Open dih-cdisc opened 6 months ago
See original Sketch for the intervention classes. We were unsure where to add formulation.
From @BSnoeijerCD email
Furthermore, per protocol I indicated what intervention information could not yet be stored in USDM. See the summary below:
Based on the above, I would say that kind of a generic interventionProperty class could work for a number of these items?
Other model issues:
Made this a more general ticket for interventions
Observational studies should not include an intervention. Types of interventions in Type CT we do not handle yet:
@dih-cdisc : see sketch below based on the following considerations:
2 more considerations to note:
Aligning to ISO IDMP
See overview document reference above (bullet 2): ISO IDMP is composed of five separate standards. The specific standards are: ISO 11238: Data elements and structures for unique identification and exchange of regulated information on Substances; ISO 11239: Data elements and structures for unique identification and exchange of regulated information on pharmaceutical dose forms, units of presentation, routes of administration and packaging; ISO 11240: Data elements and structures for unique identification and exchange of units of measurement; covered in quantity class ISO 11616: Data elements and structures for unique identification and exchange of regulated pharmaceutical Product information; ISO 11615: Data elements and structures for unique identification and exchange of regulated medicinal Product information. ISO IDMP standards apply to Human medicinal products and investigational medicinal products.
@dih-cdisc See here the proposal based on the IDMP / FHIR alignment (based on references above). The ingredients/substance part is complex in IDMP. The below proposal simplifies that part a bit as we do not have to include everything for a clinical study. However in this way it still can be mapped to IDMP/FHR. See also the notes below for the parts that I did not yet address.
For rules: note that the EMA IDMP implementation guide indicates that the StrengthLowerLimit should always be filled and the StrengthUpperLimit only when it concerns a ratio (to be verified).
Based on scrum discussions today I did the following follow-up research:
Note: page 37 of chapter 2 of the EU ISO IDMP IG indicates that "medicinal products may consist of two pharmaceutical products that correspond with two different administrable dose forms (e.g., hard capsule and cream) that form individual entities which do not need combining for administering to the patient. In this case the RMS list “Combination Package” shall be used. For clinical drugs this combination is tested and more likely seen as two different interventions. Therefore, it is probably more appropriate to use the term PharmaceuticalProduct or the corresponding FHIR name AdministrableProductDefinition instead. See https://build.fhir.org/administrableproductdefinition.html. THis class includes attributes like administrableDoseForm and routeOfAdministration. I will update the proposal accordingly tomorrow morning.
@dih-cdisc @EMuhlbradt @ASL-rmarshall @czwickl : Based on the above and other insights, please see below my full proposal for the intervention part of the model. I will go through it during the scrum today.
As discussed during scrum today:
@dih-cdisc @ASL-rmarshall @EMuhlbradt @czwickl : see first version of new UML below (not yet uploaded to Github). I have not yet moved pharmalogicalClass and productDesignation as I first like to dive into the fhir property types. Feel free to comment on names and referential integrity. To be continued on Monday.
@dih-cdisc @ASL-rmarshall : Interesting example from a Huntington Disease study. 1 intervention can include multiple tablets of different strengths. These would be different administrableProducts which implies that there should be a 0..* relationship from studyIntervention to AdministrableProductDefinition. Also, the same administrableProduct is reused across different interventions for the same study implying that for the API this should not be nested within the StudyIntervention class. See original table and my interpretation for the USDM below.
@EMuhlbradt @czwickl : See annotated UML below for you to work with.
@EMuhlbradt @czwickl : See annotated UML below for you to work with.
@BSnoeijerCD can you re-post in a different format so I can access it? I can't open png from Github for some reason.
@EMuhlbradt @czwickl Apparently the upload in the zenhub environment dose not work. Is this better?
@dih-cdisc @ASL-RMarshall @czwickl @EMuhlbradt I have gone through 7 different studies now and came to the conclusion that
We can discuss at the scrum tomorrow and hopefully decide to make these (small) changes.
@EMuhlbradt @czwickl Apparently the upload in the zenhub environment dose not work. Is this better?
Much better, thank you @BSnoeijerCD !
I made the following changes to the UML (and uploaded to Github) based on our discussions today:
@dih-cdisc @ASL-rmarshall @EMuhlbradt @czwickl Although IDMP covers manufacturer in the medicinal product definitions this might be slightly different then the provider/sourcing that is defined in most of the protocols I tested. The provider/sourcing is directly linked how it is done in the clinical study (e.g. who makes sure you get the medication). Therefore, I would not use IDMP manufacturer for this.
However, given the many instances of protocols that are including this information item, it would be beneficial to include sourcing/provider in the model. As this links to organizations and the corresponding structure, I propose to handle the sourcing/provider part of interventions in ticket #293. We need to consider then whether this would be on the level of AdministrableProductDefinition which would be beneficial if the intervention is the testing of interaction of multiple drugs or at the higher StudyIntervention level.
