• Welcome to Phoenix Rising!

    Created in 2008, Phoenix Rising is the largest and oldest forum dedicated to furthering the understanding of and finding treatments for complex chronic illnesses such as chronic fatigue syndrome (ME/CFS), fibromyalgia (FM), long COVID, postural orthostatic tachycardia syndrome (POTS), mast cell activation syndrome (MCAS), and allied diseases.

    To become a member, simply click the Register button at the top right.

Biomarkers

Persimmon

Senior Member
Messages
135
Ken Friedman stated in a video just posted by ME-CFSCommunity.com that "... there are a number of biomarkers that are almost ready to be used to be able to diagnose Chronic Fatigue Syndrome."

As best I understand, the main ME/CFS biomarker approaches are aimed at identifying:
- Genomic Abnormalities
- Immunological Abnormalities
- Exercise-Response Abnormalities
- Brain Imaging Abnormalities

QUESTION 1.
Does anyone know what, exactly, is needed for recognition of a biomarker? Kochs Postulates stipulate formal criteria that must be satisfied in order to establish a causal relationship between a microbe and a disease is there an equivalent framework for establishing the validity of proposed biomarkers?

The topic of biomarkers was peripherally addressed in a recent article in Nature (Biomarkers: portents of malignancy, Nature 471 ppS19-21, 24th March 2011). It asserts the following:
- To be useful on its own, a biomarker needs to have a sensitivity of at least 90% and a specificity of at least 90%;
- Multiple biomarkers could be used in combination. Referring to 7 proposed biomarkers for lung cancer, the authors write Together these biomarkers are expected to yield a better risk assessment than one type alone
I don't know whether these assertions are generally accepted.

QUESTION 2.
Does anyone know what sorts of biomarkers our researchers are trying to create?

I gather that there are various sorts of biomarkers that are used by the medical community, including those used as:
- A stand-alone diagnostic test;
- A diagnostic tool, but without being a stand-alone diagnostic test;
- A means of establishing how advanced a disease might be in a particular patient (ie for staging);
- A predictor of prognosis / disease development in a particular patient; or
- A means of monitoring a patients response to a treatment.

(Consequently, sometimes it is desirable for a biomarker to yield a binary (yes or no) outcome, as in the case of diagnosis; whereas sometimes a binary outcome would be unsatisfactory (eg in monitoring the course of a degenerative disease, or in monitoring whether a treatment is leading to improvement).)

Are the ME-CFS researchers looking for biomarkers all seeking a diagnostic test? A single marker or a set of multiple-markers? A binary or quantitatively graded measure? etc etc

QUESTION 3.
Do you share Professor Friedman's confidence?

I have a sneaky suspicion that the bar will be set unusually high for ME/CFS; that it would take a very clear-cut test in order to be widely accepted.
 

Snow Leopard

Hibernating
Messages
5,902
Location
South Australia
The fact is that biomarkers for any disease are a bit of a fleeting concept. Few potential biomarkers, including ones in clinical use reach 90% specificity and sensitivity on their own.

Yet 90% specificity is still poor and likely to lead to misdiagnosis. Let me explain, if you have a population of 100,000 people and 1% of them have a particular disease, if you were to test them all with a test of 90% specificity and 100% sensitivity, any person who tests positive is ~90% likely to be a false positive!

The sensitivity and specificity figures are only relevant against the group they were measured against. (in the same way that 'normal' lab ranges are non-specific and of uncertain validity for specific patients). Unfortunately this is something that many people do not understand. (including medical doctors, I believe there was a study that found that many doctors did not understand the concepts of specificity and sensitivity and the rate of false positives)

Why are blood tests used at all then? Because they are used in combination with other testing and observational diagnosis, thus increasing the specificity.

The problem is that if your group happens to be heterogeneous then there isn't going to be a single biomarker and your sensitivity will suffer.

Drugs like Ampligen or Rituximab could be very useful to find biomarkers for (responding) subgroups, provided you had the money to do the many experiments involved - there are thousands upon thousands of potential targets and much testing is needed to show reproducibility.

Once found, biomarkers can potentially be used for all of the possibilities you listed under (2).

I suspect that any biomarkers will be used in combination, rather than as a stand alone diagnostic and this will apply to biomarkers for all diseases.

