Steroid dynamics in myalgic encephalomyelitis / chronic fatigue syndrome: a case-control study [...], 2025, Thomas, Armstrong, Bergquist et al

I think you may have missed my point. I thought I was clear. I am saying that the researchers here did not gather crucial information about the participants in this study e.g. menopausal, having regular periods, taking hormonal contraceptives. They did not ensure that samples were taken at the same time in the menstrual cycle.


I don't think that is true. We know that oral contraceptives fundamentally change hormonal levels. They do that because they have to change an aspect of the way the female body works, to make it not fertile.

We know that hormonal levels vary greatly during the menstrual cycle.

So, if in one sample you have 50% of the females taking oral contraceptives, and 30% were sampled during their period, and, in the other sample, only 10% of the females were taking oral contraceptives and only 10% were sampled during their period, then the average ratio of various hormones for the cohorts will be substantially different.






I do understand this, and I am sure that securing more funds was a major motivation for publishing the study. And, that is a big part of my concern here.

The researchers have said 'Hey, here is a problem! Please give us money to study it further.' And yet, you have not established that there really is a problem, because the participants were not adequately characterised. You have drawn conclusions that may or may not be true. We cannot know if the finding is true at this point. This study's abstract makes the case for studying hormones in ME/CFS more compelling, and may result in other studies that perhaps are more likely to bear fruit being delayed.

I might try use an analogy to help. If one car in traffic breaks down then the cars behind it are also stopped in their progression. This outcome highlights a relationship in progression. This is relevant to OCP impact on controlling steroid production. Even in this instance you expect relationships to be maintained to a degree that it's statistically significant. What you are arguing is why us saying that we didn't see differences in cortisol or other steroid levels between ME and controls is weak. I agree, without accounting for the factors you highlight it's not going to be rigorous enough to say we can rule out cortisol belong low in ME because we found no such change.

Track record is a major reason everyone publishes every paper, it's the only way to survive in academia, it's why so few researchers exist in ME/CFS and such little research gets done (it's hard to build a track in this field). You are concerned we are trying to secure grants to do more research? Why? That doesn't make sense to me.
 
Last edited:
More from AI, this time on the menstrual cycle:

"The ratio of female hormones, particularly estrogen and progesterone, changes significantly throughout the menstrual cycle. Estrogen levels are generally higher during the first half of the cycle (follicular phase), including a peak before ovulation, and then drop after ovulation. Progesterone levels are low during the first half of the cycle and then increase significantly during the second half (luteal phase), after ovulation. The ratio shifts dramatically, with estrogen dominating before ovulation and progesterone becoming dominant after ovulation."

Here's an analogy. It's like you put an assortment of fruit in one basket, with a random but unknown quantity of oranges, red apples, green apples and bananas, and another assortment in another basket, possibly with quite a different mix. Then you averaged the colour of the fruit in one basket and said 'this is normal fruit colour', and said of the other basket, 'the average fruit colour is different', so the fruit in the second basket must be abnormal.


Even for the men, there are things that will change the ratio of hormones
"Hormone levels in men, particularly testosterone, fluctuate throughout the day. Testosterone levels typically peak in the morning and gradually decline throughout the day, reaching their lowest point in the evening."

There are lots of papers on how one-off and regular physical activity changes levels of various hormones. We know that higher levels of physical activity increases levels of cortisol, for example. I think differences in physical activity probably account for a lot of the reported differences in hormone levels in ME/CFS.
 
More from AI, this time on the menstrual cycle:

"The ratio of female hormones, particularly estrogen and progesterone, changes significantly throughout the menstrual cycle. Estrogen levels are generally higher during the first half of the cycle (follicular phase), including a peak before ovulation, and then drop after ovulation. Progesterone levels are low during the first half of the cycle and then increase significantly during the second half (luteal phase), after ovulation. The ratio shifts dramatically, with estrogen dominating before ovulation and progesterone becoming dominant after ovulation."

