I had aspirin and dementia filed away as a closed case, so two pairs of numbers from a new analysis stopped me cold. Among older adults whose DNA predicts the highest platelet counts, 1 percent of those on daily low-dose aspirin developed dementia, against 3.3 percent of those on placebo. In those same people, major bleeding hit 4.4 percent on aspirin and 2.1 percent on placebo.

TOP FIFTH OF PLATELET SCORES
TOP FIFTH OF PLATELET SCORES
OutcomeAspirinPlacebo
Dementia1%3.3%
Major bleeding4.4%2.1%
The same high-score group, the same pill, two outcomes moving in opposite directions. Source: ScienceDaily: Monash University release on the aspirin and platelet genetics finding (2026)

Both pairs come out of the trial that supposedly settled the question. ASPREE randomized 19,114 healthy adults, most of them 70 or older, to 100 mg of aspirin a day or a placebo. In 2020 it reported incident dementia at 6.7 versus 6.9 cases per 1,000 person-years, and the coverage ran under a headline saying aspirin showed no effect in lowering dementia risk. I read that, nodded, and never thought about it again.

Now a Monash University team led by Peter Fransquet has gone back into the trial’s DNA. Their paper, “Platelet count polygenic scores may modify the effect of aspirin in primary prevention of dementia,” appears in Alzheimer’s & Dementia and draws on 13,541 participants with genetic data, 345 of whom developed dementia over a median 4.6 years, according to Medical Republic’s account of the paper. A polygenic score adds up thousands of tiny DNA variants, each nudging a trait a hair up or down, into one number for your inherited tilt. This one estimates how high your platelet count is built to run. It isn’t the platelet count printed on your blood panel, and there is no clinical cutoff you could check yourself against. It’s also a post hoc, exploratory look at a trial that was never designed to test this.

112 hits, one signal

The search was wide. The team screened 5,182 polygenic scores, kept 1,848 after quality control, and found 112 with a nominal interaction with aspirin. By itself, 112 means little: if the conventional 5 percent cutoff was used, chance alone would produce roughly 92 hits from 1,848 tries. After correcting for all that testing, 3 survived, and all 3 were closely correlated scores for platelet count. That’s one signal measured three ways, not three confirmations. Still, of nearly two thousand inherited traits, the one left standing is tied to the exact cell aspirin is famous for hitting, and I find that harder to wave off than a stray hit on a trait with no connection to the drug.

THE GENETIC SEARCH
5,182
scores screened
1,848
left after quality control
112
nominal aspirin interactions
3
survived correction
Three correlated platelet scores survived, which amounts to one signal measured three ways. Source: Medical Republic: the ASPREE platelet-score analysis (2026)

In the top fifth of platelet scores, roughly 2,700 people split across both arms, the dementia hazard ratio was 0.28 (95% CI 0.15 to 0.52), the “70 percent lower relative risk” in the headlines. The major bleeding hazard ratio in the same group was 2.13 (95% CI 1.36 to 3.34). Spread over two arms, those dementia percentages add up to only a few dozen cases, which is why that interval is so wide. The separation showed up after about 2 to 3 years of follow-up, ScienceAlert reports.

ASPIRIN VS PLACEBO, TOP FIFTH (HR)
Figure data
ResultValue (HR)Lower boundUpper bound
Dementia0.280.150.52
Major bleeding2.131.363.34
The dementia estimate rests on a few dozen cases, which is why its interval is so wide. Source: Medical Republic: the ASPREE platelet-score analysis (2026)

The signal also didn’t run through the usual Alzheimer’s genetics. Aspirin showed no interaction with APOE or with genetic risk scores for Alzheimer’s and dementia. “What’s particularly interesting is that the signal wasn’t linked to the established genetic risk factors,” Fransquet told ScienceAlert. “Instead, it’s pointing us towards platelet biology and gives us a new avenue to investigate.”

Why would a clotting cell care about memory?

This was the question I couldn’t put down. Low-dose aspirin permanently jams an enzyme inside each platelet, the small cell fragments that pour into a nicked vessel wall and pile on to plug it. Once a platelet has met aspirin, it stays less sticky for the rest of its short life. So wait: if your genes keep the platelet supply running high, are more of those little plugs forming in the tiny vessels that feed your brain, year after year, with aspirin turning the volume down?