@BSnoeijerCD @dih-cdisc @czwickl please see the attached draft terminology for this ticket. Note that we have questions in the yellow highlighted cells. 316-444.xlsx
@dih-cdisc @EMuhlbradt @ASL-rmarshall @czwickl
Based on my experiences with real-world pharmacy data I should have known that multiple specifications of strengths (lower and upper limits) are possible for 1 ingredient. I missed that before in the EMA IDMP IG. See examples below from IDMP IG (chapter 2, Page 193):
This implies that the references strengthLowerLimit and strengthUpperLimit should be 1.. and 0.. respectively and names should be changed to strengthLowerLimits and strengthUperLimits Or maybe better: strengthsLowerLimit and strengthsUpperLimit as it refers to multiple strengths not multiple limits for the same strength.
@BSnoeijerCD @dih-cdisc @czwickl please see the attached draft terminology for this ticket. Note that we have questions in the yellow highlighted cells. 316-444.xlsx
Thank you @EMuhlbradt See my reactions in the last column. 316-444_BS.xlsx
@EMuhlbradt @dih-cdisc @czwickl AdminstrableProductProperty.types that I believe should be included:
Regarding the other optional types considered:
As discussed during scrum today:
@dih-cdisc @ASL-rmarshall @czwickl @EMuhlbradt Regarding the administrableProductDefinition as discussed last Thursday: In IDMP this is referred to as pharmaceuticalProduct which is then presented in the FHIR Resource AdministrableProductDefinition (EMA IDMP IG, Chapter 2, page 205).
The administrable pharmaceutical form is defined as follows (EMA IDMP IG, Chapter 2, page 204): The administrable pharmaceutical form refers to the pharmaceutical form for administration to the patient, after any necessary transformation of the "manufactured" pharmaceutical form has been carried out. The latter refers more to the dosing (after preparations have taken place) which we handle by the dose attribute in USDM. Also SNOMED makes a clear distinction between the pharmaceutical product and the administrable product.
Therefore, if we like to remove definition and stay close to IDMP, I would rename the class to PharmaceuticalProduct instead of AdministrableProductDefinition.
@dih-cdisc @ASL-rmarshall @czwickl @EMuhlbradt regarding pharmacologicClass see the CDISC page regarding the terminology to be used: https://www.cdisc.org/kb/known-issues/tsparm-pharmacological-class-terminology-change It looks medication - related so could link more to the product instead of the intervention which motivates that the attribute can move to the new class. Need to check some examples of this terminology on the NCI website.
@ASL-rmarshall @EMuhlbradt @czwickl @dih-cdisc See below updated annotated UML based on the discussions today. I applied the following changes:
As discussed during scrum today: I moved the reference attribute from the strength to the ingredient class:
@dih-cdisc @EMuhlbradt @czwickl @ASL-rmashall Please find attached the instance examples based on 9 randomly selected studies from clinicaltrials.gov.
Note that as administrableProducts are often repeated for different interventions, it would be good to nest them in the API at the StudyVersion level. Then they can also be reused across multiple designs with the same study drug.
@BSnoeijerCD @dih-cdisc I've been looking at the Ingredient FHIR resource. It seems that:
textPresentation
or textConcetration
)basis
attribute can be used to indicate whether the ingredient is being represented as the base or salt - which might be relevant if for reference strengths.If my understanding/interpretation is correct, this implies that:
Strength
and Quantity
(though it seems strange not to use the Range
class for this) instead of between Ingredient
and Strength
Quantity
)Quantity
)Strength
to Ingredient
. The linked ingredient would have it's own strength, which might need to be restricted to being a concentration strength, e.g. mass per mass.type
attribute/relationship for Strength
so that it's possible to differentiate presentation, concentration and reference strengths.basis
?@ASL-rmarshall : Note that I started with the IDMP and then linked to the corresponding FHIR resources which was the scope of this ticket. That probably results in the differences of the changes I defined for the model with what you proposed above. Did you check this in the IDMP IG? Furhtermore, I checked what is usually defined in a protocol/study definition - basis is not defined there. If needed we can always add later but I do not see this information available at the moment. The ingredient name can already indicate this for the greater part.
@dih-cdisc
@BSnoeijerCD I was originally looking at the modelling from the point of view of working out how it should be used and which rules would be required. I hadn't looked at the IDMP IG, but I have now. As discussed on 2024-08-16, the focus of the modelling is to accommodate what's needed to for a study definition, bearing in mind the EMA IDMP IG requirements and - if needed - with reference to FHIR resources.
I agree that basis is not required because it's not typically specified in a study definition. I also agree that a lot of the detailed FHIR modelling is not required.
