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

[wigglethemouse]

Could this 2021 paper be of interest in relation to this study.

Thread : Hypothalamic-Pituitary autoimmunity and related impairment of hormone secretions in chronic fatigue syndrome, 2021, De Bellis, Montoya et al
Paper : LINK

Results: Patients in group 1 showed a high prevalence of AHA (33%) and APA (56%) and significantly lower levels of adrenocorticotropin (ACTH)/cortisol, and growth hormone (GH) peak/insulin-like growth factor-1 (IGF-1) vs controls (all AHA/APA negative). Patients in group 1A (13 patients positive at high titers, ≥ 1:32) showed ACTH/cortisol and GH peak/IGF-1 levels significantly lower and more severe forms of ME/CFS with respect to patients in group 1B (7 positive at middle/low titers, 1:16-1:8) and 1C (10 antibody-negative patients).

 

[wigglethemouse]

What has fascinated me about hormone levels in general is that they are often higher at waking and then decline during the day. It made me wonder how the body copes with so much variation. At the back of my mind has been the thought that perhaps the body must like to maintain a ratio between certain hormones, so if they both decline as happens later in the day it is not a problem.

Now reading this thread I come back to wondering just how important is it to maintain ratios between certain hormones.........

@MelbME What does the literature or endocrinology experts say about loss of relationships between steroids? Would you expect to see certain symptoms or do any other diseases exhibit this? I guess what I'm asking is if loss of coupling is something completely new, or is a known phenomenon?

Sorry if you answered this in the paper, I'm not able to read much lately.

 

[Hutan]

Conclusions
Despite no significant differences in absolute steroid levels, network analysis revealed profound disruptions in steroid-steroid relationships in ME/CFS compared to controls, suggesting disrupted steroid homeostasis.

Collectively the results suggest dysregulation of HPA axis function and progestogen pathways, as demonstrated by altered partial correlations, centrality profiles, and steroid ratios.

These findings illustrate the importance of hormone network dynamics in ME/CFS pathophysiology and underscores the need for more research into steroid metabolism.

Chris, I'll try to explain what it's like to read those conclusions, when the data is not nearly good enough to support them and the evaluation of the literature expresses bias, as a woman with ME/CFS and the mother of a young man with ME/CFS. Maybe then you can understand why I feel so strongly about this paper and have criticised it so harshly, when equally bad papers from psychosomatic proponents so often just get a few snide posts and a bit of a laugh from us. I write these posts because I think that you can do things better.

People with ME/CFS have been told that all sorts of things are wrong with them, pretty much all of them based on inadequate data and very often poor analysis. We are routinely called neurotic, perfectionist, unwilling to make an effort, militant, fearful, hysterical and overly dramatic, unable to express emotions, overly ambitious, hypochondriacal, overactive to stress, low cortisol due to burnout, cognitively impaired, depressed, anxious, heartsink inducing, many more.... A few of the descriptions may be correct, but we need them to be solidly based in evidence and described very precisely.

Of course there is something wrong with us, we want to find out what it is. But it's a blow when a paper like this comes out, produced by people we respect, people that some of us are actually helping to fund. It can easily be interpreted as definitively saying 'people with ME/CFS have hormones that are out of whack' when there really is not good data to support that claim. The way the abstract is worded, the conclusions can easily become to be seen as an established fact. That can hurt how people with ME/CFS see themselves and how they are seen by others, in quite a number of ways. For example, the mention of progesterone pathways might lead people to think that women with ME/CFS as a whole might have difficulty with fertility. Imbalanced hormones go nicely with accusations of hysteria. Many people are ready to offer self-improvement strategies to tackle "altered HPA axes".

The paper also produces a loss of trust in the people we are relying to find the answers. It makes me feel that we are simply fodder for the advancement of careers, even though I know that that is not how you see it most of the time. I get the feeling that the researchers might think it is worth stretching the truth if it results in funding (I believe truth has been stretched to breaking point here). Ratios of hormones were not found to be significantly different, and yet the abstract says that they were. The data is not clean enough to warrant any conclusions about differences.

