[gibba999]

I think a lot of this comes down to replication, effect size, and sample size.

Yes, there are studies which show virtually everything, but in this case, we have:

- >10% of mild cases reporting long COVID brain fog (without MRIs)

- Visible correlations on MRIs with large n (cited study)

- Lots of small-scale studies / looking at specific cases

- Some understanding of a relevant mechanism-of-action (see: olfactory loss)

Together, that's about as strong evidence as you'd expect after 15 months. We have effect, we have correlation, we have case studies, and we understand why it's plausible.

The big question is whether it strikes vaccinated mild / asymptomatic cases. We don't know. There are a lot of cases like this.

[timr]

> >10% of mild cases reporting long COVID brain fog (without MRIs)

"Brain fog" is not a diagnosis. It has no definition. It has no test. Literally anyone could say they have it, and not be wrong.

It also overlaps substantially with "fatigue"...which we all know comes along with a lot of other common issues. Such as depression.

> Visible correlations on MRIs with large n (cited study)

The size of n doesn't matter if the thing you're reporting is not a meaningful metric. Here, we have a paper that has gone on a fishing expedition for a quasi-subjective metric with unknown levels of noise, which is widely "shown" to be associated with many common and uncommon issues across the research literature.

This is a low-quality data set. But yes, it is a larger low-quality data set.

> Lots of small-scale studies / looking at specific cases

Collections of anecdotes are not data.

> Some understanding of a relevant mechanism-of-action (see: olfactory loss)

...for a single symptom (loss of smell). But no, we don't know why that happens, and to the extent we do, the current best hypothesis has nothing to do with neurons, but rather, the scaffolding around those neurons.

> Together, that's about as strong evidence as you'd expect after 15 months.

Nonsense. We've been debating this "long covid" for more than a year now. There are apparently many sufferers. We could have easily conducted randomized, longitudinal, controlled trials. We have not.

The total evidence for "long covid" continues to be anecdotes and self-reported "symptoms", of indeterminate duration, amongst populations that are mostly self-selected for having "long covid". I believe that we'll eventually find out that some of these things are real, but right now, this is just hysteria.

[gibba999]

My hypothesis is that COVID leads to a Gaussian distribution in reduction in general neurological function, and that we will see a bell curve distribution reduction in e.g. general IQ.

Can you please propose a "randomized, longitudinal, controlled trials" one might conduct to figure that out?

Preferably, one which would pass an IRB review. We can't randomly infect 10,000 ethnic minorities with COVID19 anymore, which I think what you're suggesting. The Tuskegee Syphilis Study and the Nuremberg Trials took care of that for us.

Short of something like that, we work from mixed methods evidence.

As a footnote, a year isn't a long time in the world of research. That's sometimes quite literally how long it takes from when you apply for a grant to when funding lands in your account. And you're asking about a phenomenon which often occurs months later.

[timr]

> Can you please propose a "randomized, longitudinal, controlled trials" one might conduct to figure that out?

This is not a herculean problem. It's essentially the definition of any halfway decent medical study:

* pick a set of measurable endpoints from the pantheon of "long covid" symptoms that are likely to be real. Objectively measurable endpoints should be mixed in with subjective ones (e.g. "fatigue", "loss of smell", "reduced lung capacity", "heart inflammation").

* pre-register these endpoints, so that you can't go back on a fishing expedition later, when your first choices don't pan out.

* pick a set of participants at random (balancing for demographics of interest: age, co-morbidities, weight, gender, etc.)

* measure those endpoints at the start of the study so that you have a pre-trial baseline.

* follow those people over time for the endpoints of interest.

* some percentage will get infected with SARS-CoV2. verify this via testing.

* at the end of the study, compare the group that caught SARS-CoV2 with the ones who did not along the endpoints of interest. compare both groups with their own pre-trial baselines.

This is not the ONLY way of doing such a study, but it would be vastly better than any data currently reported. The biggest challenge is that it has to be done before the pandemic passes, and the number of cases drops too low to get a significant result in a reasonable period of time. The window on this is rapidly closing.

> As a footnote, a year isn't a long time in the world of research. That's sometimes quite literally how long it takes from when you apply for a grant to when funding lands in your account. And you're asking about a phenomenon which often occurs months later.

There have been multiple RCTs conducted during the pandemic, despite the bureaucratic inertia of the academy, and "long covid" is one of the biggest remaining controversies. There's no universe in which you couldn't get funding and approval for such a study in short order.

[gibba999]

>> Can you please propose a "randomized, longitudinal, controlled trials" one might conduct to figure that out?

> This is not a herculean problem. It's essentially the definition of any halfway decent medical study:

You're running in circles. That's not a randomized control trial. You'll get biases since the set of people infected with COVID19 isn't random.

This is not much better than existing studies. They're not preregistered, but that's the only upside of your methodology.

> There have been multiple RCTs conducted during the pandemic, despite the bureaucratic inertia of the academy, and "long covid" is one of the biggest remaining controversies. There's no universe in which you couldn't get funding and approval for such a study in short order.

IRBs are set up to prevent subject harm. An RCT, in this case, would involve randomly infecting people with COVID19 to eliminate the bias above. That will never fly.