In a normal world, there is no such thing as asymptomatic infected ~ only pre-symptomatic and symptomatic infected. In a normal world, you would not test just anyone, you would only test the symptomatic. In addition to "over-testing", we use testing techniques, PCR and antigen, that can produce false positives and false negatives. The PCR is especially useful because you can virtually guarantee a positive result by using a high number of cycles. It just seems to me that we have testing scenarios set up for the express purpose of producing high case numbers, while making it virtually impossible to determine VE. Based on your expose of TND, I now know that the scientific community has created another method of biasing already biased/corrupted data.
I am not a doctor/scientist. In the context of my remarks above, which were meant to be about COVID-19 only, I am very bothered by the semantics of "asymptomatic" in popular usage. In an article by Jeremy Hammond (url below), he talks about the sloppy use/misuse/misinterpretation of the term by health officials, the media, and fact-checkers, which end up contributing more to fear-mongering than a true understanding of COVID by the public.
In the article, Hammond cites WHO's systematic review of evidence of asymptomatic transmission via contact tracing studies. "The review also uses the term “asymptomatic case”. Keep in mind that this means a person who received a positive PCR test but never developed COVID-19." The WHO review concludes that asymptomatic cases and transmission are possible, but with low likelihood. My problem with that conclusion, is that proof of infection was based on a positive PCR test result. How many cycles? False positives? Have we really proven anything? It just seems more credible, in my opinion, to avoid using asymptomatic at all with regard to COVID, and just talk about pre-symptomatic and symptomatic.
I concur with all your points and share the concern about it being little more than a tool to justify imposing rules framed by those seeking a particular outcome.
Leaving the illegitimate fear mongering aside, I used the example of HIV because it wasn't previously known how it's spread. As you noted, we now think it's spread asymptomatically too. I can't vouch for the truth of any of those statements but I assume they are.
What do you think of Table C in this Omicron report. Japanese Scientists see 50% of asymptomatic RTPCR positive "cases" carrying infectious virus in the first five days. Symptomatic carry it in same amount but peak a little later. Small numbers.
Thank you! I'm glad to learn how to be wary of this kind of design. And, I especially liked your first sentence, "Assumptions are where art meets science, and, these days, the scientific literature is beginning to look like a Jackson Pollock drip-painting."
The masked maniacs are lining up to get tested because they’re jonesing on the dopamine hits that come from participating in a global fear porn operation.
I’m still waiting for Covidiots to start using fabric condoms to prevent transmission of HIV. If a fabric mask will stop Covid, why wouldn’t a blue cloth stop HIV?
Thank you for the great write-up. I'm in agreement with you about healthcare seeking behavior being normalized is a tenuous assumption but I was under the impression that this design is better at tracking asymptomatic Infections which from a VE standpoint is absolutely essential to stop transmission. I tended to think that TNDs did better at showing this risk due to better incidental coverage of infections [say getting on a plane or entering a hospital to meet a patient].
I agree that hypochondriacs will affect symptomatic VE as will government or structural biases that differentially allow people access or inaccess but I still think those could be mitigated somewhat. For example we can take the first or the last positive or negative test and discard the rest. One advantage that I've mentioned to you before of TNDs in my view is seeing the sensitivity of the outcomes to inclusion criteria where a person was initially a case then later becomes a control or vice versa. Real life has people who test positive then negative then vaccinate then positive again etc. This is an interesting case pathway because it has hidden risk that regular observational cohorts studies miss due to matching and following and not allowing switch over.
I think one other really big weakness of TNDs is the lack of any basis to control what the investigator decided is a matched pair and discarded the other. I will once again mention that I couldn't believe my eyes that a Qatar preprint that showed severe cases in first two weeks after vaccination magically had no difference in VE by the time it was printed in NEJM. The raw data shows the investigator clearly had an anomalous string of fortuitous matches that accrued to save the VE in 1 month that was not happening for so long. Seems suspect.
"I tended to think that TNDs did better at showing this risk due to better incidental coverage of infections"
If we restrict the tested patients to only those entering a hospital to meet a patient, say, then the healthcare-seeking behavior assumption may hold. But as soon as we poison the well, so to speak, and use TNDs on society at large (as occurred in the Discovery South Africa or CDC natural immunity study), then the assumption falls apart. The assumption is what gives TNDs almost quasi-causal power and taking it away completely changes the interpretation of the results. It's a limited-use study design that is being overused simply because its a quick and cheap way to get published. Here is a quote from Vandenbroucke and Pearce, which I think captures the essence of what TNDs should be and what they are not in the COVID literature: "Test-negative studies recruit cases who attend a healthcare facility and test positive for a particular disease; controls are patients undergoing the same tests for the same reasons at the same healthcare facility and who test negative."