I really hope that the study that is following on from this paper will correct the problems of this one. I hope that it will be cautious in its claims, understanding that it is very easy to create harmful myths that can suck scarce research funds into unproductive areas.

Perhaps you can tell us more about the next study and what safeguards you are building into it.

 

[Sean]

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 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.

Interesting.

 

[Hutan]

@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.

But that finding can potentially be explained by confounders. We really can't say that it's not likely, we don't know. We know that there are confounders that can even flip the correlations. One of the issues is daily fluctuations. A lot of hormone correlations change during the day.

Hormone levels fluctuate throughout the day, exhibiting circadian rhythms. For example, testosterone typically peaks in the morning and declines throughout the day, while cortisol, is highest in the morning and lowest at night.​

​

If people have different normal wake times, that alone could change the relationships for this point sample. Similarly for a variation in the time the sample was taken - all we know is that the samples were taken in the morning.

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.

Things are really complicated e.g. this from AI about the progestins in oral contraceptives, I haven't checked the details, but I think it's good enough to get the gist. The various different progestins in different oral contraceptives degrade to different suites of metabolites.

"Oral contraceptives (OCs) containing different progestins lead to varying downstream metabolites due to differences in their chemical structures and metabolic pathways. These variations can affect not only contraceptive efficacy but also influence various metabolic processes, including lipid metabolism, glucose metabolism, and potential effects on the brain.

Key Differences in Progestin Metabolism:

• Levonorgestrel (LNG):
  A potent progestin, often used in combined oral contraceptives (COCs), is metabolized to several compounds, including 17-deacetyl-noregestimate (NGMN).LNG-containing OCs can impact glucose and insulin levels, potentially contributing to insulin resistance, according to a study published by the National Institutes of Health (NIH).
• Desogestrel (DSG):
  Another common progestin, is metabolized to etonogestrel (ETG), which is a more potent progestin than DSG. DSG-containing OCs may also affect glucose and insulin levels, but with potentially different patterns compared to LNG, according to a study published by the National Institutes of Health (NIH).
• Norethindrone (NET):
  A widely used progestin, is metabolized to norethindrone acetate and other compounds. NET-containing OCs have also been shown to affect glucose and insulin levels, according to a study published by the National Institutes of Health (NIH).
• Norgestimate (NGM):
  NGM is metabolized to LNG-17-acetate and NGMN, with NGMN being the main progestogenic metabolite. NGM is known for its anti-androgenic properties and its impact on lipid and glucose metabolism, according to a study published by Taylor & Francis Online."
• Drospirenone (DRSP):
  An anti-androgenic progestin with mild diuretic effects, it is metabolized to various compounds, including dihydrodrospirenone.

If a researcher is investigating hormones with the granularity that is used here, they have to be aware of the major impact of oral contraceptives, and even the diversity of oral contraceptives. Really, I think you would have to exclude women on hormonal contraceptives - the disruption to normal hormonal levels is just way too big.

 

[jnmaciuch]

But that finding can potentially be explained by confounders. We really can't say that it's not likely, we don't know. We know that there are confounders that can even flip the correlations. One of the issues is daily fluctuations. A lot of hormone correlations change during the day.

Hormone levels fluctuate throughout the day, exhibiting circadian rhythms. For example, testosterone typically peaks in the morning and declines throughout the day, while cortisol, is highest in the morning and lowest at night.​

​

If people have different normal wake times, that alone could change the relationships for this point sample. Similarly for a variation in the time the sample was taken - all we know is that the samples were taken in the morning.



Things are really complicated e.g. this from AI about the progestins in oral contraceptives, I haven't checked the details, but I think it's good enough to get the gist. The various different progestins in different oral contraceptives degrade to different suites of metabolites.


If a researcher is investigating hormones with the granularity that is used here, they have to be aware of the major impact of oral contraceptives, and even the diversity of oral contraceptives. Really, I think you would have to exclude women on hormonal contraceptives - the disruption to normal hormonal levels is just way too big.