Here's an analogy. It's like you put an assortment of fruit in one basket, with a random but unknown quantity of oranges, red apples, green apples and bananas, and another assortment in another basket, possibly with quite a different mix. Then you averaged the colour of the fruit in one basket and said 'this is normal fruit colour', and said of the other basket, 'the average fruit colour is different', so the fruit in the second basket must be abnormal.


Even for the men, there are things that will change the ratio of hormones
"Hormone levels in men, particularly testosterone, fluctuate throughout the day. Testosterone levels typically peak in the morning and gradually decline throughout the day, reaching their lowest point in the evening."

There are lots of papers on how one-off and regular physical activity changes levels of various hormones. We know that higher levels of physical activity increases levels of cortisol, for example. I think differences in physical activity probably account for a lot of the reported differences in hormone levels in ME/CFS.

If one tree existed that made all fruit and the type of fruit created were separated by a few enzymes then that also works as an analogy for what I'm trying to describe.
 
Track record is a major reason everyone publishes every paper, it's the only way to survive in academia, it's why so few researchers exist in ME/CFS and such little research gets done (it's hard to build a track in this field). You are concerned we are trying to secure grants to do more research? Why? That doesn't make sense to me.
There are a couple of responses to this.

One is that if a funder is actually doing their job properly and checking past papers by the research team, they will look at this one and say 'this team don't seem to be thinking carefully enough', and won't fund the research you are proposing. Low quality research can also affect reputations and possibly even put funders off the whole ME/CFS research field.

The other point is that one that I already made above. That is, if a funder is convinced by the definitive language used in this abstract that there is a real problem with hormone ratios in people with ME/CFS, and is convinced that the people who did the research are the ones to back, then they may fund this research, instead of other research, perhaps put forward by researchers who make more cautious statements about their hypotheses, that may actually be more likely to move the field forward.

Look Chris, I'm a big supporter of you and your work. I want your research to get funded and I want the studies you do to be good. But this paper just was not up to standard, and that, one way or another, hurts us all. If more research is done in this area, it needs to be hugely better, with lots of attention to the many factors that affect hormone levels.
 
I don't know can someone else help? Which bit of the following do you disagree with?

Individuals have different levels of hormones at a particular point in time due to things like activity levels, oral contraceptive use, stage of the menstrual cycle and whether they are menopausal or not. They also have different ratios of hormones, again due to those things, (and also the time of day when the sample is taken).

Therefore, it stands to reason that two cohorts with different numbers of people affected by those factors will produce different average ratios of hormones. If you don't know how many people with those attributes you have in each cohort, you can't say anything certain about whether one cohort has abnormal ratios or not.
 
There are a couple of responses to this.

One is that if a funder is actually doing their job properly and checking past papers by the research team, they will look at this one and say 'this team don't seem to be thinking carefully enough', and won't fund the research you are proposing. Low quality research can also affect reputations and possibly even putting funders off the whole ME/CFS research field.

The other point is that one that I already made above. That is, if a funder is convinced by the definitive language used in this abstract that there is a real problem with hormone ratios in people with ME/CFS, and is convinced that the people who did the research are the ones to back, then they may fund this research, instead of other research, perhaps put forward by researchers who make more cautious statements about their hypotheses, that may actually be more likely to move the field forward.

Look Chris, I'm a big supporter of you and your work. I want your research to get funded and I want the studies you do to be good. But this paper just was not up to standard, and that, one way or another, hurts us all. If more research is done in this area, it needs to be hugely better, with lots of attention to the many factors that affect hormone levels.

In my experience, track record isn't a big factor in securing philanthropic grants. You're not out competing other ME/CFS researchers with quantity of publications. I think it helps to have at least a paper on ME/CFS that shows your understanding of the disease and a proven source of sample (could be a biobank or known clinician support). But after that I think philanthropy grants are usually based on reviewers knowing the area well and being intrigued by your idea. Many philanthropic grants are given to people with no track record in ME/CFS.