" But as soon as we poison the well, so to speak, and use TNDs on society at large (as occurred in the Discovery South Africa or CDC natural immunity study), then the assumption falls apart. "
You are right. It's extremely sensitive to context. Another interesting thing about that is that Qatar study when included people who tested on the port of entry, the VE dropped to -70.
This illustrates your point that it's less useful and easily misunderstood as a gauge for "vaccine effectiveness" in society at large when it's really only reflecting highly context specific VEs. I suppose young persons on a plane who land in Qatar will be infected. Not necessarily because the vaccines are bad, but rather the leaky vaccine allows transmission between people on an airport and a plane and the rules eliminate Unvaccinated from being there and exposed, so looks like being unvaccinated is protective when in reality, it's a reflection of composition of the susceptible group enhancing transmission in specific settings.
Funny enough, I just read a test-negative paper with similar methods to what you suggested (hospital only setting). I think you may enjoy going through the data in this paper. Two things stood out to me. 1) getting tested appears to have been voluntary ie., even in one setting, the assumption is violated 2) I have never seen VE be expressed as having a minimum value of 0 ;)
Thank you! The first point is absolutely correct, it overestimates VE in favor of hypochondriacs! But underestimates VE in favor of those not infected. In case VE is negative, this will mean the VE starts to look even more negative as the positive tests pile up!
Yes, I did enjoy it and thanks for sharing because that statistic is exactly why we are seeing UK trying to push boosters in healthcare workers because they are aware that the patients are dying because the vaccinated healthcare workers are infecting them.
Thanks for this fine write-up I did not realize test-taking was being used as an instrumental variable in this way but I agree it is a highly suspect assumption.
In a normal world, there is no such thing as asymptomatic infected ~ only pre-symptomatic and symptomatic infected. In a normal world, you would not test just anyone, you would only test the symptomatic. In addition to "over-testing", we use testing techniques, PCR and antigen, that can produce false positives and false negatives. The PCR is especially useful because you can virtually guarantee a positive result by using a high number of cycles. It just seems to me that we have testing scenarios set up for the express purpose of producing high case numbers, while making it virtually impossible to determine VE. Based on your expose of TND, I now know that the scientific community has created another method of biasing already biased/corrupted data.
For something like HIV, would you say asymptomatic infected exists? If they can transmit?
Yes, asymptomatic infection and asymptomatic transmission with HIV exist: https://www.cdc.gov/hiv/basics/whatishiv.html
I am not a doctor/scientist. In the context of my remarks above, which were meant to be about COVID-19 only, I am very bothered by the semantics of "asymptomatic" in popular usage. In an article by Jeremy Hammond (url below), he talks about the sloppy use/misuse/misinterpretation of the term by health officials, the media, and fact-checkers, which end up contributing more to fear-mongering than a true understanding of COVID by the public.
In the article, Hammond cites WHO's systematic review of evidence of asymptomatic transmission via contact tracing studies. "The review also uses the term “asymptomatic case”. Keep in mind that this means a person who received a positive PCR test but never developed COVID-19." The WHO review concludes that asymptomatic cases and transmission are possible, but with low likelihood. My problem with that conclusion, is that proof of infection was based on a positive PCR test result. How many cycles? False positives? Have we really proven anything? It just seems more credible, in my opinion, to avoid using asymptomatic at all with regard to COVID, and just talk about pre-symptomatic and symptomatic.
Hammond's article: https://www.jeremyrhammond.com/2020/12/15/the-big-lie-about-asymptomatic-transmission-of-sars-cov-2/
I concur with all your points and share the concern about it being little more than a tool to justify imposing rules framed by those seeking a particular outcome.
Leaving the illegitimate fear mongering aside, I used the example of HIV because it wasn't previously known how it's spread. As you noted, we now think it's spread asymptomatically too. I can't vouch for the truth of any of those statements but I assume they are.
What do you think of Table C in this Omicron report. Japanese Scientists see 50% of asymptomatic RTPCR positive "cases" carrying infectious virus in the first five days. Symptomatic carry it in same amount but peak a little later. Small numbers.
https://www.niid.go.jp/niid/en/2019-ncov-e/10884-covid19-66-en.html
Thank you! I'm glad to learn how to be wary of this kind of design. And, I especially liked your first sentence, "Assumptions are where art meets science, and, these days, the scientific literature is beginning to look like a Jackson Pollock drip-painting."