But the crucial point is that these confounders almost certainly would not produce the maintain finding here, which is that there was a stark lark of network correlations, for the reasons I already outlined. Regardless of what contraceptive you’re taking, you would still expect a partial correlation between [edit: certain closely related] metabolites.

Taking a contraceptive does not break down all regulatory relationships, neither do hormone cycles. If someone had higher progesterone from a contraceptive, they’d also be expected to have higher levels of metabolites derived from progesterone. Unless something biological was happening to interrupt enzyme function on top of that.

 

[jnmaciuch]

We all agree that contraceptives and hormone levels can change absolute levels and ratios. I think what @MelbME and I are both saying is that due to the way the network analysis is done within a highly interdependent biological system, it is largely robust to those confounding factors. Daily variations and contraceptives wouldn’t produce the findings you see here. They might flip the signs on some of the specific correlations if things were skewed enough and cause some correlations to drop out between groups. They would not cause the ME/CFS group to only have 1 significant partial correlation compared to 52 in controls

 

[forestglip]

I think what @MelbME and I are both saying is that due to the way the network analysis is done within a highly interdependent biological system, it is largely robust to those confounding factors.

They would not cause the ME/CFS group to only have 1 significant partial correlation compared to 52 in controls

I'm having trouble imagining how a specific factor would eliminate 52 (virtually all) different correlations between different metabolites. The simplest explanation to me seems like increased variability preventing significance. And I can imagine there might be increased variability in hormones due to the ME/CFS group varying more in the meds they take than the control group. We don't know if that's the case, but that seems plausible to me.

From a statistical perspective, this loss of significant relationships in ME/CFS may be due to increased variability in steroid levels within the ME/CFS group, reducing the precision of correlation estimates and making it more difficult to achieve significance with FDR correction. Sample heterogeneity which may result from clinical severity or medications, can introduce additional variability, masking significant relationships that may be present within a more homogeneous group (e.g., a control group). At the biological level, significant between-group differences indicate distinct patterns of steroid interactions, suggesting changes in the biological processes underlying steroid regulation and metabolism. These results may reflect a loss of homeostasis in ME/CFS. Likely, the results reflect a combination of increased variability, and genuine biological differences in steroid regulation and metabolism.

They basically say, these results could be due to increased variability from meds or other factors, or they could be due to a loss of homeostasis (which again it's hard to imagine affecting pretty much everything they tested), but they think it's likely a combination of both. I'm just having a hard time seeing how they can say it's likely the combination as opposed to wholly the increased variability.

 

[jnmaciuch]

I'm having trouble imagining how a specific factor would eliminate 52 (virtually all) different correlations between different metabolites. The simplest explanation to me seems like increased variability preventing significance. And I can imagine there might be increased variability in hormones due to the ME/CFS group varying more in the meds they take than the control group. We don't know if that's the case, but that seems plausible to me.


They basically say, these results could be due to increased variability from meds or other factors, or they could be due to a loss of homeostasis (which again it's hard to imagine affecting pretty much everything they tested), but they think it's likely a combination of both. I'm just having a hard time seeing how they can say it's likely the combination as opposed to wholly the increased variability.

I agree it’s due to overall increased variability, but my main point is that even medications and differences in time of day would be unlikely to be the sole source of that variability.

Like I mentioned before, medications can artificially increase or decrease the levels of certain metabolites, but wouldn’t be expected to interfere with the relationship between those increased metabolites and their immediate downstream metabolites. I think that part of the discussion you quoted is meant to cover all their bases wrt limitations, since some of the specific metabolite correlations could be chalked up to medications. But I can certainly think of processes that would lead to a string of downstream transcription factors which could affect many of the major enzymes in this system (especially since it’s not a large number of enzymes in total)

