How to read brain-health evidence

Five questions to ask of any claim

Who or what was studied? A result in mice is a hypothesis for humans, not a finding in humans. Much of the glymphatic literature began in rodents; the human studies are newer and smaller.

How many? A proof-of-concept trial of focused ultrasound plus an anti-amyloid antibody, published in the New England Journal of Medicine in 2024, involved three people. The large semaglutide trials in early Alzheimer’s enrolled 3,808. Both are legitimate science; they deserve very different amounts of weight.

What was measured? A protein level in blood is a biomarker. A change on a brain scan is an imaging finding. Slower decline on a clinical rating scale is a clinical outcome. The semaglutide trials are a textbook example of the gap: the drug moved several Alzheimer’s biomarkers by up to 10 percent and still did nothing measurable for memory or function over two years.

Compared with what? Before-and-after measurements in the same people can be informative, but a control group and, ideally, blinding guard against expectation and chance.

Has it been replicated? One study, however elegant, is a beginning. Independent replication is what turns a finding into knowledge.

Levels of evidence in this field, as of 2026

Well established in animals and supported in humans: the existence of a perivascular CSF–interstitial fluid exchange pathway (the glymphatic system); its dependence on sleep; the coupling of slow vascular waves and CSF pulses during human non-REM sleep.

Demonstrated in human trials: that antibodies which remove amyloid modestly slow decline in early Alzheimer’s; that a blood test can reliably indicate amyloid pathology in symptomatic people; that focused ultrasound can open the blood–brain barrier safely in small numbers of people.

Hypothesised, not demonstrated: that supporting the brain’s clearance or energy systems directly changes symptoms or disease course in any condition. Many groups, including ours, are working toward that evidence; nobody has it yet.

How we hold ourselves to this

Every statistic on this site carries its source and year. We say mouse when we mean mouse. We will publish bench data, including negative results, as our program produces it, and we will not describe any NeuroClearance hardware as treating or preventing anything unless and until the evidence and regulators say it does.