We have found an inconsistency with the representation of cutaneous creams within SNOMED.
In RxNorm, a Topical Cream is always over ML. We try our best to convert to Gram as that’s the most populous way it’s done in SNOMED based on our assertions.
How can an implementor know when to use ML vs. G? Is this an inconsistency, or is there any trick found within SNOMED to let an implementor know when to use the proper denominator unit? Thanks! @pwilliams
In the UK its certainly an inconsistency and normally comes from how the product has been licensed.
Ametop is weight in weight: Each gram of gel contains 40 mg tetracaine (as hydrochloride) equivalent to tetracaine base 4.0% w/w
Bifonazole is also weight in weight: The cream contains 1% w/w bifonazole.
There are very few in the UK that are w/v, though they also exist: Androfeme: One millilitre of cream contains 10 mg testosterone.
How can’t always directly convert ml to gram as one ml of a cream is not always the same density as one ml of water. We use the units that the actual product uses and don’t convert.
Can you expand on this part of the question though “How can an implementor know when to use ML vs. G? Is this an inconsistency, or is there any trick found within SNOMED to let an implementor know when to use the proper denominator unit?”
Is this for a prescribing use case and if so do you mean a pack size amount or a dosage?
I believe we needed a way to be consistent across all clinical drugs so that something weighing in at 500mg would classify as equivalent to 0.5g. If we didn’t normalize the strengths, these would show as distinct concepts because the classifier (the way we model anyway) doesn’t understand that 1000 mg === 1 g
So basically if the number is > 1000 then move to the next unit up, and if it’s < 1 then move down. If you see any cases where this rule hasn’t been followed, please do let me know.
The rules for rounding follow how Excel behaves. But that mostly relates to concentrations which are normalized to have denominator of 1. Again, so that same strength drugs can be detected eg 10mg / 5ml is the same as 20 mg / 10 ml. I understand that this causes problems when the modelling (and therefore the FSN - because it’s programatically generated) does not mirror the real world packaging and I think the DEUSG have been looking at this sort of thing.
@greynoso was asking me about this at our April Business meetings. I’ll tag @klilly and @krees also because I’d defer to them. And @mcordell because he cares.
I understand all of this and had the implementation done 7 years ago from our side.
I think this has to do with the products themselves, and possibly the source of truth they were created from. Why does the lesser numerator (125mcg) use G as the denominator unit and the greater numerator values (250mcg+) (of the same cream) use ML as the denominator units? This is where my translation is at a loss in implementation.
RxNorm is always using ML as the denominator for Topical Creams. Why the inconsistency with cutaneous creams in SNOMED? This is one example, but a good one because it’s of the same ingredients, just different strengths. There are probably others I don’t have time for right now.
Oh right, yes, sorry. Your point was that one concept uses a weight and the other a volume as a denominator, rather than anything to do with the scale. I think you’re right, it will have come in from the real world products. It would be great to bring more consistency in there, although Stuart made a good point about the difficulty of doing any algorithmic conversion.
You’d think there should be some International Standards body that could encourage manufacturers towards a common representation. Although it’s interesting that RxNorm picked volume and it sounds like the UK went weight, mostly.
For what it is worth AMT is very much as @sabbott described for the same reason - a mixture of milligram and millilitre concentration denominators.
The units used match the way the product is registered in Australia by the Therapeutic Goods Administration (FDA equivalent).
This is somewhat unsatisfying from the perspective of classification, but unfortunately the data is not always available to safely convert. So we can only represent the regulatory data we get for these products.
This issue with unit normalisation is (one of the many) reasons we deviated with AMT. And both @sabbott and @rwynne are correct- and because of this, I don’t think the core drug content is ever going to make everyone happy.
The “normalisation” @pwilliams describes actually has a negative impact the classification - when you look beyond the simple “equivalent strength” use case…
We used to have the same rules in AMT but now they only apply to descriptions.
We properly normalise ALL (not just creams) our strengths to mg and mL (where reasonable).
Both are modelled using “mg” so it’s trivial to programmatically determine 1000 > 250. Or 4x250 = 1000. etc,
The Preferred Terms use different units based on clinical safety guidelines.
Two different solutions for different audiences - Clinicians and Implementers.
Even for products like Fentanyl - that always use microgram in the descriptions (per clinical requirements). We still use milligram in the modelling (implementers like consistency and predictability).
However, our Preferred Terms are always a simple percentage (as per our clinical requirements). (Personally I think I like the idea of modeling the strength with a “%” unit also/instead (agnostic of m/v or m/m) - but there hasn’t been a community requirement for this.)
In AMT the distribution is this:
mg/mL modelling - n = 7 concepts.
mg/mg modelling - n= 161 concepts.
We clearly have a strong preference to mg/mg - the opposite to the US, which is fine.
Different jurisdictions - and why trying to agree on Clinical Drug modelling is impossible.
As noted by @sabbott and @dmcmurtrie, many regulatory agencies are at the mercy of the licensing of a product when it comes to the strength representation. In Norway, the medicine agency’s drug catalogue contains creams and ointments with both w/w% and v/v% strength representation.
