Stream: genomics
Topic: Associated Phenotype Nomenclature
Rachel Kutner (Mar 25 2020 at 21:30):
We have been discussing our implementation of the HL7v2 Clinical Genomics implementation spec with some clinicians, and received push back on using "Associated Phenotype" to describe a disease associated with a variant that may or may not be observed on the patient. They suggested using "Disease Association" instead of "Associated Phenotype" to try and distinguish between those that are observed on the patient and those that are just associated with the result.
- Is this the correct interpretation for that field in the HL7 specification?
- Is this something anyone else has heard or would support?
The concern is in confusing phenotypes actually observed on a patient vs. ones that are reported as being associated with a genomic result on the patient.
Kevin Power (Mar 26 2020 at 16:43):
The associated phenotypes is still defined as something associated to the genetic finding and not direct patient observations. So perhaps this is just something we haven't defined very well (likely in both the V2 and FHIR IG work we have done recently)?
FWIW - This is the LOINC code from the FHIR IG (and I think it is the same one in the V2?)
https://loinc.org/81259-4/
"The possible phenotype associated with the genetic variant found in this study."
Having said that, I would be supportive of having this be an official request that we consider. Anything we can to to make it easier for people to see and digest is good. Welcome input from the community on this one.
Liz Amos (Mar 26 2020 at 21:35):
I looked up what the description was in V2: "The disorder with which this variant is associated." (It was also named Probably Associated Phenotype). Perhaps that language is more clear than what is currently in the FHIR IG for General Genomic Reporting: "The 'Inherited Disease' Implication indicates the likelihood of inheritance (valueCodeableConcept) of a particular disease (the associated-phenotype) as well as how inheritance is likely to occur (mode-of-inheritance)." http://build.fhir.org/ig/HL7/genomics-reporting/general.html#genomic-implications
Kevin Power (Mar 26 2020 at 21:48):
That part of the FHIR IG will be completely redone with the implication simplification by @Jamie Jones , so if some better wording would help here, happy to take suggestions.
Liz Amos (Mar 30 2020 at 19:10):
In a related thread, I was checking out the new/proposed implication profiles and was curious why we have Associated Phenotype and Associated Cancer? This is in Somatic Reporting. Could we not just use Associated Phenotype or is there a reason Associated Cancer has it's own component?
Jamie Jones (Mar 30 2020 at 20:00):
That's a great question! We try to not give guidance to shoehorn things into LOINC concepts that are well-defined if the other concepts are out of scope of that definition. Do you feel we could use associated-phenotype instead? The concepts are treated the same here but seem to diverge in what I've seen :upside_down:.
I am hoping we could properly define a single concept that covers both uses
Patrick Werner (Mar 30 2020 at 20:02):
what are the expected coded values of these components?
Jamie Jones (Mar 30 2020 at 20:03):
Our friends (unbound) and (example) currently
Patrick Werner (Mar 30 2020 at 20:04):
i know :slight_smile:
Jamie Jones (Mar 30 2020 at 20:04):
HPO is an option
Patrick Werner (Mar 30 2020 at 20:04):
I was talking about what concepts we are expection in the value of these two components. I think this has a huge influence on wether we can merge them into one component or not
Patrick Werner (Mar 30 2020 at 20:07):
hmm HPO would cover both well, i guess. I think that oncologists would link to icd10 codes in associated cancer.
Jamie Jones (Mar 30 2020 at 20:09):
Phenopackets uses
Include codes from http://purl.obolibrary.org/obo/hp.owl where concept descends from HP:0000118 (Phenotypic abnormality)
Include codes from http://snomed.info/sct where concept descends from 404684003 (Clinical finding)
(link) https://aehrc.github.io/fhir-phenopackets-ig/ValueSet-PhenotypicFeatureType.html
Patrick Werner (Mar 30 2020 at 20:11):
Ok, ICD-10 then would fit in as well.
Patrick Werner (Mar 30 2020 at 20:12):
So we could merge them together i guess. example Binding to HPO, and some guidance on what other CSsystems could be used here.
Jamie Jones (Mar 30 2020 at 20:13):
I like it. Others?
Patrick Werner (Mar 30 2020 at 20:14):
(and i hate these obo purl links as canonicals :smirk: as they are resolving to the owl file, not to some human consumable explanations)
Jamie Jones (Mar 30 2020 at 20:17):
Perhaps we could borrow phenopacket naming and call it 'associated-phenotypic-feature-type'
Jamie Jones (Mar 30 2020 at 20:20):
Does the same merger work for Therapeutic implication?
