Borderline products and regulatory classification
Some products do not obviously belong to any regulatory framework, and until that is settled there is no template, no defined evidence package and no agency to submit to. This topic covers what a writer does with a product like that: how the frameworks are told apart, why the answer changes everything downstream, how the classification question is taken to a regulator, and how a safety programme with no standard shape is justified in writing. Where the product is a medicine and a device deliberately combined, the question is which framework covers which half and drug-device and device-drug combination products owns it; the device framework by itself is in medical devices.
Recognise the product that fits nowhere
The account that follows is the last talk on the agenda of an EMWA expert seminar on nonclinical regulatory writing, held at the association's 57th conference in Valencia in May 2024 and opened by its chair as the first time the association had put a spotlight on the subject. Its speaker is a European Registered Toxicologist who worked for fifteen years at the Dutch national institute for public health and the environment and has been a freelance scientific and medical writer since 2021, providing consultancy to governmental agencies, academic research institutes, contract research organisations, biotech companies and consultancies; her expertise is the safety assessment of borderline products, and she works in the European frameworks rather than the American ones. She is introduced under one rendering of her name and greeted under a different one by a questioner, and the transcript does not settle which is right, so she is named here by her subject rather than by a guess. The record is Vimeo's automatically generated English captions grouped by minute and not manually corrected, so institution names, regulation titles and figures in it are approximate. What the talk reports is the position as of May 2024.
A borderline product, in her working definition, is an often quite complex product for which there is uncertainty over which European regulatory framework applies. The frameworks a product can fall between include cosmetics, medical devices, medicinal products, food, supplements and consumer products, and the simplest illustration is a cosmetic cream carrying an active ingredient: cosmetic, or medicinal product?
Her worked example is a capsule of fermented red rice containing monacolin, which the captions render as monacan, an ingredient with cholesterol-inhibiting effects similar to lovastatin, itself an active ingredient in medicinal products. Asked to classify it by a show of hands, about half of the room voted food supplement and fewer than half voted medicinal product, with the remainder abstaining; the size of the audience in the room is not recorded, so the split is a demonstration of difficulty rather than a measurement. The answer is that the German authorities classified the product as a medicinal product on account of its active ingredient, the supplier disagreed and took it to court, and the European Court of Justice ruled that it was a food supplement despite the active ingredient — a decision resting mostly on the product's purpose, which was to be used as a food supplement.
The cosmetic half of that example has a definition a writer can apply directly, and applying it is a drafting decision as much as a regulatory one. Lorine Preud'homme and two colleagues at Biorius in Belgium set it out in the September 2014 issue of EMWA's journal, and their account is the position as of that date. Under the European cosmetics regulation a product is classified as a cosmetic on four criteria — function, presentation, mode of application and composition — and all but composition are evaluated from the claims made on the label, so the text is what puts the product inside or outside the category: to remain a cosmetic, the manufacturer cannot refer to treating, preventing or alleviating a disease. A product departing from the definition on any criterion can be treated as a non-compliant cosmetic and assigned to another category such as a medical device, a room fragrance or a general consumer product, and a product fitting no alternative category becomes a non-classified or borderline product whose category may then be read differently in different member states. That issue and the June 2017 medical-devices issue drawn on below were admitted from existing Markdown conversions without content repair, and their known limitations are broken ligatures, mangled multi-column tables, interleaved page headers and footers, and figure text represented as marked text. The cosmetic product information file covers what a product that stays inside the definition then needs.
How far the medical device category stretches at the ordinary end is worth knowing before arguing about the hard cases. The guest editor of the June 2017 medical-devices issue of EMWA's journal makes the point through everyday examples: there are more than 500,000 types of medical and in vitro test device on the EU market, including plasters and contact lenses alongside the x-ray machines and hip replacements that come to mind first — and medical leeches are classified as medical devices, a point for which the editorial cites the Medicines and Healthcare products Regulatory Agency's June 2013 guidance on the borderlines between medical devices and medicinal products.
