Safety, tolerability and clinical effects of... BC007 on fatigue and quality of life in patients with Post-COVID syndrome, 2025, Hohberger+

EndME

Senior Member (Voting Rights)

Safety, tolerability and clinical effects of rovunaptabin, also known as BC007 on fatigue and quality of life in patients with Post-COVID syndrome (reCOVer): a prospective, exploratory, placebo-controlled, double-blind, randomised phase IIa clinical trial (RCT)


Background

Rovunaptabin neutralises functional autoantibodies targeting G-Protein coupled receptors (GPCR-fAAbs), observed in patients with Post-COVID syndrome. As we hypothesise an improvement of PCS by rovunaptabin, the aim of reCOVer was to investigate safety, tolerability, and clinical effects of rovunaptabin in PCS patients.

Methods

reCOVer is a prospective, exploratory, placebo-controlled, double-blind, randomised phase IIa clinical investigator initiated trial with 1350 mg rovunaptabin with additional cross-over at the Universitätsklinikum Erlangen, Germany. The trial was registered in EudraCT, 2022-001781-35. Screening was done between 21·11·2023 and 25·06·2024. Eligible participants (18–80 years) showed GPCR-fAAbs, at least 3/8 defined PCS symptoms persisting ≥3 months after COVID-19 and fatigue as major symptom. Participants were randomly assigned (1:1) to either receive rovunaptabin or placebo at day 0 (d0) and d48 with a follow-up of 28 days, respectively. Primary endpoint was the number of treatment emergent adverse events (TEAE) at d28 (co-primary endpoint: TEAE at d70); secondary endpoint focused on fatigue and quality of life.

Findings

Thirty PCS patients were randomised and analysed. RCT analysis showed nine (rovunaptabin) and five TEAEs (placebo), yet without statistically significance (p = 0·1299; CI −14·80%; 63·02%); one serious adverse event, not related to treatment, was recorded. Rovunaptabin showed a neutralisation of GPCR-fAAb and a significant improvement of FACIT Fatigue Scale (effect size = 2·10, p = 0·0378), Bell score (effect size = 3·64, p = 0·0004), Fatigue Severity Scale (effect size = −2·66, p = 0·0088), and quality of life (4/8 items).

Interpretation

As this proof-of-concept study showed effects on the patient-centred endpoint PCS and a good safety profil, subsequent studies are needed to confirm these results in a larger cohort.

 
FACIT is 0 to 52, bell is 0 to 100, these effect sizes are tiny and well within the normal variance of the disease and I suspect quite likely below the usual threshold for a detectable change from the questionaries themselves. I saw the same thing being done on some studies with Chalder fatigue studies as well where they hoped we didn't know what the threshold of the questionaires was, they were P hacking to find significance and that was how they did it, this looks to be the case here as well.

The other much larger BC007 study failed. I see no reason to think another BC007 will succeed, they are depleting the GCPR antibodies and its not helping.
 
IMG_0262.png
The improvement of fatigue was apparent by the FACIT fatigue scale, Bell scale, FSS, and CCC, yet not by the Chalder fatigue scale, DSQ-PEM, and Borg scales. This data demonstrates clear differences between the different fatigue questionnaires. In addition, patients did not show a significant increase of their walking distance, as measured by the 6MWT.
So they got some minor improvement on some of the scales, but nothing in the only objective outcome measure related to functioning.
 
View attachment 27263

So they got some minor improvement on some of the scales, but nothing in the only objective outcome measure related to functioning.
None of these graphs look impressive and are surely all clinically irrelevant, despite the section bias of only showing those things that even had a positive result.

Berlin Cures has to be the happiest pharmaceutical company of all time. Not a dime invested into drug development, patients are happy to finance your studies and patients and researchers will do all the marketing for you despite trials not showing anything meaningful.
 
A Reddit user's criticism of the study:

Link to Reddit

Copying the main bulletpoint headings:

1. The reported improvements, while statistically significant, are small and clinically of no significance to patients (MCID not met)

2. The primary endpoint of the study was safety (TEAEs), not efficacy. Nevertheless, the supposed effectiveness is strongly concluded and the TEAEs are only mentioned slightly.

3. In the side effect analysis / AE analysis, there is also a lack of transparency and causal assessment.

4. Multiple testing and selection bias were not taken into account.

5. Not all outcomes or endpoints were listed in the study, which should actually be the content of the study.
 
A Reddit user's criticism of the study:

Link to Reddit

Copying the main bulletpoint headings:

1. The reported improvements, while statistically significant, are small and clinically of no significance to patients (MCID not met)

2. The primary endpoint of the study was safety (TEAEs), not efficacy. Nevertheless, the supposed effectiveness is strongly concluded and the TEAEs are only mentioned slightly.