And like I said, it’s a small cohort and a limited finding, I’m not trying to say this is anything definitive or groundbreaking. The lack of correlations in ME/CFS is almost certainly due to increased variability. My sense from modeling biological systems is just that I truly don’t think contraceptives or daily fluctuation could explain more than a portion of the difference between groups—it much more closely resembles the variability of a system where the direct gene and protein regulatory functions are being interfered with. Contraceptives and daily fluctuations affect hormone levels by the same relationships that the partial correlations are meant to detect—not by breaking down those channels

 

[Hutan]

Taking a contraceptive does not break down the regulatory relationships, neither do hormone cycles. If someone had higher progesterone from a contraceptive, they’d also be expected to have higher levels of metabolites derived from progesterone. Unless something biological was happening to interrupt enzyme function on top of that.

But different progestins in the different contraceptives degrade into different metabolites. This study included quite fine grained metabolites, but not all. So, it might be that some people were found to have very little of one particular metabolite, but actually had higher levels of another. They do fundamentally change what happens.

There aren't really 52 separate correlations. If one key hormone and its metabolites are different because of a skew in the type of person included, then that could affect a lot of correlations.

I'm having trouble imagining how a specific factor would eliminate 52 (virtually all) different correlations between different metabolites. The simplest explanation to me seems like increased variability preventing significance.

Yes, I've been playing around with the same idea.

I think too that there might be a difference in the ranges of key steroids due to non-disease factors. If the controls had a wider range of the key steroids, that is, more people with high levels and more people with low levels of the steroids (perhaps due to having menopausal women at one end and women at the point in the menstrual cycle with high steroids at the other), then the controls could be more likely to have significant correlations between steroids.

See these for example:
Female steroid levels
Pregnanolone (PNL) Controls 1.714, SD 2.650 : ME/CFS 0.624, SD 1.350
Pregnenolone (P5) controls 4.136, SD 10.392 : ME/CFS 4.205 SD 5.227

For these two key steroids, the controls had a much wider range of values.

It's like if you only sample people between ages 30 and 35 and try to look for a relationship between age and income - you probably won't find a significant correlation. But, if you sample people from age 5 to 60 - you probably will. There are lots of non-disease reasons why the ME/CFS range for the key steroids might not be very wide. Remember that there were only 16 females in each group.

I should stop spending time on this now. The data is just not good enough to do what was done with it. The uncertainty about what is causing differences between the two cohorts is too great.

Let's hope the next studies are a lot more rigorous.

 

[forestglip]

Contraceptives and daily fluctuations affect hormone levels by the same relationships that the partial correlations are meant to detect—not by breaking down those channels

The point isn't that meds would break down those channels. It's that if half the people take a med that increases the output metabolite of a pathway and half don't, with the input metabolite not being directly affected, the correlation between the input and output becomes much harder to detect.

I don't know the intracacies of what specific relationships would change due to a contraceptive versus the scope of what could potentially change through some more interesting process. Maybe that's more obvious to people like you that are more in the know.

But in any case, it wouldn't hurt to incorporate data on contraceptive use or phase of menstrual cycle, to demonstrate that it's not that, and so that they don't have to put that it might have been the meds in the limitations, right?

 

[jnmaciuch]

The point isn't that meds would break down those channels. It's that if half the people take a med that increases the output metabolite of a pathway and half don't, with the input metabolite not being directly affected, the correlation between the input and output becomes much harder to detect

That’s what I’m trying to explain, maybe the finer point is being lost in translation. Specific correlations between the supplemented hormone and its upstream precursor would break down to some extent, though you’d still expect some compensatory influence. But even with that effect, you’d still expect (for example) the relationship between progesterone and its downstream metabolites to hold up strongly.

Same for many of the other major “hub” metabolites that may or may not be affected by medications. A complete lack of those correlations means heightened variability across the whole system, including in the relationships that would not be targeted by any of those medications. That speaks to an issue in enzymes that govern the conversion between metabolites, which happens to be a small redundant set across the many steps of this pathway.

 

[forestglip]

That speaks to an issue in enzymes that govern the conversion between metabolites, which happens to be a small redundant set across the many steps of this pathway.

