The Dutch families have been known to medicine for decades. In 1990, Efrat Levy and colleagues reported in Science that hereditary cerebral haemorrhage with amyloidosis of the Dutch type is caused by a mutation in the amyloid gene, a single change at codon 693 of the amyloid precursor protein. The condition is autosomal dominant, so the children of a carrier each have a one in two chance of inheriting it, and a person can know for years, sometimes decades, that the amyloid is coming for the walls of their cerebral vessels before anything at all has happened to them. That waiting room is exactly what makes these families scientifically precious. Sporadic cerebral amyloid angiopathy is usually found only after a lobar bleed or a memory clinic referral, but in Dutch-type CAA (D-CAA) researchers can watch the disease from before the first haemorrhage. As Ellis van Etten and colleagues put it in a 2016 Neurology paper on recurrent bleeding, the hereditary form has been treated as a monogenic model for the sporadic variant.

The latest instalment is a prospective observational cohort study published in Alzheimer's & Dementia (2026;22(7)) by Rosemarie van Dort, Vera van Stek-Smits, Marieke Wermer, van Etten and colleagues, with collaborators including Steven Greenberg in Boston and Ralph Martins in Australia. They included 181 participants: 93 D-CAA mutation carriers, of whom 59 had never had an intracerebral haemorrhage and 34 had, and 88 people with sporadic CAA, 57 without and 31 with a prior haemorrhage. Everyone had annual neuropsychological assessment, and change in global cognition, memory, processing speed and executive function over five years was modelled with linear mixed models. That is not a trial and nothing was given to anyone; it is a natural history study, and it reports association, not cause.

The headline result is deliberately undramatic. Over five years the authors describe cognitive decline as "subtle but measurable". Memory and processing speed declined in the D-CAA carriers who had already had a haemorrhage. The carriers who had not had one, and the sporadic CAA group as a whole, stayed relatively stable. Global cognition showed a modest but statistically significant decline; executive function did not budge. The authors' reading is that CAA-related cognitive decline is domain-dependent and stage-dependent, and that progressive small vessel injury, rather than the acute haemorrhage itself, may be what drives deterioration. That is an interpretation of observational data from a group of people whose bleeds and biology travel together, not a demonstration of mechanism.

For a family, the shape of this matters more than the statistics. It says the decline that shows up on annual testing in this cohort was not a cliff. It says the domains that moved first were speed and memory rather than the planning-and-switching functions that clinicians often probe first. And it fits with the same group's cross-sectional work: in their 2024 report in the same journal, 20 to 30 percent of D-CAA mutation carriers already showed cognitive dysfunction in the stage before any symptomatic haemorrhage. Trouble can be present early, and it can also sit still for years. Both of those are true at once, and neither predicts any one person's next five years.

What does this license about sporadic CAA, the far commoner condition that turns up in memory clinics and stroke units? Something, but less than the framing invites. The same Leiden research programme reported in Neurology in April 2025 that D-CAA has an earlier onset and faster progression than the sporadic form while being otherwise pathologically and clinically similar, which is the whole basis for using the families as a model. Still, the groups here are small once split by haemorrhage history, they were followed at a single research programme with an annual test battery that people get better at simply by repeating it, and the comparison between hereditary and sporadic groups is a comparison between people of different ages and different routes to diagnosis. Earlier longitudinal work from Leiden, published in Stroke in 2022, found no significant cognitive change over roughly four years in a much smaller D-CAA sample while vascular reactivity worsened, a reminder that cognition is the slowest of these signals to move.

The practical value is for trial design as much as for clinics. If a treatment for vessel amyloid is ever to be tested in these families, someone has to know how fast the untreated curve falls and in which domain, otherwise a study will be too short or too small to see anything. Five years of annual testing in 181 people is the kind of unglamorous groundwork that decides whether the next decade of trials asks answerable questions.