Findings

The first chart shows each study that compared people with dementia against people without it. Pooled estimates combine many studies. The often-quoted “8 to 10 times” comes from two memory-clinic studies that compared their patients against published population rates; studies with their own matched comparison groups mostly find two to five times. In Amatniek 2006, the risk was about 87 times higher when Alzheimer’s began at 50 to 59, and about 3 times higher at 80 or over. The second chart shows every seizure rate reported by a study of people with dementia. Most fall between 1% and 15% (the middle value is 7%); rates rise when researchers follow people longer, look harder, or study inherited Alzheimer’s.

Each bar is one study that reported what share of seizures had no jerking or stiffening.

In a study that followed 453 people with Alzheimer’s disease, caregivers answered yes to a seizure screening question for 52 people. Two epileptologists then reviewed each record: 7 were judged to be seizures, 9 were not, and 36 could not be judged because the records lacked the detail needed. Since then, studies have looked at records once or at a single caregiver report; none has followed what families and care staff observe over time.

For people with dementia and no known seizures, each dot is the share whose EEG showed seizure-related activity, grouped by how long the EEG recorded. A routine EEG (under an hour) found it in 2–6% of people; 24-hour recordings found it in 22–54%, and overnight or multi-day recordings in 10–80%. The lowest figure, 10%, came from the study with the strictest expert review, so how a recording is read matters as much as how long it runs.

Vossel 2016 recorded people with Alzheimer’s disease overnight: those with seizure-related activity lost 3.9 points a year on a standard memory and thinking test (MMSE), against 1.6 points for those without. The study was small, so the size of the difference is uncertain. In Baker 2019, a caregiver’s suspicion of epilepsy at the first visit was linked to faster decline over 12 months. Seizures can also come first: in some people, epilepsy began years before memory problems were diagnosed (Vossel 2013, Cretin 2016).

One trial’s authors estimate that choosing participants whose EEG shows seizure-related activity would cut the trial needed from 468 people to 20. Finding seizures is what makes treatment testable.