I'm a boosted, healthy person living in NYC where the vaccination rate is relatively high and positivity rates have plunged. My take, helped by your post, is to loosen my self-imposed restrictions that affect quality of life. I won't look to get Omicron, but if I do, the silver lining may be a boost against a subsequent variant. I'll also still be careful and mask around high risk people. Thanks Zvi!
I'm only part of the way through, but I wanted to comment on this quote:
"But on further thought, that’s totally wrong: this pattern could be caused by beliefs causing illness, but it could also be caused by illness causing beliefs, which obviously happens all the time."
I think that this excludes the idea that both belief causes illness *and* illness causes belief, and it wouldn't be the first cycle in the human body that does this. Its the primary driver of pain. Injure a joint, feel pain, and you'll still feel pain after the joint is structurally healed. Physiological measures of depression like heart rate respond to self talk about depression [citation needed, but I know I read this a while ago], etc. You can release saliva if you think about food. You can dump all kinds of hormones into your blood stream if you think about a naked attractive person or bear attack. Arguably, you can dump mucus in your nose if you drink milk and think it causes congestion, but not if you don't [citation needed for all these, sorry].
We only disagree on the English here not the logic - when writing this I was thinking of this as and/or not xor, and my best guess is indeed that causation here is in both directions.
There is a paper [1] that suggests that "many long COVID symptoms may not be a direct result of the SARS-CoV-2 virus but may be the result of COVID-19 inflammation-induced EBV reactivation". Any thoughts on that hypothesis?
Thanks again for all the work you put into these. This was truly monstrous, but with all the work seemingly being done to keep people scared of COVID and fluff up Long COVID to keep the panic train going, it was probably necessary. Well done.
"If you’d shown me a positive serology test in January 2021 and asked me if I’d had Covid, I’d have said yes, because you just showed me the test, so I guess I must have.
"The French public did not think this way.
"I find that to be an interesting study result in its own right. People simply don’t believe such tests, or don’t even care enough to look at them in context."
In January 2021, what was the percent who'd had covid? Using CDC numbers it's maybe 8% of the US by the end of Jan 2021, but that's probably an understatement. But with a sensitivity of .87, using Bayes theorem, you are still getting P(covid|+test)=.87*X where X is percent have had covid in the past, right? I don't think X is as high as .5/.87=.57
If I were on the edge as to whether I thought I had covid in the past, suppose I had been sick after being exposed to covid (or what I thought looked like covid), then a test like this would sway me one way of the other, I guess. But if I had a really bad disease or if I never felt sick I don't think I'd change my mind. Maybe that's what the French population did. Maybe the researchers told them how to interpret the test results? Maybe your priors are much higher that you got covid without noticing? Though that seems different that saying "you just showed me the test, so I guess I must have."
Specificity is very high, though. A test that picks up on 87% of positives and only 1% of negatives, with an 8% base rate, is probably giving you a true positive.
And yes, my prior is that I had enough minor colds and what not that I'd not notice reasonably often.
"There are also issues about the serology test that was used. Generally the report is written with a mostly implicit assumption that the people who tested positive for antibodies on the serology test definitely had been previously infected, while those who tested negative had not been infected. The researcher do acknowledge that that isn’t the case – in the part of the Discussion section of their report about Strengths and Limitations, they rightly acknowledge that the antibody test might, in some cases, not actually correspond to whether people had really been infected. That’s always the case for any diagnostic test of this sort, because no such test is perfect. The researchers provide arguments to support their belief that this is not in fact a big problem in this study – but errors in diagnostic test results are often counter-intuitive, and I don’t entirely agree with the researchers’ conclusion. They do correctly point out that, with an assumed prevalence of previous infection of 4%, and the performance rates for the test that they quoted earlier in their report, they would expect 139 participants to be false negatives (that is, they had a negative antibody test results but did actually have a previous infection), and they rightly point out that this is less than 1% of those who tested negative. What they don’t do is look at false positives – with the assumptions they make, there would be rather a lot of false positives, in fact about 644 of them, which would be about four in every ten people who test positive for antibodies. That is, quite a big proportion of those who had a positive serology result would not in fact have been infected – and in a study that is comparing how closely test results align to symptoms, false positives can matter too."