That’s my question, not the paper’s finding, and the authors don’t claim to know. Medical Republic stresses that the mechanism is uncertain and that the scores come from platelet-count genetics, not from any measurement of how sticky or active a person’s platelets actually are. One wrinkle cuts both ways. Placebo takers in the top fifth developed dementia more often than the cohort overall, which hints that this genetic group carries extra risk aspirin might be blocking. Yet the continuous platelet score was not significantly linked to dementia in the placebo group or the whole cohort. One reanalysis can’t tell you whether that elevated placebo rate is biology or a small subgroup throwing off noise.

Five fewer, five more

Medical Republic converts the subgroup results into absolute terms: about 5 fewer dementia cases and about 5 extra major bleeds per 1,000 person-years, with no difference in all-cause mortality. The symmetry is tidy, but it isn’t a settled one-for-one trade. The dementia side is a single exploratory estimate with a wide interval, while the bleeding side echoes everything ASPREE already showed, and pooled “major bleeding” doesn’t tell you how many were stomach bleeds caught and treated versus bleeds inside the skull.

That bleeding history is long. The parent trial raised major hemorrhage to 8.6 versus 6.2 events per 1,000 person-years without significantly cutting cardiovascular events. The trial’s own summary records an increase in cancer mortality in the aspirin arm during the randomized phase; in the post-trial extension, cancer mortality no longer differed between the original arms. Those results helped push the US Preventive Services Task Force to a 2022 grade D, recommending against starting low-dose aspirin for primary prevention of cardiovascular disease in adults 60 and older.

Fransquet doesn’t oversell it. “People shouldn’t start taking aspirin to prevent dementia on the basis of this study without consulting with their doctor,” he said, as quoted by The Epoch Times, “particularly given the increased risk of serious bleeding.”


What bothers me is how quickly “no effect on average” turned into “no effect.” A flat average across thousands of people means the trial couldn’t find a benefit overall. It can’t show that nobody benefited, any more than it shows nobody was harmed, because a gain in one fifth of the cohort can be diluted by four fifths who get nothing. The 2020 framing was accurate for the average and silent about everyone inside it. This reanalysis doesn’t prove the subgroup benefit either. It produces a hypothesis sharp enough to deserve a trial, which is exactly what Fransquet calls for: “We now need to confirm the finding in other studies and ultimately test it in a trial designed specifically.”

Then comes the money question. ASPREE ran on US and Australian public funding, and Bayer, which supplied the pills and placebo, “had no other involvement in the trial.” Fransquet notes that if the result holds up, aspirin being “already cheap and widely available” could make personalised dementia prevention possible. That cheapness is great news for patients and an awkward fact for anyone hoping a drug company will pay for the next trial, since a generic that costs pennies leaves no new patent waiting at the end. The confirmation will most likely need the same kind of public backers who paid for ASPREE, and whether they step up for a cheap pill will show what they actually prioritize.

I came into this thinking aspirin had been tested for dementia and failed. I’m leaving thinking it was tested on average, came up flat on average, and the more useful question, whether it helps a particular kind of brain, has barely been asked. If I were 75 tomorrow, I would not start a daily aspirin because of this paper, and if a trial built for people with this genetic profile opened near me, I would enroll.

Sources

  1. Alzheimer’s & Dementia: Fransquet et al., “Platelet count polygenic scores may modify the effect of aspirin in primary prevention of dementia” (2026)
  2. Medical Republic: the ASPREE platelet-score analysis (2026)
  3. ScienceDaily: Monash University release on the aspirin and platelet genetics finding (2026)
  4. ScienceAlert: Fransquet on platelet biology and the need for a dedicated trial (2026)
  5. The Epoch Times: researchers’ bleeding caution on low-dose aspirin and dementia (2026)
  6. NeurologyLive: ASPREE 2020 dementia results, aspirin vs placebo
  7. Healio: ASPREE cardiovascular and hemorrhage results (2018)
  8. ASPREE: scientific summary, funding, Bayer in-kind supply and cancer mortality follow-up
  9. US Preventive Services Task Force: aspirin for primary prevention of cardiovascular disease (2022)