The Healthy Vaccinee Effect (HVE) is too often estimated and treated as a marginal phenomenon. In studies it is in Strengths and Weaknesses Or eliminated in a footnote: "There is no correction for the HVE, so that the figures must be interpreted carefully" or words of the same scope.
So it is not about placebo or nocebo effects (you can here More about reading). Psychological effects do not participate. It is purely about the statistical consequence of "misclassification": the transfer of people who will die in the not too long term, from the vaccinated person to the unvaccinated group.
With a simple, so simplified calculation, you can experiment yourself how the HVE effect, reduced here to non-vaccination of people with poor health and/or approaching end of life, can easily lead to a respectable vaccine effectiveness that is added to the actual effect of a vaccine.
Bottom line
Even with a non-workable vaccine, such as above, the VE (vaccine effectiveness) soon rises. If you add a vaccine that actually has 25% effectiveness with OP, then with lower assumptions you will soon come to the 80%, 90% effectiveness.
Converted to life years, the image becomes truly dramatic because many diseases mainly affect the vulnerable elderly, while the entire population is exposed to possible vaccine side effects.
It is impossible that researchers do not know this. Observational vaccinst studies that do not do their utmost to correct for the HVE aim to flatter the results. I don't see a different option than deliberation.
Addendum: rubbing the stain
Intention or misunderstanding? There are studies in which the HVE is recognized and addressed in a way that again testifies to either a lack of statistical insight or a conscious polishing of unwelcome data, namely the omission of the first weeks after the injection. In this way, one artifact is replaced by another: HVE out, Time Related Bias in...
1. Bar-on et al., 2022 (NEJM)- Protection by a Fourth Dose…
- Discussion: “We attempted to address this bias by excluding the first 7 days after vaccination from the analysis.”
https://www.nejm.org/doi/full/10.1056/NEJMoa2201570
2. Tartof et al., 2021 (The Lancet) — Effectiveness of mRNA BNT162b2 up to 6 months
- Limitations: “Limitations include the observational design, possible residual confounding, and healthy vaccinee bias, which we attempted to minimize by excluding person-time during the first week after each dose.”
https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(21)02183-8/fulltext
3. Patalon et al., 2022 (Nature Communications) — Waning Effectiveness of the Third Dose of BNT162b2
- Methods/Discussion (summarized key sentence): “We minimized [healthy vaccinee] bias by excluding the first 7 days post-vaccination from our analysis.”
https://www.nature.com/articles/s41467-022-30884-6
4. Dagan et al., 2021 (NEJM) — BNT162b2 mRNA Covid-19 Vaccine in a Nationwide Mass Vaccination Setting
- Discussion: “Limitations include unmeasured confounding and possible selection bias. Healthy vaccinee bias may play a role in the early post-vaccination period, but we addressed this by excluding person-time during the first 7 days after the first dose.”
https://www.nejm.org/doi/full/10.1056/NEJMoa2101765
5. Jackson et al., 2013 (Vaccine) — Influenza vaccine effectiveness in seniors estimated using different methods
- Discussion: “We observed that vaccinated seniors had lower risk of death and hospitalization even before influenza circulation began … To reduce the impact of this bias, we excluded outcomes that occurred during the first 14 days after vaccination. After this exclusion, effectiveness estimates were closer to expected values, but residual confounding is still possible.”
https://pubmed.ncbi.nlm.nih.gov/24095882/
At the ages of 70-80 years we see a vaccination rate of 93%. Well protected, good turnout.
However …. At 90+ we only see a vaccination rate of 84%. More than 2 times as much unaccinated. That smells of "too weak to vaccinate". That is 18,500 unvaccinated deaths. Total deceased is 27,000. So 2/3 of the elderly who died unaccinated. That doesn't necessarily have to be the same, but a strong indication
That means a much higher expectation percentage and also a much higher real mortality rate in that group, which means that the VE probably runs towards 100%. But this calculator does not work age -specific unfortunately. Then you should fill in background percentages and expectations per age. This post was accidentally published, I first wanted to go through it with you again. But that's how it is possible.
You actually mix up 2 things.
What all the hassle with Ronald Meester and Herman is about his "afterwards studies". Then you have to deal with HVE and the effectiveness etc. is very difficult to determine. Because the vaccinees have different characteristics than the non -vaccinated people as you correctly indicate in your article: the non -vaccinees have relatively many people who will soon die. But afterwards that is difficult to determine.
What should have been (and that was also half done by, among other things, Pfizer) and RCT (https://www.scribbr.nl/onderzoeksmethoden/randomized-controlled-trial/ ).
Then the 2 groups are really random and you will not be bothered by HVE at all. And then you can clean the effectiveness and side effects/death. But for (possibly sinister) inexplicable reasons, Pfizer has completely vaccinated that control group halfway. Away pure measurement.