Thank you for this clear explanation of test-negative design.
The masked maniacs are lining up to get tested because they’re jonesing on the dopamine hits that come from participating in a global fear porn operation.
I’m still waiting for Covidiots to start using fabric condoms to prevent transmission of HIV. If a fabric mask will stop Covid, why wouldn’t a blue cloth stop HIV?
Thank you for the great write-up. I'm in agreement with you about healthcare seeking behavior being normalized is a tenuous assumption but I was under the impression that this design is better at tracking asymptomatic Infections which from a VE standpoint is absolutely essential to stop transmission. I tended to think that TNDs did better at showing this risk due to better incidental coverage of infections [say getting on a plane or entering a hospital to meet a patient].
I agree that hypochondriacs will affect symptomatic VE as will government or structural biases that differentially allow people access or inaccess but I still think those could be mitigated somewhat. For example we can take the first or the last positive or negative test and discard the rest. One advantage that I've mentioned to you before of TNDs in my view is seeing the sensitivity of the outcomes to inclusion criteria where a person was initially a case then later becomes a control or vice versa. Real life has people who test positive then negative then vaccinate then positive again etc. This is an interesting case pathway because it has hidden risk that regular observational cohorts studies miss due to matching and following and not allowing switch over.
I think one other really big weakness of TNDs is the lack of any basis to control what the investigator decided is a matched pair and discarded the other. I will once again mention that I couldn't believe my eyes that a Qatar preprint that showed severe cases in first two weeks after vaccination magically had no difference in VE by the time it was printed in NEJM. The raw data shows the investigator clearly had an anomalous string of fortuitous matches that accrued to save the VE in 1 month that was not happening for so long. Seems suspect.
"I tended to think that TNDs did better at showing this risk due to better incidental coverage of infections"
If we restrict the tested patients to only those entering a hospital to meet a patient, say, then the healthcare-seeking behavior assumption may hold. But as soon as we poison the well, so to speak, and use TNDs on society at large (as occurred in the Discovery South Africa or CDC natural immunity study), then the assumption falls apart. The assumption is what gives TNDs almost quasi-causal power and taking it away completely changes the interpretation of the results. It's a limited-use study design that is being overused simply because its a quick and cheap way to get published. Here is a quote from Vandenbroucke and Pearce, which I think captures the essence of what TNDs should be and what they are not in the COVID literature: "Test-negative studies recruit cases who attend a healthcare facility and test positive for a particular disease; controls are patients undergoing the same tests for the same reasons at the same healthcare facility and who test negative."
Emphasis on the word "same".
" But as soon as we poison the well, so to speak, and use TNDs on society at large (as occurred in the Discovery South Africa or CDC natural immunity study), then the assumption falls apart. "
You are right. It's extremely sensitive to context. Another interesting thing about that is that Qatar study when included people who tested on the port of entry, the VE dropped to -70.
This illustrates your point that it's less useful and easily misunderstood as a gauge for "vaccine effectiveness" in society at large when it's really only reflecting highly context specific VEs. I suppose young persons on a plane who land in Qatar will be infected. Not necessarily because the vaccines are bad, but rather the leaky vaccine allows transmission between people on an airport and a plane and the rules eliminate Unvaccinated from being there and exposed, so looks like being unvaccinated is protective when in reality, it's a reflection of composition of the susceptible group enhancing transmission in specific settings.
Funny enough, I just read a test-negative paper with similar methods to what you suggested (hospital only setting). I think you may enjoy going through the data in this paper. Two things stood out to me. 1) getting tested appears to have been voluntary ie., even in one setting, the assumption is violated 2) I have never seen VE be expressed as having a minimum value of 0 ;)
https://www.medrxiv.org/content/10.1101/2022.01.07.22268919v1.full.pdf
Thank you! The first point is absolutely correct, it overestimates VE in favor of hypochondriacs! But underestimates VE in favor of those not infected. In case VE is negative, this will mean the VE starts to look even more negative as the positive tests pile up!
2) Qatar study preprint did that too! expressed VE as zero so I had to manually calculate. (page 24 onwards) https://www.medrxiv.org/content/10.1101/2021.08.25.21262584v1.full.pdf
Yes, I did enjoy it and thanks for sharing because that statistic is exactly why we are seeing UK trying to push boosters in healthcare workers because they are aware that the patients are dying because the vaccinated healthcare workers are infecting them.
Thanks for this fine write-up I did not realize test-taking was being used as an instrumental variable in this way but I agree it is a highly suspect assumption.