I had a whole section on serology in particular to handle the French study, and concluded this is not a concern. This doesn't change that. Also the 4% number seems crazy low to me in a 'yeah, uh huh' kinda way.
Heh, I'm dumb. P(covid|+test)=P(covid, +test)/(P(covid, +test)+P(no covid, +test))=X*.87/(X*.87+(1-X)*.025) which means X only needs to be around 2.8% to make P(covid|+test). Don't know what I was thinking. No, I guess I just wasn't thinking...
No, higher risks for a specific set of disorders known to be associated with COVID-19. And of course there aren't "controls" -- this isn't an RCT. I'm not good enough at statistics to assess whether appropriate statistical techniques were used to compensate for this not being an RCT, but I presume the Nature reviewers have such expertise (https://www.nature.com/articles/s41591-022-01689-3)
"Appropriate statistical techniques" are exactly what Zvi means by controls, if you look at what he's said about controls in his main post. He's given multiple examples of peer-reviewed studies in the post that lack adequate controls.
That said, quickly searching the study you linked, it looks like they did thoroughly attempt to include controls, and I too lack the statistical expertise to evaluate them, so would love a more thorough response if and when Zvi has time.
However, someone else linked that on the LW copy of this post, and here some of my thoughts there on why it probably doesn't change Zvi's overall conclusion:
״about future unknowns. Given how much time has now passed, and what I see as the relevant reference classes, I don’t think we need worry about this going foward, but precautionary principle does apply.״
We are only two and something years out from the first known infection with SARS-CoV-2. Over the same period, zero people with HIV will suffer AIDS or will be hospitalized for it because it takes 3-15 years to appear, nobody with Measles will suffer SSPE that takes 7 years from infection and same for Herpes Zoster, Post Polio, liver failure after Hepatitis C etc.
HIV is 100%, SSPE is lethal 1 in 600 for little kids and 1 in 1,400 for adolescence. So it's a huge risk to take for unvaccinated kids, we do not yet have vaccines for kids up to 5. How do you plan to protect them until we do?
I appreciate your taking on this very very important question about evaluating Long Covid. How do you account for studies showing brain issues in hamsters? Your post doesn’t deal enough with studies such as this showing measurable issues with brain function comparable to Alzheimer’s etc — could you address that? Surely may be something that could be more specific about those comparisons made such as to Alzheimer’s — is that a valid comparison? Otherwise your post reads too much like — hmm, haven’t been able to measure it, so let’s say it’s worth the risk/reward. https://www.scientificamerican.com/article/covid-smell-loss-and-long-covid-linked-to-inflammation1/
I am not aware of the hamster studies and they were not incorporated into my thinking. I don't think the comparisons to Alzheimer's are valid from what I've seen. There's a ton of *different* such claims and even with such a long post one can't address them all. If I see consensus that a particular additional data point is concerning, I'll take another look at whatever it may be.
I'm a boosted, healthy person living in NYC where the vaccination rate is relatively high and positivity rates have plunged. My take, helped by your post, is to loosen my self-imposed restrictions that affect quality of life. I won't look to get Omicron, but if I do, the silver lining may be a boost against a subsequent variant. I'll also still be careful and mask around high risk people. Thanks Zvi!
You are welcome. That's exactly why I wrote this.
I'm only part of the way through, but I wanted to comment on this quote:
"But on further thought, that’s totally wrong: this pattern could be caused by beliefs causing illness, but it could also be caused by illness causing beliefs, which obviously happens all the time."