You think this is mixed with RCTs. This is purely about observational afterwards research: “The HVE problem arises if we look back on a vaccination campaign and want to compare a group of vaccinates with a group of unvaccinants. With a well-designed trial, those groups are" matched ": they are similar in composition: age, gender, health status, etc.” Or am I misunderstanding you?
(matched should actually be randomized)
True.
We should actually require RCTs to be done (the gold standard). Because those "retrospect" investigations by definition have inaccurate and especially too optimistic outcomes.
That is why I didn't understand what you wanted with this article. Because that HVE really has already published many clear articles with good explanation. Also on virus varia.
Anyway. Now that Pfizer has already messed up his RCT after six months, we will have to. And maybe there are still people who don't know this yet. And can clarify your explanation with numbers ... ..
It was my intention to show that, especially with a high rate, it degenerates into enormous differences. It does not save a percent or two, it can be 50% and more. And that enormous effect is almost nowhere addressed, except with a disclaimer in the Strengths and Weaknesses: "The actual effectiveness can differ", such forms, as if it is something marginal.
With an RCT, the two groups should be about the same size by the way, then it plays much less.
Ah, with me only now falls.
You wanted to show that the HVE even measured a total, even psychological, ineffective placebo, afterwards, a VE of 50 - 80%. [N.B. A placebo really works thanks to psychological effects on the patient. A lot of research has been done on that.).
That did not arrive immediately (is that up to me or the text?). So I would adjust the title and the opening sentences.
With an RCT, the groups do not have to be the same size at all. As long as they are really random (and therefore completely similar to each other). For "really random" a sufficient size is required; But no identical size.
To reinforce the message, I would outline 5 scenarios:
1. A high rate of vaccination among the (total) population from, for example, approx. 18 years. Then the HVE is limited.
2. A high rate of vaccination among only the 60+ population (as with the flu shot and now the Corona Prick); Then the HVE "explodes". And so the HVE if you only do retrospective research.
3. And explicitly show that the HVE is increasing exponentially with the vaccination rate within one specific target group.
4. Identrate that the HVE can lead to Simpson's Paradox in an inhomogenic population: the overall VE seems positive, but this masks that the VE can vary greatly in different age groups. For example: a seemingly high overall VE of 70% can consist of a VE of 90% at 18-50 year olds (who had very low risk anyway) and a VE of only 20% or even negative at 70+ (where the vaccine was most needed). The HvE strengthens this effect because within each age group the most vulnerable is under -represented among the vaccinates.
5. Demonstrate that in groups with a very low baseline risk (such as young people) an apparently high VE can mask that the vaccine is net harmful. For example: if young people have a baseline risk of 0.01% for serious illness, a vaccine with 80% can reduce this to 0.002%-a "impressive" relative risk reduction. But if the same young people have 0.05%chance of serious side effects, then the absolute damage (0.05%) is much greater than absolute protection (0.008%). The HvE strengthens this problem because the healthiest young people are vaccinated, so that their low baseline risk is even further underestimated.
This illustrates why absolute risk reduction and number needed to treat/harm are crucial metrics, especially with low-risk populations where impressive relative figures can conceal a negative damage benefit balance.
It is nice to note that if there is a difference in VE between different age groups, the result of calculations will completely go wrong. We are working on an article to be published soon, in which we present a better method for calculating vaccination rate and VE. If, for example, you assume a VE for young people of 0% and the elderly of 33%, then the VE comes to -44% for the entire population (in our example). Extremely remarkable this negative value, but it is. With our method we will come up with a realistic value: 26%.
I'm going to read that with red ears !!!
I have now indicated in the intro that this is not about the placebo or nocebo effect. I have that too previously treated quite extensively: with regard to population, doctors/ protocols/ and psycho-indematic.
Good addition!
I have often seen that a control group is a lot smaller than the test group. I now understand why a little better, as you can see from the calculator.
If a control group only has to be "sufficiently large", why is the test group bigger? What is the statistical rationale behind it? To prove significance earlier perhaps?
A control group that is considerably smaller than the test group causes a leverage effect when transferring well to non-vaccinated. With 1/4 control and 3/4 test group, each percent transfer yields a reduction of 1/75% in the test group, plus a 3x so large dying person of 1/25% in the check. But you can easily view that effect in the calculator.
The explanation and refinements that you provide further give enough material for a syllabus 🙂 This calculator is trying to keep accessible and understandable for now, with only the most necessary parameters. And above all: in line with how the audience is informed about HVE. If it is already mentioned ...
Thank you for your input again Jan!
yes it is crystal clear!
In point 3 under "About the default values" I read "vaccinates" where, I suspect, "unvaccinated" is meant.
Under "the two most difficult" I read: "(0.4% = half of the actual death of that month you see)" - Hoebedoelu? For me it is Wartaal. Then "Alk" instead of "Al", an understandable typo.
