Ultrasound opens the brain’s barrier to Alzheimer’s drugs in tiny trials

In 2017, a team at Sunnybrook Health Sciences Centre in Toronto placed a helmet holding 1,024 ultrasound elements on the head of a person with Alzheimer’s disease, inside an MRI scanner. A dose of microscopic gas bubbles, the kind used as contrast in heart scans, went into a vein. Then the helmet focused low-frequency sound, 220 kHz, on a small block of the right frontal lobe.

With the bubbles in the bloodstream, the focused sound loosened the lining of the blood vessels in that block, the barrier that normally keeps most of what is in the blood out of the brain. MRI contrast seeped into the target at once. Within 24 hours the barrier had sealed again.

The team did this for five patients, twice each, a month apart. They reported no serious side effects, and noted that opening the barrier took less than 1% of the energy needed to burn a lesion with the same machine. Their paper, in Nature Communications in 2018, ends modestly: the results “support the continued investigation of focused ultrasound.”

Eight years on, that investigation has a sharper purpose.

Why open a door in the brain

Alzheimer’s now has drugs that remove amyloid plaque from the brain. Lecanemab, an antibody given by infusion every two weeks, won full FDA approval in 2023 after a 1,795-patient trial showed it slowed decline over 18 months. The same approval carries a boxed warning for ARIA, amyloid-related imaging abnormalities: swelling or small bleeds in the brain, usually temporary and without symptoms, rarely serious or fatal, and more common in people who carry two copies of the APOE ε4 gene.

The antibodies have to cross the blood-brain barrier to work. The idea behind pairing them with ultrasound is simple: if you open the barrier in the places where plaque is thickest, more of the drug should arrive there, and the plaque should clear faster. Perhaps, eventually, at a lower dose.

What the trials show

The early studies asked only whether the barrier could be opened safely, and how widely.

At West Virginia University, Ali Rezai’s group aimed at the hippocampus, the memory structure the disease hits early. In 10 patients followed for at least six months, Alzforum reports, plaque fell by an average 14% in the treated regions, and the patients declined at about the rate of a comparison group drawn from a national research database. In Seoul, a Korea University group led by J. W. Chang opened about 40 cubic centimeters of both frontal lobes in six patients, three times, two months apart. Plaque fell in four of them, by an average 14.9 Centiloids, a standard scale for amyloid scans. It rose in the other two. Memory scores did not change, though caregivers rated mood and behavior better in five of six.

At Columbia, a portable device opened the barrier in five of six patients, and it closed within three days in all of them. Treated regions gathered amyloid more slowly than the untreated side, but in the first patient the signal fell 1.8% at three weeks and had risen 5.9% by three months. Back at Sunnybrook, regions of a memory-related brain network were opened in nine patients with no serious side effects and a modest drop in amyloid.

Then came the drug.

In January 2024, The New England Journal of Medicine published the West Virginia team’s first combination study. Three people received six monthly infusions of aducanumab, an earlier amyloid antibody, each followed by an ultrasound session aimed at one side of the brain. The other side served as the comparison. On amyloid scans, plaque fell more in the treated regions than in the matching untreated ones. The group later put the difference at an additional 32% reduction over 26 weeks.

Alzforum’s account of the study adds the details that matter. The patients got 6 mg/kg of aducanumab, below the 10 mg/kg dose tested in its large trials. None of them carried APOE ε4, and none developed ARIA. Headache was the most common side effect. One patient’s cognition worsened during follow-up.

The team has since moved on to lecanemab. An abstract published this January describes two patients treated so far, chosen because they had shown no ARIA on lecanemab alone. According to Alzforum, after seven weeks the treated areas had cleared two and five times as much plaque as the untreated ones. The abstract itself claims only that the “preliminary results suggest safety and feasibility.”

The honest read

In June, a systematic review in Aging Clinical and Experimental Research gathered every clinical study of ultrasound barrier-opening in Alzheimer’s it could find: 27 studies and roughly 260 participants. Its first finding was what is missing. “No randomized controlled trials meeting the inclusion criteria were identified.” The barrier-opening studies enrolled between 1 and 10 people each, and every study was rated at moderate risk of bias.

On safety, the record so far is reassuring for the procedure itself. The review found “no reports of irreversible procedure-related adverse events.” Side effects were mild and passed: headache, fatigue, swelling seen only on scans. One Columbia patient had swelling with a small bleed and no symptoms.

On benefit, the review calls the changes in memory and plaque “heterogeneous and exploratory.” That is a fair word for a field where plaque goes down in four patients and up in two.

Three cautions follow from the evidence itself.

First, plaque on a scan is not memory. Lecanemab was approved on what happened to patients over 18 months, not on their scans. No ultrasound study has yet shown that clearing extra plaque in one patch of brain slows anyone’s decline.

Second, the safety question that matters most has not been tested. The antibodies’ main risk is brain swelling and bleeding, and the combination studies so far have chosen people at lower risk: no APOE ε4 in the aducanumab trio, no prior ARIA in the lecanemab pair. Whether letting more antibody into the brain raises that risk is the question the next trials need to answer with the people most likely to face it.

Third, three patients, each compared with the other half of their own brain, are the smallest possible test.

None of this makes the idea weak. Steering a drug to where it is needed, through a door that closes again within days, is a plausible way to make a modest drug work better. The case is still being built, one patient at a time.

What to watch

The West Virginia combination trial, registered as NCT05469009, plans 15 participants. Its listed goals are counts of adverse events, followed for up to five years, and it is expected to finish in July 2029. The Focused Ultrasound Foundation lists it and two other Alzheimer’s trials as fully enrolled with results pending. Alzforum reports a 15-patient Korean trial planned around larger openings, and a second Columbia study of six patients.

Sound is being tested against Alzheimer’s in another way too. Cognito Therapeutics’ 670-patient trial of a 40 Hz light-and-sound headset is due to read out by December, which we previewed in October. We have also covered ultrasound-delivered gene editing in mice and a critical review of gentler brain ultrasound.