Hydration Subsumption

For Canada's presentation Hydration Subsumption
Issue and Options

see 2025-06-05: DEUSG Virtual

Infoway Request for Drug Model enhancements_DEUSG_LP_LB_20250409 (Oslo Final) (1).pptx

https://confluence.ihtsdotools.org/download/attachments/293569471/Hydrated%20and%20Solvated%20Substances%20and%20their%20relationship%20to%20the%20description%20of%20Medicinal%20Products%2020201125.docx?api=v2

Hydration paper from 2021

Uploading new presentation for Linda:

Infoway Request for Drug Model enhancements_DEUSG_LP_LB_20250619 (Hydration).pptx

Hi Julie M. James  - Are you suggesting that we:

1. Add an |Is solvate of| relationship (or similar) to every hydrated/solvated substances, e.g. 96068000 |Amoxicillin trihydrate (substance)| is solvate of 372687004 |Amoxicillin (substance)|
2. Add the property chain: 762949000 |Has precise active ingredient| o |Is solvate of| → |762949000 |Has precise active ingredient|

If so, then I think this is a great option to consider! We're currently testing this to check that it works as expected, and if it does then we'll be happy to support this approach.

Is this what you were suggesting?

"Is solvate of" is just a specific type of "modification". The only benefit of having a new relationship is you can add that property chain - and it only affects a specific type of modification.
This softens the idea of "preciseness" if you can have something more precise. It just seems like bolting on more things, rather than fixing the underlying problem.

I'm not sure why "hydration/solvation states" got treated differently to any other modification. They're relevant for BoSS. So important (at least some of the time).
But I don't think the problem is so much about "hydration" (although that is a problem) but that the INT model/content is too precise.

374646004|Product containing precisely amoxicillin 500 milligram/1 each conventional release oral tablet (clinical drug)|
Is there a actually a real product that contains "amoxicillin" as opposed to "Amoxicillin trihydrate"?
If there is it should probably be modelled with 785686003|Amoxicillin anhydrous|.
Whatever the case there IS a problem.

The issue is the ambiguity around the definitions for some of these relationships and classes, and uncertainty about what they're supposed to be used for.
I think there's three things wrapped up here.

1) A grouper. That says these products contain the same therapeutic moeity (in whatever form) - "Has active ingredient" seems to be an attempt at that.
    The purpose of "Medicinal Product" concepts seems to be to aggregate similar (not interchangable!) products - "These are the products that contain amoxicillin"
    Useful for search. Or adverse reaction recording "patient had a reaction to some amoxicillin product".

2) Something a little more specific like clinically significant or clinically relevant (same therapeutic effect, but maybe different bioavailability or potency etc).
    Perhaps for Active Ingredient Prescribing / post coordinated prescribing instructions.

3) What is actually in the product. "The precise ingredient".
    For precise pharmacovigilance.

What we're missing is something to represent 2). Or at least 3) is being used as a proxy for - but as evidence by extensions - can be too specific.

Additionally the role chain and subrole setup with active ingredient, is modification of, and precise ingredient - obfuscate the 1)

Using the Pramipexole example from the hydration paper above.
1179006000|Pramipexole (as pramipexole dihydrochloride) 1.05 mg oral tablet|

We know the BoSS is Pramipexole

We know the "Precise Ingredient" is Pramipexole dihydrochloride (Although it's actually Pramipexole dihydrochloride monohydrate...)

There's no relationship that tells you what the "active ingredient" is. You could maybe ECL back through modification(s) and parents - but how do you know when to stop?

In this case I suspect we want:

1) active ingredient = Pramipexole

2) clinically relevant ingredient = Pramipexole dihydrochloride

3) precise ingredient (in RCDs) = Pramipexole dihydrochloride monohydrate

Though we still have the problem that the INT model uses a different BoSS to some countries (including Australia).

These are likely abstractions of the same product

Some other strengths are pretty close, and maybe with rounding can be considered equivalent too?

Note: But again the issue here isn't even the hydrate.

So the real problems are:

• Agreeing on the right level of specificity in the INT content should be. Noting that everything in INT is an abstraction (it has no branded concepts).
• Adjusting the description logic between active ingredient and precise ingredient.

I have the same question. Why should hydration, a common modification, be treated differently from other modifications? What are the business cases for modeling them differently?

The "has precise active ingredient" attribute should always contain the exact active ingredient in detail. The model aims to group products with identical active ingredients, excluding those with further modification. Therefore, the CDs modeled by “has precise active ingredient” can only subsume CDs with the exact same substance. This is why role chain is not applied to the "Has precise active ingredient" attribute.

