Causal evidence that herpes zoster vaccination prevents a proportion of dementia cases, 2023, Markus Eyting

[forestglip]

Interesting, thanks!

One possibility is that they independently reduce dementia risk by preventing their respective infections, shingles and RSV. [...] However, the fact that a protective effect is seen within a few months of vaccination (as was observed in our previous study5) argues against this possibility as it seems unlikely that enough infections would be prevented during such a short time frame to explain the magnitude of the protection against dementia.

This seems like it could be important. They thought preventing the shingles infection was helping prevent dementia, but this is pointing to a specific ingredient found in multiple vaccines instead.

 

[Hutan]

I think a big problem with these studies is that someone who gets a Shingles or RSV vaccine is not, on average, like a person who does not.

Either, the person who is vaccinated is well enough organised, capable of deciding to do something to maintain their health, making the appointment and getting themselves there. They probably regularly visit the doctor for their other health needs. In which case, they are still thinking well and are regularly seen by a doctor, and so are, at worst, some years off getting diagnosed with dementia.

And/or, the person who is vaccinated lives in a situation where someone is looking after their health needs and getting vaccinations done, and so they are also being closely monitored for signs of dementia. So again, if they had dementia, they would probably have already been diagnosed with it. I imagine in a retirement home, there is more funding for a resident with dementia, so it would be in the retirement home's interests to get the diagnosis fairly quickly.

Contrast that with someone who doesn't get the vaccine. Someone in the first stages of dementia and living on their own will be struggling to hold everything together, and probably won't be thinking about discretionary activities such as getting a vaccine. If the unvaccinated person is living with a spouse or a child who isn't able to arrange for the vaccine or doesn't believe in vaccines, then they probably aren't getting to the doctor often either, and may not be being closely monitored for signs of dementia.


If a person who is vaccinated has any signs of dementia, they probably will have already been diagnosed with it already. So, for them, the chance of being diagnosed in those next 18 months is only the chance of a new onset that moves very swiftly.

Whereas, a person who is not vaccinated may already have quite significant signs of dementia but be undiagnosed. So, for them, the chance of being diagnosed in those next 18 months is the chance of an onset in quite a lot of years prior as well as the next 18 months.

 

[forestglip]

I think a big problem with these studies is that someone who gets a Shingles or RSV vaccine is not, on average, like a person who does not.

That's what makes the natural experiment of the first study above useful. A large portion of the people who didn't take the vaccine didn't take it because they weren't allowed to.

Adults born immediately after the 2 September 1933 date-of-birth eligibility cut-off had a 47.2 percentage point higher probability (from 0.01% to 47.2%) of ever receiving the herpes zoster vaccine than those born immediately before this cut-off date.

I think what they did was more sophisticated, but a crude way to test the effect of vaccine on dementia is just seeing how much dementia occurs in the group before September 2 and the group after. The main difference is a different proportion of people who took the vaccine based on whether they were allowed to take the vaccine. Not socioeconomic factors.

 

[Mij]

I don't find it surprising that they are discovering that neurotrophic viruses may put us at a higher risk of developing dementia.

Conclusion​

In summary we find an association between symptomatic HSV-1 infection and AD using a large claims database from USA and highlighting antiherpetic therapies as potentially protective for AD. These findings place an even greater emphasis on viewing the prevention of herpesviruses as a public health priority. Further research to determine whether suppression of neurotropic viruses can alter the natural history of AD and ADRD is warranted based on consistent observational studies.
LINK

​

 

[forestglip]

See thread for new study from this group: The effect of shingles vaccination at different stages of the dementia disease course, 2025, Min Xie et al

Our finding that HZ vaccination had a beneficial effect on two different dementia-related outcomes in two different patient samples—and at two opposing ends of the disease course of dementia—thus provides promising evidence that HZ vaccination may prevent or slow the dementia disease process in a substantial proportion of individuals.

 

[Peter T]

Not read the paper yet, but can they get beyond demonstrating an association. The abstract seems to be claiming they have proved causation which the previous studies reported definitely did not.

 

[forestglip]

Not read the paper yet, but can they get beyond demonstrating an association. The abstract seems to be claiming they have proved causation which the previous studies reported definitely did not.

Are you referring to the new study linked in another thread?

It's basically the same methodology as this thread's study, but instead of looking at dementia diagnoses, they looked at two other metrics: diagnosis of mild cognitive impairment and death from dementia. So something like a sensitivity check. If the effect they saw of vaccines reducing dementia is real, then you should also see the vaccine reducing these other dementia related metrics - and they did.

Are these studies demonstrating causality? While it's still not a randomized controlled trial, it's about the closest you could get from observational data. Instead of researchers splitting a group of people into a treatment group and a control group at random, it was basically the government doing that by choosing who was allowed to get the vaccine. Though it wasn't exactly random - they split based on whether people were born before or after a specific date. But they argue that apart from whether or not they were eligible for the vaccine, these two groups did not systematically differ in any other major ways.

Then they looked at the population several years after becoming eligible for the vaccine. The birthdate cutoff was September 2, 1933. People born after could get it (and about half did) and people born before could not. They checked the dementia rates of people born on either side of that birthdate. You would expect a smooth transition in dementia rates, but they saw a sudden decrease in cases for people born right after that date. Figure 3 from this thread's study shows that sudden change:



Though it appears that the effect is really only apparent for women. This is split by sex:

 

[Peter T]

Are you referring to the new study linked in another thread?

I am not sure now as I had not realised there were two separate studies/threads until after I had commented. Brain fog quite foggy today.