As Stuart pointed out, a direct conversion is challenging and possibly imprecise, so we’ve erred on the side of caution and choose to represent the strength as given from NoMA (and thus the manufacturer) without converting anything.
Unless an international body decides to use one or the other globally, I think a normalization in this particular category of drug concepts would be a risky move - we’ll probably have to live with these differences and solve them locally for the time being.
The authoring done in the International release will be based on the product (regulatory/licensing ) information and not converted for reasons stated above.
I will however look into the issue described by Robert and the variations in strength representations for the same product. It may simply be due to different licensing bodies.
The pickle we may find ourselves in is that different jurisdictions end up with different regulatory rules when licensing products the same products. So some countries may end up with w/w and others w/v for the same product.
What happens in the international edition in those cases for CD? Do you pick one or the other (most numerous?)? Or does the international edition model different strength axioms for the different licensing around the world?
NLM had a post-doc, at the genesis of my implementation, exclaim converting a denominator of ML to G as “indefensible” (perhaps at the academic level). But when bridging cross-border medicinal products based on approved assertions by a governed body of SMEs, it shouldn’t be so scary.
This is an opportunity for consistency. And maybe not even at the denominator unit level. What if EDQM were to split cutaneous cream to more granular terms such as “liquid cutaneous cream” and, I don’t know, “solid cutaneous cream”? We would then see this as a way for implementors to get it right every time.
I seem to remember a proposal in the UK about including whether a cream was w/w, w/v, v/v in the term. Off the top of my head I can’t remember why we did not go forward with that. In SNOMED this could be modelled…potentially.
I remember thinking we should also model with %w/v and %w/w also -Though it’s more accurate - I don’t think it’s useful (and there’s other challenges).
Some example products from AMT:
These both have a strength representation of “0.05%” (though one is really 0.047%w/v)
1296401000168106 Calamine 50 mg / 1 g lotion (clinical drug)
70022011000036106 Calamine 50 mg / 1 mL lotion (clinical drug)
Both these are 5%.
I’m not sure how interested clinical end-users are in the w/v or w/w distinction?
An additional thorn - is years ago I thought it would be a good idea writing some validations for our content checking units all lined up mathematically… But abanonded that because the pack sizes don’t necessarily align with the concentration denominators
Benzoyl peroxide 40 mg / 1 g cream, 25 mL (clinical drug package)
Cetomacrogol aqueous cream + glycerol 100 mg / 1 g cream, 325 mL (clinical drug package)
Desonide 500 microgram / 1 mL ointment, 50 g (clinical drug package)
Miconazole nitrate 20 mg / 1 mL lotion, 30 g (clinical drug package)
I don’t like it. (The units can vary across brands also).
But this is the reality how these products are registered and labeled…
I think the main hurdle @rwynne is that the data just isn’t always available to be able to model the conversion. We really are at the mercy of what we get form the national regulator - if we don’t have the data points to calculate the conversion from verified data sources we can’t add it. Even if we think it is right.
@dmcmurtrie I understand your dilemma at the National level. But the MDFs in the International release are based upon EDQM. Does the MDF always have to align one to one with EDQM? Why not split the EDQM term and have two MDFs? One more solid than the other? There are two hurdles in this example. The implementor needs to know how to jump both.
@rwynne it’s an interesting suggestion. But I’m not sure splitting the MDF is the right lever, for a couple of reasons.
The MDF describes the pharmaceutical form, not the strength convention. Whether a cream is registered w/w or w/v is mostly a regulatory/labelling choice driven by what the manufacturer measured, not a distinction in the physical product. The same physical cream can end up w/w in one jurisdiction and w/v in another. Encoding that into the dose form would conflate two genuinely different concepts.
The mapping problem also relocates rather than disappears. National extensions still only have whatever the regulator gave them. They’d now have to pick between “more solid cream” and “less solid cream” based on… the strength denominator? So the implementer hurdle becomes “which MDF” instead of “which denominator” - same data dependency, different expression.
The w/v vs w/w is still explicitly modelled and therefore discoverable. For example this ECL will find the creams modelled using a strength with a volume denominator.
< 763158003 | Medicinal product (product) |:
999000031000168102 | Has concentration strength unit (attribute) | =
(*: 700000071000036103 | Has denominator units (attribute) | = < 258769000 | Unit of volume (qualifier value) |),
411116001 | Has manufactured dose form (attribute) | = << 421628006 | Conventional release cutaneous cream (dose form) |
So I think the signal implementers need is already in the modelling, even if it’s not in the FSN or the manufactured dose form.
But the key part is that you can’t assume given a certain ingredient or type of cream that it will have the same w/w or w/v or v/v strength rendering across all nations. Unfortunately the nature of the regulatory rules internationally means that can’t be assumed.
This is a bigger problem for Clinical Drug concepts internationally more broadly. Differences in regulation, prescribing, and even subsidy systems have impacts on the required content and modelling of that content from Clinical Drug down. MPF up is much more stable.