Is it possible that a contraceptive changing one hormone level might lead to downregulation or upregulation of these core enzymes so that many more hormones are affected than just progesterone?

 

[Hutan]

But even with that effect, you’d still expect (for example) the relationship between progesterone and its downstream metabolites to hold up strongly.

But not if the supplemented progesterone results in slightly different metabolites. Which it can (see my post upthread).

One thing that might be useful would be to see charts for the correlations. Maybe the data is in the supplementary files. But, regardless, there are too many unknowns.

 

[Hoopoe]

Of course there is something wrong with us, we want to find out what it is. But it's a blow when a paper like this comes out, produced by people we respect, people that some of us are actually helping to fund. It can easily be interpreted as definitively saying 'people with ME/CFS have hormones that are out of whack' when there really is not good data to support that claim. The way the abstract is worded, the conclusions can easily become to be seen as an established fact. That can hurt how people with ME/CFS see themselves and how they are seen by others, in quite a number of ways. For example, the mention of progesterone pathways might lead people to think that women with ME/CFS as a whole might have difficulty with fertility. Imbalanced hormones go nicely with accusations of hysteria. Many people are ready to offer self-improvement strategies to tackle "altered HPA axes".

I agree that it's really annoying that so many people have negative attitudes towards ME/CFS patients, but I don't think we should be trying to change their views. They will use anything against us, and if there's nothing, they will just make something up... because they simply don't want to deal with a frustrating problem and their negative attitude is how they take the distance. It has to be our fault so that their own bad behaviour is justified. There's nothing we can do, we can't force anyone to like us or take an interest in finding treatments. What we can do is not care about what someone might think and focus on concrete things.

 

[Utsikt]

The problem isn’t if it’s likely that it can be explained by confounders or not, it’s if it’s possible.

As long as it’s possible, we can’t make definitive claims.

If you disagree about that, please explain why.

 

[Utsikt]

Specific correlations between the supplemented hormone and its upstream precursor would break down to some extent, though you’d still expect some compensatory influence. But even with that effect, you’d still expect (for example) the relationship between progesterone and its downstream metabolites to hold up strongly.

How do you know that the relationships should hold for the downstream metabolites? Why is that a given?

Are there robust studies that have looked at the effect of messing with just one of the metabolites and seeing what kind of effect it has on the system as a whole? What about if you mess with five or ten of them?

We frequently see paradoxical interactions between different things in biology. If one profession should know that things don’t really behave neatly all the time, it’s biologists.

 

[Trish]

I do share the concern expressed by others about the abstract conclusion, particularly the bit I've highlighted in bold:

Conclusions
Despite no significant differences in absolute steroid levels, network analysis revealed profound disruptions in steroid-steroid relationships in ME/CFS compared to controls, suggesting disrupted steroid homeostasis.

Collectively the results suggest dysregulation of HPA axis function and progestogen pathways, as demonstrated by altered partial correlations, centrality profiles, and steroid ratios.

These findings illustrate the importance of hormone network dynamics in ME/CFS pathophysiology and underscores the need for more research into steroid metabolism.

The highlighting of HPA axis function dysregulation as a finding of note rings alarm bells for many of us because the UK therapists organisation, BACME, that causes us so much trouble by clinging to outdated ideas and practices base their 'theory' on HPA axis dysfunction. In their simplistic model they hypothesise a causative relationship to stress, and therefore, in their faulty logic, we can be fixed by de-stressing with things like mindfulness meditation and psychological therapies, and by increasing activity.

I think it's therefore vital that good biomedical scientists don't write abstract conclusions that highlight the HPA axis as central to ME/CFS unless thay have cast iron evidence of its relationship and moreover whether the correlation has a direction of causation.

I think the last sentence overstates the findings of the study. The findings illustrate an area that warrants more research, not an area that definitively has a central role in ME/CFS.

 

[jnmaciuch]

I'll do one final reply--I think I've spent too much time as it is already.