However, I think there are some tweaks that could be applied to the modelling that would simplify things a bit, align with existing USDM modelling, make it easier to create and check for conformant data, and eventually meet the EMA IDMA requirements. The following diagram shows these tweaks:
These are the 4 suggested changes:
Make numerator range a property of the strength instead of having two independent strength specifications. In the diagram, I've shown this as:
numerator
relationship to the Range
class. I changed the cardinality for the original numerator
relationship from from 1 to 0..1 because numerator can be specified as either a quantity or a range. We'd need a rule to check that a numerator is specified (either as a quantity or a range). Note that I also modified the unit
relationship for Range
to point to AliasCode
instead of Code
to be consistent with the unit
relationship from Quantity
. An alternative would be to not use the Range
class and instead have numeratorLow
(1) and numeratorHigh
(0..1) relationships to Quantity
(similar to FHIR RatioRange), but that's less in line with existing USDM modelling.strengthsUpperLimit
and strengthsLowerLimit
relationships with a single strengths
relationship.This has the following advantages:
reference
relationship on Ingredient
to references
and change the target cardinality from 0..1 to 0..*. According to both EMA IDMP and FHIR, you can have more than one reference strength.role
attribute/relationship to Ingredient
- see FHIR Ingredient.role and associated value set. This is not 100% necessary, but it would facilitate:
type
attribute/relationship to Strength
to differentiate "presentation" and "concentration". This is not 100% necessary, but it would potentially facilitate:
@dih-cdisc
@BSnoeijerCD @dih-cdisc @czwickl see the attached file containing what we hope is the draft final new terms and changes to existing for this ticket. @BSnoeijerCD there is one outstanding question in row 20 in the 'new' tab. 316-444_Draft final.xlsx
@dih-cdisc @EMuhlbradt @czwickl Erins question is about reference as well and relates to what Richard indicates above in point 2. The related citations from the IDMP IG are below: "Reference strength represents the strength (quantitative composition) of the active moiety of the active substance or of another substance used to express the strength of the product. There are situations when the active substance and active moiety are different resulting in different expression of the strength. For example, if an active substance is in the form of a salt or hydrate, the reference strength shall be expressed in terms of the mass [or biological activity in International (or other) units where appropriate] of the active moiety (base, acid or anhydrous material). The reference substance and reference strength of the active substance(s) contained in the Pharmaceutical product or Manufactured Item can be found in section 2. Qualitative and Quantitative Composition of the corresponding SmPC. The reference strength shall be provided for active substances and is repeatable. However, there are cases where the reference strength of the active substance may not be provided. In cases where the quantitative composition of the excipient(s) and/or adjuvant(s) of the medicinal product is not reported in the relevant product information (i.e., SmPC), the substance strength is to be left empty and the substance reference strength is to be completed."
See also examples in Chapter 8 of IDMPIG - starting on page 38
This indicates that indeed there can be more than 1 reference strength. Given what I put in bold above, I am not sure whether it should be per se the active (component of the) ingredient.
@dih-cdisc @EMuhlbradt @czwickl
Based on the discussions today. See the updated UML below with the two outstanding questions indicated as well.
@dih-cdisc @EMuhlbradt @czwickl I updated the list with instances and added a number of examples from IDMP Chapter 8. See attached. InstancesExamples_v2.xlsx
Note: that there was only 1 ingredient in the examples from IDMP (and none found in the selected protocols) that had a range for strength and this was an excipient and not an active ingredient. Given that and for a better understanding of the model when not IDMP related, I am now in favor of the suggested change from strengthsUpperLimit and strengthsLowerLimit relationships to 1 strengths relationship with the option to define a minValue and maxValue (using range class) for the numerator instead of value (using quantity class). The first row of this file indicates the paths according to the model version in the comment above. The second row in this files indicates the paths when this is changed.
Note: IDMP requires a reference strength for the active substance while the original strength might be missing. This information is missing in vaccine examples in the IDMP.
@dih-cdisc @EMuhlbradt @czwickl
Please see the last updated version of the UML below (also uploaded to github). The last changes are based on our discussions earlier today. I indicated all intervention changes in yellow and the new changes in orange.
@EMuhlbradt @czwickl: This is the link for the ingredient role: https://www.hl7.org/fhir/valueset-ingredient-role.html
@dih-cdisc @BSnoeijerCD Final changes are ready for QC in the Github CT deliverables folder.
Thank you @EMuhlbradt @czwickl
Based on my review I have the following comments:
Thank you @EMuhlbradt @czwickl
Based on my review I have the following comments:
- AgentAdministration should be changed to Administration - The reason for this is that it can be used more broadly for other types of interventions that can be specified as well like: behavioral therapy, diagnostic test, dietary supplement, radiation, medical procedure etc.
- Since we moved it, should the definition of pharmacologic class be changes to "The pharmacological class of the administrable product"?
- With previous links to SDTM codelists you specified that in the "Has value List" column like "Y (SDTM Terminology Codelist ... )". Should the SDTM be specified for administrableDoseForm?
- referenceSubtance: do we want to be more explicit regarding what is expected here? I put it like this in the IG: "Each active ingredient specified by its substance may have a reference substance, indicating the reference strength once the product is processed in vivo."
@BSnoeijerCD
@BSnoeijerCD @dih-cdisc @ASL-rmarshall I have updated the CT deliverables in 3-6-0 release to update the semantics (C-code, preferred term, and definition) for both 'pharmacologicClass' and 'administrableDoseForm' to reflect the AdministrableProduct class.
Formulation missing from intervention + alignment with IDMP
See JIRA 520