Will the ‘rona rot your brain?
I’m triple vaccinated but at the end of February 2022 caught COVID-19. It was almost definitely Omicron, which was the dominant strain in Melbourne at the time. My symptoms were mild and I considered this my fourth booster. When I told my ever worried mother, I received a flurry of concern and articles purporting that COVID-19 causes brain damage. I like my brain, and I had gifted the virus to my partner and 4 month old, so I was initially concerned about these reports.
The headlines were powerful:
“New study reveals what mild Covid really does to your brain”
and
“Gray matter brain shrinkage revealed in new study shows long term damage from COVID”
The opening sentence from news.com was even more terrifying.
People who have suffered mild Covid-19 are likely to have “significant” long-term effects on their brains, a new peer-reviewed study has found.
Now, although I’m not an expert in the field of virology or epidemiology, I can read a scientific paper. As someone who has published and peer-reviewed my fair share of medical research, I know how to assess the limitations of study design and causal inference in retrospective research. I scoured these articles to find the original research paper and delved into it.
What does the actual analysis tell us about whether SARS-CoV-2 causes brain damage. The answer is… nothing.
The original paper is phenomenal work with a super interesting analysis. If you care about credentialling, then it’s worth noting that it’s a team from Oxford, published in Nature. This is peak academia. The real question is, what does the actual analysis tell us about whether SARS-CoV-2 causes brain damage. The answer is… nothing. This paper does not assess for causation, it simply shows a statistically significant correlation. The importance of study design in assessing causation cannot be understated.
Study Design
The devil is in the detail. Without turning this into an epidemiology lecture, if you really want to assess true causation then, unfortunately, you need to perform a randomized controlled trial (RCT). This is the gold standard of causal inference. Failing that you need to accept that you are performing an observational (not causal) study and apply statistical methodology to try and approximate an RCT in the limit. So maybe you do a prospective cohort study and collect data for as many variables, that you can think of, which could affect the outcome. Worse still you do a retrospective cohort study. Or god forbid, you could do a case-control study. Then maybe you can perform statistical tests that control for the confounding effect of those variables. In all cases, the problem is the unknown unknowns.
As a concrete example, we know there is an association between low vitamin D levels and COVID-19 severity. The Nature paper did not have data on the Vitamin D levels of participants to include as a covariate. It is possible that the effect observed was due to an unmeasured deficiency in the COVID cohort. No statistical test can correct this (now known) unknown unknown.
The study in the Nature paper is a case-control design based on UK BioBank data, where a ‘case’ was someone who had two brain scans and COVID-19 in the intervening period. The controls also had to have 2 scans (obviously) and were matched to the cases based on “age, sex, ethnicity and time elapsed between the two scans.” The two ‘arms’ of the study were also compared across many dimensions to check for any differences. It’s all fairly robust, but ultimately, limited in proving causation. The Oxford team did the absolute best you can do to massage a case-control design into a robust statement of effect but it’s just not enough to make a claim of causation. The title of their paper is reserved, stating “associated” and not “caused” for that very reason.
But, to my absolute lack of surprise, the clickbait media ran rampant with the association, and the language slips. It’s not fun to say things are associated unless it’s JFK and the mob.
Cohort
The second main gripe I have with the reporting, is they claim this finding applied to absolutely everybody with mild COVID-19. Oi vey. The Nature paper examines the effect in a very narrow population. Crucially, the authors were not scanning people to look for this effect. This is an opportunistic study on people already getting brain MRIs as part of their medical treatment. The age range of the participants themselves was 51 to 81. So this cohort is mostly elderly and comorbid enough to be needing at least 2 brain scans during the study period. That is definitely not your average person.
Given these major limitations, the most you can conclude given the study findings is that in elderly people who need serial brain imaging to manage their chronic illnesses, a COVID-19 infection is associated with a more rapid decline in certain metrics of brain health on imaging and cognition. I don’t think that would go viral though.
Novelty of the Findings
The final irritating claim implicit in the clickbait media is that this is some new feature in viral pathogenicity that we need to be scared about. This is just not true. A recent review on cognitive decline post a viral infection makes this abundantly clear. Influenza hastens cognitive decline after infection and even the common cold (also caused by coronaviruses) has effects on cognition. Heck, UTIs, pneumonia, surgery, basically any inflammatory insult in the elderly can hasten cognitive decline.
What’s it all mean?
The reality is, it’s unclear and I dare say even unlikely that a mild infection from SARS-CoV-2 in a young healthy person will cause long-term brain damage. That’s just not the take-home message from the Nature paper.
If you have comorbidities, have a pre-existing mild cognitive impairment, or are elderly then absolutely get vaccinated and take precautions to avoid infections… of any kind!
Why write this?
I was initially going to let this all slide by and move on with my life. Then I received an email from Medium with my daily digest. Second on the list was the following hot take:
Something about this ‘independent’ media clickbait just irritated me more than the traditional digital media I started the article with. I think it has something to do with the strong opinions conveyed in the article, and the arrogance of the viewpoint expressed. It’s as if the author were pontificating in divine providence. He goes so far as to claim any possible sobering realism on interpreting the Nature paper brands you an ‘idiot.’
Well not only is this article bad scientific communication, but it’s deliberately polarising and just an unhelpful addition to an already bloated information ecosystem. It would be a slick and enjoyable piece of writing if it weren’t such an infuriatingly wrong hot take.
The most charitable interpretation of this article is that the author genuinely believes that deliberately misinterpreting the science to scare people into slowing the spread is important. The worst case is that he’s on a deadline to produce content, he only read the abstract, and pushed ctrl-v over the sexy graph. I’ll let the readers be the judge of that.
Wrap up
Scientific findings aren’t headlines, and presenting the importance of new research effectively takes finesse. On one level this is a cautionary tale of how bad reporting can create unnecessary panic for a mother and infinite irritation in her son. I worry, however, that these forces at scale create real social problems. Knowledge is power, but the illusion of knowledge is a trojan horse.
