Gut Bacteria Are Different in People With Chronic Fatigue Syndrome (TNYT)

[Daffodil]

"The scientists also discovered that people with C.F.S. had higher blood levels of lipopolysaccharides, inflammatory molecules that may indicate that bacteria have moved from the gut into the bloodstream, where they can produce various symptoms of disease."

are lipopolysaccharides in the bloodstream proof of leaky gut?

thats what everyone is saying..including people who treat HIV dementia

i cant seem to find any literature on HIV dementia patients treating leaky gut and improving..?

 

[panckage]

Converting the percentages in the table back to number of participants doesn't seem to work without ending up with half a participant, so I was wondering why.

Looking at the study itself the group sizes weren't constant. The control group contains between 34 and 39 people per test while the CFS group contains between 44 and 49 people per test.

@JaimeS may want to put a disclaimer to her post that the n=165 chart is not actually from this study

 

[dannybex]

Elevated LPS from bacteria you'd expect to see in the gut would be. I think they're just not calling it 'leaky gut' because the alt. medicine practitioners pioneered the idea of inflammation causing leaky gut and leading to food intolerances (the latter of which is still 'imaginary' in the minds of many mainstream practitioners).

Isn't it, or can't it also be leaky gut causing inflammation?

 

[JaimeS]

Isn't it, or can't it also be leaky gut causing inflammation?

The difference is then semantics. That the cause of food intolerances is proteins from those foods escaping into the bloodstream before being broken down and then 'attacked' by the immune system was the initial theory behind 'leaky gut' and food intolerance. It was also always suggested that the cause for the increased 'leaking' was inflammatory in nature. Moreover, the mechanism behind intolerances to foods doesn't affect that the phenomenon of food intolerance exists.

Maybe I'm misunderstanding you, here? Can you clarify -- are you saying that this shows there's no such thing as food intolerances?

 

[JaimeS]

@JaimeS may want to put a disclaimer to her post that the n=165 chart is not actually from this study

Thank you! I added that this is an example to show how this sort of chart works.

 

[JaimeS]

Sorry I'm getting confused by this. I thought that table b referred to percentages of participants, but it can't because when you convert the percentage figures back to actual number of participants, 5.11 percent of 87 participants (48 ill + 39 controls) is 4.4457 actual participants instead of a whole number.

So: you're definitely right that this can't be percentages of participants! Here's from the original text:

b) confusion matrix for random forest analysis (values are presented as percentage) and ROC area under the curve (AUC value for 97% OTUs collapsed at the genus level. Mean AUC ROC value for five times repeated, 10-fold cross-validation.

Can any stats experts explain this more clearly? I provided helpful links!

-J

 

[JaimeS]

So I asked a stats expert to translate that into English. This is what I got (and this is why we aren't getting whole numbers when we multiply it by the number of subjects!)

Being an old school stats person, I dislike the growing reliance on random forest -- there are a ton of caveats with it which are often ignored.

Random forest selects (randomly) the factors to classify the data. Once classified, it says how well it works on the data. Typically you take 70% of the data to use in creating the random forest, and then apply it to the last 30% of the data. A 10-fold, means you randomly shuffle the data and repeat this process 10 times. Confusion Matrix: Means how well the random forest performed.

ROC is Receiving Operator Characteristics... and deals with the concept of Sensitivity and Specificity which is like the Heisenberg uncertainty principle. You know one or the the other. It indicates how many factors you need to include to get various accuracy of results.

http://gim.unmc.edu/dxtests/roc3.htm has some interesting similar charts -- note the "These differences turn out not to be statistically different, however.", you can get good looking results that for one reason or another, are actually not reliable. A lot of machine learning types, know how to execute the techniques but lack the background to determine if the results (besides looking good) are statistically significant. Sample size in each category selected by random forest is critical!

I notice the reference to genus level, so I assume that we are talking about bacteria here.

One thing that I dislike in the chart was the x-axis were not in the same scale, which creates a false illusion to people reading it. Also the usefulness of the data is only good to determine if a person is healthy or not. A person may have a different illness and be a positive positive for CFS.

The last part I don't believe makes sense in context because the person in question hasn't read the whole study -- they think it's just the number of all bacteria rather than specific genera seen high in CFS.

-J

 

[dannybex]

The difference is then semantics. That the cause of food intolerances is proteins from those foods escaping into the bloodstream before being broken down and then 'attacked' by the immune system was the initial theory behind 'leaky gut' and food intolerance. It was also always suggested that the cause for the increased 'leaking' was inflammatory in nature. Moreover, the mechanism behind intolerances to foods doesn't affect that the phenomenon of food intolerance exists.