I can't imagine this paper, where we say we didn't see differences in hormone levels, is going to be convincing to a philanthropic funder. As a grant writer I wouldn't base a grant around this paper. We do have current projects in this space that I would though.

Number of publications is really just important for government grants when we are competing with all other diseases. The panels have diverse knowledge but not expertise in ME. For us to compete for the billions in funding we rely on track record for reviewers to feel confident we know what we are doing in an area. And you would agree that any research funded from this global pool is found money. This is a main motivation to publish beyond dissemination of information.

Yes the paper is limited and we write that out in the paper itself. I understand you're unhappy about the abstract. The intention was to highlight that steroid dynamics appeared worthwhile following up in ME/CFS. I guess it can be difficult to write from one perspective and read from another. I can try improve on this.
 
Last edited:
I don't know can someone else help? Which bit of the following do you disagree with?

Individuals have different levels of hormones at a particular point in time due to things like activity levels, oral contraceptive use, stage of the menstrual cycle and whether they are menopausal or not. They also have different ratios of hormones, again due to those things, (and also the time of day when the sample is taken).

Therefore, it stands to reason that two cohorts with different numbers of people affected by those factors will produce different average ratios of hormones. If you don't know how many people with those attributes you have in each cohort, you can't say anything certain about whether one cohort has abnormal ratios or not.

Individuals have different levels of steroids at a particular point in time due to things like activity levels, oral contraceptive use, stage of the menstrual cycle and whether they are menopausal or not. Some steroids will change in ratio differently to other steroid during different times but majority of steroids will stay in similar ratio. The outcome as a whole is that steroids are expected to maintain a relationship with each other across a cohort.
 
The intention was to highlight that steroid dynamics appeared worthwhile following up in ME/CFS. I guess it can be difficult to write from one perspective and read from another.
That isn’t what was written, though. It’s not a matter of interpretation or perspectives, it’s a matter of clearly going beyond the evidence:
These findings illustrate the importance of hormone network dynamics in ME/CFS pathophysiology
The above statement from the abstract simply isn’t true.
I think it helps to have at least a paper on ME/CFS that shows your understanding of the disease
The introductions isn’t doing you any favours then, because it goes way beyond the evidence wrt what we know about ME/CFS. As you surely know better than me, things aren’t so just because a paper says it is - and the sources for the claims you make in the introduction don’t hold up under scrutiny.
For us to compete for the billions in funding we rely on track record for reviewers to feel confident we know what we are doing in an area.
How do you think a reviewer will feel about a paper that clearly goes beyond the evidence?
 
If one tree existed that made all fruit and the type of fruit created were separated by a few enzymes then that also works as an analogy for what I'm trying to describe.
In my analogy, the colours are the different ratios that are associated with a particular category.

I'll make it more concrete.

Say that healthy women in their reproductive years on a particular type of oral contraceptive have a ratio of a couple of hormones of 1.0. (Actually different contraceptives will have different effects, and I think even the number of years that a woman is on the contraceptive can change hormone levels too).
And healthy women in their reproductive years not on oral contraceptives and at a certain point in the menstrual cycle (A) have a ratio of 0.6.
And healthy women in their reproductive years not on oral contraceptives and at another point in the menstrual cycle (B) have a ratio of 0.4.
And post-menopausal women have a ratio of 0.9

Then, if you have 15 women, and 10 are on the oral contraceptive, 3 happen to be at point A in their menstrual cycle and two are post-menopausal, then the average ratio is 0.9.

But if you have 15 women and 3 are on the oral contraceptive and 2 are at point A and all the rest are at point B in their menstrual cycle, then the average ratio is 0.55.

(I may have the maths wrong, I haven't checked, the point is the principle.)

The problem comes when you have 15 women with a disease, and 15 healthy women and no idea whatsoever whether the women were on any oral contraceptive, let alone a specific one. And you have no idea if they were getting periods and, if so, at what point in the cycle they were. So, if you get a ratio of 0.55 for the women with the disease and 0.9 for the healthy women, does this mean that women with the disease have dysregulated hormones?

