Acute blood biomarker profiles predict cognitive deficits 6 and 12 months after COVID-19 hospitalization, 2023, Taquet et al

Acute blood biomarker profiles predict cognitive deficits 6 and 12 months after COVID-19 hospitalization

Abstract
Post-COVID cognitive deficits, including ‘brain fog’, are clinically complex, with both objective and subjective components. They are common and debilitating, and can affect the ability to work, yet their biological underpinnings remain unknown.

In this prospective cohort study of 1,837 adults hospitalized with COVID-19, we identified two distinct biomarker profiles measured during the acute admission, which predict cognitive outcomes 6 and 12 months after COVID-19. A first profile links elevated fibrinogen relative to C-reactive protein with both objective and subjective cognitive deficits. A second profile links elevated D-dimer relative to C-reactive protein with subjective cognitive deficits and occupational impact. This second profile was mediated by fatigue and shortness of breath. Neither profile was significantly mediated by depression or anxiety.

Results were robust across secondary analyses. They were replicated, and their specificity to COVID-19 tested, in a large-scale electronic health records dataset. These findings provide insights into the heterogeneous biology of post-COVID cognitive deficits.


https://www.nature.com/articles/s41591-023-02525-y

 

There is quite a long article on this study by the BBC here: https://www.bbc.co.uk/news/health-66658257

 

A lot oft the big newspaper agencies have written something about the study. Science also wrote a piece https://www.science.org/content/article/clotting-proteins-linked-long-covid-s-brain-fog.

 

My first impression looking through this is it looks quite good. For n= 1,837 acute COVID-19 patients they have measured a handful of blood markers including fibrinogen, D-dimer and 4 or 5 others while they were in hospital. Then half a year later used a cognitive function test (MoCA) and a cognitive symptom severity survey (C-PSQ) (I don't know either of these), thereby providing both and objective and subjective measure of cognitive function so they claim.

The analysis they are using is Canonical Correlation Analysis (CCA). This takes the two separate sets of variables - the blood markers and the cognitive symptom assessments - and seeks to combine the variables in each of the two sets in different ratios such that when you correlate the resultant combined variables from the blood and the cognitive tests you get as high a correlation as possible (that might not be very intuitive but hopefully will become clear). There's many ways to combine the sets of variables of course so it seems they multiple test correct using some kind of permutation based method (random subsampling to see if the p values are robust).

From this they find two 'profiles' (combinations of the variables) which come out significant. The first profile is described by having a high positive weighting of fibrinogen and a high negative weighting of CRP as described in figure 2 (they show a similar thing for a second profile defined by high D-dimer):

What this is describing is that a blood profile where fibrinogen is weighted positively and CRP weighted negatively is highly correlated with a cognition profile with high subjective symptom severity (CPSQ) and low objective cognitive functioning (MoCA). It would be nice to see a 2D scatter plot where we can see the blood profile and symptom profile correlation on the two axes so we can get a feel with our own eyes what the relationship is but they don't seem to have shown that. Similarly it would have been nice if they could simply take the ratio of fibrinogen to CRP and plot that against the symptom scores directly - and the relationship should you'd hope still hold even without being in the context of this model.

What's nice is that they replicate it in a different way on another cohort derived from a big dataset of hospitalised patients across many hospitals, which ends up being a similar cohort size as before after they filter it (n=1,276). This is done by filtering the data for those with 'normal' CRP levels and then comparing patients with 'high' fibrinogen to those with 'low' fibrinogen to assess the risk of them having cognitive problems 6 months later as determined by various ICD diagnostic codes:

This is showing you have an increased probability of having a 'cognitive dysfunction' outcome over time with high fibrinogen or high D-dimer respectively. Can't help but notice that p value for the fibrinogen plot is pretty high given such a big sample size, which I guess would imply a tiny effect size. They also haven't multiple test corrected for these and they say in the text they tried doing these analyses in multiple different configurations.

 

I had a MoCA at the NIH. It's like a 15-minute or so mini-IQ with an embedded drawing task, as I recall.

 

There are plenty of PwME on anticoagulation therapies - would they not have seen an improvement in their cognitive function if what this study says is true?

 

That's the little known very delayed recall component of the test and how you get to score 31/30

 

I haven't read the paper yet, but it sounds as if the study was looking at fibrinogen and d-Dimer vs CRP during the acute illness. That might be a marker or surrogate of some underlying derangement that was associated with cognitive dysfunction 6-12 months later.

 

I thought I'd previously posted this but hadn't. See review article in thread Fibrinogen in neurological diseases: mechanisms, imaging and therapeutics (2018)

 

A critique of this study was published by Paul Garner, Alan Carson and others (but also George Davey Smith). They write:

There are problems with the methodology in this study. Firstly, the authors use levels of CRP relative to other biomarkers to make a large number of possible ratio traits. This use of ratio traits, in which the same denominator is used (CRP), puts the analysis at risk of collider bias [8]. The lower levels of CRP relative to fibrinogen and Ddimer may relate to other processes, such as less detection by individuals of symptoms of infection, and thus, a lower relative CRP as a denominator in their 2 measures could relate to greater or lesser detection or recognition of symptoms. Problems of using ratio traits have recently been discussed elsewhere [9].

The authors attempt to provide a mechanistic biological justification of the association and, by discussing the “mediation” of effects, suggest that there may be a causal link. However, it is more plausible that fibrinogen and D-dimer are simply markers of inflammation. The other flaw is that the authors suggest the associations between biomarker profiles and subjective cognitive deficits cannot be explained by pre-COVID-19 cognitive function. This is not epidemiologically sound, as the premorbid cognitive ability assessment was by retrospective reporting at the 6-monthly follow-up visit. Thus, there is high likelihood of recall bias.

Further, the patients in this study were acutely unwell, and thus, acute cerebral ischemic events with acute COVID-19 pneumonia would produce inflammatory changes. This is neither mentioned, nor did the authors acknowledge that most patients with PCC had mild COVID-19.
​

Source: https://www.sciencedirect.com/science/article/pii/S2475037924001316

 

Worth noting is that the first author’s affiliation is King’s Healthcare Partners, which run a CFS clinic that is part of the NHS’ South London and Maudsley Trust (where Prof Wessely is based) and have recently hosted a conference on FND with 200 attending clinicians where they touted GET and CBT for ME/CFS:

https://twitter.com/user/status/1773073013757342121