Apologies for such a long post, especially to those for whom English is not your mother tongue. But I wanted to be as comprehensive as I could because this topic seems to rumble on and on, which is not great for anyone.
Very few substances can be (legally!) administered directly; even a line of cocaine is probably cut with something! Olive oil as an ear drop is maybe one example of a pure substance being used as a medicine; enflurane liquid is another. So dose forms are the things that take a therapeutic substance from being just a substance to being something that can be administered to a patient: they allow us to formulate a medicine for administration.
Dose forms are therefore definitional to the description of a medicine (manufactured and/or administrable) because of their intrinsic characteristics. For the purposes of this discussion I have picked out just two characteristics:
Ā· the potency of an administration (a tablet contains xxx mg of therapeutic substance, a solution contains yyymg/mL) ā without a dose form to describe a medicine, it is impossible to know the strength of a medicinal product
Ā· the level of āpharmaceutical qualityā for a medicinal product, and hence have a sense of what would be appropriate routes for the administration
I suspect, from the threads in this post, that it is this second characteristic that is causing the discussion.
Dose forms are defined and controlled by pharmaceutical standards; they are in no sense arbitrary. A tablet must be produced by dry powder compression within a dye mould, whereas a pessary is produced by the cooling of molten material in a mould that is usually in two parts. That is why clotrimazole is presented as vaginal tablets not pessaries.
Within those pharmaceutical standards, there are considerations of pharmaceutical kinetics (an oral tablet must dissolve in x minutes in ph1-2 at 37C) and of appropriate quality (an injection must be sterile and have no pyogenicity; an eye drop solution must have a pH in a particular range). As you can see, these quality standards are quite closely related to where the dose form is intended to be used in the body, and often includes a body site in the name of the monograph for that dose form.
These standards apply to ALL dose forms regardless of the therapeutic moiety that will be incorporated into them. Therefore, when designing a model to fully define dose forms (as was originally done in SNOMED CT in 2005, long before EDGM were involved) this aspect of a dose form was named āsite prepared forā or āintended siteā. Obviously, we can debate whether that was a good name for the attribute or not, but that is the name as it exists in informatic standards now. And because the value set of concepts that can fill this characteristic is not route of administration, and not strictly body site ā really the concept is adjectival rather than noun-like (ocular drops is could be considered more correct than eye drops) and because there is no body site for parenteral, or possibly even ācutaneousā or āoralā, a separate set of concepts was authored.
Route of administration is separate clinical concept. It is concerned with the administration of a particular medicinal product and its posology (so, this solution for injection containing 10mg/mL of gentamicin is suitable for intravenous administration and intramuscular injection). Although it meets the quality standards as a parenteral dose form, gentamicin injection is not suitable for subcutaneous injection, because it will be a little irritant and it will not provide sufficient therapeutic blood levels.
Route of administration is also concerned with the prescription and administration of a particular product for a particular patient at a particular point in time ā hence we see route of administration as an attribute in most information models for prescriptions (HL7 V2, V3, FHIR etc).
Clearly, there can be a ārelationship of sortsā made between a dose formās intended site of administration and the likely route(s) of administration that a medicinal product with that dose form could be used with. But it is not absolute. The relationship is absolutely not always and necessarily true.
If I had absolutely no other options to treat a Pseudomonas conjunctival infection, I could take my 10mg/mL gentamicin solution for injection and prescribe it for a patient to use by putting drops into their eye (ocular route). Because of the pharmacopoeial standards that govern the dose form āsolution for injectionā I can know that the patient will receive a sterile solution, so no risk of introducing further infection, and there will be no particulate matter, so no risk of scratching the cornea. But, because the pharmacopoeial standards for an injection solution are not as restrictive on pH and osmolarity as for an eye drop, I can be fairly sure that the patient is going to experience a lot of uncomfortable stinging. But if I am desperate to save their sight and I have no properly formulated gentamicin eye drops, I would do it.
So, for me, I could not support having a single set of concepts to describe a dose form intended site and a medicineās routes of administration, or use a single set of concepts in a value set for prescribing and dispensing. If SNOMED CT and this group decide to go that way, for the implementations I am responsible for, I will have to branch away from SNOMED CT for this, which I would find somewhat sad, and definitely not good for my user community.
I will also offer some specific comments on the threads:
The only mapping I have ever found to be useful in clinical practice is a mapping FROM route of administration TO a group of medicinal products, based on dose form site prepared for. So the clinician selects āgentamicinā and āintravenousā andā¦.the products that can be used areā¦.the children of the MPF parenteral ā which of course is based on dose form intended site 
To have a subset of (for example) of the oral MPF medications that are liquid in their administrable form for patients with issues swallowing, it is really easy to get that from the basic dose form after any transformation to work the closed loop system. Iāve had such systems successfully deployed now for several years ā including offering clinicians the option of āauthorised routeā (for a particular actual product within a local environment) versus āgeneric route setā providing the internationally accepted posologies for the particular medication.
As Stuart says, it is possible to maintain an ontology for this, and it can end up being a lot of work to maintain; but it is also possible to maintain something very clinically relevant using good rules and tools with only a reasonable maintenance overhead, especially if the raw data is available and well structured, as exists in a good medicines knowledgebase (not knowledgebase, not terminology).
I must admit I go slightly white at a statement that says āoral means something like ātaken by the mouth for systemic effect via enteral routeā ā there are so many cases where that is not āalways and necessarily trueā to list here, but the gut disinfectants and all of the antacids are good examples. And āoromucosalā is quite a complex concept to use ā particularly because if used as a route of administration, it is a grouper concept but when used as a dose form intended site, it indicates a particular set of pharmaceutical characteristics.
And of course, I have to fundamentally disagree that route of administration is definitional for the medicinal product model; it is a clinical concept (knowledge concept) that can change over time for any one medicinal product ā so it cannot be always and necessarily true and definitional. Again, there are so many examples of medicinal products that extend (or reduce) their set of routes of administration over time that I will not list them here.
I do agree that some of the methods in the dose form model need further work, especially around āswallowā and oromucosal medicines. But that is a complex topic that is beyond the core principle of how intended site concepts should nor should not be related to route of administration concepts.
I think we also need to be careful about current content. There is no such thing pharmaceutically as a specific āintraperitoneal injectionā dose form even though EDQM possibly allows it (although not specifically with those words) ; there are injection (solutions) that can be administered as an intraperitoneal infusion or irrigation (which is actually a route plus a method). Actually, if I remember correctly, the SNOMED CT dose form hierarchy contains many concepts that are non-standard. These were not retired when the EDQM mapping was done, because retirement is always a little challenging, and there was no request to do so. But if one looks at the current SNOMED CT dose form hierarchy as if it is a fully quality assured hierarchy without applying any pharmaceutical principles, I think one could be misled. Similarly, I personally do not use the EDQM concepts āwholesaleā without applying some judgement to them, because my clinical use cases are not the same as EDQMās regulatory ones.
To Marie-Alexandreās comments about EDQM routes of administration, to me this is a separate topic. As I am emphasising, route of administration is a clinical concept, and therefore not so directly relating to pharmacopoeial standards. For our pan-European work, we go some way beyond the EDQM routes of administration because our clinical use cases need more granularity and more specificity with regard to method of administration etc.