If you don't have data about attributes that fundamentally affect hormone levels and will confound your analysis, then, unless you have found a wildly abnormal ratio, you cannot claim the cohort has dysregulated hormones.

If average levels of the hormones themselves were not found to be different between the two cohorts (as was the case in this paper), then it's very unlikely that ratios between the hormones will be wildly abnormal to the point that confounding factors can't explain the difference.
 
Last edited:
That isn’t what was written, though. It’s not a matter of interpretation or perspectives, it’s a matter of clearly going beyond the evidence:

The above statement from the abstract simply isn’t true.

The introductions isn’t doing you any favours then, because it goes way beyond the evidence wrt what we know about ME/CFS. As you surely know better than me, things aren’t so just because a paper says it is - and the sources for the claims you make in the introduction don’t hold up under scrutiny.

How do you think a reviewer will feel about a paper that clearly goes beyond the evidence?

But we do think steroid dynamics are relevant to the pathophysiology. We think we did provide some evidence for that. We have published reviews on this topic, we think there is evidence.

I think we disagree on the value of the work, which means the way we would word it differs.
 
Last edited:
Yeah, we haven't pinned down exactly what is off with this pathway. Its not the same for each person with ME/CFS or LC.
A lack of consistency in the way things are "off" is even more reason not to make a claim about hormone dysregulation in a disease. It makes it even more likely that confounders are causing differences.

Track record is a major reason everyone publishes every paper, it's the only way to survive in academia, it's why so few researchers exist in ME/CFS and such little research gets done (it's hard to build a track in this field). You are concerned we are trying to secure grants to do more research? Why? That doesn't make sense to me.
In my experience, track record isn't a big factor in securing philanthropic grants.
Unfortunately, it is true that track record often is not a big factor in securing grants. We have seen far too many researchers secure grants when a good look at their track record would suggest that they don't have the capacity to create a good hypothesis and research it. That's one of the reasons this forum exists, to try to help inform philanthropic donors about researchers in the field and about ME/CFS biology, so they don't throw money at ideas and teams that are very unlikely be productive.


I think it helps to have at least a paper on ME/CFS that shows your understanding of the disease
I agree with @Utsikt, this paper doesn't do that. I think it is worse than that. By ignoring major confounders that render the data uninterpretable, it calls a lot more things than an understanding of ME/CFS into question.

@MelbME, I'm sorry that you are having to speak for your fellow authors and perhaps you weren't deeply involved in writing the paper. I understand that you will be wanting to defend researchers you supervise. Do you understand the point I illustrated with that example of the ratios? This study has this great yawning gap of unknowns relating to major confounders.


But we do think steroid dynamics are relevant to the pathophysiology. We think we did provide some evidence for that.
I think we disagree on the value of the work, which means the way we would word it differs.

It is sounding a lot like 'but, but we really know this stuff is relevant. Therefore, we will ignore the enormous problems with the data, and write this paper up to make it sound as though steroid dynamics are relevant, so we have a better chance to get funding to study it more. Our intentions are good here. Why aren't you pleased?'.

There are three separate issues here.
1. Do the unknown confounders in this study make it impossible to draw any conclusions about whether steroid dynamics are dysregulated in ME/CFS? and
2. Is there any reason to think that steroid dynamics are important in ME/CFS?
3. Are steroid dynamics in ME/CFS a valid thing to investigate?

For 1, I don't see how the answer can be anything other than 'yes'. If I am missing something there, I'd like to understand.
I think it is possible that the reason the study write up was shelved for so long is because some people did understand how limited the data is.

I'd prefer that we don't get diverted onto the second and third questions until we have what I see as the problems with the data and with the way the study is presented in this paper dealt with.

For 2. The evidence supporting steroid dynamics, as being important in ME/CFS is pretty weak. There are holes in the evidence put forward in the paper. We can go through it and I think I and others made some comments up thread. But, whether there is prior evidence or not, that does not make it okay to ignore the fatal problems with the data in this study.