Maybe I'm misunderstanding you, here? Can you clarify -- are you saying that this shows there's no such thing as food intolerances?

Absolutely not. Just that I thought that instead of inflammation causing leaky gut it was leaky gut that caused the systemic inflammation and food intolerances...

 

[JaimeS]

Absolutely not. Just that I thought that instead of inflammation causing leaky gut it was leaky gut that caused the systemic inflammation and food intolerances...

Ah! I see.

At least the way that I was taught it -- twenty years ago btw -- inflammation damages the lining of the gut enough that larger food particles can slip out into the bloodstream. These are viewed as foreign and attacked by the immune system, such that now antibodies exist against the proteins that were in those foods. The next time these are ingested, the body 'recognizes' them as a foreign invader and attacks, eliciting an allergic reaction. Not an anaphylactic one -- an IgG reaction, not IgE.

It definitely is a 'vicious cycle' sort of thing, because the inflammation is then self-perpetuating: you get a reaction to that food from then on, or so the theory goes.

The initial, inflammatory state could be caused by a pathogen to start with.

BTW, I read a study that said "food intolerances aren't an allergy" which stated that food intolerance did did not elicit an IgE response, but newspapers reported it as "no such thing as food intolerances". The researchers were pissed -- I read their sad comments after the fact -- but the damage was done. For some reason, people weirdly believe 'if it can't hurt me, it can't hurt you'.

 

[TiredSam]

From the article in the OP:

Finding a biomarker for C.F.S. has been an ongoing goal for researchers who hope to one day develop a diagnostic test for the condition. Still, the senior author of the study, Maureen R. Hanson, a professor of molecular biology at Cornell, said the bacteria blueprint in the new study is not yet a method of definitively diagnosing C.F.S. The importance of the finding, she said, is that it may offer new clues as to why people have these symptoms.

And from the conclusion of the study:

There is no single precise alteration of the gut microbiota in all ME/CFS patients we examined, but our data converges to support the concept of a less diverse and unstable community of bacteria in the disorder. It highlights the association of specific bacterial taxa with ME/CFS, and the identification of the underlying role of this altered commensal gut microbiota could lead to novel diagnostic and therapeutic strategies that would improve clinical outcome. Future studies may also reveal additional molecular markers that could be combined with gut microbiome information to enhance the sensitivity and specificity of ME/CFS diagnostic assays.

This study may be useful in terms of identifying a direction for further interesting research, but in terms of finding a diagnostic test or biomarker it is still a million miles away from being of any use whatsoever. Just for fun I thought I’d work out how accurate a diagnostic test based on the above findings would be, and unfortunately it only comes out at an accuracy of 4.33%. ie if a random person walks into a Dr's office saying "I've been feeling a bit tired, can you do an ME test?" If the Dr performs a test based on the above findings and it comes up positive, the chances that the person really has ME are only 4.33%.

Here are the calculations:

Take the prevalence of ME as the higher IOM estimate of 2.5 million out of 318 million Americans, = 0.78%.

From table b, percentage of people with ME correctly diagnosed is 83% (52.93 / (52.93 + 11.87) x 100). Actually I make that to be 81.68%, but let them have their claimed 83% for the sake of argument.

From table b, percentage of people who don’t have ME correctly diagnosed as not having ME is 85.5% (30.09 / (30.09 + 5.11) x 100). So 14.5% of people who don’t have ME are incorrectly diagnosed as having it.

So if 100,000 random members of the population go to their Dr asking to be tested for ME, with a prevalence rate of 0.78%, 786 of them will actually have ME and 99,214 won’t.

Of the 786 people with ME, 83% of them will get a diagnosis of ME, so 652 of them.

Of the 99,214 people without ME, 14.5% of them will get a diagnosis of ME, so 14,386 of them.

So out of 100,000 random Americans, 652 + 14,386 = 15,038 of them will get a diagnosis of ME using this test. Of the 15,038 people who test positive for ME, only 652, or 4.33% of them, actually have ME, so a positive diagnosis would only have an accuracy of 4.33%.

 

[Hutan]

In practice of course it wouldn't be anywhere near that bad because if someone is given the test, they probably aren't your average random American. For a start they have to be motivated enough to go to see a doctor to find out why they have no energy. If the doctor is decent, they will have screened a bit further by checking there is PEM, doing iron tests and so on. So, the prevalence of ME in people given the test would be much higher than 0.78%.

And possibly if someone tests positive and doesn't have ME, they may have some other problem, or incipient problem.

But yes, there is some way to go. Not least, some replication of the findings is needed.