The fact is that even SNP testing for obvious genetic diseases does not have 100% specificity and sensitivity.
 

oceanblue

Guest
Messages
1,383
Location
UK
Persimmon
1. 90% specificity and sensitivity are the usual standard in CFS research, though sometimes 'AUC, Area Under the Curve', a combined measure of specificity and sensitivity is used instead.
2. Diagnostic is the only proposed use I've seen for biomarkers; people have looked at individual ones and in combination
3. I think Ken Freidman in being a tad optimistic.

Snow Leopard makes a lot of excellent points, inc. that even with 90% specificity/sensitivity there can be a 90% chance of a false positive, and that if the condition is heterogenous (which is quite possible even for International criteria ME, let alone CFS) you won't get anywhere near 90% for either.

Problems with biomarker research
I think there are two other significant problems with biomarker research to date:

1. No replication in independent samples
Most biomarkers are based on thresholds not binary choices, e.g. above a certain score for post-exertional fatigue (Lenny Jason) or below a certain threshold for Natural Killer cell cytotoxicity. But the thresholds are chosen to maximise sensitivity and specificity in the researcher's particular sample i.e. it's the maximum specificity/sensitivity for that particular group of patients and controls. Until it's replicated on an independent sample - using the same thresholds set in the original study - it doesn't mean very much. But then no replication in independent samples is the story of CFS research generally.

2. Ability to discriminate healthy from CFS isn't much use
You or I, or someone off the street could probably easily spot the difference between a healthy person and someone with suspected CFS. In other diseases, e.g. cancer or Rheumatoid Arthtitis, biomarkers are often targeted at people who might have the disease in question, but have very few symptoms i.e. they don't look very different from healthy people and a biomarker can help crystalise the diagnosis.

With CFS, the challenge is to distinguish cases of ME/CFS from other Chronic Fatigue cases - that would be of great clinical value. However, I've never seen a biomarker putting this to the test. I strongly suspect that biomarkers which struggle to hit the 90% specificity/sensitivity target in healthy vs CFS studies will do significantly worse in CFS vs a range of other types of Chronic Fatigue.

Given these problems, and those highlighted by Snow Leopard, I think it's going to be a while before we see the Holy Grail of a biomarker for a CFS. Researchers have struggled to find them even in diseases much better understood and more intensively researched than CFS.
 

heapsreal

iherb 10% discount code OPA989,
Messages
10,089
Location
australia (brisbane)
biomarkers have to be better then a questionaire asking if someone has been tired for 6 months. I dont think biomarkers or questionaires should be used on there own but together, if it picks up 90% of the people with me/cfs then its alot better then what they have now. Plus maybe we need several biomarkers as well like nk dysfunction with certain cytokines like IL6, im sure there are others they could use. IL6 would be a good biomarker to help prove sleep abnormalities along with a proper sleep study that doesnt just rule out apneas but also looks at sleep quality through sleep depth by its stages of sleep.Another test that they have found useful is the exercise treadmill test done on consecutive days. I think they need to use the questionaire along with blood test results for immune and infectious conditions and cytokines, then later the sleep study and maybe the consecutive exercise treadmill test with a cytokine study straight after the exercise tests. I dont think all the tests are necessary but if one wanted more evidence then do more. I know some of these biomarkers maybe common in other disorders but atleast its showing that the person is ill and needs treatment. eg poor sleep study with high IL6 levels, wow this person isnt sleeping very well, lets try and improve their sleep. Medications can be used and tested again to see if they are effective, i dont know anyone who has been tested to see how effective their sleep meds are.

At the end of the day we might not know the exact cause but bloody hell its easy enough to find abnormalities in us to say we are sick, surely we deserve to be treated on these abnormalities even though we dont have a 100% diagnoses and not bloody pace crap or prozac either, proper treatment.

cheers!!!
 

Levi

Senior Member
Messages
188
Initially, biomarkers should be used to define ME/CFS subsets, NOT to identify of diagnose individual patients. Then additional research should be done to study these subsets and replicate and validate the data on a larger scale. Finally, the biomarkers should be cross-referenced and grouped to the subsets and weighted using analytical statistical tools and algorithms:
http://www.flimecfsforum.com/forum/index.php?f=4&t=587&rb_v=viewtopic

Then individual patient diagnosis using biomarkers will be feasible. Its all a long way off. It will be accomplished eventually by patients themselves since researchers have no interest or funding to pursue the categorization of ME/CFS subsets.
 

svetoslav80

Senior Member
Messages
700
Location
Bulgaria
Let me explain, if you have a population of 100,000 people and 1% of them have a particular disease, if you were to test them all with a test of 90% specificity and 100% sensitivity, any person who tests positive is ~90% likely to be a false positive!