As discussed at the Oslo meeting, general abstract CDs defined by "has active ingredient" should receive expected classifications for hydration and salt forms. This should address the challenges that CDs are too precise in the international release. Note, the CDs with precise active ingredients are still valid when the differences are clinically relevant.

I think Julie M. James gave a compelling explanation as to why hydrations should be treated differently to other types of modifications at the last DEUSG meeting (i.e. due to the weaker bonds between the water molecules). This is certainly consistent with the usage in Canada, where we want the hydration on the real/trade product, but it is the only modification that we do not want on the (virtual) clinical drug.

While I agree that we need a higher level of CD grouper concept in the international edition, to allow each country to have their own specific requirements (and still subsume under this), I think Matt Cordell's three levels - i.e. L1) active moiety, L2) clinically relevant (active moiety + BOSS), L3) precise (including hydration) is missing a level in between L2 and L3 - L2.5) precise (excluding hydration). In Canada, we're using L1, L2.5, L3. And because the Canadian regulatory authority and prescribers treat hydrations differently to other modifications, we're keen to support the distinction of |is hydrate of| from the other |is modification of| relationships in the Substance hierarchy.

We're currently testing this approach on the 234 hydrate substances found in the international edition using the ECL: (< 105590001 |Substance (substance)|: 738774007 |Is modification of| = *) {{ D term = wild:"*hydrate *", type = FSN }}. We have around 310 single-ingredient products which contain these hydrate substances and will report our findings once the testing is complete.

I have some examples that may help inform our discussions
Hydration and Grouper CDs.pptx

Again, as I said earlier "the problem is (not really) about "hydration" ... but that the INT model/content is too precise."

"Hydrates" ARE imporant when it comes to strengths (BoSS relationship).
Are they clinically significant? Different therapeutic effect etc.
Different countries may want to be more/less precise based on their own requirements (often legislation/registration requirements).
Being too precise in CORE is what's causing the problem.

The only way to fix this problem is to accept there's an issue with the model/modelling. Julie's slides illustrate this.

• Amoxicillin "Poster child" - 323509004|Product containing precisely amoxicillin 250 milligram/1 each conventional release oral capsule (clinical drug)|
  There is no such thing as a product that contains "precisely Amoxicillin".
  This concept was created as a replacement in 2021 as a replacement for 783279000|Product containing precisely amoxicillin (as amoxicillin trihydrate) 250 milligram/1 each conventional release oral capsule (clinical drug)|
  Saying "precise ingredient" means that's precisely the ingredient... it's obviously not as some extensions want to be more precise and say trihydrate.
  The intent of "precise" was to stop subsumption.
  Back engineering exceptions to handle hydrates, avoids the problem.
  
  
• Alendronic acid - 783144007|Product containing precisely alendronic acid (as alendronate sodium) 10 milligram/1 each conventional release oral tablet (clinical drug)|
  Again, this product was created as a replacement for 325974004|Product containing precisely alendronic acid 10 milligram/1 each conventional release oral tablet (clinical drug)|
  Why does the INT CD need to be so "precise"? In AMT we simply say 45274011000036109|Alendronate 10 mg tablet (clinical drug)|
  RCDs could be more precise. (Though we don't get more specific in AU)
  But UK do. They have 39728011000001102|Alendronic acid 10mg tablets (product)| with a subtype 5546111000001105|Fosamax 10mg tablets (Dowelhurst Ltd)|, modelled with a "specific active ingredient=Alendronate sodium"
  The international CD being too "precise" is causing the problem.
  
  
• Tiotropium -This is a really good/interesting example.
  I'm not sure, but I suspect it's the different inhaler devices (rather than the formulation) that affects the strength vs delivery.
  The difference between strength/dose/delivery values has been something AMT has had to grapple with - at the moment we just model strength - 18ug or 13ug - and not what's delivered.
  Thus there's nothing in the terminology indicating both products are "10 ug".
  Side note: The BRALTUS example in the slides says 16ug/acutation. At least in AU it appears to be 13ug
  Here's a snippet from the BRALTUS insert in Australia

  Braltus capsules contains 13 micrograms of tiotropium (equivalent to tiotropium bromide 15.6 micrograms) as the active ingredient.
  During inhalation, 10 micrograms of tiotropium is delivered from each capsule from the mouthpiece of the Zonda inhaler.
  Braltus and Spiriva both deliver 10 micrograms of tiotropium and are equivalent.

