A number of diseases exist in which the literature describes the potential for there to exist a recognised set of associated "systemic manifestations". In an attempt to model these systemic manifestations there has been a requirement to identify a modelling pattern which lies somewhere between the concepts of "Co-occurrent" and "Due to".
The issue arose from a discussion on how we should model the systemic manifestation of Sjögren's syndrome and the thread relating to this has been included below as background.
-------------------------------
Potential issues with the use of "Co-occurrence"
Dictionary definitions of 'Co-occurrent' include:
: occurring together or at the same time - "two disorders that are frequently co-occurrent" - Merriam Webster
: the fact of two things happening at the same time - "co-occurrence - the temporal property of two things happening at the same time; "the interval determining the coincidence gate is adjustable" - TheFreeDictionary
The slides suggest the following definition within the context of SNOMED Modelling:
"Use simple co-occurrence for 2 or more conditions without a known causal or temporal relationship associated by a common predisposition"
There are potentially a number of issues with this definition:
- Clearly there is a temporal relationship as we expect the two conditions to be present at the same time
- We are stating that there is no current evidence of a causal relationship BUT
- They are associated by a "common predisposition" - how do we differentiate between a "predisposition" and a "causal relationship"?
In achieving a suitable solution to these issues there are at least 2 use cases which need addressing:
- The ability to record within the notes the patients current "disease state"
- The ability to ensure that the modelling within SNOMED CT supports appropriate classification inferences.
Use case 1 requires us to provide sufficient clinical concepts to document both the underlying disorder and its 'manifestations'. For syndromes this will be the name of the syndrome in which the necessary and sufficient manifestations are acknowledged as being present and required for the patient to be said to have that syndrome. In addition, the presence of the genetic mutation which led to the syndrome may predispose that patient to have additional manifestations, each of which must also be recorded separately to gain the full picture of the expression of the disorder in that patient.
Use case 2 requires modelling of both the necessary and sufficient manifestations, currently this modelling does not include YYY (disorder) |Due to| |Deletion of part of XXX chromosome (disorder)| (what is the reason for not adopting |due to|?). However, if we wish to record a manifestation which is predisposed by the presence of the gene abnormality, but is not necessary to classify this as YYY syndrome then I assume we would have to consider the |Due to| modelling pattern and use a description something like "ZZZ with YYY syndrome due to deletion of XXX chromosome".
So, what I am trying to say here is that for syndromes there is a direct causal relationship between the chromosome abnormality and the manifestations, however, some manifestations are necessary in order for it to be "labeled as a syndrome", while there exists a predisposition to other manifestations some of which will be expressed and others that are not. Where we need to record the existence of the additional manifestation I think the use of co-occurrent, if defined as above, does not adequately represent our understanding of the association.
We could also consider diabetes and it's 'complications. I believe the current understanding of the complications is that abnormal blood glucose regulation leads to pathological changes to the microvasculature which in turn leads to damage to the end organs, retina, kidney, nervous system and peripheral vascular system etc. Therefore, having a diagnosis of diabetes mellitus predisposes one to all of the 'complications' as a direct result of the pathological process. However, while the individual may be predisposed to all of these manifestations, a combination of good blood glucose control and lifestyle may result in only some of the complications being manifested. We choose to use |Due to| in modelling in this instance - is there a clear and reproducible difference between diabetes and genetic syndromes which supports us taking a different modelling approach?
And so to Sjögren's syndrome - There is a considerable amount of literature about Sjögren's syndrome both in the academic press, ClinicalKey (Elsevier) and UpToDate, in addition there has been a small number of large cohort studies which have looked specifically at the extra-glandular manifestations.
In essence and historically Sjögren's syndrome has been divided into "primary" and "secondary" where Primary Sjögren's syndrome is a systemic auto-immune disorder characterised by the presence of dry eyes (keratoconjunctivitis sicca) and dry mouth (xerostomia) as a consequence of lymphocytic infiltration into the lacrimal and salivary glands. (https://bestpractice.bmj.com/topics/en-gb/175#referencePop1 ) and Secondary Sjögren's syndrome in which Sjögren's syndrome is present along with another autoimmune disease such as lupus, rheumatoid arthritis and systemic sclerosis.
The presence of both Sjögren's syndrome and one of the other autoimmune diseases is a recognised entity, however, as yet there is no clear evidence that the underlying causal factors of each of the disorders is the same except that they are both due to an autoimmune process, however, they do not necessarily share the same autoimmune process. It is suggested that a combination of Sjögren's syndrome with another autoimmune syndrome may, for some clinicians be useful to record as a single statement, although my personal preference is that they should be recorded separately - (indeed I couldn't find an instance of any concept which include Sjögren's syndrome plus another autoimmune disorder in SNOMED CT!)
If we accept "Co-occurrence" as a useful concept then I would suggest an alternative definition for consideration:
"Co-occurrence" is two or more disorders which are present at the same time in anyone individual, but in which there is currently no evidence that there is a causal relationship between the disorders.