In this case you already know that 99% of the people are either healthy or have another disorder (which you're not testing for), which is not the case in real life situations where the probability of true positive is much higher.
 

Snow Leopard

Hibernating
Messages
5,902
Location
South Australia
In this case you already know that 99% of the people are either healthy or have another disorder (which you're not testing for), which is not the case in real life situations where the probability of true positive is much higher.

Of course. My example explains why SNP testing of the population may (and has) failed to be useful for any disease.

If you were to test, say 2.5% of the population (eg those presenting with a particular symptom) and only 0.5% has the disease, then about 20% of the results are still going to be false positives.
 

ramakentesh

Senior Member
Messages
534
Can you talk about biomarkers without considering more than just a few approachs to the etiology of the condition? have you considered that serum angiotensin II is elevated at least in a subset of POTS patients?

If we were to include biomarkers found in POTS (of which some can also fit the diagnostic criteria for CFS) there are many:

Elevated angiotensin II
Reduced MIBG reuptake
Reduced vasoconstrictive responses to angiotensin II infusions
Reduced beta 2 receptor vasodilation
and perhaps increased endothelial NO and hydrogen sulfide activation
 

ramakentesh

Senior Member
Messages
534
And a percentage of POTS patients test positive to autoantibodies that clog of ganglionic acetylcholine receptors in the peripheral autonomic nervous system.
 

oceanblue

Guest
Messages
1,383
Location
UK
Of course. My example explains why SNP testing of the population may (and has) failed to be useful for any disease.

If you were to test, say 2.5% of the population (eg those presenting with a particular symptom) and only 0.5% has the disease, then about 20% of the results are still going to be false positives.
And in that scenario of those with the same symptom eg chronic fatigue, the specificity and sensitivity also need to be measured on CFS vs chronic fatigue, not vs healthy. No one has done that with CFS biomarkers yet and I suspect that when they do the results will be lower specificity and sensitivity.

So although there are gains in 'prior probability' by comparing with a sick population, there are almost certainly losses too through reduced accuracy of the test when comparing against other sick patients.
 

Levi

Senior Member
Messages
188
Many biomarkers that are interesting and at least initially supported by published research are experimental, esoteric, not generally available to the public, and expensive to test for as well. Intelligent weighting of the reliability of a palette of biomarkers will include a measure of pragmatic effectiveness and availability

You night have stumbled onto a great biomarker for ME/CFS, but if it is not available to patients, it becomes a mere curiosity. Of note, if you divide patients into subsets of biomarkers such as ME1, ME2, ME3 etc., then you start out with 100% sensitivity and specificity.
 

Snow Leopard

Hibernating
Messages
5,902
Location
South Australia
Can you talk about biomarkers without considering more than just a few approachs to the etiology of the condition?

I agree, in fact the whole concept of biomarkers is excessively reductionistic. Medicine needs to move towards a more systematic approach. For example, it would be more effective to use the set of tests you listed and calculate the likelihood given realistic thresholds of specificity and sensitivity.

Why don't they do this now? The practise of medicine has long been driven by cost constraints and I believe this is one of the reasons why medical science is unrealistically biased towards 'silver bullets'.

then you start out with 100% sensitivity and specificity.

100% specificity and sensitivity for each sub-set diagnosis, but not necessarily treatment outcomes. Nature always has a way of being fuzzy at the edges.
 

WillowJ

คภภเє ɠรค๓թєl
Messages
4,940
Location
WA, USA
Yet 90% specificity is still poor and likely to lead to misdiagnosis. Let me explain, if you have a population of 100,000 people and 1% of them have a particular disease, if you were to test them all with a test of 90% specificity and 100% sensitivity, any person who tests positive is ~90% likely to be a false positive!

why is that 90% false positives rather than 10%? I'm missing something here...
 

Snow Leopard

Hibernating
Messages
5,902
Location
South Australia
I should rephrase, 90% of the positives would be false positives.

If you had 100,000 samples, 1% of them were true positives then you'd have 99,000 who were true negatives. But if the specificity is only 90%, then 10% of these samples are falsely going to test positive. With 100% sensitivity, you are going to get 10900 positive results, but only 1000 of these are true positives.

So if you had a SNP test that didn't have 100% specificity, then it won't work too well in terms of population screening. (and most SNP tests (for many diseases) have failed to be useful clinical biomarkers, so...)