We've seen people do the exact same thing with levels of cortisol. There's really nothing of importance in the absolute values of cortisol of people with ME/CFS across the extensive literature, and yet so many researchers have secured funds to study it again and again. And many papers presented null results as if they mean something.

We've seen the exact same thing with people convinced that CBT and GET fix people with ME/CFS. It's not true, and yet people present the data as if it must be, because they are so convinced by the rightness of their hypothesis.

For 3. Sure, steroid dynamics in ME/CFS are a valid thing to investigate. From the research I've seen, there are other things that I think are more important. Also, there are sensitivities around investigating steroid dynamics that any good ME/CFS researcher should be aware of. One is the history with cortisol because people are so convinced that ME/CFS is a problem of reactions to stress (often with 'and those reactions can be trained away'). Doctors rolling their eyes and dismissing symptoms of ME/CFS as essentially a mood disorder caused by 'women and their hormones' is another. That doesn't mean that the area shouldn't be studied. It does mean that researchers working on this should be aware of their own bias and committed to not making claims that cannot be robustly supported.
 
But we do think steroid dynamics are relevant to the pathophysiology. We think we did provide some evidence for that. We have published reviews on this topic, we think there is evidence.
I don’t think there is any evidence that is good enough to warrant so definitive statements. Most of the research consists of small studies with either unreplicated findings or large and sometimes contrasting differences between findings in different studies, and/or with a lot of potential confounders that have not been controlled for.
I think we disagree on the value of the work, which means the way we would word it differs.
I don’t think it’s a difference about the value of the work. It about if the data supports definitive statements - and it doesn’t yet.
 
A lack of consistency in the way things are "off" is even more reason not to make a claim about hormone dysregulation in a disease. It makes it even more likely that confounders are causing differences.



Unfortunately, it is true that track record often is not a big factor in securing grants. We have seen far too many researchers secure grants when a good look at their track record would suggest that they don't have the capacity to create a good hypothesis and research it. That's one of the reasons this forum exists, to try to help inform philanthropic donors about researchers in the field and about ME/CFS biology, so they don't throw money at ideas and teams that are very unlikely be productive.



I agree with @Utsikt, this paper doesn't do that. I think it is worse than that. By ignoring major confounders that render the data uninterpretable, it calls a lot more things than an understanding of ME/CFS into question.

@MelbME, I'm sorry that you are having to speak for your fellow authors and perhaps you weren't deeply involved in writing the paper. I understand that you will be wanting to defend researchers you supervise. Do you understand the point I illustrated with that example of the ratios? This study has this great yawning gap of unknowns relating to major confounders.





It is sounding a lot like 'but, but we really know this stuff is relevant. Therefore, we will ignore the enormous problems with the data, and write this paper up to make it sound as though steroid dynamics are relevant, so we have a better chance to get funding to study it more. Our intentions are good here. Why aren't you pleased?'.

There are three separate issues here.
1. Do the unknown confounders in this study make it impossible to draw any conclusions about whether steroid dynamics are dysregulated in ME/CFS? and
2. Is there any reason to think that steroid dynamics are important in ME/CFS?
3. Are steroid dynamics in ME/CFS a valid thing to investigate?

For 1, I don't see how the answer can be anything other than 'yes'. If I am missing something there, I'd like to understand.
I think it is possible that the reason the study write up was shelved for so long is because some people did understand how limited the data is.

I'd prefer that we don't get diverted onto the second and third questions until we have what I see as the problems with the data and with the way the study is presented in this paper dealt with.

For 2. The evidence supporting steroid dynamics, as being important in ME/CFS is pretty weak. There are holes in the evidence put forward in the paper. We can go through it and I think I and others made some comments up thread. But, whether there is prior evidence or not, that does not make it okay to ignore the fatal problems with the data in this study.

We've seen people do the exact same thing with levels of cortisol. There's really nothing of importance in the absolute values of cortisol of people with ME/CFS across the extensive literature, and yet so many researchers have secured funds to study it again and again. And many papers presented null results as if they mean something.

