associated with the packaging-level UDI-DI, even if it is not always printed on the unit itself. For most startup devices, the practical level to get right first is the primary label that the hospital scans at goods receipt and at the point of use.

RFID and direct part marking

Two specific carrier cases deserve attention because they often confuse founders.

RFID. Radio-frequency identification tags are permitted as an additional UDI carrier alongside a compliant barcode and HRI. They are not permitted as a replacement. A manufacturer that implements RFID on packaging for supply chain automation still needs the GS1 DataMatrix (or equivalent) and the HRI on the same label. The RFID tag is a supplementary channel. The reason is interoperability: not every downstream system reads RFID, and the Regulation cannot assume one is present. Where RFID is used, the encoded UDI on the tag must match the UDI in the barcode and the HRI. Inconsistency between the three is a finding during an audit.

Direct part marking. For reusable devices that require reprocessing between uses — surgical instruments, reusable probes, implantable trial components — the UDI must be applied to the device itself, not only to the packaging. This is direct part marking, and Annex VI Part C requires it for reusable devices where applicable. The marking must survive the reprocessing cycles the device is intended to go through without becoming unreadable. For surgical instruments, the typical method is laser marking of a DataMatrix directly onto the instrument in a location that does not interfere with function. The HRI requirement still applies in principle, though the physical constraints of very small instruments interact with the small-device exception below.

Direct part marking is a significant engineering task. It touches material compatibility, sterilisation validation, and usability. A founder who discovers the direct part marking obligation late in the project — after the instrument design is frozen — has a painful rework ahead. The right place to surface direct part marking is at the start of the UDI workstream, not at label design.

The small-device exception

Annex VI Part C recognises that some devices are too small to carry the UDI carrier on the device itself, or that the nature of the device makes application on the device impossible or impractical. In those defined cases, the UDI carrier may be placed on the lowest level of packaging that does accommodate it, and the obligation at the unit level is met by that packaging-level carrier.

The exception is not a general-purpose escape hatch. It applies where the size or nature of the device genuinely prevents carrier application. A manufacturer that relies on this exception should document the rationale — the physical dimensions, the functional constraints, the specific reason the carrier cannot go on the unit — and keep that rationale in the technical documentation. An auditor who sees the exception invoked without a documented rationale will ask why, and the answer has to be more than "it was inconvenient."

Very small devices such as orthodontic brackets, tiny surgical clips, or single-use instruments used in ophthalmic procedures are the kind of case the exception is built for. For these, the primary packaging carries the compliant AIDC and HRI, and the device itself does not.

Common mistakes startups make

  • Treating the DataMatrix as the whole requirement. A compliant carrier needs AIDC and HRI together. A DataMatrix alone is half the job.
  • Compressing the HRI into illegible type. Small HRI that cannot be read at arm's length fails the practical test even if the characters are technically present.
  • Putting the UDI only on the outer box. Every packaging level up to but not including the shipping container needs its own carrier with its own UDI-DI.
  • Assuming RFID replaces the barcode. RFID is an addition, not a substitute.
  • Discovering direct part marking late. Reusable devices need the UDI on the device itself. This is an engineering workstream, not a labelling workstream, and it must start early.
  • Invoking the small-device exception without documenting why. The exception is defined, and an auditor will expect a rationale.
  • Mismatched data across forms. The UDI encoded in the barcode, printed in the HRI, and written in any RFID tag must match. Any divergence is a finding.
  • Freezing the label with a print vendor who has never done MDR. A vendor who does consumer goods barcodes will not know the Application Identifier syntax or the HRI conventions. Choose a print vendor who has shipped MDR-compliant labels before.

The Subtract to Ship angle on UDI carriers

The carrier decision is a clean case for the Subtract to Ship framework for MDR because the work that survives is specific and the waste is visible.

The real work is: confirm the issuing entity choice, learn that entity's carrier format, design the label with AIDC and HRI together at every required packaging level, decide whether direct part marking applies based on whether the device is reusable, engineer the direct part marking process if it does, and document any small-device exception with a written rationale. That list is finite.

Everything else is a candidate for removal. The "symbology comparison" that evaluates linear versus 2D versus RFID in the abstract for a company that has already chosen GS1. The custom HRI typography project that produces unreadable characters in the name of visual consistency. The parallel RFID rollout that consumes a sprint for a supply chain that has no RFID readers. The speculative direct part marking investigation for a single-use device that will never be reprocessed. None of these trace back to Article 27 or Annex VI Part C. They come out.