Know what the classification decides
The classification is not a labelling question, because the frameworks impose entirely different processes. A medicinal product requires pre-market authorisation, with a body of safety data supplied and scrutinised by regulators to establish that the product is safe. A food supplement has no such procedure: the manufacturer is responsible for the safety of what it puts on the market, no dossier is submitted or evaluated in advance, and the route is essentially a notification. Enforcement falls to the member states where the product is marketed, which may ask a supplier to show its safety data — but often do not.
She reports one event that gives the gap its weight, and reports it with her own uncertainty intact: between the first and second drafts of her presentation she saw a news item that the food supplement from the quiz had been taken off the market in Japan after being linked to five deaths. She says she does not know what Japanese regulations require, that she is not sure whether the deaths related to the safety of the active ingredient or to a contamination or to something else, and only that it made her wonder whether the product had been appropriately assessed for safety before being marketed. The prose here follows her hedges deliberately; the source establishes a question, not a causal finding.
Work the definitions one at a time
Her second worked case is a European-funded project, nTRACK — the captions render the name as NTRK — that developed nanomaterials to track stem cells. The bottleneck it addressed is that once stem cells are injected, nobody knows where they go or what happens to them; the proposal was to label them with a nanoparticle having a magnetic core followed by MRI and a gold shell followed by CT, glucose molecules on the surface to help uptake into the cells, and a radioactive label added for the animal studies. Asked to classify that particle as a small molecule, a biological or an advanced therapy medicinal product, the room split across all three with a few hands each — and the speaker's answer was that everyone was wrong.
The classification work that produced that answer started somewhere unexpected: with defining the product's purpose precisely. The original project proposal was, in her phrase, all rainbow and unicorn — the nanomaterial was to track the particles over time non-invasively and quantitatively and report how the stem cells were doing — and it was too ambitious. What the team settled on after discussion was narrow and it is the narrowness that drives everything after it. The particles would label stem cells during the development of a stem cell therapy; they would be marketed as standalone products and not as part of an advanced therapy medicinal product; they would be used only before marketing authorisation, to help define the correct dose of stem cells; they would not form part of the final therapy that reached the market; and they had to be compatible with CT and MRI. A writer who cannot state the purpose in those terms cannot classify the product, which is the transferable step.
Each candidate definition was then tested against that purpose and failed for a stateable reason. A medicinal product has to have properties that modify physiological functions by metabolic or immunological action, and the particles were not intended to have any effect on the body at all — the cells would do that. The diagnostic agent subcategory did not fit either, because the particles track a treatment rather than diagnose an illness. The combined advanced therapy category did not fit because the product was to be marketed on its own. The medical device definition came closest: a material used in human beings for the purpose of monitoring, though of a treatment rather than of a disease, an injury or a physiological process, and one whose definition excludes pharmacological, immunological and metabolic means — a better fit than a medicinal product without being a good one. She notes she was working from the older device definition, which she says has not changed much. An excipient did not fit because an excipient is a constituent of a medicinal product and this was never to be part of one, and a research-use-only product did not fit because it would be used in vivo.
A short orientation on the material helps here, because the classification difficulty follows from it. A nanomaterial can be manufactured from more or less anything — iron, silica, titanium dioxide, gold, or any combination of them — with coatings and with all kinds of molecules attached to the surface, which is what makes them so complex. They are expected to revolutionise medicine because they can be given very specific targeting: load a nanomaterial with a drug, then give the particle properties that direct it to where the drug is wanted. And their scale is what makes them biologically active in the first place: a nanomaterial is in the same size range as an enzyme, slightly bigger than DNA and much smaller than a cell, so it is taken up by cells.
Take the question to a regulator, and know what each route can settle
Two consultation routes were used, and their differences are the practical content of this section. Scientific-advice routes beyond the standard procedure covers the briefing packages both of them require, and scientific-advice briefing documents how such a package is written.