With the change in vaccine eligibility, people still had to make an active decision to have it, and there was some confusion on the ground about who was eligible. As a result of the NHS promotion I had asked my GP practice for the jab, but was told I was not eligible, however several months after that my practice invited me to book in for the jab. However I guess given a 50% take up rate it makes selection bias less likely an issue.

 

[forestglip]

With the change in vaccine eligibility, people still had to make an active decision to have it, and there was some confusion on the ground about who was eligible. As a result of the NHS promotion I had asked my GP practice for the jab, but was told I was not eligible, however several months after that my practice invited me to book in for the jab. However I guess given a 50% take up rate it makes selection bias less likely an issue.

That's the thing. Selection bias doesn't really matter here.

Simple way to imagine it: there's a room of people. They all randomly decided to wear a red or blue shirt that morning (akin to being born before or after the cutoff date in the actual study).

Someone walks in and says "people wearing red, you can have the vaccine if you want. People wearing blue are not allowed."

Imagine in the red group, there is some intense selection effect. Only the people who are very unlikely to get dementia anyway end up taking the vaccine for whatever reason. This would be a problem if you were comparing the people who took the vaccine to those that didn't.

But that's not what they did. They compared the entire red group to the entire blue group. There should be just as many people who are likely to naturally get or not get dementia in both groups. So they're not actually "selecting for" healthier people in the red (vaccine) group.

Imagine there was no real effect from the vaccine. If you compare the red group to the blue group, there should be no difference in dementia rates no matter who in the red group actually was able to or decided to take the vaccine. Yet the study found a difference in dementia rates.

 

[Peter T]

Thank you @forestglip

 

[Utsikt]

From July 2023, a criticism of the preprint:

Shingles vaccine & dementia prevention: too good to be true? by Ellie Murray

The paper seems to cover the main point above that the September 1st cutoff might affect dementia rates because of differing schooling length before and after. They showed that there was no such difference in dementia surrounding the same month and day in prior birth years, only the year that included the vaccine eligibility cut off.

Is that right, though?

They say this about education:

Fourth, using data from the 2011 Census, we show in Supplementary Figs. 15–17 that there are no discontinuities across the 2 September 1933 threshold in the proportion of individuals in Wales who reached a particular level of education.

Equal level of education != equal years of education across the date threshold, because the people after the threshold presumably required one more year of education to get the same level of education.

My mom’s equivalent to a BSc took 14 years of school in total. 9+3+2, with 2 being the BSc-equivalent. Mine took 16 years due to changes in the length of the education: 10+3+3. Yet we have the same level of education (if you exclude my MSc from the calculations).

 

[forestglip]

Equal level of education != equal years of education across the date threshold, because the people after the threshold presumably required one more year of education to get the same level of education.

My mom’s equivalent to a BSc took 14 years of school in total. 9+3+2, with 2 being the BSc-equivalent. Mine took 16 years due to changes in the length of the education: 10+3+3. Yet we have the same level of education (if you exclude my MSc from the calculations).

If eligibility for school was the reason for having more or less dementia, then you'd expect to see the same effect in other years, right? Instead of splitting the groups by being born before or after September 2, 1933, what if you split groups by being born before or after September 2, 1932? The same effect due to entering school a year early should still apply.

But they say the drop in dementia cases was not seen when changing the year but otherwise repeating the same analysis:

Second, we examined whether the day–month (that is, 2 September) date-of-birth cut-off used for zoster vaccine eligibility was also used by other interventions that affect dementia risk. We did so by implementing the identical analysis as for 1 September 2013 (the actual date on which the zoster vaccine program started) for 1 September of each of the three years before and after 2013.

Thus, for example, when shifting the start date of the program to 1 September 2012, we compared those around the 2 September 1932 date-of-birth threshold with the follow-up period starting on 1 September 2012.

As an additional check that enabled us to maintain the length of the seven-year follow-up period used in our primary analyses, we shifted the program start date to 1 September of each of the 6 years preceding (but not after) 2013.

As expected, for both of these checks, we find a significant effect on dementia occurrence only for the date-of-birth cutoff (2 September 1933) that was actually used by the zoster vaccination program (Supplementary Figs. 12 and 13).

Sure, maybe something in this specific year changed in regard to education. But as far as I know, the criticism didn't suggest what that might be, so it's just speculating that some unknown thing relating to education happened and that this is more likely than the known major event of being eligible for vaccination.

 

[forestglip]

I think she might have missed that sensitivity check in the paper. Here's someone asking her if the peer reviewed version of the study checking other years allays her concerns (her post was about the preprint).

Cool! Does the lack of an effect in populations split around that same birthdate but in prior years, or in years preceding 2013 when the vaccine was introduced, allay your concern about the "September effect" of the diff in age at school onset? (Not sure if that control was part of the preprint.)

Response:

I’ll have to read the paper before I can comment

The same sensitivity check the person asked about was in the preprint too, though.

She wrote an updated critique for the peer-reviewed version which might provide some clarity, but it's behind a Substack paywall. [Edit: Nevermind, I think that's just the preprint criticism copied from Medium to Substack. She said to stay tuned for a post about the peer reviewed version, but I don't know if she ended up writing one.]

Edit: Another person asking about if it would be better if the authors did the thing that they did do:

Could your argument that confounding by number of years schooling could be evaluated by simply picking a different year (that did not affect shingles vaccine eligibility) and comparing individuals for the same two birth weeks. If the association were still observed it would not be due to vaccine.

Response:

Yes, it could have been. And I shared these comments with the authors when I wrote them.