3. In the side effect analysis / AE analysis, there is also a lack of transparency and causal assessment.

4. Multiple testing and selection bias were not taken into account.

5. Not all outcomes or endpoints were listed in the study, which should actually be the content of the study.
MCID is not meaningful if the given theory only applies to a subset of the patients.

We know that the blood test, even the functional one is lacking. There's a new one in development based on pluripotent stem cells and I hope that we will also get blood measurements with proper controls in the future.

There are simply too many unknowns regarding the blood test, for example the threshold for the detection is quite high, compared to elisa. It could also be that you need several types of auto antibodies that interplay with each other, like The combination ß2 M2. One point in case for this theory is that COVID actually can or does induce these auto antibodies in cattle and ferrets.

It could also be that you need the medicine regularly or the dose would need to be higher. Also, serum measurement can also be not ideal, since it does not necessarily directly relate to levels in tissue for example. This has been evident in car t research.

The critique about other outcomes like OCT measurement is also not well-founded in my opinion, this first result was rushed to get it out after the bankruptcy of berlin cures. However, proper analyzes and writing papers takes time. Everyone who has worked in science knows this. I would be surprised if we do not get The data of measurements like OCT in the future.

Regardless of what everyone thinks of this critique, I would suggest everyone that is interested to actually read it in detail, since the main bullet points are not a good representation in my opinion.

I would also add that the Reddit user has been part of the study and did not see a benefit.

Meanwhile one interesting thing that I find with BC 007 is that it seems to neutralize the Auto antibodies despite having a very short half life of a few minutes only. One would expect the auto antibodies to be replenished sooner or later. As already mentioned, one cannot completely trust the blood test due to the detection efficiency, but it's still quite interesting. If this is confirmed, it hints to some unknown mechanism, and if it really gets rid of nearly all antibodies, this may be interesting for other diseases as well. In the hypothetical case that it gets rid of the auto antibodies completely, including in tissue, it would have a similar effect as CAR T.
 
None of these graphs look impressive and are surely all clinically irrelevant, despite the section bias of only showing those things that even had a positive result.

Berlin Cures has to be the happiest pharmaceutical company of all time. Not a dime invested into drug development, patients are happy to finance your studies and patients and researchers will do all the marketing for you despite trials not showing anything meaningful.
Second point is totally wrong, private investments of BC have been in The double digits of millions for their phase 2 alone.
 
Hello there,

I am the reddit user who made this criticism.
Decided to register here to get in touch with you about this. And would be happy to hear your comments on my criticism…

I wrote it in first line because of the linked press releases from UK Erlangen two days ago. The claims made there are in my opinion not backed up by the provided data and are giving a lot of patients false hopes.

I have no medical background as i am an engineer but tried my best to get some points together…

A Reddit user's criticism of the study:

Link to Reddit

Copying the main bulletpoint headings:

1. The reported improvements, while statistically significant, are small and clinically of no significance to patients (MCID not met)

2. The primary endpoint of the study was safety (TEAEs), not efficacy. Nevertheless, the supposed effectiveness is strongly concluded and the TEAEs are only mentioned slightly.

3. In the side effect analysis / AE analysis, there is also a lack of transparency and causal assessment.

4. Multiple testing and selection bias were not taken into account.

5. Not all outcomes or endpoints were listed in the study, which should actually be the content of the study.
 
Second point is totally wrong, private investments of BC have been in The double digits of millions for their phase 2 alone.
Sorry, I was refering to pre-clinical research etc, I should have been clearer. They indeed had to invest millions to conduct a phase 2 study.

MCID is not meaningful if the given theory only applies to a subset of the patients.
Of course it is a meaningful argument, because the authors provide no arguments for subgroups or evidence for it. You either have to have a sufficient sample size (and the much larger study had negative results on precisely these outcomes so this argument becomes less relevant rather than the opposite) or devise a mechanistic understanding what the subgroups are. Otherwise you just always end up talking your way out by referencing subgroups "GET only works for subgroups, homeopathy only works for subgroups...." Said plainly: You cannot argue for efficacy but at same time say there can't be efficacy because of subgroups. There may not be efficacy due to subgroups, but the result is then that there is no efficacy in the population you studied, not the opposite!

MCID is not meaningful if the given theory only applies to a subset of the patients.

We know that the blood test, even the functional one is lacking. There's a new one in development based on pluripotent stem cells and I hope that we will also get blood measurements with proper controls in the future.

There are simply too many unknowns regarding the blood test, for example the threshold for the detection is quite high, compared to elisa. It could also be that you need several types of auto antibodies that interplay with each other, like The combination ß2 M2. One point in case for this theory is that COVID actually can or does induce these auto antibodies in cattle and ferrets.
These are all mute points. There's no evidence to suggest that these antibodies play a role. You cannot argue that that is because tests are ineffective because then you have no evidence in the first place (and of course ELISA is known to be able to differentiate between the antibody levels of people with an autoimmune disease caused by antibodies and healthy controls for several decades, which is largely what the narrative has been). It's like saying: "ME/CFS is caused by particle x based on my results, but my results don't show it because I can't measure particle x". Perhaps one may then argue that more research such as the things being done by Dmitry Veprintsev should have been conducted, rather than what we've seen.