The European Medicines Agency's Innovation Task Force is the informal one. A developer of an innovative product with scientific, regulatory or technical questions sends a request to the agency, and if the agency thinks the case is a fit, asks for a briefing document — around 30 pages outlining the product's properties and setting out the questions to be discussed. The meeting is nowadays virtual, the agency may invite experts from outside itself, a report is written afterwards which the agency reviews for alignment with what was discussed, and the whole thing is free of charge. It is genuinely a dialogue, which means expecting the agency to ask questions back: one it asked here was whether the particles would stay inside the cell or leave it and travel in the body, in which case the cells would not in fact be what was being tracked. Asked afterwards what can be brought to it, she describes the scope as broad — any technological, scientific or regulatory question; help thinking about why a product is or is not a diagnostic, though never a final classification; what safety studies and safety aspects will need to be investigated; and how to organise the quality assessment of an innovative product, so chemistry, manufacturing and controls questions too. The experts invited depend on what is in the briefing dossier, and the meeting is a discussion rather than a presentation of data for judgement. The panel it convenes is largely agency employees with a regulatory background, with some scientists, plus experts from universities or anyone else in the agency's network it thinks can help.
What that route cannot do is classify. Classification is a member-state responsibility resting with the national competent authorities, and the agency could not settle it — although the discussion was useful, and revealed that even within the agency there were diverging views about how the particle should be classified, with most leaning as the project did towards a medical device. Its recommendation was to choose one classification, go for it, and justify the choice in the submission. That is the operative advice in the whole talk. The task force also reported the development as highly relevant, because a great many of the developers who use it are stem cell therapy developers and the same bottleneck — where do the cells go after injection — keeps recurring.
Scientific advice at a national competent authority is the formal counterpart, and having chosen the device classification the project had to go there, because for a device it is the national authority rather than the agency that advises on the preclinical safety evaluation programme and on the clinical trial protocol. It is voluntary but formal, similar in that to the agency's own scientific advice for medicinal products, and the rule that changes how a writer prepares for it is that open questions are not allowed: you must arrive with your own proposed preclinical safety evaluation programme, and the authority tells you whether it agrees. The briefing dossier here ran to about 65 pages covering the physicochemical properties, the manufacturing process and how manufacturing quality would be evaluated — very complicated for nanomaterials — together with the specific questions and the project's own position on each. The project convened a joint meeting with two German authorities chosen for complementary experience: the institute responsible for biological medicinal products, for its experience with advanced therapies, and the authority responsible for medicines and medical devices, for its experience of clinical trials with stem cells. The captions damage both names: the first is the Paul-Ehrlich-Institut, which they give as the Paul Institute for Biological Medicinal Products, and the second is the Federal Institute for Drugs and Medical Devices, known by its German initials BfArM, which they give as B rm.
The scientific advice committee says whether it agrees with the applicant's position on each question and, where it does not, why not.
The project also wanted the authorities' advice on the regulatory status of the particles, and that was what they chose to discuss first. Their answer repeated the pattern and sharpened it. Compared with the other definitions, medical device fits best — and it is not a medical device, because it is never marketed with a medical purpose for an individual patient and is used only in the development phase. Since it would nonetheless be injected into humans, everyone agreed the preclinical programme had to be very safe, and the authorities largely agreed the programme the project proposed: pharmacokinetic and toxicity studies of the particles, of the cells, and of the combination of the two — a very large safety evaluation package. A classification can therefore be settled well enough to submit against without ever being correct, which is a different outcome from the one a writer expects to be waiting at the end of the process.
The classification confusion is not confined to Europe. Asked why the project chose medical device when, in the questioner's recollection, a very similar product had been classified in the United States as an investigational medicinal product, she answers that a workshop the project held with participants including people from the American agency was told that a similar product had in fact been classified there as an excipient — which, she concludes, is what a borderline product looks like: people do not know what to do with it. In the exchange that closes the talk the session's chair says an excipient would have been the first guess, and the captions, which do not mark the change of speaker there, go on to say that in the States something similar is classified as an excipient but treated as a medicinal product.
There is also a published record of how borderline classification questions have been answered before, and it is the first place to look. Karin Eichele's resource column in the June 2017 medical-devices issue of EMWA's journal points to the European Commission's manual on borderline products, issued and updated to harmonise interpretations and protect the single market principle, which describes borderline classification cases and serves as a decision tool for the member states — the decision itself generally falling to the national authorities.
What the manual demonstrates is that classification depends on the purpose the product serves and not only on the product. Her example is shoe covers distributed in a hospital: covers intended by their manufacturer for use in operating rooms, intensive care units or with immunodepressed patients to protect the patient from potential contamination are medical devices, while shoe covers for visitors, even in a hospital, are products for control of the environment.
