Cytokine and chemokine profiles in fibromyalgia, rheumatoid arthritis & systemic lupus erythematosus

Bob

Senior Member
Messages
16,455
Likes
34,051
Location
England (south coast)
Cytokine and chemokine profiles in fibromyalgia, rheumatoid arthritis and systemic lupus erythematosus: a potentially useful tool in differential diagnosis
Wallace DJ, Gavin IM, Karpenko O, Barkhordar F, Gillis BS.
07 Nov 2014
Rheumatol Int. 2014 Nov 7. [Epub ahead of print]
http://link.springer.com/article/10.1007/s00296-014-3172-2

Open Access. Full article available here:
http://link.springer.com/content/pdf/10.1007/s00296-014-3172-2.pdf

Abstract
Making a correct diagnosis is pivotal in the practice of clinical rheumatology. Occasionally, the consultation fails to provide desired clarity in making labeling an individual as having fibromyalgia (FM), systemic lupus erythematosus (SLE) or rheumatoid arthritis (RA). A chemokine and cytokine multiplex assay was developed and tested with the goal of improving and achieving an accurate differential diagnosis. 160 patients with FM, 98 with RA and 100 with SLE fulfilling accepted criteria were recruited and compared to 119 controls. Supernatant cytokine concentrations for IL-6, IL-8,MIP-1 alpha and MIP-1 beta were determined using the Luminex multiplex immunoassay bead array technology after mitogenic stimulation of cultured peripheral blood mononuclear cells. Each patient’s profile was scored using a logistical regression model to achieve statistically determined weighting for each chemokine and cytokine. Among the 477 patients evaluated, the mean scores for FM (1.7 ± 1.2; 1.52–1.89), controls (−3.56 ± 5.7; −4.59 to −2.54), RA (−0.68 ± 2.26; −1.12 to −0.23) and SLE (−1.45 ± 3.34, −2.1 to −0.79). Ninety-three percent with FM scored positive compared to only 11 % of healthy controls, 69 % RA or 71 % SLE patients had negative scores. The sensitivity, specificity, positive predictive and negative predictive value for having FM compared to controls was 93, 89, 92 and 91 %, respectively (p < 2.2 × 10−16). Evaluating cytokine and chemokine profiles in stimulated cells reveals patterns that are uniquely present in patients with FM. This assay can be a useful tool in assisting clinicians in differentiating systemic inflammatory autoimmune processes from FM and its related syndromes and healthy individuals.
 

Bob

Senior Member
Messages
16,455
Likes
34,051
Location
England (south coast)
I spotted this article on ProHealth, along with the following comment...
ProHealth Editor said:
Editor's comment: Investigators in this study were also key researchers in the development of the FM/a® diagnostic blood test for fibromyalgia. This study is likely further verification that the test is able to differentiate between FM, RA and Lupus.
 
Messages
5,256
Likes
32,011
Cytokine and chemokine profiles in fibromyalgia, rheumatoid arthritis and systemic lupus erythematosus: a potentially useful tool in differential diagnosis
Wallace DJ, Gavin IM, Karpenko O, Barkhordar F, Gillis BS.
07 Nov 2014
Rheumatol Int. 2014 Nov 7. [Epub ahead of print]
http://link.springer.com/article/10.1007/s00296-014-3172-2

Open Access. Full article available here:
http://link.springer.com/content/pdf/10.1007/s00296-014-3172-2.pdf

This seems a weird way to present data doesn't it Bob? I am beginning to wonder what is happening with journals but I would have assumed I would have had this rejected if I had sent it in myself. We are not told what the basis of the 'logical regression model' was but it looks as if it was designed to give a positive result for FM. So a positive result for FM is not surprising. Maybe I'm getting old but this looks more like another advert than a scientific paper.

And I know I moan a lot but we have had some really good stuff flagged up in the last week or so too!
 

user9876

Senior Member
Messages
4,556
Likes
18,067
This seems a weird way to present data doesn't it Bob? I am beginning to wonder what is happening with journals but I would have assumed I would have had this rejected if I had sent it in myself. We are not told what the basis of the 'logical regression model' was but it looks as if it was designed to give a positive result for FM. So a positive result for FM is not surprising. Maybe I'm getting old but this looks more like another advert than a scientific paper.

And I know I moan a lot but we have had some really good stuff flagged up in the last week or so too!
They use a logistic regression model which I have a vague memory as being similar in form to Minsky's perceptron work. Its not clear in the paper what they are fitting but my guess is that they are basically using it as a classifier although its very unclear what the classes are they are training against. But my guess is it is some variable that indicates a diagnosis of FM.

The fitting process gives the set of weights that are the best that can be found to discriminate between the two classes of data given a linear combination of weights and input vector which is the passed through a logistic function which acts as a kind of threshold function.

Its not surprising that they get the graph in fig 2 showing separation of data if that is what they have fitted their regression model to. It is valid to do that but they should quote results against a set of test data that has not been used in generating their regression model. At that point it would become interesting but then its a case of trying to understand the weights and what they mean in terms of the data and the biology of the situation.
 
