Trial Report Plasma cell targeting with the anti-CD38 antibody daratumumab in ME/CFS -a clinical pilot study, 2025, Fluge et al

Except this data doesn't support the idea that NK cells and severity correlate. The other two non-responders with low NK cell count were in the moderate group.
I never argued NK cells and severity correlate, my point is you cannot make any conclusion about the severity and response because of the confounding variable.

That’s the whole idea of a confounder
 
That may be the explanation, though I'd expect dara to be considerably slower than rituximab considering the expression of CD38 was usually at least an order of magnitude lower than CD20 in my old data sets.

My understanding is that killing is pretty quick. If it going to kill it has probably done it by the time the drip has come down. If not, it won't.
 
I never argued NK cells and severity correlate, my point is you cannot make any conclusion about the severity and response because of the confounding variable.

That’s the whole idea of a confounder
In order to be a confounder it would need to correlate with both the independent and dependent variable

[edit: might be worth it to drop this thread of discussion anyways, looking at the actual step count and questionnaire data I’m not sure how much the severity labels actually align with the available data anyways]
 
In order to be a confounder it would need to correlate with both the independent and dependent variable

[edit: might be worth it to drop this thread of discussion anyways, looking at the actual step count and questionnaire data I’m not sure how much the severity labels actually align with the available data anyways]
I’ll do a scatter of baseline SF36 vs baseline NK cell count and you should be able to see a minor correlation.
 
Last edited:
1758488633206.png
@jnmaciuch

There is minor trend except for the SF36 baseline 10 patient responder with 270 NK cells that is patient 01.

But anecdotally (from a small group of 5 people I asked, there is no correlation. I am mild and have NK cells 49. There is a severe person with 250)
 
Last edited:
View attachment 28496
@jnmaciuch

There is minor trend except for the SF36 baseline 10 patient responder with 270 NK cells that is patient 01.

But anecdotally (from a small group of 5 people I asked, there is no correlation. I am mild and have NK cells 49. There is a severe person with 250)
Thanks for the plot. So it seems like we can conclude that NK cells at baseline is not a significant confounder for the apparent difference between responder groups in baseline questionnaire scores.

Why the questionnaire scores don’t seem to correlate with step counts at baseline is another source of confusion though. I have a feeling that small sample size is heavily weighing on this data
 
But with a cohort this small, it introduces a seed of doubt if two conditions need to apply.

Unless there's a known link between the functionality of this type of cell and the numbers of them, e.g. more get killed off if they're tagged as poor quality?
Sorry, I missed this earlier. For lymphoid cells, cell “activation” and proliferation frequently go hand in hand—they can both be downstream consequences of the same upstream triggering event. So yes, easy to propose a possible explanation that links cell numbers and functionality if we were convinced that it was NK cell killing that mattered.

Or, alternatively, it could be a signaling pathway involving CD38 that affects many cell types, and just one of its many downstream consequences is affecting NK cell proliferation despite NK cells having little to do with responsiveness. This explanation seems more complicated on paper but in reality, most immune signals have receptors on many different cell types, to slightly different effect in each of those cell types.
 
Thanks for the plot. So it seems like we can conclude that NK cells at baseline is not a significant confounder for the apparent difference between responder groups in baseline questionnaire scores.

Why the questionnaire scores don’t seem to correlate with step counts at baseline is another source of confusion though. I have a feeling that small sample size is heavily weighing on this data
I don’t quite get what you mean, but there is some corr in the scatter and hence there is this confounding

However I do believe this correlations is due to chance due to my anecdotal evidence of polling 5 people in a group I’m in and we see the reverse (the milder guys have lower NK).

Also step counts does correlate with SF36.

1758526304241.png
 
Last edited:
However I do believe this correlations is due to chance due to my anecdotal evidence of polling 5 people in a group I’m in and we see the reverse (the milder guys have lower NK).
We've already had countless studies looking at NK cells in ME/CFS. I think we can be pretty sure if this study would have picked up a meaningful correlation (say between severity and NK cell count) the other studies should have as well, but as far as I remember apart from the "Marshall-Gradisnik story" those studies have not painted a consistent picture.
 
We've already had countless studies looking at NK cells in ME/CFS. I think we can be pretty sure if this study would have picked up a meaningful correlation (say between severity and NK cell count) the other studies should have as well, but as far as I remember apart from the "Marshall-Gradisnik story" those studies have not painted a consistent picture.
Yes, believe only correlation between severity and NK function not count
 
I’m a bit late to some of the detail of this paper but some questions I had which seem relevant to the current discussion

When is the assumption made that ADCC and NK is important (Daratumumab seems to have other mechanisms of action)? Why is there an assumption of autoantibodies being involved?

If we assume there is a feedback loop of some sort with various factors involved and this is only one leg being kicked out, should we focus on that one leg?Correlation is not causation after all so what about the influence of other, non measured factors which may support this loop?

Could the non-responders have been in PEM or had another perpetuating effect and therefore their body wasn't able to ‘recover’ or ‘reset’ during this initial dosing period. For instance how far did the different patients meed to travel? Maybe some had minor infections? Maybe some were just unlucky with which cells were hit and some cells were at different points in their lifecycle so topped up a pool of perpetuating cells? Etc

There seems to be many factors which may be involved here that I’m unsure why NK cells is being focused on other than, well, because it was measured. So I’m just a bit sceptical at this point that we can really know what’s going on given so many unknowns. Building a hypothesis is obviously useful, but thinking about other factors may be important for understanding what happens in this larger trial, especially if the assumptions being made are wrong.
 
This explanation seems more complicated on paper but in reality, most immune signals have receptors on many different cell types, to slightly different effect in each of those cell types.
Presumably this balance matters even for a particular cell type, like in the theory here? It’s not just ‘cell with this receptor gets killed’ it’s a probabilistic thing depending upon receptor makeup, how much daratumumab binds and other factors on deciding if a cell is killed?

Separately to this I was wondering if the balance of activating and inhibitory signals could somehow be a bit off for NK cells in people with ME/CFS, so wondered if this could play into things here.
 
Last edited:
I’m sorry but correlation is not causation is the biggest cop out people use here when you see a strong correlation, saying correlation not causation is just a way of saying I will ignore that correlation because I want to.
 
I’m sorry but correlation is not causation is the biggest cop out people use here when you see a strong correlation, saying correlation not causation is just a way of saying I will ignore that correlation because I want to.
A correlation so strong we have no indication it even passes a very mild multiple testing correction?
 
A correlation so strong we have no indication it even passes a very mild multiple testing correction?
Sorry but what is a multiple testing correction? Correlation is literally the scaled covariance between two variables, any statistical test on 10 data points is useless.
 
I’m sorry but correlation is not causation is the biggest cop out people use here when you see a strong correlation, saying correlation not causation is just a way of saying I will ignore that correlation because I want to.
No, it is an accurate reflection of the way so many people drawing conclusions published in papers we read misuse correlations. All scatter plots and correlation coefficients can show is whether there is an association between two sets of data. Interpretation has to be based on understanding of the nature of that data.
 
Sorry but what is a multiple testing correction? Correlation is literally the scaled covariance between two variables, any statistical test on 10 data points is useless.
We have no idea how many measurements F&M took as part of this trial. I see no reason for correcting for multiple tests here and I'm perfeectly fine with how the findings were reported, but we shouldn't forget that the probability of one measurement showing a correlation purely by chance is probably high.
 
Last edited:
Yes, and I’m saying it’s an easy cop out when you want to disregard the correlation. Just say “correlation not causation” and you’re done!
 
Back
Top Bottom