Returning now to Primary Sjögren's syndrome. There is significant literature on the systemic extra-glandular manifestations of Sjögren's syndrome in the absence of a relationship with any other well defined autoimmune disorder with a prevalences of between 10% and 70%. See table below and http://www.medsci.org/v14p0191.pdf
Under normal circumstances, and for many other disorder I think we would call these complications and model them as |Due to|. However, I would accept that the pathogenesis is not well understood and therefore the association between Sjögren's syndrome and its associated systemic manifestations is not as strong as it is for diabetes mellitus and its complications.
What we require is a set of "associations' which can be used by authors consistently and reproducibly with clear criteria for their definitions. I think we have this for |Due to| and |causative agent|. However, I do not think this is the case for "Co-occurrent" and/or a 4th category in which an association is clearly evidenced but for which, at this moment in time we are unable to evidence the cause.
I would like to suggest that |Associated with| may be a suitable candidate and would suggest the following definitions:
"Co-occurrence" - two or more disorders which are present at the same time in anyone individual, but in which there is currently no evidence that there is a causal relationship between the disorders.
"Associated with" - The presence of systemic manifestations of the primary disorder confirmed by evidence from clinical studies but at this moment in time we are unable to evidence the causal chain.
(I am sure we can come up with more succinct definitions but hopefully the above provides a basis for discussion)
I think the use of 47429007 |Associated with (attribute)| might be appropriate as later clinical studies and evidence may lead to a final decision that the association is either, 42752001 |Due to (attribute)|, 246075003 |Causative agent (attribute)| or |Co-occurrent (attribute)|.
In summary, I think I am saying:
- We need much better clarity on the distinction between the various forms of "association" to enable consistent and reproducible authoring
- I think there is an association which sits between |Due to| and my definition of |Co-occurrent| where there is clear evidence for the association without a causal relationship having been fully determined
- I think the systemic manifestations of Sjögren's syndrome fit into the category which sits between my definition of |Co-occurrent| and |Due to|
- The association between Sjögren's syndrome and Lupus, SLE RA etc. is currently at best |Co-occurrent|
5 Comments
Paul Amos
Bruce Goldberg wrote:
I would like to address some of your concerns.
1. The definition of co-occurrence – I agree that the definition should not exclude temporal relationships. I believe what I was trying to express was to exclude the after relationship. I interpret co-occurrence for the purposes of defining syndromes as 2 or more conditions that occur together at least at some point during the course of the syndrome.
2. Modeling of genetic syndromes. Quite a while ago when we talked about this in ECE (when Ed was chair) we decided that it would be too difficult to add mutations to the model but I have been thinking about this recently and I see no downside to modeling genetic syndromes with not only the physical manifestations but with the causative mutations and resulting protein abnormalities. I think I mentioned in the last ECE call, Jim Campbell has been doing some work in this area. I will try to pull up some of his proposed models and invite him on a future call.
3. Predisposition – I was interpreting this as some factor (often genetic) that causes 2 or more conditions to occur more frequently than by random chance.
4. Associated with – I am not in favor of once again broadening the usage of this attribute. We moved away from it because of the difficulty in maintaining consistency. The degree of causation is often interpreted subjectively and has lead to similar concepts modeled using either associated with or due to. I agree that just using simple co-occurrence may not fully express the nature of the association between a systemic disorder and one of it’s manifestations I think it would achieve consistency.
Example (oversimplified) Xerophthalmia due to Sjogren syndrome vs. Myalgia associated with Sjogren syndrome
Another possible definition for simple co-occurrence:
"Two or more conditions that are found together more often than would be expected by random chance"
Paul Amos
Jim Case wrote:
First of all, thanks Paul for your detailed exposition on the issue. It sheds light on the difficulties we have when we know little about the mechanisms of a systemic disease. Trying to focus on Sjögren's, let's look at the current representation in the literature. The fact that these clinical signs are known as "systemic manifestations of Sjögren's" leads one to believe there is an associated underlying process, as Bruce has so well described in his response. What we need in these instances is something more specific than ASSOCIATED WITH, which is very vague on what is meant and why we do not really want to use it anymore and something more specific than "co-occurrent" because of the notion that there is (probably) not a causal relationship between the two disorders. I think what Bruce has been trying to get to is a different beast, i.e. instead of X causing Y, we have a Z causing both X and Y. Z being the predisposing disorder (be it genetic or something else). I think this is different than something such as "Secondary Sjögren's". We currently have a relationship type that we have not really defined well and may be useful to resolve this situation.
Let's consider:
1. For true co-occurrence (no known or suspected causal relationship) we use multiple IS A relationships
2. For two disorders for which there is or may be a common underlying predisposition, we use the TEMPORALLY RELATED TO relationship and define that relationship as the co-occurrent presence of two disorders due to a known or suspected underlying predispostion.
Because we have already placed TEMPORALLY RELATED TO as a parent to BEFORE, DURING, AFTER, we need to consider whether the proposed definition would also be consistent with the subtypes. As I think about the use of these subtype relationships, I could be convinced that in all (most?) cases, there is a common underlying predisposition.