Jamie Jones (Mar 30 2020 at 20:21):
What are the downsides of combining these concepts, how does it affect search functionality? (user would have to be able to resolve that ontology)
Liz Amos (Mar 30 2020 at 21:19):
Jamie Jones said:
Perhaps we could borrow phenopacket naming and call it 'associated-phenotypic-feature-type'
I'm with you in spirit but I'm a bit wary of using the exact phrasing. The way I read the phenopacket spec is that PhenotypicFeature describes the characteristics associated with the subject/patient. This is kind of similar to what @Rachel Kutner was describing in her question. I think adding the word "Associated" helps but maybe we can define it even more.
Kevin Power (Mar 30 2020 at 23:25):
So, two questions here:
1 - Do we combine 'associated-phenotype' and 'associated-cancer' components?
2 - If we have one or two, what name(s) do we use?
My thoughts:
-
This may be naive, but honestly I am less concerned about the name we use than I am about clearly defining what we have and how it should be used. Having said that, a clear, simple name is powerful.
-
I do think deciding if one or two components is the best place to start. The more I think about it, I tend to lend towards having two. I think keeping a clear distinction between the sort of data being delivered is important (even if there is some potential overlap in code systems). I also think @Jamie Jones asked a good question - if we combine them, can consumers easily query by just what they are looking for, and then easily understand the difference when they get a response? Thoughts from @Bob Dolin @Alexander Mankovich @Bret H @Rachel Kutner @Larry Babb @Mullai Murugan ?
Before we talk about the final definition/usage guidance/name, I will stop there and see what others think of one or two.
Bret H (Mar 31 2020 at 00:15):
Perhaps change associated phenotype to 'risk of' then cancer has a
different meaning when in one bucket of the other.
Bret H (Mar 31 2020 at 12:29):
sorry not able to be on zulip much at the moment. I suggest to change the name of associated phenotype to be specific - but clinicians will never see it as a user of the data. The names are/should only be visible in the backend. However, 'risk-of' might be clearer. Regards Cancer component and codes , there's ICD-O-3 and SNOMEDCT to consider. Locking down to a code set will be tricky, but no more tricky than systems already have to deal with when it comes to terminology. @Rachel Kutner could you perhaps use a different display in your UI? can your system accommodate that? Just trying to get at where the push back is coming from...
Patrick Werner (Mar 31 2020 at 12:48):
@Bret H this was my proposal:
Patrick Werner said:
So we could merge them together i guess. example Binding to HPO, and some guidance on what other CSsystems could be used here.
No need to bind it harder than preferred i guess
Bret H (Mar 31 2020 at 12:57):
@Patrick Werner got it.
Bret H (Mar 31 2020 at 12:58):
Right now the semantics of two elements gives us the opportunity to explicitly state that the observation components are specific to a cancer. We could switch to a flag, I guess. If we use one field then one would need to know ahead of time the code system and codes of interest for oncology...
Alexander Mankovich (Mar 31 2020 at 13:11):
at first glance this seems like an ok approach..
Kevin Power (Mar 31 2020 at 13:13):
@Alexander Mankovich - Which approach seems OK? Combing them into a single field?
Alexander Mankovich (Mar 31 2020 at 13:14):
@Kevin Power combining them
Bret H (Mar 31 2020 at 13:14):
@Kevin Power you did not give him a second option
Bret H (Mar 31 2020 at 13:14):
: ^ )
Bret H (Mar 31 2020 at 13:15):
what would the field name be?
Jamie Jones (Mar 31 2020 at 13:15):
i'm more worried about the definition than the name
Bret H (Mar 31 2020 at 13:16):
and how do you know if a medication implication is referring to a risk of a specific adverse event or that the medication has reduced efficacy in treatment of the disease ?
Bret H (Mar 31 2020 at 13:16):
for theraputic implication
Bret H (Mar 31 2020 at 13:16):
Is the disease one being treated or one that is at risk for?
Bret H (Mar 31 2020 at 13:18):
in the diagnosis implication the definition is it pretty clear from context, right? but for the therapeutic implication we could either be giving information about a disease that is being treated or a condition that there is a risk of.
Jamie Jones (Mar 31 2020 at 13:19):
i was pushing to call them 'contextual-phenotype' and/or 'contextual-cancer' for therapeutic. Different context changes the concept.
Kevin Power (Mar 31 2020 at 13:20):
I was just typing the same thing @Jamie Jones -- I think 'context' is the important aspect. Not sure that answers @Bret H's question about 'treatment-of' or 'risk-of' question however.
Bret H (Mar 31 2020 at 13:20):
also, you can see that there has been a history of explicitly calling out oncology related genetic findings. Jeremy Warner was a large proponent. Shoot an email to the list-serv to get more input on dispensing with this.
Jamie Jones (Mar 31 2020 at 13:21):
We didn't land on a representation of adverse event phenotypes
Kevin Power (Mar 31 2020 at 13:21):
Yea, that is related, but perhaps we should hold off on "adverse event phenotype" during the simplification phase.
