Columbia University: Insights from Metabolites Get Us Closer to a Test for Chronic Fatigue Syndrome A study led by researchers at the Center for Infection and Immunity (CII) at Columbia University’s Mailman School of Public Health has identified a constellation of metabolites related to myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). Combining this data with data from an earlier microbiome study, the researchers now report they can predict whether or not someone has the disorder with a confidence of 84 percent. The paper on which this article is based is also discussed on this thread here.
Very interesting read!! Isn't a .836 predictive score pretty decent considering that ME/CFS is believed to include a heterogeneous group of patiens?
This website ranks AUC (area under the curve) scores like this: ( a score of .50 is like a coin toss ) .90-1 = excellent (A) .80-.90 = good (B) .70-.80 = fair (C) .60-.70 = poor (D) .50-.60 = fail (F) [This table also shows up in this 2009 paper on diagnostic accuracy. ] I've also seen it said that predictive test scores should not be judged in absolute terms, but rather as relative to the current best predictive test, so, if the best test is .70, then a test that manages .80 would be considered a big leap forward.
Hmm. Going by this article (and what I may or may not have read from the actual paper, my memory is hazy), I think that they are saying that they can separate out the patients and the controls used in the microbiome study and this metabolic study (I think they were the same people used in both) quite well on the basis of some selected parameters identified as important from both studies. While that is interesting, I would not say that it's a predictive model yet or that it means that we are close to lab tests. Given that they are applying a model developed from the results from one set of people to the same set of people and given they have lots of parameters to choose from, I would be very surprised if their model was not good at separating out patients and controls in that set of people. I could be wrong (hopefully I am), but I don't see any mention of applying this model (which includes quite a large number of parameters) to a completely new set of patients and controls and finding that it successfully separates out the two groups. If indeed they have not done that, then they are doing what we have criticised others of doing, which is making their findings seem a lot more significant than they are at this stage. This problem was discussed in detail in the thread https://www.s4me.info/threads/treat...th-sodium-dichloroacetate-comhaire-2018.2885/ where Frank Comhaire was doing something similar in principle, and was rightly slammed for it. Edit: I have re-read the CII study now. They definitely have not tested their model on an independent cohort. This is from their report: It erodes trust when researchers make claims of biomarkers and lab tests being imminent on the basis of very preliminary data.
And this is a bit of a stretch too. The researchers could separate out patients from controls on the basis of the 10 metabolites showing the most difference - with a certainty of around 82.0%. By adding in the 8 most useful bacterial species in the trial participants' faecal samples, they increased the certainty to 83.6%. To say that the inclusion of the prevalence of the bacteria species improved the certainty of the model is true, but bearing in mind that this is an unverified predictive model, an improvement in certainty of 1.6% is pretty unexciting. Perhaps it just sounds more sexy to have a model based on microbiome and metabolome parameters? I'm not criticising the paper so much as the unwarranted spin of this article. Here's the thread on the paper itself: https://www.s4me.info/threads/insig...ehensive-metabolomics-2018-lipkin-et-al.4873/
The same thought crossed my mind, but I'm not so sure that it is easy to find a group of bacterial and metabolite differences in a large proportion of the ME/CFS patient and in very few controls. I assume you have to find the entire group of differences in (most) ME/CFS patients and the entire group has to be absent in (most) controls. That seems a lot less of a matter of chance than finding one difference between patients and controls. [But what do I know?] Also, if it was there, I missed a comparison between the "ME with IBS" and an "IBS without ME" group* to see if the differences might all be attributable to IBS. *There was an "ME without IBS" group.
I have the same hesitation about the predictive modelling aspect of this research, especially if they haven't tested their predictions on a separate cohort. I've just been looking at some of the Workwell 2day CPET information, and they get 95% predictive accuracy from this testing. Not as useful, in the sense that you have to put people through the test that sets off PEM, but it would be interesting if they could test the metabolomic/microbiomic data on patients who have done 2 day CPET. And I'm also left with the conundrum - If you diagnose people in two ways - 1. by a doctor interview and symptom questionnaire 2. by metabolomics And you get an 85% overlap, which is the inaccurate one? Maybe the metabolomics are 100% accurate and the 15% who didn't show up on the metabolomics had been misdiagnosed by the doctor, or maybe the doctor was 100% accurate, but not everyone with ME has the same metabolic signature. Or am I talking bollocks? As for developing an animal model, I'm skeptical. Maybe give them an infection and while they have the infection, over-exercise them instead of letting them rest, then separate out the ones that fail a 2 day CPET, if you can do such a thing with mice.
I don't think it would be very hard. The number of metabolites they tested for was over 500. It's not that surprising that 10 of them were quite different in the patients compared to the controls. Even then, those 10 metabolites still only separated patients from controls with moderate certainty. And remember, it's not as if every patient had abnormal levels of all ten of those metabolites. A single patient might just have abnormal results for a few of those metabolites. We have seen a number of metabolite studies now that tell some generally similar stories but the detail of the metabolites with the biggest differences between controls and patients I think varies quite a bit. This makes it quite unlikely that this study is the last word on the combination of metabolites that identifies ME. I agree with others who suggest that more factors need to be standardised in these metabolite studies - e.g. all women, no medications or supplements, standard diet for a month before samples are taken, same time of menstrual cycle.
What we need is a means to do that test in vitro, with a tissue sample. So the whole organism doesn't have to go through the trauma.
I think that's where the Ron Davis nanoneedle, or the Seahorse technology with blood cells comes in - finding that there's something wrong with the aerobic metabolism in our cells, rather than having to test our whole system's aerobic metabolism with exercise.
I am more than sceptical about animal models. They DO NOT WORK. I spent years studying this. I am sceptical about anyone who says they are going to use animal models. Now I remember why I was sceptical about Ian Lipkin.
This post and several following have been merged since it introduces another article about the same research. https://www.medicalnewstoday.com/articles/322414.php ''We're getting close to the point where we can develop animal models that will allow us to test various hypotheses, as well as potential therapies. For instance, some patients might benefit from probiotics to retune their gastrointestinal microflora or drugs that activate certain neurotransmitter systems." Well if that's all he can come up with...probiotics??! That makes me feel really hopeful
Well - if animal models are all he can come up with... They don't work, and they are pointlessly cruel.
No, they have not. See this thread. https://www.s4me.info/threads/colum...for-chronic-fatigue-syndrome.4947/#post-88857
How many potential biomarkers have we at this stage. Time for another break from the world of M E happenings
https://twitter.com/user/status/1023246864437657600 https://twitter.com/user/status/1023261631709683712
Job Advert at Columbia University; "The Center for Infection and Immunity (CII) in the Mailman School of Public Health (MPSH) at Columbia University - one of the world's most advanced academic centers focused on microbial surveillance, discovery and diagnosis (https://www.mailman.columbia.edu/research/center-infection-and-immunity) has an immediate opening for a Data Manager to join a team of researchers working on understanding the role of infectious and immune factors in the development of brain disorders across the life course. Under the supervision of Dr. Mady Hornig, Director of Translational Research at CII and Associate Professor of Epidemiology at MSPH, who specializes in research on gut-brain-immune interactions in autism, attention-deficit/hyperactivity disorder (ADHD), Pediatric Autommune Neuropsychiatric Disorders Associated with Streptococcal infection (PANDAS), mood disorders, myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS) and cognitive aging, the incumbent will be responsible for preparation of quality-controlled data sets comprising clinical and biological data from cases and controls across several large projects" https://www.snagajob.com/job-seeker/jobs/job-details.aspx?postingid=509696421