We've seen the exact same thing with people convinced that CBT and GET fix people with ME/CFS. It's not true, and yet people present the data as if it must be, because they are so convinced by the rightness of their hypothesis.

For 3. Sure, steroid dynamics in ME/CFS are a valid thing to investigate. From the research I've seen, there are other things that I think are more important. Also, there are sensitivities around investigating steroid dynamics that any good ME/CFS researcher should be aware of. One is the history with cortisol because people are so convinced that ME/CFS is a problem of reactions to stress (often with 'and those reactions can be trained away'). Doctors rolling their eyes and dismissing symptoms of ME/CFS as essentially a mood disorder caused by 'women and their hormones' is another. That doesn't mean that the area shouldn't be studied. It does mean that researchers working on this should be aware of their own bias and committed to not making claims that cannot be robustly supported.

1. The confounders are relevant to steroid levels but have little relevance to relationships between steroids.
2. Yes the study shows something we did not expect in a cohort of 25, the loss of relationships between steroids.
3. This loss of relationship requires follow up investigation, if people have this data we want them to analyse the interconnecting relationships, it's also worth a prospective study to validate.

I think you might be assuming it's something that is being confounded because of lack of information on cycle, etc. I keep saying I don't see these as big confounders for relationships between steroids across a network, it would be a big confounder for the direct comparison of steroids levels (eg. cortisol) between ME and controls, which is what I think you might be thinking of. You'd have 100+ connections across the steroid pathway, estradiol and progesterone flipping during a menstrual cycle is just 1 relationship, even if that flipped you'd still have 100+ relationships that don't and that's just in one person, it wouldn't matter statistically across a cohort. I'm just using this as a simple example. Now this finding could be due to an outlier cohort, that's why it needs to be validated in another cohort.

"That doesn't mean that the area shouldn't be studied. It does mean that researchers working on this should be aware of their own bias and committed to not making claims that cannot be robustly supported."
I agree and this needs to be true of all research, not just of research that have been misused in the past. Nothing is stopping people from misusing any research in the future. I still think some of my early metabolomics work gets misused to sell vitamins.
 
Last edited:
Thank you @MelbME for taking the time to respond to criticisms and questions. I like your summary, it helps me to understand what the paper is about and what it might be pointing to that is interesting and needs further study:
1. The confounders are relevant to steroid levels but have little relevance to relationships between steroids.
2. Yes the study shows something we did not expect in a cohort of 25, the loss of relationships between steroids.
3. This loss of relationship requires follow up investigation, if people have this data we want them to analyse the interconnecting relationships, it's also worth a prospective study to validate.
 
1. The confounders are relevant to steroid levels but have little relevance to relationships between steroids.
That is not true.

I've given you a number of examples where ratios fundamentally change due to things other than ME/CFS status, age and sex - things that this study did not have information on. Things like stage in menstrual cycle or being on an oral contraceptive do affect relationships between steroids, not just progesterone (although the abstract does specifically mention the progesterone pathway). Upstream steroids are affected too.

Screenshot 2025-08-04 at 7.20.13 am.png

Here's just one example, a different way of expressing my ratio example above. This chart suggests that ratios of progesterone, estrogen and testosterone are close to 1 to 1 in Phase 1 of the menstrual cycle. On other days, the ratios are different, often extremely different. If you don't know how many of your female participants are at each point in the cycle, you can't know if your mean ratio is indicative of abnormal steroid networks in the individuals in the ME/CFS cohort.

Here's a comment about the impact of oral contraceptives:
Given that breast cancer risk increases with hormonal exposure, our finding that four widely prescribed formulations more than quadruple progestin exposure relative to endogenous progesterone exposure is cause for concern. As not all formulations produce the same exposures, these findings are pertinent to contraceptive choice.