I am guessing that a lot of the confusion originates from understanding of the method used. A partial spearman correlation is a rank-based (not value-based) correlation which effectively regresses out the effects of other measured variables to try to quantify the direct influence of one variable on another as much as possible. Unlike other statistical comparisons which might be affected by these confounders (like the tests for between-group abundance differences), it is looking at covariance. This has the following benefits:

1. Confounders which change the abundance of any of the metabolites will not, on their own, change the number of significant partial correlations. The reason is precisely because of the following point, which I also stated previously:

But not if the supplemented progesterone results in slightly different metabolites. Which it can (see my post upthread).

Progesterone (and any other hormone which can be supplemented) is readily converted into other metabolites. So if what you're assessing is the extent to which levels of A directly correlate with levels of B in a relationship where A--> B, having some people with higher levels of A from contraceptives doesn't obscure that relationship. You still have a situation where the samples with high levels of A will also have high levels of B, and the samples with low levels of A will also have low levels of B. The fact that some of those high levels of A will be due to contraceptives does not change the [edit: downstream] relationship that is being studied here.

[edit: you’d still expect some number of downstream relationships to hold even if the relationship between progesterone and its own upstream metabolite are confounded.]

2. The regulatory effects that certain metabolites will exert on enzymes in this pathway are still going to be the same regulatory effect regardless of relative abundance. So if metabolite D downregulates the enzyme which mediates A --> B, you're still going to see that correlation between A and D regardless. In a rank-based intra-group correlation, if D has a strong negative influence on A-->B, the samples with high levels of D (even through contraceptive use) are going to have low levels of B, and the samples with lower levels of D are going to have higher levels of B, regressing out the influence of A.

To that point:

Is it possible that a contraceptive changing one hormone level might lead to downregulation or upregulation of these core enzymes so that many more hormones are affected than just progesterone?

Only a few of the hormone intermediaries are bioactive. Yes, supplementing some of those bioactive hormones would probably lead to upregulation or downregulation of some of those enzymes which might affect something at a different point in the pathway. Which is what the "partial" part of partial spearman correlations aims to address--it is assessing the strength of the relationship between A-->B while accounting for the levels of D, E, F, etc. If A-->B is so strongly influenced by D that the relative abundance of A doesn't strongly correlate with the relative abundance of B, then the A--> B relationship is not significant but the D-->B relationship is.

That's what the 52 significant relationships in the control group tells us--that even with the interrelatedness of the system and the general influence of outside factors, there are still a bunch of strong direct relationships in the healthy group. The fact that nearly nothing could be picked up in the ME/CFS group speaks to a strong influence beyond what can be attributed to the confounders that were brought up in this thread.

---

So the takeaway is, as I said before, the way that the network analysis was done makes it pretty robust to those particular confounders.

I think a larger sample size would probably make the analysis more robust to even the very slim possibility that the control group happened to be extremely homogenous in terms of age and time that the samples were taken and contraceptive use and diet and ACTH levels and any of the huge number of additional confounders that cannot be adequately controlled for (but can be expected to zero each other out in a large enough sample size). The focus on contraceptives and hormone cycles may be because they are the most obvious confounders for absolute hormone levels and ratios, but I'd caution against extending that assumption to the network analysis.

And I do agree that the text does make an inadequately supported jump to HPA axis function and whether this dysregulation would actually be central to the pathophysiology of ME/CFS. Could the analysis have taken into account a few more variables and been more careful in the interpretation? Sure. Can the primary finding be explained away by the confounders listed in this thread? I don't think it actually can.

 

[wigglethemouse]

One universal finding in metabolomics studies in ME/CFS is that there is something going on with lipids but it seems we are not quite sure what this is and the reason for it.

Steroid hormones are lipids. If we were to consider the lack of correlation between certain steroid metabolites in ME/CFS is true, could this finding be related to the wider lipid changes that have been found. For example, could there be an issue with lipid transport into cells and would that mess with the lack of correlations?