There’s recently been [increased] interest in the possibility that a virus may be involved in the causal pathways of dementia. Most attention has been on herpesviruses, incl the VZV that causes chickenpox and shingles, partly because of their lifelong latency. In Wales starting in Sept 2013, the shingles vaccine was rolled out using an exact DOB cutoff. Those born on or after Sept 2 1933 were eligible, while those born earlier weren’t and remained ineligible for life. We analyzed EHR data for all of Wales with DOB exact to the week. We first show that just a one-week difference in age causes a massive (47 %-point) difference in the probability of ever getting the vaccine.
This is paradise for getting at CAUSATION rather than correlation!
After all, there’s no reason why those born one week before Sept 2 1933 should differ in anything from those born just one week later, EXCEPT for this diff in vacc receipt. This is just like a clinical trial: two exchangeable groups with only one getting the intervention.
So, no need for the usual heroic assumption of perfect info on all confounders that make those who get the vacc different from those who don’t. This is crucial, and why our study is fundamentally different to other (all correlational) analyses in this area.
What did we find?
We first tried to replicate the vaccine’s known effect from clinical trials that it prevents shingles and PHN. It does – no surprise here but nice proof of concept.
Now we test our real hypothesis: does getting vaccinated [decrease] your chance of a dementia diagnosis years later?
It does! We estimate that over a 7-year follow-up period, getting vaccinated averts one in five new dementia diagnoses. In the paper, we demonstrate through extensive robustness checks that it is essentially impossible that this finding is due to confounding.
Given how robust this finding is across a large variety of specifications and series of robustness checks, it’s also extremely unlikely to be due to pure chance.
We then show that the vaccine has no effect on any other common causes of morbidity or mortality. This is unlike correlational analyses that usually suffer from bias because healthier/more health motivated folk with better healthcare access are more likely to get vaccinated.
Shingles is more common in women + [increased] recognition that the causes of Alzheimer’s may differ by sex (
http://bit.ly/3HqGh8k). We find strong protective effects of the vaccine for women but none for men, and that this diff is driven by Alzheimer’s (not vascular) dementia.