The critique about other outcomes like OCT measurement is also not well-founded in my opinion, this first result was rushed to get it out after the bankruptcy of berlin cures. However, proper analyzes and writing papers takes time. Everyone who has worked in science knows this. I would be surprised if we do not get The data of measurements like OCT in the future.
Selective reporting of outcome measures is seems like bad practice and I think should be heavily critiqued as it has been elsewhere. Especially when the authors mention explicitly mention microcirculation in the paper and hypothesize about the role of microcirculation as part of the paper and cite previous work done on one person, but don't mention the actual data from the study. It would have been sufficient to simply include the following in the supplementary material: "Other outcome measures are still being analysed and will be part of upcoming publications."

I would also add that the Reddit user has been part of the study and did not see a benefit.
Sure. That will have influenced them to write this, but shouldn't impact their arguments.
 
Last edited:
Sorry, I was refering to pre-clinical research etc, I should have been clearer. They indeed had to invest millions to conduct a phase 2 study.


Of course it is a meaningful argument, because the authors provide no arguments for subgroups or evidence for it. You either have to have a sufficient sample size (and the much larger study had negative results on precisely these outcomes so this argument becomes less relevant rather than the opposite) or devise a mechanistic understanding what the subgroups are. Otherwise you just always end up talking your way out by referencing subgroups "GET only works for subgroups, homeopathy only works for subgroups...." Said plainly: You cannot argue for efficacy but at same time say there can't be efficacy because of subgroups. There may not be efficacy due to subgroups, but the result is then that there is no efficacy in the population you studied, not the opposite!


These are all mute points. There's no evidence to suggest that these antibodies play a role. You cannot argue that that is because tests are ineffective because then you have no evidence in the first place (and of course ELISA is known to be able to differentiate between the antibody levels of people with an autoimmune disease caused by antibodies and healthy controls for several decades, which is largely what the narrative has been). It's like saying: "ME/CFS is caused by particle x based on my results, but my results don't show it because I can't measure particle x". Perhaps one may then argue that more research such as the things being done by Dmitry Veprintsev should have been conducted, rather than what we've seen.



Selective reporting of outcome measures is seems like bad practice and I think should be heavily critiqued as it has been elsewhere. Especially when the authors mention explicitly mention microcirculation in the paper and hypothesize about the role of microcirculation as part of the paper and cite previous work done on one person, but don't mention the actual data from the study. It would have been sufficient to simply include the following in the supplementary material: "Other outcome measures are still being analysed and will be part of upcoming publications."


Sure. That will have influenced them to write this, but shouldn't impact their arguments
Unfortunately I cannot answer in a detail level (blood tests) which would be needed since I don't agree.
I'm sorry, but my illness is limiting me too much and I'm already writing too much again (sorry this is really not a bland excuse, believe me).

Everyone has his or his own favorite theory, and I think I've been enough in science to judge it. My impression is that also this forum has its personal favorites, eg linked to the work of clinicians which are part of this forum. Which is completely fine and reasonable.
I just think that this research is being down talked too much.
At least if you compare it to other research. I'm not talking about the quality of the research, but more if it's worth to follow the cues and invest more into this topic.

I agree that they have been made false promises with the results of this BC 007 study. However, I still think it's one of the most evidenced theories combining all The little cues. Yes, evidence is still low, but for nearly everything else it's a lot lower. So Im comparing in a relative way.

Regarding your first point, I think we have to wait and see till this new blood test is out and till we finally get proper measurements with a larger control group. And then we can continue discussing.
You also can't compare the phase two, since there are far too many unknowns like different outcomes.
It is also not unusual for a study to fail its phase two or phase 3, while still making it to market with another new study. Just look at the new fibro drug.

The critique said that the data leads to the conclusion that the treatment did not offer any clinical benefit to the patients. That conclusion cannot be drawn, since it assumes that the patient's group is homogeneous. But that is not true if a subgroup only suffers from this root cause of autoimmunity. So the conclusion can simply not be made. I'm also not arguing that we have seen a significant effect of the treatment, we need to wait and see for further research. I'm just saying that the claim in the critique cannot be made.

I agree they could have written a sentence that everything like OCT is following up, but I think this is more like a nothing burger.
 
Everyone has his or his own favorite theory, and I think I've been enough in science to judge it. My impression is that also this forum has its personal favorites, eg linked to the work of clinicians which are part of this forum. Which is completely fine and reasonable.