Bret H (Mar 31 2020 at 13:22):
@Kevin Power that would be bad
Kevin Power (Mar 31 2020 at 13:24):
@Bret H - we have (tried) to keep the scope of the simplification to what we can represent today, and logging JIRAs to track new concepts. So I am not saying we shouldn't do it, I am suggesting we log a new JIRA (or find the one where this was requested) and work on it from there.
Jamie Jones (Mar 31 2020 at 13:25):
i logged it before, it's linked in monday's notes
Kevin Power (Mar 31 2020 at 13:26):
ah ha... FHIR-26380 describe adverse effects on Implication
Currently have ‘associated-phenotype’ only
Kevin Power (Mar 31 2020 at 13:28):
I suppose given that mantra, perhaps for the 'simplification' work, we keep our two components, leave them named as they are, and log a new JIRA to consider everything we have talked about here?
Bret H (Mar 31 2020 at 13:28):
just taking a look at the implications of losing the Cancer specific call out. Losing the specific oncology call-out would mean that to find oncology related genomics one would 1) start with the tumor specimen and look for all genetic reports where the tumor is the specimen (not so bad conceptually), or 2) need to know all the codesystems and codes that one would need to search in the 'field relating connection with a disease or symptom and the genetic finding'
Kevin Power (Mar 31 2020 at 13:29):
I don't like putting stuff off, but we have to find a way to cut this off and move forward.
Alexander Mankovich (Mar 31 2020 at 13:36):
@Bret H to your point im not sure how many uses there are for collecting ALL the oncology related findings. usually I think one would be looking for specific cancer types, requiring some knowledge of the code systems anyway
Bret H (Mar 31 2020 at 13:37):
yeah, that would be true even with an explicit cancer field. Right? (one would need the code/codesystem ahead of time) seems like working from the tumor would be the better path for querying ?
Alexander Mankovich (Mar 31 2020 at 13:39):
I'd think so. another thought: i do wonder if by combining the components we might be losing out on distinctions between phenotypes and the diseases they may indicate especially in retrospective studies
Bret H (Mar 31 2020 at 13:50):
can you say more?
Rachel Kutner (Mar 31 2020 at 13:55):
Bret H said:
sorry not able to be on zulip much at the moment. I suggest to change the name of associated phenotype to be specific - but clinicians will never see it as a user of the data. The names are/should only be visible in the backend. However, 'risk-of' might be clearer. Regards Cancer component and codes , there's ICD-O-3 and SNOMEDCT to consider. Locking down to a code set will be tricky, but no more tricky than systems already have to deal with when it comes to terminology. Rachel Kutner could you perhaps use a different display in your UI? can your system accommodate that? Just trying to get at where the push back is coming from...
@Bret H Yes, we definitely can change the display name for this field. I wanted to make sure I interpreted the field correctly and see if there is a clinical consensus for a term that can be used with less confusion for clinicians on the front end.
I was hoping this work group would be able to suggest some possible alternatives which we could explore with clinicians to find a clearer, more specific alternative. Additionally, if a field is introduced in the future that is supposed to encompass phenotypes that have been reported on a patient and is associated with a result or finding, I'd like to make sure we don't choose a term that could be confusing between the two.
Bret H (Mar 31 2020 at 13:57):
Thanks @Rachel Kutner ! phenopackets was mentioned earlier they have terms like 'Phenotypic feature' might have less baggage.
Kevin Power (Mar 31 2020 at 14:05):
@Rachel Kutner To be clear, your interpretation of this field is correct, so that is good.
The rest of this thread has been a lot of brain storming about some alternatives we could/should consider. Any thoughts on the ideas being discussed here?
Bret H (Apr 05 2020 at 17:37):
back to brainstorming: single filed for a phenotype linked either through causation or risk link, is consistent with VICC. I think the crux is defining the relationship of the condition/disease with the variant and the rest of the statement being made with the profile. for example, with the theraputic implication is the variant 'known to cause' or 'related to risk of'
Rachel Kutner (Apr 27 2020 at 15:50):
Is Associated Cancer a diagnostic field for somatic samples, or is it a field that is meeting the same concept as Associated Phenotype to specify a germline risk for the specified cancers?
If the purpose is to provide a diagnostic field for somatic samples, that sounds like a clearer case for differentiation since one indicates confident diagnostic assertion of a phenotype, the other indicates risk of a phenotype that the patient may or may not already exhibit.
Kevin Power (Apr 27 2020 at 16:36):
@Rachel Kutner - We talked about this (sort of) in the FHIR subgroup today. The short answer is that we need to figure out an appropriate way to distinguish between 'risk of' and 'context of' when it comes to imlications. @Bret H logged the following tracker - J#26945 which I think gets to the heart of your question?
Last updated: Apr 12 2022 at 19:14 UTC
