Ok, you got me to dig into the study. Really what’s most noticeable to me is that the dementia rate in the ADHD cohort was 13.2% but only 7% in the general population, before any kind of statistical massaging, which I’d have expected to push even further towards correlating ADHD with dementia given it’s much more diagnosed now than in the past and most people with dementia are older.

But the way they’re controlling for stimulants is very poor. To begin with, out of the 100k in the study, only 730 had ADHD and of that only 163 were in the psychostimulant cohort. So to establish statistical significance regarding dementia, which only affected 96 people out of the 730 with ADHD, would be almost impossible if the relationship were mild. And you see that in the huge P-value for the infra-ADHD comparison (0.74) which shows a slight decrease in dementia for stimulant-takers. Basically there is too little data to take that seriously at all.

The comparison across ADHD and no-ADHD populations regarding psychostimulants is deeply flawed too. They’re bucketing [0] dopaminergic stimulants with modafinil (much less dopaminergic) and even nootropics like racetams - literally prescribed to heal brain damage and treat or prevent dementia in former Soviet Union countries, with an entirely different safety profile and mechanism compared to dopaminergic stimulants. Having trouble finding data on racetam legality and use in Israel, so they may not be in scope given the medication would need to be formally prescribed IIUC. But that leaves the remainder as likely taking prescribed dopaminergic stimulants and modafinil, which for those without ADHD are mostly taken by those with narcolepsy. So the non-ADHD psychostimulant users are likely strongly biased towards narcoleptics with the remainder being those taking drugs for the purpose of treating or preventing cognitive decline, and the intra-cohort ratio is 1.1:22.3 so roughly only 5% of the ADHD-stimulant group is comparable to the nonADHD-stimulant group assuming similar, uncorrelated incidence of narcolepsy.

In other words, their two attempts to control for stimulant use either have a uselessly low amount of data giving a p-value of 0.74, or are confounded by comparing against narcoleptics taking stimulants’ dementia rates (not healthy people taking stimulants; they wouldn’t have been prescribed stimulants so long without having ADHD or narcolepsy) with the possible wildcard of some of these “psychostimulant” users taking non-dopaminergic stimulants or even non-stimulant nootropics.

[0] https://en.m.wikipedia.org/wiki/ATC_code_N06