I think you might be assuming it's something that is being confounded because of lack of information on cycle, etc. I keep saying I don't see these as big confounders for relationships between steroids across a network, it would be a big confounder for the direct comparison of steroids levels (eg. cortisol) between ME and controls, which is what I think you might be thinking of.
No, I'm absolutely not talking about direct comparisons of individual steroid levels between ME and controls. Your study found no differences between levels of individual steroids in the ME/CFS and healthy cohorts, and that is fine.

Collectively the results suggest dysregulation of HPA axis function and progestogen pathways, as demonstrated by altered partial correlations, centrality profiles, and steroid ratios.
You don't have enough information to be concluding that. I understand all the things you have been saying, that you want to secure funding, that you want to establish track records, that you want to demonstrate that your teams have been doing things. Those are reasonable things. But none of that justifies making claims that the data does not support. The results could possibly suggest the things you propose, but the more likely scenario is that confounders account for the variations reported in this paper. Without information about the confounding factors, we can't know what is going on.
 
Last edited:
I haven't been able to read the whole paper, and anyway my extremely ancient maths/stats undergrad degrees are not up to following the details of the analysis.
From the parts I have skimmed, this part from page 15 on the PDF and table 10 seems the most interesting and worthy of further research:

In healthy physiology, steroid hormones form an interdependent network, where changes in one hormone aretypically met with compensatory shifts in others. Thisdynamic coordination, mediated by feedback loops,enzyme driven conversions, receptor density, and circadian timing [47], allows the endocrine system to maintaininternal stability across a range of physiological demands.Partial correlation networks may capture this underlyingcoherence, highlighting how hormones change togetherrather than in isolation. In ME/CFS, however, the striking loss of these correlations suggests that this adaptivecommunication has broken down. This network collapse may be driven by biological disturbances observedin ME/CFS, including chronic inflammation [48] (which Thomas et al. Journal of Translational Medicine (2025) 23:829 Page 16 of 19can regulate steroid hormone metabolism and action)

So they are not looking at between group comparisons of ratios of hormones, which they agree don't reach significance once corrected for multiple comparisons.

The network analysis does something different. It does correlations within each group between pairs of hormones, and then compares the strong pairwise partial correlations found in the healthy group with the much less strong correlations found in the ME/CFS group. They say this suggests a some sort of disruption in the internal stability of the hormone interactions in ME/CFS.

Have I got this right?
 
1. The confounders are relevant to steroid levels but have little relevance to relationships between steroids.
2. Yes the study shows something we did not expect in a cohort of 25, the loss of relationships between steroids.
3. This loss of relationship requires follow up investigation, if people have this data we want them to analyse the interconnecting relationships, it's also worth a prospective study to validate.
Thank you for the discussion and explanations.
 
I had not fully read the study, because to me it was obvious that if there is data that will be deeply confounded and no information with which to unravel the confounding, then you have no basis on which to make conclusions.

But, I have read it now.



*****Two group comparison of absolute steroid levels
No significant differences in absolute (untransformed) steroid levels were obtained when comparing the two groups (ME/CFS & CTRL) using Mann-Whitney U Tests (Table 2, Fig. 1)
So, as we have already noted, no differences in absolute levels of the steroids, including in groups of just males, or just females.



*****Comparison of steroid: steroid ratio levels between ME/CFS and control groups
After adjusting for multiple comparisons using the false discovery rate (FDR), these differences did not retain statistical significance (Table 6).
So, no differences in steroid ratios between ME/CFS and control groups, overall, and in the male and female subsets.


*****Partial spearman correlation analysis of steroid interrelationships and concentration network analysis
This is a regression analysis, basically asking the question of how correlated levels of steroids are with levels of another steroid. They found relationships between a lot of the steroids in the controls, but not in the people with ME/CFS.
There were 57 significant partial correlations that were demonstrated to be different between groups (ME/CFS vs. CTRL) (Table 10).
There are potentially interesting things there, but, confounders could easily explain the differences. I've looked at some of the pairs of steroids levels of which were shown to be correlated in the control group. There are plenty of reasons why pairs might not be correlated that are unrelated to disease, e.g. times during the menstrual cycle when the relationships are different. It's just not possible to conclude that there are ME/CFS-related differences here.