I don't see the forum as having favourite theories. I don't know of any clinicians on the forum working on ME/CFS so I am not sure what that refers to. I certainly don't have a favourite theory. But I do have views on theories that don't add up.

The idea that anti-GCPR cause ME/CFS or LC symptoms does not add up. If they did then certain levels of antibodies would pretty much always be associated with symptoms.

So if you compared antibodies above level L in 100 patients and 100 controls you would get a figure something like:

Patients +ve: 37
Controls +ve: zero, or maybe 2 if you allow for some technical problems.

The actual findings are more like:

Patients +ve: 37
Controls +ve: 33

Although some statistically significant differences have been reported, that is not what you need to make a theory viable. If X causes Y then no healthy people without Y will have X.

It is true that autoantibodies are a bit more complicated because our tests have never been that good at picking out pathogenic antibodies from antibodies that bind similarly in tests but don't cause problems. So things are messier. The control rate above could be as much as 7, but in diseases where we have other evidence for antibodies being pathogenic rates in patients are at least five times greater than in controls, especially if you choose a high threshold level. the data we have on anti-GPCR antibodies say firmly that these are not likely to be pathogenic. I have never seen control data on other randomly sampled autoantibodies in these patients but for all sorts of reasons you are likely to get a statistically significant difference for just about any auto-antibody in a group of ill people compared to controls.

And as far as I can see the 'blocking' power of the drug is based on the same assays that perform so poorly on discriminating patients.
 
I don't see the forum as having favourite theories. I don't know of any clinicians on the forum working on ME/CFS so I am not sure what that refers to. I certainly don't have a favourite theory. But I do have views on theories that don't add up.

The idea that anti-GCPR cause ME/CFS or LC symptoms does not add up. If they did then certain levels of antibodies would pretty much always be associated with symptoms.

So if you compared antibodies above level L in 100 patients and 100 controls you would get a figure something like:

Patients +ve: 37
Controls +ve: zero, or maybe 2 if you allow for some technical problems.

The actual findings are more like:

Patients +ve: 37
Controls +ve: 33

Although some statistically significant differences have been reported, that is not what you need to make a theory viable. If X causes Y then no healthy people without Y will have X.

It is true that autoantibodies are a bit more complicated because our tests have never been that good at picking out pathogenic antibodies from antibodies that bind similarly in tests but don't cause problems. So things are messier. The control rate above could be as much as 7, but in diseases where we have other evidence for antibodies being pathogenic rates in patients are at least five times greater than in controls, especially if you choose a high threshold level. the data we have on anti-GPCR antibodies say firmly that these are not likely to be pathogenic. I have never seen control data on other randomly sampled autoantibodies in these patients but for all sorts of reasons you are likely to get a statistically significant difference for just about any auto-antibody in a group of ill people compared to controls.

And as far as I can see the 'blocking' power of the drug is based on the same assays that perform so poorly on discriminating patients.
Maybe the term clinicians was too precise, I mean people involved in studies like decode me.
But also former clinicians with experience in the field, like you.
But that is simply my personal impression and it is also totally fine.

Regarding the blood test, there are just too many unknowns.
Elisa by different private labs, functional rat tests, in the future pluripotent stem cells.
It could be that only a combination of AABs is pathogenic, or that a factor x needs to be in play which is also hypothesized.

I just think giving the current data one cannot say that these auto antibodies are likely not pathogenic. We simply don't know yet.

And since there is a lot of additional data in the pipeline, like OCT, blood cell deformability, MRT data, The new functional stem cell test I think it's better to wait.

Could this whole topic have been researched a lot better? Definitely yes. Like having a proper study with controls and the functional rat-based test. But in this field of science you have to live with many circumstances like very limited funds and personal, technical limitations like not being able to do the blood test by yourself as a lab, The politics of funding and probably many more not public obstacles.
When The first results of the study were announced on a conference, it was quite clear to me that these were very rushed due to the bankruptcy of Berlin cures. that should simply be kept in mind.

edit
Also I think the evidence that we have for pathogenic auto antibodies in other diseases is mostly related to inflammatory autoimmune diseases. Which is very different from this theory of auto antibodies.
So there could be other factors in play, I just doubt that one can simply carry over the experience from inflammatory autoimmune diseases.
 
Last edited:
And as far as I can see the 'blocking' power of the drug is based on the same assays that perform so poorly on discriminating patients.
I think for the functional test we do not have any control data except for two persons which were positive.
Of course that is not good.
Though as weird as it sounds we cant say it performs poorly in discriminating, because we basically have never tested it.

I think what you refer to has been made with cell trend Elisa. A lab which I personally don't trust at all, after the debacle of the pots study with all controls being threshold positive.

This will be my last post for longer time, I need to rest now and take a step back im sorry.
 
Back
Top Bottom