Reality Check — Is your UDI carrier compliant?

  1. Does every label carry both an AIDC (barcode or DataMatrix) and an HRI (human-readable) form of the UDI at the same time?
  2. Is the HRI legible at arm's length without a magnifier?
  3. Do the AIDC and HRI encode the same UDI-DI and the same applicable UDI-PI elements?
  4. Does every packaging level up to but not including the shipping container have its own UDI-DI on its own label?
  5. If the device is reusable and must be reprocessed, have you engineered direct part marking that survives the reprocessing cycles?
  6. If you use RFID, is it in addition to the compliant barcode and HRI, not instead of them?
  7. If you invoke the small-device exception, have you documented the specific physical or functional reason the carrier cannot go on the unit?
  8. Has your print vendor produced MDR-compliant labels for a device before, or are you the first project they are learning on?

If you cannot answer six or more cleanly, the carrier design is not yet ready for the Notified Body review.

Frequently Asked Questions

Is a GS1 DataMatrix on its own enough to meet the UDI carrier requirement? No. Annex VI Part C requires the UDI carrier to include both AIDC (machine-readable) and HRI (human-readable) forms on the same label. A DataMatrix alone is the AIDC half. The printed characters beside it are the HRI half. Both are required at the same time.

Can RFID replace the barcode? No. RFID is permitted as an additional carrier alongside a compliant barcode and HRI, not as a replacement. The AIDC barcode and the HRI remain mandatory. Where RFID is used, the UDI encoded on the tag must match the UDI in the barcode and in the HRI.

Which barcode formats are accepted? The accepted formats are those defined by the issuing entity the manufacturer has chosen under Article 27(2). For GS1 users, the common formats are GS1-128 linear and GS1 DataMatrix 2D. HIBCC, ICCBBA, and IFA each define their own carrier formats. A manufacturer cannot pick a symbology outside its chosen entity's rules.

Where must the UDI appear across packaging levels? On the label of the device and on every higher packaging level up to but not including the shipping container. Each level has its own UDI-DI because each level is a different trade item.

What is direct part marking and when does it apply? Direct part marking is the application of the UDI directly on the device itself, typically by laser marking a DataMatrix. Annex VI Part C requires it for reusable devices where applicable — the marking must survive the intended reprocessing cycles without becoming unreadable. Single-use devices do not need direct part marking.

Can I use the small-device exception for my product? Only where the size or nature of the device genuinely prevents carrier application on the device itself. In those defined cases, the carrier may move to the lowest packaging level that accommodates it. The rationale must be documented in the technical file so an auditor can see why the exception applies.

Does the Commission regulate carrier standards directly? The Commission designates the issuing entities and sets the framework through the MDR and through Commission Implementing Regulation (EU) 2021/2078 on EUDAMED. The carrier formats themselves are defined by the entities, not the Commission, within the general framing provided by the ISO/IEC 15459 family on unique identification.

Sources

  1. Regulation (EU) 2017/745 of the European Parliament and of the Council of 5 April 2017 on medical devices, Article 27 (Unique Device Identification system), Article 27(4) (UDI on the label and higher packaging levels), and Annex VI Part C (the UDI system, including carrier requirements, AIDC and HRI, packaging-level obligations, direct part marking for reusable devices, and small-device exceptions). Official Journal L 117, 5.5.2017.
  2. Commission Implementing Regulation (EU) 2021/2078 of 26 November 2021 laying down rules for the application of Regulation (EU) 2017/745 as regards the European Database on Medical Devices (Eudamed). OJ L 426, 29.11.2021.
  3. ISO/IEC 15459 family — Information technology — Automatic identification and data capture techniques — Unique identification (general international framing within which the issuing entities' carrier formats sit).
  4. Issuing entity carrier documentation — GS1 General Specifications (GS1-128, GS1 DataMatrix, Application Identifiers), HIBCC HIBC standard, ICCBBA ISBT 128 standard, IFA PZN documentation. Confirm current versions with each entity directly.

This post is part of the EUDAMED, UDI and Registration category in the Subtract to Ship: MDR blog. Authored by Felix Lenhard and Tibor Zechmeister. For the upstream entity decision, read this post alongside how to obtain a UDI. For the downstream database submission, read UDI database registration in EUDAMED.