I think that this excludes the idea that both belief causes illness *and* illness causes belief, and it wouldn't be the first cycle in the human body that does this. Its the primary driver of pain. Injure a joint, feel pain, and you'll still feel pain after the joint is structurally healed. Physiological measures of depression like heart rate respond to self talk about depression [citation needed, but I know I read this a while ago], etc. You can release saliva if you think about food. You can dump all kinds of hormones into your blood stream if you think about a naked attractive person or bear attack. Arguably, you can dump mucus in your nose if you drink milk and think it causes congestion, but not if you don't [citation needed for all these, sorry].
We only disagree on the English here not the logic - when writing this I was thinking of this as and/or not xor, and my best guess is indeed that causation here is in both directions.
There is a paper [1] that suggests that "many long COVID symptoms may not be a direct result of the SARS-CoV-2 virus but may be the result of COVID-19 inflammation-induced EBV reactivation". Any thoughts on that hypothesis?
[1] "Investigation of Long COVID Prevalence and Its Relationship to Epstein-Barr Virus Reactivation" https://www.ncbi.nlm.nih.gov/labs/pmc/articles/PMC8233978/
Interesting. Sample sizes involved are quite small, but I'd like to see a bigger and properly controlled replication.
Thanks again for all the work you put into these. This was truly monstrous, but with all the work seemingly being done to keep people scared of COVID and fluff up Long COVID to keep the panic train going, it was probably necessary. Well done.
I'm confused by this:
"If you’d shown me a positive serology test in January 2021 and asked me if I’d had Covid, I’d have said yes, because you just showed me the test, so I guess I must have.
"The French public did not think this way.
"I find that to be an interesting study result in its own right. People simply don’t believe such tests, or don’t even care enough to look at them in context."
In January 2021, what was the percent who'd had covid? Using CDC numbers it's maybe 8% of the US by the end of Jan 2021, but that's probably an understatement. But with a sensitivity of .87, using Bayes theorem, you are still getting P(covid|+test)=.87*X where X is percent have had covid in the past, right? I don't think X is as high as .5/.87=.57
If I were on the edge as to whether I thought I had covid in the past, suppose I had been sick after being exposed to covid (or what I thought looked like covid), then a test like this would sway me one way of the other, I guess. But if I had a really bad disease or if I never felt sick I don't think I'd change my mind. Maybe that's what the French population did. Maybe the researchers told them how to interpret the test results? Maybe your priors are much higher that you got covid without noticing? Though that seems different that saying "you just showed me the test, so I guess I must have."
Specificity is very high, though. A test that picks up on 87% of positives and only 1% of negatives, with an 8% base rate, is probably giving you a true positive.
And yes, my prior is that I had enough minor colds and what not that I'd not notice reasonably often.
Have you read Kevin McConway's response to this study?
https://www.sciencemediacentre.org/expert-reaction-to-study-looking-at-the-association-of-self-reported-covid-19-infection-and-sars-cov-2-serology-test-results-with-persistent-physical-symptoms/
This paragraph in particular is useful:
"There are also issues about the serology test that was used. Generally the report is written with a mostly implicit assumption that the people who tested positive for antibodies on the serology test definitely had been previously infected, while those who tested negative had not been infected. The researcher do acknowledge that that isn’t the case – in the part of the Discussion section of their report about Strengths and Limitations, they rightly acknowledge that the antibody test might, in some cases, not actually correspond to whether people had really been infected. That’s always the case for any diagnostic test of this sort, because no such test is perfect. The researchers provide arguments to support their belief that this is not in fact a big problem in this study – but errors in diagnostic test results are often counter-intuitive, and I don’t entirely agree with the researchers’ conclusion. They do correctly point out that, with an assumed prevalence of previous infection of 4%, and the performance rates for the test that they quoted earlier in their report, they would expect 139 participants to be false negatives (that is, they had a negative antibody test results but did actually have a previous infection), and they rightly point out that this is less than 1% of those who tested negative. What they don’t do is look at false positives – with the assumptions they make, there would be rather a lot of false positives, in fact about 644 of them, which would be about four in every ten people who test positive for antibodies. That is, quite a big proportion of those who had a positive serology result would not in fact have been infected – and in a study that is comparing how closely test results align to symptoms, false positives can matter too."
