Headline (one of many): “Long COVID may be triggering Alzheimer’s-like changes in the brain: new study”. Story:
The effects of long COVID rage on.
Over 20 million Americans are believed to have the debilitating post-infection condition, suffering symptoms such as severe fatigue, shortness of breath, chest pain, palpitations, dizziness and muscle pain.
Now, research from NYU Langone Health might explain why some patients experience incessant “brain fog” and memory issues long after a COVID infection.
The researchers propose that long COVID may trigger changes in the brain that resemble the biological processes seen in diseases like Alzheimer’s…
“Physical, molecular and clinical evidence suggests that a larger CP may be an early warning sign of future Alzheimer’s-like cognitive decline,” added Ge…
Rage on is what I do about bad statistics. Doesn’t help me either.
Anyway, the peer-reviewed paper is “Choroid plexus alterations in long COVID and their associations with Alzheimer’s disease risks” by Pang and others in Alzheimers & Dementia.
The idea of “long covid” is disputable. The symptoms are vague, often relying on self-report, and there are no agreed upon set of biomarkers. For instance, see this paper: “How methodological pitfalls have created widespread misunderstanding about long COVID” or this article “Flawed body of research indicates true ‘long covid’ risk likely exaggerated.”
In this study, those who had “long covid” were identified by “neurological complaints temporally related to the SARS CoV2 infection”. Complaints. So it seems self-report.
Our crew gathered 86 “long covid” patients, 67 with some evidence of recent but cleared infections, and a mere 26 who didn’t show evidence of being infected recently. These were their “healthy controls.” About two thirds (in each category) were women.
Among other things, people were given the “Mini-Mental” exam then hooked them up to an MRI and took models of the images. Gaussian something or others. Most forget, or don’t know, what they’re seeing on MRI pics are model output.
The Mini-Mental is a proven reliable guide to mental states, but only in those suffering strokes and the like. A woman who is stroking out won’t be able to answer even the first question “What is the (year) (season) (date) (day) (month)?”, let alone be able to draw the somewhat complex image the test gives you. A person who had covid three months ago, or even “long covid”, ought to score perfectly (see for yourself at the link). Unless they are being dramatic.
So what we this test supposed to prove?
The three groups were not even close to the same health-wise. Here’s their Table 1.
I couldn’t fit the diabetes result for some reason, but the percents were 15%, 6%, and 4%.
Look how much the “long covid” drinks and smokes. Look how fat they are. Look at their high blood pressure and diabetes rates. Not surprisingly, the healthy group is healthiest. What does it mean?
This—drumroll—any outcomes which are correlated with ill health will show the non-healthy groups as looking, well, non-healthy. And the healthy group will look healthy.
How would you know the cause of some, and only some, slight difference in scores, like the Mini Mental, was because of “covid” given these facts? Well, you cannot.
They gave their groups Neuropsychological batteries, and couldn’t really get the Ps wee enough between the groups. Meaning “no real difference.”
Pictures like this were given:
The CBP is blood flow, and ChP volume is “Choroid plexus (ChP) enlargement is a neuroimaging biomarker of neuroinflammation and neurodegeneration.”
The HC are healthy controls. A very small sample: recall small samples are much more variable than large (a fundamental result in stats). Which means it doesn’t look like there’s much going on in the way of differences between the covid groups, except maybe for the HC groups. But its small sample sort of kills the idea of knowing any real difference, especially for blood flow (the second pic).
Point is, even supposing these differences are real, and that most don’t get “long covid”, the covid sufferers group will eventually become proud members of the healthy control, so there’s nothing really to worry about. Unless “long covid” turns out to be a real thing. Perhaps. But we still have better indictions that other causes are responsible (drinking, smoking, hypertension, diabetes).
Finally are pictures like this. These are why I am doing the Class:
It does really matter to us what these are, but they are outputs of the MRI model (the volume or blood flow) with other bio-measures. They think they have found that the markers are well predicted by the volume, and have drawn regressions with the little gray shaded regions meaning “95% confidence intervals.” But you can see there is little confidence 95% of the points fall in those grey windows. Barely any do.
I have drawn, masterfully, red lines over the areas where roughly 95% of the observables lie.
The dots, remember, are real people. We want models to predict real observables (people). Take, say, the upper right figure. Put your finger on the figure at a CBF of 40. Then move your figure-finger between the red lines. Given the model and data, and whatever other assumptions the researchers used, there’s roughly a 95% chance the TNF (whatever that is) of some new person lies between those red lines for people with CBP = 40.
Which is..? Right. Almost every possible value of TNF. The models’ predictive ability of actual observables is weak sauce.
(This is the difference between parametric and predictive uncertainty. See this Class.)
I wouldn’t have bothered myself, let alone you, dear readers, with any of this, because it’s just not that interesting. But since it made headlines, as these things do, I wanted you to learn how to investigate the claims behind the headlines. Most are not worth worrying about.
Here are the various ways to support this work:
- Subscribe at Substack (paid or free)
- Cash App: $WilliamMBriggs
- Zelle: use email: matt@wmbriggs.com
- Buy me a coffee
- Paypal
- Other credit card subscription or single donations
- Hire me
- Subscribe at YouTube
- PASS POSTS ON TO OTHERS
Discover more from William M. Briggs
Subscribe to get the latest posts sent to your email.
