People with both atrial fibrillation and Alzheimer’s disease who took a newer blood thinner lost 0.23 fewer points per year on a standard 30-point memory test than people taking no blood thinner at all, according to a Karolinska Institutet study of 7,308 Swedish patients published August 12, 2026 in the European Heart Journal. The same group also had fewer strokes, fewer fractures and fewer deaths, with no rise in serious bleeding.
Key takeaways
- The memory difference was real but small: 0.23 Mini-Mental State Examination points a year versus no treatment, and 0.21 points a year versus warfarin.
- The newer anticoagulants here are the NOACs: apixaban, rivaroxaban, dabigatran and edoxaban. They were linked to about 19% lower death rates and about 34% lower stroke rates than taking nothing, without more major bleeding.
- This is registry data, not a trial, so the doctors chose who got treated. That alone can produce this result.
What the study found
The researchers used the Swedish Registry for Cognitive/Dementia Disorders to find people diagnosed with Alzheimer’s disease between May 2007 and December 2020 who already had atrial fibrillation. Of the 7,308 people who qualified, 3,341 took no anticoagulant, 2,277 took warfarin, and 1,690 took a NOAC. NOAC stands for non-vitamin K oral anticoagulant, the newer class of blood thinners that includes apixaban (sold as Eliquis), rivaroxaban (Xarelto), dabigatran (Pradaxa) and edoxaban (Lixiana or Savaysa). Each one blocks a single step in the clotting process directly, rather than interfering with vitamin K the way warfarin does, so they do not require routine blood monitoring or the dietary limits warfarin demands. Memory was tracked with the Mini-Mental State Examination.
NOAC users declined more slowly than untreated patients by 0.23 MMSE points a year (95% CI 0.11 to 0.36), and more slowly than warfarin users by 0.21 points a year (95% CI 0.10 to 0.33). Both gaps are small, and the statistics say the true size is very likely somewhere between roughly a tenth and a third of a point a year.
The harder outcomes moved in the same direction. Compared with no anticoagulant, NOAC use was tied to about 19% lower death rates (hazard ratio 0.81), very likely between 9% and 28% lower; about 34% lower rates of ischemic stroke or a clot lodging in an artery elsewhere in the body (HR 0.66), very likely between 18% and 47% lower; and about 21% lower fracture rates (HR 0.79), very likely between 3% and 36% lower. Major bleeding did not go up (HR 1.05).
Warfarin looked worse on safety. It was linked to about 12% lower death rates but a 31% higher rate of major bleeding (HR 1.31), very likely between 9% and 56% higher. Warfarin also carries a known fracture signal in older adults, and it is the drug whose effect shifts with vitamin K intake.
Dr. Kumar’s take
A fifth of an MMSE point a year is not something a family would notice. Over five years it adds up to roughly one point on the test. Any headline that says these drugs protect the brain is running well past the data.
The bigger problem is who ends up on treatment. In real practice, the AFib patient with Alzheimer’s who gets anticoagulated is the one who is steadier on their feet, has family managing the pills, and is further from the end of the disease. The patient left untreated is often the frailer one. That difference, not the drug, is the plain explanation for a slower slide on a memory test, and the authors say so directly: this is observational, and factors affecting both treatment choice and outcome could not be fully analyzed. Statistical weighting narrows that gap but does not close it.
What the study does speak to is a decision made in clinics every day. Anticoagulation is routinely withheld from people with dementia because of fall risk and bleeding fear. Here, the untreated group had more strokes, more deaths and, notably, more fractures, the exact harm the withholding is meant to prevent. The defensible reading is not that blood thinners preserve memory. It is that reflexively leaving a dementia patient with AFib on nothing may be the riskier call.
What it means for you
If a family member has both AFib and Alzheimer’s and is on no blood thinner, ask why. “They have dementia” is not a reason by itself, and “they might fall” deserves an actual fall-risk conversation rather than a reflex.
If they are on warfarin, ask whether one of the NOACs, apixaban, rivaroxaban, dabigatran or edoxaban, fits instead. This study, like earlier work, points to less major bleeding with the newer drugs. Kidney function, cost and drug interactions still decide it case by case.
Do not start, stop or switch an anticoagulant based on a memory argument. The stroke and bleeding math is what should drive that decision, and it always has been.