I had a whole section on serology in particular to handle the French study, and concluded this is not a concern. This doesn't change that. Also the 4% number seems crazy low to me in a 'yeah, uh huh' kinda way.
Heh, I'm dumb. P(covid|+test)=P(covid, +test)/(P(covid, +test)+P(no covid, +test))=X*.87/(X*.87+(1-X)*.025) which means X only needs to be around 2.8% to make P(covid|+test). Don't know what I was thinking. No, I guess I just wasn't thinking...
A large study (US Dept Veterans Affairs, but pre-vaccine): https://arstechnica.com/science/2022/02/covid-raised-heart-risks-63-for-a-year-after-infection-in-study-of-11m-people/
Pretty sure this is selection and insufficient controls. Oh look, higher risk for basically everything, uh huh, yeah.
No, higher risks for a specific set of disorders known to be associated with COVID-19. And of course there aren't "controls" -- this isn't an RCT. I'm not good enough at statistics to assess whether appropriate statistical techniques were used to compensate for this not being an RCT, but I presume the Nature reviewers have such expertise (https://www.nature.com/articles/s41591-022-01689-3)
"Appropriate statistical techniques" are exactly what Zvi means by controls, if you look at what he's said about controls in his main post. He's given multiple examples of peer-reviewed studies in the post that lack adequate controls.
That said, quickly searching the study you linked, it looks like they did thoroughly attempt to include controls, and I too lack the statistical expertise to evaluate them, so would love a more thorough response if and when Zvi has time.
However, someone else linked that on the LW copy of this post, and here some of my thoughts there on why it probably doesn't change Zvi's overall conclusion:
https://www.lesswrong.com/posts/mh3xapTix6fFtd3xM/the-long-long-covid-post?commentId=iyRCL332J7eLvxNqq
״about future unknowns. Given how much time has now passed, and what I see as the relevant reference classes, I don’t think we need worry about this going foward, but precautionary principle does apply.״
We are only two and something years out from the first known infection with SARS-CoV-2. Over the same period, zero people with HIV will suffer AIDS or will be hospitalized for it because it takes 3-15 years to appear, nobody with Measles will suffer SSPE that takes 7 years from infection and same for Herpes Zoster, Post Polio, liver failure after Hepatitis C etc.
HIV is 100%, SSPE is lethal 1 in 600 for little kids and 1 in 1,400 for adolescence. So it's a huge risk to take for unvaccinated kids, we do not yet have vaccines for kids up to 5. How do you plan to protect them until we do?
I appreciate your taking on this very very important question about evaluating Long Covid. How do you account for studies showing brain issues in hamsters? Your post doesn’t deal enough with studies such as this showing measurable issues with brain function comparable to Alzheimer’s etc — could you address that? Surely may be something that could be more specific about those comparisons made such as to Alzheimer’s — is that a valid comparison? Otherwise your post reads too much like — hmm, haven’t been able to measure it, so let’s say it’s worth the risk/reward. https://www.scientificamerican.com/article/covid-smell-loss-and-long-covid-linked-to-inflammation1/
I am not aware of the hamster studies and they were not incorporated into my thinking. I don't think the comparisons to Alzheimer's are valid from what I've seen. There's a ton of *different* such claims and even with such a long post one can't address them all. If I see consensus that a particular additional data point is concerning, I'll take another look at whatever it may be.
"but I do continue to think a lot of it is ‘blame whatever is wrong with me on Covid.'"
Fuck you Zvi. Just fuck you.
I know you don't have it. I do.
I'm never gonna subscribe to you now.
If you weren't such a great writer and fellow Tribal member I would hunt you down and torch you online for saying something so profoundly ignorant.