Messages
5,256
Likes
32,011
They use a logistic regression model which I have a vague memory as being similar in form to Minsky's perceptron work. Its not clear in the paper what they are fitting but my guess is that they are basically using it as a classifier although its very unclear what the classes are they are training against. But my guess is it is some variable that indicates a diagnosis of FM.

The fitting process gives the set of weights that are the best that can be found to discriminate between the two classes of data given a linear combination of weights and input vector which is the passed through a logistic function which acts as a kind of threshold function.

Its not surprising that they get the graph in fig 2 showing separation of data if that is what they have fitted their regression model to. It is valid to do that but they should quote results against a set of test data that has not been used in generating their regression model. At that point it would become interesting but then its a case of trying to understand the weights and what they mean in terms of the data and the biology of the situation.
Yes, this was my thought - that they would need to report the extraction of the of the threshold criteria on a test group and then apply it again. Otherwise there are no p values because of the large Bonferoni correction needed.

But I think there is a much deeper problem here. What they have done is find a threshold that classifies patients the same way as they are classified by symptoms. That would be reasonable in a perceptron situation because that is about how a system classifies. But the point of a diagnostic test is not to classify patients the same way as by symptoms. It is to give you different information - otherwise it just tells you what you already know. There is a common misconception amongst doctors that a perfect diagnostic test correlates perfectly with symptoms but if you think about it a test that does that is useless. I good test is a biomarker of a process that you think might be causing the symptoms. Sometimes it correlates very well with symptoms, maybe 90%, and then it is useful for finding the 10% that did not have the process after all. Sometimes it correlates rather weakly , so you can find the 10% with the process amongst masses of others who can be reassured.

My thought is that in general a good biomarker for a process is one that makes some sort of sense in terms of a process. The combination of high calcium and high alkaline phosphatase is a good marker for hyperparathyroidism because the process raises the calcium through a mechanism that involves bone turnover and alkaline phosphatase. You do not want a biomarker that just marks a symptom pattern because that symptom pattern might have a dozen different causes. I am afraid I think the logic behind this study, as reported, is fundamentally upside down.

If the researchers have found some results on cytokines in people with FM then we should look at those results presented in an intelligible format and sit down and think what they might perhaps tell us about processes that they might be biomarkers for - maybe to show that there are three sorts of fibromyalgia. If, for instance, a high IL-8 crops up quite often then it might be useful to do studies on 'high IL-8 FM' in the hope of showing that they formed some sort of homogeneous group in terms of process. That would advance science. But to study people who went above some best fit threshold that happened to identify the same people as you identify by symptoms does not seem like scientific sense to me.

I seriously worry that 'clinical science' is becoming just a way for people to reinforce their preconceptions and thereby prevent them from coming to understand anything. Still, there always was 60% chaff with the grain. Even if it is now 90% there is still some interesting stuff coming through.
 

barbc56

Senior Member
Messages
3,657
Likes
4,964
It looks like the authors of the study cited above published another study in 2012, Unique Immuologic Patterns in Fibromyalgia. The study was highly criticized.
http://www.ncbi.nlm.nih.gov/pubmed/?term=Unique immunologic patterns in fibromyalgia

Here's a critique of that study which resulted in a patent for a test.

In March 2013, EpicGenetics started selling the ‘test’ for $744 to physicians and directly to patients (http://thefmtest.com/about-the-fm-test/). This test is unlikely to be reimbursed by insurance companies. In the BMC Pathology paper they `declared that they have no competing interests.’ Either they forget to tell the reviewers they were about to market the ‘test’ or they concealed it. There was a lot of money involved for the company and Dr. Gillis, and it is almost certain that the reviewers would have taken note of this fact. If the authors just ‘forgot’ and asserted no conflict, they still had a lot of time to correct the error and to tell the journal.

Cytokine levels are abnormal in many physical and mental conditions. The authors studied none of those conditions. They provided almost no information about patient selection or many other vital data for a comparison trial. The CONSORT statement on publishing trials offers guidelines on the reporting of data. They didn’t follow these guidelines. As a clinical study, it was very poorly planned and carried out. As a pathology report it might pass. But the data did not in any way addresses the validity and reliability of their ‘test’ to diagnose fibromyalgia. I conclude that they (EpicGenetics and Dr. Gillis) were trying to fool people into buying the scientifically unproven $744 test
http://www.fmperplex.com/2013/02/25/junk-science-junk-ethics/

Are the same criticisms for that study applicable to this most recent study?

I wonder if these authors are going to use the result to patent another test as they did in the first study?.Possibly a more "refined" test whose validity is also questionable?

My foggy brain is having difficulty trying to sort this through. Maybe I missed something obvious?

Barb

ETA I crossed post with Dr. Edwards which made this a bit clearer.
 
Last edited:

lansbergen

Senior Member
Messages
2,512
Likes
2,685
The combination of high calcium and high alkaline phosphatase is a good marker for hyperparathyroidism because the process raises the calcium through a mechanism that involves bone turnover and alkaline phosphatase.
I have been thinking for a long time something is wrong with the calcium system. As I could not find anything useful I moved on but Levamisole influences both calcium and alkaline fosphatase. .