The same logic produces cases that look absurd and are not. A discussion of YouTube videos for treating insomnia — relaxing noises intended to help a patient calm down and fall asleep, not materially different from meditation videos, which attract no such discussion — would formally meet the criteria for a medical device because of the claim that they treat an illness. Eichele's point in raising it is not the video but the direction of travel: with the increasing number of health applications on mobile devices, the classification of eHealth tools might become a regulatory issue, and she expected the regulations to go on being updated to address unsolved classification and assessment issues.
Expect a non-standard safety programme, and justify it
The resulting programme is far from the standard list of studies the seminar's earlier speakers had set out, and the reason is physical. Nanoparticles behave very differently from small molecules: a combination of chemical and physical properties — not composition alone, but size, coating and surface charge — determines where they go in the body and how they behave there, and that also breaks a number of the standard toxicity studies. The clearest broken test is the Ames test, which the captions have as the AIMS test and which adds material to bacteria and looks for mutations: nanomaterials do not cross the bacterial wall, so the result is a false negative. She calls it a well-known example that a lot of people are nevertheless unaware of, and notes that regulators are becoming more aware of this class of problem.
The guidance landscape is correspondingly thin and the useful material is in other frameworks. For nanomedicinal products there were no guidances specific to nanomaterials, only reflection papers the European agency published some years earlier for particular types. For medical devices there is an ISO guidance on nanomaterials that accounts for the artefacts that arise in testing them. And a great many OECD toxicity study guidelines developed for chemicals and pesticides, where nanomaterials are also used, are useful for the nano-specific considerations a safety evaluation has to take into account. Asked afterwards whether the system should have been treated as a nanomedicine, given that the agency now recognises nanomedicines and there are guidelines for preparing a marketing authorisation application for them, she answers that as far as she knows the agency has published only the reflection papers — the project had itself encouraged it to produce nanomedicine guidelines — and that there is nothing amounting to a box to tick declaring a product a nanomedicine, whereas medical devices do have a nanomaterials guidance. The questioner supplied a link and the exchange was not resolved in the session.
The thinness of that guidance landscape was not new in 2024. Simone Lerch, of nspm ltd in Switzerland, described it in the September 2014 issue of EMWA's journal already drawn on above. She recorded that nanomedicines were being regulated by both the European and the American agency under the same procedures used for small-molecule pharmaceuticals or medical devices; that a lack of definitions for nanomaterials and limited standard nomenclature and reference material were themselves a difficulty for the regulators; and that the draft guidance papers both agencies had issued were expected to be precursors of nanomedicine-specific requirements. Read the speaker's answer above against that expectation: a decade later she knew of nothing from the European agency beyond the reflection papers. Nanotechnology and nanomedicine carries the science and the exposure evidence behind these products.
Write for a field that has no template
Her closing advice to writers of borderline and innovative products is a short list, and she expects to need it more often as advanced material technologies produce more complex products that reach nanomedicine and medical devices.
Get involved at the earliest stages of product development, which she frames explicitly as the same point the seminar's first speaker made about beginning with the end in mind, so that you can follow the discussions on primary purpose and regulatory classification as they happen rather than inheriting their conclusions. Organise the early informal dialogues — the innovation task force, and scientific advice — for anything at all innovative, without expecting definitive answers: the task force is partly there to learn from you about the products coming its way, and what you get from it is the regulator's way of thinking about your class of product and therefore what information you will have to generate. Find any information you can on similar materials or products and how their safety evaluations were executed.
Then expect the development programme itself to be non-standard, containing studies you have not seen before, and expect to justify it thoroughly — a point she attributes to an earlier speaker in the session, that anything non-standard has to be explained. Be extra critical of the results of non-standard studies: whether they look too good to be true or simply odd, and whether the material could be interfering with the test, which is a question worth putting to a toxicologist specialising in these materials. Make every assumption and every uncertainty explicit, because the regulators will pick them up and ask about them anyway. And be aware that standard file formats and test methods may be incompatible with what you are working on.
Her own summary of the compensation is worth keeping: compared with regulatory medical writing this field is a lot less standardised, and that also means there is room to be more creative in it.