There were some analyses done to try to predict cohort on the basis of the steroid levels. It is noted that predictive power is poor. Looking at the heat map, differences in steroid levels vary a lot by individual.
 
Screenshot 2025-08-04 at 9.03.29 am.png

Have a look at the second comparison there for androstenedione (andro) and androstenedione (AED). Both are precursors to testosterone. They can both be found in body building supplements. We don't know whether people were taking supplements.

If we look at the levels data for Androstenedione (Andro) and Androstenedione (AED) in Tables 3 and 4, the levels for androstenedione (andro) aren't even given. We only have the levels for the AED version. I don't know why not all of the steroid levels were reported.

Screenshot 2025-08-04 at 9.21.54 am.png

Also, have a look at the variability in the level of androstenedione (AED) in Table 4. The controls have a mean of 0.75 and a standard deviation of 0.49. The ME/CFS group have a mean of 1.21 and a standard deviation of 1.369. There's huge variation, even in the controls. The same is true of other steroids. That variation means that the small sample sizes are really a problem - the specific findings could easily be different in another sample.
 
I haven't been able to read the whole paper, and anyway my extremely ancient maths/stats undergrad degrees are not up to following the details of the analysis.
From the parts I have skimmed, this part from page 15 on the PDF and table 10 seems the most interesting and worthy of further research:



So they are not looking at between group comparisons of ratios of hormones, which they agree don't reach significance once corrected for multiple comparisons.

The network analysis does something different. It does correlations within each group between pairs of hormones, and then compares the strong pairwise partial correlations found in the healthy group with the much less strong correlations found in the ME/CFS group. They say this suggests a some sort of disruption in the internal stability of the hormone interactions in ME/CFS.

Have I got this right?
That looks correct to me.

@Hutan 's concerns are all valid for comparisons of absolute levels and hormone ratios--even if there was a positive finding in that respect, I probably would still be skeptical.

But I do think the main finding about a complete lack of partial correlations in the ME/CFS cohort is striking, and it's a finding that cannot be easily explained by confounders. The reason is that even if different proportions of the groups were on contraceptives or were at different points in their cycle, you would still expect some strong internal correlations.

Progesterone, estrogen, etc. are not static within these networks--they're going to get converted into other metabolites, and they're also going to exert regulatory effects on the enzymes that mediate these pathways. So even if [edit: different numbers of people were] taking progesterone, you would expect to see correlations between progesterone and its immediate downstream metabolites, as well as the effects of progesterone suppressing the production of other metabolites in a feedback loop.

The high variability in the ME/CFS group is likely what's driving the breakdown of the network, and that in itself is the interesting finding here. What this indicates is that something is throwing a big sledgehammer into a system that usually has strong internal regulation. Where you'd normally expect two metabolites that are separated by one enzyme to be strongly correlated with each other, suddenly they're not--and this is seen across the whole system. What this strongly suggests to me, coming from a lab that studied transcriptional regulation, is the presence of a transient signal in the ME/CFS group that exerts a stronger regulatory effect on multiple enzymes in this pathway than the normal internal feedback loops. Whatever it is could be immune, it could be endocrine, I can really only guess.

It's still technically possible that the aforementioned confounders can mimic these results, and I think the discussion does acknowledge that. But in order for that to happen, my sense is that you'd have to have an extremely homogenous control group in terms of being similar ages, sampled at similar points in the cycle, or all being on similar contraceptives, and then simultaneously have massive variability for those factors in the ME/CFS group. It's possible, but unlikely that things shook out that way.

The best way to validate this particular finding would probably be to take serial samples from the same participants, so that you're only assessing the strength of correlations within the individual and then comparing the overall network between groups. But it would probably be quite difficult logistically to get enough samples from the same person to get a robust correlation.
 
Last edited:
Back
Top Bottom