Anyway, curious as to your thoughts on this. I would like to avoid ASSOCIATED WITH if at all possible, but agree with Paul that we need something in between co-occurrent and DUE TO.
Paul Amos
Bruce Goldberg wrote:
Using temporally related to, to represent an association between 2 conditions in which one condition may precede, follow or occur during the course of the others seems like a sensible approach but it still does not address the underlying predisposing cause. In that case perhaps including a due to relationship to 55446002 |Genetic mutation (finding)|, a specific genetic mutation or unknown predisposing factor (as a new concept) in addition to temporally related to would be more robust. Another option would be to have a new attribute, has predisposition with same values as above.
Getting back to temporally related to, this would be an ideal way to represent the association of the different components of a syndrome. Modeling these as simple co-occurrence assumes that all of the components are present and we have made a compromise by including the most common conditions as always present in the definition. I am now wondering about including multiple temporally related to relationships to define multi-component syndromes.
X,Y,Z syndrome
isA X
temporally related to Y
temporally related to Z
----------------------------
After playing around with modeling using temporally related, the issue appears to be that this role is symmetric (as is "Associated with":
X after Y is not the same as Y after X
X after Y is the same as Y before X
X temporally related to Y is the same as Y temporally related to X. The problem is that:
X temporally related to Y isA X and Y temporally related to X isA Y.
This is probably why this attribute has only been used to define perioperative complications for which a disorder is temporally related to a procedure.
I think we either need a new attribute of has predisposition or we just use associated with as a bad compromise.
Paul Amos
I am on annual leave this week and so I am posting my initial thoughts here as I will not be at tonights ECE meeting.
I would agree with Bruce that use of "temporally related to" does not address the issue of the underlying predisposing cause. While I accept that the "manifestation" my occur before, during or after the "main event" the most common temporal relationships are likely to be during or during and after.
There is also a potential issue with using HAS PREDISPOSITION. Given the need to provide an FSN which accurately supports the modelling we would probably need to include the word "predisposition" within the description as we do for DUE TO. This leaves us with a problem as the clinician does NOT want to record that the patient has a predisposition to suffer a specific manifestation but rather the patient has YYY finding or disorder(the manifestation) as a realised predisposition of XXX disorder.
I think the syndromes which have a clear genetic mutation should be modelled with a DUE TO the identified mutation as part of the definition of the syndrome and where necessary to add the additional features which are due to the genetic mutation but do not form a necessary part of the defined syndrome.
The question I would like to raise is whether we feel that we should be restrained from using ASSOCIATED WITH simply because of our historical relationship with its use? Surely, if we were starting with a blank sheet we would be happy to use CO-OCCURRENT, ASSOCIATED WITH and DUE TO with the definitions of:
CO-OCCURRENT - For true co-occurrence (no known or suspected causal relationship) we use multiple IS A relationships
ASSOCIATED WITH - Co-occurrent presence of two disorders due to a suspected underlying predisposition/cause. Use the ASSOCIATED WITH attribute but no IS_A to the primary disorder or cause
DUE TO - Co-occurrent presence of two disorders in which there is clear documented evidence of a causal relationship - Use the DUE TO attribute but no IS_A to the primary disorder or cause.
By taking this approach we should be able to clearly distinguish between ASSOCIATED WITH and DUE TO consistently. Also as with so called diabetic complications we are recording the complication as a separate concept from the underlying cause. So we record someone has diabetes and we separately record that the patient has a diabetic cataract rather than a senile cataract. Recording a systemic manifestation of Sjögren's syndrome as "Arthralgia associated with Sjögren's syndrome" is simply recording the manifestation.
This would require us to review the current concepts which have within any of their descriptions "associated with" for appropriate and consistent usage but I think that is required irrespective of whether we adopt my proposed solution or some other solution.
Bruce Goldberg
To summarize the issues as I see it, we want to distinguish between two types of disorder associations, one in which a finding or disorder is a manifestation of a multi system disorder for which the manifestation and the underlying disorder may share a common pathological process and morphology and the other in which there are 2 discrete disorders that are found together more than would be expected by chance alone and may be related by a common predisposition. I think you are suggesting to use associated with for the former and simple co-occurrence for the latter. Jim seems to be suggesting to use simple co-occurrence for the former and temporally related to for the latter. In a separate e-mail, I expressed my concern about using temporally related to which is essentially a symmetric property to relate two conditions where neither one is secondary to the other. I would like to propose the following solution:
For the situation where a disorder or finding is a manifestation of a multisystem disorder, create a new attribute, is manifestation of (note there is currently a link assertion “417318003 |Is manifestation of (link assertion)|”.
Another option would be to use as Jim suggested temporally related to where the manifestation is temporally related to the underlying systemic disorder. A manifestation may precede and/or occur during the underlying disorder but would never as far I know occur after the underlying disorder has resolved.
For two distinct disorders that may share a common predisposition, represent as simple co-occurrence with a due to relationship to the predisposing factor which may be one or more findings or disorders such as a mutation, substance, organism, physical force or event or some combination of the above. If the predisposing factor is unknown, then a due to relationship would not be included.