Regarding the last question: I think that in three months more vulnerable die than 3 times the dead Kwetfaren of one month, because that one month is the "first" month. When on day 0, for example, 100 people get the prognosis "I will give you at most half a year", it seems more likely that after one month of maybe 3, after two months 11 and after three months 28, etc., there are about 15 those who are recovering every month. The data to formulate a workable position for something like that seems to be difficult to acquire ...
Thank you for the improvements, indeed: the non-vaccinated people. And Alk => Al.
[edited response because the calculator has been adjusted, AT]
The actual death per month is 0.8% of the population. How much of this can healthcare staff or a doctor provide? Half? Then that 0.4% will not get a puncture, I mean that.
These are indeed complex considerations. Because when are you going to have the injection? What if someone still has six months? Or already at 3 months? You mention six months. Why would only 3 be added in the first month and more later? The mortality rate is the same every month. It is about the correctness of the assessment per individual.
But actually the point is that you can easily reach 50% to 80% with somewhat plausible estimates. It is not a edge phenomenon; It is an important factor that is just included in the real VE for many studies. HVE is "forgotten".
That is a good Ansatz and the built -in calculator is very beautiful and useful (I have not yet checked the calculations).
What seemed confusing to me: Ve usually indicates how well a vaccine protects against a certain disease, see for example https://www.sciencedirect.com/topics/immunology-and-microbiology/vaccine-efficacy. HvE is usually mentioned when comparisons of death on all causes and that is what you look at here. In recent studies, however, the term VE is also used against all causes of death; It would certainly be enlightening for occasional readers to clarify that aspect and also specify it in this article.
Thanks Harald. It is indeed complex. In reality it also intertwines:
- HVE plays just as well with (afterwards) measuring protection against a disease. If someone with an autoimmune problem is not vaccinated, he/she will also come to the unvaccinants with a much greater chance of becoming sick.
- If someone dies, is it then the disease or underlying suffering?
- According to reports (NIVEL, RIVM), the vaccinations also have a beneficial effect on all causes of death.
When measuring an effect on a disease, many more disturbing effects come. When is someone "sick"? With a positive test? With certain antibodies? Doctor's visit, hospitalization? Is everyone exposed to the pathogenic (season, location) on the same extent?
"Died or not" is a harder parameter that excludes all those uncertainties, so it is also better suited for viewing the "dry" HVE effect.
For me it falls outside the scope of this article. I wanted to investigate whether the HVE is not a marginal effect, especially with a high rate (that was my intuition). I see that an overestimation of dozens of percent is almost inescapable and with some good (or evil) you want to come to really high protection of 80% or more. With a placebo ...
I am not a mathematician and I don't know anything about probability. My expectation was Nattevingerwerk, based on the following reasoning: If a doctor gives a quantum people at most half a year, the number of cases in which he is right will increase with time. Now that I think about it, I see that it is probably nonsense. It is probably the case that the prognosis is more accurate after six months than after one month, but that of course says nothing about the time within those six months when the dead fall. It is also full of subjective factors. Insureless nonsense, in short, and, as you rightly point out: it doesn't matter much. What you want to know is: Total number of un sprayed deaths in a certain period, and the percentage of those dead that is un sprayed because of weak condition or low life expectancy - nothing more. Two bundles of data, both polluted and hardly find out: the first, simple fact (number of unsprushed deaths) due to the ruined registration (only two to four weeks after the sprayer booked as sprayed), the second because the motives for non-pricks are in the heads of doctors and dead patients. A thousand relatives survey? Impossible work. What remains are a few objective criteria: age, hospitalization, stay in the retirement home, dementia - of those things.
That's why I started with 1 month. That makes the impact of the effect clear. The longer the period, the harder - and the lower the effect, you can see that in the Caplan-mox Curves. It is no longer predicted, the longer term.
I have simplified it enormously now. Did you just get a message about that I had posted a comment above?
And!
In this opinion piece I read the reverse HVE for the first time...
https://www.trouw.nl/opinie/nee-coronavaccins-veroorzaken-geen-kanker~bf5e785e/
Pretty nice piece! However, he says “there is no known mechanism that would underlie this imaginary risk.” That is of course not correct. That mechanism does exist, several in fact, but no one knows yet how often it actually happens.
But if the total number of cancer diagnoses has indeed not increased as he says (but maybe that's not true, I don't know) then something does indeed seem wrong.
Still looked it up. The number of cancer diagnoses increased by 10% in 2021 and remains more or less at that level. There's more to say about it (the 2020 dip and all). The last few years are expectations, not observations. But the fact that they set expectations so high says it all.
That idea is not new, when vaccinating younger risk groups it was quickly admitted (in England if I remember correctly) that they died relatively more, but that was completely blamed on their poorer health before the injection.
Moreover, in many countries the number of diagnoses of certain types of cancers has indeed increased significantly, and articles have also been published that explain how this is possible as a result of vaccination.