  SPIRIVA

  Spiriva contains 18 micrograms of tiotropium (equivalent to tiotropium bromidemonohydrate 22.5 micrograms) as the active ingredient.

  AMT doesn't mention the specific salts at all.
  This is an exceptional strength modelling scenario. But not really about "hydrate subtypes".
  
  
• SOLVATION - deemed clinically irrelevant so excluded.
  But if extensions wanted it, they could. No problem.
  So long as the INT content isn't too precise.
  
  
• Amlodipine  - This diagram implies that a "grouper" is something different. It's not. Any concept that subsumes something else is "grouping".
  Having "Precise CDs" is what's causing the problem. 
  • AU has 23517011000036108|Amlodipine 5 mg tablet (clinical drug)|
  • UK has 39732011000001102|Amlodipine 5mg tablets (product)|
  • Ireland has 267451000220105|Product containing only amlodipine 5 milligram/1 each conventional release oral tablet (clinical drug)|
    (This subsumes the 3 specific salts just as the example in the slide)
  • US (RxNorm) has 197361|amLODIPine 5 MG Oral Tablet|

What does “Amlodipine (substance)" mean? - It means Amlodipine... Or to be really pedantic Amlodipine unmodified.
What would "Amlodipine 5 mg tablet" mean - It means "a tablet containing (only) 5 mg of amlodipine" - see the Irish FSN.

I can't speak for the other countries, but in Australia a prescriber would just choose "Amlodipine 5 mg tablet" (the clinical drug).
And the pharmacist could dispense any of the specific salts (RCD).

Ireland have just modelled their CD with |Has active ingredient| rather than |Has precise active ingredient| to get the expected subsumption.
Again, the Clinical Drugs are inherently abstractions. The are not Real (clinical drugs). Thus saying "precise" is generally unnecessary (there might be some exception).

• Amitriptyline - "Amitriptyline 25 mg tablet" doesn't exist in SCT, because it doesn't exist - Amitriptyline always comes as Amitriptyline hydrochloride.
  IF there was an Amitriptyline (not HCL) 25 mg tablet, there wouldn't be a "CD Grouper" because they have different strengths. Though they would both be subsumed by 778301005|Amitriptyline only product in oral dose form|
  
  Now, I see the UK has 42291211000001105|Amitriptyline 25mg tablets| - but it is modelled with "Has specific active ingredient=Amitriptyline hydrochloride"
  I'd say this is because Amitriptyline always+only comes as hydrochloride - so it's easier to just say/write "Amitriptyline" and there's no clinical ambiguity.
  Perhaps their FSN is inaccurate. But the PT reflects clinical use.
  
  And I'm certainly not saying there's anything unusual about the UK approach.
  See this Australian website - Amitriptyline 25mg Tablets
  Our "RCD" for this product is 742031000168109|Amitriptyline (Alphapharm) 25 mg tablet| - subsumed by 22788011000036102|Amitriptyline hydrochloride 25 mg tablet|
  
  The modification is important (because it's against the BoSS). But we don't HAVE to say "Has precise active ingredient=Amitriptyline hydrochloride"
  "Has active ingredient=Amitriptyline hydrochloride" or even "Has active ingredient=Amitriptyline" could be just as applicable for an abstract concept like CD.
  With the added level of "what people expect".
  
  
• Irbesartan - I had no luck finding a "Irbesartan hydrochloride product". So ...
  But the question - "How could a user distinguish that the Irbesartan CD here is a different type of concept from the Amlodipine grouper concept?"
  This is an implementation consideration - and refsets are the easiest answer. e.g. "a prescribing refset".
  We don't allow users to select any "CD" within Australia - only the ones relevant to Australia. This is something that would apply to EVERY country.
  This is about implementation support.

• Perindopril (not in the slides) but another example like Tiotropium.
  "Perindopril erbumine 4 mg tablet" and "Perindopril arginine 5 mg tablet" - both contain the same amount of "Perindopril".
  This isn't obvious from the model (any model AFAIK). But again, comes down to what use cases are supported and implementation advice.
  (Multiple options, but beyond this topic).

So in summary. This problem is NOT an issue with hydrations. It's a problem with the INT model/content being more specific than some extensions want.
The solution is to drop "has precise ingredient" from CDs in core/INT and just use "Has active ingredient" or something like "Has specific active ingredient" (that sits between HasAI and HasPI, with an appropriate property chain).

(Thanks to Dion McMurtrie & Michael Keary for helping me with the details on this)